An Unusual TOM20/TOM22 Bypass Mechanism for the Mitochondrial Targeting of Cytochrome P450 Proteins Containing N-terminal Chimeric Signals*

Previously we showed that xenobiotic-inducible cytochrome P450 (CYP) proteins are bimodally targeted to the endoplasmic reticulum and mitochondria. In the present study, we investigated the mechanism of delivery of chimeric signal-containing CYP proteins to the peripheral and channel-forming mitochondrial outer membrane translocases (TOMs). CYP+33/1A1 and CYP2B1 did not require peripheral TOM70, TOM20, or TOM22 for translocation through the channel-forming TOM40 protein. In contrast, CYP+5/1A1 and CYP2E1 were able to bypass TOM20 and TOM22 but required TOM70. CYP27, which contains a canonical cleavable mitochondrial signal, required all of the peripheral TOMs for its mitochondrial translocation. We investigated the underlying mechanisms of bypass of peripheral TOMs by CYPs with chimeric signals. The results suggested that interaction of CYPs with Hsp70, a cytosolic chaperone involved in the mitochondrial import, alone was sufficient for the recognition of chimeric signals by peripheral TOMs. However, sequential interaction of chimeric signal-containing CYPs with Hsp70 and Hsp90 resulted in the bypass of peripheral TOMs, whereas CYP27 interacted only with Hsp70 and was not able to bypass peripheral TOMs. Our results also show that delivery of chimeric signal-containing client proteins by Hsp90 required the cytosol-exposed N-terminal 143 amino acids of TOM40. TOM40 devoid of this domain was unable to bind CYP proteins. These results suggest that, compared with the unimodal mitochondria-targeting signals, the chimeric mitochondria-targeting signals are highly evolved and dynamic in nature.

and mitochondria, vary markedly in terms of amino acid sequence, hydrophobicity, and secondary structure and interact with distinctly different sets of carrier proteins, receptors, and protein translocator complexes (3)(4)(5)(6)(7)(8). These observations have led to a widely accepted view that most protein targeting signals are unimodal in nature, which in turn restricts the destination of a given protein to a single subcellular compartment.
ER-targeted proteins contain a distinct N-terminal hydrophobic signal for binding to a signal recognition particle, which in turn targets the emerging nascent chains to the ER (9). With certain exceptions (1,10), ER targeting is thought to be cotranslational. Current models of protein targeting imply that the ER or mitochondrial destination of a protein is determined at the pretranslational level by virtue of the signal sequence that the protein carries. Many mitochondria-and ER-targeted proteins are encoded by a distinct set of genes. In a limited number of cases, characteristic mitochondrial targeting sequences are generated by differential expression of the gene using either alternate transcription/translation start sites or differential splicing of the primary transcripts (11)(12)(13). Thus, proteins targeted to different cytoplasmic organelles carry a specific signal that directs the protein to a specific organelle.
In contrast to the prevailing dogma, studies from our and other laboratories have shown that a number of proteins are bimodally targeted to different subcellular compartments (14 -20). Our results showed that different CYPs, glutathione S-transferase (GST) isoforms, and the amyloid precursor protein (APP) are targeted to mitochondria in addition to their well established subcellular locations in the ER, cytosol, and plasma membranes, respectively (21,22). Bimodal targeting of CYPs and APP was facilitated by the N-terminal chimeric signal, which consists of a cryptic mitochondria-targeting signal immediately flanking the N-terminal ER-targeting signal. By contrast, a C-terminal cryptic mitochondria-targeting signal was critical for the mitochondrial translocation of cytosolic GSTs (21). We also showed that mitochondrial targeting of these proteins required the activation of a cryptic mitochondrial signal either by sequence-specific processing by a cytosolic endoprotease as in the case of CYP1A1 (14) or by PKA-mediated phosphorylation of Ser residues at positions 128 or 129 as in the case of CYP2B1 and CYP2E1 (16,18,23). In contrast, phosphorylation of a tandem PKA/PKC (protein kinase C) tar-get site close to the C terminus was involved in the activation of the mitochondria-targeting signal of GSTA4-4 (21).
Proteins with canonical mitochondria-targeting signals specifically bind to the outer membrane protein import complex, TOM, as the critical initial step in mitochondrial protein targeting (13, 24 -27). The TOM complex consists of nine subunits, of which TOM70, TOM40, TOM22, and TOM20 are the major components. The C termini of TOM20 and TOM70 and the N terminus of TOM22 are exposed to the cytosol and provide receptor sites for binding to client proteins. The small TOMs (TOM7, TOM6, and TOM5), on the other hand, are associated with the channel-forming TOM40 protein, which is embedded in the outer membrane (13,24,26). Current models suggest that the N-terminal signal regions of pre-proteins bind to the cytosol-exposed domains of TOM20 and TOM22 through both hydrophobic and ionic interactions. TOM20 and TOM70 exhibit overlapping substrate specificity. TOM22 is believed to regulate the gating activity of the TOM40 channel (28 -31). A recent study suggests the involvement of Hsp70 in the targeting of phosphorylated CYP2E1 to TOM40 (18). However, the precise mode of transport of chimeric signal-containing proteins to mitochondria and the interaction of chimeric signals with members of the TOM complex remain unclear. A number of studies have observed low levels of mitochondrial protein import by mitochondria pretreated with proteases or those depleted of peripheral TOMs (28,(32)(33)(34)(35), suggesting a distinct role for these peripheral TOMs in the mitochondrial import of proteins. However, the role of these peripheral TOMs in the mitochondrial import of proteins containing chimeric signals remains unclear.
