The Mitochondrial Apoptosis-induced Channel (MAC) Corresponds to a Late Apoptotic Event*

We have investigated the mechanism responsible for mitochondria permeabilization occurring during cell apoptosis. We have developed an in vivo model of apoptotic rat liver. Mitochondria appeared as an homogenous population in control liver. On the contrary, mitochondria varied in size, morphology, and the matrical density in apoptotic liver. Mitochondria were purified from control and apoptotic livers. In control conditions, a single mitochondrial population was identified; whereas three populations of mitochondria were purified from apoptotic liver. Our data show that these apoptotic populations correspond to early, intermediate, and late apoptotic mitochondria, which are characterized by an increasing extent of permeabilization of their outer membrane and a gradual enrichment in oligomerized Bax protein. Remarkably, a new ionic channel was observed in apoptotic but not in control mitochondria.

Cell apoptosis is of primary importance in the control of organ and tissue development, as well as in cell organization and homeostasis (reviewed in Ref. 1). Defaults in the execution or regulation of the apoptotic program have been involved in proliferative as well as degenerative pathologies (reviewed in Ref. 2).
Apoptosis results from highly coordinated biochemical events (reviewed in Ref. 3), among which the recruitment of pro-apoptotic members of the Bcl-2 family, such as Bax or Bak at the mitochondria constitutes a major checkpoint conditioning the engagement of cell degradation (4). It is now well admitted that mitochondria undergo general permeabilization upon their interaction with Bax or Bak proteins (5). Following this, apoptogenic proteins of the mitochondrial intermembrane space are released into the cytosol, thus triggering the activation of caspase-dependent and caspase-independent degradation of the cell compounds. Interestingly, the release of proapoptotic mitochondrial factors occurs with different kinetics (6,7) indicating that mitochondrial outer membrane (MOM) 1 permeabilization proceeds through kinetically and perhaps mechanistically distinct steps.
Despite the wealth of in vitro and in vivo investigations, the molecular mechanism responsible for Bax-dependent MOM permeabilization is still controversial and several, although non-exclusive, hypothetical models have been considered (8). In an early model, Bax has been described as an efficient activator of the mitochondrial permeability transition pore (PTP) (9,10). Accordingly, Bax-induced PTP activation leads to an outer membrane rupture through mitochondrial swelling. Supporting this, modulators of PTP activity modulate Bax-induced mitochondrial permeabilization (reviewed in Ref. 11). Furthermore, Bax can promote new channel activity in artificial membrane bearing adenine nucleotide translocators (12) and can also modulate adenine nucleotide translocator channel activity (13). However, in several cases of Bax-mediated cell apoptosis, mitochondrial swelling is not observed and has been reported to be independent of a PTP activation (14 -16). An alternative model has been developed in which Bax inserts MOM and forms a pore facilitating intermembrane space protein diffusion through MOM. This is supported by reports describing Baxinduced channel formation into planar lipid bilayer and liposomes (17)(18)(19)(20)(21)(22), although Bax-induced pore permeability toward cytochrome c is not clearly demonstrated (23). In another hypothesis, Bax modifies resident MOM channels. Strengthening this, an increase in VDAC conduction properties has been reported upon its interaction with Bax (20,24,25). Recently, patch clamp investigations of MOM electrophysiological properties revealed a new channel (the so-called MAC for mitochondrial apoptosis-induced channel) in mitochondria purified from an apoptotic cell line and bax-expressing yeast were investigated (26). MAC corresponded to a large conductance, and cytochrome c interacting channel (26 -28). Accordingly, MAC has been postulated to correspond to the channel responsible for cytochrome c efflux through MOM during cell apoptosis.
In the present work, we have investigated the relevance of MAC in a more physiological model of apoptosis induced in rat liver. Our results show that in vivo MAC is a functional component of the envelope of apoptotic mitochondria, the biophysical and pharmacological properties of which are established here for the first time. However, because MAC is specifically observed in late apoptotic mitochondria, we propose that MAC does not correspond to the structure responsible for the early mitochondrial permeabilization observed in cell apoptosis.

