Cloning and Functional Expression in Yeast of Two Human Isoforms of the Outer Mitochondrial Membrane Channel, the Voltage-dependent Anion Channel*

The voltage-dependent anion channel (VDAC) of the outer mitochondrial membrane is a small abundant protein found in all eukaryotic kingdoms which forms a voltage-gated pore when incorporated into planar lipid bilayers. VDAC is also the site of binding of the metabolic enzymes hexokinase and glycerol kinase to the mitochondrion in what may be a significant metabolic regulatory interaction. Recently, there has been speculation that there may be multiple forms of VDAC in mammals which differ in their localization in the outer mitochondrial membrane and in their physiological function. In this report, we describe the identification and characterization of two human cDNAs encoding VDAC homologs (HVDACl and HVDACS). To confirm VDAC function, each human protein has been expressed in yeast lacking the endogenous VDAC gene. Human proteins isolated from yeast mitochondria formed channels with the characteristics expected of VDAC when incorporated into planar lipid bilayers. In addition, expression of the human proteins in a liquid scintillation counter. Human Expression-Expression of each HVDAC isoform was assessed by Northern blot analysis of total RNA and by PCR amplifi- cation of transcripts present in human cell lines and tissues. RNAs prepared from the indicated human cell lines and human liver, heart, hippocampus, pituitary, and thyroid were obtained from M Grompe, Department of Molecular and Medical Genetics, Oregon Health Sci- ences University, and R. Rehfuss, D. Grandy, and R. Cone, Vollum Institue. Human cell lines examined were Molt4B (T cells), 293 (kidney), SY5Y (neuroblastoma), JEG-4 (placenta), LX-1 (lung), HepG2 (liver), SKN (neuroblastoma), HL60 (pre-T cell), SKN-SH (neuroblastoma), MCF70 (breast carcinoma), and NB5 (neuro- blastoma). For PCR studies, first strand cDNA was synthesized from total RNA using oligo(dT) primer and Moloney murine leukemia virus reverse transcriptase. Primers for specific amplification of HVDACl were CAAGTATCAGATTGACCCTGAC and AACTT-TAACCTGGAGGGCTAAC. Primers for detecting HLVDAC 2 were CTTGGACATCAGGTACCAACTG and AACCAGCTAACAAA-GAACTGTC. Amplifications were performed for 50 cycles, with a denaturing temperature of 96 "C for 30 s, annealing at 60 "C for 30 S, and elongation at 72 "C for 30 s. Amplified fragments were separated on 1.2% agarose gels, Southern-blotted, and probed sequentially at high stringency with the entire cDNA encoding HVDACl or HVDACB. Probes were labeled by the random-primed incorporation of [32P]dCTP.

The voltage-dependent anion channel (VDAC) of the outer mitochondrial membrane is a small abundant protein found in all eukaryotic kingdoms which forms a voltage-gated pore when incorporated into planar lipid bilayers. VDAC is also the site of binding of the metabolic enzymes hexokinase and glycerol kinase to the mitochondrion in what may be a significant metabolic regulatory interaction. Recently, there has been speculation that there may be multiple forms of VDAC in mammals which differ in their localization in the outer mitochondrial membrane and in their physiological function. In this report, we describe the identification and characterization of two human cDNAs encoding VDAC homologs (HVDACl and HVDACS). To confirm VDAC function, each human protein has been expressed in yeast lacking the endogenous VDAC gene. Human proteins isolated from yeast mitochondria formed channels with the characteristics expected of VDAC when incorporated into planar lipid bilayers. In addition, expression of the human proteins in such strains can complement phenotypic defects associated with elimination of the endogenous yeast VDAC gene. Since VDAC is the site of binding of hexokinase to the outer mitochondrial membrane, the binding capacity of each VDAC isoform expressed in yeast mitochondria was assessed. When compared with the binding of hexokinase to mitochondria lacking VDAC, the results show that mitochondria expressing HVDACl are capable of specifically binding hexokinase, whereas mitochondria expressing HVDACS only bind hexokinase a t background levels. The expression of each human cDNA has been assessed by Northern blot and polymerase chain reaction techniques. With one exception, each is expressed in all human cell lines and tissues examined.
