BH3 Domain of BAD Is Required for Heterodimerization with BCL-XL and Pro-apoptotic Activity*

BAD interacts with anti-apoptotic molecules BCL-2 and BCL-XL and promotes apoptosis. BAD is phosphorylated on serine residues in response to a survival factor, interleukin-3. Phosphorylated BAD cannot bind to BCL-XL or BCL-2 at membrane sites and is found in the cytosol bound to 14-3-3. We report here that deletion mapping and site-directed mutagenesis identified a BH3 domain within BAD that proved necessary for both its heterodimerization with BCL-XL and its death agonist activity. Substitution of the conserved Leu151 with Ala in the BH3 amphipathic α-helix abrogated both functions. The BAD Leu151 mutant was predominantly in the cytosol bound to 14-3-3. The BH3 domain of BCL-2 also proved important for BCL-2/BAD interaction. These results establish a critical role for a BH3 domain within BAD and provide evidence that BAD may function as a death ligand whose pro-apoptotic activity requires heterodimerization with BCL-XL.

The BCL-2 family of proteins comprised of anti and pro-apoptotic molecules constitutes a critical, intracellular decision point within a common cell death pathway (1). The ratio of antagonist (BCL-2, BCL-X L , MCL-1, and A1) to agonist (BAX, BAK, BCL-X S , and BAD) molecules dictates whether a cell will respond to a proximal apoptotic stimulus (1,2). Membership in the family was first defined by homology in two conserved regions, the BH1 and BH2 domains. Mutational analysis of BCL-2 identified key residues within BH1 and BH2 domains required for both heterodimerization with BAX and repression of cell death (3). However, other BCL-X L mutants lost heterodimerization with BAX but still retained some death repressor activity, suggesting that these two functions were separable (4). An additional domain, BH3, had also been noted in BCL-2 proteins and proved essential for the pro-apoptotic function of BAK (5)(6)(7). Moreover, the death-promoting molecules BIK and BID possess only a BH3 domain without identifiable BH1 and BH2 domains, arguing for the importance of BH3 in death agonists (8 -10). BAD, initially identified by its interaction with BCL-2 and BCL-X L , is a distant BCL-2 family member, bears only the most universally conserved amino acids within BH1 and BH2 domains, and lacks the typical hydrophobic C-terminal signalanchor. The presence of BAD counters the anti-apoptotic effect of BCL-X L or BCL-2 perhaps by directly inhibiting them or by displacing the pro-apoptotic BAX molecule (11). BAD represents a bridging molecule interconnecting signal transduction pathways from extracellular survival factors with the BCL-2 intracellular checkpoint on cell death. BAD is phosphorylated on two serine residues embedded in canonical 14-3-3 binding sites in response to a survival factor, IL-3. 1 Phosphorylated BAD does not bind BCL-X L and is sequestered in the cytosol bound to 14-3-3, a specific phosphoserine-binding protein. Substitution of the serine phosphorylation sites indicated that phosphorylation of BAD inactivated the molecule to promote cell survival (12). In the present study, we utilize deletion mapping and site-directed mutagenesis to define a BH3 domain in BAD required for heterodimerization with BCL-X L and its pro-apoptotic activity. Substitution of the conserved Leu 151 in this predicted amphipathic ␣-helical BH3 domain abrogated both functions.
Binding 32 P-labeled BCL-2 to BAD Deletion Mutants-Total cell lysates were prepared from BL21 cells (Novagen), which contained a pET17b vector with BAD deletion constructs fused in frame with T7 gene 10 following induction with isopropyl-1-thio-␤-D-galactopyranoside (0.1 mM) for 1 h. Lysates (40 g) were size fractionated by SDSpolyacrylamide gel electrophoresis and transferred to a nitrocellulose membrane. The membrane has hybridized with 32 P-labeled GST-BCL-2 according to the protocol of Blanar and Rutter (13).
In Vitro Transcription-Translation and in Vitro Binding Assay-Wt and mutant BAD and BCL-2 proteins were generated by an in vitro transcription-translation system (Promega). In vitro binding assays were performed as described previously (12).
