Movement of Bax from the Cytosol to Mitochondria during Apoptosis

Except as noted, reagents were purchased from Sigma Chemical Co. (St. Louis, MO).

Full-length and COOH-terminal truncated forms (ΔCT) of human Bcl-2, Bcl-X_L, and Bax were synthesized by PCR amplification and cloned into the pcDNA-3 mammalian expression vector (Invitrogen, Carlsbad, CA). A similiar strategy was used for GFP-fusion constructs, using the C3-EGFP plasmid (Clontech Laboratories, Inc., Palo Alto, CA). The cDNAs for human bcl-2, human bcl-X_L, and human bax were gifts of C.M. Croee (Jefferson Cancer Center, Philadelphia, PA), C.B. Thompson (HHMI and University of Chicago, Chicago, IL), and S.J. Korsmeyer (HHMI and Washington University, St. Louis, MO), respectively. The bcl-2 ΔCT, bcl-X_L ΔCT, and bax ΔCT forms constructed encode for proteins which have a deletion of 23, 25, and 21 amino acids at their COOH-terminal ends, respectively. For full-length constructs of each gene, oligonucleotides placing a restriction enzyme site immediately adjacent to the start codon (forward primers) or stop codon (reverse primers) were synthesized (GIBCO BRL, Gaithersburg, MD). For the truncated (ΔCT) forms, a stop codon was introduced by the reverse primers. PCR was performed using Pfu polymerase (Stratagene, La Jolla, CA) to minimize errors. PCR fragments were purified, digested with restriction enzymes, and ligated into the expression vectors. The bcl-X_L, bcl-X_LΔCT, bcl-2, and bcl-2 ΔCT genes were subcloned between the XhoI and XbaI sites in the polycloning regions of both pcDNA3 and C3-EGFP. The bax and bax ΔCT genes were subcloned between the HindIII and EcoRI sites in the polycloning regions of both pcDNA3 and C3-EGFP. All constructs were sequenced using the Sequenase 2 kit (United States Biochemical Corp., Cleveland, OH).

L929 murine fibrosarcoma and Cos-7 green monkey renal epithelia cell lines (American Type Culture Collection, Rockville, MD) were grown in Minimum essential medium, Eagle's in Earle's balanced salt solution and DME, respectively, each supplemented with 2 mM glutamine, 1× nonessential amino acids, 2.5 mM sodium pyruvate, 100 U/ml penicillin, 100 μg/ ml streptomycin (all from Biofluids Inc., Rockville, MD), 50 μg/ml gentamicin (GIBCO BRL), and 10% heat-treated FCS. Cells were cultured at 37°C in 5% CO₂ and then split twice weekly using 0.05% trypsin/0.02% versene (Biofluids Inc.). In preparation for transfection, cells were plated at a density of 2 × 10⁵ cells per well in six-well tissue-culture plates (Becton Dickinson Labware, Bedford, MA). Either one (Cos-7) or two (L929) days later, cells were transiently transfected using the cationic lipid LipofectAmine (GIBCO BRL) as described by the manufacturer, using 2 μg of plasmid DNA and either 8 (L929) or 10 μl (Cos-7) of LipofectAmine per well. After 5 h in serum-free medium, an equal volume of medium containing 2× normal FCS was added. 1 d after transfection, the cells were washed once and placed in normal growth media. Transfected cells were then cultured for an additional 24 h, and visualized by microscopy. For cotransfection experiments (see Fig. 7), the above procedure was followed, except that a total of 4 μg of plasmid DNA was used per well: 1 of C3–EGFP–Bax and 3 of either pcDNA3–Bcl-2 or pcDNA3–Bcl-X_L.

