The N-terminal Domain of Rat Liver Carnitine Palmitoyltransferase 1 Contains an Internal Mitochondrial Import Signal and Residues Essential for Folding of Its C-terminal Catalytic Domain*

We have previously shown that the first 147 N-terminal residues of the rat liver carnitine palmitoyltransferase 1 (CPT1), encompassing its two transmembrane (TM) segments, specify both mitochondrial targeting and anchorage at the outer mitochondrial membrane (OMM). In the present study, we have identified the precise import sequence in this polytopic OMM protein. In vitroimport studies with fusion and deletion CPT1 proteins demonstrated that none of its TM segments behave as a signal anchor sequence. Analysis of the regions flanking the TM segments revealed that residues 123–147, located immediately downstream of TM2, function as a noncleavable, matrix-targeting signal. They specify mitochondrial targeting, whereas the hydrophobic TM segment(s) acts as a stop-transfer sequence that stops and anchors the translocating CPT1 into the OMM. Heterologous expression in Saccharomyces cerevisiae of several deleted CPT1 proteins not only confirms the validity of the “stop-transfer” import model but also indicates that residues 1–82 of CPT1 contain a putative microsomal targeting signal whose cellular significance awaits further investigation. Finally, we identified a highly folded core within the C-terminal domain of CPT1 that is hidden in the entire protein by its cytosolic N-terminal residues. Functional analysis of the deleted CPT1 proteins indicates that this folded C-terminal core, which may belong to the catalytic domain of CPT1, requires TM2 for its correct folding achievement and is in close proximity to residues 1–47.

In mammals, the mitochondrial carnitine palmitoyltransferase (CPT, 1 EC 2.3.1.21) 1 is the key regulatory enzyme of long chain fatty acid oxidation (1). This enzyme catalyzes the conversion of long chain acyl-CoA to acylcarnitines, which permits, in cooperation with the carnitine/acylcarnitine translocase and the CPT2, their transport from the cytoplasm into the mitochondrial matrix to undergo ␤-oxidation. CPT1 is tightly regulated by its physiological inhibitor, malonyl-CoA, the first committed intermediate of fatty acid biosynthesis (1). The liver mitochondrial CPT1 isoform is anchored into the outer mitochondrial membrane (OMM) in an N cyto -C cyto orientation via two ␣-helical hydrophobic transmembrane (TM) segments (TM1, residues 48 -75; TM2, residues 103-122). Its N terminus (residues 1-47) and its large C-terminal domain (residues 123-773) are facing the cytosol, whereas the loop connecting TM1 and TM2 is exposed in the intermembrane space (2,3). Apart from mitochondria, microsomes and peroxisomes also contain membrane-bound malonyl-CoA-sensitive CPTs (4,5), which share similar functional properties with the mitochondrial CPT1, have an identical molecular mass of about 88 kDa, and were immunoreactive with antibodies raised against distinct linear epitopes of the mitochondrial CPT1 (6). Whether these enzymes are identical or similar is still a matter of debate. This raises the crucial question of how the mitochondrial CPT1 is specifically imported into the OMM and whether multiple or hierarchical targeting sequences could exist within a single polypeptide allowing distinct subcellular locations.
Nuclear-encoded mitochondrial proteins are synthesized as precursors in the cytosol and harbor signals that mediate primarily, via a specific interaction with the outer mitochondrial receptors (Tom complex), their mitochondrial targeting and specify their intramitochondrial sorting (7). The targeting signals in matrix-destined preproteins are cleavable N-terminal presequences, positively charged and that have the potential to adopt amphipathic ␣-helices (8). By contrast, integral OMM proteins are synthesized as noncleavable proteins (9) and therefore are targeted to mitochondria by means of internal signals. How this is accomplished is still not clear, although clues have begun to emerge from studies of bitopic proteins, such as the Saccharomyces cerevisiae Tom70 and Tom6 and the mammalian Bcl-2 protein. Their targeting and insertion into the OMM have been shown to be mediated by their unique hydrophobic TM segment that functions as a "signal anchor sequence" selective for the OMM (10 -12). An alternative to the signal anchor sequence model is the combination of a matrixtargeting signal with a hydrophobic stop-transfer sequence. Primarily based on the import studies of artificial bitopic chimeric proteins (13), this model has been shown to be valid for the Neurospora crassa bitopic Tom22 protein (14). Very few investigations have been performed on the nature of the targeting and/or topogenic signals of integral polytopic OMM proteins over the past 15 years. These proteins fall into two classes, namely those that contain transmembrane ␤-sheets, such as porin and the yeast Tom40, and those with ␣-helical hydrophobic TM segments. In the case of porin and Tom40, limited information regarding structural determinants of these ␤-barrel proteins is available (15)(16)(17)(18)(19), but the precise nature of their targeting signals remains unclear. Bearing in mind the structural difference between the two classes of polytopic OMM proteins, it would appear unlikely that the targeting and/or * 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. topogenic signals operate in a uniform manner. Thus, the rat liver CPT1 could be a useful model to study the mechanisms involved in mitochondrial targeting and membrane insertion of OMM proteins containing more than one ␣-helical hydrophobic TM segment.
