Topological Analysis of the Peripheral Benzodiazepine Receptor in Yeast Mitochondrial Membranes Supports a Five-transmembrane Structure*

The peripheral benzodiazepine receptor, implicated in the transport of cholesterol from the outer to the inner mitochondrial membrane, is predicted by hydropathy analysis to feature five membrane-spanning domains, with the amino terminus within the mitochondrial periplasm and the carboxyl terminus in the external cytoplasm. We have tested these structural predictions directly by immunodetection of c-Myc-tagged peripheral benzodiazepine receptor on intact yeast mitochondria and by specific labeling in yeast membranes of cysteine residues introduced by site-directed mutagenesis. The combined results support the model originally proposed with some minor but important modifications. The theoretical model predicted relatively short α-helical domains, only long enough to span a phospholipid monolayer, whereas the results presented here would support a model with extended α-helices sufficiently long to span an entire membrane bilayer, with concomitant shorter loop and tail regions.

The peripheral benzodiazepine receptor, implicated in the transport of cholesterol from the outer to the inner mitochondrial membrane, is predicted by hydropathy analysis to feature five membrane-spanning domains, with the amino terminus within the mitochondrial periplasm and the carboxyl terminus in the external cytoplasm. We have tested these structural predictions directly by immunodetection of c-Myc-tagged peripheral benzodiazepine receptor on intact yeast mitochondria and by specific labeling in yeast membranes of cysteine residues introduced by site-directed mutagenesis. The combined results support the model originally proposed with some minor but important modifications. The theoretical model predicted relatively short ␣-helical domains, only long enough to span a phospholipid monolayer, whereas the results presented here would support a model with extended ␣-helices sufficiently long to span an entire membrane bilayer, with concomitant shorter loop and tail regions.
The peripheral (or mitochondrial) benzodiazepine receptor (PBR) 1 possesses a benzodiazepine binding site that is clearly distinct from the modulator site of the neurotransmitter ␥-aminobutyric acid receptor. PBR is present in most, if not all, tissues and is particularly abundant in the outer membrane of mitochondria. PBR has been suggested to be required for the transport of cholesterol from the outer to the inner mitochondrial membrane where steroid biosynthesis takes place (1). The receptor also appears to play a key role in modulating mitochondrial electrophysiology, which suggests its implication in the side effects of benzodiazepine pharmacology (for recent review, see Ref. 2). An outer membrane sensory protein of the proteobacterium Rhodobacter sphaeroides has recently been shown (3) to have a close structural and functional relationship with the PBR, supporting the hypothesis that mammalian mitochondria are of photosynthetic bacterial origin. The identification of an 18-kDa protein in human tissues (4) followed by the isolation of the corresponding cDNA (5) allowed us to produce recombinant human PBR in Saccharyomyces cerevisiae (6,7), an organism normally devoid of binding sites for PBR li-gands, thus opening up new avenues for the study of PBR structure-activity relationships.
Hydrophobicity analysis of the amino acid sequences of the rat (8) (rPBR), murine (9) (mPBR), bovine (10) (bPBR), and human (5) (hPBR), together with the positive inside rule (11) led to a two-dimensional membrane topological model comprising an intramitochondrial short amino-terminal region and five putative amphipathic ␣-helices linked by hydrophilic loops leading to an extramitochondrial carboxyl-terminal tail (12). A similar pentahelix topology has been found at 2.2-Å resolution for a crystallized apolipoprotein (13) and from gene fusion studies of the cytochrome c terminal oxidase complex of Escherichia coli (14). A three-dimensional model for PBR was proposed and studied using molecular dynamics simulations (12). It was concluded from this model that the ␣-helices were too short to cross an entire bilayer membrane but corresponded approximately to one phospholipid layer. The work described here was aimed at determining whether the PBR model was correct or needed to be refined.
