Identification by site-directed mutagenesis and chemical modification of three vicinal cysteine residues in rat mitochondrial carnitine/acylcarnitine transporter.

The proximity of the Cys residues present in the mitochondrial rat carnitine/acylcarnitine carrier (CAC) primary structure was studied by using site-directed mutagenesis in combination with chemical modification. CAC mutants, in which one or more Cys residues had been replaced with Ser, were overexpressed in Escherichia coli and reconstituted into liposomes. The effect of SH oxidizing, cross-linking, and coordinating reagents was evaluated on the carnitine/carnitine exchange catalyzed by the recombinant reconstituted CAC proteins. All the tested reagents efficiently inhibited the wild-type CAC. The inhibitory effect of diamide, Cu(2+)-phenanthroline, or phenylarsine oxide was largely reduced or abolished by the double substitutions C136S/C155S, C58S/C136S, and C58S/C155S. The decrease in sensitivity to these reagents was much lower in double mutants in which Cys(23) was substituted with Cys(136) or Cys(155). No decrease in inhibition was found when Cys(89) and/or Cys(283) were replaced with Ser. Sb(3+), which coordinates three cysteines, inhibited only the Cys replacement mutants containing cysteines 58, 136, and 155 of the six native cysteines. In addition, the mutant C23S/C89S/C155S/C283S, in which double tandem fXa recognition sites were inserted in positions 65-72, i.e. between Cys(58) and Cys(136), was not cleaved into two fragments by fXa protease after treatment with diamide. These results are interpreted in light of the homology model of CAC based on the available x-ray structure of the ADP/ATP carrier. They indicate that Cys(58), Cys(136), and Cys(155) become close in the tertiary structure of the CAC during its catalytic cycle.

overlap extension method (13) and the High Fidelity PCR system (Roche Applied Science). The PCR products were purified by the Gene Clean Kit (La Jolla Pharmaceutical Company), digested with NdeI and HindIII (restriction sites added at the 5Ј end of forward and reverse primers, respectively), and ligated into the pMW7 expression vector. Single or double tandem fXa recognition sites were inserted in positions 65-72 of the four-Cys replacement mutant C23S/C89S/C155S/C283S by replacing the sequence REGITGLY with IEGRIEGR using the overlap extension method (13). All mutations were verified by DNA sequencing, and, except for the desired base changes, all of the sequences were identical to that of rat CAC cDNA. The resulting plasmids were transformed into E. coli C0214 (11). Bacterial overexpression, isolation of the inclusion body fraction, and solubilization and purification of the wildtype CAC and mutant CAC proteins were performed as described previously (11).
Reconstitution of CAC and CAC Mutants into Liposomes-The recombinant proteins were reconstituted into liposomes in the presence of 13 mM carnitine, as described previously (11). The external substrate was removed from proteoliposomes on Sephadex G-75 columns.
Transport Measurements and Effect of SH Oxidizing, Cross-linking, and Coordinating Reagents-Transport at 25°C was started by adding 0.1 mM [ 3 H]carnitine to proteoliposomes and terminated by the addition of 1.5 mM N-ethylmaleimide (11). In controls, the inhibitor was added together with the labeled substrate, according to the inhibitor stop method (14). Finally, the external substrate was removed by chromatography on Sephadex G-50 columns, and the radioactivity in the liposomes was measured (14). The experimental values were corrected by subtracting control values. All of the transport activities were determined by taking into account the efficiency of reconstitution (i.e. the share of successfully incorporated protein). To study the effect of diamide, Cu 2ϩ -phenanthroline, or phenylarsine oxide on the transport activity of wild-type CAC and CAC mutants, the proteoliposomes were preincubated with each reagent (under the conditions indicated in the legends to figures 1, 2, 3, 4, and 6) before transport was started. In the case of Sb 3ϩ , this reagent was added to the reconstitution mixture as potassium antimony tartrate.
Digestion with fXa Protease and Electrophoresis Analyses-The mutant containing fXa recognition sites was reconstituted into liposomes, as described above, and incubated in the presence or absence of 2 mM diamide for 30 min at room temperature. Next, the proteoliposomes were passed through Sephadex G-200, ultracentrifuged at 110,000 ϫ g for 90 min at 4°C, resuspended in fXa digestion buffer (50 mM Tris-HCl, pH 8.0, and 1 mM CaCl 2 ) and 2% Triton X-100, and digested with fXa protease (4 g/ml) for 24 h at 4°C. Finally, the samples were analyzed by SDS-PAGE in the presence or absence of dithioerythritol.
