Mechanistic Differences in the Membrane Activity of Bax and Bcl-xL Correlate with Their Opposing Roles in Apoptosis

Abstract

Based on their membrane-permeabilizing activity in vitro, it has been proposed that Bax-like proteins induce cytochrome c release during apoptosis via pore formation. However, antiapoptotic Bcl-2 proteins, which inhibit cytochrome c release, also display pore activity in model membranes. As a consequence, a unified description that aligns the pore activity of the Bcl-2 proteins with their apoptotic function is missing. Here, we studied the mechanism of membrane binding, oligomerization, and permeabilization by pro- and antiapoptotic Bcl-2 members at the single-vesicle level. We found that proapoptotic Bax forms large, stable pores via an all-or-none mechanism that can release cytochrome c. In contrast, antiapoptotic Bcl-xL induces transient permeability alterations in pure lipid membranes that have no consequences for the mitochondrial outer membrane but inhibit Bax membrane insertion. These differences in pore activity correlate with a distinct oligomeric state of Bax and Bcl-xL in membranes and can be reproduced in isolated mitochondria. Based on our findings, we propose new models for the mechanisms of action of Bax and Bcl-xL that relate their membrane activity to their opposing roles in apoptosis and beyond.

Introduction

The members of the Bcl-2 family are multifunctional proteins. They are major regulators of apoptosis, and they have recently also been associated with a number of additional roles, including the regulation of mitochondrial energy metabolism and dynamics, cellular Ca²⁺ homeostasis, autophagy, and the unfolded protein response at the endoplasmic reticulum (ER) (1). Bcl-2 homologs, including Bid, Bax, and Bcl-xL, have very similar cytosolic folds and share the rare ability to switch from water-soluble to membrane-inserted conformations. Indeed, most functions of the Bcl-2 proteins are performed when associated with cellular membranes, but the molecular mechanisms involved remain unclear. For this reason, understanding how the Bcl-2 proteins behave in lipid membranes is crucial to identifying the molecular mechanisms behind their biological functions.

Currently, the regulation of apoptosis by the Bcl-2 proteins is their best-characterized function. Bax and Bak induce permeabilization of the mitochondrial outer membrane (MOM) that releases cytochrome c (cyt c) during apoptosis (2). They are activated by some BH3 proteins like Bid and inhibited by the antiapoptotic Bcl-2 proteins, like Bcl-xL. Despite intense research, the mechanism of Bax-induced MOM permeabilization remains controversial. During apoptosis Bax translocates to the mitochondria and undergoes a conformational change (3). This leads to its extensive insertion into the MOM, where it oligomerizes and forms a pore that is responsible for cyt c release (4,5). The pore hypothesis was proposed based on structural similarities between the soluble forms of the Bcl-2 proteins and the channel-like translocation domain of diphtheria toxin (6). Indeed, Bax exhibits pore activity in vitro. In the presence of cleaved Bid (cBid), Bax releases fluorescent molecules (including cyt c and dextrans of several hundred kDa) from liposomes with high efficiency and without altering vesicle integrity (7–9). In addition, Bax forms large channels in vesicles and in black lipid membranes with properties of lipid pores, also known as toroidal pores (8,10).

However, antiapoptotic Bcl-2 proteins also permeabilize model membranes (11–16). Current models for the action of the Bcl-2 proteins in apoptosis do not explain these observations. To reconcile the pore activity of antiapoptotic Bcl-2 proteins with their prosurvival function in apoptosis, some ideas proposed are that these proteins regulate other channels, such as ceramide, voltage-dependent anion, or ER channels (17–20). Their inability to oligomerize could also play a role. However, the mechanistic connection between oligomerization and pore formation remains unexplained (21). The lack of experimental evidence integrating the membrane activity of pro- and antiapoptotic proteins according to their function has hindered resolution of this long-standing problem.

Our aim in this work was to provide a better understanding of the general mechanisms of action of the Bcl-2 proteins and to tackle concretely the paradox between the pore activities of pro- and antiapoptotic Bcl-2 proteins in model membranes and their opposing roles in apoptosis. For this purpose, we used a new method based on giant unilamellar vesicles (GUVs). This single-vesicle approach provides an unprecedented level of detail in the analysis of the membrane-binding, oligomeric state and pore activity of Bax and Bcl-xL. For this purpose, a chemically defined system that avoids unwanted effects of unknown components and allows the use of advanced microscopy techniques is essential to discern the mechanistic differences between Bax and Bcl-xL membrane activity.

Our results provide, to our knowledge, novel information about the molecular mechanism of membrane interaction and pore formation of Bax and Bcl-xL. cBid binding to membranes induces the insertion and activation of both Bax and Bcl-xL. On the one hand, Bax follows an all-or-none mechanism of membrane permeabilization. It is important to note that, as we show, it oligomerizes while forming stable pores. On the other hand, Bcl-xL is unable to self-assemble and induces a graded permeabilization of the vesicles due to transient alterations of the membrane permeability. Moreover, we found that once at the membrane Bcl-xL inhibits Bax membrane insertion. These findings are consistent with the roles of these two proteins in apoptosis: while Bax self-organizes into long-lived pores large enough to release cyt c, Bcl-xL transient permeabilization of pure lipid bilayers has no consequences in the context of the MOM, which is naturally permeable to molecules <5 kDa. The physiological relevance of these observations was further confirmed in a new MOM-permeabilizing assay using isolated mitochondria. Interestingly, these features of the membrane activity of Bax and Bcl-xL are likely to be relevant also for their nonapoptotic roles. Based on our findings, we propose to our knowledge, new models for the membrane activity of Bax and Bcl-xL that incorporate biophysical considerations and, for the first time, integrate the different effects of these proteins on MOM permeability during apoptosis.