We have previously shown that the N-terminal domain (residues 1-147) of CPT1 contains all of the information for mitochondrial targeting, OMM insertion, and membrane orientation (20). In the present study, we have identified the precise internal import signal of CPT1 by using two complementary approaches. In vitro import assay of fusion and/or deletion CPT1 proteins shows that its import into the OMM does not occur by a "signal anchor" but by a "stop-transfer" mechanism with the involvement of an internal matrix-targeting signal located immediately downstream of TM2. Heterologous expression in S. cerevisiae of several deleted CPT1 proteins not only confirms the validity of this model but also allows new insights into the folding of the C-terminal catalytic domain of CPT1. The functional importance of certain residues within the Nterminal domain for maintenance of a putative catalytic core is discussed.
pCPT1⌬83-148 -DNA encoding CPT1-(1-82) was amplified by PCR using the same 5Ј-primer previously used to generate pCPT1-(1-147)-DHFR and the 3Ј-primer 5Ј-CGG GAT CCT GCC CAG GGA GGG-3Ј, introducing a BamHI restriction site. The EcoRI-BamHI PCR fragment was subcloned into pGEM4-CPT1⌬3Ј deleted from its large EcoRI-BglII fragment which codes for the first 147 amino acids of CPT1. The resulting pCPT1⌬83-148 encodes a protein in which amino acid 82 is fused to residue 149 and possesses one extra amino acid (Arg) in the joining region.
pCPT1⌬31-148 -This construct was obtained by excising the BglII-BglII fragment from pGEM4-CPT1⌬3Ј and re-ligating the plasmid. This results in a CPT1 protein in which amino acid 30 is fused to amino acid 149.

In Vitro Synthesis of Precursor Proteins and Import into Mitochondria
Radiolabeled precursor proteins were synthesized by in vitro transcription-translation using the TNT® SP6-coupled reticulocyte lysate system (Promega) in the presence of [ 35 S]methionine (Amersham Pharmacia Biotech) according to the manufacturer's protocols. Isolation of purified rat liver mitochondria was performed as described previously (20). In vitro import of radiolabeled proteins into mitochondria was carried out for 30 min at 4 or 30°C in import buffer as in Ref. 20. Dissipation of the membrane potential (⌬⌿) by 1 M of carbonyl cyanide m-chlorophenylhydrazone (CCCP; Sigma), mitochondrial pretreatment with trypsin (Sigma), postmitochondrial treatment with trypsin, alkaline extraction with 0.1 M Na 2 CO 3 , as well as analysis of the import reactions by SDS-PAGE and fluorography were performed as described previously (20).

Submitochondrial Localization of Imported Proteins
After import, samples were split into 5 eq aliquots (70 g of mitochondrial protein). Mitochondria were centrifuged at 12,000 ϫ g for 5 min at 4°C, washed in KCl buffer (250 mM sucrose, 10 mM Hepes, 80 mM KCl, pH 7.6), and resuspended (0.5 mg of protein/ml). Mitochondria were then diluted 10-fold either in KCl buffer (nonswollen mitochondria) or in a swelling buffer (20 mM Hepes-KOH, pH 7.4) and subjected to trypsin (100 g/ml) when indicated. After a 20-min incubation on ice, 2.8 mg/ml soybean trypsin inhibitor (STI; Sigma) was added, and sam-ples were further kept on ice for 10 min. Mitochondria/mitoplasts were reisolated by centrifugation and washed in EDTA buffer (250 mM sucrose, 10 mM Hepes, 1 mM EDTA, pH 7.6). Mitoplasts were then solubilized by 0.5% (v/v) Triton X-100 (Sigma) in the absence or presence of trypsin (400 g/ml). After centrifugation, both the pellet and the supernatant were incubated 5 min at 65°C to inhibit trypsin, and the solubilized proteins present in the supernatant were trichloroacetic acidprecipitated. Samples were submitted to SDS-PAGE, blotted onto nitrocellulose, and analyzed by fluorography. The efficiency of swelling and solubilization of the mitoplasts was assessed by immunostaining.

Miscellaneous Methods and Chemicals
Protein concentrations were determined by the method of Ref. 24 with bovine serum albumin as standard. Swelling of yeast mitochondria and Western blotting were performed as described previously (21). The antisera used were against the rat liver CPT1 (1/3000 -1/10,000), the yeast cytochrome b 2 (1/1000), and the yeast mtHSP70 (1/10,000). CPT activity was assayed at 30°C as palmitoyl-L-[methyl-3 H]carnitine formed from L-[methyl-3 H]carnitine (200 M; 10 Ci/mol) and palmitoyl-CoA (80 M) in the presence of 1% bovine serum albumin (w/v) as described previously (21). Malonyl-CoA concentration was 150 M. When indicated, mitochondria were solubilized by 0.5% Triton X-100 as described in Ref. 20 and centrifuged at 16,000 ϫ g for 10 min at 4°C to sediment the insoluble membrane residues, and the supernatants were used for CPT assay.

Statistics
Results are expressed as means Ϯ S.E. Statistical analysis was performed using the Mann-Whitney U test.

