The M r 18,000 Subunit of the Peripheral-type Benzodiazepine Receptor Exhibits Both Benzodiazepine and Isoquinoline Carboxamide Binding Sites in the Absence of the Voltage-dependent Anion Channel or of the Adenine Nucleotide Carrier*

The peripheral type benzodiazepine receptor (PBR) binds benzodiazepines such as RO5-4864 and isoquinoline carboxamide derivatives such as PK11195. This receptor includes anM r 18,000 isoquinoline-binding subunit predominantly located in mitochondrial mem- branes. This protein has been found to copurify with two other mitochondrial proteins, namely the outer membrane voltage-dependent anion channel (VDAC), also known as mitochondrial porin, and the inner membrane adenine nucleotide carrier. In vitro reconstitution experiments suggested that the PBR was a multimeric complex in which the isoquinoline binding site was on the M r18,000 subunit, denoted pk18, whereas the benzodiazepine binding site required the association of this subunit with VDAC to be expressed. Untransformed cells of the yeast Saccharomyces cerevisiaeare devoid of specific binding sites for isoquinolines and benzodiazepines, whereas yeast cells transformed with a pk18-expressing vector exhibit RO5-4864 and PK11195 binding sites that are pharmacologically identical to those of the PBR. To clarify the role of VDAC and of the adenine nucleotide carrier, if any, in the constitution of the benzodiazepine binding site, yeast host strains were constructed in which the corresponding genes had been knocked out. Mitochondria prepared from pk18-producing cells devoid of either VDAC or adenine nucleotide carrier exhibit both benzodiazepine and isoquinoline carboxamide binding sites with little or no change in theK d values as compared with the wild-type background. These results rule out the contention that VDAC is indispensable for establishing the benzodiazepine binding site and are in agreement with the hypothesis that the M r18,000 subunit carries both the isoquinoline carboxamide and benzodiazepine binding domains.

Benzodiazepine (Bz) 1 binding sites generally fall into two main structurally distinct types. The first is found on the ␥-aminobutyric acid A receptor, the location of which is restricted to the central nervous system. The second is found in peripheral and in central tissues and is widely known as the peripheral-type benzodiazepine receptor (PBR) (1,2). The PBR, predominantly located in mitochondrial membranes, displays pharmacological characteristics distinct from the central type. Isoquinoline carboxamide (IQ) derivatives like PK11195 that are high affinity ligands of PBR are ineffective in binding to the central type receptor. The subset of Bz molecules that binds to each type of receptor is also markedly different. For instance, RO5-4864 is a Bz that binds to PBR, at least in human and rat (3) and also to the central receptor, but with a much lower affinity. Purification of rat kidney mitochondrial PBR in a form that retained reversible binding properties yielded a protein fraction composed of three polypeptides of M r 18,000, 30,000, and 32,000. The two latter subunits were identified as the adenine nucleotide carrier (ADC) and the voltage-dependent anion channel (VDAC), respectively (4). Bzs like [ 3 H]flunitrazepam and [ 3 H]AHN-086 were shown to photolabel the M r 30,000 and M r 32,000 polypeptides nonspecifically (5, 6), whereas [ 3 H]PK14104, a fluorinated IQ derivative, specifically photolabeled the M r 18,000 subunit, denoted pk18 (7). It was postulated that the benzodiazepine binding site resided on ADC and/or VDAC, while pk18 carried the IQ binding site (8). Recently, Garnier et al. (9) showed that a purified recombinant construct consisting of pk18 fused to the Escherichia coli maltose-binding protein (MBP) bound the IQ molecule PK11195 but not RO5-4864, a ligand of the Bz family, in an in vitro reconstruction system. RO5-4864 binding sites were restored by adding purified VDAC to the hybrid MBP-pk18 construct, whereas VDAC by itself did not show any Bz or IQ binding. The authors concluded that a functional interaction of pk18 with VDAC was necessary to form the Bz binding site. They suggested that the Bz recognition site was constituted by an association of pk18 and VDAC components or, alternatively, that the site might be situated on either pk18 or VDAC and conformational changes induced by the partner was required to make it accessible for Bz binding.
