Lipidic pore formation by the concerted action of pro-apoptotic BAX and tBID

Unidad de Biofísica (Centro Mixto Consejo Superior de Investigaciones Cientificas– Universidad del Pais Vasco/Euskal Herriko Unibertsitatea), Universidad del Pais Vasco/Euskal Herriko Unibertsitatea (UPV/EHU), P.O. Box 644, 48080 Bilbao, Spain. Serono Pharmaceutical Research Institute, Serono International S.A., 14, chemin des Aulx, CH1228 Plan-les Ouates, Geneva, Switzerland. Drug Discovery Program, H. Lee Moffitt Cancer Center & Research Institute, Tampa, Florida 33612 Huntsman Cancer Institute and Departments of Internal Medicine and Oncological Sciences, University of Utah, Salt Lake City, UT, 84112 Institut für Biologie, Molekulare Biophysik, Humboldt-Universität Berlin, Invalidenstr. 42, D10115, Berlin, Germany


9
FD, λex was 490, and λem was 520 nm (slits, 4 nm); for ANTS/DPX, λex was 350 nm, and λem was 520 nm (slits, 8 nm). A 515 nm cut-off filter was placed between the sample and the emission monochromator to avoid scattering interferences. The extent of marker release was quantified on a percentage basis according to the equation: [(F t -F 0 /F 100 -F 0 ) X 100], where F t is the measured fluorescence of protein-treated LUV at time t, F 0 is the initial fluorescence of the LUV suspension before protein addition, and F 100 is the fluorescence value after complete disruption of LUV by addition of C 12 E 8 (final concentration, 0.5 mM). Unless otherwise stated, lipid concentration was 20 µM.
Assays of BAX Binding to and Insertion into the LUV Membrane-To measure the amount of BAX bound to LUV, a method was used based on the fact that lipid-associated protein, but not free protein, floats in D 2 0-based KHE buffer. Briefly, proteins and LUV were incubated together at 37 0 C in D 2 O-based KHE buffer (final volume of reaction mixture, 100 µl), followed by centrifugation of the mixture for 1 hour at 100 000 X g, room temperature. Under these conditions, LUV remained in the upper fraction of the buffer, whereas free BAX protein sedimented. The top 20-µl fraction of the gradient, corresponding to the lipid-associated fraction, was subjected to SDS-PAGE in 15% Tris-glycine gels, followed by BAX visualization by Western blotting using N20 anti-BAX polyclonal antibody (Santa Cruz Biotechnology, Inc.).
To discriminate between protein inserted into and only peripherally adsorbed to the membrane surface, the same protocol as above was followed except that a second incubation was performed for 30 minutes at pH 11.5, the alkaline pH being maintained during sample centrifugation. Upon alkaline pH incubation, the fraction of protein integrated into the membrane hydrophobic matrix Hercules, CA). Samples of 300 µl were loaded onto the column, followed by collection of 2-ml elution fractions. Aliquots of individual fractions were subjected to SDS-PAGE in 15% Trisglycine gels, followed by visualization of BAX using N20 anti-BAX antibody.

Measurements of LUV Size by Quasi-Elastic Light Scattering (QELS)-Vesicle size was
determined by QELS at a fixed angle of 90 o and room temperature, using a Malvern Zetasizer 4 instrument (Malvern, UK). A 64-channel correlator was used capable of estimating particle sizes in the range from 5 nm to 5000 nm. Data were analyzed by the cumulant method using Malvern Application Software. The hydrodynamic radius of the particle was obtained from the first cumulant, whereas the polydispersity of the sample was obtained from the second cumulant.
Assays of Lipid Transbilayer Movement in LUV-Assays of lipid transbilayer motion were done using 1-lauroyl-2-(1'pyrenebutyroyl)-sn-glycero-3-phosphocholine (pyPC), as described before (41,42). Briefly, pyPC dissolved in an ethanolic solution was added externally to the liposome suspension at a final concentration of 5 mol % (pyPC/total lipid), leading to its incorporation only in the external monolayer of the membrane. When pyPC translocates from the external by guest on September 1, 2017 http://www.jbc.org/ Downloaded from membrane monolayer to the internal membrane monolayer the probe is diluted, leading to a decrease in the ratio of fluorescence intensities of pyPC excimers (I E ) and pyPC monomers (I M ).
