Review Article
Gating Model
Auto-Directed Insertion
Ultra-Steep Voltage Dependence



VDAC, a Channel in the Outer Mitochondrial Membrane

Marco Colombini*, Elizabeth Blachly-Dyson#, and Michael Forte#

* Department of Zoology, University of Maryland, College Park, MD 20742

# Vollum Institute, Oregon Health Sciences University, Portland, OR 97201
Address correspondence to Marco Colombini
TEL: 301-405-6925
FAX: 301-314-9358

Colombini, M., Blachly-Dyson, E. and Forte, M.  1996.  VDAC, a channel in the outer mitochondrial membrane.  In: "Ion Channels" Vol. 4 (Narahashi, T. ed.) pp 169-202. Plenum Publishing Corp., New York, NY.



1. Background

2. Fundamental Channel Properties

2.1. Two Gating Processes

2.2. Asymmetric Structure Yields Symmetrical Behavior

2.3. Selectivity of Open and Closed States

2.4. Biophysical Basis for Selectivity

3. Structure of VDAC Channels

3.1. Genes Encoding VDAC

3.2. Conservation of Primary and Secondary Structure

3.3. Site-directed Mutagenesis of Charged Residues in Yeast VDAC

4. Molecular Analysis of the Gating Process

4.1. Magnitude of the Structural Change Associated with Voltage Gating

4.2. Location of the Mobile Domain

4.3. Identification and Localization of the Voltage Sensor

4.4. Activation Energy and the Gating Mechanism

4.5. Ion Flow and the Gating Process

5. Comparison of Proposed Gating Mechanisms

5.1. Corking the Bottle

5.2. Forming a Jelly Roll

5.3. Tilting the Strands

5.4. Deforming the Cylinder

6. Regulation

6.1. Amplification of the Voltage Dependence

6.2. Action of the VDAC Modulator

6.3. Action of Nucleotides

6.4. Colloidal Osmotic Pressure

6.5. Inhibition by Metal Hydroxides

7. Auto-Directed Insertion

8. Role of VDAC in Cell Function

9. Conclusions

10. Acknowledgments

11. References


Proteins that form aqueous channels in membranes generate conduction pathways with a variety of shapes and sizes. Perhaps the largest channel-forming protein is the 2MDa ryanodine receptor while he smallest may be gramicidin. However, the size of the conducting pathway is uncorrelated with the amount of protein mass needed to make up the structure as demonstrated by the fact that some of the narrowest conducting pathways are produced by very large amounts of protein (e.g. 0.3 MDa for the Na+/K+/Ca++ channel family). By contrast, the focus of this review, the voltage dependent anion channel (VDAC) of the mitochondrial outer membrane, produces one of the largest aqueous pathways from a single 30 kDa protein. VDAC also demonstrates that functional complexity does not seem to correlate well with amount of protein used to form a channel. VDAC has a small amount of protein mass but displays complex behavior. It has two voltage-gating processes, can be controlled by metabolites and regulatory proteins, is able to form complexes with other proteins/enzymes, and responds to the protein concentration of the cytoplasm (Colombini, 1994). Thus, many functions are packed into a single, relatively small VDAC protein.

The molecular structure of channels is also quite variable. The only structures known with atomic resolution are gramicindin and some member of the bacterial porin family. The former forms a variety of structures but the membrane channel is a b helix (for discussion see Durkin et al., 1990) and the latter b barrels (Weiss et al., 1990). Other channel-forming proteins seem to form the transmembrane aqueous pathway in different ways: using a helices (the cystic fibrosis chloride channel, Akabas et al., 1994b), a combination of a helices and extended regions (the nicotinic acetylcholine receptor, Akabas et al., 1994a), extended regions only (K+ channels, Yellen et al., 1991), and b -barrel/a -helix combinations (VDAC). This variety of structures is likely to increase as more detailed structural information is obtained about additional channels.

A goal of studying membrane channels is to interrelate electrical properties of the channel, the molecular rearrangements that underlie these properties, and the in vivo physiological function of individual channel types. In the case of VDAC, as with other channels, more is known about the properties of the channel than about its physiological role. Most investigators agree however that VDAC is the main pathway by which metabolites cross the outer mitochondrial membrane. These metabolites can reach a molecular weight of almost 1000 and therefore the permeability pathway must be large. If VDAC functions to control the outer membrane's permeability for these molecules and ions, VDAC channels may work quite differently than channels whose function is to control the permeability of small alkali-metal ions. The goal of this review is to outline our current understanding of VDAC: the properties of the channel as assayed following reconstitution into planar phospholipid membranes, the molecules that generate these channels, the nature of the conformational transitions underling basic channel properties, and the regulatory interactions that either influence VDAC's function and/or that may represent important functional interactions in vivo.


All VDAC channels described to date have a remarkably conserved set of biophysical properties. Whether isolated from yeast or humans, VDAC proteins form channels with a roughly similar single channel conductances ( 4 nS in 1 M KCl), open channel selectivity (2:1 preference for the conduction of Cl- over K+ for a gradient of 1M vs 0.1 M KCl) and voltage dependent conductance (Colombini, 1989). In addition, each channel closes to lower conducting or "closed" states when either positive or negative potentials are applied. Thus, although the details listed below are generated primarily from study of VDAC isolated from fungi like yeast or Neurospora crassa, the same general properties are observed for VDAC isolated from essentially all eukaryotes.

2.1. Two Gating Processes

When reconstituted into planar phospholipid membranes, VDAC channels are open most of the time at low voltages (» 10 mV) although they undergo rare transitions to lower conducting states (termed closed states). Thus, when channels insert into the membrane from the aqueous phase at low potentials, one typically observes current increments forming a staircase effect (Fig. 1). There is open-channel noise which manifests itself more and more as the number of reconstituted channels increases. Typically, channels insert as single discrete conducting units, defined as a single channel, into such membranes. It is unlikely that the single channel is actually the insertion of two channels side-by-side because when a single channel closes, the drop in conductance is usually more than half the original conductance. Smaller insertion events are interpreted as channels inserting in one of the closed states. Channels inserting with low conductance have been observed to open and then behave like normal fully open channels. When channels do not behave like the majority behave, it is usually assumed that they are damaged, modified, or improperly folded.

When the membrane potential is increased, typically above 30 mV, VDAC channels undergo transitions to lower conducting ("closed") states (Schein et al., 1976; Colombini, 1989). Although both the open and closed states are permeable to simple salts (the former more permeable than the latter), their permeability to mitochondrial metabolites (mostly organic anions) is dramatically different (see Table I). Thus, from a physiological perspective, the terms open and closed are quite appropriate. Closed channels will reopen but, in VDAC from most species, they tend to remain closed until the voltage is reduced. In addition, although there is a single open state, there does not appear to be a single closed state but a variety of possible closed states, some more stable than others. For example, immediately after closing, channels reopen more readily than after they have been closed for a longer time suggesting that structural rearrangements take place in order to achieve more stable closed conformations. These rearrangements are sometimes visible as transitions between states shortly after the membrane potential is elevated. Indeed, while the voltage is kept constant, the channel often returns to the highest conducting state (the open state) before closing again to a closed state with a different conductance (Fig. 2). In addition, there are electrically silent changes that can be detected only by the fact that the rate of reopening of the channels is reduced with increasing time in the closed state. In contrast to other channel types, channel closure is observed at both positive and negative potentials (Schein et al., 1976). This is true not only in multi-channel membranes but also in single-channel membranes (Fig. 3) demonstrating that a single channel forms one conductance pathway which responds to both positive and negative potentials.

2.2. Asymmetric Structure Yields Symmetrical Behavior

The ability of individual channels to behave in a symmetrical manner seemed best explained by a symmetrical structure. Initial indications supported the conclusion that VDAC was a homodimer of 30 kDa subunits. The detergent-solubilized VDAC from rat liver seemed to be a dimer (Linden and Gellerfors, 1983) and there is a linear dependence of the number of channels reconstituted into a planar membrane with the amount of detergent-solubilized protein added to the aqueous phase (Roos et al., 1982). However more exacting experiments indicate that a single VDAC channel is formed by a single VDAC protein.

The ability of VDAC channels from N. crassa to make 2-dimensional crystals was exploited to estimate the mass of one channel. Fig. 4 shows the surface topography of such a crystal (upper panel) and a view of its internal structure after freezing in vitreous ice (lower panel). Mannella estimated that there was not enough mass per pore in these crystals for each channel to be composed of a dimer of 30 kDa proteins (Mannella, 1986, 1987). Consistent with this conclusion, Thomas and coworkers (1991) measured the mass of an area of the crystalline array (using tobacco mosaic virus as a standard) by Scanning Transmission Electron Microscopy. Since the number of channels per unit area was known, it was possible to estimate that each channel in an array had an effective mass of 44 kDa. Since 1 VDAC polypeptide is 30 kDa, one channel can only contain 1 polypeptide. The extra mass is likely to be due to the association of phospholipids and/or sterols with the VDAC protein. Consistent with this idea, there is evidence that sterols are part of the basic structure of VDAC (Pfaller et al., 1985).