In view of the paucity of crystallographic data for membrane proteins several techniques have been developed to investigate their topology. Each of these techniques relies on the localization of the hydrophilic loop and tail regions by proteolytic degradation, by antibiotic selection, or by various labeling strategies. In the present report we describe two quite distinct labeling approaches we have adopted to investigate the PBR model. The first made use of c-Myc epitope insertions with subsequent immunodetection; in the various loop regions the epitope was flanked by pentaglycine to improve antibody accessibility, an innovation used successfully for topological studies of bacteriorhodopsin (15). In the second method, designed to obtain a more refined topological analysis, we mutated the PBR to introduce cysteine residues into various regions that were subsequently visualized by sulfhydryl labeling techniques (16). From the results obtained we were indeed able to confirm the pentahelical PBR structure; but by taking into consideration other published investigations, we propose a model in which the transmembrane helices are longer than those previously thought to exist.
Construction of c-Myc-tagged PBR-Fusions of the human PBR with the c-Myc epitope were carried out by the overlap extension polymerase chain reaction method (19) using the hot-start procedure. Taking c-Myc c1 as a typical example, a sense primer situated on the expression * 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.
§ To whom correspondence should be addressed. Tel.: 33-5-6100-4064; Fax: 33-5-6100-4001; E-mail: evelyne.liauzun@tls1.elfsanofi.fr. 1 The abbreviations used are: PBR, peripheral benzodiazepine receptor; r, m, b, and hPBR, rat, murine, bovine, and human PBR, respectively; ELISA, enzyme-linked immunosorbent assay; PBS, phosphatebuffered saline; 5FM, fluorescein-5-maleimide; BM, benzophenone-4maleimide; c and m, cytoplasmic and intramitochondrial locations, respectively. vector pEMR971 upstream from the PBR initiation codon, 5Ј-AATA-GAGTGCCAGTAGCGAC (400 ng) and antisense primer 5Ј-CCTCC-TCGGAGATCAGCTTCTGCTCGCCTCCGCCTCCACCGggcgtaccagcggagac (400 ng) (the underlined nucleotides are part of the c-Myc epitope, lowercase nucleotides are those of PBR) in 50 l of buffer containing 10 ϫ Pfu buffer (Stratagene, 5 l, 2 mM final MgCl 2 ), dNTP (1 l of 5 mM, 50 M final), were capped with an Ampliwax bead (Perkin-Elmer) and heated at 70°C for 5 min followed by 5 min at 20°C. A solution (50 l) containing pEMR971 (100 ng), 10 ϫ Pfu buffer (5 l), and Pfu DNA polymerase (Stratagene, 2.5 units) was added, and the mixture was heated at 95°C for 2 min and then cooled to 50°C for 1 min. Amplification was carried out for 10 cycles at 94°C for 1 min; 55°C for 1 min, and 72°C for 1 min using a Techne 1 thermocycler. Exactly the same procedure was carried out with an antisense primer downstream from the PBR stop codon in pEMR971 5Ј-GAGTCGACGG-GATCCCGTT (400 ng) and 5Ј-AAGCTGATCTCCGAGGAGGACCTG-GGCGGTGGTGGAGGCggcctgcagaagccctcg (400 ng) (underlined and lowercase nucleotides as described above). The two amplified products were purified on 2% agarose, extracted from the gel using a Sephaglass Bandprep purification kit (Pharmacia Biotech Inc.), and approximately equal quantities of amplicons were fused using the same hot-start procedure with Pfu polymerase and five cycles of the same cycle conditions. Finally, the sense and antisense primers described above (400 ng of each) were added, and 15 cycles using the above cycle conditions were carried out. The full-length PBR-c-Myc fusion was digested with XhoI/ BamHI and cloned into the corresponding site in pEMR971. The aminoand carboxyl-terminal c-Myc fusions lacking the pentaglycine flanks were obtained by the insertion of double-stranded synthetic oligonucleotides into appropriate restriction sites. The sequences of these and subsequent constructs were verified by the dideoxynucleotide chain termination method.