The degree of pyPC transbilayer redistribution (q) was estimated from measured I E /I M ratios, using a calibration curve obtained with LUV containing different molar amounts of pyPC, as described before (41).

Results and discussion
tBID, but not BIM EL , works in conjunction with BAX to form large membrane lesions in LUV-To elucidate the mechanism(s) by which BID and BIM assist BAX in the release of mitochondrial intermembrane apoptogenic factors, we first used an in vitro reconstitution system consisting of mitochondria freshly isolated from rat liver (which contain minimal amounts of BAX and BAK) (43) and various combinations of recombinant purified pro-apoptotic proteins.
We used physiologically-relevant concentrations of full-length monomeric BAX (BAX) (44), together with increasing concentrations of tBID or a form of BIM previously shown to induce BAX-dependent mitochondrial cytochrome c release in cultured cells (BIM EL ) (19). Soluble fractions were separated from mitochondria by centrifugation, and assayed for the presence of freed cytochrome c by immunoblotting. As shown in Fig. 1A, the combination of BAX with either tBID or BIM EL led to release of mitochondrial cytochrome c, whereas none of these proteins alone released substantial cytochrome c under conditions tested (Fig. 1A, and data not shown). However, tBID released mitochondrial cytochrome in the presence of BAX with ~5-fold higher potency compared to BIM EL (Fig. 1A).
Next, we wished to test whether in the absence of additional proteins, tBID and/or BIM EL can assist BAX to release macromolecules of the size of cytochrome c (~14 kDa), as well as other larger pro-death proteins released from mitochondria during apoptosis, i.e. Smac/DIABLO (~23 kDa), EndoG (~30 kDa), Omi/HtrA2 (~37 kDa), and AIF (~57 kDa) (3). To this aim, we shown that the combination of BAX and cBID can elicit the release of 2000-kDa dextrans from pure lipid vesicles (35). We found that the combination of BAX and tBID is much more potent in the release of vesicular FD-70 than the BAX+cBID combination (Fig. 1B). This finding is consistent with evidence showing that BID possesses an auto-inhibitory N-terminal domain which remains attached to the apoptogenic C-terminal part of the molecule after caspase-8 cleavage (15). In sharp contrast to the situation found with tBID/cBID, addition of BIM EL together with BAX resulted in virtually no vesicular FD-70 release, even at concentrations of BIM EL 10-fold higher than those used in the tBID/cBID assays (Fig. 1B).
Anti-apoptotic proteins such as BCL-2 antagonize the release of mitochondrial pro-death factors during apoptosis (1)(2)(3)(4)7). To analyze this issue in our in vitro reconstitution systems, isolated rat liver mitochondria and LUV were incubated with BCL-2 prior to the BAX+tBID treatment. BCL-2 effectively blocked the release of mitochondrial cytochrome c as well as that of LUV-entrapped FD-70 induced by the BAX+tBID mixture ( Fig. 1, C, D). Moreover, the finding that BCL-2 inhibition occurred with a similar dose-dependence in both experimental setups adds credit to the physiological significance of the LUV system.
BH3-only proteins are thought to heterodimerize with other BCL-2 family members through insertion of their BH3 domain into a hydrophobic groove localized on the surface of binding partners (2,4,9). The three-dimensional structure of BAX revealed that a carboxylterminal helix folds into the groove to which BH3 ligands presumably bind (45). To test whether this could explain the lack of effect of the BAX+BIM EL mixture in LUV, we used a deletion mutant of BAX lacking its carboxyl-terminal hydrophobic region (BAX∆C). Additionally, we In summary, these results indicate that cBID and, more efficiently, tBID cooperate with BAX to form membrane openings in pure lipid vesicles large enough as to allow transit of mitochondrial intermembrane apoptogenic proteins. In sharp contrast, BIM EL lacks this capacity.
