INTRODUCTION

The BCL-2 family of proteins consists of inhibitors and inducers of programmed cell death or apoptosis (1-3). Inhibitors include the BCL-2 (4, 5) and BCL-X_L proteins (6) and inducers include BAX, BAK, and BCL-X_S (6-10). These proteins have been shown to form a network of homo- and heterodimers (11). A number of studies suggest that dimer formation is essential for the biological activity of these molecules. For example, mutagenesis data have demonstrated a correlation between BCL-2 activity and the ability to form heterodimers with BAX (12). However, other experiments with BCL-X_L suggested that dimerization with Bax was not necessary for biological activity (13).

Bad was originally cloned from mouse cDNA by its ability to bind to BCL-2, both in yeast two-hybrid interactions and by direct biochemical interaction (14). It was subsequently shown to interact more strongly with BCL-X_L than with BCL-2, and in functional studies it antagonized the protective effect of BCL-X_L. To date, the human homologue has not been reported. The sequences of the BCL-2 family proteins show several regions of clustered conserved residues, termed by some investigators BH-1 to BH-4 domains (12, 15, 16). The crystal and NMR structures of BCL-X_L show a potential binding pocket on the surface of the molecule formed by a combination of the BH-1, BH-2, and BH-3 domains (17). The BH-3 domains of BAX and BAK have been characterized extensively in structure/function studies by mutagenesis. It was shown that the BH-3 domains of BAX and BAK are necessary and sufficient to induce apoptosis and to mediate heterodimer formation with BCL-2 and BCL-X_L (16, 18, 21). More recent NMR studies have defined the binding mode of the BH-3 domain of Bak to the binding pocket in BCL-X_L (19). Our own recent results indicate that peptides encompassing the BH-3 domains of BAX and BAK are capable of completely inhibiting the various dimerization reactions of full-length BCL-2, BCL-X_L, and BAX (20). To date, domains in BAD necessary for binding to BCL-2 and BCL-X_L have not been identified. Furthermore, a series of BCL-2 and BCL-X_L mutants have been described which have all lost the ability to interact with BAX or BAK. Functional analysis of these mutants has demonstrated that some of these mutants retain anti-apoptotic activity while others are are inert, leading to seemingly conflicting conclusions regarding the role of heterodimers in biological activity (12, 13). These mutants, however, had not been evaluated for binding to BAD. In the present study we analyze three issues relating to the biology of BAD. First, we cloned the human homologue of BAD and have evaluated its biological function and binding properties. Second, we have established that BAD contains a functional BH-3 domain and that this sequence is responsible for its observed heterodimerization with BCL-X_L. And third, we have compared BAD to BAX and BAK regarding their respective abilities to bind to active and inactive mutants of BCL-2 and BCL-X_L. The results indicate that these mutants that have lost the ability to bind to BAX and BAK still bind avidly to BAD. The implications of these results concerning the relationship of heterodimer formation to biological activity are discussed.

EXPERIMENTAL PROCEDURES

The full-length human BAD sequence was amplified by PCR¹ and subcloned into the yeast two-hybrid vectors pGilda and pJG4-5 (27) to express BAD as a fusion protein with either the DNA-binding protein LexA or the transcriptional activator protein B42. For the PCR reactions we used the following primers: 5-ATCAGTGAATTCACTATGTTCCAGATCCCAGAC-3 and 5-ATCGATCTCGAGTCACTGGGAGGGGGCGGAGCT-3. The PCR fragments were digested with EcoRI and XhoI and subsequently subcloned. Human BAK was PCR amplified and subcloned as an EcoRI/XhoI fragment into vector pJG4-5 using the following primers: 5-ATCAGTGAATTCACTATGGCTTCGGGGCAAGGCCC-3 and 5-ATCGATCTCGAGTCAGTTCAGGATGGGACCATTGC-3. All other yeast two hybrid plasmids were described previously (21).

For quantitative -galactosidase activity assays, each bait vector was cotransformed with its respective prey plasmid and the reporter plasmid p18-34 into the yeast strain EGY191 (Mata ura3 trp1 his3 LEU2::pLEXop1-LEU2). Three colonies were assayed in duplicates as described (21).