A peptide corresponding to amino acids 41–54 of human Bcl-2 (GAAPAPGIFSSQPGC; Peptide Technologies Corp., Gaithersburg, MD) was conjugated to maleimide-activated keyhole limpet hemocyanin (KLH) (Pierce Chemical Co., Rockford, IL) through the cysteine residue according to the protocol provided by the manufacturer. Mice were immunized with the peptide–KLH conjugate and splenocytes from immunoreactive mice were fused by PEG 4,000 to murine NS-1 myeloma cells and selected with histone acetyltransferase (HAT) medium as previously described (15). The anti–human Bcl-2 antibody was designated αhBcl-2 8C8.

Transfected cells from two wells of a six-well plate were washed once with PBS and then lysed in cold solubilization buffer (PBS, 25 μg/ml phenylmethylsulphonyl fluoride, 1 μg/ml leupeptin, and 1 μg/ml aprotinin) containing 1% Triton X-100. For Western blotting, 20 μl of each sample were separated by electrophoresis using 12% SDS-PAGE gels. Gels were electrotransferred onto Immobilon-P (Millipore Corp., Waters Chromatography, Milford, MA) membranes. The blots were blocked in PBS/0.05% Tween 20 containing 5% fetal bovine serum and incubated with the appropriate antibody: anti–human Bcl-2 8C8, anti-universal Bcl-X_L 2H12 (15), or anti–human Bax 2D2 monoclonal antibodies (15). Primary antibody binding was detected by blotting with sheep anti–mouse F(ab') linked to horseradish peroxidase (Amersham Corp., Arlington Heights, IL), followed by band visualization using enhanced chemiluminescence as described by the manufacturer (Amersham Corp.).

Cells used for confocal microscopy were transfected as described above on 1.5 thickness glass coverslips pretreated with 5 μg/ml poly-l-lysine. After 24–48 h, cells were prepared for microscopy by incubation with 20 ng/ ml of a mitochondrion-specific dye (Mitotracker Red CMXRos; Molecular Probes Inc., Eugene, OR) and/or 100 ng/ml bis-benzamide for 30 min, as indicated in individual experiments. After pretreatment, cells were transferred to a sealed mounting chamber containing fresh medium; in indicated experiments, this medium was supplemented with 1–3 μM staurosporine (STS). Images were collected on a microscope (model LSM 410 with a 40× 1.2 NA Apochromat objective; Carl Zeiss, Thornwood, NY). The 488- and 568-nm lines of a krypton/argon laser were used for fluorescence excitation of GFP and Mitotracker Red CMXRos, respectively, and the 364-nm line of an argon laser was used for excitation of bis-benzamide dye. The temperature of the specimen was maintained between 35° and 37°C with an air stream incubator.

To investigate the mobility of the GFP-fusion proteins, the fluorescence within a region of a cell was photobleached by scanning the region with the 488-nm laser line at full power (no attenuation) and then the entire cell was scanned at 10-s intervals with low laser power (10% power, 0.3% attenuation) to monitor the recovery of fluorescence. The fraction of the fluorescent protein that was mobile was calculated as described (4) by comparing the fluorescence intensities in two regions of the cell, one inside the photobleach zone and the other elsewhere, before and 100 s after the photobleach (mobile fraction = ratio after 100-s recovery/ratio before photobleach). The time course of fluorescence recovery was determined by photobleaching a 16-μm strip within a flattened area of the cell and then monitoring the recovery of fluorescence within the strip (200 images at 0.46-s intervals).

To assess the effects of the full-length and ΔCT forms of Bcl-2, Bcl-X_L, and Bax, on cell viability, a cotransfection assay was performed in L929 or Cos-7 cells. Wells plated as described above were cotransfected with one vector construct containing a given gene in pcDNA3, and the second vector, EGFP (Clontech Laboratories Inc.), containing a gene for a red-shifted GFP. A 3:1 molar ratio of the former to the latter was used, with a total of 4 μg of DNA per well. Otherwise, cells were transfected as above. At 36–48 h after transfection, cells were washed twice, fresh culture medium was added, and the number of GFP-positive cells in a defined field was determined by examination with a fluorescence microscope. Apoptosis was then induced by the addition of medium containing STS at the concentrations indicated, and the number of GFP-positive cells within the same field was assessed at regular intervals over the next 24–42 h. To test the effect on bioactivity of adding GFP to the NH₂ termini of the Bcl-2 family members, a similar assay was performed on cells transfected with 2 μg of the appropriate full-length or ΔCT gene fused to GFP inserted into C3-EGFP. For each construct, the assay was repeated three or four times. To normalize for variations in transfection efficiency, changes in viability are expressed as percentage decrease in the number of fluorescent cells within a given field over time.