Complementary DNA encoding pk18 has been isolated from various species (9 -11) including Homo sapiens (12). The deduced amino acid sequence of the human clone corresponds to a polypeptide of 169 residues that displays five putative transmembrane regions. Wild-type bakers' yeast, Saccharomyces cerevisiae, which is devoid of pk18, presents no specific binding sites for IQ and Bz. Recombinant human pk18 was produced in yeast (13)(14)(15)(16), and mitochondria were extracted from the yeast cells presented both IQ and Bz binding sites that were pharmacologically identical with PBR expressed in animal cells (13). [ 3 H]PK14105 photolabeling of mitochondrial proteins extracted from these recombinant cells revealed a band corre-sponding to an M r 30,000 polypeptide which was identified as the yeast VDAC1 gene-encoded mitochondrial porin. By means of a series of deletion and site-directed mutagenesis experiments it was demonstrated that a particular residue of the pk18 sequence, namely Val 154 , was crucial for the expression of the Bz ligand RO5-4864 binding site (16). A single change M154V mutation was sufficient to confer RO5-4864 binding capability to the bovine version of pk18 which was otherwise unable to bind this ligand (16). These results indicated that recombinant pk18 was not only involved in, but also absolutely required for, Bz binding. In this report, we address the corollary question of whether VDAC, ADC, or both, are necessary for forming the Bz binding site with pk18. This issue can be addressed in S. cerevisiae since, in this microorganism, mutants devoid of functional VDAC or ADC remain able to proliferate under certain conditions. It is therefore possible to knock out the VDAC-and ADC-encoding genes in a pk18-producing strain, and examine the effects, if any, of these genetic alterations on Bz and IQ binding. Two VDAC isoforms exist in yeast. They are synthesized from two distinct genes, namely VDAC1 (17) and VDAC2 (GenBank TM accession no. P40478). 2 ADC isoforms are encoded by the genes AAC1, AAC2, and AAC3. The AAC2 gene product appears to be the more abundant ADC isoform in cells grown under aerobic conditions (18,19). In this report, we describe Bz and IQ binding experiments performed on mitochondrial extracts prepared from cells that produce pk18 in the absence of VDAC and ADC.
Expression Plasmids-Plasmids pEMR1043 and pEMR1130 are yeast vectors that express human PBR. Both plasmids derive from pEMR971 whose construction and structure were described by Joseph-Liauzun et al. (15). pEMR1043, like pEMR971, is a 2-m-based replicating vector that contains different elements as follows: an oriT sequence required for the in vivo DNA transfer by trans-kingdom conjugation between yeast and E. coli cells, a PBR expression cassette including the galactose-regulated GRAP1 promoter and the PGK terminator (24) and finally URA3 as the selectable marker. pEMR1043 differs from pEMR971 only by the length of the oriT-encompassing fragment which was about 200 bp shorter in the former plasmid. pEMR1130 differs from pEMR1043 only by replacing TRP1 by URA3 as the selectable marker. Plasmid pEMR1179 is a low copy centromeric vector which derives from pFL39 (25). Plasmid pEMR1179 contains the neomycin gene of Tn903 which confers G418 resistance (250 -500 g/ ml), and the AAC2 gene expressed under the control of the PGK promoter.
Yeast Transformation-Two methods were used to transform yeast strains. One was by electroporation essentially as described by Becker and Guarente (26). The other method was by trans-kingdom conjugation experiments (27) as follows. S17-1 cells were transformed to ampicillin resistance using DNA of an oriT-containing yeast/E. coli shuttle plasmid. A transformed colony was used as a source of donor cells, cells were grown in ampicillin-containing LB and harvested in mid-exponential phase of growth. Cells were then resuspended in 10 mM MgSO 4 at a density of 10 9 /ml. Yeast recipient cells were grown overnight in YPG, pelleted, and resuspended in 10 mM MgSO 4 at a density of 10 8 /ml. Donor and recipient cells were mixed in equivalent amounts, the mixture was then poured onto a filter placed on non-selective YPG-agar medium and left to incubate at 30°C for 4 -16 h. The cells were then resuspended in 10 mM MgSO 4 and an aliquot was spread onto selective YNBG medium at 30°C. Ura ϩ or Trp ϩ exconjugant colonies appeared 3 days later at a frequency of approximately 10 Ϫ5 per recipient cell.