Binding Detection of the Tagged PBR-30 g of total mitochondrial protein (determined according to Bradford (20)) was separated by SDS-polyacrylamide gel electrophoresis (15% gel). Proteins were transferred from polyacrylamide gels to nitrocellulose sheet using a semidry transfer cell (Atto) for 1-1.5 h. Blots were overlaid with a specific antiserum, either an anti-hPBR carboxyl-terminal peptide antibody (5) or the antic-Myc 9C10 antibody, and then incubated with an alkaline phosphatase-conjugated anti-antibody or a horseradish peroxidase-conjugated anti-antibody (Jackson), and detection was carried out accordingly.
c-Myc Antibody Titration by ELISA-Back-titration ELISA was performed essentially as described (21). Briefly, freshly prepared intact or sonicated mitochondria (Bioblock 72442, power 40, duty cycle 50%, 2 min) were serially diluted (5-120 g/ml) in 0.4 M sucrose-containing buffer and incubated with anti-c-Myc 9C10 monoclonal mouse antibodies (a gift from Dr. B. Pau, diluted from 1:20,000 to 1:60,000 in sucrosecontaining buffer) or anti F 1 -ATPase rabbit antibodies (dilution of 1:40,000) overnight at 4°C. Mitochondria were pelleted at 15,000 rpm at 4°C for 1 h. The supernatants containing the unbound antibodies were titrated by ELISA. Microtiter plates (ICN) were coated with sonicated mitochondria extracted from the cells harboring the expression plasmid that gave the highest level of c-Myc-tagged PBR. Mitochondria were washed twice with PBS-T (phosphate-buffered saline with 0.05% Tween 20) and incubated for 1 h with 2% bovine serum albumin in PBS to saturate nonspecific sites. After two washes using PBS-T, 100 l of the supernatant-containing antibodies was added to each suspension, which was then incubated at room temperature for 2 h and finally washed five times with PBS-T. Alkaline phosphatase-conjugated antimouse Ig (Jackson) or alkaline phosphatase-conjugated anti-rabbit Ig (Calbiochem), diluted 5,000-fold, was added to each well and incubated at room temperature for another 2 h. After five washes in PBS-T, 100 l of p-nitrophenyl phosphate (Sigma 203) dissolved in 0.1 M Tris-HCl, pH 9, 50 mM MgCl 2 , was added. The phosphatase reaction was left to develop at 37°C for 40 min, and the absorbance was determined at 405 nm using an automatic reader.
Construction of Various PBR Cys Mutants-As a template for sitedirected mutagenesis, we constructed the single-stranded vector (mp19-bPBR). This vector contains the entire bPBR coding sequence, substituted with Val-154 to ensure Ro5-4864 binding (6), inserted into the M13mp19 polylinker region. A similar vector contained a sandwich hybrid PBR devoid of Cys residues coding for the amino-terminal extremity of the bPBR gene to residue 41, the middle of the hPBR between residues 42 and 147, and the carboxyl-terminal extremity of the bPBR(Val-154) mutant from residue 148. Cys replacements were obtained either by site-directed mutagenesis or by polymerase chain reaction using oligonucleotides corresponding to the mutated sequence as the primer. Annealing with the mutagenic oligonucleotides, filling, ligation, and recovery of the mutated strand were performed as recommended by the manufacturer (Amersham) and were essentially as described previously (6). All site-directed and polymerase chain reaction-generated mutants were sequenced over the entire coding region.
Expression of the Cys-PBR Variants and Modifications with Fluorescein-5-maleimide (5FM) or Benzophenone-4-maleimide (BM)-Mitochondria were isolated from yeast cells grown in minimal YNB (0.67% yeast nitrogen base) medium supplemented with glucose and required amino acids or complex YP (1% yeast extract, 2% Bacto-peptone) supplemented with galactose plus glycerol to 2% for induction of PBR expression. Labeling experiments were carried out according to Zhou et al. (16) with the following modifications. Mitochondria (400 g of protein) were resuspended in 0.6 M mannitol-containing buffer (22). Broken mitochondria were obtained by swelling intact mitochondria in water (containing 0.2% Triton X-100 where noted) and sonicated for 20 s at 0°C. 5FM (Interchim) was added to a final concentration of 1 mM (100-fold dilution of a stock solution), and samples were incubated at 0°C for different periods. Aliquots of 0.25 ml were withdrawn and quenched with cysteine to a final concentration of 200 mM. Samples were concentrated by trichloroacetic acid precipitation as follows. 0.25 ml of 50% trichloroacetic acid was added, and the mixture was kept on ice for at least 4 h. Tubes were spun down for 10 min, and pellets were rinsed with acetone. The resulting pellets were resuspended in 30 l of Laemmli buffer to which 1 l of 100 mM Tris-HCl, pH 9.5, had been added. Samples were boiled for 5 min, and 10 l was loaded onto a 15% polyacrylamide gel electrophoresis with 0.1% SDS. After electrophoresis, the gel was placed on an ultraviolet transilluminator to visualize the fluorescence. The gel was subsequently either colored by Coomassie Blue or used for electroblotting to quantify the amount of PBR as described above, using the anti-PBR antibody.
