The Cytosolic Domain of Bcl-2 Oligomerizes to Form Pores in Model Mitochondrial Outer Membrane at Acidic pH

Abstract

The three dimensional structures of both pro-apoptotic Bax and anti-apoptotic Bcl-2 are strikingly similar to that of pore-forming domains of diphtheria toxin and E. coli colicins. Consistent with the structural similarity, both Bax and Bcl-2 have been shown to possess pore-forming property in the membrane. However, these pore-forming proteins use different mechanisms to form pores. While Bax and diphtheria toxin form pores via oligomerization, the colicin pore is formed only by colicin monomers. Although oligomers of Bcl-2 proteins have been found in the mitochondria of both healthy and apoptotic cells, it is unknown whether or not oligomerization is involved in its pore formation. To determine the mechanism of Bcl-2 pore formation, we reconstituted the pore-forming process of Bcl-2 using purified proteins and liposomes. We found that Bcl-2 pore size depended on Bcl-2 concentration; and smaller entrapped molecules released faster than larger ones from liposomes at a given Bcl-2 concentration. Moreover, the rate of dye release mediated by pre-formed wild-type Bcl-2 oligomers or by the mutant Bcl-2 monomers with a higher homo-association affinity was much higher than that by wild-type Bcl-2 monomers. Together, it is suggested that oligomerization is likely involved in Bcl-2 pore formation.

Introduction

Bcl-2 family proteins are key regulators of apoptosis, functioning either as suppressors or as promoters (1, 2). All the Bcl-2 family members contain at least one of the four Bcl-2 homology (BH) motifs. According to the function and sequence homology, Bcl-2 family is divided to three subfamilies. The anti-apoptotic Bcl-2 subfamily proteins such as Bcl-2, Bcl-XL, Bcl-w, Bfl-1 and Mcl-1 contain all four BH motifs. The pro-apoptotic Bax subfamily members such as Bax and Bak display homology in BH1-3 motifs, whereas the other pro-apoptotic BH3-only subfamily proteins such as Bid, Bim and Bad share the sequence homology only in the BH3 motif (3-7).

Despite an overall divergence in amino acid sequence and function, the three-dimensional structures of several Bcl-2 family proteins including members from all three subfamilies are remarkably similar to each other and to that of pore-forming domains of diphtheria toxin (8) and E. coli colicins (9, 10), consisting of two hydrophobic core helices surrounded by five to eight amphipathic helices and their connecting loops (11-15). Consistent with the structural similarity, Bcl-2 family proteins including anti-apoptotic Bcl-2 and pro-apoptotic Bax have been shown to form pores to alter the mitochondrial membrane permeability during apoptosis. However, unlike Bax to form large cytochrome c-permeable pores, Bcl-2 forms much smaller pores in the membrane; which may account for the opposite functions of these two proteins (16).

Pore-forming proteins use different mechanisms to form pores in the membrane in spite of the structural similarity. Oligomerization of Bax was found to be required for Bax to form pores in liposomes, to release cytochrome c from mitochondria, and to have the deleterious effect on cells (17-21). Oligomerization of other pore-forming proteins such as diphtheria toxin (22) and perfringolysin O (PFO) (23) was found to facilitate their pore formation. However, a different pore-forming model has been proposed for colicins in which colicin pore is formed by monomeric colicin proteins (24). Although oligomers of Bcl-2 proteins have been found in the mitochondria of both healthy and apoptotic cells (19), it is unclear if Bcl-2 forms pores via oligomerization or only by monomers. To address this question, in this study we reconstituted the pore-forming process of Bcl-2 in vitro using the MOM-liposomes loaded with encapsulated fluorescent dyes and purified cytosolic domain of Bcl-2 (Bcl-2ΔTM). We found that Bcl-2 formed pores with various sizes depending on the protein concentration, suggesting that the pore may be formed via oligomerization of Bcl-2 monomers. Moreover, studies on the kinetics of dye release showed that the small entrapped dyes released faster than the large ones, and that the rate of dye release mediated by pre-formed wild-type Bcl-2 oligomers or by the mutant Bcl-2 monomers with a higher homo-association affinity was much higher than that by wild-type Bcl-2 monomers, further supporting the model that oligomerization is likely involved in Bcl-2 pore formation. In addition, Bcl-2 oligomerization both in solution and in membrane was directly observed by chemical crosslinking.

Experimental Procedures

Materials

All phospholipids and lipid analogs were purchased from Avanti Polar Lipids (Alabaster, AL). Fluorescent dye Cascade Blue (CB, M_r ∼ 0.5 kDa), CB-labeled dextran of 3 kDa or 10 kDa, and rabbit anti-CB polyclonal antibody were obtained from Molecular Probes. Bismaleimidohexane (BMH) was from Pierces.

