Ca2+-dependent Control of the Permeability Properties of the Mitochondrial Outer Membrane and Voltage-dependent Anion-selective Channel (VDAC)*

Cell function depends on the distribution of cytosolic and mitochondrial factors across the outer mitochondrial membrane (OMM). Passage of metabolites through the OMM has been attributed to the voltage-dependent anion-selective channel (VDAC), which can form a large conductance and permanently open a channel in lipid bilayers. However, recent data indicate that the transport of metabolites through the OMM is controlled in the cells. Recognizing that the bilayer studies had been commonly conducted at supraphysiological [Ca2+] and [K+], we determined the effect of Ca2+ on VDAC activity. In liposomes, the purified VDAC displays Ca2+-dependent control of the molecular cut-off size and shows Ca2+-regulated Ca2+ permeability in the physiological [Ca2+] range. In bilayer experiments, at submicromolar [Ca2+], the purified VDAC or isolated OMM does not show sustained large conductance but rather exhibits gating between a nonconducting state and various subconductance states. Ca2+ addition causes a reversible increase in the conductance and may evoke channel opening to full conductance. Furthermore, single cell imaging data indicate that Ca2+ may facilitate the cation and ATP transport across the OMM. Thus, the VDAC gating is dependent on the physiological concentrations of cations, allowing the OMM to control the passage of ions and some small molecules. The OMM barrier is likely to decrease during the calcium signal.

Mitochondria were recognized first as the main source of cellular ATP production, but over recent years their role has also been established in many aspects of cell physiology and pathophysiology (1). For example, to regulate cell survival, mitochondria retain proteins that induce apoptosis upon release to the cytosol (2)(3)(4)(5) and participate in calcium signaling by rapidly accumulating and releasing Ca 2ϩ (6 -9). Tight control of the transport of small molecules and ions across the inner mitochondrial membrane (IMM) 2 has been established, whereas, based on early studies of isolated mitochondria, the outer mitochondrial membrane (OMM) was perceived to form a barrier only for large molecules.
The voltage-dependent anion-selective channel (VDAC)/mitochondrial porin is the most abundant protein of the OMM. When reconsti-tuted in artificial membrane systems, such as liposomes and planar bilayer membranes (black lipid membrane (BLM)), the VDAC commonly appears as a large channel. At low potentials (Ͻ30 mV), the VDAC was found to be mostly fully open (ϳ4.5-nanosiemens conductance in 1 M KCl) and weakly anion-selective, whereas at high potentials, the VDAC showed a switch to cationic selectivity and closed to approximately half of the original conductance (classical "closed" states) (10 -13). Since the molecular cut-off of the VDAC is Ͼ3,000 Da in the fully open state, the common belief was that the OMM is freely permeable to small molecules and ions. However, during the past decade, several lines of evidence raised questions about the large size pore-forming activity of the VDAC in natural membranes and the consequent high permeability of the OMM (14). Patch clamp measurements of OMM fractions and intracellular patch clamping of native mitochondria did not show high conductance channel activity (15,16) unless ionic strength or [Ca 2ϩ ] was elevated to supraphysiological levels (17). Within a cell, the OMM may form a barrier against the transport of some small molecules (e.g. ADP (18,19)). Furthermore, in certain apoptosis models, the OMM permeability has been shown to be restricted for metabolites (20). Recently, selective permeabilization of the OMM by tBid, a proapoptotic protein (21), and by overexpression of VDAC (22) was reported to enhance the inositol 1,4,5-trisphosphate-induced mitochondrial matrix [Ca 2ϩ ] ([Ca 2ϩ ] m ) signal. These results suggested that the OMM Ca 2ϩ permeability could limit rapid Ca 2ϩ uptake to the mitochondria during physiological calcium signals. Furthermore, using targeted enhanced yellow fluorescence protein, lower pH was recorded in the mitochondrial intermembrane space (IMS) than that in the cytosol (23). A possible explanation for this phenomenon is that the VDAC did not permit rapid equilibration of the protons across the OMM.
To understand the reason for the apparent discrepancy between reconstituted and nonreconstituted data, we focused our interest on the experimental parameters of the reconstituted systems. We recognized that in the liposome and BLM experiments, typically supraphysiological [Ca 2ϩ ] were used (0.1-5 mM [Ca 2ϩ ] in BLM experiments and unbuffered or poorly buffered calcium in liposome experiments), presumably to increase membrane stability. Also, the bilayers were regularly bathed in a 1 M salt solution (e.g. KCl). To establish the VDAC behavior in the presence of physiological (submicromolar) and elevated [Ca 2ϩ ] as well, we set up both liposomal transport and BLM experiments with purified VDAC and OMM and fluorescence imaging of mitochondrial ion transport in single cells.

Isolation and Quality Control of VDAC and OMM from Rat Liver-
Mitochondria isolated from rat liver were used for purification of VDAC as previously described (24). For the isolation of OMM, mitochondria were further purified using self-performing Percoll gradient centrifugation (30% Percoll in 250 mM sucrose, 1 mM EGTA, and 2.5 mM Tris/HCl, pH 7.4). After a washing step, the pellet was resuspended and incubated in hyposmotic potassium phosphate buffer (swelling medium, 10 mM KH 2 PO 4 , pH 7.4) for 30 min on ice. Subsequently, a shrinking medium (32% sucrose, 30% glycerol, 10 mM MgCl 2 ) was added to the suspension (33% volume). After 30 min of shrinking, the mitochondrial membranes were disrupted by ultrasonication (4 ϫ 1 min of irradiation, 1-min break between each run). The resulted material was spun down at 12,000 ϫ g for 10 min. The supernatant contains a mixture of mitochondrial membrane vesicles and was used as a reference for the further purified membranes (marked as Mito in Fig. 1). To separate the OMM, contact point (CP), and IMM fractions, a discontinuous sucrose gradient was used (from bottom to top, 70, 45.6, 34.2, and 26% sucrose steps, 200,000 ϫ g for 240 min). The OMM was concentrated in the interface between the 26 and 34.2% steps, the CP was collected between 34.2 and 45.6%, and the IMM was collected between 45.6 and 70%. The protein concentration was determined in each fraction. To evaluate the purity of the isolated VDAC and mitochondrial membrane fractions, SDS-PAGE was performed, and the proteins were visualized with either silver staining or Western blotting (antibodies from the following sources: VDAC (Affinity Bioreagents); Bcl-x L (Transduction Laboratories), and Bax (Upstate Biotechnology, Inc.)). The membrane fractions were also tested for monoaminooxidase and succinate dehydrogenase enzyme activity, which were used as an OMM and an IMM marker, respectively.