Materials and Methods

Protein production and labeling

Full-length mouse Bid and human Bax, and single-cysteine mutants Bid C30S and Bax S4C, C62S, and C126S, as well as full-length human Bcl-xL C151A and S4C, were expressed in Escherichia coli and purified and, in the case of cBid, cleaved as described in Desagher et al. (3) and Bleicken et al. (5). We used the same purification strategy for full-length Bcl-xL. Protein quality was checked by sodium dodecyl sulfate polyacrylamide gel electrophoresis and liquid chromatography-mass spectroscopy (micrOTOF LC, Bruker Daltonics, Bremen, Germany). Alexa (Invitrogen, Darmstadt, Germany) or Atto (Atto TEC, Siegen-Weidenau, Germany) dyes were covalently attached to the cysteines of the mutants as described by the manufacturer. Proteins labeled with Alexa488 or Atto488 are named cBid_G, Bax_G, or BcL-xl_G, and proteins labeled with Alexa633 or Atto655 are named cBid_R, Bax_R, or Bcl-xL_R. cBid was labeled with Alexa dyes and Bax and Bcl-xL with Atto dyes. Bovine cyt c (Sigma, Munich, Germany) was labeled at lysines with Alexa488 (cyt c-Al488) or Alexa633 (cyt c-Al633). Excess of label was removed with desalting columns (BioRAD, Munich, Germany). The activity of the labeled proteins was controlled with calcein release assays (Fig. S1).

Calcein release assays

Egg PC and heart CL were from Avanti Polar Lipids (Hamburg, Germany). Large unilamellar vesicles were prepared as in García-Sáez et al. (22). Lipids were mixed in chloroform, dried to thin films, and resuspended at 4 mg/ml in 80 mM calcein (Sigma) neutralized with NaOH, followed by repeated freezing and thawing and extrusion through 100-nm polycarbonate membranes. Nonencapsulated calcein was removed with PD10 columns (GE Healthcare, Garching, Germany). Calcein release was detected on an Infinite M200 plate reader (Tecan, Mainz, Germany) by the increase in fluorescence intensity.

Composition of the MOM-like lipid mixture

The lipid mixture mimicking the MOM composition was prepared as in Lovell et al. (9) and Bleicken et al. (23) with 46% egg PC (L-α-phosphatidylcholine), 25% egg PE (L-α-phosphatidylethanolamine), 11% bovine liver PI (L-α-phosphatidylinositol), 10% 18:1 phosphatidylserine, and 8% cardiolipin (all percentages w/w).

Membrane binding and permeabilization

Giant unilamellar vesicles (GUVs) were produced by electroformation (12). Lipid mix (5 μg) in chloroform were spread on the platinum wires of the electroformation chamber. After solvent evaporation, the wires were immersed in 300 mM sucrose and electroformation proceeded for 2 h at 10 Hz at room temperature followed by 30 min at 2 Hz. Cyt c-Al488 or free Al488 and Bcl-2 proteins in phosphate-buffered saline were mixed in LabTec chambers (NUNC) blocked with casein. GUVs were added to the sample at a ratio of 70/300 μL. In pore-stability experiments, 20 μL of free Al633 or cyt c-Al633 were added after 90 min and images were collected after 30 min.

Pore-stability experiments on isolated mitochondria

Wild-type mitochondria were isolated from Saccharomyces cerevisiae DS1-2b (Y2197). To obtain matrix-targeted green fluorescent protein (mt-GFP) mitochondria, S. cerevisiae DS1-2b was transformed with a plasmid containing mt-GFP (pVT100U-mtGFP). Yeast cultures were grown at 30°C in rich YPG media (3% (v/v) glycerol as carbon source) to an optical density between 1 and 2. Yeast mitochondria were prepared as in Meisinger et al. (24). After cell homogenization, mitochondria were isolated by differential centrifugation, and the crude mitochondrial pellet was resuspended in ice-cold buffer (250 mM sucrose, 1 mM EDTA, and 10 mM MOPS-KOH, pH 7.2) at 8–10 mg/ml protein.

For experiments, mitochondria with or without mt-GFP were used. As a first dye, Dex_G, (Al488-labeled 10 kDa Dextran) was mixed with buffer (10 mM HEPES, 100 mM KCl, 1 mM EGTA, and 200 mM sucrose, pH 7.5) and the Bcl-2 proteins before mitochondria were added. After 45 min at 37°C, the second dye, Dex_R (Al647-labeled 10 kDa Dextran), was added and samples were incubated for 10 min at 37°C before adding agarose to 1%. In experiments using mt-GFP, no first dye was added, as the green channel was used for the mt-GFP. To label fluorescently the mitochondria in experiments without mt-GFP, Bcl-xl_R was added instead of the second dye. Images were collected after gel hardening. In the indicated experiments agarose was added at a temperature close to gel hardening, so that the dextrans outside the mitochondria were diluted and the dye could not equilibrate with the external medium.