Taking a different approach, Peng and coworkers (1992b) tried to produce VDAC channels composed of two different VDAC polypeptides by expressing wild type genes and genes containing site-directed mutations that result in channels with altered selectivity, together in yeast. If VDAC channels are dimers, one would expect that three kinds of channels would be present in these cells: 1) wild-type dimers, 2) mutant dimers, and 3) mixed dimers with an intermediate selectivity. When individual channels from such cells were examined, channels with wild-type selectivity and channels with mutant selectivity were observed with almost equal frequency. No channels were found with intermediate selectivity. If dimers had formed either as two adjacent channels, two channels in tandem, or two semi-cylindrical halves forming one large pore, channels with intermediate selectivity would have been detected. It is possible that the failure to observe "hybrid"channels reflects some unknown process related to the translation of VDAC transcripts and the assembly and targeting of newly formed VDAC proteins to mitochondria which prevents "mixing" of different VDAC proteins, although this is unlikely. Statistically, enough channels were examined to reduce the probability of having missed hybrid channels by random chance to 1 in 107.

Thus, these findings agree with the electron microscopy experiments in the conclusion that one channel is composed of only one 30 kDa polypeptide chain. Since the primary structure of VDAC proteins does not contain palindromic sequences nor any significant structural repeats, it is very likely that the molecule is fundamentally asymmetric. In fact, the inherent structural asymmetry of the channel can be readily observed as functional asymmetry when channels are reconstituted in unusual lipids such as diphytanoylphosphatidylcholine (Ming Yao Liu, unpublished observations) or when point mutations are engineered at particular sites (Zizi et al., 1995). Thus, while the presence of two symmetric gating processes in planar phospholipid membranes has been conserved throughout evolution, it is possible to establish experimental conditions in which gating becomes more asymmetric.

2.3. Selectivity of Open and Closed States

The open channels differ from closed channels not only in conductance but in selectivity. The open state prefers anions, Cl- over K+ by a factor of 5 (Table) at 0.1 M salt concentrations (measurements made at higher salt concentrations and steeper salt gradients yield lower values, cf. Colombini, 1989). The closed states generally favor cations (Colombini, 1980b; Benz et al., 1990) although the selectivities vary depending on the closed state that is achieved (Zhang and Colombini, 1990). When organic anions are examined, channel closure results in large drops in the selectivity for these ions (Table). The selectivities for succinate and citrate drop dramatically and thus VDAC becomes a poor conduit for these anions. Higher voltages or the addition of agents that induce VDAC closure (see section 6) result in channels entering states of much lower conductance. These lower conducting states have not been well studied but there is evidence that the selectivity for cations is even greater in these states (Peng and Colombini, unpublished). This large change in selectivity for small ions and organic compounds is consistent with the idea that the transition from open to closed channel is associated with a large change in the overall structure of the protein (see section 4).

2.4. Biophysical Basis for Selectivity

Traditionally, the study of channel selectivity has focused on the ability of these molecules to distinguish among similar small ions such as the alkali metals, the alkaline earths and the halides. The work of Eisenman and coworkers (1975) has demonstrated however the importance of ion dehydration on selectivity; the energy needed to dehydrate the ion is just as important as the binding energy to specific sites on the protein. More recently the discussion has focused on how one can get specific binding and, at the same time, fast throughput.

In the case of proteins that form large aqueous pores, selectivity among ions whose size is much smaller than that of the pore must rest largely on more long-range forces than direct atomic interactions (electrostatics, dipole, or van der Waal's). In the case of VDAC, the large pore size (2.5 to 3 nm in diameter) means that ions located in the center of the pore may feel a rather different electrostatic environment compared to those moving closer to the walls of the channel. In fact, the distribution of electrostatic charge on the walls of the pore may be uneven, resulting in preferred conduits. The extent of the inhomogeneity depends on the ionic strength and the effect depends on the size of the ion. Thus, ion flow through large channels is likely to represent the composite of a number of a distinctly different processes when compared to ion flow through narrow channels.

The selectivity of a channel for different ions is often estimated by determining the zero-current potential (reversal potential) in the presence of an ion gradient. By modifying a theory developed by Teorell (1953) for ion flow through ion-exchange membranes, a theoretical description of ion flow through VDAC was achieved (Zambrowicz and Colombini, 1993). This large channel theory (LCT) provides values for the reversal potential under various conditions of ion activities and activity gradients. The theory accounts at least qualitatively and often quantitatively for the reversal potentials observed under different conditions. Hence, the model is likely to reflect to some extent the way the ions actually flow through VDAC.

LCT divides the cylindrical plug of solution in the channel into two compartments: an outer shell of solution close to the wall of the channel that contains immobile charge and a central cylindrical compartment that is totally devoid of the effects of fixed charge. These two pathways are in parallel and current may even flow in opposite directions in the two compartments. The physical size of each compartment depends on the ionic strength. The thickness of the cylindrical shell is defined by the cylindrical equivalent of the Debye length: i.e. the distance at which the surface potential decays to 1/e of the value at the surface.

This rather crude framework accounts for some unusual observations. The reversal potential does not increase monotonically with increase in activity gradient of the salt but goes through a maximum (Fig. 5). The reversal potential varies with ionic strength even if the activity ratio does not. It goes from one value at low ionic strength to another at high ionic strength. At low ionic strength it is dominated by the properties of the cylindrical shell and thus reflects mainly the charge on the walls of the channel. At high ionic strength it reflects the properties of the central cylinder and thus the properties of the ions making up the salt (e.g. the difference in ion mobility).

Thus, ion flow through VDAC's large aqueous pore is probably more complex than one might have at first imagined. The selectivity change accompanying channel closure is therefore a result of not only a change in the net charge on the wall of the channel, but also a change in the diameter of the pore causing the effect of cylindrical shell to become more dominant. This model is consistent with many of the results obtained by analysis of VDAC channels containing site-directed mutations as described in detail below.


3.1. Genes Encoding VDAC

VDAC genes have now been isolated from mammals and plants, as well as from fungi. While originally only single sequences were reported from the fungi Neurospora crassa and Saccharomyces cerevisiae, it is now clear that many species have multiple VDAC genes. The spectrum of genes encoding VDAC isoforms has been most completely characterized in humans. Human cDNAs representing two different VDAC genes (HVDAC1 and HVDAC2, Blachly-Dyson et al., 1993) have been identified. HVDAC1 encodes the protein purified from human B-lymphocytes by Kayser et al. (1989). The proteins encoded by HVDAC1 and HVDAC2 are only 75% identical at the amino acid level, and HVDAC2 contains an 11 residue amino-terminal extension relative to HVDAC1. Both human VDACs have less than 30% sequence identity with yeast VDAC. In addition, a cDNA that differs from HVDAC2 at the 5' end (Ha et al., 1993) has been described. This transcript may represent a splicing variant of HVDAC2, or perhaps an incompletely spliced transcript, since it contains the 5' end sequences of HVDAC2 within its 5' untranslated region. In addition, it contains a single nucleotide deletion relative to HVDAC 2 near the 3' end of the coding region, resulting in a frameshift which alters the translation of the C-terminal end of the protein. Two additional human sequences highly homologous to HVDAC1 have been identified by polymerase chain reaction (Blachly-Dyson et al., 1994), but it is not know whether these represent expressed genes. A protein highly homologous to HVDAC1 and a cDNA highly homologous to HVDAC2 have been isolated from rat brain (Bureau et al., 1992). In addition, an HVDAC1-like cDNA has been isolated from bovine brain (Dermietzel et al, 1994). Three mouse VDAC genes, with 65-75% sequence identity to each other have also been cloned (Craigen et al., 1994).

VDAC genes have also been isolated from several plants. VDAC genes from pea and maize were found to have 58% sequence homology with each other, while each of them was only about 25% identical to fungal or human VDAC (Fischer et al., 1994). cDNAs representing two different potato VDAC genes have been characterized (Heins et al., 1994). These two isoforms have about 75% identity to each other, 25% with human VDAC and less than 25% with fungal VDACs. A wheat VDAC sequence is also available in the database. In addition to these genes, a VDAC gene has been cloned from Dictyostelium discoideum (Troll et al., 1992), and expressed sequence-tagged sites from C. elegans appear to encode a VDAC gene from this species.

3.2. Conservation of Primary and Secondary Structure.

Although, as indicated above, the VDAC molecules of different species have very little sequence conservation, the channel gating and selectivity properties are highly conserved. Both HVDAC1 and HVDAC2 channels (expressed in yeast cell lacking the endogenous VDAC gene) have single channel conductances and ion selectivities (K+ vs. Cl-) that are virtually indistinguishable from those of yeast VDAC (Blachly-Dyson et al., 1993). Likewise, the plant VDACs (Blumenthal et al., 1993) and Dictyostelium VDAC (Troll et al., 1992) form channels with very similar properties, although the sequences are highly divergent.