Preparation of Plasmids and Proteins

Construction of pET22b(+)-based plasmid encoding His₆-tagged Bcl-2ΔTM with Gly¹⁵⁴ and Gly¹⁵⁵ replaced by two alanines (designated G154A/G155A) was done using appropriate primers and overlapping PCR-based mutagenesis as described (16, 25). Expression and purification of (His)₆-tagged Bcl-2ΔTM and mutants was done basically as described before (16, 25).

Preparation of Liposomes

Liposomes of MOM lipid composition and with CB or CB-dextrans encapsulated were prepared by the extrusion method as described (16). The plain liposomes without fluorophores were prepared similarly except that no fluorophores were in the citric-phosphate buffer (CPB) used for resuspension of the lipids and no gel filtration chromatography was needed after the extrusion.

Assay of CB or CB-dextran Release from Liposomes by Fluorescence Quenching

Steady-state fluorescence emission intensity was measured using the SLM-8100 spectrofluorometer, as described (16). Briefly, the sample contained liposomes with entrapped CB dyes and anti-CB antibodies in 250 μl of CPB buffer (pH 5.0). The initial emission intensity (F₀) was determined after equilibration of the sample at 25 °C for 5 min. Purified Bcl-2ΔTM protein was then added and incubated with liposomes for 5 hours; then the emission intensity (F) was measured in end-point measurements. In kinetics measurements, F was measured at each time point right after the addition of Bcl-2 proteins. Triton X-100 was then added to 0.1% for complete disruption of liposomes and release of all CB dyes. The emission intensity (F_t) was measures again. The quenching efficiency was calculated by ΔF_{Bcl-2 ΔTM}/ΔF_Triton, where ΔF_{Bcl-2 ΔTM} = F₀ − F and ΔF_Triton = F₀ − F_t.

Monitoring Bcl-2-liposome association by flotation and Bcl-2 homo-oligomerization by chemical cross-linking

1 mM liposomes and 4μM Bcl-2ΔTM protein were incubated in 138 μl of CPB containing 10 mM EDTA for 1 hour at 25°C at pH 5.0. After the addition of 20 μM bismaleimidohexane (BMH), the sample was kept at 25 °C for another 30 min. The chemical reaction was stopped with 50 mM β-mercaptoethanol (β-Me).The resultant sample was then mixed with 162 μl of 2.5 M sucrose to bring the final sucrose concentration to 1.35 M. The resulted sample was added to a centrifuge tube and overlaid with 500 μl of 0.8 M sucrose and then 200 μl of 0.25 M sucrose. The sample was centrifuged at 436,000 × g for 6 hrs at 4 °C. Four 250 μl fractions were drawn, starting from the top of the gradient, resulting in fractions S1-S4. The pellet was re-suspended in 250 μl of CPB buffer, resulting in fraction P. The distribution of the liposomes among these fractions was estimated by measuring the radioactivity of the ¹⁴C that resulted from the trace of [¹⁴C]-PC incorporated into the liposomes. Proteins from each fraction were precipitated by 25% TCA and analyzed by SDS-PAGE and Coomassie Brilliant Blue.

Results

Experimental design

To determine the mechanism of Bcl-2 pore formation, we first need to analyze the properties of Bcl-2 pore formation including the protein-dependence of Bcl-2 pore size and the kinetics of Bcl-2-pore-mediated dye release. We encapsulated fluorophores into liposomes surrounded by antibodies against the fluorophores, followed by the addition of purified Bcl-2ΔTM proteins. The Bcl-2ΔTM-pore-mediated dye release from the liposomes was monitored by quenching of fluorescence emission of the fluorophores. Since both fluorophores and antibodies do not diffuse across the liposomal membrane spontaneously, the quenching only occurs when the fluorophores are small enough to leak through the pore formed by the Bcl-2ΔTM protein.

By using fluorophores of different sizes within the liposomes and incubating various concentrations of Bcl-2ΔTM with liposomes, we will be able to determine the relationship between the pore size and the protein concentration, which will tell us whether the oligomerization is involved in Bcl-2 pore formation. If Bcl-2 pores are formed only by monomers, these pores should have a uniform size that will be independent of Bcl-2 concentration. On the contrary, if the pore size is positively dependent on Bcl-2 concentration, it would suggest oligomeric nature of the pore. Moreover, the mechanism of Bcl-2 pore formation can be further investigated by comparing the release kinetics of different sized fluorophores, at a given protein concentration by which both small and large fluorophores can be released from liposomes. If pores are formed only by monomeric proteins, the entrapped fluorophores of different sizes should be released from liposomes at the same rate that is dictated by the rate of pore formation, a constant at the given protein concentration. In contrast, if the oligomerization is involved in the pore formation, the release rate of small fluorophores should be faster than that of large ones since the formation rate of small oligomeric pores is faster than that of large oligomeric pores.