Two expression vectors were used to express human BCL-2, BCL-X_L, BAX, and BAD in bacteria. The C-terminal membrane spanning domain present in BCL-2, BCL-X_L, and BAX was deleted. pGEX-4T-1 (Pharmacia) was used for expression with N-terminal GST tags. The construct for expression of GST-BCL-2 coding for amino acids 1-218 of human BCL-2 was a generous gift from John Reed. For the expression of GST-BCL-X_L, a fragment coding for amino acids 1-211 was obtained from a full-length BCL-X_L cDNA template by PCR with plaque forming units of polymerase (Stratagene). The primer sequences were: 5-AGTATCGAATTCATGTCTCAGAGCAACCGG-3 and 5-TACAGTCTCGAGCTAGTTGAAGCGTTCCTGGCCCT-3 with EcoRI as a 5-cloning site and XhoI as a 3-cloning site. The construct for expression of GST-BAX coding for amino acids 1-170 was obtained in analogous fashion. pET-15b (Novagen) was used for the expression of proteins with N-terminal 6-histidine tags. For the expression of 6-histidine-BCL-X_L, a fragment coding for amino acids 1-211 was obtained by PCR. The fragment contains an 5 NdeI site and 3 XhoI site as well as an additional C-terminal Glu-Glu-Phe epitope tag. The primer sequences were: 5-ATCCGTCATATGTCTCAGAGCAACCGG-3 and 5-TACAGTCTCGAGTCAAAATTCCTCGTTGAAGCGTTCCTGGCC-3. 6-Histidine-BAX (amino acids 1-170) was constructed in an analogous fashion, but without a Glu-Glu-Phe tag with the following primers: 5-ACGTACCATATGGACGGGTCCGGGGAG-3 and 5-TACAGTCTCGAGCTACCACGTGGGCGTCCCAAA-3. For the expression of 6-histidine-BAD, a fragment with an 5 NdeI site and 3 XhoI site was obtained by PCR using the following primers: 5-GGGAATTCCATATGTTCCAGATCCCAGAG-3 and 5-TACAGTCTCCAGTCACTGGGAGGGGGCGGAGCT-3.

Bacteria were grown in 3-liter batches in shaker flasks. Protein expression was induced by addition of isopropyl--D-thiogalactopyranoside. Expressed proteins were purified by one-step affinity chromatography using His-Bind® metal chelation resin (Novagen) or glutathione-Sepharose (Pharmacia) columns on an FPLC system (Pharmacia) essentially following the manufacturer's instructions. The purity of proteins obtained by these methods was assessed by SDS-polyacrylamide gel electrophoresis. Protein concentrations were determined routinely using a bicinchoninic acid method (Pierce Inc.) with a bovine serum albumin (BSA) standard. Expressed proteins were considered sufficiently pure when no contaminating bands could be detected in Coomassie-stained gels loaded with approximately 1 µg of protein per lane as described in detail previously (20).

Peptides were synthesized by standard solid-phase methodology by Alpha Diagnostic, San Antonio, TX. Molecular weight was confirmed by mass spectrometry. Purity was considered acceptable when greater than 70% by high performance liquid chromatography analysis. Peptides not meeting this criterion were high performance liquid chromatography purified. The following peptides were synthesized: human BAD 102-127, NLWAAQRYGRELRRMSDEFVDSFKK (25-mer); human BAD 103-123, NLWAAQRYGRELRRMSDEFVD (21-mer); human BAD 108-127, QRYGRELRRMSDEFVDSFKK (20-mer); human Bad 108-123, QRYGRELRRMSDEFVD (16-mer); human BAK 72-87, GQVGRQLAIIGDDINR (16-mer). The following peptides with amino acid exchanges in conserved positions served as controls: human BAD 103-123, NLWAAQRYGREARRMSREFVD (21-mer); human BAD 108-127, QRYGREARRMSREFVDSFKK (20-mer); human BAD 108-123, QRYGRELRRMRDEFVD (16-mer); human BAK 72-87, GQDGRQEAIIRRLINR (16-mer).

Purified 6H-BAD, 6H-BAX, 6H-BCL-2, or 6H-BCL-X_L was diluted to 4 µg/ml in phosphate-buffered saline (PBS) and coated onto 96-well microtiter plates (50 µl/well; Immunosorb, Nunc) for 18 h at 4 °C. All other incubations were performed at room temperature. The plates were washed two times with PBS containing 0.05% (v/v) Tween-20 (PBS-T) then blocked with 150 µl of 2% (w/v) BSA in PBS for 2 h. Subsequently, the plates were washed two times with PBS-T and incubated with a range of concentrations (20-0.0001 µM) of GST-BCL-2, GST-BCL-X_L, GST-BAX (or versions with the described mutations) in PBS-T containing 0.5% (w/v) BSA (50 µl/well) and the interaction was allowed to proceed for 2 h before washing the plates five times with PBS-T. The wells were incubated for 1 h with 50 µl of the primary antibody, a mouse anti-GST monoclonal antibody (7E5A6) (a generous gift of John Reed) used at 1 ng/ml in PBS-T plus BSA, before being washed five times with PBS-T. An alkaline phosphatase-conjugated goat anti-mouse antibody (ImmunoResearch Laboratories) was added (50 µl/well) and incubated for 1 h, and the plates were then washed five times. p-Nitrophenyl phosphate (Kirkegaard and Perry) at 4 mg/ml in 10 mM diethanolamine (pH 9.5) containing 0.5 mM MgCl₂ was used as the enzyme substrate; the reaction was allowed to progress for 15 min then stopped by the addition of 0.4 M NaOH (50 µl/well). The optical density at 405 nm was determined in a spectrophotometer (Molecular Devices).

