Genetic Demonstration That the Plasma Membrane Maxianion Channel and Voltage-dependent Anion Channels Are Unrelated Proteins*

The maxianion channel is widely expressed in many cell types, where it fulfills a general physiological function as an ATP-conductive gate for cell-to-cell purinergic signaling. Establishing the molecular identity of this channel is crucial to understanding the mechanisms of regulated ATP release. A mitochondrial porin (voltage-dependent anion channel (VDAC)) located in the plasma membrane has long been considered as the molecule underlying the maxianion channel activity, based upon similarities in the biophysical properties of these two channels and the purported presence of VDAC protein in the plasma membrane. We have deleted each of the three genes encoding the VDAC isoforms individually and collectively and demonstrate that maxianion channel (∼400 picosiemens) activity in VDAC-deficient mouse fibroblasts is unaltered. The channel activity is similar in VDAC1/VDAC3-double-deficient cells and in double-deficient cells with the VDAC2 protein depleted by RNA interference. VDAC deletion slightly down-regulated, but never abolished, the swelling-induced ATP release. The lack of correlation between VDAC protein expression and maxianion channel activity strongly argues against the long held hypothesis of plasmalemmal VDAC being the maxianion channel.

Purinergic signaling is a widespread phenomenon of general biological significance, and mechanisms accounting for ATP release from cells remain a contentious issue (1). We have recently identified a maxianion channel as a nanoscopic pore (2) well suited to function as the ATPreleasing pathway (3,4). This pore accounts for the swelling-induced ATP release from mouse mammary C127 cells (5,6) and NaCl-dependent ATP-mediated signaling from macula densa to mesangial cells during tubuloglomerular feedback in the kidney (7). In addition, the same pore is operational in swelling-, ischemia-, and hypoxia-induced ATP release from neonatal rat cardiomyocytes (8).
The maxianion channel has been observed in a wide variety of cell types and exhibits roughly uniform behavior (9), suggesting that it has a general physiological function. Although the molecular identity of the maxianion channel is not yet firmly established, it is widely held that a voltage-dependent anion channel (VDAC) 2 located in the plasmale-mma that normally functions in the mitochondrial outer membrane (10 -12) is the most likely candidate protein. This hypothesis was based on the similarity of shared biophysical properties, such as the large unitary conductance and bell-shaped voltage dependence of the maxianion channel and mitochondrial VDAC (13)(14)(15)(16)(17)(18). Corroborating this idea, numerous groups have reported the presence of VDAC protein in the plasma membrane of various cell types (13)(14)(15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26). A possible mechanism for targeting of the same protein to different locations has been suggested by Buettner et al. (14). They reported the existence of an alternative first exon in the murine vdac-1 gene that encodes a different leader peptide at the N terminus, a signal that purportedly targets the protein to the plasma membrane through the Golgi apparatus. Under this scenario, the signal peptide is eventually cleaved away to produce a plasmalemmal VDAC (pl-VDAC) protein identical to the mitochondrial protein. Other mechanisms involving alternative mRNA untranslated regions have also been hypothesized to explain the extramitochondrial localization of porin (27).
The "maxianion channel ϭ pl-VDAC" hypothesis is widely accepted (e.g. see Ref. 17). We thoroughly tested this hypothesis by gene deletion and gene silencing. In mammals, three isoforms of mitochondrial porin, VDAC1, VDAC2, and VDAC3, have been cloned (13, 28 -35). The deletion of vdac genes was shown to result in various physiological dysfunctions (36). These include male infertility (37), disrupted fear conditioning and spatial learning (38), enhanced apoptosis (39), and growth retardation and increased fatigue (36). If the maxianion channel is a pl-VDAC, then deletion and/or silencing of the VDAC genes would be expected to eliminate the channel activity and abolish the maxianion channel-mediated ATP release. Here we present the results of such a study.