Oligomerization is involved in the pore formation of membrane-bound Bcl-2ΔTM

Using above fluorescence approach, we examined the relationship between the pore size and concentration of Bcl-2ΔTM. We encapsulated free Cascade Blue (CB) dye (0.5 kDa), or CB-conjugated 3 kDa or 10kDa dextrans into liposomes. Various concentrations of purified Bcl-2ΔTM protein were then incubated with liposomes at pH 5.0. As shown in Fig. 1, more 0.5 kDa CB dyes were released than CB-3 kD dextrans at a given concentration of Bcl-2ΔTM, and no significant release of CB-10 kD dextrans was observed even at the highest protein concentration tested, demonstrating the concentration-dependence of Bcl-2ΔTM pore size: a probable relationship to oligomerization.

Fig. 1. The dependence of release of CB-dextrans on Bcl-2ΔTM concentration.

Release of 0.5 kDa Cascade Blue (CB) dyes, 3 kDa or 10 kDa CB-dextrans from 100 μM liposome by various concentrations of Bcl-2ΔTM at pH 5.0, was shown as the extent of the fluorescence quenching by the anti-CB antibody located outside of the liposome. The extent of Bcl-2ΔTM-induced release is determined by the value of ΔF_Protein/ΔF_Triton after a 5-hour incubation (see Materials & Methods). Data shown are averages from three independent experiments with S.D. (error bars).

To further test the model for Bcl-2ΔTM pore formation, the release kinetics of entrapped CB dyes of different sizes was examined. As shown in Fig. 2, even though both 0.5 kDa CB dye and 3 kDa CB-dextran were released to the similar extent by 0.8 μM Bcl-2ΔTM, the small dye was released at a higher rate than the large one. The time for half of maximal release (t_1/2) for 0.5 kDa CB dye or 3 kDa CB-dextran was ∼140 seconds and 1500 seconds, respectively. The 10 kDa CB-dextran was not released even after hours of incubation, ruling out the possibility that prolonged exposure of liposomes and Bcl-2ΔTM can cause specific damage in liposomal membranes. These kinetics data thereby further support the model that Bcl-2ΔTM proteins form pores via oligomerization.

Fig. 2. The release kinetics of entrapped molecules with different sizes.

The release kinetics of CB dextrans from 100 μM liposome by 0.8 μM Bcl-2ΔTM at pH 5.0 was shown for: 0.5 kDa CB (○); CB-labeled 3 kDa dextrans (●); CB-labeled 10kDa dextrans (□); or the negative control reaction containing liposomes loaded with 0.5 kDa CB in the absence of Bcl-2ΔTM proteins (■). Data shown are averages from three independent experiments with S.D. (error bars).

Homo-oligomerization of Bcl-2ΔTM significantly enhances its pore formation

Next, by using different form of Bcl-2ΔTM, we examined the effect of homo-oligomerization on Bcl-2ΔTM pore formation. During purification of Bcl-2ΔTM protein, we observed that Bcl-2ΔTM protein was eluted from a Superdex 200 gel filtration column mainly in the monomeric form and an oligomeric form (Fig. 3A). The protein in both fractions is >90% pure according to SDS-PAGE analysis followed by Coomassie Blue staining (Fig. 3B); and the identity of purified protein was confirmed by Western Blot using antibody specific for human Bcl-2 (Fig. 3C). The Bcl-2ΔTM monomer and oligomer are quite stable since more than 95% of the protein was eluted in the same fraction after a second chromatography using the same column (data not shown). This provided us an opportunity to test which form of Bcl-2ΔTM protein is more efficient in pore formation. As shown in Fig. 3D, both monomeric and oligomeric Bcl-2ΔTM proteins can release the same amount entrapped 3 kDa CB-dextran molecules from the liposomes after 5 hr incubation. However, the kinetics of dye release by the two forms of Bcl-2ΔTM is different. The rate of release mediated by the oligomeric Bcl-2 ΔTM is much higher than that by the monomeric one with t_1/2 being ∼120 sec for the oligomer and ∼1500 sec for the monomer. These results demonstrate that, although both Bcl-2ΔTM oligomer and monomer are competent for the pore formation, the oligomer is much more efficient in the pore formation than the monomer, indicating that the oligomerization of Bcl-2ΔTM is an intermediate step during the pore forming reaction.

Figure 3. Effect of Bcl-2ΔTM homo-association on pore formation.