Receptor Expression and Mitochondria Isolation-Precultures and cultures for PBR expression experiments were done as follows. Precultures were performed in minimal YNBG medium supplemented so as to fulfill the auxotrophic requirements of the strain. Cells were then harvested and resuspended in complex YP medium containing 2% glycerol, 2% galactose, and 1% ethanol. In the cases where mutants did not grow on glycerol and ethanol, these were replaced by 2% raffinose. Mitochondria were isolated as already described by Farges et al. (14). The mixtures were placed in 1.7-cm 2 cell culture dishes, and samples were exposed to UV light (365 nm) for 15 or 60 min, depending on the photolabeling ligand (IQ or Bz, respectively) and analyzed by autoradiographically revealed SDS-polyacrylamide gel electrophoresis. , no other addition (total binding) or the unlabeled ligand at the desired concentration or at 10 mM (nonspecific binding). Each sample was incubated for 1 h at 0°C, and then rapidly filtered through Whatman GF/C filters pretreated with 0.3% polyethylenimine and washed 3 times with 5 ml of ice-cold Tris/HCl buffer. The filters were then dried and suspended in 5 ml of Beckerman Ready Solv. E.P. mixtures for determination of filter bound radioactivity.

RESULTS
Photolabeling of PBR in vdac Ϫ and aac Ϫ Mutants-A photoactivable Bz ligand, namely [ 3 H]flunitrazepam, was used to label mitochondrial proteins extracted from wild-type S. cerevisiae cells. Several bands of variable intensities were revealed (Fig. 1). The two major bands apparently corresponded to M r 30,000 and M r 32,000 proteins. The photolabeled M r 30,000 protein was absent from preparations from vdac mutant strains and probably corresponded to VDAC. Similarly, the M r 32,000 protein was absent from extracts from the adenine nucleotide carrier-defective strains and probably corresponded to the AAC2 gene product. Labeling of both proteins was not inhibited by the addition of excess cold flunitrazepam and could consequently be regarded as nonspecific. Extracts of pEMR1043-transformed cells exhibited an additional band faintly labeled by [ 3 H]flunitrazepam in a reversible manner that apparently corresponded to a M r 18,000 polypeptide. This band, likely that of pk18, was also found in pk18-producing vdac cells devoid of the major VDAC, but was hardly detectable in extracts from pk18-producing cells that were devoid of all three AAC gene products. This result confirmed that pk18 was specifically labeled by [ 3 H]flunitrazepam, albeit poorly.
[ 3 H]PK14105 nonspecifically labeled a M r 30,000 polypeptide (Fig. 2). This photolabeled protein was not detected in vdac mutant extracts, suggesting that it was yeast VDAC. Gel digestion and sequence analyses of this photolabeled protein confirmed its identity as the VDAC1 gene product (not shown). pk18-Producing cells exhibited a strongly labeled M r 18,000 polypeptide. Labeling was inhibited by cold PK14105, as expected. This result confirmed that pk18 was specifically labeled by [ 3 H]PK14105, as already described (5,13).

Binding of RO5-4864 and PK11195 in Mitochondria-enriched Fractions of VDAC-lacking Cell Extracts-RO5-4864
and PK11-195 binding analyses were performed on mitochondrial extracts from various pk18-expressing yeast strains, some of which were defective in ADC and VDAC (Fig. 3,A and B, and Table I). The fact that no VDAC isoform was present had no influence on the K D values estimated for both ligands. An approximately two-fold decrease in the maximal binding capacity for both ligands could be observed in the pk18-expressing cells that were devoid of any VDAC gene product, namely EL302 pEMR1130 (Table I), as compared with W303 cells transformed by the same plasmid (Table I). Such a decrease in pk18 expression was confirmed by Western blot experiments (not shown). This might simply reflect a difference in plasmid copy number between the strains, since this parameter can vary from clone to clone in a host cell genotype-independent manner (24,28).  (Fig. 3, A and B, and Table I) as compared with an ADC-producing strain. A slight decrease in RO5-4864 affinity could be suspected in cells devoid of the three isoforms, this decrease could, at least partially, be corrected by expressing the AAC2 gene. Thus, even though yeast ADC was clearly not required to form the binding sites of PK11195 and RO5-4864, it might indirectly enhance the affinity of pk18 for the Bz ligand.