vdac2 gene silencing was performed by RNA interference in dKO fibroblasts. The following oligonucleotides, targeting three different regions of the mouse vdac2 gene, were used to generate the hairpinencoding inserts for the three RNAi constructs: 5Ј-TTTTCTCGAGC-CGCCTCGGCTGTGATGTTGACTTCAAGAGA-3Ј and 5Ј-CCTT-GGTACCTTCCAAAAACCTCGGCTGTGATGTTGACTCTCTT-GAA-3Ј (RNAi-1 construct); 5Ј-TTTTCTCGAGCCGCTTTGCAGT-CGGCTACAGGTTCAAGAGA-3Ј and 5Ј-CCTTGGTACCTTCCA-AAAACTTTGCAGTCGGCTACAGGTCTCTTGAA-3Ј (RNAi-2 construct); 5Ј-TTTTCTCGAGCCGTGGGACAGAATTTGGAGGA-TTCAAGAGA-3Ј, and 5Ј-CCTTGGTACCTTCCAAAAATGGGAC-AGAATTTGGAGGATCTCTTGAA-3Ј (RNAi-3 construct). The underlined sequences are 19-nucleotide target sequences corresponding to the regions of murine vdac2. The oligonucleotides were annealed, extended with the Klenow fragment of DNA polymerase, and cloned into a pBluescript II KSϩ vector (Stratagene, La Jolla, CA) containing the mouse RNAase P H1 promoter and a puromycin resistance gene. Fugene 6 (Roche Applied Science)-mediated transfection of fibroblasts was performed with these three RNAi constructs as well as with the control construct containing only the RNAase P H1 promoter and puromycin resistance gene. Clones stably expressing hairpin siRNAs were selected with 5 g/ml puromycin (Sigma). The effectiveness of VDAC2 down-regulation by each RNAi construct was studied by Western blotting with specific anti-VDAC2 antibodies.
For patch clamp experiments, the cells were grown on glass coverslips and were transferred to the chamber immediately before experiments.
Solutions-The normal Ringer solution for patch clamp contained 135 mM NaCl, 5 mM KCl, 2 mM CaCl 2 , 1 mM MgCl 2 , 5 mM Na-HEPES, 6 mM HEPES, 5 mM glucose (pH 7.4; 290 mosmol/kg H 2 O). For selectivity measurements, 135 mM NaCl in Ringer solution was replaced with 135 mM of tetraethylammonium or sodium glutamate. The pipette solution for inside-out experiments was either normal Ringer solution or solution containing 100 mM KCl, 2 mM MgCl 2 , and 5 mM HEPES (pH 7.4, adjusted with KOH). The pipette solution for outside-out experiments contained 120 mM NaCl, 5 mM KCl, 2 mM CaCl 2 , 1 mM MgCl 2 , 5 mM Na-HEPES, 6 mM HEPES, 10 mM EGTA (pH 7.4, adjusted with NaOH, pCa 7.7, 280 mosmol/kg H 2 O). For K/Cl permeability measurements, the bath solution contained 1000 mM KCl, 2 mM MgCl 2 , and 5 mM HEPES (pH 7.4, adjusted with KOH). The bath solution for bromide and acetate permeability measurements contained 1000 mM KBr or potassium acetate, 2 mM MgCl 2 , and 5 mM HEPES (pH 7.4, adjusted with KOH or acetic acid, respectively). For ATP permeability measurements, 100 mM Na 2 ATP solution was used after the pH was adjusted to 7.4 with NaOH. The hypotonic solution for ATP release measurements contained 92 mM NaCl, 5 mM KCl, 2 mM CaCl 2 , 1 mM MgCl 2 , 5 mM Na-HEPES, 6 mM HEPES, 5 mM glucose (pH 7.4, 210 mosmol/kg H 2 O, 72% hypotonicity). GdCl 3 was stored as a 50 mM stock solution in water and added directly to Ringer solution immediately before each experiment. 5-Ni-tro-2-(phenylpropylamino)-benzoate, glibenclamide, 4-acetamino-4Јisothiocyanostilbene, arachidonic acid, indomethacin, nordihydroguaiaretic acid, and clotrimazole were purchased from Sigma and were added to Ringer solution immediately before use from stock solutions in Me 2 SO. Me 2 SO did not have any effect when added alone at a concentration of less than 0.1%.