A. Homo-association of Bcl-2ΔTM was examined by gel filtration chromatography. The A₂₈₀ was monitored for each eluted fraction and plotted. The elution positions for protein standards are indicated at the top of the plot. B & C. The purity or identity of Bcl-2ΔTM from both 200kDa-fraction (lane1) and 25kDa-fraction (lane2) was analyzed by 15% SDS-PAGE and Coomassie staining (B), or by Western Blotting using a polyclonal antibody specific for human Bcl-2 (C), respectively. D. Release kinetics of 3 kDa CB-labeled dextrans from 100 μM liposome at pH 5.0, by 0.8 μM of wild-type Bcl-2ΔTM monomers (●); oligomers (○); or protein buffer (□); and E. by 0.8 μM of wild-type Bcl-2ΔTM monomer (●); G154A/G155A monomer (○); or protein buffer (□). Data shown are average from three experiments with S.D. (error bars).

To further test the effect of homo-association of Bcl-2ΔTM on its pore formation, we generated a mutation (G154A/G155A) in the helix 5 of Bcl-2 which has a 7-fold higher homo-association affinity than the wild type protein based on the measurement of tryptophan anisotropy (26), and then compared the pore-forming activity of monomeric wild-type Bcl-2ΔTM and its mutant. As shown in Fig. 3E, the mutant had a much faster kinetics than the wild type, with a t_1/2 about 4 times shorter than that for the wild type, further supporting the conclusion that homo-association of Bcl-2ΔTM facilitates its pore formation.

Bcl-2ΔTM oligomerization can be detected by chemical cross-linking, which is enhanced by membrane association

To examine whether oligomers could be observed directly, a homobifunctional sulfhydryl-reactive cross-linker BMH was used to crosslink Bcl-2ΔTM proteins. The cross-linking is via the only cysteine (Cys158) in each Bcl-2ΔTM protein; therefore only the dimer adduct can be generated even though some proteins are in oligomeric form as shown by the size-exclusion chromatography (Fig. 3A).

After incubation of Bcl-2ΔTM and liposomes in presence of BMH at pH 5.0, the sample was fractionated by a 3-step sucrose gradient float-up centrifugation. As shown in Fig. 4, a BMH-dependent Bcl-2ΔTM dimmer adduct was detected both in membrane and solution. The specificity of BMH reaction was demonstrated by that no dimer adduct was observed even in the presence of BMH when a cysteine-null mutant of Bcl-2ΔTM (Cys 0) was used (data not shown). Therefore, two Bcl-2ΔTM proteins can be cross-linked via their cysteine side chains by the thio-specific BMH both in solution and membrane. Furthermore, data in Fig. 4 also showed that the dimerization activity of membrane-bound Bcl-2ΔTM was much higher than that of soluble Bcl-2ΔTM in solution, suggesting that membrane association facilitates the homo-oligomerization of Bcl-2ΔTM.

Fig. 4. The oligomerization of Bcl-2ΔTM was detected by chemical cross-linking.

After incubation of Bcl-2ΔTM in presence (A) or absence of liposomes (B) at pH 5.0, the sample was cross-linked by BMH and fractionated by a step sucrose gradient float-up centrifugation. The liposome-bound protein migrated into the upper supernatant fractions (S1-2), whereas the soluble protein remained in the bottom fractions (S3-4). The aggregated protein was in the pellet fraction (P). Data shown are representative Coomassie stained SDS-PAGE gels of three independent experiments. The distribution of liposomes in the gradient was monitored by counting the ¹⁴C radioactivity and the results are indicated below each lane as the percentage of total liposomes (A).

Discussion

Our study here indicates that the Bcl-2 pore size is dependent on the protein concentration, and the release kinetics of small and large molecules is different at a given Bcl-2 concentration, suggesting that the Bcl-2 pore is likely to be formed through homo-association of multiple Bcl-2 molecules in the membrane. In addition, compared with wild-type Bcl-2 monomer, the pre-formed wild-type Bcl-2 oligomers and the mutant Bcl-2 (G154A/G155A) monomers with a higher homo-association affinity display a faster pore-forming kinetics, further supporting that oligomerization is involved in Bcl-2 pore formation. Moreover, Bcl-2 oligomerization can be directly observed both in solution and in membrane.

One of the current strategies for anti-cancer drug development is using BH3 memitics to bind to Bcl-2-like proteins and neutralize their survival function. Previously we reported the in vitro activity of Bcl-2 to form small pores might be correlated with its anti-apoptosis function in vivo (16). Together with the finding that oligomerization is involved in Bcl-2 pore formation, it is reasonable to consider that oligomerization might be one of the molecular mechanisms by which Bcl-2 inhibits apoptosis. Therefore, discovering small molecules to inhibit the homo-association of Bcl-2 may be a novel anti-tumor therapeutic approach.