In the present study, we investigated the requirement for peripheral TOM proteins for the translocation of CYP MT2a (ϩ5/1A1), CYP MT2b (ϩ33/1A1) (the N-terminal truncated forms of CYP1A1), CYPMT4, and CYPMT5 (phosphorylated forms of CYP2B1 and 2E1, respectively) that contain N-terminal chimeric signals. We found that the activated chimeric signals of these proteins can bypass the peripheral TOM20 and TOM22 receptors. Our results also suggest that differential interaction of chimeric signals with Hsp70 and Hsp90 might be the key discriminating factor in the TOM20/TOM22 bypass mechanism of mitochondrial protein targeting.
Construction of Expression Plasmids-Rat CYP2E1, 2B1, ϩ5/1A1, and ϩ33/1A1 cDNAs were generated by reverse transcriptase-PCR and cloned in pGEM7zf (Promega Biotech, Madison, WI) plasmid vectors. Full-length rat liver cDNA to TOM40 was purchased from ATCC (IMAGE clone). N-terminal truncated TOM40 cDNA (⌬143TOM40) was generated by PCR amplification using full-length TOM40 DNA as a template. Both full-length and truncated TOM40 constructs were cloned in pGEM7zf vector and pET vectors to carry out expression in vitro and bacterial cells, respectively.
Disruption of Outer Membrane Receptors by Protease Treatment-Rat liver mitochondria were isolated by differential centrifugation in H medium (5 mM HEPES (pH 7.4), 210 mM mannitol, 70 mM sucrose, 1 mM EDTA), as described previously (38) and washed three times with H medium. Mitochondria were subjected to limited Pronase treatment (30 g/mg protein) at 4°C for 30 min. Following the addition of a 10-fold excess of protease inhibitor, the mitochondrial pellet was reisolated by passing through 1.2 M sucrose and used for protein import as described below.
In Vitro Import of Labeled Proteins into Mitochondria-In vitro translation products were generated in the transcriptionlinked translation RRL system (Promega, Madison, WI) in the presence of added [ 35 S]Met (40 Ci/50-l reaction, 1175 Ci/mmol, PerkinElmer Life Sciences) according to the manufacturer's recommended protocol. In indicated experiments, CYP2B1 and CYP2E1 translation products were phosphorylated for 45 min at 30°C by supplementing the translation mix with 10 units/100 l of PKA (Sigma) and 100 M ATP before being used for import assays. Import reactions were performed in 200-l final volumes containing 200 g of mitochondria at 28°C for 60 min. In some experiments the reaction volume for import assays was increased to 500 l using 500 g of mitochondria. After import, the samples were treated with 200 g/ml trypsin for 20 min at 4°C. Trypsin-treated and untreated samples were mixed with trypsin inhibitor (10ϫ molar excess), and mitochondria were recovered by sedimentation through 1 M sucrose as described previously (18). Mitochondrial proteins were solubilized in 2ϫ Laemmli sample buffer (42) for 10 min at 75°C and analyzed by SDS-PAGE and fluorography.
Effects of Immunoinhibition of Membrane Translocases on Mitochondrial Import-Rat liver mitochondria in H medium containing 1 mM phenylmethylsulfonyl fluoride were incubated with antibodies to TOM20, TOM22, TOM70, TOM40, porin, or preimmune serum at 30°C for 30 min. The reaction mixtures were centrifuged at 5,000 ϫ g for 5 min to remove excess antibody, and the resultant mitochondria were washed twice with H medium and used in import assays.
Immunodepletion of Hsp70 and Hsp90 from the Rabbit Reticulocyte Lysate-For immunodepletion, RRL (200 l each) was incubated with 20 g of Hsp70 and/or Hsp90 antibodies overnight (12-16 h) at 4°C. Immune complexes were pulled down by incubation with protein A-agarose (1 h at 4°C) followed by centrifugation at 5000 ϫ g for 5 min. In parallel experiments, RRL was incubated with preimmune IgG (20 g/200 l RRL) as antibody controls. The depleted RRL were used for in vitro translation as described above.