EXPERIMENTAL PROCEDURES
Animals and Chemicals-Escherichia coli LPS serotype 0127:B8, D-galN, apo-cytochrome c, cytochrome c, lysozyme, cyclosporin A, and atractyloside were from Sigma. All salts and buffers used here were analytical grade.
In Vivo Liver Apoptosis-200 -250 g of BD IX rats received an intraperitoneal injection of LPS (50 mg kg Ϫ1 body weight) and D-galN (1 g⅐ kg Ϫ1 body weight) solubilized in sterile NaCl (treated rats) or an injection of sterile NaCl (0.9%, w/v) only (sham-treated rats). 5 h and 30 min after the beginning of the treatment, sham-treated and treated rats were euthanized by a lethal injection of ketamine (10 mg/ml) and xylazine (0.8% w/v). Animals were starved from diet but provided with fresh water within the 6 h preceding their euthanasia.
Biochemistry-Purification of liver mitochondria was performed at 4°C as described earlier (29). In brief, liver was harvested immediately after animal euthanasia and was homogenized with 5 volumes of an extraction buffer containing 250 mM sucrose, 5 mM Hepes-K, pH 7.0, in an Elvehjem motor driven Teflon pestle (20 strokes, 1,000 rpm). Liver homogenate was cleared from cell debris and nuclei by a 1,200 ϫ g centrifugation (20 min) and the crude mitochondrial fraction was pelleted by a 8,700 ϫ g centrifugation (15 min). After two washes by 300 mM mannitol, 10 mM MOPS (pH 7.0), crude mitochondrial fraction was layered onto the top of a discontinuous Percoll TM gradient prepared in 300 mM mannitol, 10 mM MOPS (pH 7.0) and consisting of 2 ml of Percoll TM 70%, 3 ml Percoll TM 30%, 2 ml of Percoll TM 18%, and 2 ml of Percoll TM 10%. Gradients were centrifuged (9,000 ϫ g, 45 min), and the different mitochondrial fractions were collected at the different Percoll TM interfaces. Mitochondria were washed twice with 300 mM mannitol, 10 MOPS (pH 7.0) and were kept at Ϫ80°C after their freezing in liquid N 2 . The liver cytosolic fraction was obtained from liver homogenate cleared from intracellular membranes by centrifugation (100,000 ϫ g, 90 min). Proteins into the different samples were determined using bovine serum albumin as a standard.
DEVDase activity was measured on liver cytosolic fractions originating from sham-treated and treated rats, using the fluorogenic caspase-3-specific substrate (Ac-DEVD-AMC, Promega), added in the presence of Triton X-100 (0.01%, v/v) (29). The assay consisted in a quintuplet measurement of DEVDase activity performed for 3 sham-treated and 3 treated rats. The data corresponded to the mean Ϯ S.E. of the results and were expressed in arbitrary units.
Microscopy-Fresh livers were rinsed in phosphate-buffered saline (PBS) at 4°C. They were then fixed in phosphate-buffered saline ϩ paraformaldehyde (4% w/v, 4°C, overnight) and further incubated during 24 h at room temperature. After embedding in paraffin, tissue was cut in 5-m sections. Sections were stained with hematoxylinphloxin and observed with a conventional visible light microscope (Olympus BH2).
For electronic microscopy, liver samples from control and treated rats were fixed in sodium cacodylate (100 mM, pH 7.2) containing 3% (w/v) OsO 4 , rinsed by 300 mM sodium cacodylate (pH 7.2) and 64.5 mM NaCl, then dehydrated with ethanol at increasing grades. Samples were washed by propylene oxide before their inclusion in Epon 812. Ultrathin slices were stained by uranyl acetate and lead citrate. Observations were performed onto a TEM JEM 1010 (Philips).
Electrophysiology-Mitochondrial membranes were harvested (100,000 ϫ g, 30 min) from mitochondria submitted to an osmotic down-shock in 10 mM Na 2 HPO 4 /NaH 2 PO 4 (pH 7.0). Giant proteoliposomes were formed as described earlier (30) by mixing 10 -50 g of mitochondrial membrane proteins with 1 mg of azolectine (soybean type IV, Sigma). The mixture was frozen (liquid N 2 ) and thawed (room temperature) 3 times and 10-l aliquots were spotted onto a 96-well plate. Samples were dehydrated under vacuum at room temperature (3 h) and rehydrated overnight (4°C) in the presence of 150 mM KCl, 20 mM MgCl 2 , 10 mM Hepes-K (pH 7.0). The proteoliposomes formed in these conditions were layered onto a 35-mm Petri dish containing 50 mM KCl, 10 mM Hepes-K (pH 7.0).