tBID can work in concert with BAX to induce membrane damage in a BH3-independent manner-It is well established that the BH3 domain of tBID is crucial for its bioactivity. To address this, we used two BH3 mutants of BID with reduced affinity for BAX: cBid mIII-2 ( 93 IGDE 96 →AAAA) and cBid mIII-3 (G 94 →A) (46,47). We compared the effects of increasing concentrations of cBID, cBID mIII-2 and cBID mIII-3 on the release of FD-70 from LUV treated with fixed amounts of BAX. cBID mIII-2 showed ~3-4-fold reduced efficiency in vesicular FD-70 release relative to cBID, whereas cBID mIII-3 showed similar efficiency in FD-70 release relative to cBID (Fig. 2). Previous studies showed that BID mIII-3 and BID mIII-2 possess at least 10-fold lower affinity for BAX (47). This observation, together with the paradoxical finding that exposure of the BH3-binding cleft of BAX did not increase but actually decreased tBID/cBID-assisted vesicular contents release, prompted us to investigate the possibility that tBID utilizes BH3-independent mechanisms to aid BAX in membrane damage induction.
In healthy cells BAX exists in a monomeric inactive state (2). BAX activation during apoptosis has been related to a change in conformation of the protein leading to its mitochondrial membrane insertion and oligomerization, and tBID has been suggested to trigger both events through binding to BAX via its BH3 domain (47)(48)(49). Specific detergents such as octylglucoside (OG) can induce BAX membrane integration and oligomerization (22,34,35,38), and several groups have shown that such OG-activated BAX (OG-BAX) permeabilizes both outer mitochondrial (35,38,50,51) as well as pure lipid membranes (13,22,34,35,38). Moreover, OG-activated BAX and tBID-activated BAX display similar degrees of membrane integration and oligomerization in LUV (22, and data not shown). Thus, we decided to use OG-BAX as a tool to test whether tBID can collaborate with BAX through mechanisms other than BH3triggered BAX membrane integration and oligomerrization. OG-BAX alone released LUVentrapped FD-70 efficiently (data not shown), consistent with previous findings (35).
Remarkably, when LUV were treated with suboptimal doses of OG-BAX, subsequent addition of tBID, but not BIM EL or tBID buffer alone, led to additional vesicular FD-70 release (Fig. 3A).
Since OG-BAX inserted in the LUV membrane in an alkali-resistant manner to the same degree with or without tBID treatment ( One possibility to explain the above observations is that a membrane-inserted portion of tBID assists OG-BAX in membrane damage induction. A domain was identified in BID, distinct from BH3, which possesses potential for membrane insertion and targets the molecule to CLcontaining membranes, the so-called cardiolipin-binding-domain (CBD) of BID (27,28). When FD-10-loaded vesicles were pre-treated with sub-optimal OG-BAX doses, subsequent addition of recombinant CBD induced marker release without increasing the degree of BAX membraneinsertion or multimerization ( Fig. 3B-E). CBD did not induce substantial FD-10 release in OG-BAX-treated LUV in which cardiolipin had been substituted for by equimolar amounts of phosphatidylserine (data not shown), consistent with the selective preference of CBD for CL (27,28). Of note, however, CBD was less efficient than tBID in eliciting FD-10 and FD-70 release from OG-BAX-treated LUV (Fig. 3C). Also of interest, CBD did not cause vesicular contents release from LUV pre-treated with monomeric BAX (Fig. 3B), in agreement with the notion that the BH3 domain of BID (absent in CBD) is required for activation of BAX membrane insertion and oligomerization.
Collectively, these results suggest that tBID works in concert with BAX to form large membrane openings in LUV through two distinct mechanisms: (1) activation of BAX integral membrane insertion and oligomerization via physical interaction of the BH3 domain of tBID with BAX, and (2) a heterodimerization-independent mechanism via non-BH3 regions of tBID which include, but are not limited to, the CBD.
Lipidic pore formation by the concerted action of BAX and tBID-The finding that the membrane lesion caused by BAX in conjunction with tBID/cBID did not discriminate permeants according to size (Fig. 1E) raised the possibility that pro-apoptotic proteins may act by solubilizing LUV in a "detergent-like" manner. For example, BAX together with tBID/cBID may induce micellization of the membrane of LUV, breaking it up into small fragments or disklike structures as seen with a number of membrane-disrupting antimicrobial peptides, including melittin (52). Evidence for such a detergent-like action can be obtained by analysing changes in the size distribution of LUV using QELS. This approach was utilized here to evaluate whether pro-apoptotic proteins caused FD-10 release through vesicle fragmentation. Results were compared to those obtained with melittin, used as a positive reference. As shown in Table I (Table II). DOG possesses an intrinsic curvature more negative than DOPE (57), and DOG inhibited vesicular FD-70 release to a higher extent than DOPE (Table II) (Table II).