The voltage-dependent anion channel (VDAC, also known * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. The nucleotide sequencels) reported in this paper has been submitted to the GenBankTM/EMBL Data Bank with accession number(s) LO6132 (H VDAC1) and LO6328 (H VDAC2). ** To whom correspondence should be addressed. Tel.:  as mitochondrial porin)' of the outer mitochondrial membrane is a small (-30 kDa) abundant protein found in all eukaryotic kingdoms which forms a large (-3 nm) voltagegated pore when incorporated into planar lipid bilayers (Colombini, 1989;Adams et al., 1991). Physiologically, VDAC is thought to function as the primary pathway for the movement of adenine nucleotides through the mitochondrial outer membrane, thus controlling the traffic of these essential compounds to and from the mitochondria as well as the entry of other substrates into a variety of metabolic pathways. VDAC has also been shown to be the binding site for hexokinase and glycerol kinase (Felgner et al., 1979;Linden et al., 1982;Fiek et al., 1982;Ostlund et al., 1983;Nakashima et al., 1986). Binding of these enzymes may allow them preferential access to mitochondrial ATP (Fiek et al., 1982;Ostlund et al., 1983;Seltzer and McCabe, 1984;Kaneko et al., 1985), although recently this view has been questioned (Kabir and Nelson, 1991). The association of hexokinase and glycerol kinase with mitochondria is dynamic, varying in different tissues, during development, and with respect to the metabolic state of the cell, suggesting that the association of these enzymes with the outer membrane constitutes a significant metabolic regulatory interaction (reviewed in Adams et al., 1991). Consistent with this notion, binding of these enzymes to VDAC may occur specifically at contact sites between the inner and outer mitochondrial membranes, thus linking cytoplasmic metabolism and ADP production as regulated by these enzymes with the regulation of mitochondrial respiration and oxidative phosphorylation in the mitochondrial matrix (Kottke et al., 1988Brdiczka, 1990). In addition, cells in highly malignant tumors have an increased percentage of mitochondrially bound hexokinase when compared with normal cells (Nakashima et al., 1986), and it appears that VDAC is part of a complex forming the mitochondrial benzodiazepine receptor (McEnery et al., 1992).
Since VDAC is distributed throughout the outer membrane and hexokinase and glycerol kinase may be bound to mitochondria through VDAC only at contact sites, there has been speculation that there may be multiple forms of VDAC in mammals which differ in their localization within the outer mitochondrial membrane and in physiological function (Dorbani et al., 1987;Brdiczka, 1990) What appears to be a single VDAC protein in bovine (DePinto et al., 1987;De Pinto et al., 1991), rat (Roos et al., 1982Colombini, 1983), and human ' The abbreviations used are: VDAC, voltage-dependent anion channel; PCR, polymerase chain reaction: bp, base pair(s); MES, 4morpholineethanesulfonic acid.
1835 (Towbin et al., 1989;Thinnes et al., 1989) tissues has been characterized, but VDAC genes have been cloned only from t h e fungi Saccharomyces cerevisiae (Mihara and Sato, 1985; and Neurospora crassa (Kleene et al., 1987). Comparison of the fungal proteins indicates that the amino acid sequence of VDAC is not highly conserved; the two fungal proteins share only 30% identity, although the two types of channels have very similar, if not identical, electrophysiological characteristics in planar lipid bilayers. Given this low level of sequence conservation, even within fungi, it is likely to be difficult to generate probes for the identification of mammalian VDAC genes from knowledge of the fungal sequences. In order to investigate the diversity of VDAC proteins in mammals and to begin to define the physiological significance of individual isoforms, should they exist, we have attempted to identify human VDAC cDNAs. I n this report, we describe two cDNAs representing the transcripts of two different human VDAC genes. To confirm VDAC function, each protein has been expressed in yeast cells that lack the endogenous yeast VDAC gene. Human proteins isolated from yeast mitochondria formed channels with the expected characteristics of VDAC channels when introduced into planar phospholipid bilayers and can complement phenotypic defects associated with elimination of the endogenous yeast VDAC gene. When expressed in yeast, the two isoforms appear t o differ in their ability to bind rat brain hexokinase. T h e expression of each transcript has also been assessed by Northern blot and PCR techniques. With one exception, both transcripts appear to be expressed in a wide variety of tissues.