Cys modification by BM followed the same strategy except that BM prepared in dimethylformamide to a stock concentration of 0.25 M was added to give a final concentration of 2 mM in the mitochondrial preparation and was heated at 30°C for 20 -120 min. The reaction was terminated by the addition of 2% 2-mercaptoethanol.

RESULTS
The c-Myc Epitope Scanning Procedure-The topological model of the PBR, as proposed previously by Bernassau et al. (12) on the basis of the amino acid sequence, is shown in Fig. 1. To study the model we first adopted a c-Myc epitope-scanning procedure. At the cDNA level, the 10 amino acid c-Myc epitope was inserted into the various loops of PBR between hydrophobic amino acids ( Fig. 1) to give four constructions denoted c-Myc c1, c2, m1, and m2 (c and m denoting a cytoplasmic and an intramitochondrial location, respectively). To make the epitope more accessible to antibodies it was flanked at each end by 5 glycine residues (15). We also placed the c-Myc epitope either at the amino terminus or carboxyl terminus of the polypeptide, without the flanking glycines. After transfection into yeast we first ensured that expression had occurred and that the PBR had translocated into the yeast mitochondrial membranes in a correctly folded form. This was done by testing the ability of the membrane to be recognized by anti-c-Myc antibodies using Western immunoblotting and by binding experiments.
All of the constructs but one could be detected (Fig. 2). The c-Myc m2 fusion could not be detected, and subsequent Western blot analysis of the cytoplasmic fraction (not shown) revealed a total absence of expression of this construct. The different intensities observed in Fig. 2 reflect construct expression levels because an anti-PBR antibody (5) gave similar results (data not shown). Nearly all of the tagged receptors bound [ 3 H]PK11195 with an affinity similar to that of the wild type PBR, c-Myc c1 binding being a little attenuated. In contrast, the [ 3 H]Ro5-4864 binding varied, the lowest affinity being as-sociated with the amino-terminal-tagged receptor. The addition of an alternative, hemagglutinin-derived epitope at the amino terminus or inserted into the m2 region gave results similar to those with the c-Myc fusions (data not shown).
Having ensured that the modified PBR was essentially similar to the wild type PBR, we next looked at the location of the epitope using the reverse ELISA technique. Either predominantly intact or fully lysed mitochondria were mixed at various concentrations with a known amount of anti-c-Myc antibodies. Mitochondria-bound antibodies were pelleted, and free antibodies were assayed in the supernatants by ELISA. As a con-trol we used antibodies against an integral inner membrane protein, F 1 -ATPase. The results presented in Fig. 3 clearly show that intact mitochondria-dependent antibody depletion was obtained with c-Myc c1, c-Myc c2, and c-Myc at the carboxyl terminus, but not with c-Myc at the amino terminus or c-Myc m1. Antibodies raised against F 1 -ATPase were more depleted with lysed than with intact mitochondria, as would be expected; the incomplete depletion may be attributable to a lack of efficiency of the antibodies. In line with initial experiments, the depletion of antibodies from the supernatant was less efficient with the amino-terminal c-Myc construct than with the others. However, the experiments tended to support the model schematized in Fig. 1. Unfortunately, the lack of expression of the c-Myc m2 construct meant that the putative m2 loop could not be verified by this procedure. Furthermore, the c-Myc insertion technique could obviously not be used to verify amino acids thought to form transmembrane regions. Therefore, in subsequent experiments we adopted a chemical labeling method.