The Mitochondrial Targeting Signal of CPT1 Resides within
Residues 97-147-To investigate the precise location of the mitochondrial targeting signal of CPT1 within its N-terminal domain, we first asked whether each half of this domain could play an equivalent role in this process. For this purpose, CPT1-(1-82) and CPT1-(97-147) were fused to a non-OMM-related protein, the cytosolic mouse DHFR (Fig. 1A). CPT1-(1-147)-DHFR was used as a positive control protein, since CPT1-(1-147) allowed import of DHFR into the OMM in a temperatureand trypsin-sensitive surface receptor-dependent manner, the latter being exposed on the cytosolic face of mitochondria (20). Radiolabeled CPT1-(1-147)-DHFR, CPT1-(1-82), and CPT1-(97-147) were synthesized in vitro and used to perform import reactions to determine their specific import requirements. In all cases, efficiency of import into the OMM was assayed using the alkaline extraction method, which removes all proteins that are not integrated into the membranes (10,20). A significant amount of alkaline-resistant CPT1-(1-82)-DHFR and CPT1-(97-147)-DHFR was recovered at 30°C when mitochondria were added in the import reaction (Fig. 1B, compare lanes 2 and 5), suggesting that TM1 and TM2 of CPT1 allowed membrane insertion of CPT1-(1-82)-DHFR and CPT1-(97-147)-DHFR, respectively. Like CPT1-(1-147)-DHFR, the inserted CPT1-(1-82)-DHFR and CPT1-(97-147)-DHFR were totally digested by exogenous added trypsin (Fig. 1B, compare lanes 5 and 6), confirming their insertion into the OMM with DHFR exposed to the cytosol. However, the efficiency of OMM insertion of CPT1-(1-82)-DHFR was decreased by 75% in comparison to what was observed for CPT1-(1-147)-DHFR and CPT1-(97-147)-DHFR. Moreover, analysis of their import requirements showed that, like CPT1-(1-147)-DHFR but unlike CPT1-(1-82)-DHFR, CPT1-(97-147)-DHFR was imported into the OMM in a temperature- (Fig. 1B, compare lanes 4 and 5) and trypsin-sensitive receptor- (Fig. 1B, compare lanes 5 and 7) dependent manner. Thus, the second half of the N-terminal domain of CPT1 conferred to DHFR the ability to interact with the mitochondrial receptors and to be specifically inserted into the OMM. Similarly, deletion of residues 83-148 in CPT1 abrogated in vitro the ability of the protein to be imported into rat liver mitochondria in a temperature-and trypsin-sensitive receptor-dependent manner, whereas deletion of residues 1-82 did not (data not shown). These results suggest that the signal sequence of CPT1 mediating mitochondrial targeting may reside within residues 97-147.
The Transmembrane Segments of CPT1 Do Not Function as Signal Anchor Sequences-To determine whether one or both of the TM segments of CPT1 function as a mitochondrial signal anchor sequence, we fused CPT1-(48 -75), CPT1-(97-122), and CPT1-(48 -122) to DHFR ( Fig. 2A). When import reactions were performed in the absence of added mitochondria, a small amount of CPT1-(48 -75)-DHFR, CPT1-(97-122)-DHFR, and CPT1-(48 -122)-DHFR was recovered as alkaline-resistant forms (Fig. 2B, lane 2), likely due to protein aggregation. Import in the presence of mitochondria led to an increase in the level of these alkaline-resistant forms (Fig. 2B, compare lanes 2 and 4) that were totally digested by trypsin (Fig. 2B, compare  lanes 4 and 5). Thus, the presence of TM1 and/or TM2 allowed insertion of DHFR into the OMM, the latter facing the cytosol. However, in contrast to CPT1-(1-147)-DHFR and CPT1-(97-147)-DHFR (Fig. 1B), there was no significant changes in the amount of inserted CPT1-(48 -75)-DHFR, CPT1-(97-122)- TM segments of CPT1 are indicated by black squares. Numbers denote residues of CPT1 after which the fusion to DHFR occurs. B, import of CPT1 deletion proteins. Import of CPT1 fusion proteins was carried out for 30 min at 4 (lane 4) or 30°C (lanes 2, 3, 5, 6, and 7) in the absence (ϪM; lanes 2 and 3) or presence of freshly isolated rat liver mitochondria (80 g) either directly (lanes 4 -6) or after a pretreatment with 50 g/ml trypsin for 20 min on ice to remove the mitochondrial surface receptors (ϪR; lane 7). Following import, samples were centrifuged and washed in KCl buffer. The samples were then split into two equivalent aliquots and submitted to 0.1 M Na 2 CO 3 (pH 11.5; 0.2 mg of protein/ml) for 30 min on ice either directly (ϪTrypsin; lanes 2, 4, 5, and 7) or after a trypsin treatment (300 g/ml) (ϩTrypsin; lanes 3 and 6). After centrifugation at 100,000 ϫ g for 30 min at 4°C, integral membrane proteins were recovered in the pellet, and samples were analyzed by SDS-PAGE and fluorography. 10%, percentage of the amount of radiolabeled proteins added to each import reaction (lane 1).
DHFR, and CPT1-(48 -122)-DHFR when import was performed with trypsin-pretreated organelles (Fig. 2C, compare lanes 4  and 6). These results show that insertion of TM1 and/or TM2 into the OMM does not require the trypsin-sensitive surface receptors in contrast to CPT1. Thus, targeting of CPT1 to mitochondria is not mediated by its TM segments, ruling out the hypothesis that they may serve as signal anchor sequences. Therefore, the mitochondrial targeting signal of CPT1 must reside in regions flanking the TM segments.