We examined the location of Bcl-2, Bcl-X_L, and Bax in individual living cells by using GFP fusion constructs. Fusion constructs containing GFP have been successfully used to study the subcellular localization of a number of proteins (16, 26, 33). The gene encoding a high fluorescence, red-shifted GFP variant (5) was fused to the 5′ termini of cDNAs encoding human Bcl-2, Bcl-X_L, or Bax. L929 murine fibrosarcoma and Cos-7 green monkey kidney epithelial cells were transfected with the fusion constructs encoding GFP–Bcl-2, GFP–Bcl-X_L, and GFP–Bax.

The localization of these transiently expressed GFP– linked fusion proteins was visualized in living cells using confocal microscopy. In both Cos-7 (Fig. 1, a and b) and L929 cells (Fig. 2, a and b), GFP alone has a diffuse distribution, filling the entire cell as previously reported (28). GFP–Bcl-2 displays a punctate pattern of distribution, primarily in areas near the nucleus, in both Cos-7 (Fig. 1, e and f) and L929 (Fig. 2, e and f) cell lines. This distribution matches the localization determined previously for Bcl-2 by immunofluorescence (24, 41), and suggests association with an intracellular organelle or organelles. GFP–Bcl-X_L displays a bright punctate distribution similar to that of GFP– Bcl-2 superimposed upon a fainter diffuse background (Figs. 1 and 2, i and j), suggesting both soluble and membrane-bound populations in individual cells. This is consistent with the dual localization of Bcl-X_L in both the soluble and membrane pellets in cellular fractionation experiments (14, 15). In contrast to the localization of GFP–Bcl-2 and GFP–Bcl-X_L, GFP–Bax has a diffuse distribution similar to that of GFP alone in both Cos-7 (Fig. 1, m and n) and L929 (Fig. 2, m and n) cell types.

We examined the GFP fusion protein expression by Western blot analysis of transiently transfected Cos-7 cells 36 h after transfection. Immunoblotting of these lysates confirmed that GFP–Bcl-2 (Fig. 3, middle, arrow a), GFP– Bcl-X_L (Fig. 3, bottom, arrow a), and GFP–Bax (Fig. 3, top, arrow a) are of appropriate size. Low molecular weight bands (lane c) representing native Bcl-2, Bcl-X_L, and Bax were detected in all lysates. An additional, unexpected protein band (arrow b) of intermediate molecular weight was detected for each expression construct (Fig. 3). This intermediate molecular weight form was also found when COOH-terminal truncated forms of Bcl-2 (lane c), Bcl-X_L (lane e), and Bax (lane a) fused to GFP were expressed and immunoblotted. When compared to full-length forms, all three COOH-terminal truncated proteins ran at the expected slightly lower molecular weights (Fig. 3). The COOH-terminal truncated constructs display slightly lower molecular weights for both the high molecular weight and intermediate molecular weight bands, indicating that the intermediate weight bands likely represent proteins shortened from the NH₂-terminal, GFP end, of the fusion protein. Thus, the intermediate size bands may result from expression of transcripts initiated at an internal start codon within the GFP gene or proteolytic fragments lacking part of the GFP molecule, and thus they may or may not fluoresce.