VDAC sequences contain no long stretches of hydrophobic amino acid residues that could form transmembrane a helices. However, when these sequences are examined for secondary structure motifs, a common pattern emerges. In all the VDAC sequences analyzed, the N-terminal end contains a sequence that can form an amphiphilic a helix. Downstream sequences all have 12 or more peaks of strong alternating hydrophobic/hydrophilic segments which could contribute to the formation of a "sided" b barrel structure with a hydrophilic inner surface facing the pore and a hydrophobic outer surface buried in the membrane. This can be easily quantitated and visualized in a plot called a b pattern (Fig. 6) These homologous sequence patterns have led to a general model for the structure of the VDAC channel in which the pore of the channel is formed by the hydrophilic side of a cylindrically curved b sheet that has a hydrophobic side facing the membrane interior (Blachly-Dyson et al., 1989; Song and Colombini, 1995).

3.3. Site-directed Mutagenesis of Charged Residues in Yeast VDAC

To test this model, extensive site-directed mutagenesis was performed, changing the charge of amino acid residues throughout the sequence of the molecule. Charge changes at positions lining the pore would be expected to alter the selectivity of the open channel; charge increases should increase anion selectivity, while charge decreases (or increases in negative charge) should decrease or reverse the anion selectivity. While a number of the mutations had no effect on selectivity, residues were found throughout the molecule which, when the charge was changed, had the expected effect on selectivity (Fig. 7). These residues were found within the putative amino-terminal a helix and twelve of the proposed transmembrane b strands (Blachly-Dyson et al., 1990). This led to a refinement of the general model outlined above in which the VDAC channel is formed by the amino-terminal a helix and a b barrel consisting of the twelve b strands whose position is now constrained by these functional studies of mutant channels (Fig. 8). (Note: While residues whose charge affects selectivity must line the pore, and thus must reside in transmembrane strands, transmembrane strands that do not line the pore may not affect selectivity). Thus, many charged residues distributed throughout the length of the protein sequence contribute to the overall charge of the cylindrical shell in a roughly additive manner to determine the selectivity of the open channel.


By definition, voltage-gated channels must undergo a voltage-dependent conformational change. This can be achieved by coupling the gating process to the motion of charges ("gating charges") relative to the electric field or the alignment of dipoles with the field. Two approaches have been used to identify the portion of the VDAC molecule that moves in response to an applied field. First, the site-directed mutations were used to determine whether the same residues that contributed to the selectivity of the open channel also contributed to the selectivity of the closed (low conducting) state of the channel. Residues that contributed to the selectivity of one state, but not the other, were candidates for being part of the "gate" that moves when the channel opens. Second, these mutations were examined for their effect on the gating of the channels. Changing a charged residue that contributes to the gating charge (i.e. a charge that "senses" the applied field during gating) should affect channel gating parameters in predictable ways. Both of these approaches indicated that a large portion of the protein moves during channel gating.

4.1. Magnitude of the Structural Change Associated With Voltage Gating

The VDAC channel is a large pore that undergoes large changes in both size and selectivity during channel closure (see above). Thus, it should not be surprising that large conformational changes may be involved in channel gating. While the large conductance change associated with VDAC closure suggests a large structural change, there is often a poor correlation between conductance and pore size (Finkelstein, 1985), so it is important to determine the size of the open and closed state pores by other methods. A good assessment of open pore size is obtained by using non-electrolytes to probe the steric barrier. The ability of inulin, polyethylene glycol (PEG 3400), and dextran (Zalman et al., 1980) to pass through the open state of VDAC (Colombini, 1980b), indicate a large open pathway for the narrowest portion of the channel. The flexibility of PEG and dextran may tend to overestimate the pore size of the open state, but inulin is not as flexible. Electron microscopy of negatively-stained images of VDAC channels in 2d crystalline arrays indicates pore diameters of 2.4 to 3 nm. Estimates based on access resistance considerations (Vodyanoy et al., 1992) yield a pore diameter of 2.4 after correction for the size of the permeating ion. Hence, a reasonable estimate of the pore diameter is 2.5 to 3 nm.

Closer agreement exists for the pore diameter of the low-conducting "closed" state. Functionally, this state is just barely permeable to gamma cyclodextrin (Colombini et al., 1987). This rigid spheroid is 1.9 nm in diameter. This is almost the same estimate obtained by electron microscopy of negatively-stained arrays (Mannella and Guo, 1990). These estimates must be viewed with caution however since the closed state for both estimates was induced by the addition of König's polyanion, a synthetic polymer that induces channel closure at very low concentrations. While all indications favor the conclusion that the states induced by this polymer are the same as those induced by the electric field, this may not be totally correct.

To estimate the change in pore volume during closure, Zimmerberg and Parsegian (1986) used macromolecules that could not penetrate the pore of VDAC, to induce a tension within the channel that favored the closed state. From the energy change induced in the molecule, these investigators calculated a change in the volume of the channel upon closure of between 20 and 40 nm3. This is consistent with a large, global change in the structure of the pore as opposed to a local, shutter-like, constriction at one point in the channel. Together, these studies lead to the conclusion that closure of the channel results in a rather large change in effective pore size.

4.2. Location of the Mobile Domain

Functional experiments on channels reconstituted into planar membranes indicated that regions of the channel move across the membrane during the gating process. Addition of the reagents aluminum hydroxide (Zhang and Colombini, 1990) or succinic anhydride (Doring and Colombini, 1985) to one side or the other of the membrane caused asymmetric changes in the behavior of the channels in the membrane. The changes observed depended on the state of the channel (open or closed) when the reagent was added. These results were best explained by the translocation of some groups all the way through the membrane.

In general, the motion of a protein domain can be detected if, in its new environment, its effect on a property of the protein is quite different. Since the selectivity of the channel changes rather dramatically upon closure, the electrostatic nature of the ion-conducting pathway must change radically. This could result from a change in the residues forming the protein wall of the pore; new residues could be introduced or existing residues could be removed. Since channel closure also results in a reduction in the effective diameter of the pore, a removal of residues forming the wall of the open channel seems more likely.

Having identified regions that form the wall of the pore in the open state, Peng and coworkers sought to determine if some of these no longer affected the channel in the closed state. Such regions would be part of the mobile domain. Using the battery of mutant proteins with the single amino-acid substitutions at locations either within the pore or outside the pore, Peng et al. (1992a) identified residues that influenced the selectivity of the channel in the open state and not in the closed state (Fig. 9). In addition there were residues that still affected the selectivity of the closed state but to a lesser extent than their effects on the selectivity of the open state. Both of these classes of mutations were considered to identify mobile regions in the channel. Most of these mutations fell in the N-terminal end of the channel, specifically the a helix and nearby 3 b strands in the model in Fig. 8. One mutation was at position 282 at the C-terminus. Since in a "barrel"-like structure this transmembrane segment is expected to be adjacent to the a helix, this mutation could still be part of a contiguous mobile domain. However, a mutation at position 152, in the middle of a non-mobile region, also affected the selectivity of the channels in the open state but not the closed state. Subsequent studies have indicated that this region of the protein contributes in unique ways to the gating process and in more recent refinements of the general structural model has been repositioned to reside in mobile regions (see Fig. 10).

Thus, these results are consistent with the motion of regions forming part of the wall of the channel in the open state, out of the channel proper or at least away from the ion stream. If the region that moves has a net positive charge, this change would result in a reduction in the net positive charge on the walls of the pore and a reduction in the effective pore radius. It would also result in a reduction in the volume of water within the pore. All these predictions are consistent with experimental observations (see above).

4.3. Identification and Localization of the Voltage Sensor

Based on these results, we hypothesized that the domains that affect selectivity in the open, but not the closed VDAC channel, correspond to or overlap the voltage sensor, and that this voltage sensor is a moiety with a net positive charge that moves perpendicular to the membrane out of the channel wall during channel closure. The hypothesis that this mobile domain contains the voltage sensor makes very clear predictions: 1) If the charge on any part of this domain is altered it must affect the steepness of the voltage-dependence of the channel. Steepness is reflected in the parameter n in a two-state model defined by the equation

ln[(Gmax - G)/(G - Gmin)]=(nFV - nFV0)/RT

where V0 represents the voltage at which half the channels are closed; G, Gmax and Gmin are the conductance at any voltage, V, the maximum conductance and the minimum conductance respectively; and F, R, and T are the Faraday constant, the gas constant, and the absolute temperature respectively; 2) n should be increased if the charge in sensing domains is made more positive and decreased if it is made more negative; 3) The change in n induced by such a change in charge must be proportional to the magnitude of the charge change and the fraction of the electric field through which the charge moves.

These predictions were tested by examining the effects on the parameter n of mutations that changed the charge of specific residues. Amino-acid substitutions that changed the charge at eight positions in the N-terminal a helix, three nearby b strands, or the C-terminal b strand increased or decreased n if the site was made more positive or more negative, respectively (Thomas et al., 1993). These sites matched (dark bars in Fig. 9) the sites that seem to move based on selectivity changes (section 4.2 above). This correspondence by two different approaches provides strong evidence that these regions are moving through the field during the gating process. The lack of correspondence at position 248, a site close to the mouth of the channel, could be explained by the residue moving out of the ion stream without moving through a significant portion of the electric field.