Other Methods-SDS-PAGE was performed according to Laemmli (15) as described previously (12). The amount of recombinant protein was estimated on Coomassie Blue-stained SDS-polyacrylamide gels by a Bio-Rad GS-700 Imaging Densitometer equipped with the software Bio-Rad Multi-Analist, using bovine serum albumin as standard. The extent of incorporation of the recombinant protein into liposomes was determined as described by Phelps et al. (16), with the modifications reported in Ref. 12. N-terminal sequencing was carried out as described previously (17). The amino acid sequences were aligned with ClustalW (version 1.7). The homology model of the CAC was built by the Swiss-Model protein modeling server (18 -20) using the x-ray structure of the carboxyatractyloside-ADP/ATP carrier complex as a template (21).

RESULTS
Transport Activity of CAC Mutants-As reported previously, all of the single Cys mutants exhibited transport activities similar to the wild-type protein (12). Table I shows the transport activities of the mutants containing more than one substitution. Most of these mutants had transport activities ranging from ϳ20 to 110% of the wild-type CAC. The mutants C23S/ C58S/C136S, C58S/C136S/C155S, C23S/C58S/C136S/C155S, and C23S/C58S/C89S/C155S were more or less inactive (transport activity Ͻ5% of the wild type). A comparable loss of activity was observed previously for the C-less CAC and for the five-replacement mutants C23S/C58S/C89S/C136S/C283S and C23S/C58S/C89S/C155S/C283S (12). All of these virtually inactive mutants are characterized by substitution of at least three Cys residues among the four residues Cys 23 , Cys 58 , Cys 136 , and Cys 155 , independently of the presence or absence of Cys 89 and Cys 283 ( Table I). Fig. 1 shows the effect of the SH-oxidizing reagent diamide (22), at concentrations from 0.1 to 0.5 mM, on single CAC Cys mutants. Diamide strongly inactivated the wild-type protein, reaching nearly complete inhibition of transport activity at 0.5 mM. The mutants C23S, C58S, C89S, C155S, and C283S were inactivated similarly to the wild-type CAC. Only C136S was much less sensitive to the reagent. Because protein inactivation caused by diamide is used to monitor the formation of S-S bridge(s) (4), these results indicate that Cys 136 is involved in the formation of S-S bridge(s). To gain information about the second Cys residue involved in the disulfide bridge(s), double mutants were tested for their sensitivity to diamide concentrations up to 2 mM. In a first set of double mutants, Cys 136 was substituted together with a second Cys residue ( Fig. 2A). The double mutants C89S/C136S and C136S/C283S showed a sensitivity to diamide that was virtually indistinguishable from that exhibited by the single mutant C136S, indicating that Cys 89 and Cys 283 are not involved in S-S formation. The sensitivity to diamide of the double mutant C23S/C136S and in particular of C58S/C136S was lower than that of the single mutant C136S; a complete loss of sensitivity to diamide was observed with the mutant C136S/C155S, indicating that the presence of Cys 136 and/or Cys 155 is essential for the formation of S-S bridges either with each other or in combination with other Cys residues. The dependence on diamide of the transport activity exhibited by the mutants C23S/C58S, C23S/C155S, and C58S/C155S is shown in Fig. 2B. C23S/C58S and C23S/C155S showed a sensitivity to diamide similar to or only slightly lower than that of the wild type, whereas C58S/C155S had a much lower sensitivity to diamide than the wild type, indicating that an S-S bridge is formed between Cys 58 (but not Cys 23 ) and Cys 155 . We then analyzed the effect of diamide on CAC mutants in which only two of the four Cys residues 23, 58, 136, and 155 were present (Fig. 3). C23S/C58S/C89S/C283S showed a sensitivity  2B) was found for the triple mutants C23S/C58S/C89S and C23S/C58S/C283S and that the addition of 15 mM dithioerythritol after treatment with 2 mM diamide led to a complete recovery of the activity of all the mutants described (not shown). In addition, the effect of diamide on mutants C23S/ C58S/C136S, C58S/C136S/C155S, C23S/C58S/C136S/C155S, and C23S/C58S/C89S/C155S was not tested because of their very low transport activity.