(G/A)TCNGG(G/A)TC(T/G/A)AT(T/C)TG
(G/A)TA(T/C)TT (codons L y~* '~-A l a~~~, antisense). The first amplification reactions were performed for five cycles with a denaturing temperature of 96 "C for 30 s, annealing 37 "C for 60 s, and elongation a t 72 "C for 30 s. The products of this reaction were subjected to 45 additional cycles of amplification with a denaturing temperature of 96 "C for 30 s, annealing 50 "C for 60 s, and elongation at 72 "C for 60 s. Amplified fragments were subcloned into pBS(M13-) (Stratagene) and sequenced by standard dideoxy nucleotide chain termination methods using Sequenase (U. S. Biochemical Corp.). Fragments containing HVDACl sequences were used to screen a human pituitary lambda gtl0 cDNA library (provided by D. Grandy, Vollum Institute) a t high stringency. Hybridizations were carried out in 5 X SSC (1 X SSC = 0.15 M NaC1,0.015 M sodium citrate), 50% formamide at 42 "C and filters washed in 0.1 X SSC at 65 "C. Restriction fragments from positive phage were subcloned and sequenced as described above.
HVDAC2 containing cDNAs were identified by low stringency hybridization of a human liver Xgtll cDNA library (provided by N. Kennaway, Oregon Health Sciences University). Duplicate filters were probed with restriction fragments containing the entire coding sequence of HVDACl either at high stringency or at low stringency (hybridization = 5 X SSC, 30% formamide, 37 "C.; washes = 1 X SSC, 50 "C.) Clones that hybridized at low stringency but not at high stringency were identified, subcloned, and sequenced as described above.
Expression of HVDACl and HVDAC2 in Yeast-For expression of human VDAC proteins in yeast, yeast VDAC sequences between the EcoRV and NsiI sites  within the yeast VDAC gene were replaced with corresponding human sequences. Expression in each case was driven by the yeast VDAC promoter, and each protein was expressed as a fusion protein in which the NH, terminus of the human protein was replaced by the first 9 amino acids of yeast VDAC. For HVDAC1, oligonucleotide-directed mutagenesis was used to an create an EcoRV site at codon 10 (C(127) to A and T(129) to C (Fig.  l)), converting this codon from Leu to Ile. The modified HVDACl gene was cut with EcoRV and DraI and inserted into the yeast VDAC gene between the EcoRV and NsiI sites. The resulting fusion gene encodes amino acids 1-9 of yeast VDAC followed by amino acids 11-283 of HVDACl. For HVDAC2, PCR was used to generate an EcoRV site at nucleotide positions 120-125. The resulting PCR product was digested with EcoRV, and the 850-bp fragment was inserted into the yeast VDAC gene between the EcoRV and NsiI sites. This construct encodes amino acids 1-9 of yeast VDAC followed by amino acids 22-294 of HVDAC 2.
Both constructs were inserted into single copy yeast plasmids and introduced into yeast lacking the endogenous yeast VDAC gene (Blachly-Dyson et al., 1990) by transformation. VDAC was prepared from yeast mitochondrial membranes and characterized electrophysiologically following insertion into planar lipid bilayers as described previously (Blachly-Dyson et al., 1990).