There are, however, difficulties in understanding the motion of the sensor. The substitution of lysine for aspartate at position 30 (D30K) had no significant effect on n. This substitution also affected the selectivity of VDAC in both the open and closed states indicating that this residue is not moving out of the channel upon channel closure. Thus, although these results are consistent with each other, it is hard to see how the nearby regions of the protein can move out of the channel without this strand also moving. D51K also did not generate the expected changes in n. Located in the loop region between the second and third b strands, this residue is positioned outside the pore in current models since D51K does not affect open-state selectivity and would not have to traverse the electric field if the mobile domain moved toward it. Changes in a nearby residue, K46E, decrease the voltage dependence of gating processes at both positive and negative potentials indicating that the domain containing K46 and D51 may move in both directions. D51K would be expected to increase the voltage dependence of at least one of the processes. The D51K mutation had no effect on n however. D51K thus would be expected to increase the voltage dependence of at least one of the processes. These findings point out the fact that although current models are valuable tools for understanding and summarizing results obtained in functional studies, the models are almost certainly simplified representations of structures and changes that are undoubtedly more complex.

Amino-acid substitutions that changed the charge at other locations generally had no effect on the steepness of the voltage dependence. A notable exception was glutamate 152 on b -strand #7 (Fig. 8). This region is in the middle of the molecule between strands that are proposed to remain fixed during the gating process. E152K increased the steepness of the voltage dependence but did so in an asymmetric manner, i.e. it increased the voltage dependence of only one of the two gating processes. The asymmetry can be explained if E152 is located near one end of a transmembrane strand. More difficult to explain is the fact that this strand has a net negative charge and reducing this negative charge by the glutamate to lysine substitution increases the voltage dependence. This means that normally this negatively charged strand moves toward the negative side of the membrane. This seems possible only if it is coupled to the motion of positively-charged regions moving in the same direction. However, the adjacent b strands in Fig. 8 do not move based on selectivity and voltage-dependence measurements on charge substitutions in these regions. It was therefore proposed (Zizi et al., 1995) that this strand might be located elsewhere such as within the mobile region (as in Fig. 10 upper panel). This relocation is possible because of the long loop regions connecting this strand to the nearby transmembrane segments. More importantly, the relocation of the strand containing residue 152 into the mobile domain provides constraints on the movement of this domain. Because the strand is linked to non-mobile regions, it may act as a tether that is stressed by channel closure and this stress would be released by the reopening process (Fig. 10 lower panel). Such a proposal would account for the rapid, sub-millisecond, rates of channel opening (Schein et al., 1976; Colombini, 1979).

If the strand containing residue 152 does in fact move normal to the membrane upon channel closure, the nearby residue glutamate 145 is likely to move as well. Changing this residue to a positive charge should increase the voltage dependence of closure. When glutamate 145 was mutated to lysine (E145K), an asymmetrical increase in voltage dependence was observed, similar to the effect of the E152K mutation (Fig. 11). Furthermore, if residues 145 and 152 are located at opposite ends of a transmembrane strand, E145K and E152K should increase the voltage dependence of the two different gating processes, respectively. This was tested (Zizi et al., 1995) by generating the double mutant. E145K/E152K showed elevated steepness of voltage dependence in response to both positive and negative potentials (Fig. 11), consistent with this strand moving in one direction at positive potentials and the other at negative potentials.

Thus, while the details of the gating process need to be clarified, the voltage sensor is clearly a region of the protein forming part of the wall of the pore. This region moves out of the channel upon channel closure and in the process traverses part of the electric field, resulting in a voltage-dependent change that accounts for the voltage dependence of the conductance.

4.4. Activation Energy and the Gating Mechanism

The large conformational change proposed for the gating process raises the question of the size of the activation energy. While VDAC closure is a slow process, the opening is fast (sub-millisecond). Can such a large conformational change occur in such a short time?

The movement of the domain illustrated in Fig. 12 requires the cleavage of the hydrogen bonds connecting transmembrane b strands. On closure, polar side chains facing the aqueous phase within the pore would face a similar environment on the membrane surface. Apolar groups facing the hydrocarbon chains of the phospholipids need not be removed from this environment but slide along to a location near the membrane surface. Thus it is not unreasonable to postulate that breaking the hydrogen bonds between the mobile domain and the rest of the structure may be the major contribution to the energy barrier.

There is no general agreement as to the stabilizing effect of hydrogen bonds on protein folding. Although estimates are very crude, it is generally agreed that hydrogen bonds need to be formed within the protein structure but the strength of these intramolecular hydrogen bonds as compared to that of hydrogen bonds with water is not very different, perhaps 4 kJ/mole more stable. Since there are two edges where hydrogen bonds between b strands must be broken in order to move these strands relative to each other during the closing process, the proposed VDAC gating mechanism would require approximately 20 hydrogen bonds to be broken for channel closure. Only half as many need to be broken for channels to open from the closed state since only one set of hydrogen bonds need be cleaved to reinsert the portion that moved out into the wall of the channel . Consistent with this mechanism, estimates of the enthalpy change for the transition from the open to the closed states are 35 to 40 kJ/mole (Pavlin and Colombini, in preparation), or roughly 10 hydrogen bonds. Thus, this difference in activation energy can account for the 1000 fold difference in the rate of channel closure compared to channel opening (Colombini, 1979) (taking reasonable values for the energy to break a hydrogen bond).

4.5. Ion Flow and the Gating Process

Another unusual observation that can be explained by the proposed gating mechanism is the observed shift along the voltage axis of the switching region of the voltage-gating processes in the presence of a salt gradient. When the salt concentration on both sides of the membrane is the same, the two gating processes generally occur at the same magnitude of the membrane potential (VDAC reconstituted into soybean phospholipid membranes). However, in the presence of a salt gradient (1 M vs 0.1 M) there is a pronounced shift in that higher negative potentials and lower positive potentials on the high-salt side are required to close the channels. It appears that moving the sensor against the salt gradient required more energy than moving it down the gradient. These observations lead to the hypothesis that some of the kinetic energy from the flow of salt through the channel was imparted to the mobile domain favoring motion down the gradient. This was tested and confirmed by using salts of different mass and size (Zizi, Byrd and Colombini, in preparation). Thus, as illustrated in Fig. 13, the bias that ion flow imparts to the gating process can be explained by the transfer of kinetic energy to the mobile domain. This notion is consistent with the proposed gating mechanism since residues forming the ion conduction pathway also form part of the voltage sensor or gate.


Section 4 focused on one proposed mechanism for voltage-gating in VDAC. Very different mechanisms have been proposed by Mannella and others (Mannella, 1990; Mannella et al., 1992; Adams and McCabe, 1994). In addition the actual mechanism may differ from any yet proposed. Here, we consider how the current experimental evidence may exclude other possible mechanisms (see Fig. 14).

5.1. Corking the Bottle

In analogy to the mechanism that seems to underlie slow inactivation of Na+ and K+ channels, the idea of plugging the channel with a protein domain from the surface seems simple and straight forward. To be consistent with the conductivity of "closed" VDAC channels, this would need to be a porous cork. However, even with such a modification, this mechanism has difficulty accounting for a number of experimental observations including:

i) the large volume change associated with channel closure

ii) the fact that certain residues that influenced selectivity in the open state no longer did so in the closed state

iii) the ability of residues that influence channel selectivity in the open state to influence the steepness of the voltage dependence

iv) the voltage dependence of the gating process.

A modified proposal by Mannella (1990) tries to circumvent some of these problems. He proposes that the a helix enters the channel resulting in a reduced pore volume, thereby overlying regions of the wall of the pore and masking the effect of charges at these sites on the selectivity of the resulting closed channel. The difficulties with this proposal are:

i) The a helix is proposed to lie on the surface of the membrane in the open channel, away from the ion stream. Yet charge substitutions at two sites in the helix (D15K and K19E) affect the selectivity of the open state of the channel, indicating this region is in intimate contact with the ion stream. In the closed state, these sites on the a helix have less, not more, affect on selectivity, in contrast to the expected results if the helix enters the channel during closure.

ii) While the a helix of VDAC from some species has a net charge, in some cases there is no net charge (e.g. in yeast VDAC). Thus there is no apparent way for enough charge to move through the field to account for the voltage dependence of the gating process.

iii) Since one surface of the a helix is hydrophobic, the movement of this strand from the membrane surface to the inside of a polar channel would require a high activation energy.

iv) It is unclear how putting an a helix in the channel would mask the effect of charged residues on ion flow. The formation of ion pairs would require residues of opposite charge to interact with the existing residues on the walls of the pore. In order for the a helix to dramatically change the selectivity of the channel (favoring cations), it would have to be a negatively-charged structure. In fact, the a helices of VDAC channels from a wide variety of species are either neutral or positively charged (Song and Colombini, 1995). Thus, this mechanism could not account for the change in selectivity associated with the gating process.

v) this mechanism cannot account for two distinct gating processes in each VDAC channel.