Interaction of CYPs with Cytosolic Hsp70 and Hsp90-Interaction of CYP proteins with cytosolic Hsp70 under in vitro and in vivo conditions was tested by co-immunoprecipitation. For studying in vitro interactions, nascent proteins were synthesized in the presence (for CYP2E1 and CYP2B1) and absence (for ϩ/51A1, ϩ33/1A1, and CYP27) of PKA in the RRL system. Interaction of CYP proteins with cytosolic Hsp70 or Hsp90 present in the lysate under these conditions was estimated by co-immunoprecipitation. Reaction mixtures were immunoprecipitated with polyclonal antibodies against the respective CYP proteins. One portion of the immunoprecipitate was resolved by SDS-PAGE on a 10% gel and subjected to immunoblot analysis with anti-Hsp70 and -Hsp90 antibodies. The other portion was subjected to SDS-PAGE, and gels were subjected to autoradiography.
Bacterial Expression of TOM40 and Binding of CYPs to Reconstituted TOM40-Full-length and truncated ⌬143TOM40 expressed in Escherichia coli BL21 cells were purified from inclusion bodies to near homogeneity and used for reconstitution in liposomes. Liposomes were prepared from azolectin (Sigma, type IVS) in MOPS-Tris (pH 6.9) buffer using a Branson sonifier as described previously (39). Liposomes were freeze-thawed three times and solubilized by the addition of n-octyl glucopyranoside (6% v/v) on ice for 30 min. Solubilized liposomes (10 mg) were mixed with 9 -300 g of TOM40 and diluted 2-fold with 10 mM MOPS-Tris (pH 6.9) containing 0.5 mM phenylmethylsulfonyl fluoride and 0.5 mM dithiothreitol. Reconstitution was performed by dialysis of the mixture at 4°C for 24 h against the same buffer. 35 S-Labeled and phosphorylated RRL-translation products (80,000 -100,000 cpm) were incubated for 1 h at 30°C with vesicles containing full-length and truncated TOM40 in a buffer containing 20 mM KCl and 2% bovine serum albumin. Liposomes were reisolated by sedimenting through a 0.4 M sucrose layer by centrifugation at 125,000 ϫ g for 1 h. Client 35 S-labeled proteins bound to TOM40 vesicles were eluted with 10 mM Tris (pH 7.4) containing 10 -80 mM NaCl. Liposomes and eluted proteins were subjected to SDS-PAGE on a 12% gel, and the amount of radioactivity in the liposomal and post-liposomal fractions was determined by radiometric imaging through the STORM system (GE Healthcare).
Measurement of Pore Function of Liposome-reconstituted TOM40 Protein-The orientation and function of reconstituted TOM40 pore was investigated by measuring the release of [ 14 C]sucrose and retention of [ 3 H]dextran (70 kDa) from the proteoliposomes containing wild-type (WT) or N-terminal truncated ⌬143 TOM40 proteins as described (40). As a negative control, purified P4501A1, which does not form a channel, was used following reconstitution in lipid vesicles. Reconstituted vesicles were suspended in 2 mM HEPES buffer (pH 7.4) containing 20 mM NaCl, 2 mM MgCl 2 , 0.3 mM NaN 3 , and 2 mM sucrose. Reaction mixtures were incubated with equal amounts (ϳ100,000 cpm) of [ 14 C]sucrose (7.5 mCi/mmol, American Radiolabeled Chemicals Corp.) or [ 3 H]dextran (55 mCi/mmol; American Radiolabeled Chemicals Corp.) at room temperature for 30 min. Following vortexing, tubes were placed in a bathtype sonic oscillator for 10 s. The tubes were incubated at 45°C for 30 min and allowed to cool down to room temperature. The reaction mixture was passed through a Sepharose 4B column; the vesicles were eluted with 10 mM HEPES-KOH buffer (pH 7.4) containing 100 mM NaCl, 0.1 mM MgCl 2 , and 0.3 mM NaN 3 and assayed for the amount of radioactivity retained by using a Beckman 5000 liquid scintillation counter.
Preparation of Mitochondria from tom-deleted Yeast Strains-Mitochondria were isolated from both the control and tomdeleted yeast strains essentially as described previously (41). Steady-state levels of mitochondrial proteins were analyzed by separation via SDS-PAGE and immunoblotting.