Hexokinase Binding-Hexokinase-1 for binding studies was purified from rat brain according to the method of Chou and Wilson cells containing HVDACl, HVDAC2, or lacking VDAC (Blachly-(1972). For binding studies, mitochondria were prepared from yeast Dyson et al., 1990) and suspended in sucrose medium (0.25 M sucrose, 0.01 M Hepes, pH 7.4) to a protein concentration of 2 mg/ml. From this suspension, 0.1-ml aliquots were incubated for 30 min in an ice bath with increasing activity levels of hexokinase in the presence of 10 mM MgC12 and 10 mM glucose. Following incubation, the mitochondria were separated from the supernatant by centrifugation in a Microfuge at 16,000 X g. The mitochondrial pellet was resuspended in 0.02 ml of phosphate buffer (10 mM, pH 7.2) supplemented with 10 mM glucose. Bound hexokinase activity was determined using radioactively labeled glucose (gl~cose-D-[3-~H], 16.8 Ci/mM) by the method of Guggenheim et al. (1980) with the following modifications; 5 el of the resuspended pellet was incubated for 10 min at 37 "C in triethanolamine buffer (50 mM, pH 7.2) containing 6 mM ATP, 8 mM MgCl', and 2 mM glucose (ratio of unlabeled to radioactively labeled glucose = 70:l). The reaction was terminated by the addition of 25 pl of ice-cold ethanol and subsequent centrifugation. A 20-4 aliquot of the supernatant was spotted on DE-81 filter paper, washed sequentially with water and 1 M glucose, and radioactivity counted in a liquid scintillation counter.
Human Expression-Expression of each HVDAC isoform was assessed by Northern blot analysis of total RNA and by PCR amplification of transcripts present in human cell lines and tissues. RNAs prepared from the indicated human cell lines and human liver, heart, hippocampus, pituitary, and thyroid were obtained from M Grompe, Department of Molecular and Medical Genetics, Oregon Health Sciences University, and R. Rehfuss, D. Grandy, and R. Cone, Vollum Institue. Human cell lines examined were Molt4B (T cells), 293 (kidney), SY5Y (neuroblastoma), JEG-4 (placenta), LX-1 (lung), HepG2 (liver), SKN (neuroblastoma), HL60 (pre-T cell), SKN-SH (neuroblastoma), MCF70 (breast carcinoma), and NB5 (neuroblastoma). For PCR studies, first strand cDNA was synthesized from total RNA using oligo(dT) primer and Moloney murine leukemia virus reverse transcriptase. Primers for specific amplification of HVDACl were CAAGTATCAGATTGACCCTGAC and AACTT-TAACCTGGAGGGCTAAC. Primers for detecting HLVDAC 2 were CTTGGACATCAGGTACCAACTG and AACCAGCTAACAAA-GAACTGTC. Amplifications were performed for 50 cycles, with a denaturing temperature of 96 "C for 30 s, annealing at 60 "C for 30 S, and elongation at 72 "C for 30 s. Amplified fragments were separated on 1.2% agarose gels, Southern-blotted, and probed sequentially at high stringency with the entire cDNA encoding HVDACl or HVDACB. Probes were labeled by the random-primed incorporation of [32P]dCTP.

RESULTS AND DISCUSSION
Identification and Characterization of HVDACl and HVDAC2"HVDACl encoding sequences were identified by use of two pairs of nested primers based on the published amino acid sequence of VDAC purified from human B-lymphocytes (Thinnes et al., 1989;Kayser et al., 1989). First strand cDNA was prepared from transcripts expressed in a human B cell hybridoma and amplified in two stages using the nested primers to obtain a -450-bp fragment. The se-quence of this fragment indicated that it contained an open reading frame encoding a polypeptide with a sequence identical to amino acids 83-230 of the published human VDAC protein sequence. This fragment was then used to screen a human pituitary cDNA library. One hybridizing clone (designated HVDACl) was identified and characterized. The sequence of the insert (Fig. 1) contains an open reading frame which encodes a polypeptide identical to the complete published protein sequence of the purified VDAC protein (Thinnes et al., 1989), with the addition of an amino-terminal methionine, the most 5' ATG codon in this cDNA. Since no in-frame stop codons exist in the putative 5"untranslated region, we cannot exclude the possibility that the nascent VDAC protein contains an amino-terminal extension that is subsequently cleaved, although yeast and Neurospora VDAC lack cleaved presequences, as do other mitochondrial outer membrane proteins. The open reading frame is followed by an 854-bp 3"untranslated sequence containing a consensus polyadenylation signal (AATAAA) at nucleotide 1788.