Binding of RO5-4864 and PK11195 in Mitochondria-enriched Fractions of ADC-lacking Cell Extracts-Strain
Binding of RO5-4864 and PK11195 to PBR Produced in the  1 and 2; recombinant-pk18-producing wild-type cells, namely EMY761 pEMR1043, lanes 3 and 4; recombinant-pk18-producing vdac1 mutant, namely EL301 pEMR1030; lanes 5 and 6, recombinant pk18 and AAC2p-producing triple aac mutant, namely EL105 pEMR1030 and pEMR 1179 (lanes 7 and 8); recombinant pk18-producing triple aac mutant EL105 pEMR1030 (lanes 9 and 10). 2. [ 3 H]PK14105 photolabeling of mitochondria from wild type and mutant yeast cells. Mitochondria were isolated from various strains and photolabeled with [3H]PK14195 in the absence (lane 1 and even numbers) or in the presence of unlabeled PK11195 (see "Experimental Procedure"). Strains were as follows: wild-type control, lane 1; pk18-producing wild-type strain, namely M3 pEMR971 (lanes 2 and  3); pk18-producing vdac mutant, namely M22 pEMR971, (lanes 4 and 5); pk18-producing aac mutant, namely EL105 pEMR1130 (lanes 6 and 7) and pk18 and Aac2p producing aac mutant, namely EL105 pEMR1130 ϩ pEMR1179 (lanes 8 and 9). Absence of Cytochrome Oxydase Subunit IV-Similar experiments were performed using a strain that was devoid of any cytochrome oxidase subunit IV, a mitochondrial polypeptide supposedly not linked to PBR. We checked whether the disruption of the COX4 gene that encoded this cytochrome oxidase subunit could affect the affinity of PBR for both types of ligand. As shown on Table I, mitochondria extracted from a pk18producing cox4 gene disruptant exhibited PK11195 and RO5-4864 binding with normal K D values, as expected.
Competition Experiments-We checked whether PK11195 binding to PBR could be inhibited by RO5-4864, and reciprocally, in the absence as in the presence of VDAC or AAC. No influence of VDAC or AAC on the binding inhibition could be detected (Tables II and III). DISCUSSION Photolabeling of PBR using [ 3 H]PK14105 has revealed a polypeptide of M r 18,000 in a variety of mammalian cells, as well as in recombinant pk18-expressing yeast cells. Transfection experiments with pk18-encoding cDNA in mammalian cell lines result in increased numbers of both IQ and Bz binding sites. Similarly, pk18-expressing vectors confer both IQ and Bz binding capability to yeast transformants (13). These expression experiments strongly suggest that pk18 by itself contains both the IQ and Bz binding domains. However, there have been several lines of evidence that suggest the association of pk18 with other proteins, including VDAC and ADC. First, mammalian proteins of M r similar to VDAC and ADC could be photolabeled with [ 3 H]PK14105 and various photoactivable Bz ligands (1,5). However, photolabeling appears to be nonspecific. Second, VDAC and ADC were observed to copurify with pk18 (4) and a VDAC-like protein was shown to be associated to the ␥-aminobutyric acid A receptor when copurified on a benzodiazepine affinity column (29). These last results suggested a relationship between VDAC and Bz receptors. The fact that VDAC could be labeled by flunitrazepam and that specific reagents that inhibit VDAC function were able to inhibit PBR ligand binding suggested that the Bz binding domain of PBR was on VDAC (8). Further evidence suggesting a role of VDAC in the formation of the benzodiazepine binding site of PBR came from studies using a recombinant chimeric protein consisting of the maltose-binding protein fused to pk18 (MBP-pk18). Purified MBP-pk18 failed to exhibit any R05-4864 binding sites in a liposome-based reconstitution system, whereas the addition of VDAC restored binding. It is worth noting that purified VDAC by itself did not exhibit any IQ or Bz binding sites, which suggested that Bz binding required the interaction of MBP-pk18 with VDAC in this reconstitution system. In yeast, as already found in mammals, both VDAC and ADC are nonspecifically labeled by irreversible Bzs. However, these proteins are labeled even in wild-type cells which lack pk18 and which do not express any specific IQ and Bz binding sites. It is therefore probable that photolabeling of VDAC and AAC by Bz ligands is irrelevant to PBR.