Effect of Diamide on CAC Cys Mutants-
Effect of Cu 2ϩ -phenanthroline and Phenylarsine Oxide on CAC Four-Cys Replacement Mutants-We also tested the effect of Cu 2ϩ -phenanthroline, another reagent that, like diamide, oxidizes vicinal SH residues to S-S bridges (23) In another set of experiments, the effect of phenylarsine oxide was investigated. This reagent does not induce the formation of S-S bridges but reacts with two Cys residues close enough to be cross-linked with the reagent (24). Fig. 4B shows that the wild-type CAC, C23S/C58S/C89S/C283S, and C23S/ C89S/C155S/C283S were strongly inhibited by phenylarsine oxide. C23S/C89S/C136S/C283S and C58S/C89S/C155S/C283S were inhibited to a lesser extent, and C58S/C89S/C136S/C283S and especially C89S/C136S/C155S/C283S were affected poorly by phenylarsine oxide up to 1 mM. Taken together, these results indicate that Cys 136 and Cys 155 can form an S-S bridge not only with each other but also with Cys 58 and perhaps with Cys 23 .
Effect of Sb 3ϩ on CAC Cys Mutants-To obtain further support for the existence of vicinal SH groups in the CAC molecule, the effect of Sb 3ϩ was investigated. Sb 3ϩ is known to form a tricoordinate complex with a cluster of three Cys residues in the tertiary structure of proteins (25,26). We first found that 0.8 mM Sb 3ϩ completely inhibited the activity of the wild-type CAC, indicating that Sb 3ϩ can form a complex with three of the Cys residues present in the carrier protein. Then we compared the effects of the reagent, at a concentration giving half-maximal inhibition of the wild-type protein (0.2 mM), on CAC mutants in which one or more of the four Cys residues 23, 58, 136, and 155 were substituted (Fig. 5). All of the four-replacement mutants tested (containing only two Cys residues) were insensitive to Sb 3ϩ . The double mutant C23S/C283S (containing Cys 58 , Cys 89 , Cys 136 , and Cys 155 ) was strongly inhibited by Sb 3ϩ , similarly to the wild-type protein. In contrast, little inhibition by Sb 3ϩ was observed with the double mutants in which at least one of the three cysteines 58, 136, and 155 was substituted. Among the single mutants, C23S, C89S, and C283S were markedly inhibited similarly to the wild-type CAC, whereas C58S, C136S, and C155S were only slightly inhibited by Sb 3ϩ . These results point to the importance of each of the three cysteines 58, 136, and 155 in the formation of the tricoordinate complex with Sb 3ϩ .
The Chemical Modifications Involve Only Cysteines Belonging to a Single Polypeptide Chain-It is important to know whether the cysteine pairs of the CAC oxidized by diamide and Cu 2ϩ -phenanthroline and cross-linked by phenylarsine oxide, as well as the three cysteines coordinated by Sb 3ϩ , are located on the same polypeptide or on different polypeptide subunits. This question was investigated by subjecting the wild-type CAC and its mutants to SDS-PAGE, in the absence of dithioerythritol, after treatment with the above mentioned reagents. The results obtained showed no significant variation in the apparent molecular mass of any protein tested with or without preincubation with the chemical modifiers (not shown). This result clearly indicates that diamide, Cu 2ϩ -phenanthroline, phenylarsine oxide, and Sb 3ϩ do not cause dimerization of CAC and its mutants, and therefore all the cysteines involved in the modifications described above are located on a single polypeptide subunit.
Disulfuric Linkage in the C23S/C89S/C155S/C283S CAC Mutant with Inserted fXa Recognition Sites-In other experiments, double tandem fXa recognition sites were inserted in positions 65-72 of the C23S/C89S/C155S/C283S mutant. After incubation with fXa protease, this mutant was proteolyzed to an appreciable extent, giving rise to two fragments (Fig. 6, lane  1). As deduced from their N-terminal sequencing, the smaller fragment, with a molecular mass of ϳ7.5 kDa, extended from the N terminus to the cleavage site, and the larger fragment, with a molecular mass of ϳ25 kDa, extended from the cleavage site to the C terminus. Interestingly, if the mutant was treated with diamide before the incubation with fXa protease, no cleavage was observed (Fig. 6, lane 2), indicating that under these conditions, the two peptides are linked by a diamide-induced S-S bridge. To support this conclusion, aliquots of the proteins that had been treated with fXa protease without (Fig. 6, lane 1) and with (lane 2) diamide were supplemented with dithioerythritol before being subjected to SDS-PAGE (lanes 3 and 4). After the dithioerythritol addition, the two peptides also appeared in the diamide-treated sample (Fig. 6, lane 4), i.e. the S-S bridge formed by diamide was reduced. It is noteworthy that the smear of the protein bands shown in Fig. 6 was caused by the presence of phospholipids derived from the proteoliposomes from which the CAC proteins were extracted. Furthermore, it is possible that internal disulfide bridges contributed slightly to the apparent molecular mass in the absence of dithioerythritol (Fig. 6, lane 2). This also explains why the mutant C23S/ C89S/C155S/C283S often appears as two close bands after gel electrophoresis, due to either spontaneous oxidation of a fraction of the protein or incomplete reaction with diamide. It should be mentioned that after reconstitution into liposomes, the mutant containing double tandem fXa recognition sites showed much lower activity than the wild-type CAC (Table I) and that the same mutant containing only one fXa recognition site showed higher transport activity but was poorly cleaved by fXa protease (not shown, see also Ref. 27).