Measurement of Dye Release and Ca 2ϩ Uptake by Plain and VDACreconstituted Liposomes-Soy lecithin (L-␣-phosphatidylcholine (Soy 20%) from Avanti Polar Lipids or 1,2-diacyl-sn-glycero-3-phosphocholine from Sigma) was washed with 10 mM EGTA to remove Ca 2ϩ contamination and was recrystallized. Liposomes were prepared by sonicating 50 mg/ml recrystallized lipid in a buffer containing 100 mM KCl, 20 mM Hepes/KOH, pH 7.2, 325 M EGTA, and 5 mg/ml dextran. After sonication, incorporation of purified VDAC and loading with fura2FA (50 M) or with 10-kDa fura-dextran (250 g/ml) was carried out by subjecting these vesicles to three freeze-thaw cycles. The liposomes were washed three times and were used promptly after the last washing step.
Dye release experiments were carried out essentially as described previously (25). Briefly, 30-l aliquots of the plain and VDAC-reconstituted liposomes were added to 720-l aliquots of 100 mM KCl, 20 mM Hepes/KOH, pH 7.2, and EGTA 325 M supplemented with various amounts of CaCl 2 to set different [Ca 2ϩ ]. After incubation for 0, 5, or 10 min at room temperature, the liposome suspensions were centrifuged at 35,000 ϫ g for 10 min at 4°C. Subsequently, the fura fluorescence released to the supernatant and the residual liposomal fraction in the solubilized pellet (Triton 0.3%) was measured (excitation, 340 nm; emission, 500 nm). The concentration of CaCl 2 was increased to 2 mM in every sample before the fluorescence measurement.
For the Ca 2ϩ transport measurements, the fura-dextran-loaded, VDAC-reconstituted, or plain liposomes were resuspended in a fluorometer cuvette in a buffer containing 100 mM KCl, 20 mM Hepes/KOH, pH 7.2, and 325 M EGTA at room temperature. To attain various [Ca 2ϩ ] levels, CaCl 2 (100, 200, 300, 400, or 500 M) was added. The time course of the fura fluorescence was monitored in a PTI DeltaRam fluorometer (excitation, 340 and 380 nm; emission, 500 nm). At the end of the measurement, to permeabilize every liposome for Ca 2ϩ , ionomycin (5 M), a Ca 2ϩ ionophore, was applied. Calibration of the fura ratio in terms of nM [Ca 2ϩ ] was added carried out as described previously (26).
For calculation of the initial rate of the VDAC-mediated Ca 2ϩ entry, first a curve was fitted to the time course of the intraliposomal [Ca 2ϩ ] in each condition. The nonreconstituted background displayed a quasilinear kinetics, whereas the VDAC-reconstituted samples gave an excellent fit with the combination of an exponential and a linear component ( y ϭ y 0 ϩ a ϫ (1 Ϫ e Ϫb(x Ϫ c) ) ϩ d ϫ (x Ϫ c)). The derivation of the exponential tag on x ϭ c gives the initial rate that is equal to a ϫ b.
Planar Lipid Bilayer Experiments-Purified asolectin (freshly recrystallized from stock purchased from Sigma or Avanti Polar) bilayers were prepared by painting a decane solution across a hole of a delrin cup (Warner Instruments). The experimental medium contained 0.1 M KCl or 1 M KCl plus 20 mM Hepes, pH 7.2. The preparation and storage of the medium was carefully controlled to minimize the risk of Ca 2ϩ contamination. Under these conditions, the [Ca 2ϩ ] was 500 -600 nM. In order to exclude Ca 2ϩ contamination, at the end of several experiments, the content of the cup was transferred to a fluorometer cuvette, and the [Ca 2ϩ ] was evaluated by fura2 (0.25 M) fluorescence measurement. The bilayer was painted, applying a drop of lipid to the hole by the means of a Teflon rod. After a stable membrane appeared (around 80 picofarads) and was recorded for 5 min, isolated VDAC or OMM vesicles were added to the cis partition. The additions of 0.5-6 l of solubilized VDAC protein (VDAC-prep) or OMM vesicles (OMM-prep) were made in the vicinity of the membrane. The content of the partitions was stirred by magnetic bars. Usually, the channel activity appeared within 10 min. If not, the addition of VDAC-prep was repeated. The total amount of added protein was 0.3-25 ng. In a few cases when the VDAC-prep was added only once, single channel activity appeared. The experiments were carried out at room temperature (22-24°C).