The CPT1 Protein Is Targeted to Mitochondria by an Internal Matrix-targeting Signal-Hydrophobic cluster analysis is an efficient method for predicting secondary protein structure and segmentation (25). This method predicts two putative amphiphilic ␤-strands within residues 1-32 (␤1, residues 8 -14; ␤2, residues 19 -23) and three amphipathic ␣-helices (residues 33-47, 76 -102, and 123-147) flanking the TM segments (26). To test whether these regions have the capacity to target a reporter protein to mitochondria, we constructed a series of proteins in which CPT1-(1-47), CPT1-(1-32), CPT1-(33-47), CPT1-(76 -102), and CPT1-(122-147) were fused to DHFR (Fig.  3A). We used the Su9-DHFR protein, which consists of the presequence of N. crassa F 0 -ATPase subunit 9 preceding DHFR (27), as a positive control for protein import into the mitochondrial matrix. In all cases, the efficiency of the import of these radiolabeled fusion proteins into mitochondria was estimated by determining their protection toward trypsin proteolysis. Su9-DHFR was imported into the mitochondrial matrix, where it became processed to its mature-form size, which was inaccessible to exogenously added trypsin (Fig. 3B, compare lanes 4  and 8). Following import at 30°C in the presence of mitochondria, CPT1-(1-47)-DHFR, CPT1-(1-32)-DHFR, and CPT1-(76 -102)-DHFR were recovered in association with mitochondria but were almost totally digested by exogenous trypsin (Fig. 3B,  compare lanes 4 and 8). This indicated that these CPT1 residues were unable to drive DHFR into the mitochondria. By contrast, CPT1-(33-47)-DHFR and CPT1-(122-147)-DHFR became largely insensitive to added trypsin (Fig. 3B, compare  lanes 4 and 8), demonstrating that they were at least translo-cated across the OMM. Acquisition of the protease protection occurred in a temperature-and receptor-dependent manner (Fig. 3B, compare lane [3][4][5] and needed the presence of a membrane potential (⌬⌿) (Fig. 3C, compare lanes 3 and 4). To ascertain the location of CPT1-(33-47)-DHFR and CPT1-(122-147)-DHFR to the matrix, mitochondria were subfractionated by a swelling procedure in the presence of trypsin. As expected from previous studies (21), the endogenous CPT1 protein was resistant to trypsin digestion in intact mitochondria (due to a folded state of its large cytosolic C-terminal domain) (Fig. 3D,  lane 3) and was partially degraded into an 83-kDa fragment when the OMM was disrupted upon swelling (Fig. 3D, f1 fragment). In contrast, the imported CPT1-(33-47)-DHFR and CPT1-(122-147)-DHFR remained resistant to trypsin proteolysis, as was the endogenous matrix-soluble HSP70 protein (mtHSP70). When the inner membrane of mitoplasts was further solubilized by Triton X-100, the imported CPT1-(33-47)-DHFR and CPT1-(122-147)-DHFR were totally degraded by trypsin, whereas a proteolytic fragment was generated for mtHSP70 (Fig. 3D, lane 8), confirming their localization into the matrix. These results indicate that residues 33-47 and 122-147 behave as matrix-targeting signals. Surprisingly, when residues 1-32 were added to CPT1-(33-47)-DHFR, import of DHFR into the mitochondrial matrix was totally inhibited (Fig. 3B). Thus, the matrix-targeting function of residues 33-47 was abrogated by the first 32 amino acids of CPT1. This explained why residues 33-47 did not support import of CPT1-(1-82)-DHFR (Fig. 1B) or CPT1⌬83-147. To study further the role of residues 122-147 in mitochondrial targeting, import of CPT1⌬121 was analyzed (Fig. 3A). In contrast to CPT1⌬31-148 (data not shown) or CPT1⌬150 (20), CPT1⌬121 was efficiently imported into mitochondria in a process that was dependent upon the temperature, the presence of trypsinsensitive surface receptors, and a membrane potential (Fig. 3,  B and C). Swelling experiments confirmed that the imported CPT1⌬121 was located into the matrix compartment (Fig. 3D). These results demonstrate that residues 122-147 function as a matrix-targeting signal, driving the import of a reporter protein into the matrix in the absence of any of the TM segments of CPT1. As soon as one TM segment was present, such as in CPT1-(97-147)-DHFR, protein translocation across the mitochondrial membranes was arrested, leading to the insertion of the corresponding protein into the OMM (see Fig. 1B). To confirm the essential role of residues 122-147 in mitochondrial targeting, we compared the receptor dependence of the import of CPT1 and CPT1⌬123-148 (Fig. 4). In contrast to the fulllength CPT1, deletion of residues 123-148 within CPT1 abrogated the ability of the protein to be imported in a trypsinsensitive receptor-dependent manner (Fig. 4B, compare lanes 3  and 4). Thus, residues 123-147 specify in vitro the mitochondrial targeting of CPT1, whereas its hydrophobic TM segment(s) likely acts as a stop-transfer sequence that stops and anchors the translocating protein into the OMM.