For the peptide inhibition studies the binding assay was carried out as above except for the following modifications. After the 96-well plates were coated with the appropriate proteins, the wells were then preincubated with increasing concentrations of peptide (0.004-80 µM) before addition of the GST-tagged protein at a constant concentration (80 nM) throughout. This concentration was determined in preliminary experiments to be on the rate-limiting part of the binding curves.

It has been shown that transfection of BCL-2 can confer resistance to staurosporine-induced death in GM701 cells which can be measured as chloramphenicol acetyltransferase production by surviving cells derived from a cotransfected chloramphenicol acetyltransferase expression plasmid (22). We have adapted this method to the use of BCL-X_L as a survival gene and luciferase as a reporter gene. In brief, GM701 cells were seeded 24 h before transfection at 3 × 10⁵ per 60-mm plate. Various test plasmids were co-transfected with the pGL3-Luc plasmid (Promega Corp.) by the calcium phosphate method for 5-6 h. 24 h post-transfection the apoptosis inducing agent staurosporine was added to one of two duplicate plates for each test plasmid at the final concentration of 1 mM. Fifteen hours after the staurosporine addition, cell extracts were prepared and luciferase activity was measured by standard methods (Promega Corp.). The percentage of the remaining luciferase activity after staurosporine addition compared with untreated cells was used as a measurement of viability of transfected cells. 293 cells were used for an assay with a morphological readout. Cells were seeded 24 h before transfection at 2 × 10⁵ per well in 6-well plates. Various test plasmids were co-transfected with the CMV-gal plasmid (CLONTECH) by the calcium phosphate method for 5-6 h. 24 h after transfection, cells were stained with 5-bromo-4-chloro-3indoyl--D-galactoside. Blue cells were scored as round (dead) or flat (alive) under the microscope.

RESULTS

The GenBank data base of expressed sequence tags was searched with the cDNA sequence of mouse bad. An expressed sequence tag with significant sequence homology which did not include the 5-end of mouse bad was identified (expressed sequence tag no. 171,560, GenBank accession number H18135). The corresponding plasmid was obtained from the IMAGE Consortium (info{at}image.llnl.gov) and used as a probe to screen human cDNA libraries derived from substantia nigra (Stratagene number 936210 in  ZAP II), bone marrow (CLONTECH number HL 1168x in  DR2), and placenta (Stratagene number 937225 in UNI-ZAP-XR). Sequences of several positive clones were obtained from each library. The open reading frame of the clone that extended furthest on the 5-end, derived from the human bone marrow library, is shown in Fig. 1A. It encodes a protein of 168 amino acids with 74% identity to mouse Bad (Fig. 1B). Human BAD lacks a stretch of 42 amino acids on its 5-end in this alignment. The sequence surrounding the start methionine satisfies the Kozak criteria (23) for translation initiation sites (A at position 3) (data not shown). None of the analyzed cDNA clones showed any homology to mouse bad upstream of this start codon. Like mouse bad, human BAD does not encode a C-terminal hydrophobic domain for membrane insertion which is present in other BCL-2 family proteins.

[View Larger Version of this Image (49K GIF file)]

The human BAD cDNA was cloned into the eukaryotic expression vector pCIneo (Promega) and was expressed in 293 cells in transient transfection assays to assess its ability to directly induce apoptosis. In this assay system we use cotransfection of -galactosidase to identify the transfected cell population. Apoptotic cells are then identified by their typical rounded and shrunken morphology as a percentage of total transfected cells. This methodology has been used successfully in a number of cell types including 293 cells (16, 24, 25). Compared with control plasmid, expression of human BAD led to a significant increase in apoptotic cells (Fig. 2A). A second set of experiments was performed in GM701 cells which undergo apoptosis following treatment with staurosporine (22, 26). Staurosporine-induced death is inhibited by BCL-X_L, and we sought to determine whether human BAD can abolish this BCL-X_L effect. Co-transfection of BAD with BCL-X_L completely abolishes the protective effect seen with BCL-X_L alone (Fig. 2B).