Localization of Cysteine Residues in Mutated bPBR and bPBR-hPBR Sandwich Hybrids Using 5FM and BM-The fluorophore 5FM (16) and BM (23) react with Cys residues in polar and nonpolar environments, respectively. 5FM will also react with Cys in the presence of detergent, which allows it to penetrate into lipophilic regions (16). Subsequent detection of fluorescence indicates whether the Cys is situated in a polar loop region or within a lipidic membrane region. We used the bPBR and a bPBR-hPBR-bPBR sandwich hybrid for these experiments because in previous work we found that the recombinant bPBR is expressed more than the hPBR and would, therefore, be easier to detect by fluorescence in yeast mitochondria (6,18). Because the hPBR contains 2 Cys residues at positions 19 and 153 and bPBR contains a unique Cys residue situated in a putative intramitochondrial region at position 135 in m2, a sandwich hybrid PBR containing no Cys was constructed by replacing a bPPR fragment by the corresponding hPBR fragment, the points of fusion being indicated in Fig. 1. This was used as a template for introducing Cys into various regions of the receptor ( Fig. 1 and Table I). In addition, all of the receptors contained a valine at position 154 (bPBR(Val-154)) to ensure Ro5-4864 binding (6). Each of the constructs described, except the Cys-61 mutant, bound both radiolabeled PK11195 and Ro5-4864 to the same extent as the wild type PBR (data not shown), indicating that they were inserted correctly into the mitochondrial membrane.
Using intact or lysed mitochondria, the bPBR(Val-154) and the various mutants were mixed with 5FM and allowed to react for 4 min at 0°C as described (16). Labeling was monitored by measuring the fluorescence associated with 18-kDa protein bands separated by SDS-polyacrylamide gel electrophoresis (Fig. 4). As expected, no 18-kDa band from intact mitochondria bearing PBR devoid of Cys residues was labeled (no Cys in Fig.  4, B and G). In contrast, the Cys-41 bPBR mutant (Fig. 4, A, B, and G) could be detected in intact mitochondria as early as 1 min after the addition of the reagent (Fig. 4A). These results indicated the specificity of the reagent for the PBR cysteines, at the same time showing the ready accessibility of Cys-41. Panel A also shows that the Cys-3 insertion mutant was not labeled within 4 min, but it was labeled rapidly in broken mitochondria after 4 min (Fig. 4C). In intact mitochondria, however, some labeling of this mutant took place after 10 min (data not shown). These results indicate that the reagent could penetrate through intact mitochondria after a certain time and also that the Cys-3 insertion mutant was clearly in an intracellular hydrophilic environment. Other mitochondrial proteins were also labeled, in particular abundant proteins of about 30 kDa. Among the latter were the voltage-dependent anion channel and the adenine nucleotide carrier because these bands disappeared from the yeast mutants aac Ϫ and vdac Ϫ (24) lacking the genes encoding these proteins (Fig. 4O). Interestingly, the VDAC protein was labeled in intact mitochondria (i), and the AAC protein could only be labeled in broken mitochondria (b and bt), which reflects the fact that AAC is an inner membrane protein.
Importantly, bPBR(Val-154) with its unique Cys at position 135 failed to react with 5FM in either intact or broken mitochondria under the conditions used for labeling the other Cys and could only be labeled when the mitochondria were broken in the presence of Triton X-100 (Fig. 4L). These results show that Cys-135 in the bPBR is apparently situated in a transmembrane region rather than in m2, but the bPBR may be an exception since the PBR of other species have an Arg at this position (Fig. 1). The Cys-135 mutant was also labeled by BM in intact mitochondria (data not shown). The neighboring, fully conserved Ser in position 130 is clearly in an accessible internal loop because Cys-130 could be labeled with 5FM in broken mitochondria (Fig. 4K).
Subsequent experiments aimed at firmly establishing transmembrane and loop domains used the bovine-human-bovine sandwich receptor into which a single Cys was substituted. The results are shown in Figs. 4 and 5 and summarized in Table I for the 5FM labeling. The Cys residues that were labeled rap- idly by 5FM in intact mitochondria were at positions 26, 41, 102, 106, and 168 (Fig. 4, F, G, J, and N), indicating that their cytoplasmic location was compatible with the model. In contrast, Cys-75 and Cys-130 were only labeled in broken mitochondria (Fig. 4, I and K), indicating a polar, intramitochondrial location.