Single channel experiments were conducted using 20 -25 M⍀ patch clamp pipettes filled with 150 mM KCl, 10 mM Hepes-K (pH 7.0). After sealing, the proteoliposome membrane was excised and rapidly air/ water interfaced to prevent further membrane vesicularization. The membrane patch was superfused using an homemade local perfusion system containing different media. Single channel currents were monitored using the voltage-clamp configuration of the patch clamp technique. Applied voltage (V m ) corresponded to (V pip Ϫ V bath ) with V bath virtually grounded to 0 mV. Currents were amplified using an Axopatch 200B apparatus (Axon Instruments, Union City, CA) and digitized using a Labmaster A/D converter (Scientific Solution). Current traces were stored on a digital audio tape DTR 1202 (Bio-Logic, Pont de Claix, France), then played back with a low-pass filter frequency between 1.2 and 2.5 kHz (Burr-Stroud EF5-20 filter). Sampling frequency adjustment (between 4 and 7.5 kHz) and current analysis were performed by using pClamp 9.0 software (Axon Instruments).
Single channel conductance (␥) was calculated from the currentvoltage (I/V) relationship. Theoretical diameter of the pore (Ø) was deduced from the following equation, where ␥ is the channel conductance (in S), is the resistivity of the solution (ϳ0.8 ⍀.m), and l is the minimal length of the pore (7 ϫ 10 Ϫ9 m), which corresponds to the membrane width. Ion channel selectivity was deduced from the current reverse potential value (E rev ) introduced into the Goldman-Hodgkin-Katz equation. Channel open probability (P o ) was measured on current traces lasting at least 1 min. Data Analysis-Images were generated from scanned pictures using an Epson 2120 photo deluxe scanner (1200 dpi resolution) and Photoshop 6.0 (Adobe) software. Numerical and statistical data were analyzed using Sigma-Plot and Sigma-Stat (Jandel Scientific) software.

In Vivo LPS and D-galN Treatment Induces Liver Apoptosis
That Is Related to the Mitochondrial Pathway-In mouse, the intraperitoneal injection of LPS and D-galN has been shown to promote fulminant hepatitis through the recruitment of the  tumor necrosis factor-␣/tumor necrosis factor receptor pathway, Bid protein proteolytic activation, Bax protein targeting to mitochondria, and the subsequent release of cytochrome c (31,32). In the present study, we have verified that rat liver developed a massive apoptosis upon LPS ϩ D-galN treatment. Liver parenchymal organization exhibited strong histological alterations with large foci of inflammation (Fig. 1, A and B) upon the treatment by LPS ϩ D-galN. Vacuolized cells as well as cells with fragmented nuclei could be identified and numerous apoptotic bodies were also observed in these conditions (Fig. 1, A and B). The cleavage of the pro-apoptotic protein Bid was clearly observed in liver cytosol originating from treated rats (Fig. 1C). The Bid/t-Bid ratio decreased in treated rats (1.56 Ϯ 0.26, n ϭ 3) when compared with that of sham-treated rats (4.5 Ϯ 0.52, n ϭ 3). Finally, a ϳ3 times enhancement of caspase-3 activity was observed in the cytosolic fraction of treated rats when compared with that of sham-treated ones (Fig. 1D). Taken as a whole, our data show that rat liver underwent a general mitochondrial apoptosis upon LPS ϩ D-galN treatment, thus, establishing for the first time a rat model of liver cell apoptosis. Different Mitochondrial Populations Are Produced during in Vivo Liver Apoptosis-The ultrastructure of control and apoptotic livers was examined (Fig. 2, A and B). Control hepatocytes contained homogenous mitochondria regarding their size, morphology, and matricial electronic density ( Fig. 2A). On the contrary, a large mitochondrial pleomorphism could be seen in apoptotic hepatocytes (Fig. 2B). We identified a co-existence of normally sized mitochondria with undamaged envelope and numerous inner membrane crests with swollen mitochondria exhibiting no inner membrane crest and an empty matrix (Fig. 2B).