The cis and trans sides were coupled to electrodes via agar bridges. The electrodes were connected to CV203B headstage (Axon Instruments). For data acquisition, an Axopatch 200B (Axon Instruments) amplifier was used. The output signal of the amplifier was digitized by a Digidata 1200 digitizer. The headstage and chamber were placed in a small aluminum box (Faraday cage) that was closed during the measurements. The box was grounded to the headstage, as is recommended by the Axon Instruments. Noise level was monitored at the readout of the patch amplifier. The activity was monitored and recorded by means of an IBM-compatible PC. The amplifier was used in patch mode. A data point was obtained in every 60 s (the high rate was needed to detect the flickering bursts). Current recordings were filtered by a built-in low pass, 4-pole Bessel filter of the amplifier (2 or 5 kHz). In some cases, the signal was also digitally filtered (described in legend to Fig. 4). The signal was also collected essentially unfiltered for data analysis. When an ion channel appeared, first, it was recorded for 10 -30 s at various voltages, and then it was recorded for at least 5 min at Ϫ20 or ϩ20 mV. CaCl 2 (100 M) and EGTA/Tris (200 or 400 M, pH 7.4) were added to both sides of the membrane. After stirring, the recording was continued for at least 5 min. For the online recording and offline analysis of traces, the PCLAMP 9.0 program package (Axon Instruments) was used.
Imaging Experiments-RBL-2H3 mucosal mast cells were cultured and plated for experiments as previously described (27). Cells (4.5 ϫ 10 6 in 250 l) were transfected with plasmid DNA in suspension using electroporation (cytochrome c-GFP (cyto c-GFP) (28) and mitochondrial matrix-targeted DS Red (mito-DS Red); Clontech), keeping total DNA at 30 g, and then were plated to coverslips. Imaging was carried out 24 -36 h after transfection. For imaging of [Ca 2ϩ ] m , the cells were loaded with fura2FF/AM (5 M) for 50 -60 min at 37°C in extracellular medium composed of 121 mM NaCl, 5 mM NaHCO 3 , 10 mM Na-HEPES, 4.7 mM KCl, 1.2 mM KH 2 PO 4 , 1.2 mM MgSO 4 , 2 mM CaCl 2 , 10 mM glucose, and 2% (w/v) bovine serum albumin, pH 7.4, in the presence of 0.003% pluronic acid. After dye-loading, the cells were rinsed with Ca 2ϩ -free extracellular buffer containing 100 M EGTA. Permeabiliza-tion was carried out in 1 ml of ICM (120 mM KCl, 10 mM NaCl, 1 mM KH 2 PO 4 , 20 mM Tris-HEPES, and 1 g/ml each antipain, leupeptin, and pepstatin at pH 7.2) supplemented with saponin (30 -50 g/ml) and 20 M EGTA/Tris in the incubation chamber for 5 min (35°C). Subsequently, cells were put to fresh ICM containing 10 M EGTA/Tris. Confocal imaging of cyto c-GFP and mito-DS Red was conducted using a Bio-Rad Radiance system, whereas fluorescence imaging of pH IMS and [Ca 2ϩ ] m was carried out using an excitation wavelength switch (DG4; Sutter) and a high quantum efficiency cooled CCD camera (Pluto; Pixelvision) fitted to an Olympus IX70 microscope. At confocal imaging, cyto c-GFP and mito-DS Red were excited using the 488-and 568-nm lines of a krypton/argon ion laser source, respectively, recording 512 ϫ 512-pixel image pairs at 0.3 s Ϫ1 . At the fluorescence imaging experiments, cyto c-GFP was excited at 480 nm, and fura2FF was excited at 340 and 380 nm, and images or image pairs were recorded at 1 s Ϫ1 . Image analysis was performed using custom-made software (26). Fluorescence (averaged from pixels covering whole cell areas) was determined separately for each single cell (35-50 cells/field) on the field.
[ Experiments were carried out with 3-6 different membrane or cell preparations, and the data are shown as mean Ϯ S.E. Significance of differences from the relevant controls was calculated by Student's t test.

RESULTS
Characterization of Isolated VDAC and OMM-To study VDAC activity, both VDAC-prep and OMM-prep were isolated from rat liver mitochondria (Fig. 1). The procedure of the isolation of the OMM also yielded IMM and CP vesicles. Silver staining revealed a single band at ϳ32 kDa in the VDAC-prep, whereas among several bands, the dominant one was also the 32-kDa band in the OMM-prep (Fig. 1A, top). The enrichment of the 32-kDa band in the OMM-prep was apparent when compared with the mitochondrial homogenate (Mito), IMM, and CP fractions (Fig. 1A, top). Immunostaining with an anti-VDAC antibody showed intense reaction with the 32-kDa band in the VDAC-prep and also in the submitochondrial membrane fractions (Fig. 1A, middle). The OMM-prep displayed robust immunoreactivity relative to the mitochondrial homogenate, IMM, or CP fractions (Fig. 1A, middle).
Several Bcl-2 family proteins reside in the mitochondrial membranes to control the OMM permeabilization. Bcl-x L , an antiapopototic Bcl-2 family protein, was present in the homogenate but was not detectable in the VDAC-prep (Fig. 1A, bottom). By contrast, the OMM-prep was enriched in Bcl-x L (Fig. 1A, bottom). Similar to the Bcl-x L , Bcl-2 and Bax were also undetectable in the VDAC-prep and abundant in the OMMprep (data not shown). The activity of the monoaminooxidase (MAO), which is frequently used as an OMM marker enzyme, showed a 15-20fold enrichment in the OMM-prep (Fig. 1B, gray columns), whereas Note that Bcl-x L appeared in the mitochondrial homogenate and was relatively high in the OMM-prep but was not detectable in the VDAC-prep. B, the activity of the monoaminooxidase (MAO), an OMM marker enzyme, and succinate dehydrogenase (SDH), an IMM enzyme, in the submitochondrial fractions. Note the 15-20-fold enrichment of MAO and the hardly detectable level of SDH in the OMM-prep. C, VDAC-prep was reconstituted into asolecithin liposomes that were also loaded with either fura2 (Ͻ1 kDa) or fura-dextran (10 kDa). The localization of the dye inside the liposomes was evidenced by the fact that added Ca 2ϩ promptly increased the fura2 fluorescence ratio only in the presence of a Ca 2ϩ ionophore. Liposomes were incubated in the presence of 100 mM KCl and 100 M CaCl 2 for 10 min at room temperature, and the released dye was measured fluorometrically. The bar charts show that the VDAC-reconstituted liposomes became permeable for fura2, whereas the fura-dextran was not released. Data are representative for at least three separate experiments.
succinate dehydrogenase (SDH), an IMM enzyme, was present in the IMM fraction but was hardly detectable in the OMM-prep (Fig. 1B, open columns). Collectively, these data suggest that the only major protein in the VDAC-prep is VDAC and that the OMM-prep is highly enriched in VDAC. The OMM-prep has little contamination with IMM components but contains OMM proteins, including several Bcl-2 family proteins.