The osmolality of all solutions was measured using a freezing point depression osmometer (OM802; Vogel).
Electrophysiology-Patch electrodes were fabricated from borosilicate glass capillaries using a laser micropipette puller (P-2000; Sutter Instrument, Novato, CA) and had a tip resistance of around 1.5-2 megaohms when filled with pipette solution. Membrane currents were measured with an EPC-9 patch clamp system (Heka-Electronics, Lambrecht/Pfalz, Germany). The membrane potential was controlled by shifting the pipette potential (Vp). The membrane potential is reported as ϪVp. Currents were filtered at 1 kHz and sampled at 5-10 kHz. Data acquisition and analysis were done using PulseϩPulseFit (Heka-Electronics). Whenever the bath Cl Ϫ concentration was altered, a salt bridge containing 3 M KCl in 2% agarose was used to minimize variations of the bath electrode potential. The liquid junction potentials were measured according to Ref. 42 and were corrected when necessary. All experiments were performed at room temperature (23-25°C).
ATP Release and Total ATP Assay-Bulk extracellular ATP concentration was measured by the luciferin-luciferase assay (ATP Luminescence Kit; AF-2L1, DKK-TOA Co., Tokyo, Japan) with an ATP analyzer (AF-100, DKK-TOA), as described previously (5, 6) with modifications. Briefly, the cells were cultured to confluence in 12-well plates for time course experiments and in 24-well plates for end point steady-state ATP release measurements. Culture medium was totally replaced with isotonic Ringer solution (1000 l for 12-well and 425 l for 24-well plates, respectively). Cells were incubated for 60 min at 37°C, and 100 l of the extracellular Ringer solution was removed and used as a control sample for background ATP release. A hypotonic challenge was then applied by gently removing most of the remaining extracellular solution (875 l for 12-well and 300 l for 24-well plates, respectively) and adding the hypotonic solution (1000 l for 12-well and 400 l for 24-well plates, respectively). The solution was gently mixed by careful rocking, and the plate was placed into the incubator at 37°C. At the specified time points, the plate was carefully rocked again to ensure homogeneity of the extracellular solution, and samples (20 l for 12-well and 50 l for 24-well plate, respectively) were collected from each well for the luminometric ATP assay. ATP concentration was measured by mixing 20 -50 l of sample solution with 500 l of normal Ringer solution and 50 l of luciferinluciferase reagent. At this ratio, ionic salt sensitivity of the luciferase reaction (43) was negligible. When required, drugs were added to the hypotonic solution to give the final concentrations as indicated. Since Gd 3ϩ has been reported to interfere with the luciferase reaction (43), we supplemented the luciferin/luciferase assay mixture with 600 M EDTA when the samples contained Gd 3ϩ . Other drugs employed in the present study had no significant effect on the luciferin-luciferase reaction.
Data Analysis-Single channel amplitudes were measured by manually placing a cursor at the open and closed channel levels. The reversal potentials were calculated by fitting I-V curves to a second-order polynomial (2) or measured directly from ramp currents. Permeability ratios were calculated using the Goldman-Hodgkin-Katz equation as previously described (5). The value of P K /P Cl was calculated from the KCl gradient experiments, and this was then used for Br Ϫ /Cl Ϫ and acetate/ Cl Ϫ permeability ratio calculations. Data were analyzed in Origin 5-7 (OriginLab Corp., Northampton, MA). Pooled data are given as means Ϯ S.E. of n observations. Statistical differences of the data were evaluated by analysis of variance and the paired or unpaired Student's t test where appropriate and considered significant at p Ͻ 0.05. Statistical differences in slopes of linear fits were evaluated by analysis of covariance using StatsDirect software (StatsDirect Ltd., Cheshire, UK) and considered significant at p Ͻ 0.05.