Mitochondria were purified through density gradient. Control mitochondria resulted in a single population (the so-called HCM for heavy control mitochondria), at the 1.051/1.096 g/ml density interface (Table I). It represented 83.2 Ϯ 4.8% of the crude fraction, the remaining material being dispersed into the gradient. Apoptotic mitochondria were separated in three populations, namely HAM for heavy apoptotic mitochondria, LAM1 for light apoptotic mitochondria fraction 1, and LAM2 for light apoptotic mitochondria fraction 2 (Table I). HAM density corresponded to that of the HCM fraction, whereas LAM1 and LAM2 fractions exhibited lower densities (Table I). HAM, LAM1, and LAM2 fractions accounted for 66.5 Ϯ 5.0, 16.3 Ϯ 3.4, and 10.2 Ϯ 4.6% of the crude apoptotic mitochondrial fraction, respectively (Table I). The purification procedure did not generate a physical alteration of HCM and HAM fractions, because their re-loading onto the density gradient resulted in the recovery of a single population located at the 1.051/1.096 g/ml density interface (data not shown).
HCM, HAM, LAM1, and LAM2 corresponded to the entire mitochondria delimited by a double membrane envelope (Fig. 2, C-F). HCM and HAM exhibited conventional crest-forming inner membranes and a dense matrix (Figs. 2, C and D). These mitochondria did not differ in size because of similar mean diameter values (Table I). LAM1 were complex with different sizes, with different outer and inner membranes patterns, and with different matrix electronic densities (Fig. 2E). LAM2 corresponded to a swollen population (Fig. 2F). The mean diameter of LAM2 was 1.5 Ϯ 0.24 m (Table I). We observed a drastic reduction of the LAM2 matricial content (Fig. 2D) when compared with those of HCM and HAM (Fig. 2, C and D). Ruptures of the outer membrane were detected in the majority of LAM2 mitochondria (Fig. 2F).
A biochemical characterization of the different mitochondrial populations was realized (Fig. 3). In agreement with previous data (33), we have observed a weak but significant association of Bax with control mitochondria (Fig. 3A). Bax was also present in apoptotic mitochondria (Fig. 3A). Using the F1-ATPase ␤ subunit as an internal standard (Fig. 3A) (33), the amount of Bax was quantified (Fig. 3B). Our results show an enrichment of Bax protein in apoptotic mitochondria when compared with that observed in control mitochondria (Fig. 3B). Because Bax oligomerization has been shown to be involved in Bax-induced mitochondrial permeabilization (33)(34)(35), cross-linking experiments were realized on the different populations of mitochondria purified from control and apoptotic livers. Our results show that Bax was mainly present as a monomer in the HAM population (Fig. 3C), whereas most of Bax protein associated with LAM2 was recovered in a ϳ80-kDa oligomer (Fig. 3C). The oligomerization state of Bax associated with LAM1A was complex, we detected Bax in two oligomers of ϳ40 and ϳ80 kDa.
Bax insertion and oligomerization were paralleled with MOM permeabilization, because apoptotic mitochondria exhibited cytochrome c and Smac/Diablo contents lower than those found in control (Fig. 3, A and B). Interestingly, we observed a progressive reduction in the cytochrome c content of the different apoptotic populations. Indeed, cytochrome c in HAM, LAM1, and LAM2 corresponded to ϳ48, ϳ25, and ϳ10% of that found in the HCM population. Thus, our data established a direct correlation between the amount and the oligomerization state of mitochondrial Bax and the extent of mitochondrial permeabilization in our conditions. Taken as a whole, these data strongly indicate that HAM, LAM1, and LAM2 populations represent different chronological steps of Bax-related MOM permeabilization that was generated in vivo.