The result with the Cys-3 insert led us to make two further mutants near the amino terminus. Cys-12 and Cys-19 were not labeled with 5FM in intact mitochondria, but they were in the presence of detergent (Fig. 4, D and E). These mutants were also partly labeled with BM in intact mitochondria (Fig. 5), thus placing them firmly in a lipophilic environment. An exactly similar result was obtained with the Cys-61 mutant (Figs. 4H and 5), in accordance with a transmembrane location. The BM reagent was found to be satisfactory for topological studies, paradoxically because of the incompleteness of the reaction, since the presence of two closely associated bands on the gel used to separate the labeled proteins (Fig. 5), representing the unlabeled protein and the heavier, labeled protein does allow one to visualize a positive reaction easily. Of all of the constructs, only the Cys-61 mutant failed to bind Ro5-4864, per-haps attributable to a distortion of the binding sites. The Cys-153 bPBR mutant was also in a lipophilic environment since it was labeled with 5FM only in broken mitochondria in the presence of Triton (Fig. 4M) and in intact mitochondria with BM (data not shown). This last result is interesting inasmuch as the mutation adjoins the valine at position 154 previously shown (6) to be directly implicated in Ro5-4864 binding. Val-154 would appear to be at the interface of the fifth transmembrane region and cytoplasm. DISCUSSION Hydropathic analysis of the 169 amino acids that constitute the rat (8), bovine (10), human (18), and mouse (9) PBR reveals the presence of five hydrophobic regions in each of the receptors. Alignment of the sequences (6) clearly shows either identical amino acids in these putative transmembrane regions or conservative replacements. The recently described outer membrane sensory protein of the proteobacterium R. sphaeroides has also been shown recently (3) to have a close structural and functional relationship with the PBR. Therefore, it is a reasonable assumption that the topology of these receptors is identical. There are now several lines of evidence that support the general model proposed by Bernassau et al. (12), according to which the amino terminus of the PBR points toward the interior of the mitochondrial outer membrane, and five transmembrane regions lead to a highly charged carboxyl terminus exposed to the cell cytoplasm. From extensive site-directed mutagenesis modifications of hPBR and bPBR followed by expression in yeast and binding experiments in intact yeast mitochondria, we recently showed (6) that Glu-29, Arg-32, and Lys-39 are in a loop in the cell cytoplasm, c1 (Fig. 1) and that they somehow affected Ro5-4864 binding. Furthermore, we postulated that the carboxyl-terminal region was also in the cytoplasm, perhaps close to c1 (6). More direct evidence for the cytoplasmic location of the carboxyl terminus has come from experiments with an antibody raised against a synthetic peptide from this region of the PBR (25). Although we previously presented PBR models with five transmembrane regions (6, 12), we did not exclude other possible structures, notably one in which both the amino-and carboxyl-terminal regions were located in the cytoplasm (6). The results presented here support the former model.
To establish membrane protein topology various techniques are employed, notably the production of chimeras incorporating a foreign reporter in putative loop regions to establish their location with respect to the membrane and the use of membrane-impenetrable reagents that will chemically modify only exposed residues at specific sites in the protein and, conversely, lipophilic reagents that only react with residues in a membrane environment. In the first set of experiments we inserted c-Myc epitopes into the putative outer and inner loop regions of PBR. Antibodies raised against c-Myc were used as membrane-impenetrable labeling reagents. Various c-Myc-tagged PBR were translocated into the mitochondria of recombinant yeast cells, in a correctly folded form according to binding data with PK11195 and Ro5-4864. After the localization of the epitope in intact or permeabilized mitochondria, it was clear that the results supported the general five-transmembrane structure predicted by the model in Fig. 1. In particular, the aminoterminal region and m1 did indeed appear to be inside the mitochondria. However, the m2 loop could not be detected by this method, and Western blot analysis indicated the absence of this fused receptor in the mitochondria.