Mitochondrial permeabilization was additionally measured by batch experiments in a fluorimeter. After incubation at 37°C the mitochondria were separated by centrifugation (15,000 × g, 10 min). The supernatant was removed and the solid pellet carefully washed with 100 μl buffer and afterwards resuspended in 200 μl buffer. The fluorescence intensity of the mitochondrial fraction was measured in a fluorescence spectrometer (Cary Eclipse, Varian, Darmstadt, Germany) (excitation, 345 nm; slit width, 5 nm; emission spectra, 650–750 nm).

Confocal microscopy and fluorescence correlation spectroscopy

All experiments were performed on an LSM710 microscope with a C-Apochromat 40× 1.2 water immersion objective (Zeiss, Oberkochen, Germany). Excitation light came from Ar-ion (488 nm), HeNe (561 nm), or HeNe lasers (633 nm). A spectral beam guide was used to separate emitted fluorescence. Images were processed with ImageJ (http://rsbweb.nih.gov/ij/), as described below.

We performed two-focus scanning fluorescence cross-correlation spectroscopy (FCCS) measurements at 22°C using a Confocor 3 module. Photon arrival times were recorded with a hardware correlator Flex 02-01D/C (http://correlator.com). We repeatedly scanned the detection volume with two perpendicular lines across a GUV equator (the distance between the two lines, d, was measured by photobleaching on a film of dried fluorophores). Data analysis was performed with our own software (see García-Sáez et al. (12)). We binned the photon stream in 2 μs and arranged it as a matrix such that every row corresponded to one line scan. We corrected for membrane movements by calculating the maximum of a running average over several hundred line scans and shifting it to the same column. We fitted an average over all rows with a Gaussian and we added only the elements of each row between −2.5 s and 2.5 s to construct the intensity trace. We computed the autocorrelation and spectral and spatial cross-correlation curves from the intensity traces and excluded irregular curves resulting from instability and distortion. We fitted the auto- and cross-correlation functions with a nonlinear least-squares global fitting algorithm, as in García-Sáez et al. (12). FCCS was corrected for cross talk (5%) and for labeling degree (100% for Bax_G, 80% for Bax_R, 90% for Bcl-xL_G, and 70% for Bcl-xL_R).

Image analysis

The percentage of GUV filling was calculated as

$$\left[\frac{\left({\mathit{F}}_{\mathit{t}}^{in}-{\mathit{F}}_0\right)}{\left({\mathit{F}}_{\mathit{t}}^{out}-{\mathit{F}}_0\right)}\right]\times 100,$$ (1)

where F_tⁱⁿ and F_t^out are the average fluorescence intensities inside and outside a GUV, and F₀ is the background fluorescence at time t. We arbitrarily set the threshold for a nonpermeabilized GUV to <15%. Several hundreds of GUVs were analyzed per experiment.

To quantify protein binding to GUV membranes, the fluorescence intensity at the vesicle rim (F_rim) was calculated with Image J using the plug-in radial profile plot. For Fig. 1, intensity was plotted as F_rim/F_back, where F_back is the background intensity outside the GUVs.

Figure 1.

Bax and Bcl-xL binding to GUVs depends on cBid and CL. (A and B) Confocal images and radial profiles of individual GUVs made of PC/CL 80:20 (mol/mol) after 60 min incubation at room temperature, showing Bcl-xL_R binding to GUVs in the absence and presence of cBid_G (A) and cBid_G binding to GUVs in the absence and presence of Bcl-xL_R (B). Scale bar, 25 μM. (C–F) Bcl-2 proteins Bax_G (C), cBid_R (D), Bcl-xl (E), and cBid_G (F) binding to GUVs with different lipid compositions. M indicates a lipid mixture mimicking a MOM containing 6% CL (mol/mol).

In the kinetics experiments, images were recorded every 20 s and changes in the fluorescence intensity inside GUVs were analyzed over time as

$$F_t^N=\left[\frac{\left({\mathit{F}}_{\mathit{t}}^{\text{in}}-{\mathit{F}}_0\right)}{\left({\mathit{F}}_{\mathit{t}}^{\text{out}}-{\mathit{F}}_0\right)}\right],$$ (2)

where F_t^N is the normalized fluorescence intensity at time t.

To calculate the initial A₀ and relaxed A_relax permeabilized area of individual GUVs, as well as the relaxation time τ_relax, we used a multiexponential fitting described by Eq. 7:

$$F{\left(t\right)}_{\text{in}}^N=1-e^{\frac{-t}{{\tau }_{\text{flux}}\left(t\right)}},$$ (3)

where the influx rate, τ_flux, decreases with time (the initial pore size relaxes to a smaller structure) according to

$${\tau }_{\text{flux}}\left(t\right)=\frac{V}{A\left(t\right)\times \frac{D}{m}},$$ (4)

where V is the vesicle volume, D the dye diffusion coefficient, m the membrane thickness, and A the total permeabilized area, which varies with time according to

$$A\left(t\right)=A_{\text{relax}}+\left(A_0-A_{\text{relax}}\right)\times e^{\frac{-t}{{\tau }_{\text{relax}}}}.$$ (5)

Membrane thickness was assumed to be 4.5 nm. The diffusion coefficient of cyt c-al488 was 196 μm²/s (196 ± 27 μm²/s) as calculated by fluorescence correlation spectroscopy (FCS).