5.2. Forming a Jelly Roll

The possibility that the b barrel could be reduced in size by overlapping the ends of the wall, has been proposed (Mannella, 1990). Since the N and C-terminal ends seem to be involved in the structural change, this has some appeal. However, it is unclear how such a structural change could result in charge motion through the electric field and thus be voltage dependent. It is also difficult to see how the apolar groups facing the hydrocarbon portion of the membrane could be induced to now face the polar protein surface that faced the aqueous pore

5.3. Tilting the Strands

Following the mechanism proposed for gating of gap junction channels (Unwin and Zampighi, 1980), one mechanism of closure might involve the straightening of transmembrane b strands. In the open channel, transmembrane b segments must be severely tilted (roughly 60o in the current model) given the dimensions of the channel formed by a single VDAC protein. In addition to not accounting for voltage dependence and the selectivity change associated with gating in VDAC, such a conformational change would require that every hydrogen bond between each of the staves of the barrel be broken. This would be an unlikely event and should have a high enthalpic activation energy.

5.4. Deforming the Cylinder

One might imagine that the cylindrical channel might be deformed in such a way that the cross-section would look like an ellipse. However this proposal suffers from some of the same shortcomings encountered by the tilting of strands. In addition, the high curvatures produced by such a change would result either in strained hydrogen bonds between some strands or actual broken bonds. If broken, it is hard to see how these could be satisfied by the apolar membrane interior.

While the mechanism illustrated in Fig. 12 is consistent with the vast majority of the experimental data and explains many of the special properties of VDAC channels, there a few apparent inconsistencies. In addition to those already mentioned, Mannella and Guo (1990) have shown that closure results in only small changes in the packing of the channels into 2-dimensional arrays. One might expect that the reduction in the overall diameter of the pore upon closure would result in a tighter packing. However, it is known that a considerable amount of phospholipid is present in the ordered arrays and, because lipids move laterally at a very rapid rate, it is not unreasonable to expect that phospholipids have filled in any space generated by the closing process. The spacing of the channels themselves may be determined more by the nature of the surface regions of the channel that simply the packing together of cylindrical proteins. Indeed, the packing of VDAC channels is not a simple hexagonal close-packing, indicating specific interactions between individual channels.


In addition to two voltage-gating processes, there are a number of other agents and factors that affect the probability of finding VDAC channels in either an open or a closed state. Whether cells actually use these to regulate VDAC and thus the permeability of the mitochondrial outer membrane, has yet to be demonstrated directly. However, a number of these regulatory interactions have been found to be highly conserved in VDAC channels from widely different sources, implying that these mechanisms are maintained by selective pressure. Thus, it is likely that these observations define important regulatory interactions in vivo.

6.1. Amplification of the Voltage Dependence

The parameter, n, reflecting the steepness of the voltage dependence of VDAC reconstituted into planar membranes varies from 2 to 5 depending on the source of VDAC and the experimental conditions. This is comparable to the voltage dependence of voltage-gated channels responsible for action potentials. The n value of VDAC can be dramatically augmented by treating with small amounts of polyanions. Thus, n is increased 10 fold by the presence of 100 nM dextran sulfate (500 kDa) and it can be increased by another factor of 2 at higher concentrations (Mangan and Colombini, 1987) (Fig. 15). A variety of polyanions, from polyaspartic acid to RNA have similar effects (Colombini et al., 1989). Stronger effects are observed with simple polymers and the potency of the effect increases with increasing charge density on individual polymers, indicating that the structure of specific polyanions may determine the extent of interaction with VDAC.

The polyanions appear to act by favoring the closed state of the channel, rather than by a voltage-dependent blockage. The drop in conductance upon channel closure is virtually the same whether voltage alone is used to close the channel or voltage-dependent closure is augmented by the addition of dextran sulfate. There is no sign of a decrease in single-channel conductance or of channel flickering in the presence of polyanions.

The ultra-steep voltage-dependence induced by the polyanions is consistent with the voltage-dependent increase in the partitioning of the polyanion into the access-resistance region at the mouth of the channel (Mangan and Colombini, 1987). There, the polyanion presumably interacts with the positively-charged voltage sensor to increase the likelihood of these domains translocating out of the membrane and closing the channel. The fact that the polyanion acts preferentially from the negative side of the membrane is consistent with this proposal.

Polyanions with hydrophobic regions show much more complex behavior. The best studied is König's polyanion: a copolymer of methacrylate, maleate, and styrene in 1:2:3 ratio with average molecular weight of 10,000. In addition to increasing the voltage dependence as dextran sulfate does, it binds and induces closure even in the absence of a membrane potential (Colombini et al., 1987). The effect seems to be progressive in that, with time, the channels tend to enter states of lower and lower conductance. When König's polyanion was used to increase the probability of closure, VDAC-containing vesicles showed a reduction in permeability to non-electrolytes consistent with a reduction in channel diameter but the channel was still permeable to cyclodextrin (1.9 nm in diameter). This is not consistent with voltage-dependent block by a leaky plug but simply an increase in the probability of the channel entering a closed state.

To date, there are no examples of other channels responding to polyanions in this way. Thus the structure of VDAC and its gating process may be specifically inclined to respond to these polyanions. The possibility that these polyanions mimic the action of a natural substance led to the discovery of a protein (the VDAC Modulator) that acts somewhat like the polyanions.

6.2. Action of the VDAC Modulator

The VDAC modulator is a soluble protein that induces VDAC closure in planar phospholipid membranes when added at very low concentrations (Holden and Colombini, 1988). It is very sensitive to proteases, requires the presence of DTT for stability (Liu et al., 1993), and migrates with an apparent molecular weight of approximately 100,000 by gel filtration chromatography. It has been identified in mitochondria from mammals, fungi, and higher plants (Liu and Colombini, 1991) and localized to the intermembrane space of N. crassa mitochondria (Holden and Colombini, 1993). There also appears to be more than one VDAC modulator activity in calf liver, one more potent than the other (Liu et al., 1993). When added to VDAC reconstituted into planar membranes, the modulator binds very tightly since it is not washed out easily by perfusing the aqueous compartment. Functionally, the modulator not only increases the voltage dependence of the channels (Liu and Colombini, 1992b) but also tends to keep the channels in closed states, inducing channels to enter states of very low conductance which are usually achieved only at high voltages. The modulator acts from both sides of VDAC channel incorporated into planar membranes. As was the case with dextran sulfate, the modulator acts when the modulator-containing side is made negative. When added to mitochondria with intact outer membranes, the VDAC modulator reduces the permeability of the outer membrane and inhibits mitochondrial metabolic activities that require the flux of metabolites through the outer membrane (Liu and Colombini, 1992a). When the outer membrane is damaged, the inhibition is eliminated.

If the modulator is located only in the intermembrane space, its ability to bind from outside the mitochondria seems puzzling. It is possible that the modulator or related molecules may also be present in the cytoplasm, although this has yet to be demonstrated. However, since VDAC can gate in response to either positive or negative potentials, the modulator might be able to bind to it on either side of the channel if the same domain is exposed during each gating process.

6.3. Action of Nucleotides

When nucleotides (ATP, ADP, cAMP, cGMP, NADH, NADPH, and NAD+) were added to VDAC channels reconstituted into phospholipid membranes, only NADH and NADPH has significant effects on the function of the channel, NADH being an order of magnitude more potent (Xiaofeng Xu unpublished observations). NADH doubles the voltage dependence of the channel with a KD estimated to be in the low m M range (Zizi et al., 1994). ATP has been reported to bind to human VDAC (Florke et al., 1994) and even to a peptide consisting of only the 35 N-terminal amino acids. However, it has no effect on the function of VDAC reconstituted into phospholipid membranes (Zizi et al., 1994). The addition of NADH and NADPH to mitochondria with intact outer membranes results in a 6 fold reduction in the permeability of the outer membrane to ADP (Lee et al., 1994). The effect of this on mitochondrial respiration is most pronounced at physiologically relevant ADP concentrations. These results are consistent with NADH inducing VDAC closure.

6.4. Colloidal Osmotic Pressure

The presence in the medium of macromolecules that cannot enter VDAC's aqueous pore results in an imbalance in the activity of water that can only be corrected by decreasing the hydrostatic pressure in the pore. This reduced pressure favors channel closure if the volume of the pore in the closed state is reduced. This was the line of reasoning used to estimate the volume change upon channel closure (Zimmerberg and Parsegian, 1986). Thus, VDAC is sensitive to the concentration of macromolecules and will be influenced by the concentration of these molecules in the cell. The influence of the colloidal osmotic pressure on isolated mitochondria is consistent with the closure of VDAC channels by increasing osmotic pressure. Mitochondrial functions are greatly inhibited by the addition of 10% dextran (Gellerich et al., 1993).

6.5. Inhibition by Metal Hydroxides

Neutral metal trihydroxides strongly inhibit the ability of reconstituted VDAC channels to close in response to a membrane potential (Zhang and Colombini, 1989). The best studied is aluminum trihydroxide (Dill et al., 1987) formed by the addition of aluminum chloride to a buffered solution, and is by far the major species at physiological pH. It is, however, metastable and should eventually precipitate. Other metals in the same group (Ga, In) have a similar effect, as do transition metals that form the a neutral trihydroxide. The potency of the different metal trihydroxides depends on the relative amount of the trihydroxide formed under given conditions (e.g. a particular pH) vs. other possible hydrated species.