[View Larger Version of this Image (28K GIF file)]

Mouse Bad has been shown to form dimers with BCL-2 and BCL-X_L. We tested the interactions of human BAD by two methods: yeast two-hybrid and in vitro interaction assays. For the yeast two-hybrid analysis the human BAD cDNA was cloned into the plasmids pGilda and pJG4-5 of the LexA-based two-hybrid system (27). BAD was tested against BCL-2, BCL-X_L, BCL-X_S, BAK, BAX, and human BAD. Liquid o-nitrophenyl--D-galactopyranoside assays were used as a quantitative readout. BAD interacted most strongly with BCL-X_L and more weakly with BCL-2, when cloned either in the activation domain plasmid (Fig. 3A) or the DNA-binding domain plasmid (data not shown). BAD did not form homodimers, nor did it interact with BCL-X_S, BAX, or BAK (Fig. 3A).

[View Larger Version of this Image (21K GIF file)]

The yeast two-hybrid data demonstrating dimerization of BAD with BCL-2 and BCL-X_L were corroborated by direct biochemical binding experiments (Fig. 3B). In this format, one potential binding partner is absorbed as a histidine-tagged protein onto the surface of polystyrene plates. The second protein is added in liquid phase as a GST-fusion protein. A monoclonal antibody to GST is used as a reporter to detect binding of the liquid phase protein (20). Robust saturable binding was detected between BAD and BCL-X_L as well as BAD and BCL-2 (Fig. 3B), but no significant interaction was detected between BAD and BAX.

Recent reports have shown that the BH-3 domains of Bax and Bak are necessary and sufficient for dimerization with BCL-X_L as well as induction of apoptosis (16, 18, 21). In preliminary experiments, we noted that BAX was capable of blocking the binding of BAD to BCL-X_L, suggesting that BAD also bound to BCL-X_L via a BH-3 domain (data not shown). To assess whether BAD/BCL-X_L binding reflects a BH-3-mediated interaction, we evaluated whether the peptides derived from the BH-3 domains of BAK (Fig. 5, right panel) and BAX (not shown) could disrupt BAD/BCL-X_L heterodimer formation in biochemical interaction assays. A clear dose-dependent inhibition of BAD binding indicated that this dimerization was dependent on a BH-3 interaction. Motivated by these results, we next examined the human BAD sequence for the presence of BH-3 like domains. An area partially overlapping with a region previously suggested to be a BH-1 domain (14) showed significant homology to other BH-3 domains (Fig. 4). We then tested whether peptides derived from this putative BH-3 domain could inhibit heterodimer formation in in vitro binding assays. A series of peptides covering the predicted BH-3 domain of BAD was tested for their ability of block the formation of various BH-3-dependent heterodimers. The inhibition of the dimerization between BCL-X_L and BAX or BAD is shown in Fig. 5. Peptides of 25 and 21 residues in length showed complete inhibition, the 20-mer peptide showed weaker activity and the effect of the 16-mer peptide was barely detectable. Control peptides with mutations in the most conserved residues of this region had no significant effect. For comparison, a peptide derived from the BH-3 domain of BAK was also included. The effect of a 16-mer BAK peptide on BAX/BCL-X_L shown in Fig. 5 is about 5-fold less potent than previously shown (20), possibly due to a variation in the quality of a new preparation of peptide. The IC₅₀ values for inhibition of binding by BAD and BAK-derived peptides are clearly dependent on the binding partners involved. Both peptides were more potent in inhibiting the BAX/BCL-X_L than the BAD/BCL-X_L interaction.

[View Larger Version of this Image (38K GIF file)]

[View Larger Version of this Image (50K GIF file)]

A series of mutations in the BH-1 and BH-2 domains of BCL-2 and BCl-X_L that led to loss of interaction with BAX or BAK in immunoprecipitation assays have recently been described. Two such mutations in BCL-X_L had retained biological activity (BCL-X_L F131V+D133A and BCL-X_L G148E+G187A)(13). In contrast, a different set of mutations namely BCL-2 (G145A) and BCL-X_L (G138A) had lost biological activity (12, 14), but retained the ability to form homodimers (12, 21).² We have recently shown that these mutations that have lost the ability to interact with BAX or BAK all are able to bind to a truncated form of BAX that contains the BH-3 domain but not the BH-1 and BH-2 domains (21). We therefore measured the ability of these mutants to form heterodimers with BAD in a quantitative fashion using both solid-phase binding assays as well as yeast two-hybrid assays. All of the mutations had significant effects on the heterodimer formation with BAX as expected (Fig. 6B and Ref. 21). Surprisingly, all tested mutants strongly dimerize with human BAD both in yeast two-hybrid and plate binding analysis (Fig. 6, A and B). These data indicate that the mutations retain intact binding pockets for a BH-3 domain.