Results

Bax and Bcl-xL compete for membrane association induced by cBid

To compare the permeabilizing activity of Bax and Bcl-xL, it is essential to analyze their membrane association. We directly visualized and quantified protein binding to membranes by the increase in fluorescence intensity on the rim of individual GUVs (Fig. 1, A and B, and Fig. S3). cBid enables Bax, as well as Bcl-xL, insertion into the MOM and into model membranes (12,21). In agreement with its role as a receptor (7,12), we observed that membrane binding of cBid (labeled with Al633 or Al488 (cBid_{R or G})) depends on cardiolipin (CL) and increases with CL concentration (Fig. S2). Interestingly, cBid_{R or G} binding to a lipid mixture that mimics the MOM (see Materials and Methods) and contains 6% CL (mol/mol) was too weak to be detected using this method (Fig. 1 D). This weak binding is in line with previous results using different methods (23).

In the case of Bax (labeled with Atto488 (Bax_G) or Atto 655 (Bax_R)), its association depended on the presence of both cBid and CL (Fig. 1, C and D, and Fig. S3). Using the mixture mimicking the MOM, Bax_{G or R} bound to GUVs, but it was close to the detection limit, and with 20% CL the contrast was still weak. By increasing CL content to 33%, we clearly visualized Bax_{G or R} membrane association (Fig. 1 C). When Bcl-xL was present, Bax_{G or R} binding to GUVs was reduced (Fig. 1 C), as in experiments using bulk assays (21). However, cBid_{R or G} binding was not significantly affected (Fig. 1 D). This shows that Bcl-xL can inhibit Bax action by blocking its membrane association, probably by sequestration of cBid on the membrane, as their most favorable interaction takes place within the lipid bilayer (12).

Bcl-xL_{R or G} binding to CL-containing membranes was promoted by cBid_{G or R} and, vice versa, Bcl-xL_{R or G} increased cBid_{G or R} concentration at the membrane (Fig. 1, A, B, E, and F, and Fig. S3). In agreement with this, increasing the CL concentration enhanced Bcl-xL binding to the membrane. Interestingly, for the MOM-like lipid mixture, cBid_G binding to the membrane was only clearly detected in the presence of Bcl-xL (Fig. 1 D). In contrast to the effect of Bcl-xL on Bax_{G or R} binding, the presence of Bax did not affect the association of Bcl-xL_{R or G} with GUVs (Fig. 1 E).

Bax and Bcl-xL permeabilize membranes with different mechanisms

Membrane permeabilization was directly visualized by monitoring the entry of a dye into the lumen of the GUVs initially devoid of dye. With this method, the kinetics and the extent of permeabilization can be quantified for individual vesicles in the sample. This allows the identification of subpopulation effects, which is not possible in traditional bulk experiments. Moreover, preferences for certain membrane types can be directly assessed by mixing GUVs of different compositions in a situation closer to the coexistence of membranes in a cell.

We mixed GUVs composed of PC (green) and PC/CL 8:2 (red) in a buffer containing cyt c-Al488 (green) and Bcl-2 proteins (Fig. 2 A). In the presence of cBid and Bax or Bcl-xL, we found permeabilization of vesicles containing CL, but not pure PC, illustrating the specific effect of CL on the membrane activity of these Bcl-2 proteins. In the presence of CL, only the combination of cBid with Bax, Bcl-xL, or a mixture of the two caused a dramatic increase in the fraction of permeabilized vesicles (Fig. 2 B). A similar result was obtained when the GUVs were produced with the MOM-like lipid mixture (Fig. 2 B). The only detectable difference between PC/CL 8:2 and the MOM-like lipid mixture is permeabilization at a lower, but still present, rate by cBid and Bcl-xL in the latter. It is important to note that membrane permeabilization usually occurred without vesicle disruption. Bax was faster than Bcl-xL at permeabilizing the vesicles in the sample and, unlike its antiapoptotic counterpart, tended to burst part of the GUVs after long incubation times.

Figure 2.

Bax and Bcl-xL follow different mechanisms of membrane permeabilization. (A) The internalization of cyt c-Al488 into the lumen of the individual vesicles indicates membrane permeabilization. Green GUVs composed of PC 0.05% DiO and red GUVs of PC/CL 8:2 0.05% DiD. Images were taken 90 min after mixing the components. Scale bar, 75 μm. (B) Fraction of nonpermeabilized GUVs in the presence of Bcl-2 proteins. Gray bars correspond to PC GUVs, red bars to GUVs containing 20% CL, and blue bars to a mixture mimicking the MOM. (C) Distribution of degree of filling of the GUVs after 90 min incubation with cBid/Bax (red bars) and cBid/Bcl-xL (blue bars). GUVs were composed of PC/CL 8:2 0.05% DiD. In each of three experiments, a minimum of 250 vesicles were analyzed. (D and E) Time-lapse images of the permeabilization of GUVs (grayscale) to cyt c-al488 (green) induced by cBid/Bax (D) or cBid/Bcl-xL (E). Scale bar, 25 μm. (F and G) Filling kinetics measured by the increase of fluorescence intensity inside of individual GUVs (different colors) incubated with cBid/Bax (F) or cBid/Bcl-xL (G). Normalized data are shown in dots and fitting curves in lines. (H and I) Estimated radius of the initial (H) and relaxed (I) total pore area of individual GUVs after incubation with cBid/Bax (red) or cBid/ Bcl-xL (blue). Outliers are shown in gray. Protein concentration was 10 nM for cBid, 20 nM for Bax,, and 50 nM for Bcl-xL.