Single Maxianion Channels in Wild-type Mouse
Fibroblasts-In the cell-attached (on-cell) configuration, no single channel events were observed from wild-type mouse embryonic fibroblasts (WT-MEFs). However, after excision of the patch membrane, single channel events with a large amplitude could be readily recorded (Fig. 1A). The channels exhibited voltage-dependent inactivation at both positive and negative potentials greater than Ϯ20 mV. The unitary current-to-voltage (I-V) relationship was linear in the range of Ϯ50 mV, with a reversal potential of about 0 mV in symmetrical conditions with Ringer solution in the bath and pipette and a slope conductance of 400.3 Ϯ 3.8 pS (Fig. 1B). The single channel amplitude was insensitive to the replacement of NaCl with tetraethylammonium-Cl. However, the inward currents greatly decreased, and the reversal potential shifted to a negative value of Ϫ33.7 Ϯ 1.6 mV when NaCl in the bath solution was replaced with sodium glutamate (Fig. 1B). This result indicates that the large conductance channel in WT-MEFs is anion-selective with a permeability ratio of glutamate/Cl Ϫ of 0.23 Ϯ 0.02. This value is close to P glutamate /P Cl ϭ 0.22 obtained for the maxianion channel in mammary C127 cells (4).
In order to test the ATP permeability of the maxianion channels in WT-MEFs, we recorded the I-V curves in response to ramp pulse from Ϫ50 to ϩ50 mV. The ramp I-V relationship was linear in symmetric ionic conditions with Ringer solution both in the pipette and in the bath, with reversal potential at 0 mV. When all anions in the bath were replaced with 100 mM ATP 4Ϫ , large outward currents (carried by Cl Ϫ from the pipette solution) as well as small inward currents (presumably carried by ATP 4Ϫ from the bath solution (see Ref. 5) were consistently observed. The average reversal potential shifted to Ϫ18.2 Ϯ 1.2 mV (n ϭ 15 from five different patches), giving P ATP /P Cl ϭ 0.083 Ϯ 0.005.
The maxianion channel in WT-MEFs was insensitive to Gd 3ϩ ions (50 M) added from the intracellular side of excised patches (n ϭ 5; data not shown) but completely blocked (n ϭ 6) when the drug was applied from the extracellular side in the outside-out configuration (Fig. 1D). Another well known modulator of the maxianion channel, arachidonic acid (20 M), potently inhibited the channel activity when added from the cytosolic side (n ϭ 7; Fig. 1D).
The biophysical properties of the WT-MEF maxianion channel, such as the single channel conductance, voltage-dependent inactivation, ATP permeability, and sensitivity to Gd 3ϩ and arachidonate, are very similar to those observed previously in C127 cells (5,6), kidney macula densa cells (7), and cardiomyocytes (8).
In a separate set of experiments, we determined the ionic selectivity of WT-MEF maxianion channels in conditions close to those commonly used for mitochondrial VDAC proteins reconstituted in lipid bilayers. When 10-fold KCl gradient (100 mM in the pipette versus 1000 mM in the bath) was applied to the maxianion channel-containing patches, the reversal potential was 40.6 Ϯ 1.5 mV (Fig. 1B). This value is much greater than the value of 10.2 mV reported for the mitochondrial porin under the same conditions (45). The calculated permeability ratio P Cl /P K ϭ 13.5 Ϯ 2.3 also largely exceeded the P Cl /P K of 1.7-1.9 for mitochondrial porins from different sources (45,46). Replacing 1000 mM KCl in the bath with equimolar KBr or potassium acetate resulted in the reversal potentials of 41.2 Ϯ 1.5 and 28.6 Ϯ 0.5 mV, respectively (Fig.  1B). The calculated permeability ratio P Br /P Cl ϭ 1.01 Ϯ 0.06 was close to that for mitochondrial VDAC (1.02) as derived from the partitioned ionic conductances in Ref. 47. However, P acetate /P Cl of 0.58 Ϯ 0.01 was greater than P acetate /P Cl ϭ 0.41 for mitochondrial VDAC (47) (14), was based on the proposed existence of an alternative first exon in the murine vdac-1 gene. The hypothetical alternative exon encodes a leader peptide at the N terminus that targets the protein to the plasma membrane through the Golgi apparatus. Therefore, deleting the vdac gene would interrupt this process and consequently abolish plasmalemmal VDAC expression and maxianion channel activity. We have generated MEFs lacking the vdac1 gene and assayed the expression of VDAC1 protein using specific antibodies. Western blotting demonstrated a complete absence of VDAC1 protein in these cells ( Fig. 2A, a). However, when the same cells were assayed electrophysiologically, they exhibited maxianion channel activity indistinguishable from WT-MEFs, including activation by patch excision, voltage-dependent gating, linear current-voltage relationship, and unaltered single channel conductance and K ϩ to Cl Ϫ selectivity (Fig. 2B). The channel occurrence in vdac1 Ϫ/Ϫ fibroblasts was only slightly less (ϳ60% of channel incidence) than in wild-type cells (ϳ70%).