Residues 122-147 Exert a Retention Force on the OMM Surface-We have previously reported that the N-terminal domain of CPT1 participated in the determination of the N cyto -C cyto membrane topology (20) (Fig. 5B). As shown in Fig. 2B, residues 48 -122 alone, encompassing the two TM segments and the connecting loop, seemed to be sufficient for attainment of the N cyto -C cyto topology of CPT1. In the case of S. cerevisiae Tom70, the N in -C cyto orientation of the protein was reversed when its first 10 residues were replaced by a strong matrixtargeting signal (28). Our aim was to test whether amphiphilicity of the regions flanking the TM segments of CPT1 could, in addition to the TM1-TM2 pairing, be an important determinant for conferring protein topology. It has been shown that the  Fig. 1. Numbers denote residues of CPT1 where the fusion to DHFR occurs. B, TM segments of CPT1 do not allow a specific import of DHFR into the OMM. Import of radiolabeled proteins was performed as described in Fig. 1. unique TM segment of Tom70 (residues 11-29, here termed as OM) allowed specific N in -C cyto insertion of a reporter protein into the OMM (20, 28). Our strategy was to determine whether residues 1-47 or 122-147 of CPT1 could reverse the membrane orientation of this OM segment. For this purpose, CPT1-(1-47) and CPT1-(122-147) were fused to the OM domain preceding the DHFR moiety (Fig. 5A). As expected, the radiolabeled fusion proteins CPT1-(1-47)-OM-DHFR and CPT1-(122-147)-OM-DHFR were efficiently membrane inserted following an import reaction (Fig. 5B, compare lanes 2 and 3). Like CPT1-(1-147)-DHFR, the inserted CPT1-(1-47)-OM-DHFR was totally digested by exogenously added trypsin (Fig. 5B, lane 4). This implied that DHFR was located on the cytosolic face of mitochondria and indicated that residues 1-47 did not cause retention of the N terminus of the protein on the cytosolic face of the OMM. These results were in agreement with the observed N in -C cyto membrane topology of CPT1-(1-82)-DHFR (Fig. 1B). Surprisingly, the membrane-inserted CPT1-(122-147)-OM-DHFR was resistant to trypsin treatment of intact mitochondria (Fig. 5B, lane 4), even in the absence of a membrane potential (data not shown). Upon swelling of mitochondria in the presence of trypsin, CPT1-(122-147)-OM-DHFR became totally digested by the protease (Fig. 5B, lane 5), whereas mtHsp70 remained protease-protected (data not shown). These results suggested that DHFR was located in the intermembrane space compartment and that CPT1-(122-147) was able to reverse the N in -C cyto orientation of OM-DHFR. These experiments emphasize that CPT1-(122-147) but not CPT1-(1-47) may participate together with residues 48 -122 in the determination of the membrane topology of CPT1.
Heterologous Expression of Various Deleted CPT1 Proteins in S. cerevisiae-Yeast cells, a system devoid of endogenous CPT activity, represent a suitable model to study the structurefunction relationships of the rat liver CPT1 (21,29,30). To validate the importance of residues 123-147 of CPT1 for mitochondrial targeting in an in vivo setting, CPT1⌬82, CPT1⌬83-148, CPT1⌬121, CPT1⌬123-148, CPT1⌬31-148, and the fulllength CPT1 were expressed in S. cerevisiae. Immunodetection of the yeast-expressed proteins in crude homogenates showed that proteins of the predicted sizes were expressed, except for CPT1⌬31-148 (Fig. 6A, lane 7). Thus, in agreement with our in vitro import experiments, deletion of residues 31-148 led to a protein that was unable to be targeted correctly to mitochondria and hence might be rapidly degraded within the cells.  Fig. 1A. Numbers denote residues of CPT1 after which the fusion to DHFR occurs. B, import of CPT1-(33-47)-DHFR, CPT1-(122-147)-DHFR, and CPT1⌬121 is temperature-and mitochondrial receptor-dependent. Import of radiolabeled proteins was carried out as described in Fig. 1. Following import, mitochondria were washed, centrifuged, and split into 2 eq aliquots, and trypsin (200 g/ml) treatment was performed when indicated. After inactivation of the protease with STI (4 mg/ml), mitochondria were reisolated, submitted to SDS-PAGE, and analyzed by fluorography. p, precursor; m, mature form of Su9-DHFR. 10%, percentage of input lysate of each radiolabeled protein (lane 1). C, ⌬⌿ dependence of import of CPT1-(33-47)-DHFR, CPT1-(122-147)-DHFR, and CPT1⌬121. Import of the radiolabeled proteins was performed in the presence (ϪCCCP; lanes 2 and 3) or in the absence (ϩCCCP; lane 4) of a membrane potential (⌬⌿). Following import at 30°C, untreated mitochondria were split into 2 eq aliquots. The first one was washed and directly submitted to SDS-PAGE (lane 2). The second one as well as CCCP-pretreated mitochondria were subjected to trypsin treatment (ϩTrypsin; lanes 3 and 4) as described in B. Samples were then analyzed by SDS-PAGE and fluorography. 10%, percentage of input lysate of each radiolabeled protein (lane 1). D, the imported CPT1-(33-47)-DHFR, CPT1-(122-147)-DHFR, and CPT1⌬121 are located in the matrix. After import of the radiolabeled proteins, their submitochondrial localization was determined as described under "Experimental Procedures." Mitochondria (lanes 2 and 3) and mitoplasts (lanes 4, 5, and 7) were incubated in the absence (ϪTrypsin (1); lane 2) or presence of 100 g/ml trypsin (ϩTrypsin (1); lanes 3-5 and 7). All samples were reisolated by centrifugation and washed in EDTA buffer supplemented with STI, except for lane 7. The first 3 aliquots (lanes 2-4; T, total) were directly analyzed by SDS-PAGE, and the last two were solubilized by 0.5% Triton X-100 (ϩTX-100; lanes 5 and 7) at 4°C for 10 min. Trypsin (400 g/ml) was added when indicated (ϩTrypsin (2); lane 7), and samples were kept on ice for another 10 min. Samples were centrifuged to recover the pellet (P; lanes 5 and 7) and the supernatant (S; lanes 6 and 8) that was trichloroacetic acid-precipitated. Samples were submitted to SDS-PAGE, blotted onto nitrocellulose, and analyzed by fluorography. Immunostaining with the endogenous CPT1 and the mtHSP70 was then performed. 10%, percentage of input lysate of each radiolabeled protein (lane 1). f* denotes fragment of HSP70 generated by trypsin treatment of solubilized mitoplasts. f1 and f2 are CPT1-processed species.