SDS-PAGE and Immunoblot Analysis-Proteins were resolved by SDS-PAGE and transferred to nitrocellulose membranes for immunoblot analysis as described previously (42). Polyclonal antibodies against CYP2E1, CYP2B1, CYP1A1, and monoclonal antibody to CYP27, as well as monoclonal antibodies against cytosolic Hsp70 and Hsp90 (Sigma), were used. Immunoblots were developed using a chemiluminescence SuperSignal ULTRA kit (Pierce), and the blots were imaged and quantitated using a Fluor-S imaging system (Bio-Rad). Fig. 1A shows the N-terminal chimeric signals of CYP 1A1, CYP 2B1, and CYP 2E1 that encompass the N-terminal transmembrane domain and the cryptic mitochondria-targeting signal sequences. We have shown previously that mitochondrial import of these CYPs is dependent on transmembrane potential and energy (14, 16 -18). In the present study, we initially tested whether the major peripheral TOM receptors were needed for the mitochondrial import of these CYPs. For this purpose, rat liver mitochondria were subjected to limited Pronase treatment, which is known to remove cytosol-exposed regions of the proteins (32). Fig. 1B shows that Pronase treatment degraded most of the cytosol-exposed regions of TOM70, TOM20, and TOM22. This treatment protected the 7-, 3-, and 7-kDa putative transmembrane segments that were detected by antibodies raised against full-length TOM70, TOM20, and TOM22, respectively. Under these treatment conditions nearly full-length 40-kDa TOM40 protein and the 30-kDa porin protein were protected. Although not shown, the mitochondrial inner membrane protein, cytochrome c oxidase I (CcOI), and matrix protein Hsp70 were also fully protected. Using these protease-treated mitochondria, we tested the import efficiency of CYP proteins. CYP27 (43), which contains a canonical (N-terminal cleavable) mitochondria-targeting signal, was used as a positive control. The results of in vitro import showed that nearly intact 2B1 protein was imported and rendered resistant to Pronase treatment, whereas 2E1 and ϩ5/1A1 were not imported into Pronase-treated mitochondria (Fig. 1C). Fig. 1D shows that ϩ33/ 1A1 was imported efficiently into Pronase-treated mitochondria, whereas CYP27 was not. These results suggest that different chimeric signal-containing proteins have different requirements for peripheral TOM proteins.

Distinctive TOM Protein Requirements for the Import of Chimeric Signal-containing CYPs-
The results of import with protease-treated mitochondria were further ascertained using two parallel approaches. In the first approach, we used immunoinhibition of individual TOM proteins with saturating levels of specific antibodies. As shown in Fig. 2, A-E, the TOM40 antibody blocked the import of all five proteins in rat liver mitochondria. Import of CYP2B1 and ϩ33/1A1 was not blocked by antibodies to the peripheral TOM proteins, TOM70, TOM22, and TOM20 (Fig. 2, A and B). Import of CYP 2E1 and ϩ5/1A1 was blocked by TOM70 antibody ( Fig. 2, D and E) but not by antibodies to either TOM20 or TOM22. Also, the import of ϩ33/1A1 and 2E1 in the presence of TOM22 antibodies was incomplete as seen by relatively shorter protease-protected products (Fig. 2, B and D). The import of CYP27 was affected by all antibodies tested (Fig. 2C). Antibody to porin was used as a negative control and had no effect on the import of any of the proteins tested (Fig. 2

, A-E).
The specificity of the TOM40 antibody effect observed above was ascertained using Bcl-XL protein, which is targeted to the mitochondrial outer membrane in the absence of functional TOM40 protein (44). Fig. 2F shows that, as expected, TOM40 antibody and also porin antibody had no significant effect on the membrane association of Bcl-XL. The transmembrane orientation of the protein is apparent from its relative insolubility in Na 2 CO 3 . Furthermore, trypsin treatment caused substantial reduction in protein level as expected of a protein which is mostly exposed outside of the outer membrane (Fig. 2F).
In the second approach, we used mitochondria from yeast strains that contained selectively deleted tom genes. As seen from Fig. 3A, ϩ33/1A1 was imported into yeast mitochondria that were devoid of TOM70, TOM20, and TOM22 proteins at the same level as in WT mitochondria. However, the proteaseprotected fragment in TOM22-deficient mitochondria was shorter by almost 3-4 kDa. Similarly, CYP2B1 was imported into mitochondria devoid of all three TOM proteins at a level similar to that in WT mitochondria (Fig.  3B). On the other hand, ϩ5/1A1 and CYP2E1 were not imported efficiently by mitochondria deficient in TOM70, suggesting a critical requirement for TOM70 in the import of these two proteins (Fig. 3,  C and D). Additionally, the protease-protected fragment of CYP2E1 was shorter (ϳ1 kDa) in mitochondria deficient in TOM22 (Fig. 3D). Finally, CYP27 was not imported significantly in mitochondria deficient in any of the three TOMs (Fig.  3E). The topology of incompletely translocated CYP2E1 and ϩ33/1A1 in tom22⌬ mitochondria were tested by treatment with digitonin prior to treatment with trypsin. The results shown in Fig. 3F reveal that digitonin treatment rendered both proteins completely sensitive to trypsin digestion, suggesting that imported CYP2E1 and ϩ33/1A1 in tom22⌬ mitochondria are located in the intermembrane spaces. The intermembrane space location of 2E1 and ϩ33/1A1 in Fig. 3F was further confirmed in Fig. 3G, showing that cytochrome c, an intermembrane space protein, is sensitive to both digitonin and digitonin plus trypsin treatment (Fig. 3G) in these mitochondria. However, mitochondrial inner membrane protein cytochrome c oxidase subunit IVi1 (CcOIVi1) and matrix protein Hsp70 were resistant to these treatments.
As shown in Fig. 4, A-C, the rate of in vitro import of ϩ33/ 1A1 in tom70⌬, tom20⌬, and tom22⌬ mitochondria up to 50 min of incubation were nearly similar and compared well with the rate of import into WT yeast mitochondria. The rate of import of CYP2E1 was nearly comparable in the WT, tom22⌬, and tom20⌬ mitochondria, although there was no significant import into tom70⌬ mitochondria. The results shown in Figs. 2-4 suggest that different chimeric signal-containing proteins have notably different requirements for peripheral TOMs. These results also show that TOM22 may play a significant role in the complete translocation of CYP ϩ33/1A1 and 2E1 proteins through TOM40.