The very existence of maxianion channels in vdac1 Ϫ/Ϫ cells decisively refutes the "maxianion channel ϭ pl-VDAC" hypothesis as formulated for the vdac1 gene. However, the hypothesis can still be modified by replacing VDAC1 protein with one of the two other isoforms (VDAC2 and VDAC3) and hypothetically presuming the existence of an alternative leader peptide. To test this possibility, we first generated the cells lacking the vdac3 gene and double-deficient cells lacking both the vdac1 and vdac3 genes. The vdac3 Ϫ/Ϫ cells do not express VDAC3 protein, and VDAC1/VDAC3 dKO cells lack both VDAC1 and VDAC3 proteins, as demonstrated by Western blots in Fig. 2A (a and b). However, both vdac3 Ϫ/Ϫ and dKO cells displayed maxianion channels with normal biophysical properties (Fig. 2, C and D) and with maxianion channel occurrence (ϳ50% for vdac3 Ϫ/Ϫ and ϳ70% for dKO cells) similar to WT-MEF and vdac1 Ϫ/Ϫ cells. The only possible remaining candidate for pl-VDAC protein as a maxianion channel is the VDAC2 isoform. We therefore tested the vdac2 Ϫ/Ϫ cells (39). These cells lack the VDAC2 protein (see Western blots in Fig. 3A). However, when we tested the maxianion channel activity, it was found to be normal (Fig. 3B) with approximately the same occurrence in about 70% of membrane patches.
Thus, our results strongly suggest that none of the three individual isoforms of VDAC protein can be responsible for the maxianion channel activity in mouse fibroblasts. However, theoretically, it is still possible that the different isoforms replace each other when one (e.g. VDAC1), which normally functions as a wild-type maxianion channel, is deleted. VDAC1/VDAC2/VDAC3 triple knockout cells would give a definitive answer to this question. However, it was not possible to produce such mutant cells due to embryonic lethality associated with the absence of VDAC2 (39). Therefore, we employed a gene silencing strategy using RNA interference to reduce the expression of VDAC2 protein in dKO cells lacking VDAC1 and VDAC3. We generated three different constructs designated RNAi-1, RNAi-2, and RNAi-3 (see "Experimental Procedures" for details) and obtained three different clones with substantially depleted VDAC2 expression, as determined by Western blotting (Fig. 3A). When these cells were tested by patch clamp, the maxianion channel activity was normal in all three clones of VDAC2 depleted cells, and the number of maxianion channels per patch was not statistically different from each other or from WT-MEF or wild-type adult fibroblast cells (Fig. 3C). Therefore, we conclude that maxianion channel activity is unrelated to the expression of VDAC1, VDAC2, and VDAC3 proteins. It is also highly unlikely that VDAC2 (the only possible candidate for maxianion channels in VDAC1/VDAC3 double knockout cells) can replace VDAC1 (maxianion channel in the original form of the hypothesis) or VDAC3 (an alternative candidate for maxianion channel in vdac1 Ϫ/Ϫ cells) in VDAC-deficient cells.