Analysis of the subcellular distribution of the yeast-expressed full-length CPT1 shows that, by contrast to our previous reports (20,21), the amount of CPT1 recovered in the microsomal fraction represents about 27% of the total expressed protein (Fig. 6, B and C). One possible explanation for this discrepancy was that immunoblotting was performed more stringently in the present experiment, allowing increased sensitivity in the detection of the expressed protein. Recovery of CPT1 in microsomes was not due to a mitochondrial contamination of the microsomal fraction, as shown by the subcellular distribution of the mtHSP70 that was representative of all the constructs (Fig. 6B). Moreover, microsomal recovered CPT1 exhibited CPT activity (2.49 Ϯ 0.27 nmol/min/mg of protein) that was almost totally (93%) inhibited by 150 M malonyl-CoA (0.17 Ϯ 0.02 nmol/min/mg of protein), ruling out that CPT1 was recovered as aggregated protein. When analyzing the subcellular distribution of the various deleted CPT1, we kept in mind that CPT1 harbors two TM segments, and hence partial deletions within the N-terminal domain might affect targeting and membrane anchorage differentially. Deletion of residues 83-148 led to a complete reverse subcellular distribution when compared with the full-length CPT1 since about 93% of CPT1⌬83-148 was recovered in the microsomal fraction (Fig. 6, B and C). Upon deletion of 123-148, both mitochondrial and microsomal fractions contained equivalent amounts of the expressed protein, underlying the absence of privileged mitochondrial targeting (Fig. 6, B and C). Conversely, deletion of residues 1-82 or 1-121 did not alter mitochondrial targeting, whereas their recovery in the microsomal fraction was almost totally abolished (Fig. 6, B and C). These results confirmed our in vitro experiments ( Fig. 1 and Fig. 3) and supported the conclusion that residues 123-148 were essential for mitochondrial targeting of CPT1. The finding that deletion of residues 1-82 abrogated microsomal location of CPT1 whereas CPT1⌬83-148 was most exclusively recovered in this fraction was puzzling, since both proteins contain a TM segment that could allow their nonspecific anchorage at the microsomal membranes. The present observation suggests the presence of a putative microsomal targeting signal within residues 1-82 of CPT1.
To examine the submitochondrial localization of the expressed deleted CPT1 proteins, intact or swollen yeast mitochondria were submitted to trypsin treatment (Fig. 7). The integrity of the outer and inner mitochondrial membrane was checked by the inaccessibility of endogenous cytochrome b 2 (intermembrane space protein) and mtHSP70 to trypsin proteolysis, respectively. Upon trypsin treatment of intact mitochondria, CPT1⌬82, CPT1⌬83-148, and CPT1⌬123-148 were digested by the protease, and a proteolytic fragment of about 60-kDa was generated (f2 fragment) that remained membraneanchored and detected by our CPT1 antibody raised against residues 317-430 (Fig. 7). These results indicated that (i) these deleted CPT1 proteins were anchored into the OMM with their C-terminal domain exposed to the cytosol, (ii) residues 123-148 are not essential for achievement of the correct membrane topology of CPT1, and (iii) a highly folded core exists within the cytosolic C-terminal domain of CPT1. However, the generation of the f2 fragment was less efficient in the case of CPT1⌬83-148, suggesting a partial unfolding of the C-terminal domain of this deleted protein. By contrast, CPT1⌬121 remained trypsinprotected even upon swelling and became totally digested by the protease when the inner membrane of mitoplasts was solubilized by Triton X-100 (Fig. 7). This showed that CPT1⌬121 was efficiently imported into the mitochondrial matrix but did not harbor a folded core. As shown in Fig. 3D for the native rat liver CPT1 protein, no trypsin-resistant 60-kDa fragment was generated in intact yeast mitochondria expressing CPT1 (Fig.  7). Consequently, the trypsin cleavage site at the cytosolic C terminus, previously observed for CPT1⌬82, was inaccessible in the entire protein. It became unmasked only after the cleavage by trypsin of the loop connecting TM1 and TM2 (f1 fragment), allowing the protease to generate the 60-kDa fragment (f2 fragment) (Fig. 3D and Fig. 7). These results indicate the existence of a highly folded core within the cytosolic C-terminal domain of CPT1 that is hidden by the presence of residues FIG. 4. CPT1-(123-148) mediates interaction with the mitochondrial trypsin-sensitive receptors. A, residues 123-148 were deleted within CPT1 (CPT1⌬123-148). Zigzag lines and black squares were used as in Fig. 1A. B, import of CPT1⌬123-148 is receptorindependent. Import of radiolabeled proteins was carried out for 30 min at 30°C in the absence (ϪM) or presence (ϩM) of mitochondria either directly (lanes 2 and 3) or after a pretreatment with trypsin (Pre-Trypsin; lane 4). Following import, all the samples were submitted to alkaline extraction and analyzed by SDS-PAGE and fluorography. 10%, percentage of input lysate of each radiolabeled protein (lane 1). 5. CPT1-(122-147)  1-82. Moreover, among the deleted CPT1 proteins, only CPT1⌬82 was functionally active but showed a decreased malonyl-CoA sensitivity ( Table I). The absence of CPT activity in intact yeast mitochondria expressing CPT1⌬121 could be due to its matrix location (Fig. 7). To determine whether CPT1⌬121 was still functionally active despite its unfolding, CPT activity was measured in solubilized yeast mitochondria expressing CPT1⌬121 or CPT1 as positive control. Whereas solubilization of mitochondria by 5% Triton X-100 inactivated the yeastexpressed CPT1 (20), we found that 0.5% Triton X-100, which is the concentration used for determining the submitochondrial localization and the trypsin resistance of CPT1⌬121 (Fig. 7), allowed CPT1 to be solubilized in an active and malonyl-CoAsensitive form, when compared with intact mitochondria (Table  I). By contrast to CPT1, the solubilized CPT1⌬121 was totally inactive ( Table I), suggesting that unfolding of the protein led to its inactivation. In conclusion, these results confirm our in vitro import experiments and demonstrate that residues 123-148 function in vivo as a matrix-targeting signal specifying the mitochondrial targeting of CPT1, whereas its TM segment(s) acts as an anchoring signal allowing CPT1 insertion into the OMM.