By comparing the fluorescence inside the individual GUVs with that of the background, one can calculate the filling degree after a certain incubation time. Two situations are distinguished. In the all-or-none mechanism of membrane permeabilization, the vesicles in the sample exhibit two states, impermeable or totally permeabilized. Alternatively, the individual GUVs in the sample can exhibit a varying degree of filling according to the graded mechanism. In Bax-permeabilized GUVs, the distribution of filling degrees is typical of an all-or-none mechanism both for the PC/CL 8:2 and the MOM-like lipid mixture (Fig. 2 C and Fig. S4). In contrast, the vesicle population in the samples incubated with Bcl-xL spans a large range of filling degrees, following a graded mechanism of permeabilization, again found in both lipid mixtures.

We quantified the permeabilization kinetics of single vesicles in PC/CL 8:2 (Fig. 2, D–G, Movie S1, and Movie S2). They were permeabilized stochastically, but we aligned the starting point of permeabilization for clarity. The filling rate of GUVs incubated with Bax was very fast, and most vesicles reached 50% filling after a few seconds. This process was slower in GUVs permeabilized by Bcl-xL, which in some cases did not reach the totally permeabilized state. To properly compare the kinetics, we calculated the totally permeabilized area/vesicle induced by Bax or Bcl-xL as described previously (25,26) (Fig. 2, H and I). The initial filling proceeded faster than at later stages. This has been extensively investigated by Fuertes et al. (25), who showed that the two-component filling kinetics of GUVs can be explained by the relaxation of the initial membrane openings to smaller pores. In the case of Bax, the calculated average radius of the initial total pore area is 72 ± 28 nm, which relaxed to an average size of 7 ± 3 nm. In contrast, the area permeabilized by Bcl-xL was much smaller, with an average initial radius of 11 ± 4 nm that relaxed to 2 ± 1 nm. This indicates that the total size of the membrane lesions induced by Bax is significantly larger than those promoted by Bcl-xL and that the kinetics of membrane permeabilization by Bax are faster than those of its antiapoptotic counterpart.

Unlike Bcl-xL, Bax forms long-lived membrane pores that correlate with homooligomerization

The results above suggest that the pores induced by Bax evolve into structures that allow the passage of large molecules across the membrane, whereas those induced by Bcl-xL tend to close after a certain time. To investigate this in more detail, we evaluated the stability of the permeability alterations induced by Bax and Bcl-xL. After incubating the GUVs in a buffer containing Al488 and the proteins of interest, we added Al633 to see whether the vesicles permeabilized to Al488 were still permeable to the second marker (Fig. 3, A and B) (25,27). With this system we cannot distinguish whether the individual pores are opening and closing or whether they stay continuously open during the experiment, but we can monitor whether the membrane is still in a permeabilized state after the incubation time. In the control (Fig. 3 E), only a small fraction of the few permeabilized vesicles were permeable to the second marker. In the presence of Bax, the vast majority of vesicles filled with Al488 were also permeable to Al633 (Fig. 3 C), demonstrating that Bax forms pores for a long period of time. In contrast, the number of vesicles permeable to both dyes was much lower for Bcl-xL (Fig. 3 D), indicating that the membranes closed after a certain period of time. We obtained similar results with fluorescent cyt c and lipid mixture mimicking the MOM (Fig. S5). Our findings show that Bax and Bcl-xL exhibit differences in stability of the membrane perturbations they induce (Fig. 3 F): whereas Bax induces long-lived pores, Bcl-xL promotes transient alterations of the membrane permeability.

Figure 3.

Unlike Bcl-xL, Bax induces stable pores in lipid membranes. (A and B) GUVs in the presence of cBid and Bax or Bcl-xL after 2 h incubation. PC/CL 8:2 0.05% DiI GUVs (second panel) and Al488 (first panel) were added simultaneously with the proteins, and Al633 (third panel) was added after the 90-min incubation. Merged images are shown in the fourth panel. (C–E) Filling degree of individual, permeabilized GUVs to Al488 (green bars) and Al633 (red bars) after incubation in the presence of cBid/Bax (C), cBid/Bcl-xL (D), or without proteins (E). (F) Fraction of GUVs with stable pores (>50% filling with both Al488 and Al633) from five independent experiments. Error bars represent the SD. Unpaired two-tailed t-test, p value < 0.001.