Genomic Southern blots probed with this cDNA at reduced stringency indicated that related VDAC sequences were likely to be present in humans (data not shown). To identify additional members of the family, a 750-bp Hind111 fragment of HVDACl containing most of the coding region was used to probe a human liver X cDNA library at both high and low stringency. Five clones that hybridized at low, but not at high, stringency were identified and characterized. The restriction map and nucleotide sequence of the ends of these clones indicated that they contained overlapping 5' sequences. Four of the clones were identical at the 3' end, each ending in a poly(A) sequence at position 1235, although no consensus (AATAAA) site is present. The fifth clone contained an additional 168 nucleotides, including a consensus polyadenylation site at position 1387 followed by multiple A residues.
MET C y s 110 P r o P r o Ser Tyr N s A s p Leu G l y L y a V a l N a A r g A s p Ile P h e A s n L y s G l y P h e G l y P h e G l y L e u V a l L y s Leu A s p V a l L y s

of Outer Mitochondrial Membrane Channel
The identification of two classes of cDNA differing in the extent of 3"untranslated sequences suggests the use of alternate polyadenylation sites. A combined cDNA sequence (HVDAC2) representing the longest 5' and 3' sequences is shown in Fig. 1. This sequence contains a single long open reading frame that extends from nucleotide 40 to nucleotide 947. Within this open reading frame, two ATGs are found near the 5' end, one at nucleotide 63 and a second at position 96. The ATG at position 63 shows a good (8/9) match with the start codon consensus sequence ( C C A / G C C E ( G ) ) consistent with the proposition that this is the translational start site, although the second ATG corresponds to the position of the proposed start codon in HVDACl and the start codons of the N . crassa and S. cerevisiae genes (Fig. 2).
Expression of HVDACs in Yeast and Physiological Characterization-One remarkable characteristic of VDAC channels is the extreme evolutionary conservation of basic channel properties. VDAC channels from yeast to mammals share similar (a) single channel conductances (roughly 4.5 nano-Siemens in 1 M KCl), ( b ) ion selectivity (about 2:l preference for chloride over potassium), and (c) voltage sensitivity as measured by the steepness, n, of the voltage-dependent conductance changes (Benz, 1985;Colombini, 1989). In addition, all VDAC channels are symmetrical with respect to gating properties, i.e. VDAC channels are in their high conducting or "open state" at 0 mV and "close" to low conducting states in response to the application of both positive and negative potentials. In order to confirm that each putative human VDAC cDNA encodes a protein with the characteristics expected of VDAC, human VDAC cDNA constructs were generated containing the promoter, 5"untranslated region, and first nine codons of the yeast VDAC gene fused to HVDACl at codon 11 and HVDAC2 at codon 22 (see "Materials and Methods"). These constructs were introduced into yeast cells lacking the endogenous yeast VDAC gene, mitochondria from the resulting transformants were isolated, and VDAC was purified from the mitochondrial membranes (Blachly-Dyson et al., 1990). Purified samples contained a single protein band of appropriate molecular weight as assessed by silver-stained SDS-polyacrylamide gels (data not shown). These proteins were then introduced into synthetic phospholipid bilayers and their electrophysiological properties tested (Blachly-Dyson et al., 1990).
As shown in Table I, the both human VDAC cDNAs when expressed in yeast form channels on reconstitution whose properties are characteristic of VDAC from other sources (i.e. single-channel conductance and selectivity). There were, however, qualitative differences between HVDACl and HVDAC2 channels. HVDACl produced typical VDAC channels with characteristic steepness of voltage dependence ( n values) of 2.2 f 0.4 (positive potentials, three determinations) and 2.4   , Blachly-Dyson et al. (1990), or Peng et al. (1992. Selectivities calculated from reversal potential values according to the Goldman/Hodgkin/Katz equation.

Yeast
Single-channel con-4.1 ? 0.1 (6)  Methods"). The resulting strains were then streaked on media containing 2% glycerol as the sole carbon source and incubated at the indicated temperatures. f 0.3 (negative potentials, three determinations; Fig. 3). In contrast, HVDACZ channels often inserted as low conducting events suggestive of subconductance states. With time, channels with properties characteristic of VDAC open channels were observed with n values of 2.5-3.0 at both positive and negative potentials (two determinations; Fig. 3). These n values for HVDACl and HVDACZ compare well with those obtained for yeast VDAC (2.5 k 0.3, positive potentials, seven determinations; 2.4 f 0.5 negative potentials, seven determinations). However, after 20 min to 1 h of testing, HVDACZ channels appeared to lose their voltage dependence, whereas the voltage dependence of HVDACl channels persisted. Preliminary results suggest that expression of native HVDAC2 cDNAs (i.e. lacking NH2-terminal yeast sequences) in yeast results in the production of a very low level of channel activity with more stable voltage dependence. Thus, the instability of voltage dependence observed for HVDACZ channels may be an artifact produced by excision of native NH2-terminal domains and/or fusion of yeast NH2-terminal sequences.