The action of aluminum hydroxide resembles what one would expect when the voltage sensor is neutralized. However, aluminum trihydroxide does not affect the selectivity of the channel in the open conformation indicating that the net charge on the wall of the pore (including that in the mobile domain) is unchanged (Dill et al., 1987). In addition, experiments indicate that the aluminum binding site moves through the membrane in the opposite direction to that of the voltage sensor (Zhang and Colombini, 1990). Thus, the aluminum binding may result in the effective movement of positive charge in the opposite direction to the sensor thus reducing the net charge transferred.


Experimental observations demonstrating that VDAC channels insert into phospholipid membranes in an oriented manner has led to the proposal that these channels have the property of "auto-directed insertion" (Zizi et al., 1995). While a preferential insertion direction is not surprising, the evidence indicates that for VDAC the direction of insertion of the first channel determines the direction of insertion of virtually all other channels. If so, an inserted VDAC channel must: 1) interact with the other channels as they insert; 2) must provide directional information that constrains the direction of insertion; and 3) must accelerate the rate at which channels insert.

Experimental evidence in support of auto-directed insertion was obtained by examining the voltage-gating properties of two yeast VDAC mutants each possessing a single amino-acid substitution that rendered the channel asymmetric in its behavior. The asymmetric behavior allowed one to determine the direction of insertion. The degree of asymmetry was the same whether 1 or 100 channels were present in the membrane, indicating that, in the multi-channel membrane, virtually all the channels were inserted in the same direction. More important, the direction of the insertion varied from membrane to membrane in an apparently random fashion. This led to the hypothesis that the first insertion is random and subsequent insertions are determined by the direction of the first insertion. Various possible causes for the apparent randomness of the membrane to membrane variation in the direction of insertion, were investigated. Neither the sign of the membrane potential during the insertion process nor the side of the membrane to which the VDAC-containing sample was added had any significant influence on the direction of insertion. Thus, the most likely cause of this variation is the random nature of the first insertion.

The proposal that the pre-inserted channels interact with channels attempting to insert into the membrane requires that insertions at the location of the pre-inserted channel be highly favored over insertions at other locations. Since the entire membrane had a surface area 109 times greater than that of a single VDAC channel, in order for channels to preferentially insert next to the pre-inserted channel over anywhere else on the membrane, the preference must be greater than 109. To explain the observation that the insertion direction in a single experiment is almost perfect, the preferential insertion should be at least an order of magnitude greater. Thus, the most straight-forward interpretation of the experiments is that VDAC channels both catalyze and determine the direction of insertion of fellow channels. This property of auto-directed insertion has important implications in protein targeting and the maintenance of oriented VDAC channels in the outer membrane of mitochondria.


Several enzymes (hexokinase, glucokinase and glycerol kinase) that use ATP produced by mitochondria have been found to associate with the outer surface of the mitochondrial outer membrane by binding to VDAC protein. This localization presumably gives the enzymes preferential access to mitochondrially generated ATP. In addition, creatine kinase of the mitochondrial intermembrane space has been found to associate with VDAC on the inner surface of the outer membrane.

Hexokinase associates with mitochondria to varying extents, depending on tissue type, metabolic state and developmental stage, with especially high levels of mitochondrially bound hexokinase seen in highly glycolytic tumor cells (Reviewed in Adams et al, 1991). The outer membrane binding site for hexokinase has been identified as the VDAC protein, and hexokinase binds to purified VDAC reconstituted into liposomes (Linden et al., 1982, Fiek et al.,1982). This binding is inhibited by glucose 6-phosphate, the product of hexokinase activity (Fiek et al., 1982), although the fraction of mitochondrially bound hexokinase that can be released with glucose-6-phosphate treatment varies between species (Kabir and Wilson, 1994). The binding of hexokinase to outer membranes is thought to increase the accessibility of mitochondrially generated ATP to the enzyme, and hexokinase bound to mitochondria preferentially uses mitochondrial, as opposed to exogenous ATP (Rasschaert and Malaisse, 1990). Binding of hexokinase and glucokinase to the mitochondria in pancreatic b cells and adipocytes may be involved in regulation of blood sugar. Hexokinase in adipocytes and glucokinase in an insulinoma cell line (RINm5F) coimmunoprecipitate with a VDAC isoform found in these cells, and purified VDAC from adipocytes and pancreatic b cells mediates the binding of liver hexokinase to liposomes in a glucose-6-phosphate dependent manner. Treatment of RINm5F cells or adipocytes with the drug glimepiride, which lowers blood sugar in vivo, reduces the fraction of glucokinase or hexokinase that can be coimmunoprecipitated with VDAC from the respective cell types, and also reduces the ability of VDAC purified from the treated cells to bind liver hexokinase (Müller et al., 1994).

The N-terminal 15 amino acid residues of hexokinase are required for binding to mitochondria (reviewed in Adams et al., 1991). The region of VDAC required for binding hexokinase probably includes glutamate 72. Treatment of mitochondria with N, N'-dicyclohexylcarbodiimide (DCCD), which covalently modifies VDAC and no other outer membrane proteins, inhibits hexokinase binding (Nakashima et al, 1986). Recently, glutamate 72 has been identified as the site of DCCD modification (dePinto et al., 1993). The fact that the lipid-soluble reagent, DCCD, reacted with E72 but water-soluble reagents did not (Nakashima et al., 1986) indicates that E72 is located in an hydrophobic environment. The amino-terminal domain of hexokinase is hydrophobic, and is accessible to hydrophobic reagents only when HK is bound to mitochondria (Xie & Wilson , 1990). Thus it is likely that both this N-terminal domain and VDAC's E72 are either buried in the membrane or in a hydrophobic pocket of the protein.

A portion of cellular glycerol kinase has also been shown to associate with mitochondria (Yilmaz et al., 1987). Glycerol kinase from rat liver cytosol binds to purified VDAC in liposomes, but only when hexokinase binding has been inhibited by addition of glucose-6-phosphate (Fiek et al., 1982; Müller et al., 1994). This suggests that the two enzymes bind to the same site, or to overlapping sites on VDAC.

The mitochondrial form of creatine kinase, which phosphorylates creatine using mitochondrial ATP, is located in the intermembrane space, and is concentrated in contact sites between the inner and outer membranes. Mitochondrial creatine kinase exists as dimers which can further associate into cube-like homo-octamers of 45 kDa subunits (Schlegel et al., 1988). The octameric form can mediate binding between inner mitochondrial membrane lipids and outer membrane lipids (Rojo et al., 1991). Mitochondrial creatine kinase also binds VDAC, and binding to VDAC favors formation of octamers from dimers (Brdiczka et al., 1994). Binding to VDAC may increase the access of creatine kinase to cytoplasmic creatine, which presumably enters the intermembrane space through the VDAC channel. In addition, binding to creatine kinase alters the properties of VDAC channels; when VDAC-octamer complexes insert into lipid bilayers, the single-channel conductance was reduced to about half the level of uncomplexed VDAC (Brdiczka et al., 1994). It is unclear whether these are closed channels, or whether association with creatine kinase partially blocks the pore.


Study of the VDAC ion channel has shown that this small molecule can perform a wide variety of functions. It forms large anion-selective pores that can close to smaller cation-selective pores in two distinct voltage-dependent gating processes. Thus, this molecule can serve as a model of how the relevant charged groups forming a voltage "sensor" couple to other regions of a protein to produce the atomic rearrangements generating channel gating in response to voltage changes. One remarkable feature of all VDAC channels is the high conservation of basic channel properties. VDAC proteins from yeast, plants or humans, are not highly conserved at the primary sequence level yet form channels with essentially identical single-channel conductance, ion selectivity and voltage-dependence in phospholipid membranes. Using yeast VDAC, the functional consequences of a large number of charge changes introduced by site-directed mutagenesis have been investigated. These functional studies have allowed the development of a model of the transmembrane topology of the yeast VDAC protein in the open state, defined regions of the protein that are removed from the pore during channel closure and identified specific residues forming the voltage sensor. Additional studies have demonstrated that gating can be modulated by a variety of factors, including NADH and a protein found in the mitochondrial intermembrane space. The VDAC protein also specifically binds to cellular kinase enzymes, presumably giving them preferential access to metabolites transported through the VDAC channel. Thus, it is likely that VDAC forms the co-ordination point for a complex consisting of a number of molecules and regulation of VDAC's gating properties is likely to depend not only on external factors, such as NADH, but on the dynamic association with a variety of other proteins that constitute a significant metabolic regulatory interaction. Clearly, understanding the function implications of these associations is critical to understanding VDAC's function in the cell. These issues, as well as further refinements of existing models of the transmembrane topology of the VDAC channel and the conformational transitions associated with voltage gating, are likely to form the focus of future analysis of this unique protein.


This work was supported by grants from the Office of Naval Research (N00014-90-J-1024) and the National Institutes of Health (GM 35759).