The GUV system also offers the possibility of directly measuring the oligomeric state of the membrane-inserted conformation of Bax and Bcl-xL by FCCS, a technique with singlemolecule sensitivity that quantifies dynamic codiffusion of biomolecules online (28). To investigate the oligomeric state of Bax and Bcl-xL under our experimental conditions, we quantified complex formation between Bax_R and Bax_G, or Bcl-xL_R and Bc-xL_G, added to unlabeled GUVs in the presence of cBid (Fig. 4). The amplitude of the FCS autocorrelation curves is inversely proportional to the concentration of proteins, in this case red and green Bax or Bcl-xL in the membrane (Fig. 4, A and B, green and red lines). In line with Fig. 1, the concentration of Bax in the membrane was lower than that of Bcl-xL. In contrast, the amplitude of the FCCS curves (Fig. 4, A and B, blue lines) is directly proportional to the number of particles in a complex containing at least one red and one green protein. While fluorescent Bax showed positive cross-correlation indicative of oligomerization, Bcl-xL showed no significant cross-correlation, showing the inability of the protein to self-assemble. In GUVs with higher concentrations of Bax, we detected bright dots (large oligomers or aggregates) with colocalized red and green Bax molecules (100% cross-correlation). For technical reasons, these GUVs were not included in the FCS analysis due to their high brightness and spatial heterogeneity. Fig. 4 C shows, for several individual vesicles, the percentage of Bax or Bcl-xL particles that form part of a complex within the GUV membrane, as calculated from the FCCS experiments.

Figure 4.

Bax, but not Bcl-xL, oligomerizes in the membrane of GUVs. (A and B) Confocal images, auto- (orange and green), and cross-correlation (blue) curves from two-focus scanning FCCS measurements in the membranes of GUVs using Bax_G/Bax_R (A) or Bcl-xL_G/Bcl-xL_R (B) in the presence of cBid. Straight lines correspond to fitted curves and dotted lines to the raw data. Scale bar, 10 μm. (C) Percent complex between Bax and Bcl-xL molecules in individual GUVs. Unpaired two-tail t-test, p value < 0.001.

Bax also induces stable permeabilization in isolated mitochondria

The detailed level of analysis of membrane permeabilization obtained in GUVs cannot be achieved with higher biological systems. However, the MOM contains a much more complex protein-lipid composition than the GUVs. To test whether our results are valid under more physiological conditions, we developed a new method to study the stability of membrane permeabilization in isolated mitochondria. This method provides the complexity of a mitochondrial microenvironment. As a control that we obtain mainly mitochondria and not other organelles, we isolated mitochondria labeled with mt-GFP parallel to wild-type mitochondria (Fig. S6). To avoid the unwanted effects of endogenous Bcl-2 proteins, we used mitochondria from S. cerevisiae.

To follow MOM permeabilization, we added fluorescent 10 kDa dextrans (Dex_R or Dex_G) to the external buffer, a strategy similar to that used with the GUVs (Dex_G at t = 0; Dex_R at t = 45 min). After 55 min incubation at 37°C, the mitochondria remained impermeable in the buffer control (Fig. 5 C). In batch experiments, we compared the amount of Dex_R in sedimented mitochondria after incubation with cBid/Bax or cBid/Bcl-xL or buffer control (Fig. 5, A and B). We measured the emission spectra and compared the fluorescence intensities of the different samples (Fig. 5, A and B). Incubation with cBid/Bax led to a higher Dex_R concentration in the mitochondria compared to the buffer control, whereas cBid/Bcl-xL showed no significant difference. These results demonstrate that cBid/Bax but not cBid/Bcl-xL form stable pores in mitochondria, similar to our observations for the GUVs.

Figure 5.

Bax stably permeabilizes isolated yeast mitochondria. (A) Fluorescence spectra of pelleted yeast mitochondria incubated with DexR and buffer (black); 50 nM cBid and 100 nM Bax (red); or 50 nM cBid and 200nM Bcl-xL (blue). (B) Mean ± SD of the fluorescence intensity in the pellet fraction at 667 nm (from five independent experiments). (C–G) Confocal images of isolated yeast mitochondria after incubation with buffer (C), Bcl-xL/cBid (D), or Bax/cBid in the presence of Dex_G (added at time 0) (E–G) and Dex_R (C–F) or Bcl-xL_R added after 45 min incubation (G). In F agarose was added at a temperature suitable to fast gel hardening >1 min, so that the dextrans in the medium are diluted and no equilibration with the dextrans inside the mitochondria can take place. Images in C–G are (left to right) the transmission image, Dex_G (green), Dex_R or Bcl-xL_R (red), and a merge of the red and green channels. Protein concentrations were 50 nM cBid, 100 nM Bax, 200 nM Bcl-xL, or 50 nM Bcl-xL_R (in G). Scale bar, 5 μm.

Moreover, we used a confocal microscope to visualize the dextran entrance into single mitochondria (Fig. 5 and Fig. S7). Under control conditions and in the presence of cBid and Bcl-xL, Dex_G and Dex_R were only found outside the mitochondria (Fig. 5, C and D, and Fig. S7, A and C), indicating that most mitochondria were not permeabilized. On the contrary, when incubated with Bax and cBid, both dextrans could enter the mitochondria (Fig. 5 E and Fig. S7 B). We used a dilution strategy of the medium outside the mitochondria that led to a higher dextran concentration inside the permeabilized organelles (Fig. 5 F and Fig. S7). As an additional control, we added Bcl-xL_R at the end of the incubation to label the MOM (Fig. 5 G and Fig. S7 D). The images also show that cBid and Bax induce a stable permeabilization of the MOM, in agreement with the batch experiments and the assays with GUVs.