Deletion of the VDAC gene from yeast leaves the cells viable, but results in temperature-sensitive growth on glycerol-based media Dihanich et al., 1987;Dihanich 1990).2 The physiological basis of the ability of strains lacking VDAC to conditionally utilize nonfermentable carbon sources is currently unknown. The exact phenotype resulting from elimination of the VDAC gene appears to be highly strain-dependent,2 and in some situations, a new channel has been reported to be present in the outer membrane that can functionally compensate for the lack of VDAC at low temperature (Dihanich et al., 1989: Michejda et al., 1989. In all cases, growth defects are corrected by reintroduction of a plasmid-based yeast VDAC gene. As shown in Fig.   4, introduction of the HVDACl or HVDACZ constructs described above also complements the temperature-sensitive growth defect resulting from elimination of the endogenous VDAC gene. Thus, the human proteins we have identified not only produce channels with the characteristics observed for all VDAC channels, but the human genes can also functionally * E. Blachly-Dyson and M. Forte, unpublished data.
replace the yeast gene in yeast cells. We are confident then that these cDNAs encode human VDAC proteins.
One conserved characteristic of VDAC is its ability to provide a binding site for hexokinase to the outer mitochondrial membrane. T o determine if the two human isoforms differ in their ability to bind this molecule, mitochodria prepared from yeast strains expressing HVDACl, HVDAC2, or lacking any VDAC were compared. As shown in Fig. 5 , the two isoforms differ in their ability to bind rat brain hexokinase; HVDACl binds hexokinase, whereas HVDAC2 only binds to the background levels observed in mitochondria lacking VDAC. This background level rat brain hexokinase binding is similar to the binding of mitochondria containing wild-type yeast VDAC (data not shown), presumably reflecting the species dependence of binding, since yeast VDAC has been demonstrated to bind yeast hexokinase (Forte et dl., 1987). The observation that the two VDAC isoforms differ in their ability to bind hexokinase has clear implications for mechanisms of respiratory control involving the binding of this enzyme to the outer membrane, thereby allowing it preferential access (Adams et al., 1991) to mitochondrially generated ATP.
Expression of Genes Encoding HVDAC Isoforms-The potential for differential expression of HVDACl and HVDAC2 in a variety of human tissues and cell types was assessed by Northern blot analysis and specific PCR amplification. By Northern blot, both a 2-kilobase pair HVDACl transcript and a 1.3-kilobase pair HVDACZ transcript are present a t varying levels in three of the four cell lines tested and in RNA prepared from human thyroid (Fig. 6, A and B ) . Two
HVDAC2 transcripts differing by 167 bp are expected due to alternate polyadenylation. If both are present, they are not resolved in this gel. HVDACl and HVDAC2 transcripts appear to be absent in the NB5 human neuroblastoma cell line. For PCR amplification, first strand cDNA was prepared from RNA isolated from a variety of tissues and cell lines and amplified using pairs of primers specific for the two isoforms (Fig. 7, A and B ) . HVDAC2 was detected as a 393-nucleotide amplification product in all tissues and cell types tested, including NB5 cells, where the level was likely to be too low to be detected by the less sensitive Northern blot technique. HVDAC 1 was also detected by PCR in most tissues and cell types, with the exception of NB5 cells.