Transfection, Viability Assay, and Cellular Fractionation-Stable transfectants were generated in FL5.12 BCL-X L cells as described previously (12). Transient transfections were performed in BAD-deficient murine embryonic fibroblasts with pcDNA3-derived constructs. The luciferase reporter plasmid (0.1 mg) was mixed with 0.05 mg of various constructs and 3 ml of lipofectAMINE (Life Technologies, Inc.) in a volume of 0.5 ml added to murine embryonic fibroblast cells for 5 h. Cells were lysed 18 -20 h later, and luciferase assays were performed using a standard substrate (Promega). Luciferase activities were quantified by a luminometer (OptocompII, MGM Instruments Inc.). An assessment of cell viability was displayed as the relative luciferase activity of a test construct compared with the control pcDNA3 plasmid (10). The viability of FL5.12 cells was measured by propidium iodine exclusion. FL5.12 BCL-X L /BAD cells were fractionated to separate cytosol from crude membrane as described previously (12).
Immunoprecipitation and Western Blots-Immunoprecipitation and Western blots were performed as described previously (12). The BAD * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

RESULTS
BH3 domain of BAD is required for cell death, and heterodimerization-BCL-2 interaction cloning had identified three independent clones of Bad, the smallest encoding only residues 141-204, the C terminus (11). To define the minimal region in BAD essential for its interaction with BCL-2, we generated a nested set of deletion mutants (Fig. 1A). These mutant BAD proteins were expressed in vitro and tested for their ability to interact with BCL-2 protein. Removal of an additional 12 amino acids (BAD construct 152-204) abrogated binding to BCL-2 (Fig. 1B), whereas deletion of 32 amino acids from the C terminus (BAD 141-172) still bound BCL-2. Therefore, a small 31-amino acid region (141-172) is both sufficient and essential for BAD to heterodimerize with BCL-2. This includes a sequence (151-159) with homology to BH3 domains found in other pro-apoptotic molecules ( Fig. 2A). The BH3 domain of BAD is predicted to be an amphipathic ␣-helix (Fig. 2B).
A transient transfection assay was used to assess the role of various regions of BAD in promoting apoptosis. A large Nterminal deletion mutant (127-204) with an intact BH1/BH3 region and a small C-terminal deletion mutant that removed the BH2 domain (1-181) were nearly as effective as Wt BAD in promoting cell death (Fig. 3). In contrast, deletion of the BH1/ BH3 region (deletion of amino acids 142-165) diminished death-promoting function substantially. This same construct (BAD ⌬142-165) also failed to bind BCL-2 or BCL-X L (Fig. 3B and data not shown). Thus, the BH1/BH3 region (142-165) is required for both heterodimerization and death agonist activity.
To further dissect this BH1/BH3 region we used site-directed mutagenesis to substitute individual amino acids (143-153) and assess the effect of their substitution. Of note, the greatest impact was not from substitutions in the region homologous to BH1 as had been noted for BCL-2 and BCL-X L (4, 3, 15, 16). Instead, replacement of Leu 151 of the BH3 domain with alanine (L151A) reduced the binding of mutant BAD with either BCL-2 or BCL-X L by more than 90%. In contrast, BAD G148A binding with BCL-2 was reduced approximately 50%, whereas its interaction with BCL-X L was only minimally affected (Fig. 4A).
We also assessed the domains in BCL-2 most involved in heterodimerization with BAD. An in vitro binding assay revealed that GST-BAD still interacts with slightly reduced efficiency to the BH1 mutant mI-3 (G145A) and weakly to the BH2 mutant mII-1 (W188A) but not at all to the BCL-2 BH3 mutant (L97A) (Fig. 5). Thus, BH3 plays a prominent role in heterodimerization for both the death agonist and antagonist.