In contrast, Bcl-xL and cBid did not induce mitochondrial permeabilization (Fig. 5 D and Fig. S7 C), not even for dextran added at t = 0. This result varies from the results found on GUVs and suggests that the microenviroment of the mitochondria alters the membrane-permeabilizing activity of Bcl-xL.

Discussion

To better understand the membrane activity of pro- and antiapoptotic Bcl-2 proteins, and to integrate this with their function in apoptosis, here we have carried out a comparison at the single vesicle levle of their membrane binding, oligomerization and permeabilizing activity.

In our binding studies, we found that cBid promotes association of both Bax and Bcl-xL with CL-containing membranes. To perform this function, cBid had to be able to bind to the target membrane, and this binding depended on the membrane’s CL content. This supports the role of CL as a receptor for cBid (29) and suggests that cBid itself functions as a receptor for Bax and Bcl-xL. Bcl-xL membrane binding induced by cBid was more efficient than that of Bax, indicating that the differences in pore activity between the two proteins are not due to a more efficient binding of Bax to the membrane.

Interestingly, Bcl-xL, but not Bax, increased cBid binding to CL-containing GUVs, which could be explained by the strong tendency of these two proteins to form complexes within the lipid bilayer (12). In addition, we found that Bax did not affect Bcl-xL binding to GUVs. These findings highlight a difference in the respective mechanisms of action of Bax and Bcl-xL, probably due to the inability of Bax to compete with Bcl-xL for cBid binding. One could hypothesize that cBid/Bcl-xL complex formation would free cardiolipin molecules, which would then be able to act again as receptors and induce the membrane binding of additional cBid molecules. Moreover, as long as Bcl-xL sequesters cBid at the membrane, the latter is likely unable to interact with Bax, thus providing an explanation of how Bcl-xL can inhibit Bax binding to the membrane and therefore also its apoptotic action (also described by Billen et al. (21)).

Bax followed an all-or-none mechanism, with the liposome population divided into completely filled or intact vesicles (27). This implies that as Bax binds to the membrane, its impermeability is retained up to a threshold concentration (30,31), at which point the opening of one or more pores that allow complete vesicle filling is induced. Interestingly, the kinetics of single-vesicle filling shows a two-component kinetics, with an initial larger permeabilized area that relaxes to a smaller size. This has been observed already for pores induced by Bax-derived peptides (25), and it may be a general feature of proteins inducing lipid pores, also known as toroidal pores. It is important to note that the average total size of Bax pores in conditions close to equilibrium is 7 ± 3 nm, large enough to allow the passage of proteins like cyt c across the membrane. Therefore, our findings demonstrate that Bax has the ability to form long-lived pores that are large enough to release proteins across pure lipid membranes, as well as in isolated mitochondria. This reproduces the processes involved in MOM permeabilization during apoptosis.

Interestingly, the features of membrane permeabilization by Bcl-xL were very different from those exhibited by Bax. Bcl-xL followed a graded mechanism of permeabilization, with most vesicles only partially filled. This is typical for small and transient alterations of membrane permeability, which do not allow the vesicle interior to reach equilibrium with the external medium. Although the initial total pore area was considerable and allowed the passage of proteins the size of cyt c, these alterations evolved to very small structures, which do not allow the passage of molecules ≥2 nm radius, and to eventual closure. Moreover, we also found that Bcl-xL was unable to induce a stably permeabilized membrane. This shows that the activity of Bcl-xL observed in artificial membranes is the result of transient alterations of the bilayer permeability likely associated with membrane insertion, which disappear as the system evolves toward equilibrium. In agreement with this, we did not find permeabilization of isolated mitochondria, which have a very different microenvironment compared to the pure lipid bilayer of GUVs. However, once in the membrane, Bcl-xL is able to exclude Bax from the membrane and thereby prevent MOM permeablization. Moreover, membrane-inserted Bcl-xL can act as an interaction partner with other proteins related to nonapoptotic functions. In this context, interaction of subapoptotic amounts of Bax with Bcl-xL could hinder binding of Bcl-xL to other proteins and thus have an effect on alternative functions not related to membrane permeabilization.

Models for Bax and Bcl-xL membrane activity

Computational studies using molecular dynamics have shown that lipid pores have a rather disordered toroidal shape (32). Direct evidence for such pore structures was first described for a peptide derived from helix 5 from Bax (33). According to a continuum model, the energy, E, of a pore in a lipid membrane can be described as

$$E\left(r\right)=\mathit{2}\pi \gamma r-\pi \sigma r^2,$$ (6)

where r is the radius of the pore, γ is the line tension associated with the energy cost of the highly curved membrane at the pore edge (due to imperfect lipid packing), and σ is the membrane tension, which originates from a stress source and represents the potential for mechanical work of pore expansion (34). As a result, pores are metastable structures, with the membrane tension tending to enlarge the pore and line tension tending to close it.

The mechanism of pore-forming polypeptides usually involves two steps: 1), the introduction of membrane stress that decreases the activation energy of pore opening (35); and 2), a decrease of line tension to stabilize the open state of the pore. Once formed, the pore itself constitutes a mechanism to relieve the membrane tension, which leads to relaxation of the pore size (25,36) and to final closure. The asymmetric insertion of these polypeptides into the accessible membrane leaflet introduces additional area, which leads to membrane thinning (37,38). This raises the membrane tension up to a level at which the opening of a pore is no longer unfavorable. In our experiments, membrane insertion of Bcl-xL and Bax constitutes an asymmetric attack on the lipid bilayer that introduces membrane tension and induces pore opening in both cases (Fig. 6). In the case of Bax, the additional effect of oligomerization effectively produces a local increase of the membrane tension that increases the efficiency of pore opening compared to that of Bcl-xL, which does not self-associate.