Homology Model of CAC Based on the X-ray Structure of the ADP/ATP Carrier-The rat mitochondrial CAC is homologous with the bovine mitochondrial ADP/ATP carrier for which a crystal structure is available (21). In fact, the sequences of the two carriers can be aligned unambiguously (Fig. 7); they share  Fig. 1.   FIG. 5. Effect of Sb 3؉ on CAC mutants. The reconstitution mixture was preincubated for 2 min with 0.2 mM Sb 3ϩ , which was added as potassium antimony tartrate. The resulting proteoliposomes were assayed for transport as described under "Experimental Procedures" and as described in the Fig. 1 legend, except that the incubation time was 10 min. WT, wild type.  2 and 4) or without (lanes 1 and 3) diamide and then with fXa protease. Subsequently, the samples were subjected to SDS-PAGE in the presence (lanes 3 and 4) or absence (lanes 1 and 2) of dithioerythritol (DTE). Other conditions were as described under "Experimental Procedures." 19.6% identical amino acids and 39.6% highly conserved residues. Interestingly, Cys 58 and Cys 155 of the rat CAC are conserved in the bovine ADP/ATP carrier corresponding to Cys 56 and Cys 159 , respectively. We therefore built a homology model of CAC (Fig. 8) based on the three-dimensional structure of the carboxyatractyloside-ADP/ATP carrier complex. Fig. 8 highlights the positions of the six cysteine residues of CAC, which are grouped at the matrix side (Cys 58 , Cys 136 , and Cys 155 ) and at the cytoplasmic side (Cys 23 , Cys 89 , and Cys 283 ) of the membrane-embedded CAC protein. DISCUSSION The resolution of the tertiary structure of mitochondrial carriers by x-ray crystallography is still in its infancy. So far, only the structure of the ADP/ATP carrier-carboxyatractyloside inhibited complex has been determined after crystallization of the purified carrier from bovine mitochondria (21). Therefore, especially for a protein like CAC, which is present in a very minute amount in mitochondria, the site-directed mutagenesis and chemical modification approach still represents a useful tool to obtain structural and dynamic information and to define structure-function relationships.
We found previously that formation of disulfide(s) in the CAC protein, purified from mitochondria, leads to transport inhibition, and we therefore concluded that at least two of the six Cys residues present in the CAC amino acid sequence are in close proximity (4). Identifying the specific residues that can form disulfides or can be cross-linked has important implications on the structure and dynamic properties of the CAC, because the protein segments that contain these Cys residues have to be sufficiently close or become close to each other during the conformational changes accompanying the translocation of the substrate through the protein. The functional analysis described above of one-, two-, and four-Cys replacements mutants of CAC, as well as of the wild-type recombinant protein, clearly shows that Cys 58 , Cys 136 , and Cys 155 can be oxidized by diamide and Cu 2ϩ -phenanthroline or cross-linked by phenylarsine oxide. Therefore the residues 58, 136, and 155 of CAC are close or become close in the tertiary structure of the protein during its catalytic cycle. This conclusion is supported by the observation that Sb 3ϩ , a reagent forming a tricoordinate complex with three cysteines (25,26), inhibits only the CAC mutants containing the above mentioned cysteines. It is also supported by the observation that the CAC mutant C23S/C89S/ C155S/C283S, containing the cysteines 58 and 136 and two tandem fXa recognition sites in positions 65-72, is not cleaved by fXa protease into two fragments if pretreated with diamide.
The CAC and the ADP/ATP carrier are homologous and must have the same overall structure. Therefore, we discuss our results in light of the homology model of CAC (Fig. 8) based on the available x-ray structure of the ADP/ATP carrier (21). It is particularly interesting to compare the three-dimensional structure with our Cys mutagenesis and chemical modification data because the former can only provide static information, whereas the latter can give a more dynamic view of the structure.