Electrophysiological Analysis of Mitochondrial Populations-We have investigated the electrophysiological properties of the different populations of apoptotic mitochondria and compared them to those of control mitochondria. Because the LAM1 population composition appeared heterogenous on electron micrographs, we focused our study on HCM, HAM, and LAM2. A voltage-dependent anionic channel (VDAC) was observed in 28.6, 29.9, and 28.1% of the cases for HCM, HAM, and LAM2, respectively (Table II). These values were not statistically different. VDAC biophysical properties are given in Fig. 4. VDACs of HCM, HAM, and LAM2 displayed similar longlasting transitions from open state to low conductance substrates (Fig. 4A). The number of subconducting states increased as the membrane became more polarized (Fig. 4A). This was observed at both positive and negative membrane potentials (Fig. 4A). From I/V m curves established in symmetrical (150/150 mM KCl) and asymmetrical (150/50 mM KCl) conditions (Fig. 4B), we calculated ␥ sym ϭ 702 Ϯ 18 pS (n ϭ 16) and ␥ asym ϭ 663 Ϯ 41 pS (n ϭ 28), ␥ sym ϭ 694 Ϯ 23 pS (n ϭ 18) and ␥ asym ϭ 638 Ϯ 27 pS (n ϭ 26), ␥ sym ϭ 678 Ϯ 35 pS (n ϭ 25) and ␥ asym ϭ 648 Ϯ 38 pS (n ϭ 29), for HCM, HAM, and LAM2, respectively. Furthermore, we determined P Cl /P K ϭ 1.97, P Cl /P K ϭ 2.02, and P Cl /P K ϭ 2.31 for HCM, HAM, and LAM2, respectively (Fig. 4B). P o /V m curves displayed typical bell-shaped relationships (Fig. 4C) and the parameters of the voltage dependence are given in Table III. VDAC gating charges were found to be similar for control and apoptotic mitochondrial populations in both positive and negative potential ranges. Calculation of V 0.5 show no differences between control and apoptotic populations (Table III), although we observed a ϩ10 mV right-shift for LAM2 VDAC in the positive potential range when compared with VDAC found in HCM and HAM populations. Finally, VDAC activation energy values were found to be similar when control and apoptotic mitochondria were considered (Table III).  Table III. Straight lines correspond to linear fits of the data (r 2 Ͼ 0.98).
A larger conductance (ϳ1 nS) channel could also be recorded in both control and apoptotic mitochondrial populations in the presence of 2 mM Ca 2ϩ . Current recordings exhibited transitions from closed to open states with a major substate located at ϳ50% of the full open state (Fig. 5A). Channel conductances were found to be 0.97 Ϯ 0.16 (n ϭ 3), 1.02 Ϯ 0.094 (n ϭ 3), and 0.94 Ϯ 0.142 nS (n ϭ 2) for HCM, HAM, and LAM2, respectively. Such a channel was inhibited by cyclosporin A (Fig. 5C). These properties are in agreement with those of the PTP channel (36,37).
A new ion channel was specifically observed in LAM2 (Table  II). In reference to that reported previously (26 -28), we considered it as a MAC for mitochondrial apoptosis-induced channel. MAC accounted for 17.5% of the channels recorded for LAM2 (Table II). MAC-related activity was never detected in HCM or HAM although the number of giga-ohm seals tested for HCM, HAM, and LAM2 was similar (Table II). MAC induced large currents at any of applied potentials (Fig. 6A). As membranes became more electrically polarized, the current flow through MAC rapidly decreased, although not to zero, indicating that channel conductance substates were favored as the membrane voltage increases. The current-voltage relationship was found to be linear between Ϫ60 and ϩ60 mV (Fig. 6B). We determined ␥ ϭ 3.05 Ϯ 0.65 nS (n ϭ 12) and ␥ ϭ 2.75 Ϯ 0.44 nS (n ϭ 18), in symmetrical (150/150 mM KCl) and in asymmetrical (150/50 mM KCl) conditions, respectively. Rarely did we observed larger conductance levels (␥ ϭ 5 nS, n ϭ 2, and ␥ ϭ 7 nS, n ϭ 1). From the E rev measurement (Fig. 6B), we determined a P Cl /P K ϭ 1.68 Ϯ 0.45, indicating that MAC poorly discriminated between K ϩ and Cl Ϫ . MAC was clearly voltagedependent with a bell-shaped P o -voltage relationship (Fig. 6C). Voltage-gating parameters are given in Table IV. Voltage dependence was almost symmetrical for positive and negative membrane potentials and the number of gating charges found to modulate the MAC opening was close to 6 (Table IV).