Changes in neither the degree of BAX membrane insertion nor its oligomeric status could provide a complete explanation for the effects of non-bilayer lipids on vesicular FD-70 release (data not shown). Based on analogous results, we previously proposed that OG-activated BAX induces lipidic pore formation by promoting positive membrane monolayer curvature (34). We  (Table II).  Fig. 1B and Fig.  5A). Next, the degree of pyPC transbilayer redistribution between outer and inner monolayer (q) was determined. The results obtained with different combinations of apoptotic proteins correlated with those obtained in analogous experiments of vesicular FD-70 release (Fig. 5B).
Last, the influence of membrane monolayer curvature on pyPC transbilayer movement was analyzed. As in the case of vesicular contents release, positive and negative membrane monolayer curvature increased and decreased, respectively, the degree of pyPC transbilayer motion induced not only by the BAX+tBID mixture, but also by the OG-BAX+tBID mixture (Table II) (Table II).
Taken together, these observations strongly suggest that the release of vesicular contents induced by BAX and tBID in LUV is mechanistically related to lipid transbilayer movement, and add further support to the notion that BAX and tBID cooperate at different stages of a molecular pathway culminating with lipidic pore formation.

Studies with biomembrane mimetic LUV of outer mitochondrial membrane contact site
composition-BAX (59, 60), tBID (61) and BCL-2 (62, 63) have been shown to localize at zones of close proximity between the outer and the inner mitochondrial membranes, the so-called mitochondrial membrane contact sites. Thus, we decided to examine the effect of these proteins in LUV bearing the phospholipid composition determined for the outer mitochondrial membrane at such contact sites (64). Neither BAX nor tBID alone had any effect on such LUV (data not shown). However, addition of the two proteins together caused vesicular FD-70 release and pyPC transbilayer redistribution with similar time courses and comparable susceptibilities to BCL-2 inhibition (Fig. 6A and Table III). cBID functioned together with BAX in a similar manner causing both FD-70 release and lipid transbilayer movement, although with reduced potency compared to tBID (Fig. 6B and Table III). Once again, however, BIM EL was unable to cooperate with BAX to induce any of these effects in LUV of outer mitochondrial membrane contact site composition (Table III).
Next, we sought to determine whether CL is required for the membrane actions of BAX and tBID. To this aim, LUV were prepared of outer mitochondrial membrane contact site composition but excluding all acidic lipids (-), and with CL being substituted by PS, phosphatidylinositol (PI), or phosphatidylglycerol (PG) (at same mole ratios of acidic lipids) (Fig. 6A). The BAX+tBID mixture did not permeabilize LUV of outer mitochondrial membrane contact site composition in which all acidic lipids had been excluded. Substitution of CL by PI or PS caused much reduction in the release of vesicular FD-70 elicited by BAX together with tBID. Interestingly, when CL was substituted by PG, the BAX+tBID mixture caused vesicular FD-70 release with only slightly lower efficiency as compared to CL-containing LUV (Fig. 6A).
Moreover, pyPC transbilayer redistribution accompanied the release of vesicular FD-70 in PGcontaining LUV, and both processes were inhibited by BCL-2 ( Figure 6A, and Table III). Thus, we concluded that PG can replace CL in formation of large lipidic pores by BAX and tBID. This finding may be of physiological relevance considering that (i) PG and CL are structurally-and biosynthetically-related lipids primarily localized to mitochondrial membranes (65), (ii) mitochondrial PG levels increase early during Fas and radiation-induced apoptosis (21,66), and (iii) BAX can induce cytochrome c release in mitochondria from mutant yeast lacking CL, but containing increased PG levels (67).
What remains to be addressed is the underlying mechanism of CL and PG selectivity in the concerted action of BAX and tBID. Because CL, PG, PS and PI, all posses near-zero intrinsic curvature under conditions tested here (68), membrane curvature changes implicated in lipidic pore formation are unlikely to be the source of the CL/PG preference over PS/PI.