DISCUSSION
In the present study, we identify the import signal sequence that specifies mitochondrial targeting of an OMM protein harboring two ␣-helical hydrophobic TM segments, the rat liver CPT1. The concordance of the two approaches used clearly shows that neither TM1 nor TM2 of CPT1 constitutes a signal anchor sequence selective for the OMM, in contrast to the bitopic Tom70 and Bcl2 proteins. This emphasizes that unknown features other than simple hydrophobicity alone may explain why a TM segment functions as an OMM signal anchor sequence. The present data show that the mitochondrial targeting of CPT1 is mediated by residues 123-147. These residues exhibit several characteristic features of mitochondrial matrix-targeting signals. We also identify a second noncleavable matrix-targeting signal within residues 33-47. However, the presence of residues 1-32 totally abrogates the matrix targeting function of residues 33-47, both in vitro and in vivo. Epitope Val 14 -Lys 29 of CPT1 has been shown to be sterically masked within the native protein, unless the extreme N terminus of the protein is proteolytically cleaved (3,6). Residues 1-32 might form a tight loop structure in which the hydrophobic sides of the two highly conserved amphipathic ␤-sheets are facing, Gly and Pro residues present between these two ␤-sheets acting as structural breakers (26). We propose that the amphipathic ␣-helical residues 33-47 are in contact with this tight ␤-sheet loop. Such a physical interaction should lead were analyzed by SDS-PAGE and immunoblotting with the rat liver CPT1 and yeast mtHSP70 antibodies. Results are representative of three to four different experiments, and the subcellular distribution of mtHSP70 is representative for all constructs. C, bands from Western blots from three to four different experiments, as in B, were quantified by scanning densitometry. Since mitochondria, microsomes, and cytosol accounted, on average, for 9, 13, and 78% of the total homogenate protein, the signal detected in the mitochondrial and microsomal fractions has to be corrected accordingly to analyze the ratio of CPT constructs targeted to mitochondria versus microsomes. The recovery of the expressed protein in the mitochondrial (gray bars) and microsomal (open bars) fractions was expressed as percentage of total expressed protein.
FIG. 7. Submitochondrial localization of the expressed deleted CPT1 proteins. Mitochondria (50 g of protein) isolated from yeast cells expressing CPT1, CPT1⌬82, CPT1⌬83-148, CPT1⌬121, or CPT1⌬123-148 were incubated for 30 min at 4°C under either iso-(ϪSwelling) or hypo-osmotic (ϩSwelling) conditions in the absence (ϪTrypsin) or presence (ϩTrypsin) of trypsin (10 g/ml). After addition of STI, samples were sedimented, washed, electrophoresed on SDS-PAGE, and analyzed by Western blot using anti-CPT1 antibody. When Triton X-100 (TX-100; 0.5% v/v) was present during the swelling procedure, the sample was directly centrifuged at the end of the incubation. Both pellet (P) and supernatant (S) fractions were incubated 5 min at 65°C to inhibit trypsin, and solubilized proteins present in the supernatant were trichloroacetic acid-precipitated before being analyzed by SDS-PAGE and immunoblotted. Marker proteins were cytochrome b 2 (Cyt.b2) for the intermembrane space and mtHSP70 for the matrix. f1 and f2 denote the respective 83-and 60-kDa fragments of CPT1 generated by trypsin. Results are representative of three to four different experiments.
to an embedding of epitope Val 14 -Lys 29 and should mask the matrix targeting function of residues 33-47.
The present results strongly support our working model of the import pathway for CPT1 (Fig. 8). Initially, the newly synthesized CPT1 is targeted to mitochondria by the means of its internal import sequence (residues 123-147). Subsequently, the protein interacts with the OMM import machinery, as suggested by the inhibition of its import in trypsin-pretreated mitochondria. The determination of the precise component(s) involved in this process awaits further analysis. Whereas the protein was specifically imported into the matrix in the absence of any TM segment (CPT1⌬121), the presence of TM2 proximal to residues 123-147 (CPT1⌬82) led to an OMM insertion of the protein. This clearly shows that TM2 at least acts as a stoptransfer sequence that arrests protein translocation during import across the OMM. At this stage, our working model includes two possible variations. In the "step by step model" integration of the TM segments would occur sequentially (Fig.  8a), whereas pairing of TM1 and TM2 may be a prerequisite before membrane insertion in the "single concerted step model" (Fig. 8b). The hairpin structure will exhibit a higher hydrophobic moment that would favor bilayer integration. Such a concerted partitioning of the TM segments has been described for the ␤-barrel Tom40 (19) and for the inner mitochondrial membrane carrier proteins (31). As reported for Tom40 (19), denaturation of the radiolabeled CPT1 precursor with urea partially decreased the efficiency of its import 2 that favors the single concerted step model. Although further experimental evidences are required to discriminate between these two possibilities, the present work clearly demonstrates that the TM segments of CPT1 act as stop-transfer sequences that arrest protein translocation during import across the OMM.