When the VDAC-prep was reconstituted into asolecithin liposomes that were incubated in the presence of 100 mM KCl and 100 M CaCl 2 , the liposomes became permeable for fura2, a 0.8-kDa fluorescent probe (Fig. 1C). However, larger molecules, such as fura2 conjugated to 10-kDa dextran (fura-dextran; Fig. 1C) or fluorescein isothiocyanate conjugated with cytochrome c (14 kDa) was not released (25). These results indicate that the VDAC-prep inserted into the liposome membrane and in the presence of 100 M CaCl 2 , formed pores that allowed for the transport of small molecules.
Ca 2ϩ Dependence of the Permeability of Liposomes Reconstituted with VDAC-To determine whether Ca 2ϩ exerts any effects on the VDACmediated transport of small molecules, the fura2 release from plain and VDAC-prep-reconstituted liposomes (VDAC-liposomes) was evaluated at varying Ca 2ϩ concentrations (Fig. 2). The liposomes were exposed to Ͻ20 nM, 2-3 M, and Ͼ1 mM [Ca 2ϩ ], respectively, and were incubated for 0 or 5 min at room temperature. Measurement of the liberated and the residual liposomal fluorescence provided information about the rate of fura2 release in the various conditions. No significant effect of Ca 2ϩ was observed on the release of fura2 from the plain liposomes (Fig. 2, left). At subphysiological [Ca 2ϩ ] (Ͻ20 nM), no VDACmediated fura2 release occurred, but a dose-dependent release was observed at the higher Ca 2ϩ concentrations tested (Fig. 2, right). Importantly, the VDAC-liposomes were extensively washed prior to exposure to Ca 2ϩ , eliminating the possibility that the effect of Ca 2ϩ was mediated by insertion of surplus VDAC. The low permeability of the VDACliposomes for fura2 indicates that at low [Ca 2ϩ ] and physiological [K ϩ ], the VDAC might form a barrier against small molecules. Furthermore, the permeation of small molecules through the VDAC seems to increase when [Ca 2ϩ ] is increased to the micromolar range.
To test the effect of [Ca 2ϩ ] on the Ca 2ϩ permeability of VDAC, liposomes were loaded with a fura-conjugated dextran (10 kDa), which did not exit the VDAC-liposomes even at 100 M [Ca 2ϩ ] (Fig.  1C). The Ca 2ϩ influx was monitored at varying extravesicular [Ca 2ϩ ] (Fig. 3). Ca 2ϩ influx was initiated by the addition of Ca 2ϩ . The kinetics of the Ca 2ϩ influx for plain and VDAC-reconstituted liposomes at two different concentrations is shown in Fig. 3, A and B. The addition of 200 and 500 M CaCl 2 increased the [Ca 2ϩ ] to 500 nM and 4 M, respectively. The fura excitation ratio showed a rapid and dose-dependent rise in luminal [Ca 2ϩ ] in the VDAC-reconstituted liposomes, indicating Ca 2ϩ permeation through the VDAC. To further characterize the [Ca 2ϩ ] dependence of the VDAC-mediated Ca 2ϩ entry, the initial rate was calculated at each extraluminal [Ca 2ϩ ] (for a description, see "Materials and Methods"). The luminal [Ca 2ϩ ] in the nonreconstituted liposomes displayed a quasilinear rise, whereas in the VDAC-liposomes, it was described by the sum of a linear and an exponential component. Thus, the exponential component seems to describe the VDACdependent Ca 2ϩ influx into the reconstituted population of liposomes (the yield of reconstitution was 20 -30%). In Fig. 3C (Fig. 4A), it showed the classical hallmarks of the VDAC, such as the characteristic size (single channel full conductance of about 450 pS in 100 mM KCl), symmetrical voltage dependence (practically permanent, full open state at lower voltages (ϩ20 mV, top) and closures on higher voltages (Ϫ50 mV, middle)), a plethora of subconductance states, and slow gating kinetic (bottom). Thus, these results have reproduced the core findings of previous BLM studies of isolated VDAC.
In the absence of CaCl 2 (in a nominally Ca 2ϩ -free buffer, [Ca 2ϩ ] ϳ600 nM), we recorded channels in 48 experiments. In 43 cases, only small conductances were found (Fig. 4B). In four cases, no channel activity appeared, but the addition of 100 M CaCl 2 induced gating. In one case, we observed channel activity similar to the activity found in the presence of CaCl 2 (not shown). However, it remains a possibility that the chamber was contaminated with CaCl 2 in this particular case. In Fig. 4B, a short multichannel recording of VDAC activity at ϳ600 nM [Ca 2ϩ ] is presented. At 20 mV, a nearly closed substate (ϳ30 pS) was the dominant conductance that was interrupted by short opening events. The amplitudes were typically smaller than the half-size of the full conductance (see histogram). Occasionally, amplitudes above the full conductance of a single channel (480 pS) also appeared, indicating the presence of more than one channel in the membrane (not shown). This activity is in sharp contrast to the permanent full conductance opening that was observed in the high calcium condition at 20 mV (Fig. 4A). It is of significance that the most frequent subconductance states observed in submicromolar [Ca 2ϩ ] medium are smaller in size than the "classical closed states" that were established at high voltage (Ͼ30 mV) in 1 M KCl and high [Ca 2ϩ ]. In Fig. 4C, a recording of a single channel is presented to illustrate the gating kinetics. In the 300-s period, seven series of events appeared. These events exhibited burstlike behavior as exemplified in the upper trace of the inset. Intraburst events are shown in the second trace of the inset, whereas the third and fourth traces exemplify the interburst activity. These data further support the notion that the gating kinetics of VDAC is slow.