Interestingly, recent studies showed that during Fas-induced apoptosis the amount of mitochondrial CL decreases whereas those of monolysocardiolipin, lysophosphatidylglyerol and lysophosphatidylcholine increase coincidentally with cytochrome c release (21,31). Considering that lysolipids, in general, possess positive intrinsic curvature (54), it is tempting to speculate that during apoptosis such lysoderivatives rather than their parental species may be sequentially and/or cooperatively implicated in lipidic pore formation by BAX and tBID/cBID.
Finally, we performed additional studies to gain more insight on the mechanism of action of BIM EL . Since no evidence was found that BIM EL can assist BAX in the release of large molecular weight dextrans from LUV, yet BIM EL was able to assist BAX in cytochrome c release from isolated rat liver mitochondria, we reasoned that additional endogenous mitochondrial factors may be required to reconstitute the BIM EL -BAX cooperation in LUV. It has been proposed that BH3-only proteins can be subdivided into two classes: those that can bind to and activate BAX-type proteins directly, and those that can render cells more susceptible to apoptogenic stimuli through binding to and inactivation of BCL-2-type proteins (8,10).
Although it has been difficult to capture BIM EL -BAX complexes in vivo (8,20,(69)(70)(71)(72), evidence indicates that BIM EL can bind to BCL-2 in several cellular contexts (69)(70)(71)(72). Thus, we assessed whether, by virtue of its putative capacity to bind to BCL-2, BIM EL could reverse BCL-2 inhibition of LUV permeabilization by the concerted action of BAX and cBID. BCL-2 exerted a dose-dependent inhibition on the release of vesicular FD-70 elicited by the BAX+cBID mixture in LUV of outer mitochondrial membrane contact site composition (Fig. 6B). Addition of equimolar concentrations of BIM EL compared to cBID did not affect vesicular FD-70 release.
However, augmentation of BIM EL concentration led to a progressive increase of vesicular FD-70 release, and at a 6-fold higher concentration of BIM EL relative to cBID a complete reversal of the BCL-2 inhibition was achieved (Fig. 6B). BIM EL did not enhance the release of vesicular FD-70 mediated by the BAX+cBID combination in the absence of BCL-2, arguing that BIM EL acts through BCL-2 rather than through BAX and/or cBID.
Concluding remarks-In summary, we showed that tBID/cBID and BIM EL use different mechanisms to aid BAX in making lipid vesicles permeable to macromolecules as large as prodeath proteins released from mitochondria during apoptosis. Several findings support the idea that tBID/cBID assists BAX directly in LUV permeabilization through formation of large, lipidcontaining pores: (i) the observation that BAX together with tBID/cBID induce the release of all molecules entrapped within LUV irrespective of their size; (ii) the strong coupling between this vesicular contents release and the redistribution of lipids from one monolayer of the membrane to the other, (iii) the dependence of both vesicular contents release and lipid transbilayer redistribution on membrane monolayer curvature, and (iv) the lack of effect of pro-apoptotic proteins on LUV size, which argues against a detergent-like mechanism of action. An additional important highlight of our study is that tBID utilizes BH3-dependent as well as BH3-independent mechanisms to assist BAX in LUV permeabilization. Based on these observations, we propose a new model in which tBID aids BAX to elicit outer mitochondrial membrane permeabilization not only by triggering BAX oligomerization and membrane insertion, but also by increasing outer membrane curvature in such a way as to allow formation/expansion of a positively-curved lipidic pore.
In contrast to tBID/cBID, our results with LUV suggest that BIM EL aids BAX in outer membrane permeabilization indirectly, through binding to and neutralization of BCL-2-type proteins. However, we cannot rule out the possibility that other isoforms of BIM or a posttranslationally modified form of BIM EL may assist BAX in a more direct manner. Additionally, it is possible that mitochondrial components unrelated to BCL-2 family proteins (e.g. VDAC) play an important role in BIM EL bioactivity (12). Further studies are required to test these possibilities, as well as to examine the applicability of the lipidic pore model and its postulates for the increasing number of molecules proposed to collaborate with BAX-type proteins in permeabilization of the outer mitochondrial membrane during apoptosis (36,(73)(74)(75).