Sequence Comparisons-Alignment of the amino acid sequences encoded by the two human clones indicates that they are identical a t 211/283 positions (75%), with no introduction of gaps and with an 11-amino acid amino-terminal extension in HVDAC2 relative to HVDACl. The majority of the amino acid differences between HVDACl and HVDAC2 are conservative substitutions (e.g. Ser for Thr, Lys for Arg). The most dramatic difference between the two sequences is the presence of 2 acidic residues (G1u2" and Asp2") in HVDAC2 at the positions corresponding to two adjacent basic residues (Lys'" and Lys20L) in HVDACl. In addition, there are seven positions where one sequence contains a charged residue and the other an uncharged one (E36/C47, R93/Q104, H122/C133, Q166/R177, D176/G187, D230/T241, and Q282/E293).
Models of the VDAC channel have been developed based on the fungal sequences and functional studies. VDAC sequences contain many stretches of alternating hydrophobic and hydrophilic residues, which could form sheets with hydrophobic residues on one side and hydrophilic residues on the other. In the models, such a / 3 sheet forms a cylinder in the membrane with the hydrophobic side facing the lipid bilayer and the hydrophilic side forming the lining of the aqueous channel Blachly-Dyson et al., 1990;De Pinto et al., 1991;Mannella et al., 1992). Fig. 8 shows an evaluation of the potential of regions of the yeast and human VDAC proteins to form transmembrane p strands (Blachly-Dyson et al., 1989). In spite of the low level of sequence identity between human and fungal VDAC sequences, the overall pattern of putative transmembrane, sided p sheets, is preserved, suggesting that the human proteins can assume a conformation similar to that proposed for the yeast VDAC channel.
Current models of the yeast VDAC transmembrane topology have been further refined by defining protein domains forming the walls of the open VDAC channel based on the location of residues that alter the selectivity of the channel when their charge is changed by site-directed mutagenesis (Blachly-Dyson et al., 1990, Peng et al., 1992. When human VDAC sequences are aligned with the yeast sequences as in Fig. 2 and arranged into one NH2-terminal transmembrane CY helix and 13 transmembrane p strands, as proposed for the yeast molecule, the human channels contain a striking excess of negatively charged residues in putative transmembrane regions and an excess of positively charged residues in loop regions when compared with the yeast protein. If the selectivity of VDAC channels is due to the overall charge in the pore as determined by residues lining the wall of the channel, consistent with studies on the yeast VDAC, then the human channels should be decidedly cation selective. Cation-selective yeast VDAC channels can be created by site-directed mutagenesis of residues that are proposed to line the transmembrane pore. However, as documented in Table I, human VDAC appears to form channels in planar bilayers whose selectivity is indistinguishable from the yeast channel. This observation suggests two alternatives. First, it is possible that the models developed for yeast VDAC do not apply to the human counterparts. As demonstrated in Fig. 8, however, the overall pattern of potential transmembrane / 3 strands in the yeast and human sequences is very similar, suggesting that they share similar overall structures. More likely then is the possibility that the exact position of one or more transmembrane strands, relative to the linear protein sequence, differs between yeast and human VDACs, such that regions that are aligned (in Fig. 2) with yeast loop domains form transmembrane strands in the human sequence. In the human channel, these "loop" regions contain excess positive charge and thus could balance the excess negative charge found in other transmembrane regions. For example, residues 104-122 of the human sequence (HVDAC1) contain 5 additional positively charged residues and a stronger overall transmembrane sided fl sheet pattern (Fig. 7) than the corresponding region of the yeast protein. It should be possible to test these alternatives by the construction of chimeric human/yeast VDAC channels and by site-directed mutagenesis of specific residues of human VDAC genes expressed in yeast.
Immunocytochemical studies using antibodies directed to the NH2-terminal 19 residues of HVDACl have indicated that antibody cross-reacting molecules are present in the plasma membrane as well as the mitochondria (Thinnes et al., 1989;Babel et al., 1991). The identification of two human VDAC genes which encode proteins that are identical in 13 positions over this region raises the possibility that such antibodies recognize both isoforms. Thus, although both proteins can be directed to the mitochondria in yeast, in human cells one form may also be preferentially expressed at the plasma membrane. Consistent with this notion, numerous reports have documented the presence of plasma membrane channels that have physiological characteristics similar to VDAC (Blatz and Magleby, 1983). Determination of the subcellular localization of each isoform awaits the development of reagents (i.e. antibodies) that specifically identify each isoform.