Binding of Chimeric Signal-containing Proteins to Hsp70 Is a Prerequisite for Binding to Hsp90-A marked difference in the requirement for different TOM proteins suggested a possible bypass mechanism for the import of certain chimeric signalcontaining proteins. These results prompted an investigation into the role of other cytosolic factors on the ability of chimeric signals to bypass peripheral TOMs on their way to mitochondrial entry. Recent studies from our laboratory and others indicated a possible role for cytosolic Hsp70 in the import of mitochondria-targeted proteins (18). In this study, we investigated FIGURE 1. Import of CYP proteins into protease-treated mitochondria. A, schematic representation of chimeric signals of CYP1A1, 2B1, and 2E1 with N-terminal ER-targeting domain followed by the positively charged mitochondria-targeting domains. Arrows over the CYP1A1 schematic indicate processing sites at the 4th and 32nd residues. B, isolated rat liver mitochondria were treated with or without Pronase as indicated under "Experimental Procedures." The TOM components (full-length and fragments) were detected by immunoblot analysis (30 g of protein each) using subunit-specific antibodies. C and D, 35 S-labeled CYPs translated in a RRL system were used for import into Pronase-treated (P-mito) and untreated control (C-mito) mitochondria. The extent of import was determined by resistance to treatment with trypsin (200 g/ml); mitochondria were pelleted through 1 M sucrose, and 250 g of protein in each case was analyzed by SDS-PAGE on a 14% gel and subjected to fluorography. p and m indicate the precursor and mature form of CYP27, respectively. the ability of different proteins to bind Hsp70 and Hsp90 by a co-immunoprecipitation method. In vitro translated CYP ϩ5/1A1, ϩ33/1A1, 2B1, 2E1, and CYP27 in RRL were immunoprecipitated with CYP-specific antibodies, and the immunoprecipitates were then probed with Hsp70 and Hsp90 antibodies using an immunoblot method. The immunoblot shown in Fig. 5A reveals that CYPϩ5/1A1, ϩ33/1A1, 2E1, and 2B1 interacted with both Hsp70 and Hsp90, whereas CYP27 interacted only with Hsp70. Additionally, the S128A and S129A phosphorylation site mutants of CYP2B1 and CYP2E1 (S/A mutants), respectively, interacted weakly with Hsp70 but did not interact with Hsp90. These latter results are consistent with our observation that PKA-mediated phosphorylation is important for efficient mitochondrial import of CYP 2B1 and 2E1 (16,18,23). The fluorogram shown in Fig. 5B shows the levels of 35 S-labeled client proteins in the immunoprecipitates shown in Fig. 5A.
The possibility of a ternary complex formation between the client protein and Hsp70 and Hsp90 was tested by co-immunoprecipitation. Fig. 5C shows that immunoprecipitation of RRL containing 35 S-labeled ϩ33/1A1 with Hsp70 antibody resulted in the pulldown of the client protein and Hsp90. Immunopre-cipitation of RRL devoid of client protein with Hsp70 antibody, on the other hand, immunoprecipitated only Hsp70. These results confirm the formation of a ternary complex consisting of the client protein and the two Hsp chaperones.
We next investigated whether the binding of various CYPs to Hsp70 and Hsp90 are interdependent or independent of each other. Various CYPs were translated in RRLs depleted of either Hsp90 or Hsp70 and subjected to immunoprecipitation with CYP-specific antibodies (Fig. 5, E and F). The immunoprecipitates were then probed with Hsp-specific antibodies by immunoblot analysis. Fig. 5D shows the levels of Hsp70 and Hsp90 in control and immunodepleted RRL used in this experiment. Fig.  5E shows that in the absence of Hsp90, all of the CYPs tested were able to interact with Hsp70. Surprisingly, these CYPs failed to co-immunoprecipitate Hsp90 in Hsp70-depleted lysates (Fig. 5F). However, supplementation with purified Hsp70 restored CYP binding to Hsp90. The only exception was CYP27, which did not show any significant interactions with Hsp90 either in complete or Hsp70-depleted extracts. As seen in Fig. 5G, the level of translation of all CYP proteins was comparable in Hsp70-depleted reticulocyte lysates. Although not shown, Hsp90 depletion also did not adversely affect the translation of CYP proteins. These results show that the interaction between chimeric signal-containing proteins and Hsp90 is dependent on the presence of Hsp70, suggesting a degree of cooperativity between the two factors.