Adams, V., Griffin, L., Towbin, J., Gelb, B., Worley, K. and McCAbe, E. R. B., 1991, Porin interaction with hexokinase and glycerol kinase: metabolic microcompartmentation at the outer mitochondrial membrane, Biochem. Med. Metabol. Biol. 45: 271-291.

Adams, V., and McCabe, E.R.B., 1994, Role of porin-kinase interactions in disease, in: Molecular Biology of Mitochondrial Transport Systems, NATO ASI Series, Volume 83 (M. Forte and M Colombini, eds.), Springer-Verlag, Berlin, pp. 357-377.

Akabas, M.H., Kaufmann, C., Archdeacon, P., and Karlin, A., 1994, Identification od acetylcholine receptor channel-lining residues in the entire M2 segment of the alpha subunit, Neuron 13: 919-927.

Akabas, M.H., Kaufmann, C., Cook, T.A., and Archdeacon, P., 1994, Amino acid residues lining the chloride channel of the cystic fibrosis transmembrane conductance regulator, J. Biol. Chem. 269: 14865-14868.

Benz, R., Kottke, M., and Brdiczka, D., 1990, The cationically selective state of the mitochondrial outer membrane pore: a study with intact mitochondria and reconstituted mitochondrial porin, Biochim. Biophys. Acta 1022: 311-318.

Blachly-Dyson, E., Baldini, A., Litt, M., McCabe, E. R. B. and Forte, M., 1994, Human genes encoding the voltage-dependent anion channel (VDAC) of the outer mitochondrial membrane: mapping and identification of two new isoforms, Genomics 20: 62-67.

Blachly-Dyson, E., Peng, S., Colombini, M. and Forte, M., 1989, Probing the structure of the mitochondrial channel, VDAC, by site-directed mutagenesis: a progress report, J. Bioenerg. Biomembr. 21: 471-483.

Blachly-Dyson, E., Peng, S., Colombini, M. and Forte, M., 1990, Selectivity changes in site-directed mutants of the VDAC ion channel: structural implications, Science 247: 1233-1236.

Blachly-Dyson, E., Zambrowicz, E. B., Yu, W.-H., Adams, V., McCabe, E. R. B., Adelman, J., Colombini, M. and Forte, M., 1993, Cloning and functional expression in yeast of two human isoforms of the outer mitochondrial membrane channel, the voltage-dependent anion channel, J. Biol. Chem. 268: 1835-1841.

Blumenthal, A., Kahn, K., Beja, O., Galun, E., Colombini, M., and Breiman, A., 1993, Purification and characterization of the voltage-dependent anion-selective channel protein from wheat mitochondrial membranes, Plant Physiol. 101: 579-587.

Brdiczka, D., Kaldis, P. and Wallimann, T., 1994, In vitro complex formation between the octamer of mitochondrial creatine kinase and porin, J. Biol. Chem. 269: 27640-27644.

Bureau, M. H., Khrestchatisky, M., Heeren, M. A., Zambrowicz, E. B., Kim, H., Grisar, T. M., Colombini, M., Tobin, A. J. and Olsen, T. W., 1992, Isolation and cloning of a voltage-dependent anion channel-like Mr 36,000 polypeptide from mammalian brain, J. Biol. Chem. 267(12): 8679-8684.

Colombini M., 1979, A candidate for the permeability pathway of the outer mitochondrial membrane, Nature (Lond) 279: 643-645.

Colombini, M., 1980a, Pore size and properties of channels from mitochondria isolated from Neurospora crassa, J. Membr. Biol. 53: 79-84.

Colombini, M., 1980b, Structure and mode of action of a voltage-dependent anion-selective channel (VDAC) located in the outer mitochondrial membrane, Ann. N.Y. Acad. Sci. 341: 552-563.

Colombini, M., 1989, Voltage gating in the mitochondrial channel, VDAC, J. Membr. Biol. 111: 103-111.

Colombini, M., 1994, Anion channels in the mitochondrial outer membrane, in: Current Topics in Membranes, Volume 42 (W. Guggino, ed.), Academic press, San Diego CA, pp. 73-101.

Colombini, M., Holden, M.J., and Mangan, P.S., 1989, Modulation of the mitochondrial channel VDAC by a variety of agents, in: Anion Carriers of Mitochondrial Membranes, (A. Azzi, K.A. Nalecz, M.J. Nalecz, and L. Wojtczak, eds.), Springer-Verlag, Berlin, pp. 215-224.

Colombini, M., Yeung, C. L., Tung, J., and König, T., 1987, The mitochondrial outer membrane channel, VDAC, is regulated by a synthetic polyanion, Biochim. Biophys. Acta 905: 279-286.

Craigen, W. J., Lovell, R. S. and Sampson, M. J., 1994, Structure and expression of mouse mitochondrial voltage dependent anion channel genes, Amer. J. Human Genet. 55: (3) supplement.

DePinto, V., Jamal, J. A. and Palmieri, F., 1993, Location of the dicyclohexylcarbodiimide-reactive glutamate residue in the bovine heart mitochondrial porin, J. Biol. Chem. 268: 12977-12982.

Dermietzel, R., Hwang, T.-K., Buettner, R., Hofer, A., Dotzler, E., Kremer, M., Deutzmann, R., Thinnes, F. P., Fishman, G. I., spray, D. C. and Siemen, D., 1994, Cloning and in situ localization of a brain-derived porin that constitutes a large-conductance anion channel in astrocytic plasma membranes, Proc. Natl. Acad. Sci. USA 91: 499-503.

Dill, E.T., Holden, M.J., and Colombini, M., 1987, Voltage gating in VDAC is markedly inhibited by micromolar quantities of aluminum, J. Membr. Biol. 99: 187-196.

Doring, C., and Colombini, M., 1985, The mitochondrial voltage-dependent channel, VDAC, is modified asymmetrically by succinic anhydride, J. Membr. Biol. 83: 87-94.

Durkin, J.T., Koeppe II, R.E., and Anderson, O.S., 1990, Energetics of gramicidin hybrid channel formation: a test for structural equivalence. Side-chain substitutions in the native sequence, J. Mol. Biol. 211:221-234.

Eisenman, G., and Krasne, S., 1975, The ion selectivity of carrier molecules, membranes and enzymes. In MTP International Review of Science, Biochemistry Series, Volume 2 (C.F. Fox, ed.) Butterworths, London, pp 27-59.

Fiek, C., Benz, R., Roos, N. and Brdiczka, D., 1982, Evidence for identity between the hexokinase-binding protein and the mitochondrial porin in the outer membrane of rat liver mitochondrial, Biochim. Biophys. Acta 688: 429-440.

Finkelstein, A., 1985, The ubiquitous presence of channels with wide lumen and their gating by voltage, Ann. N.Y. Acad. Sci. USA 456: 26-32.

Fischer, K., Weber, A., Brink, S., Arbinger, B., Schünemann, D., Borchert, S., Heldt, H. W., Popp, B., Benz, R., Link, T.-A., Eckerskorn, C. and Flügge, U., 1994, Porins from plants: molecular cloning and functional characterization of two new members of the porin family, J. Biol. Chem. 269(41): 25754-25760.

Florke, H., Thinnes, F.P., Winkelbach, H., Stadtmuller, U., Paetzold, G., Morys-Wortmann, C., Hesse, D., Sternbach, H., Zimmermann, B., and Kaufmann-Kolle, P., 1994, Channel active mammalian porin, purified from crude membrane fractions of B lymphocytes and bovine skeletal muscle, reversibly binds adenosine triphosphate (ATP), Biol. Chem. Hoppe Seyler 375: 513-520.

Gellerich, F.N., Wagner, M., Kapischke, M., Wicker, U., and Brdiczka, D., 1993, Effect of macromolecules on the regulation of the mitochondrial outer membrane pore and the activity of adenylate kinase in the inter-membrane space, Biochim. Biophys. Acta 1142: 217-227.

Ha, H., Hajek, P., Bedwell, D. M. and Burrows, P. D., 1993, A mitochondrial porin cDNA predicts the existence of multiple human porins, J. Biol. Chem. 268: 12143-12149.

Heins, L., Mentzel, H., Schmid, A., Benz, R. and Schmitz, U. K., 1994, Biochemical, molecular and functional characterization of two different porins from potato mitochondria, J. Biol. Chem. 264 (42): 26402-26410.

Holden, M.J., and Colombini, M., 1988, The mitochondrial outer membrane channel, VDAC, is modulated by a soluble protein, FEBS Lett. 241: 105-109.

Holden, M.J., and Colombini, M., 1993, The outer mitochondrial membrane channel, VDAC, is modulated by a protein localized in the intermembrane space, Biochim. Biophys. Acta 1144: 396-402.

Kabir, F. and Wilson, J. E., 1994, Mitochondrial hexokinase in brain: coexistence of forms differing in sensitivity to solubilization by glucose-6-phosphate on the same mitochondria, Arch. Biochem. Biophys. 310 (2): 410-416.

Kayser, H., Kratzin, H. D., Thinnes, F. P., Götz, H., Schmidt, W. E., Eckart, K. and Hilschmann, N., 1989, II. Charakterisierung und primärstruktur eines 31-kDa-porins aus menschlichen B-Lumphozyten (porin 31HL), Biol. Chem. Hoppe-Seyler 370: 1265-1278.