Calcium Induces Increased VDAC Activity and Transition to Full Opening-The experiments shown in Fig. 4 have revealed the strong impact of steady 100 M (A) and submicromolar (B and C) [Ca 2ϩ ] on VDAC activity. We investigated whether the addition of calcium induces transition between the low conductance and the classical high conductance state. In 19 experiments, after the channel activity was recorded at ϳ600 nM [Ca 2ϩ ], the [Ca 2ϩ ] was raised to 50 or 100 M. In 16 experiments, the channel activity was clearly affected by the Ca 2ϩ addition (in four cases, gating appeared only after the Ca 2ϩ addition), whereas in three experiments, there was virtually no change. Two types of responses to Ca 2ϩ were seen: an immediate increase in activity and, after a lag period, an eventual transition to full opening. To illustrate both effects of Ca 2ϩ , in the upper row of Fig. 5A, the complete traces recorded at ϳ600 nM [Ca 2ϩ ] (i) and after elevation of [Ca 2ϩ ] (ii) in a multichannel experiment are shown. In the first 15 s of [Ca 2ϩ ] elevation, the activity was augmented, and at ϳ16 and ϳ115 s, the conductance was largely increased. The size of the increase at ϳ16 s is consistent with the opening of one VDAC to the full conductance state, and the increase at ϳ115 s indicates the opening of a second channel. At the top of the second opening, additional activity appeared, which demonstrated the presence of at least one more VDAC. To better visualize the initial effect of Ca 2ϩ on the channel activity, a short segment of the records obtained both at ϳ600 nM [Ca 2ϩ ] and 10 s of the [Ca 2ϩ ] elevation is also shown on an expanded time scale (iv and v). The individual events seem to be shorter and attain larger conductances after the [Ca 2ϩ ] rise. The average current was increased from 0.86 to 1.51 pA in the initial period of the Ca 2ϩ effect and to 10.62 pA after the opening of the first channel to full conductance (Fig. 5A, vii). All-point amplitude histograms revealed a Ca 2ϩ -induced shift in the amplitude spectrum from low to higher values (iii and vi). The histogram of the whole trace displays two additional peaks corresponding to the first and second full openings (ϳ10 and ϳ20 pA, respectively; iii). Amplitude analysis discriminated five different subconductance levels in the control trace and two more after Ca 2ϩ addition. The open probability of the first level (0.63 pA, ϳ31 pS) decreased from 0.85 to 0.22, and the second (1.11 pA, ϳ55 pS) and third (1.82 pA, ϳ91 pS) increased from 0.12 to 0.37 and from 0.01 to 0.31, respectively. The two Ca 2ϩ -dependent conductances were 4.42 and 9.47 pA (ϳ220 and ϳ470 pS, respectively). The latter corresponds to the full conductance. However, the related open probabilities are almost negligible (0.01 and 0.02). Collectively, these results indicate that the channel became faster and individual events became shorter (v), and the average current increase is due to the more frequent occupancy of higher subconductance levels. However, the more frequent gating (and sporadic visits to full conductance) reflects instability of the channel that is later stabilized as classical VDAC with a full open state at low voltage. Along the line of the models that consider the VDAC as the OMM component of the permeability transition pore (PTP), this tran- sition may reflect the VDAC opening when the VDAC participates in the Ca 2ϩ -induced permeability transition.
In other multichannel bilayers, different subconductances were involved, and the lag time of the transition to full conductance varied from experiment to experiment. In two cases, when Ca 2ϩ caused an increase in channel activity but without full opening up to ϳ6 min, we added EGTA to lower the [Ca 2ϩ ] back into the submicromolar range. One example is shown in Fig. 5B. Ideal histograms of the plot were obtained from the event list of about ϳ6 min long in the control (black), Ca 2ϩ -exposed (red), and EGTA-buffered periods ( green). The histograms resolved six subconductances. The Ca 2ϩ -induced frequency increase is reflected in the elevation of the number of the events of the third, fourth, and fifth subconductance levels (red). Restoration of the submicromolar [Ca 2ϩ ] by EGTA led to a return to the pre-Ca 2ϩ subconductance pattern (green). This result demonstrates that the Ca 2ϩinduced activation of the VDAC was fully reversible.
Taken together, the electrophysiology data complement and extend the results of the liposome experiments. The liposome experiments showed that at submicromolar [Ca 2ϩ ], the VDAC exerts a restricted permeability to Ca 2ϩ and a relatively small molecular cut-off size. These results were unexpected, based on previous observations of a large pore size of the VDAC in high [Ca 2ϩ ] environment. The BLM experiments revealed, however, that under low [Ca 2ϩ ] conditions, the VDAC gates almost exclusively with small subconductances. In addition, the Ca 2ϩinduced reversible increase in channel activity, more specifically the higher conductance levels, seems to provide a mechanism for the Ca 2ϩdependent permeability increase in the VDAC-liposomes. The next question is whether the small conductance gating and the calcium sensitivity of the isolated VDAC are physiologically relevant phenomena. As an approach to this question, we investigated the ion channel activity in the natural outer mitochondrial membrane (OMM-prep).
Calcium Controls VDAC Activity in Natural Mitochondrial Outer Membrane-First, noncontact OMMs were fused into BLM membranes and were exposed to 1 M KCl and 100 M CaCl 2 . The records showed "classical VDAC"-like activity, symmetrical voltage dependence (practically permanent open state at lower voltages and closures at higher voltages), numerous subconductance states, and slow gating kinetics. In Fig. 6A, the upper graph illustrates the voltage-dependent closure in a multichannel experiment (five channels, single channel full conductance of about 4.5 nanosiemens in 1 M KCl). The lower graph with an expanded scale shows several substates. Both traces display slow gating kinetics.