We subsequently employed cysteine labeling techniques with the polar 5FM and the apolar BM reagents, techniques that can only be used if the target protein is produced at a high level. For this reason, we turned to bPBR and a bPBR-hPBR hybrid, receptors that can be produced in high amounts in yeast mitochondria. After ensuring that a PBR variant lacking Cys in its sequence could not be modified with 5FM or BM, we used this technique to ascertain whether various Cys residues were located in aqueous or apolar environments. Because it has been reported (26) that hydrophilic reagents are capable of crossing lipid membranes, care must be taken in using them to draw conclusions about receptor topography, unless kinetic aspects are considered. We found that 5FM treatment of intact mitochondria labeled all of the Cys residues predicted to be in three of the domains, c1, c2, and carboxyl terminal, within 4 min, clearly establishing an extramitochondrial, cytoplasmic localization of these regions. This period was therefore chosen for the 5FM experiments because a longer treatment could label intramitochondrial residues. However, the latter residues also reacted within 4 min after we lysed the mitochondria. Significantly, the Cys-135 residue predicted to be in the m2 loop was labeled by 5FM only in the presence of Triton X-100. This result showed this residue to be localized not in an accessible loop region but inside the fifth transmembrane helix. However, the other species contain an Arg at this position, suggesting that the 135 residue lies near the hydrophilic face of the phospholipid and also indicating some flexibility regarding the extent to which the hydrophobic domains are embedded in the membrane. In contrast, residue 130 is clearly situated in the m2 loop region.
In the original model ( Fig. 1) it was proposed that around 18 amino acids formed each of the five transmembrane regions and that the PBR reached only halfway between the outer and inner surfaces of the outer mitochondrial membrane (12). The Cys labeling results with 5FM would indicate, however, that the intramitochondrial residues are in a hydrophilic environment and, therefore, that the transmembrane regions are longer than hydrophobicity analysis would suggest. The exact FIG. 4. Cys labeling with 5FM. Mitochondria were labeled and processed as described under "Experimental Procedures." Unless otherwise indicated all labelings were performed for 4 min on intact (i) or broken (b) mitochondria or mitochondria broken in the presence of Triton X-100 (bt). Arrows indicate the position of the 18 FIG. 5. Cys labeling with 5FM or BM. Intact mitochondria containing the sandwich PBR with Cys-12, Cys-19, and Cys-61 were labeled with 5FM (ϩ) or with BM (ϩ) and detected by Western immunoblot as described under "Experimental Procedures." No PBR refers to mitochondria without PBR. The arrows indicate the unlabeled PBR and the more slowly migrating labeled PBR. beginning and end of transmembrane regions are notoriously difficult to establish, as was pointed out recently in an electroncrystallographic refinement of bacteriorhodopsin at 3.5-Å resolution (27). However, several theoretical methods to predict such regions exist. It has been postulated (11) that Arg/Lys patches occur at the cytoplasmic ends of receptors present in plasma membranes, their positive charges serving to anchor the receptor to the negatively charged phospholipids. No Arg/ Lys patches occur in the PBR of the four species known at present, and it is uncertain whether such anchors exist. However, an alignment of the sequences (6) reveals several perfectly conserved Arg or Lys at positions 32, 39, 69, 103, and 166. In addition, by accepting His in the analysis, positions 27, 43, 46, and 162 can be added. All but two of these positive charges (69 and 76/77) are found in or near the putative extramitochondrial regions or in the cytoplasm (Fig. 1). Our results show that residue 26, just before the positively charged residue 27, is clearly in the cytoplasm. Arg-103 in c2 also appears to be far from the membrane, the nearby positions 102 and 106 being shown here to be in the cytoplasm. Finally, we found Cys-153 to be located in the fifth transmembrane region of the hPBR, but the bPBR has an Arg at this position which, as in the case of Cys-135 discussed earlier, would again suggest that the residue is at the membrane surface. The adjacent Val-154, important for Ro5-4864 binding, is also near the membrane-cytoplasm interface.
It is interesting to note that all four PBR sequences presently available exhibit fully conserved glycines, a residue often involved in structures that terminate ␣-helices (28), near the proposed carboxyl ends of each of the first three putative transmembrane regions; but it must be clearly stated that no evidence exists to show that the transmembrane regions have an ␣-helical structure. Finally, several of the structural features discussed here are to be seen in the R. sphaeroides protein described by Yeliseev et al. (3).
In conclusion, we have obtained experimental data that strongly support the theoretical topographical PBR model shown in Fig. 1 (6, 12), but the data lead us to suggest that some refinements should be made to the model. In particular, because the putative intramitochondrial loops are accessible to hydrophilic reagents, the transmembrane domains may be longer by 3-4 residues thereby completely traversing the outer mitochondrial membrane.