[View Larger Version of this Image (49K GIF file)]

DISCUSSION

To begin our analysis of BAD function and molecular interaction, we cloned a cDNA for human BAD based on an expressed sequence tags sequence identified by homology with the previously described mouse sequence. Sequence identity of 74% was found between the mouse and the human clones and is in line with the general level of sequence conservation between these two species, although it is significantly lower than the values observed between mouse and human sequences for BCL-2, BCL-X_L, and BAX. The lack of a 42-amino acid fragment at the N terminus of the human protein compared with mouse prompted us to do an extensive search for possible splice variants containing this sequence. The isolation and sequencing of 25 cDNA clones from three different cDNA libraries did not yield any clones with homology to the missing sequence. Attempts to isolate the sequence by anchor PCR from libraries and from first strand cDNA from Jurkat cell RNA were also unsuccessful. Although we have not found any indication for the presence of a longer BAD cDNA with these methods, final resolution will come from protein analysis of wild type BAD protein produced in cells compared with proteins produced from the isolated cDNA.

Human BAD was biologically active in two types of transient transfection assays. In 293 cells direct induction of apoptosis by BAD was observed. Although these types of assays are commonly used to assess the apoptosis promoting activity of transfected genes, they suffer from the drawback that numerous proteins are able to induce apoptosis when expressed to high enough levels independent of their biological role under physiologic conditions. An example of this phenomenon is the artificial expression of extracellular proteases such as trypsin or chymotrypsin. We therefore confirmed the activity of BAD in a luciferase reporter-based survival assay in GM701 cells and demonstrated that BAD is able to counteract the activity of BCL-X_L, consistent with data previously obtained with mouse Bad in FL 5.12 cells (14).

Human BAD was found to form heterodimers with BCL-X_L and BCL-2 both in yeast two-hybrid interactions and in solid-phase binding assays. As in the mouse system, the interaction with BCL-X_L was generally observed to be stronger. There was no indication for homodimer formation nor was there detectable interaction between BAD and BCL-X_S, BAX or BAK. The observation that binding of BAD to BCL-X_L could be inhibited by peptides derived from the BH-3 domains of BAK and BAX prompted us to look for a BH-3 like domain in the sequence of BAD. Computer alignments of BAD with the full-length sequences of BCL-2 family proteins show weak homologies with the BH-1 and BH-2 domains of these proteins as published for mouse Bad (14). However, when the BH-3 domains of BCL-2 family proteins are aligned with BAD, an area of obvious homology emerges in the same region that shows homology with BH-1 domains. Peptides synthesized from this region are able to completely inhibit the heterodimerization of BAD, indicating that this motif is the interacting domain in dimer formation. As there are no other recognizable homologies with the BH-1 and BH-2 domains of BCL-2 family proteins, our data suggest that BAD contains only a BH-3 domain as is the case for the BCL-2-binding proteins BIK (28), BID (29), and HRK (30). Since BH-1 and BH-2 domains are critical to the formation of the hydrophobic binding groove in BCL-X_L, our results suggest that BAD does not contain such a binding site. This is in accord with the inability to observe BAD/BAD homodimers since BH-1 and BH-2 domains are required for all homodimer reactions studied to date (12, 20). It is also in accord with the failure to observe strong binding of BAD to BAX and BAK, as human BAX (21, 31) and BAK (9) do not form homodimers.

The published data regarding the BCL-2 G145A mutation has been interpreted to suggest that the heterodimerization site on BCL-2 must be intact for biological activity. These data are are seemingly in conflict with the BCL-X_L F131V+D133A and BCL-X_L G148E+G187A mutations that indicate that the heterodimerization site on BCL-X_L can be destroyed but biological activity is retained. Surprisingly, here we show that all mutants with supposedly destroyed heterodimerization sites are still able to strongly bind BAD. The peptide inhibition data discussed above indicate that BAD, BAK, and BAX form heterodimers by the same BH-3-dependent mechanism. Additional recent data with dimerization blocking peptides derived from BAX indicate that BCL-2 family homodimers are also mediated through BH-3 interactions (20). Taken together, these data indicate that the analyzed mutations do not destroy the dimerization domain but rather alter its selectivity. Thus, generalizations regarding the ability of specific mutants to bind pro-apoptotic versus anti-apoptotic BCL-2 family members are not warranted.