In 100 mM KCl and submicromolar [Ca 2ϩ ], there were always various small conductance gatings found (30 experiments). The activity appeared in burstlike fashion, which is exemplified in Fig. 6B (see sample on extended time scale). It is also shown in Fig. 6B that the Konig's polyanion, an inhib- itor of the VDAC, practically eliminated the OMM channel activity, although some rare events remained, verifying the presence of the channel (10 of 10 experiments). The OMM-prep is highly enriched in VDAC (Fig. 1), and upon fusion to BLM, it provided classical VDAC activity in 1 M KCl and appeared as a polyanion-sensitive channel, suggesting that the OMM-derived channel represents VDAC activity in the natural membrane. Along this line, the small conductance gating in 100 mM KCl seems to apply to the VDAC in the natural membrane at the physiological ion concentrations.
In 17 experiments, CaCl 2 (100 M) was added to BLM reconstituted with OMM in the presence of 100 mM KCl and submicromolar [Ca 2ϩ ]. In 12 cases of the 14 experiments that could be evaluated, the channel amplitude spectrum and gating frequency were affected by Ca 2ϩ (Fig.  6C). Thus, a Ca 2ϩ -dependent shift to larger subconductance states was shown for both the VDAC-prep and the OMM-prep. Interestingly, Ca 2ϩ induced permanent opening to full conductance of the OMMderived channels only in 1 M KCl (Fig. 6A), whereas the VDAC-prep yielded full conductance in the presence of either 1 M or 100 mM KCl (Fig. 4A). Whereas the VDAC-prep comprises only VDAC, the OMM-prep contains numerous proteins that may interact with the VDAC to control the channel activity (Fig. 1A). The high ionic strength may be required for the dissociation of some of these modulators from the VDAC. In two experiments, no Ca 2ϩ effect was observed. Collectively, these studies indicate that the VDAC activity is controlled by Ca 2ϩ in the natural OMM environment.
[Ca 2ϩ ] c Controls the H ϩ and Ca 2ϩ Permeability of the OMM in Cells-To validate the biophysical data in a cellular paradigm, we sought for an approach that allows detection of materials entering mitochondria through the VDAC directly in the IMS. EGFP and its variants have been used for pH measurements (29), and EGFP has been targeted to the IMS as a cytochrome c fusion protein (28). We set out to generate a pH gradient between cytosol and IMS and to record the protons that traverse the OMM in the IMS using cyto c-GFP (Fig. 7A, i).
Confocal images show the distribution of both cyto c-GFP (green) and mito-DS Red (red) in a permeabilized RBL-2H3 cell in Fig. 7A, ii. Cyto c-GFP appeared in the structures positive for the red mitochondrial marker. In response to tBid, a proapoptotic Bcl-2 family protein that selectively permeabilizes the OMM, complete loss of cyto c-GFP occurred, whereas the mito-DS Red fluorescence remained unchanged (Fig 7A, ii, graph and images on the right). Thus, cyto c-GFP is confined to the IMS in the transfected RBL-2H3 cells.
To avoid any changes in cyto c-GFP fluorescence that might be evoked by H ϩ extrusion through the IMM, the permeabilized cells were not provided with any substrates, and utilization of the endogenous substrates was prevented by the addition of rotenone (10 M), antimycin A (5 g/ml), and oligomycin (5 g/ml). Thapsigargin (2 M), an inhibitor of the SERCA pump, was also added to prevent Ca 2ϩ handling by the endoplasmic reticulum. To minimize the photodamage, the time course of the cyto c-GFP fluorescence was recorded at relatively low spatial resolution using a high quantum efficiency CCD camera (Fig. 7A,  iii). When the pH in the cytosol was lowered from 7.2 to 6.0 by the addition of 8 millinormal HCl, a rapid and large decrease in the GFP fluorescence occurred, indicating that the protons traversing the OMM can be detected by cyto c-GFP in the IMS (Fig 7A, iii, middle, shown in  red). The cytosolic acidification-induced decrease in GFP fluorescence was effectively reversed after restoring the cytosolic pH by the addition of KOH, illustrating the reversibility of the pH-dependent change in fluorescence (Fig. 7A, iii, right, shown in blue). In the next set of measurements (Fig. 7A, iv), the permeabilized cells were preexposed to either Ͻ20 nM or 4 M [Ca 2ϩ ] c , and the time course of the HCl-induced fluorescence change was recorded. Ca 2ϩ by itself did not affect the cyto c-GFP fluorescence (not shown, n ϭ 11) and did not affect the magnitude of the HCl-induced decrease in fluorescence (Fig. 7A, iv, 120 s, n ϭ   11). However, as the initial rates of the F GFP decrease show in Fig. 7A, iv, the kinetics of the HCl-induced fluorescence decrease was considerably faster in the presence of elevated [Ca 2ϩ ] c (10 s, red versus black column). This result indicates that the proton permeation across the OMM was facilitated by Ca 2ϩ .