The distribution of charges on either side of a membrane anchor is responsible for the orientation of proteins of both the bacterial inner membrane and the endoplasmic reticulum membrane (32,33). However, residues 1-47, 76 -102, and 123-147 of CPT1 all bear an identical net positive charge of ϩ3. Therefore, another putative topogenic determinant could have been the amphiphilicity of the regions flanking the TM segments that may exert on the mitochondrial surface a "retention signal" functionally similar to those created and analyzed by Shore and coworkers (13,28). The fact that residues 1-82 of CPT1 can adopt either a N in -C cyto or N cyto -C in topology in the CPT1-(1-82)-DHFR and full-length CPT1 suggests that residues 1-47 do not contain any topogenic information. Indeed, CPT1-(123-147), but not CPT1-(1-47), was able to exert in vitro such a retention signal since its fusion to the unique TM segment of the yeast Tom70 led to the inversion of its membrane topology. However, deletion of residues 123-147 within the N-terminal domain of CPT1 did not alter its membrane topology. Although we cannot exclude that CPT1-(123-147) may participate in the process of membrane insertion, our results indicate that the presence of both TM1 and TM2 is sufficient for achievement of the correct N cyto -C cyto topology of CPT1.
Besides the identification of the import signal sequence specifying the import of CPT1 into the OMM, the present study shows that a minor proportion of the yeast-expressed CPT1 was recovered as a functional enzyme in microsomes. Deletion of the first 82 N-terminal residues abolished completely this microsomal targeting, the resulting protein being recovered only into the mitochondrial fraction, whereas deletion of residues 83-148 had the opposite effect. Similar results were also obtained by Zammit and co-workers 3 by using another yeast expression system (Pichia pastoris). Subcellular distribution of isoenzymes is usually achieved by the expression of two (or more) closely nuclear-related genes. However, the product of a single gene can be targeted to different locations due to the use of alternative transcription-translation initiation sites, alternative splicing, or multiple targeting signal sequences (7). Although the identity of the microsomal CPT1 remains obscure, our results suggest that residues 1-82 of the mitochondrial CPT1 may contain a putative microsomal targeting signal. Further work is required to determine whether the microsomal CPT1 corresponds to a misrouting of the mitochondrial isoen-   zyme, and if the mitochondrial and microsomal CPT1s are encoded by a single gene. If it is not the case, the reminiscence of such a signal within the mitochondrial enzyme might result from the evolution of an ancestral CPT1 gene. Finally, the present study allows new insights into the folding of the C-terminal catalytic domain of CPT1. Following mitochondrial targeting and OMM insertion, CPT1 must fold correctly to attain its native functional conformation that is characterized by a highly folded state resistant to trypsin proteolysis (Fig. 8c). Our current findings indicate the existence of a highly folded core in its cytosolic C-terminal domain, as emphasized by the generation of a trypsin-resistant 60-kDa fragment upon trypsin treatment of intact mitochondria expressing CPT1⌬82. Possible trypsin sites occur C-terminal to Arg-595 or -598 and Lys-631 or -634. Trypsin was able to digest the full-length CPT1 at these sites only when the loop connecting TM1 and TM2 was previously cleaved by the protease. This suggests that the highly folded core within the cytosolic Cterminal domain of CPT1 is hidden in the native protein by its cytosolic first N-terminal residues. Several lines of evidence support the idea that this folded domain may belong to the catalytic core of CPT1. First, CPT1 remained active and malonyl-CoA-sensitive when solubilized by a low Triton X-100 concentration that maintained its C-terminal domain folded. Second, deletion of residues 1-82 did neither alter the folded core nor the catalytic activity but decreased the malonyl-CoA sensitivity. This is in agreement with previous studies suggesting that the catalytic domain of CPT1 resides within its cytosolic C-terminal domain, whereas its extreme N terminus is important for malonyl-CoA sensitivity (3,20,34). Third, deletion of residues 1-121 or 83-148 that encompassed TM2 altered both folding of the C-terminal domain and CPT activity. The fact that residues 171-186 contain the (LI)PX(LVP)P(IVTA)PX-(LIVM)X(DENQAS)(ST)(LIVM)X 2 (LY) motif, which corresponds to the carnitine/choline acyltransferase family signature 1 (2), may explain why CPT1⌬123-148 was inactive despite the presence of the folded core. Indeed, deletion of residues 123-148 led to a shift of residues 171-186 to the OMM that likely alters the catalytic activity of the enzyme independently of the trypsin-resistant folded core.
The present study confirms our previous observation that the N-terminal domain of CPT1 is essential to maintain an optimal conformation for catalytic function (20). We show here that TM2 is essential to achieve the correct folding of this putative catalytic core and that residues 1-47 may be in close proximity to this domain, preventing trypsin from having access to residues at position 595, 598, 631, or 634. The extreme protease resistance that characterizes the native CPT1 is likely due to intramolecular interactions between either the cytosolic N-and C-terminal domains of the enzyme or between TM1 and TM2. Additionally, it could result from an oligomerization of the enzyme, as reported for porin (16). It is now clear that ␣-helical TM segments of membrane proteins can participate in highly specific interactions that drive their folding and/or oli-gomerization and contribute to an increasingly diverse set of functional roles (35,36). Whether the TM segment(s) of CPT1 fulfill such functional interactions with either each or with other OMM proteins needs to be determined.