To determine if the fusion of GFP to Bcl-2, Bcl-X_L, and Bax affected their bioactivities, we compared the viability of either L929 or Cos-7 cells transfected with wild-type cDNA constructs to those transfected with GFP-fusion constructs after induction of apoptosis with STS. As shown in Fig. 4, the protective activities of GFP–Bcl-2 (Fig. 4,A) and GFP–Bcl-X_L (Fig. 4,B) in L929 cells are comparable to those of the wild-type forms. Likewise, the activity of GFP–Bax appears to accelerate cell death of Cos-7 cells as effectively as does wild-type Bax (Fig. 4 C). Thus, adding GFP to the NH₂-terminus of Bcl-2, Bcl-X_L, or Bax does not block their abilities to influence apoptosis. However, the addition of a large protein to the NH₂ terminus may alter the potency of Bcl-2 family members to regulate apoptosis or influence their intracellular localization.

To investigate the identity of the organelle or organelles to which the GFP–Bcl-2 and the majority of the GFP–Bcl-X_L localize, we treated transiently transfected Cos-7 cells with Mitotracker Red CMXRos. As shown in Fig. 5 (a–c), the pattern of GFP–Bcl-2 distribution coincides very closely with that of the mitochondrial marker, demonstrating that most of the GFP–Bcl-2 is found associated with mitochondria. The punctate fraction of the GFP–Bcl-X_L also distributes to mitochondria (Fig. 5, d–f). These results match earlier findings localizing Bcl-2 and Bcl-X_L to the membrane of mitochondria using other imaging modalities (19, 24). In constrast to GFP–Bcl-2 and GFP–Bcl-X_L, GFP– Bax does not localize to mitochondria in healthy Cos-7 cells (Fig. 5, g–i).

The similarity of the distribution of GFP–Bax to the distribution of GFP led us to investigate whether GFP–Bax is free to diffuse within the cytoplasm, as GFP has been shown to be (34). The mobilities of GFP–Bax and GFP were compared by photobleaching the fluorescence in a small region of the cell and then monitoring the recovery of fluorescence due to movement of unbleached molecules into the bleached zone. Both GFP and GFP–Bax moved rapidly into the bleached zone (Fig. 6). The time courses for fluorescence recovery of GFP and GFP–Bax were essentially identical (Fig. 6 B), and were consistent with diffusion kinetics. The proteins diffused too rapidly relative to the rates at which we could photobleach and monitor recovery to allow us to accurately determine diffusion constants.

Photobleaching experiments also were performed on cells expressing GFP–Bcl-2 (Fig. 6). In contrast to GFP and GFP-Bax, GFP-Bcl-2 appeared to be largely immobile. The mobile fraction (percentage recovery) at 100 s after photobleaching (calculated as described in Materials and Methods) was 9 to 13% for GFP–Bcl-2 (n = 3) as compared with 45 to 85% for GFP–Bax (n = 4), and 82 ± 2% for GFP alone (34). The low mobility of GFP–Bcl-2 supports the conclusion that it resides in discrete organelles (mitochondria) rather than in a continuous membrane system such as the endoplasmic reticulum (where fluorescence recovers rapidly after photobleaching [4]).

Recent studies have proposed that Bax (32) and Bad (40) require Bcl-2 to associate with organelle membranes. To test the possibility that GFP–Bax localization to organelles is limited by the endogenous level of either Bcl-2 or Bcl-X_L, vectors expressing either Bcl-2 or Bcl-X_L were cotransfected with the GFP–Bax vector. As shown in Fig. 7,A, coexpression of Bcl-2 at high levels, verified by immunoblotting of cell lysates (Fig. 7 B), did not alter the cytoplasmic distribution of GFP–Bax. Similar results were obtained when Bcl-X_L was coexpressed with GFP–Bax (data not shown). Thus, we find Bax is a soluble, cytosolic protein in healthy cells, which does not bind organelles even in the presence of elevated levels of Bcl-2 or Bcl-X_L.