To explore whether ATP permeation across the OMM was also enhanced by Ca 2ϩ in the above described model, the ATP-driven Ca 2ϩ accumulation was monitored by imaging [Ca 2ϩ ] m (Fig 7B, i). Specifically, RBL-2H3 cells were loaded with fura2FF using a protocol that results in compartmentalization of the dye in the mitochondrial matrix (26,27). The fura2FF-loaded cells were permeabilized in the buffer used in the cyto c-GFP imaging experiments, except that oligomycin was omitted. Under this condition, no driving force for the mitochondrial Ca 2ϩ uptake remained. Consistently, Ca 2ϩ ([Ca 2ϩ ] c ϭ 4 M) failed to induce any considerable [Ca 2ϩ ] m elevation (Fig. 7B, ii, red). However, the addition of ATP to the Ca 2ϩ -pretreated cells evoked a rapid and large [Ca 2ϩ ] m increase (Fig. 7B, ii, red). By contrast, when Ca 2ϩ was added at the same time as ATP, only a slow [Ca 2ϩ ] m rise appeared ( Fig  7B, ii, black trace and bar charts). Thus, elevation of [Ca 2ϩ ] c before ATP addition led to an increase in the rate of the ATP-driven mitochondrial Ca 2ϩ accumulation. VDAC inhibitors could not be used to confirm the involvement of the VDAC in this process, since DIDS and the Konig's polyanion caused loss of the ⌬⌿ m and, at higher concentration, cytochrome c release (25). However, earlier studies have established that the VDAC serves as the primary transport mechanism for both nucleotides  and Ca 2ϩ across the OMM (30,31). Although the studied process involved the transport of ATP and Ca 2ϩ across both the OMM and IMM, the function of the adenine nucleotide translocase is affected by [Ca 2ϩ ] c only in the millimolar range (32), and the uniporter also shows a relatively modest sensitization at the tested [Ca 2ϩ ] c (33). Considering these points, the massive increase in the rate of [Ca 2ϩ ] m rise indicates the possible enhancement of the VDAC-mediated passage of ATP and/or Ca 2ϩ across the OMM by Ca 2ϩ . This result complements the observation that Ca 2ϩ stimulates the proton transport across the OMM. Furthermore, both the Ca 2ϩ -dependent stimulation of the proton and the ATP/Ca 2ϩ entry serve as evidence that the observations made in the liposome and bilayer experiments are relevant for mitochondrial membrane transport in the cell.

DISCUSSION
The present study revealed [Ca 2ϩ ] as a regulator of VDAC activity, providing a novel mechanism for the control of OMM permeability to ions and small molecules. We have shown that 1) in the 100 nM range of [Ca 2ϩ ], the VDAC permits only a slow flux of fura2 (0.8 kDa) and Ca 2ϩ in liposomes and 2) under similar ionic conditions, in BLM measurements, the VDAC displays small conductance substates, which are below the classical "closed" states. Furthermore, we showed that elevation of [Ca 2ϩ ] to the micromolar range 1) largely enhanced the flux of fura2 and Ca 2ϩ by liposomes and 2) induced higher subconductances and sustained opening to full conductance of the VDAC in the BLM. In addition, OMM fused to lipid bilayers displayed VDAC-like channel activity and sensitivity to a VDAC inhibitor. The OMM channel activity was increased when Ca 2ϩ was added. Finally, an effect of Ca 2ϩ on the mitochondrial transport of cations and ATP was exposed in a cellular paradigm. Thus, the VDAC seems to account for the channel activity in the OMM and displays regulation by Ca 2ϩ both in the highly purified form and in the natural outer membranes. Regulation of VDAC activity by Ca 2ϩ may exert an effect on several aspects of the bioenergetics.
Our results provide a clue to the apparent discrepancy between the classical electrophysiology work that established the large conductance and almost continuous opening of the VDAC at low voltages (10,13,34) and some recent studies of the OMM in the intracellular milieu, which showed the presence of small conductances and restricted permeability to ions or metabolites (17,18,21,22). The classical electrophysiology on isolated VDAC was carried out in the presence of high CaCl 2 (100 M-1 mM) or unbuffered [Ca 2ϩ ], whereas in the studies of VDAC in the intracellular milieu, the channel was exposed to ϳ100 nM cytoplasmic [Ca 2ϩ ]. In light of our data on the Ca 2ϩ sensitivity of the VDAC activity, the difference in [Ca 2ϩ ] can be a source of the above differences in VDAC activity. However, in addition to the different Ca 2ϩ levels, some other factors are also important for the divergent properties of the VDAC in previous works. In the present study, permanent large conductance opening of the isolated VDAC could be triggered by high [Ca 2ϩ ] in 100 mM KCl, whereas a similar response in the noncontact OMM could be evoked only in the presence of 1 M KCl. The OMM-prep contains VDAC and several other proteins that interact with VDAC. Both a soluble mitochondrial protein and hexokinase-1 have been shown to decrease the VDAC conductance (35,36), whereas Bax was reported to increase the pore size (37,38). Thus, the high salt concentration could be required to disrupt a protein-protein interaction that modulates the VDAC activity in the OMM. Importantly, we used noncontact OMM, and the protein-protein interactions of the VDAC are different in the noncontact and contact site regions of the OMM. As to the VDAC at the contact sites, evidence has been presented that the VDAC is a component of the PTP complex (39), and the sustained full conductance opening of the VDAC has been linked to the opening of the PTP (40,41).
Based on the above data, we predict that at the resting level of cytoplasmic [Ca 2ϩ ], the VDAC activity and the OMM permeability for Ca 2ϩ and small molecules is regulated. The role of VDAC in the transport of substrates through the OMM was underscored by the inhibition of mitochondrial metabolic function in the presence of agents that induce the classical "closed" states (35,42). Evidence has also been provided for Ca 2ϩ permeation through the VDAC (31). However, in the cell, the OMM Ca 2ϩ permeability seems to limit the calcium signal propagation to the mitochondria (21), and this restriction can be alleviated by overexpression of the VDAC (22). Importantly, mitochondrial Ca 2ϩ uptake is often driven by [Ca 2ϩ ] spikes and oscillations that consist of short lasting events and require instantaneous activation of the Ca 2ϩ transport. The relatively slow release of fura2 from VDAC liposomes at low Ca 2ϩ also indicates that small molecules, like substrates, do not necessarily freely traverse the OMM. Notably, restriction of the ADP diffusion through the OMM has been documented (18). However, the transport of substrates is driven by relatively sustained gradients, and the organization of metabolism into complexes may also support channeling of metabolites. One might also speculate that a slow release of protons through the OMM may result in more effective chemiosmosis (43). The Ca 2ϩ -induced enhancement of the OMM permeability supports rapid Ca 2ϩ delivery to the uniporter, the channel that mediates Ca 2ϩ uptake via the IMM (44). Through this mechanism, Ca 2ϩ -dependent activation of the VDAC facilitates mitochondrial Ca 2ϩ uptake and Ca 2ϩ buffering and may also synergize with the Ca 2ϩ -dependent activation of the mitochondrial dehydrogenases and other enzymes to increase ATP production.