Recombinant bakers' yeast constitutes an attractive model for elucidating whether VDAC, ADC, or other proteins are indeed required to form the Bz binding site of PBR in addition to pk18. Yeast wild-type cells contain various mitochondrial isoforms of VDAC and ADC but are devoid of pk18 and do not present any specific IQ and Bz binding sites. pk18 can be efficiently produced in yeast cells (13,15), where it is targeted to mitochondrial membranes (30). pk18 production in yeast results in the formation of IQ and Bz sites that are pharmacologically similar to authentic PBR. In yeast, mitochondrial functions are dispensable for proliferation, which enables the isolation of mutant strains completely devoid of VDAC or ADC.
The fact that such pk18-expressing cells lacking VDAC exhibit RO5-4864 binding sites clearly demonstrates that VDAC and ADC are dispensable for the formation of the Bz-specific binding site. Furthermore, a genetic variation affecting a single amino acid residue in the pk18 sequence has been previously shown to be sufficient for restoring or abolishing Bz binding expression in recombinant yeast cells. It is therefore highly probable that both the Bz and IQ binding sites are on the same protein, i.e. pk18. These two binding sites are nevertheless not fully overlapping, as revealed by various chemical, genetic and physiological studies (14,16,(31)(32)(33). Our results with VDAClacking cells do not support one of the hypotheses proposed by Garnier et al. (9), namely that the Bz recognition site might be partly on pk18 and partly on VDAC. Garnier et al. based their hypothesis on the observation that in in vitro reconstruction experiments the purified VDAC and the purified chimeric construct MBP-pk18 did not exhibit any Bz binding site unless mixed together. That result, in the light of the work we present here, might be interpreted in a different manner. The pk18 domain of the chimeric construct might have adopted a conformation incompatible with Bz-binding, either as a consequence of the fusion with the bacterial protein, or as a result of a partial denaturation event occurring at some step during the purification procedure. The addition of VDAC may have corrected the conformational defect by interacting with the pk18 domain; VDAC would then play a chaperone-like role in this reconstituted system. Quite similarly, a pk18-conformationmodulating role could possibly be proposed for yeast ADC in our system, since in the absence of the major ADC isoform, the K D and K i values estimated for the Bz ligand tend to slightly increase and the flunitrazepam-photolabeled M r 18,000 polypeptide is hardly detectable. Thus, our results do not rule out the alternative hypothesis of Garnier et al. postulating that proteins distinct from pk18 might indirectly modulate the Bz binding site on PBR (i.e. pk18, according to our results) by interacting with this latter polypeptide. The Bz binding domain might correspond to a more restrictive conformational state of the pk18 homopolymer, as compared with the IQ binding domain, this Bz-binding-competent conformation might be stabilized or modulated by other proteins interacting with this complex. Recent results have shown that modifications to the PBR topography are induced upon addition of choriogonadotropin to MA10-Leydig cells, resulting in an increase in both the pk18 cluster size and PBR ligand binding (34). This observation confirms that PBR receptors can exist under distinct morphological states. However, since VDAC, whether from yeast or mammals, is dispensable for the Bz binding site of pk18 produced in yeast cells, the conclusion that the Bz binding site of PBR resides, even partly, on VDAC cannot stand. On the con-trary, our results support the assumption that both the IQ and Bz binding sites are fully contained within the pk18 sequence.