Effect of Deletion of VDAC Genes on Swelling-induced ATP Release-Our previous studies strongly suggest that the maxianion channel serves as a pathway for ATP release induced by osmotic cell swelling and other stimuli from several different types of cells (3). Therefore, we assayed the release of ATP induced by 72% hypotonicity from VDAC-deficient cells. WT-MEFs responded to this stimulation with a time-dependent release of ATP into the extracellular milieu (Fig.  4A). This ATP release was inhibited by anion channel blockers 4-acetamino-4Ј-isothiocyanostilbene and 5-nitro-2-(phenylpropylamino)-benzoate, by the maxianion channel blocker Gd 3ϩ , but not by glibenclamide, an inhibitor of cAMP-activated cystic fibrosis transmembrane conductance regulator and volume-sensitive outwardly rectifying (VSOR) Cl Ϫ channels (Fig. 4B). These latter channels are also considered as candidates for the ATP release pathway (3,4). Arachidonic acid had a significant inhibitory effect on ATP release. This inhibition was enhanced in the presence of a mixture of inhibitors, which suppresses the consumption of free arachidonate by cytosolic oxygenases (6). The total ATP content of WT-MEF cells was not significantly altered by arachidonic acid at p Ͻ 0.05 (n ϭ 12), ruling out the possibility that the observed inhibition of ATP release was due to modulation of the mitochondrial permeability transition (48). The pattern of ATP release from WT-MEFs was very similar to that observed in our earlier studies with mammary C127 cells (5,6) and neonatal cardiac myocytes (8), where maxianion channels play a key role in the release of ATP.
Fibroblasts deleted for vdac1, vdac2, and vdac3 genes displayed slightly reduced mass release of ATP compared with the wild-type cells (Fig. 4C). However, the observed difference did not reach statistical significance when analyzed by the analysis of variance test. Thus, none of these genes alone can be entirely responsible for the osmotic ATP release from these cells. When VDAC1/VDAC3-double-deficient cells were tested with and without VDAC2 protein depletion by RNA interference, the swelling-induced ATP release levels were not statistically different within this group (dKO, dKO-RNAi-1, dKO-RNAi-2, and dKO-RNAi-3), suggesting that the decreased VDAC2 protein expression in dKO fibroblasts (Fig. 3A) did not result in the comparable decrease in ATP release.
It should be noted, however, that the ATP release from dKO cells themselves was significantly lower than that from WT-MEFs when compared by unpaired t test (Fig. 4D). This result is probably due to impaired ATP production by mitochondria lacking an important ATP transport pathway in their mitochondrial outer membrane (36). To verify this possibility, we measured the total ATP content in the cell lysates. The total ATP content of dKO cells was only 78.5 Ϯ 4.5% (n ϭ 9) of that for wild-type adult fibroblast cells (significant at p Ͻ 0.05). Thus, we believe that reduced ATP release from dKO cells and its clones derived by RNAi is mainly due to the decreased ATP gradient caused by reduced mitochondrial ATP production and not due to reduced activity of an ATP-releasing maxianion channel, which was normal in these cells (Fig. 3C).

DISCUSSION
VDAC, or mitochondrial porin, resides in the outer membrane of mitochondria and when reconstituted in lipid bilayers displays large single channel conductance and bell-shaped voltage-dependent inactivation with maximal open probability at around 0 mV (10 -12). These biophysical properties are similar to those observed for maxianion channels in patch clamp experiments (3,4,9). Probing the mitochondrial porin by the nonelectrolyte partitioning method yielded a value for the pore radius of 1.5 nm for the fully open state of the channel in lipid bilayers (49). This figure is very close to the cut-off size of around 1.3 nm obtained in our experiments with the maxianion channel (2). Using the asymmetric PEG application method, Carneiro et al. (50,51) described an asymmetrical pore for mitochondrial VDAC with a radius of ϳ1 and ϳ2 nm for its cis-and trans-entrances, respectively (cis designates the side of the bilayer into which the protein was added). This asymmetry parallels the asymmetry of the maxianion channel demonstrated in experiments using a one-sided application of PEG: 1.16 and 1.42 nm for radii of intracellular and extracellular vestibules (2). Electron microscopic images have demonstrated that mitochondrial porin has an inner radius of ϳ1.4 nm (52), which is very close to the value obtained by polymer size exclusion for VDAC in lipid bilayers (44 -46) and for the maxianion channel in our patch-clamp study (2). Thus, the structural features of mitochondrial porin and the ATP-conductive maxianion channel do converge. Maxianion channels also resemble (5) mitochondrial porins with respect to their open channel block by ATP (53) and ATP conductivity (53)(54)(55).