Roles of Hsp70 and Hsp90 in Modulating the TOM Bypass Mechanism and Translocation-In a previous study, we showed that the binding of CYP2E1 to Hsp70 was important for the mitochondrial targeting of the protein (18). In an extension of this study, we have now investigated the roles of Hsp70 and Hsp90 chaperones in the mitochondrial targeting of different chimeric signal-containing proteins and their propensity to bypass peripheral TOM proteins. For this purpose, we translated CYP2B1, ϩ5/1A1, and CYP27 proteins in RRL depleted of Hsp70, Hsp90, or both. The translated products were used for import into mitochondria isolated from various TOM proteindeleted yeast cells. In each case, the effects of added Hsp70 and/or Hsp90 were tested. The results of an in vitro import assay (Fig. 6A) revealed that depletion of Hsp70 and Hsp90 completely abolished CYP2B1 import into WT as well as TOM70-, TOM20-, and TOM22-deleted mitochondria. The addition of Hsp90 alone did not rescue the import defect. However, the addition of Hsp70 alone was able to restore import of CYP2B1 into both WT and tom70⌬ mitochondria. Supplementation with both Hsp70 and Hsp90 was required for restoration of the import of CYP2B1 into both tom20⌬ and tom22⌬ mitochondria. Although not shown, ϩ33/1A1 showed a similar pattern of import into TOM-deficient mitochondria.
The import of CYP ϩ5/1A1 into tom-deficient mitochondria was somewhat different in that the addition of Hsp70 and Hsp90 did not restore the import of this protein. Furthermore, both Hsp70 and Hsp90 were required for import of this protein into tom20⌬ and tom22⌬ mitochondria (Fig. 6B), whereas mitochondria from tom70⌬ cells were unable to import ϩ5/1A1 under all conditions tested. Although not presented, CYP2E1 showed an identical requirement for TOM70 for import. CYP27 showed a distinctly different requirement (Fig.   FIGURE 2. Effects of antibody blocking of peripheral TOMs on the mitochondrial import of proteins with different chimeric signals. Isolated rat liver mitochondria were preincubated with anti-rat TOM40, anti-rat TOM20, anti-rat TOM22, anti-rat porin, or anti-rat TOM70 antibodies at 30°C for 30 min. Mitochondria were reisolated and used for the import of 35 S-labeled CYP2B1 (A), ϩ33/1A1 (B), CYP27 (C), CYP2E1 (D), and ϩ5/1A1 (E). In F, mitochondria preincubated with antibodies to porin and TOM40 were used of the import of 35 S-labeled Bcl-XL protein. Mitochondria were extracted with alkaline Na 2 CO 3 as described previously (14,15). Details of import and protease treatment were as described under "Experimental Procedures." Mitochondrial proteins (250 g each) were resolved by SDS-PAGE, and the gels were subjected to fluorography. 6C). First, Hsp90 depletion did not have a significant effect on the import of CYP27 into WT mitochondria. Second, deletion of tom70, tom20, or tom22 almost completely inhibited the import of CYP27 (Fig. 6C). These results taken together suggest that Hsp70 and Hsp90 are important factors that facilitate the presentation of different chimeric signals to the TOM complex. They also play a major role in the ability of chimeric signals to bypass peripheral TOMs. The results also suggest different requirements for peripheral TOMs in terms of the import of the four chimeric signal-containing proteins tested.

Targeting of CYP Proteins to Mitochondria in Intact Cells
Involves TOM Bypass Mechanism-The occurrence of a TOM bypass mechanism for the mitochondrial targeting of chimeric signal-containing CYPs was confirmed by the expression of various CYPs in control (parent strains of tom70 and tom20 null mutant) and tom null mutant yeast cells. Immunoblot analysis of mitochondrial proteins in Fig. 7A shows that mutant strains deficient in a given peripheral TOM, for example, tom70⌬ cells, expressed intact peripheral TOM20 and TOM22 proteins at levels comparable with the wild-type yeast. Thus, the deficiency in these strains was restricted to the specified gene product.
The tom22 null mutation is known to induce o status. To assess the effects of the o mutation on mitochondrial-protein targeting, we expressed different CYP cDNAs in the respective parent strain made o by ethidium bromide treatment (Rho 0 control). As shown in Fig. 7B, CYP ϩ33/1A1 was targeted efficiently to mitochondria in both controls and tom null mutant cells. The protein was inefficiently targeted to the ER, as expected of a protein that lacks the SRP (signal recognition particle)-binding signal domain. CYP ϩ5/1A1, on the other hand, showed a distinct requirement for TOM70, as no detectable protein was found in the mitochondrial compartment of tom70⌬ cells. These results essentially confirm the in vitro data (Fig. 6), which suggest the distinctive nature of chimeric signals of CYP2E1 and ϩ5/1A1 requiring TOM70 for mitochondrial translocation. Although not shown, CYP2B1

. Import of CYP proteins into yeast mitochondria with deleted TOM subunits.