In the present work, the VDAC activity was studied in the Ͻ20 nM to 100 M range of [Ca 2ϩ ]. Clearly, the physiological range of bulk cytoplasmic [Ca 2ϩ ] is confined to the lower end of this [Ca 2ϩ ] span (from 50 nM to 1-2 M). Thus, it is of significance that we documented Ca 2ϩ -induced changes in VDAC permeability for Ca 2ϩ and fura2 in this narrower range. Furthermore, it has been demonstrated that a fraction of the mitochondria is localized in the vicinity of endoplasmic reticulum Ca 2ϩ release channels and may be exposed to substantially higher local [Ca 2ϩ ] than the bulk [Ca 2ϩ ] elevation (27,45,46). Strikingly, elementary Ca 2ϩ release events, known as Ca 2ϩ sparks, have been found to suffice to activate the low affinity Ca 2ϩ uptake in adjacent mitochondria (47). Based on the activation of mitochondrial Ca 2ϩ uptake, we have calculated that the OMM may be exposed to at least 10 -20 M [Ca 2ϩ ] during the inositol 1,4,5trisphosphate receptor or ryanodine receptor-linked calcium signal (27). Thus, at least for the mitochondria that are localized close to the sarcoendoplasmic reticulum, the physiological [Ca 2ϩ ] range extends to the full range used in the present work.
The VDAC belongs to the family of ␤-barrel membrane proteins that are mostly confined to the OMM in mammalian cells (48 -50). Some hallmarks of the VDAC activity have been firmly established, including the symmetrical voltage dependence, the voltage dependent closure, the presence of numerous substates, slow gating kinetics, and the modulation of activity by the polyanion and DIDS. However, the electrophysiologic analysis of the VDAC has been hindered by the technical challenge posed by the mitochondrial patch clamp recording and also by the slow kinetics of the channel. For many ion channels, 1-s recording provides a sufficient number of events for statistical analysis, whereas the VDAC requires 10 -30-min recordings to produce a similar number of events. In the present work, we documented VDAC activity in the entire range, from the completely closed to the fully open state, at physiological Ca 2ϩ and K ϩ levels. As to the identification of the VDAC, we have reproduced the above described hallmarks of the VDAC for this channel activity and also reproduced the sustained full conductance opening and closure to half-conductance level, when the VDAC-prep was exposed to the conventional recording conditions (high [Ca 2ϩ ] and [K ϩ ]). Thus, under physiological ionic conditions, the VDAC can adopt conformations that appear as a completely closed state or subconductances smaller than the classical "closed" states.
Data collection for relatively long periods allowed us to visualize bursts in VDAC activity, which seem to last for tens of seconds, further illustrating the slow gating of this channel, and to analyze some properties of the events at both submicromolar and 100 M Ca 2ϩ levels. Our data suggest that Ca 2ϩ serves as an inducer of the large conductances. The Ca 2ϩ sensitivity of the VDAC could be conferred by either a Ca 2ϩsensitive enzyme that targets the VDAC or by calmodulin that may bind both Ca 2ϩ and the VDAC or by a direct effect of Ca 2ϩ on the VDAC. Since the effect of Ca 2ϩ on the VDAC was observed in a highly purified VDAC-prep and because Ca 2ϩ binding to purified and solubilized VDAC has also been shown (31), a Ca 2ϩ binding site on the VDAC may be primarily considered as the mediator of the Ca 2ϩ effect. Although no EF-hand seems to be present in the sequence of the mitochondrial VDAC, further sequence analysis seems to be warranted, since a motif search with the Gibbs sampler algorithm indicates the presence of conserved sequences that show considerable homology with parts of the EF-hand (not shown). Similar to the natural composition of the OMM, in the present study, cardiolipin was present in the liposomes and bilayers, which may bind Ca 2ϩ and also confer Ca 2ϩ sensitivity to the VDAC. Importantly, the Ca 2ϩ -induced shift to large conductances appeared promptly, but the permanent opening to full conductance was typically delayed. A possible mechanism is the dimerization of the VDAC (40,51). The Ca 2ϩ -dependent patterns described here may be modulated in the cell by the generation of electrical potential across the OMM (52); by the pH (53); by the interaction of VDAC with other ions and small molecules, including NADH (54), purine nucleotides (55)(56)(57), and superoxide (25); and with a range of proteins that have been listed above. Along the line of a commonly proposed model, the VDAC interacts with the adenine nucleotide translocase and cyclophilin D to form the PTP, which is primarily regulated by [Ca 2ϩ ] m . This regulation was envisioned to depend on Ca 2ϩ sensing by cyclophilin D, but recent data show that the Ca 2ϩ sensitivity of the permeability transition is not lost in cyclophilin D knock-out mice (58). Thus, Ca 2ϩ sensing by the adenine nucleotide translocase (32) or by the VDAC (although the VDAC is not exposed to [Ca 2ϩ ] m ) may also be of significance for the permeability transition.
The main point of the present work is that the VDAC may display the full range of potential activities in the cell, depending on the Ca 2ϩ concentration. Extension of the previously described permanent large conductance activities to the range of small conductances has major implications for mitochondrial membrane transport. Gating of VDAC by Ca 2ϩ may present a novel mechanism for the calcium signal to control both mitochondrial and cellular functions.