From the present study, we conclude that the maxianion channel activity in mouse fibroblasts does not correlate with the presence of any one of the three individual isoforms of VDAC, as established in single gene knockout experiments. Moreover, simultaneous deletion of vdac1 and vdac3 genes and vdac2 gene silencing in VDAC1/VDAC3-doubledeficient cells did not notably affect the maxianion channel activity, unequivocally establishing that the maxianion channel protein is not encoded by any of the vdac genes. This result contradicts the long held hypothesis that the maxianion channel represents a plasmalemma VDAC protein (13)(14)(15)(16)(17)(18). Our data, however, do not exclude the possibility that VDAC protein(s) could be, in fact, targeted to the plasma membrane (e.g. by the mechanism proposed by Buettner et al. (14) or via mRNA untranslated regions, as speculated by Bathori et al. (27)). If VDAC retargeting does occur, the plasmalemmal VDAC proteins may perform some other functions, such as being a receptor for plasminogen kringle 5 (56) or a trans-plasma membrane NADH-ferricyanide reductase (57), activities that are unrelated to the maxianion channel activity. The mass release of ATP could be modulated (17) indirectly by VDAC protein expression, possibly due to an altered rate of ATP production by mitochondria, or by affecting other ATP-releasing pathways. Another conclusion from these studies is that mitochondrial outer membrane permeability can be effectively ablated, yet the cell remains viable. This may reflect the limited dependence of fibroblasts on oxidative metabolism, as underscored by the ability to create rho 0 cells by mitochondrial DNA depletion (58).
It should be noted that, although both proteins transport ATP, the similarities in single channel properties between VDAC and the maxianion channel are rather superficial, and closer inspection reveals crucial differences. Specifically, the single channel conductance of the maxianion channel saturates at 617 pS (K d ϭ 77 mM) and 640 pS (K d ϭ 112 mM) with increases in the chloride concentration in skeletal muscle "sarcoballs" (59) and L6 myoblasts (60), respectively. In contrast, the VDAC single channel conductance may reach levels of over 10 nS at high salt concentrations without any saturation (61), suggesting a principally different mechanism of ionic transport in these two pores. The same inference can be drawn by comparing the K ϩ to Cl Ϫ selectivity of these two channels. Under the same 10-fold KCl gradient, the maxianion channel generated a reversal potential of about 40 mV, whereas no more than ϳ10 mV was observed for mitochondrial VDAC (45), indicating that the maxianion channel is much more selective for chloride over potassium. Although the overall ranking of anionic permeability was similar for both channels (Br Ϫ Ͼ Ͼ Cl Ϫ Ͼ acetate), the numeric value of the permeability ratio was notably different for acetate. The permeability ratio for glutamate to Cl Ϫ of 0.23 for the WT-MEF maxianion channels was also different from the value of P glutamate /P Cl ϭ 0.4 reported for mitochondrial porin (62).
Voltage-dependent gating is considered to be a common property for the two channels. Mitochondrial VDAC is known to retain ϳ40% of its initial conductance in the so-called "closed" state, which is cation-selective (10,12,46). Like many other channels, the maxianion channel also displays subconductance states. However, normally it closes completely at high positive and negative voltages (e.g. see Fig. 1C). Voltage-dependent modulation of ionic selectivity has never been reported for maxianion channels, supporting our conclusion that the maxianion channel and mitochondrial VDAC are unrelated proteins. The "maxianion channel ϭ pl-VDAC" hypothesis has been stimulating. We believe, however, that a fresh start to molecular identification of the maxianion channel, an important gateway of purinergic signaling, should be undertaken. Particular attention to other candidates, such as connexin hemichannels or orthologs of the tweety gene found in the flightless locus of Drosophila (63) is warranted.