A-E, mitochondria were isolated from wild-type, tom70⌬, tom20⌬, or tom22⌬ yeast strains. 35 S-Labeled proteins (ϩ33/1A1 (A), 2B1 (B), ϩ5/1A1 (C), 2E1 (D), CYP27 (E)) were incubated with mitochondria from either wild-type or TOM deletion strains as indicated. Following import, mitochondria were treated with trypsin, reisolated, and analyzed by SDS-PAGE and fluorography. The radioactive proteins bound to mitochondria and the proteaseprotected fractions were quantified for calculations of the percent import (values presented below the lanes). F, 35 S-labeled ϩ331A1 and 2E1 were imported into tom22⌬ mitochondria as described for A-E and resolved on a higher resolution gradient gel. In G, trypsin-and digitonin-treated fractions were also analyzed for marked proteins by immunoblot analysis using the indicated antibodies. A and D, 4 indicates the shorter protected fragment. FIGURE 4. Rates of in vitro import of ؉33/1A1 and CYP2E1 proteins into mitochondria from tom-deleted yeast strains. Import of 35 S-labeled proteins was carried out as described in the legend for Fig. 3 for 0 -50 min. Trypsin-treated mitochondria were subjected SDS-PAGE analysis (A, B, D, and E), and the gels were imaged through a STORM radiometric imager. The rate of import as a function of time was plotted in C and F. behaved similarly to ϩ33/1A1 and was targeted efficiently to mitochondria in tom70⌬ cells. Furthermore, the mitochondrial targeting of all three proteins occurred efficiently in tom20⌬ and tom22⌬ cells, although the overall level of expression in both mitochondrial and microsomal compartments was relatively low in tom20⌬ cells. Surprisingly, the level of expression of ϩ5/1A1 and 2E1 was generally higher in both Rho 0 control and tom22⌬ cells, possibly reflecting strain specific differences.
The topology of mitochondria-and microsome-associated CYPs in tom null mutants was tested by limited digestion with trypsin followed by immunoblot analysis. As shown in Fig. 7C, mitochondria-associated CYPs were resistant to trypsin digestion, whereas microsome-associated CYPs were sensitive to trypsin treatment (Fig. 7D). In this respect, mitochondria-associated CYPs exhibited resistance to trypsin similar to the matrix-associated protein pSU9-DHFR used as a control. Confirming the results in Figs. 3, 4, and 6, CYP2E1 and ϩ5/1A1 were not targeted to mitochondria significantly in tom70⌬ cells. Also, the trypsin-protected fragments of CYP2E1 and ϩ331A1 were shorter in tom22⌬ cells compared with o control cells. These results confirm that mito-chondria-associated proteins are probably localized inside the inner membrane compartment, whereas the microsome-associated proteins are exposed externally to the ER in a protease-accessible orientation. These results together confirm the occurrence of the TOM20/TOM22 bypass proteintargeting mechanism in vivo.
We used pSU9/DHFR as a control for matrix-targeted protein, because Stan et al. (45) have shown that the protein is efficiently transported in tom70⌬ cells. In contrast, results from other groups (46,47) show that deletion or depletion of TOM70, TOM20, and TOM22 affects mitochondrial import of pSU9/DHFR. However, we did not observe significant inhibition of mitochondrial import of this protein in tom70⌬ and tom22⌬ cells under either in vivo (Fig. 7C) or in vitro (not shown) conditions.
Association of Hsp90 with Reconstituted TOM40 during Delivery of ϩ33/1A1 Protein-It is known that a number of mitochondria-destined proteins with canonical targeting signal require functional TOM70 receptor protein for import. In these cases Hsp90 has been shown to physically associate with TOM70 for the delivery of the client protein to the translocase complex (48,49). The in vitro and in vivo experiments in this study  suggested that in the presence of Hsp70 and Hsp90, ϩ33/1A1 and CYP2B1 proteins are translocated to mitochondria in the absence of TOM20, TOM22, and TOM70. To understand the molecular basis of this bypass mechanism, we tested the ability of Hsp90 to bind to purified and reconstituted TOM40 under in vitro conditions. Our preliminary results with antibody pulldown assays showed that 35 S-labeled ϩ33/1A1 efficiently bound to WT TOM40 but not to ⌬143 TOM40 lacking the N-terminal cytosol exposed Pro-rich domain. We have therefore used purified and reconstituted WT and ⌬143 TOM40 proteins in lipid vesicles for binding in these studies. Fig. 8A (left panel) shows the Coomassie Bluestained gel profiles of purified WT and ⌬143 TOM40 proteins. The reconstituted lipid vesicles were recovered by sedimentation at 125,000 ϫ g for 1 h, and both the vesicular fraction (Fig.  8A, middle panel) and the supernatant representing "free" proteins (right panel) were analyzed by polyacrylamide gel electrophoresis. The middle and right panels (Fig. 8A) show that a large fraction of proteins in both cases is associated with lipid vesicles and a relatively small