Pharmacological regulation of MAC was assayed (Fig. 7). The channel was neither sensitive to atractyloside (100 M, n ϭ 3), nor to cyclosporine A (1 M, n ϭ 5). Millimolar concentrations of Ca 2ϩ did not modify MAC electrophysiological properties (not shown). On the contrary, cytochrome c stabilized the channel (5 cases over 10 trials) in a very low conducting substate (Fig. 7), which reflects a block of MAC by cytochrome c.   Fig. 4D and considering that ln where z␦ corresponds to the number of gating charges and V 0.5 represents the applied potential required for half-maximal P o . Consequently, the activation energy corresponded to Ϫz␦FV 0.5 .
Gating charges Activation energy (kJ⅐mol Ϫ1 ) This block was fully reversible, although long lasting periods of washing were necessary for an entire recovery of MAC activity. Finally, we did not observe any modification in MAC activity in the presence of apocytochrome c (n ϭ 4) or lysozyme (n ϭ 5) (Fig. 7).

A Mitochondrial Heterogeneity Occurs After an in Vivo
Induction of Liver Apoptosis-The in vitro incubation of purified mouse liver mitochondria with the pro-apoptotic protein Bid promotes a time-dependent alteration of the mitochondrial ultrastructure, which results in the segregation of different mitochondrial morphological classes (38). Similar observations have been made in other in vitro models of apoptosis (reviewed in Ref. 39).
In the present study, we establish for the first time that an in vivo induction of rat liver apoptosis promotes a mitochondrial pleomorphism. Mitochondrial populations described here share similar properties with those described in vitro (38). HCM corresponds to class I mitochondria with intact inner and outer membranes, a dense matrix, and a high content in periplasmic proteins. HAM is related to early apoptotic mitochondria with a permeabilized MOM, whereas HAM morphology is identical to that of the HCM. LAM2 strongly resembles late apoptotic mitochondria, with swollen morphology, a highly permeabilized envelope, and an empty matrix. LAM1 is heterogenous, with normally sized and swollen mitochondria. LAM1 protein content is between HCM or HAM and LAM2. Thus, it is likely that LAM1 corresponds to an intermediate population of apoptotic mitochondria.
MAC Properties-In apoptotic cultured cells and Bax-expressing yeast, MAC defines a slightly cationic 2-nS conductance channel (26,27). We have used conditions for channel recordings identical to those described in these studies. Our data confirm MAC as an active mitochondrial channel. Furthermore, our study establishes that MAC observation is not restricted to in vitro models of cell apoptosis but is also present in mitochondria after induction of apoptosis in vivo.
MAC properties are unique and unrelated to those of other MOM resident channels. Indeed, the biophysical parameters calculated for MAC (this study), VDAC (this study and Refs. 40 and 41), and for the protein-conducting TOM channel (42)   Table III.