As shown in the homology-modeled structure of the CAC, the three important cysteines, Cys 58 , Cys 136 , and Cys 155 , are clustered at approximately identical distances from the membrane/ aqueous interface near the mitochondrial matrix side. In the three-dimensional structure, the distances between the side chains of these cysteines (for instance, 9.4 Å between Cys 136 and Cys 155 ) are greater than those required for the chemical modifications described in this study. However, it is possible that Cys 58 , Cys 136 , and Cys 155 become closer during some stage of the catalytic cycle. The presence of the substrate carnitine, inside and outside of the proteoliposomes, probably facilitates the cross-linking of the CAC cysteine pairs observed here. It has indeed been shown that on binding the substrate, the carrier-substrate complex undergoes a conformational change, causing the binding center to switch between the external and internal states (reviewed in Ref. 28). Therefore, substrates may confer on the above mentioned cysteines the mobility necessary for gaining sufficient proximity to be cross-linked. We demonstrated previously that Cys 136 is accessible from the extraliposomal (cytosolic) side to the membrane-impermeant reagents sodium(2-sulfonatoethyl)-methane-thiosulfonate and p-(chloromercuri)benzene-sulfonate and that this interaction is prevented by the presence of the CAC substrate (12). Interestingly, the homology model of CAC shows that Cys 136 (corresponding to Thr-138 of the ADP/ATP carrier) protrudes into the large water-accessible cavity that, in the crystallographic structure of the ADP/ATP carrier-carboxyatractyloside complex, is exposed toward the cytosolic side of the mitochondrial membrane and is occupied by the inhibitor or possibly by the cytosolic ADP (21).
The effects of the various modification reagents on Cys 23containing mutants are relatively minor, and the kinetics of the cross-linking reactions are also clearly different for Cys 23 -containing mutants than for the mutants containing at least two of the three cysteines 58, 136, and 155 (Figs. 2 and 4). These results alone make it unlikely that Cys 23 can be oxidized or cross-linked with Cys 58 , Cys 136 , or Cys 155 . This possibility is definitively ruled out by the homology model of CAC. Cys 23 is pointing into the lipid phase at the cytoplasmic side of the carrier protein, and its distance from Cys 58 , Cys 136 , and Cys 155 (24.2, 21.3, and 24.2 Å, respectively) is too great to be crosslinked with one of these cysteines, unless drastic structural changes take place. In this study, some reconstituted mutants containing Cys 23 , but with various mutations in the other native cysteines, have considerably lower activity than the wildtype protein (for example 19.8% of the wild-type activity for C58S/C155S). Because none of the native cysteines (including Cys 23 ) are important for the transport activity per se (12), the decreased activity of these mutants may be caused by structural/folding instability. A plausible interpretation of the results, which seemed to suggest involvement of Cys 23 in disulfide bridge(s) with Cys 136 or Cys 155 , is that in certain recombinant mutants containing Cys 23 and mutations in other native cysteines, there may be a portion of misfolded molecules that give anomalous cross-linking. This would in fact agree with previous results showing that C23S mutants have low yields of reconstitution into liposomes, which is likely because of structural/folding instability (12).
The CAC contains two other Cys residues, Cys 89 and Cys 283 . Our mutagenesis and chemical modification data, which show that neither Cys 89 nor Cys 283 can be oxidized or cross-linked with Cys 58 , Cys 136 , or Cys 155 , can be explained easily by the CAC homology model, because they are located at the cytoplasmic side of the carrier and are Ͼ20 Å from the three important cysteines Cys 58 , Cys 136 , and Cys 155 . However, we cannot exclude the possibility that Cys 89 and Cys 283 may form a disulfide bridge with each other or with Cys 23 without loss of transport activity.
Another result of this work is that at least two of the three important Cys residues of the rat CAC (Cys 58 , Cys 136 , and Cys 155 ) have to be present in reduced form for efficient activity of transport. It can be excluded, however, that these cysteines have a catalytic function because they are not conserved in evolutionarily distant organisms such as S. cerevisiae and A. nidulans. It is likely that their oxidation, cross-linking, and coordination by Sb 3ϩ rigidifies the protein structure, causing reduction in transport activity. SH groups interact both with hydrophilic residues (SH, OH, NH 2 ) via hydrogen bond (29) and with hydrophobic residues (30). It is possible that in the rat CAC, two thiol groups of the three cysteines 58, 136, and 155 dynamically switch from hydrophilic to hydrophobic interactions, being structurally and dynamically important, probably because they are involved in one or more conformational changes during the translocation process.