Figure 6.

Models for membrane activity of Bax and Bcl-xL. (A) The membrane tension associated with Bax binding to the accessible membrane leaflet is locally elevated by oligomerization. This increases the efficiency of pore opening, which is then stabilized by a decrease in line tension promoted by Bax. Bax oligomerization most likely also increases the line-tension attenuation and thus, pore stability. (B) Upon membrane binding, Bcl-xL also increases the membrane tension, although less efficiently. When the membrane tension is sufficiently high, a pore opens. As the membrane tension is relieved, and since Bcl-xL does not attenuate the line tension, the pore will close.

Once the pore is open, it will evolve toward closure unless line tension is decreased. We showed previously that helix α5 of Bax exhibits such an attenuation of line tension correlating with the formation of stable pores (25,30), in agreement with what is observed here for full-length Bax. Bax oligomerization also has an effect on pore stabilization, increasing the efficiency of line-tension relief, as it ensures a high number of pore-forming proteins at the pore rim. In contrast, Bcl-xL pores evolved toward a closed state over time, indicating the inability of this protein to relieve line tension. Based on Bcl-xL inhibition of Bax self-assembly, previous models have made a link between Bax oligomerization and its ability to permeabilize the MOM (21). However, to our knowledge, this is the first time that the role of oligomerization has been included in the molecular mechanism of stable pore formation by Bax.

Implications in apoptosis and other functions

These observations are relevant in the context of biological membranes, which modulate their membrane tension (39). In the MOM, naturally permeable to molecules up to 5 kDa, the increase of membrane tension associated with Bcl-xL insertion is dissipated with small, transient membrane reorganizations that do not affect the overall MOM permeability. In contrast, Bax insertion and oligomerization ensures a local increase of membrane tension, leading to the opening of a pore that is efficiently stabilized by the line-tension-attenuating effect of the protein. As a result, MOM permeability is altered and the apoptotic factors are released into the cytosol, as we found in the studies of pore stability with isolated mitochondria.

Despite the lack of information regarding the mechanisms involved, the membrane-associated form of the Bcl-2 proteins is crucial for their nonapoptotic functions. The results reported here should be taken into account when investigating these alternative roles. For example, Bcl-xL increases the probability of the open state of voltage-dependent anion channel (40) and regulates ER Ca²⁺ content by modulating the activity of IP3R (41). This ability of Bcl-xL to modulate the permeability of the MOM to small molecules is compatible with our results and may have implications for mitochondrial metabolism. Moreover, Bcl-xL also interacts with Beclin-1 at the ER to regulate autophagy via mechanisms that are likely related to Ca²⁺ homeostasis (1,42). Another example is mitochondrial dynamics. Bax has been shown to colocalize at the mitochondrial fission sites with proteins regulating mitochondrial fusion and fission, Mfn2 and Drp1, and to modulate their activities (43). On the other hand, Drp1 promotes Bax apoptotic activity via a mechanism based on membrane curvature (44). In this context, the ability of Bax to stabilize open pores by decreasing the line tension at the pore edge (as proposed in our model) is likely to also play a role in stabilizing the highly curved membrane generated at the mitochondrial fission and fusion sites.

Our in-depth comparison of the membrane activity of Bax and Bcl-xL reveals profound differences in the molecular mechanisms of action of these proteins that are relevant to their biological functions. This explains why Bcl-xL, despite permeabilizing model membranes, does not induce MOM permeabilization in apoptotis, as its homolog Bax does. In contrast to Bax and their bacterial analogs, Bcl-xL does not behave as a pore-forming protein in the complex microenvironment of the MOM, as observed in the experiments with isolated yeast mitochondria. Bcl-xL is nevertheless able to insert into intracellular membranes and interact with other proteins from the Bcl-2 family and beyond to regulate apoptosis, as well as to fulfill other functions. However, it does not induce the permeabilization of these intracellular membranes. In a different membrane environment, like the protein-free liposomes, Bcl-xL does exhibit membrane-permeabilizing activity. In this sense, our results integrate, for the first time, the behavior in membranes of highly homologous proteins like Bax and Bcl-xL with their different apoptotic functions.

Conclusions

To summarize, our comparison of membrane activity of Bax and Bcl-xL at the single-vesicle level demonstrates fundamental differences in the membrane interaction and organization of these two proteins. On the one hand, Bax and Bcl-xL insert into membranes in a cBid-dependent manner and thereby disturb membrane integrity. On the other hand, once inserted, Bax and Bcl-xL behave quite differently. Although Bax oligomerizes to form stable pores that are large enough to release proteins, Bcl-xL induces small, transient alterations of the membrane permeability without consequences for the MOM. Moreover, Bcl-xL is able to inhibit Bax membrane insertion and therefore MOM permeabilization. These differences integrate the membrane activity of these proteins in model membranes with their different effects on MOM permeability during apoptosis.