Endogenous Bax and some of the endogeneous Bcl-X_L migrate from the supernatant to the membrane pellets in cell fractionation experiments performed on thymocytes undergoing apoptosis (14). To investigate this phenomenon in living cells, Cos-7 cells were transiently transfected with native GFP, GFP–Bcl-2, GFP–Bcl-X_L, and GFP–Bax. Cells expressing each construct were treated with 1 μM STS, and after 6 h both STS-treated and untreated cells were examined by confocal microscopy. STS-treated cells underwent cell shrinkage and blebbing of the cell membrane, a morphology indicative of apoptosis (Fig. 1, compare b, f, j, and n with d, h, l, and p). GFP–Bax markedly changed its intracellular distribution, relocating within cells during apoptosis from a diffuse (Fig. 1,m) to a punctate pattern (Fig. 1,o). This change was not observed with GFP alone (Fig. 1,c). No changes in the patterns of distribution of either GFP–Bcl-2 (Fig. 1,g) or GFP–Bcl-X_L (Fig. 1,k) were observed in treated cells beyond those related to changes in cell size and shape. It is possible that redistribution of soluble GFP–Bcl-X_L is reduced to some extent in this system due to the ability of overexpressed GFP–Bcl-X_L to inhibit apoptosis (Fig. 4,B). Similiar experiments performed using another cell line, L929 fibroblasts, confirmed that GFP–Bax undergoes a shift from a diffuse to punctate distribution pattern during apoptosis whereas the localization patterns of GFP–Bcl-2 and GFP–Bcl-X_L remain essentially unchanged (Fig. 2).

Mitochondrial staining of GFP–Bax–expressing Cos-7 cells undergoing apoptosis suggests that much of the GFP–Bax attaches to or enters mitochondria (Fig. 8, a–c). However, as there are also sites of Bax condensation that do not stain for mitochondria, it is possible that GFP–Bax is also localizing to other organelles upon induction of apoptosis or that mitochondria lose their Mitotracker Red CMXRos staining during apoptosis.

The redistribution of Bax to organelles induced by STS treatment is associated with a reduction in Bax mobility (Fig. 6). FRAP experiments on cells in which GFP–Bax had a punctate (mitochondrial) localization indicated that the mobile fraction was 10 to 30% (n = 2), as compared with 45 to 85% in untreated cells. Thus, the punctate form of GFP–Bax is less mobile than the diffuse form of GFP–Bax consistent with the model that Bax redistributes from a cytoplasmic to membrane-bound localization during apoptosis.

To relate the timing of the shift in GFP–Bax localization to other morphological events that occur during apoptosis, individual Cos-7 cells expressing GFP–Bax were followed by confocal microscopy over the course of STS-induced cell death. The change in GFP-Bax localization within individual cells occurred swiftly, over the course of ∼30 min after initiation (Fig. 9, c–e). Various cells in a given field may begin the process of GFP–Bax redistribution at different times, as seen in Fig. 9. At the time point when GFP–Bax begins redistribution, little alteration in the overall structure of the cell is evident when viewed by differential interference contrast (DIC) microscopy (Fig. 9,h), or by interference reflection microscopy that shows the cell– substrate interaction (data not shown). At 10–30 min after the initiation of Bax redistribution, cell shrinkage first became apparent by DIC imaging (Fig. 9, i and j, arrows).

Cells expressing GFP–Bax were incubated with the nuclear stain bis-benzamide (Hoechst 33342), which labels DNA and is commonly used to detect the nuclear changes characteristic of apoptosis. At low concentrations, this stain can be used to study the location of DNA within living and even dividing cells without affecting cell viability (8). After bis-benzamide staining, cells were then treated with STS to induce apoptosis. As shown in Fig. 10, there is no evidence of nuclear changes when GFP-Bax initially begins redistribution. At ∼3 h after the initiation of the GFP–Bax transition, a more uneven pattern of bis-benzamide staining can be seen in the nuclei of cells (Fig. 10,e), suggestive of nuclear condensation. This gradually progresses to nuclear fragmentation, clearly apparent by 12 h after initiation of GFP–Bax redistribution (Fig. 10 f). Therefore, the shift in GFP–Bax distribution appears to be an early event in apoptosis, preceding several established morphological changes associated with apoptosis.