To assess the death-promoting action of the BAD point mutants, we stably expressed them in the hematopoietic cell line FL5.12 BCL-X L . Clones with similar levels of Wt and mutant BAD as well as BCL-X L were identified (Fig. 4B) and tested for viability after IL-3 withdrawal. Mutants of BAD G148A and BAD R149A within the BH1-like region like Wt BAD reversed the protective effect of BCL-X L ; however, a BH3 mutant BAD L151A could no longer promote cell death (Fig. 6C). The 7B2 anti-BCL-X L mAb co-precipitated the Wt BAD, BAD G148A, and somewhat less efficiently BAD R149A, but failed to coprecipitate substantial amounts of BAD L151A with BCL-X L (Fig. 4C). Consistent with this, a markedly increased amount of BAD L151A was present in the supernatant of this immunoprecipitate. This provides in vivo confirmation (Fig. 4C) for the loss of BAD L151A binding to BCL-X L noted in vitro (Fig. 4A). FIG. 6. The BAD BH3 mutant L151A is mostly localized to cytosol bound to 14-3-3 and has lost its death-promoting activity. A, effects of BAD mutations on intracellular distribution. FL5.12BCL-X L clones expressing Wt or mutant BAD were fractionated into crude membrane (CM) and cytosol (Cyt), which were analyzed by Western blot using an anti-BAD Ab (10929). B, BAD mutants ability to interact with 14-3-3. Western blot detection of the amount of 14-3-3 co-immunoprecipitated with BAD (IP, 2G11 mAb) from the cytosolic fraction of FL5.12 BCL-X L clones expressing Wt or mutant BAD. C, comparison of Wt and mutant BAD on cell viability after IL-3 withdrawal. The viability of FL5.12 BCL-X L clones expressing Wt or mutant BAD was assessed by propidium iodine exclusion at 24, 48, and 72 h after withdrawal of IL-3. Two independent sets of clones selected for comparable levels of BAD were tested and showed similar results. Each point represents the mean Ϯ SD of triplicate assays.
Because intracellular localization of BAD is associated with its functional activity (12), we tested whether BAD L151A had also altered its subcellular distribution. The majority of BAD L151A was present in the cytosolic fraction with a more prominent upper band (Fig. 6A, lane Cyt, L151A), which represents the hyperphosphorylated form of BAD. Furthermore, anti-BAD mAb 2G11 co-precipitated significantly more 14-3-3 protein associated with BAD L151A than with Wt BAD or the other mutants (Fig. 6B). These data indicate that BAD L151A that is incapable of binding to BCL-X L is also functionally inactive and localized to the cytosol where it is bound to 14-3-3.

DISCUSSION
A combination of deletion mapping and site-specific mutagenesis identified a BH3 domain in BAD as critical for both its heterodimerization with BCL-2 or BCL-X L and its deathpromoting activity. The short BH3 motif LRRMSDEFE proved most similar to BAX ( Fig. 2A). Molecular modeling of the BH3 domain of BAD revealed a classic amphipathic ␣-helix in which the critical Leu 151 resides on the hydrophobic face (Fig. 2B). The close proximity of the BH3 motif of BAD to the previously noted BH1-like YGR core residues is somewhat unusual ( Fig.  2A), and these YGR residues may reside within the same ␣-helix. Of note, BAK contains VGR residues in the N terminus of its amphipathic ␣2-helix proximal to the highly conserved BH3 motif ( Fig. 2A). A mutant BAK R76A peptide demonstrated decreased binding to BCL-X L (17). This also suggests that the BAD structure will vary from the BCL-X L hydrophobic pocket generated by BH1, BH2, and BH3 (15). Of note substitution of Gly 149 of BAD affected its binding to BCL-2 much more than to BCL-X L indicating a difference in the contact sites of these anti-apoptotic molecules.
These data for BAD coincide with a growing body of evidence that implicates BH3 domains in pro-apoptotic molecules. The BH3 region of BAX and BAK are necessary for binding to BCL-X L and promoting apoptosis (5)(6)(7)16). A detailed NMR analysis of wild type and mutant peptides of the BH3 amphipathic ␣2-helix of BAK indicated critical interactions with BCL-X L through both hydrophobic and electrostatic interactions (17). Moreover, the death agonists BIK and BID possess only the BH3 region, arguing that it represents the minimal death domain (8 -10).
One model of BAD regulation holds that phosphorylated BAD would be inactive sequestered in the cytosol by 14-3-3, whereas nonphosphorylated BAD would be the active form bound to BCL-X L inhibiting it and/or releasing pro-apoptotic BAX (12). However, this prior study of BAD phosphorylation did not address another model in which unbound, free BAD might serve as an active death agonist. The current data disfavor this model in that the minimal L151A point mutation of BH3 indicates that a BAD molecule that is unable to bind BCL-X L or BCL-2 is heavily phosphorylated, bound to 14-3-3 in the cytosol, and functionally inactive. The prominence of the BH3 domain for BAD function combined with only minimal conservation of BH1 and BH2 motifs argue that BAD is more closely related to BID and BIK. Like BID, BAD lacks the C-terminal signal anchor sequence and may represent a death ligand that inhibits the membrane receptors BCL-2 and BCL-X L .