Ϫ29.5 ϩ28.5 Gating charges 5.9 5.7 Activation energy (kJ⅐mol Ϫ1 ) 16. 8 15.7 differ from one channel to another. PTP has been reported to participate in mitochondrial permeabilization through functional interactions of Bax with the adenine nucleotide translocator (9,12,13). Although, we have no direct evidence that PTP is not in part or totally responsible for MAC activity, we show that MAC is pharmacologically distinct from PTP, because modulators of PTP channel activity have no effect on MAC. In vitro Bax incubation with immunopurified VDAC induces a large conductance channel (i.e. 2.3 nS) (20), which is unlikely to be related to MAC because the Bax-VDAC channel is voltage independent and cytochrome c-insensitive, even in the presence of millimolar concentrations of cytochrome c (20). Moreover, Rostovtseva et al. (43) have characterized an absence of functional interaction between Bax and native VDAC. Finally, if MAC results from a Bax-VDAC interaction in our conditions, then a decrease of VDAC frequency of observation, between control and apoptotic mitochondria, is expected. Our study shows that this is not the case. MAC conductance is close to 3 nS. To our knowledge, this defines the largest channel so far identified in living animal cells. Assuming that the conduction domain of MAC corresponds to a 7-nm length cylinder, one can deduce a pore diameter of 4.6 nm. This value is consistent with the diffusion of cytochrome c, a roughly spherical protein whose diameter is close to 3.0 -3.5 nm (44). Mitochondrial peptide-sensitive channels and SEC 61 (two protein-conducting channels) display strong blocks of channel conductance when exposed to their respective substrates (i.e. mitochondrial or endoplasmic reticulum pre-proteins) (45)(46)(47), indicating that the transported protein plugs the channel. We show that a block of conductance is observed when MAC is superfused in the presence of cyto-chrome c. Thus, our data strongly suggest that cytochrome c inserts into the MAC channel. Finally, because apo-cytochrome c or lyzozyme, which exhibit isoelectric point close to that of cytochrome c, have no effect on MAC activity, we propose that MAC/cytochrome c interactions result from structural, rather than purely electrostatic, determinants. MAC is voltage-gated. This contradicts the initial MAC characterization (26) but is supported by a later report (27). The existence of an electrical potential across the outer membrane is controversial (48). However, the opening of one or several MAC during mitochondrial permeabilization will certainly dissipate such membrane potential, thus promoting the channel in its full open conformation. MAC is a nonselective channel. Ion recognition by an ionic channel is based on electrical charges segregation by a region delineating a selectivity filter, as deduced from channel atomic structure (49). The estimated diameter of MAC is consistent with the diffusion of cytochrome c, the diameter of which largely exceeds those of Cl Ϫ and K ϩ ions (50). Thus, our observation that MAC is not selective is not surprising.
MAC Activity Correlates with Bax Oligomerization-Using physiological conditions for an in vivo induction of apoptosis, we have been able to detect Bax oligomers in apoptotic mitochondria. To our knowledge, this establishes the first correlation between the modification of Bax quaternary structure and the extent of permeabilization of apoptotic mitochondria in vivo. Very high molecular weight Bax oligomers have been observed in previous in vitro studies (33,55). These are not detected in our conditions. This indicates that their association with mitochondria is lost during mitochondrial purification, or more likely that in vivo apoptosis triggers physiological amounts of Bax that are further addressed to the mitochon- drial membranes, thus preventing its coalescence into large aggregates.
In our study, Bax oligomers increase in size from 20 to 40 kDa and finally to 80 kDa in early, intermediate, and late apoptotic mitochondria, indicating that Bax oligomerization is time-related and is likely to occur after Bax targeting to mitochondria. The different sizes of Bax oligomers formed in vivo are in good agreement with those formed in acellular conditions (33). Several studies have demonstrated the strong permeabilization potency of oligomerized Bax to permeabilize lipid bilayer (43,(51)(52) as well as purified mitochondria (33,(53)(54)(55)(56). In our study, MAC activity is observed in apoptotic mitochondria that exhibit Bax oligomers but not Bax monomers. Similarly, MAC is not observed when Bax is present as monomer. Thus, MAC activity can be correlated with Bax oligomerization in apoptotic mitochondria.
MAC Is a Late Apoptotic Channel-The contribution of MAC in MOM permeabilization can be questioned. Quantitatively, the value of MAC conductance determined in our study is in good agreement with a pore diameter large enough to allow cytochrome c diffusion. We show that MAC formed in our in vivo conditions can interact specifically with cytochrome c. These data would be consistent with an involvement of MAC in MOM permeabilization during cell apoptosis. However, MAC activity is never seen in the early permeabilized mitochondrial population. Thus, we conclude that MAC cannot be responsible for the early permeabilization of apoptotic mitochondria, and that MAC corresponds to a channel that contributes lately in the apoptotic process. Based on our results and on those published elsewhere (26 -28), it is impossible for the moment to establish whether MAC results from a long time evolution of an initial permeabilizing structure, or whether MAC is the consequence of late feedback steps occurring at the mitochondrial level long after mitochondrial permeabilization. Further studies will be devoted to MAC structure/function relationships to understand MAC molecular origin and its contribution in the overall apoptotic process.