The COOH-terminal hydrophobic region, spanning the last 20–24 amino acids of Bcl-2, Bcl-X_L, and Bax is thought to act as an anchor localizing these proteins to the membranes of organelles. As our results suggest that Bax localizes to organelles only upon induction of apoptosis, we investigated the role of the COOH terminus in Bax trafficking. ΔCT constructs of GFP–Bcl-2 (Fig. 11, a and c), GFP-Bcl-X_L (Fig. 11, e and g), and GFP–Bax (Fig. 11, i and k) all display a diffuse distribution in healthy cells, consistent with localization throughout the cytosol, when expressed in either Cos-7 or L929 cell lines. Induction of apoptosis causes no apparent redistribution of any of the ΔCT constructs in either cell line (Fig. 11). This indicates that the COOH terminus may play a pivotal role not only in anchoring Bcl-2 and Bcl-X_L to organelles in healthy cells, but also in mediating the organelle association of Bax observed in cells undergoing apoptosis.

If Bax insertion into organelles is required for death promoting activity, then Bax ΔCT would be expected to have diminished potential to enhance apoptosis relative to wild-type Bax. The bioactivity of ΔCT constructs of Bax, Bcl-2, and Bcl-X_L was compared to that of the full-length genes in a cell viability assay both with and without GFP fused to the NH₂ terminus of each protein upon induction of apoptosis with STS. As illustrated in Fig. 4, A and B, for both Bcl-2 and Bcl-X_L, deletion of the COOH-terminal hydrophobic domain resulted in a decrease in the apoptosis protection in L929 cells. For Bax, the deletion of the COOH-terminal domain also resulted in a sharp decrease in the ability of both wild-type and GFP-linked protein to accelerate apoptosis in both Cos-7 (Fig. 4 C) and L929 cells (data not shown). These results indicate that the COOH-terminal hydrophobic domain is important for Bax function, linking the ability of Bax to redistribute to organelles during apoptosis with its death-accelerating activity.

Bax promotes programmed cell death in mammalian cells (18, 27). However, Bax expression per se, at least at physiological levels, is not lethal to cells. For example, sympathetic neurons express high levels of bax mRNA, yet will not undergo apoptosis unless deprived of growth factors (6). As initially characterized, even cells stably overexpressing Bax proliferate normally (27), and Bax only accelerates cell death after an external signal, such as interleukin-3 withdrawal. As protein levels of Bax do not change during apoptosis (14, 27), it seems plausible that a posttranslational activation of Bax occurs during apoptosis. Our findings indicate that Bax activity may be regulated at the level of intracellular localization, with Bax moving to the membranes of organelles after induction of apoptosis, but before nuclear fragmentation. Elimination of 21 amino acids from the COOH terminus prevents Bax redistribution and abrogates Bax apoptosis-promoting activity, correlating organelle binding with bioactivity.

In contrast to cytosolic GFP–Bax, GFP–Bcl-2 displays a punctate distribution in healthy cells that coincides very strongly with a mitochondrial marker. GFP–Bcl-X_L demonstrates a mixed pattern, suggesting that although much of the protein is associated with mitochondria, a significant amount is found diffusely throughout the cytosol. In cell fractionation studies, this soluble Bcl-X_L found in healthy cells redistributes to associate with the membrane during apoptosis (14). No change in distribution is observed for either GFP–Bcl-2 or GFP–Bcl-X_L after the addition of STS. However, the levels of GFP–Bcl-X_L in the transfection experiments are sufficient to delay or even block apoptosis (Fig. 4); thus, GFP–Bcl-X_L may actually prevent alterations in the intracellular milieu that would lead to its redistribution. Further experiments will be necessary to explore the intracellular distribution of GFP–Bcl-X_L during apoptosis, perhaps by using apoptosis pathways that are Bcl-X_L insensitive.