Lee, A.-C., Zizi, M., and Colombini, M., 1994, b -NADH decreases the permeability of the mitochondrial outer membrane to ADP by a factor of 6, J. Biol. Chem. 269: 30974-30980.

Linden, M., Gellerfors, P., 1983, Hydrodynamic properties of porin isolated from outer membrane of rat liver mitochondria, Biochim. Biophys. Acta 736:125-129.

Linden, M., Gellerfors, P. and Nelson, B. D., 1982, Pore protein and the hexokinase-binding protein from the outer membrane of rat liver mitochondria are identical, FEBS Letters 141: 189-192.

Liu, M.Y., and Colombini, M., 1991, Voltage gating of the mitochondrial outer membrane channel VDAC is regulated by a very conserved protein, Am. J. Physiol. 260: C371-C374.

Liu, M.Y., and Colombini, M., 1992a, Regulation of mitochondrial respiration by controlling the permeability of the outer membrane through the mitochondrial channel, VDAC, Biochim. Biophys. Acta 1098: 255-260.

Liu, M.Y., and Colombini, M., 1992b, A soluble protein increases the voltage dependence of the mitochondrial channel, VDAC, J. Bioenerg. Biomembr. 24: 41-46.

Liu, M.Y., Torgrimson, A., and Colombini, M., 1993, Characterization and partial purification of the VDAC-channel-modulating protein from calf liver mitochondria, Biochim. Biophys. Acta 1185: 203-212.

Mangan, P., and Colombini, M, 1987, Ultrasteep voltage dependence in a membrane channel, Proc. Natl. Acad. Sci. USA. 84: 4896-4900.

Mannella, C.A., 1986, Mitochondrial outer membrane channel (VDAC,porin): Two-dimensional crystals from Neurospora, Meth. Enzymol. 125: 595-610.

Mannella, C.A., 1987, Electron microscopy and image analysis of the mitochondrial outer membrane channel, VDAC, J. Bioenerg. Biomembr. 19: 329-340.

Mannella, C.A., 1990, Structural analysis of mitochondrial pores, Experientia 46: 137-145.

Mannella, C.A., and Guo, X.W., 1990, Interaction between the VDAC channel and a polyanionic effector, Biophys. J. 57: 23-31.

Mannella, C.A., Forte, M., and Colombini, M., 1992, Toward the molecular structure of the mitochondrial channel, VDAC, J. Bioenerg. Biomembr. 24: 7-19.

Müller, G., Korndörfer, A., Kornak, U. and Malaisse, W. J., 1994, Porin proteins in mitochondria from rat pancreatic islet cells and white adipocytes: identification and regulation of hexokinase binding by the sulfonylurea glimepiride, Arch. Biochem. Biophys. 308: 8-23. (erratum: IBID 313:382)

Nakashima, R. A., Mangan, P. S., Colombini, M. and Pedersen, P. L., 1986, Hexokinase receptor complex in hepatoma mitochondria: evidence from N, N'-dicyclohexylcarbodiimide-labeling studies for the involvement of the pore-forming protein VDAC, Biochemistry 25: 1015-1021.

Peng, S., Blachly-Dyson, E., Forte, M., and Colombini, M., 1992a, Large scale rearrangement of protein domains is associated with voltage gating of the VDAC channel, Biophys. J. 62:123-135.

Peng, S., Blachly-Dyson, E., Forte, M., and Colombini, M., 1992b, Determination of the number of polypeptide subunits in a functional VDAC channel from Saccharomyces cerevisiae, J. Bioenerg. Biomembr. 24:27-31.

Pfaller,R., Freitag, H., Harmey, M.A., Benz, R., and Neupert, W., 1985, A water-soluble form of porin from the mitochondrial outer membrane of Neurospora crassa, J. Biol. Chem. 260: 8188-8193.

Rasschaert, J. and Malaisse, W. J., 1990, Hexose metabolism in pancreatic islets: preferential utilization of mitochondrial ATP for glucose phosphorylation, Biochim. Biophys. Acta 1015: 353-360.

Rojo, M., Hovius, R., Demel, R. A., Nicolay, K. and Wallimann, T., 1991, Mitochondrial creatine kinase mediates contact formation between mitochondrial membranes, J. Biol. Chem. 266: 20290-20295.

Roos, N., Benz, R., and Brdiczka, D., 1982, Identification and characterization of the pore-forming protein in the outer membrane of rat liver mitochondria, Biochim. Biophys. Acta 686: 204-214.

Schein, S.J., Colombini, M., and Finkelstein, A., 1976, Reconstitution in planar lipid bilayers of a voltage-dependent anion-selective channel obtained from Paramecium mitochondria, J. Membrane Biol. 30:99-120.

Schlegel, J., Wyss, M., Schürch, U., Schnyder, T., Quest, A., Wegmann, G., Eppenberger, H. M. and Wallimann, T., 1988, Mitochondrial creatine kinase from cardiac muscle and brain are two distinct isoenzymes but both form octameric molecules, J. Biol. Chem. 263 (32): 16963-16969.

Song, J.M., and Colombini, M., 1995, Indications of a common folding pattern for VDAC channels from all sources, J. Bioenerg. Biomembr. (in press)

Teorell, T., 1953, Transport processes and electrical phenomena in ionic membranes, Prog. in Biophys. Biophys. Chem. 3: 305-369.

Thomas, L., Blachly-Dyson, E., Colombini, M., and Forte, M., 1993, Mapping of residues forming the voltage sensor of the VDAC channel, Proc. Natl. Acad. Sci. USA. 90: 5446-5449.

Thomas, L., Kocsis, E., Colombini, M., Erbe, E., Trus, B. L. and Steven, A. C., 1991, Surface topography and molecular stoichiometry of the mitochondrial channel, VDAC, in crystalline arrays, J. Struct. Biol. 106: 161-171.

Troll, H., Malchow, D., Müller-Taubenberger, A., Humbel, B., Lottspeich, F., Ecke, M., Gerisch, G., Schmid, A. and Benz, R., 1992, Purification, functional characterization, and cDNA sequencing of mitochondrial porin from dictyostelium discoideum, J. Biol. Chem. 267 (29): 21072-21079.

Unwin, P.N.T., and Zampighi, G., 1980, Structure of the junction between communicating cells, Nature 283: 545-549.

Vodyanoy, I., Bezrukov, S.M., and Colombini, M., 1992, Measurement of ion channel access resistance, Biophys. J. 61: A114.

Weiss, M.S., Wacker, T., Weckesser, J., Welte, W., and Schulz, G.E., 1990, The three-dimensional structure of porin from Rhodobacter capsulatus at 3 A resolution, FEBS Lett. 267: 268-272.

Xie, G. and Wilson, J. E., 1990, Rat brain hexokinase: the hydrophobic N-terminus of the mitochondrially bound enzyme is inserted in the lipid bilayer, Arch. Biochem. Biophys. 276: 285-293.

Yellen, G.M., Jurman, M., Abramson, T., and MacKinnon, R., 1991, Mutations affecting internal TEA blockade identify the probable pore-forming region of a K+ channel, Science 251: 939-942.

Yilmaz, M. T., Sener, A. and Malaisse`, W. J., 1987, Glycerol phosphorylation and oxidation in pancreatic islets, Molec. Cell. Endocrinol. 52: 251-256.

Zalman, L. S., Nikaido, H., and Kagawa, Y., 1980, Mitochondrial outer membrane contains a protein producing nonspecific diffusion channels, J. Biol. Chem. 255: 1771-1774.

Zambrowicz, E.B., and Colombini, M., 1993, Zero-current potentials in a large membrane channel: a simple theory accounts for complex behavior, Biophys. J. 65: 1093-1100.

Zhang, D.W., and Colombini, M., 1989, Inhibition by aluminum hydroxide of the voltage-dependent closure of the mitochondrial channel, VDAC, Biochim. Biophys. Acta 991: 68-78.

Zhang, D.-W., and Colombini, M., 1990, Group IIIA-metal hydroxides indirectly neutralize the voltage sensor of the voltage-dependent mitochondrial channel, VDAC, by interacting with a dynamic binding site, Biochim. Biophys. Acta 1025: 127-134.

Zimmerberg, J., and Parsegian, V.A., 1986, Polymer inaccessible volume changes during opening and closing of a voltage-dependent ionic channel, Nature (Lond) 323: 36-39.

Zizi, M., Forte, M., Blachly-Dyson, E., and Colombini, M., 1994, NADH regulates the gating of VDAC, the mitochondrial outer membrane channel, J. Biol. Chem. 269: 1614-1616.

Zizi, M., Thomas, L., Blachly-Dyson, E., Forte, M., and Colombini, M., 1995, Oriented channel insertion reveals the motion of a transmembrane beta strand during voltage gating of VDAC, J. Membrane Biol. 144: 121-129.

Please send comments and contributions to Dr. Marco Colombini

Last modified: April 23, 2004 12:41

Page created and maintained by Swapnil Sharma