For all three of the Bcl-2 family members examined here, the COOH-terminal deletion eliminates the ability of GFP-tagged proteins to associate with organelles. The dependency of Bax redistribution on the COOH-terminal hydrophobic domain suggests that a conformational change or other modification might occur in Bax during apoptosis, allowing the previously inaccessible or blocked hydrophobic tail to insert into membranes. It is striking that the translocation domain of diphtheria toxin, which has structural homology to Bcl-X_L, displays a similar ability to change from a soluble to membrane-associated state. In the case of the diphtheria toxin translocation domain, this insertion is driven by a conformational change triggered by low pH (7, 30). It remains unclear how Bax insertion into membranes is controlled. One interesting candidate is phosphorylation: a regulatory mechanism involving phosphorylation has recently been described that influences both localization and activity of the apoptosis promoter Bad (40), and phosphorylation of Bcl-2 has been suggested to regulate its function (10, 17, 22).

Our findings conflict with previous reports that suggest that Bax is constitutively localized to organelle membranes (for review see reference 29). One possible explanation for this discrepancy may be that in earlier work, Bax localization was examined in contexts where overexpression of Bax alone was sufficient to cause apoptosis, without the overt induction of apoptosis. In a study examining Bax localization in a transformed BRK cell line, Bax overexpression led directly to cell death (11). A recent paper on Bax localization to mitochondria (39) reported that the expression of fusion proteins containing full-length Bax was lethal to yeast cells, and removal of the COOH terminus abolished the ability of Bax to kill yeast cells and localize to yeast mitochondria. We have found that cationic lipid transfection procedures can themselves trigger apoptosis (Wolter, K.G., and X.-G. Xi, unpublished observations). Deletion of the COOH-terminal tail from GFP-Bax eliminated this increase in transfection-associated death. High levels of full-length GFP–Bax expression may, therefore, lead directly to the death of transfected cells, particularly if this expression closely follows the cellular stress of the transfection procedure. Perhaps some previous observations of organelle-located Bax were made on cells entering an apoptosis program. These reports and our own data concur in suggesting that Bax localization to organelles is a vital aspect of its death promoting ability.

However, our findings on the role of Bcl-2 in Bax localization clearly differ from those of Shibasaki and co-workers (32), who observed a redistribution of soluble Bax to a punctate pattern in BHK cells expressing high levels of Bcl-2. In our systems, we see no evidence for a localization of Bax to membranes driven by either Bcl-2 or Bcl-X_L coexpression. This is consistent with earlier findings that Bax and Bcl-X_L do not form heterodimers in healthy cells (15).

Recent reports indicate that Bcl-2 family members control the release of cytochrome C and apoptosis-inducing factor from mitochondria during apoptosis and contribute to cell death by caspase activation (20, 21). The overexpression of Bcl-2 has been reported to block the mitochondrial release of cytochrome C, and in doing so prevents apoptosis (38). These results are interesting, given the structural connection made between Bcl-X_L and the diphtheria toxin translocation domain. This domain is capable of forming transmembrane ion channels, and is furthermore believed to multimerize to form pores that mediate the translocation of the catalytic domain of the diphtheria toxin across membranes. Bcl-X_L and Bcl-2 have been reported to form transmembrane channels (23, 31). In this context, it is possible that upon Bax insertion into membranes, Bax may form channels or pores allowing for the release of factors such as cytochrome C from within the mitochondria to propagate the apoptotic pathway. This model does not preclude the interaction of Bax with Bcl-2 and Bcl-X_L upon membrane insertion. In the future, it will be crucial to elucidate the trigger regulating Bax insertion into organelle membranes and the molecular consequences of membrane-bound Bax.

We would like to thank C. Croce, C. Thompson and S. Korsmeyer for the gifts of Bcl-2, Bcl-X_L, and Bax genes, respectively. We thank R. Kocher, J. Castelli, X.-H. Liu, C. Ng, and P. Johnson for assistance, and L. Marrone (all from National Institutes of Health [NIH], Bethesda, MD), and K. Wood (Trevigen Inc., Gaithersburg, MD) for reading the manuscript.

ΔCT

COOH-terminal truncation

DIC

differential interference contrast

GFP

green fluorescent protein

STS

staurosporine