fraction is recovered in the supernatant fraction. Another control experiment, in Fig. 8B, shows that  35 S-labeled wild-type, ϩ5/1A1, ϩ33/1A1, 2E1, 2B1, and CYP27 proteins and phosphorylation site mutant 2E1S/A (S129A CYP2E1), and 2B1S/A (S128A CYP2B1) proteins in RRL were subjected to immunoprecipitation (IP) with CYP-specific antibodies (Ab) as indicated. Half of the immunoprecipitate was analyzed by SDS-PAGE and probed with anti-Hsp70 and -Hsp90 antibodies (A). The other half was analyzed by SDS-PAGE and fluorography (B). In C, RRL with 35 S ϩ 33/1A1 and that incubated with empty vector were immunoprecipitated with Hsp70 antibody. Half of the immunoprecipitate was subjected to immunoblot analysis using Hsp70 plus Hsp90 antibodies (upper panel), and the other half was resolved on the gel and subjected to fluorography (lower panel). D, levels of Hsp70 and Hsp90 in immunodepleted RRL. E, various CYP proteins were translated in Hsp90-depleted RRL and immunoprecipitated with CYP-specific antibodies. The immunoprecipitates were subjected to immunoblot analysis with anti-Hsp70 plus anti-Hsp90 antibodies. F, various CYP proteins were translated in Hsp70-depleted or Hsp70-replenished RRL and immunoprecipitated with CYPspecific antibodies. The immunoprecipitates were probed with a combination of anti-Hsp70 plus -Hsp90 antibodies. Immunodepletion of lysates was carried out as described under "Experimental Procedures." G, autoradiogram representing the 35 S-labeled products of CYPs translated in reticulocyte lysates depleted of Hsp70. JULY 11, 2008 • VOLUME 283 • NUMBER 28

JOURNAL OF BIOLOGICAL CHEMISTRY 19775
WITHDRAWN only vesicles containing TOM40 protein binds efficiently to ϩ33/1A1 in complete RRL, whereas vesicles devoid of TOM40 fail to bind to the protein indicating the specificity.
The proper orientation of reconstituted TOM40 and ⌬143 TOM40 proteins and the presence of TOM40 pore was tested using [ 14 C]sucrose and [ 3 H]dextran (ϳ70kDa) as permeable and impermeable substrates, respectively. As shown in Table 1, reconstituted vesicles with CYP1A1 showed nearly complete retention of both 14 C and 3 H radioactivity, suggesting no pore activity. Vesicles containing increasing amounts of intact TOM40 as well as ⌬143 TOM40 proteins showed increasing release of [ 14 C]sucrose but complete retention of [ 3 H]dextran, suggesting efficient pore activity.
In Fig. 8C, we tested the efficiencies of two types of chimeric signal-containing proteins, ϩ5/1A1 and ϩ33/1A1, for binding to reconstituted vesicles containing WT TOM40 protein. 35 S-Labeled client proteins translated in intact RRL were incubated with lipid vesicles at 30°C for 20 min; the vesicles were recovered by centrifugation at 125,000 ϫ g for 1 h, washed twice with buffer containing 10 mM NaCl, and used for analysis. The autoradiogram in Fig. 8C, left panel, shows that DHFR, a cytosolic protein and ϩ5/1A1 did not bind significantly to TOM40 vesicles. The inability of ϩ5/1A1 protein to bind to TOM40 may be due to a lack of TOM70 in the vesicle. The CcOVb subunit (Fig.  8C, left panel) and also ϩ33/1A1 (right panel) bound efficiently to TOM40 in the absence of other TOMs. These results are consistent with our in vitro and in vivo results that only ϩ33/ 1A1, but not ϩ5/1A1, can bypass all of the peripheral TOMs. In Fig. 8D we ascertained the roles of Hsp70 and Hsp90 for the ϩ33/1A1 protein binding to reconstituted TOM40 vesicles using an Hsp70-and Hsp90-depleted RRL system. 35 S-Labeled ϩ33/1A1 protein was bound to TOM40 vesicles, and the vesicles were reisolated, washed twice, and analyzed by SDS gel electrophoresis. The autoradiogram in Fig. 8D shows that ϩ33/1A1 did not bind to TOM40 in the absence of Hsp70 or Hsp90. Also, both Hsp90 and Hsp70 induced the binding in a concentration-dependent manner. These results further confirm the need for both Hsp70 and Hsp90 for support of the bypass pathway.
In Fig. 8E we tested whether Hsp90 could physically associate with the reconstituted TOM40 vesicles for delivery of the ϩ33/1A1 protein and whether the binding requires the cytosol-exposed region of TOM40. In this experiment, HSP90-depleted RRL was used for translating ϩ33/1A1 and Hsp90 proteins together in the presence of [ 35 S]Met and used for binding to vesicles containing TOM40 and ⌬143 TOM40 proteins. The vesicles were recovered and washed sequentially with buffers containing 10, 40, and 80 mM NaCl, and both the eluted proteins and that remaining with the vesicles (resid-FIGURE 6. Role of Hsp70 and Hsp90 in the import of CYP proteins into mitochondria with deleted TOM subunits. CYP proteins were translated in RRL depleted of Hsp70, Hsp90, or both and used for in vitro import into wildtype or TOM subunit-deficient mitochondria. RRL were immunodepleted with anti-Hsp70 and -Hsp90 antibodies as described under "Experimental Procedures." 35 S-Labeled 2B1 (A), ϩ51A1 (B), and CYP27 (C) were used for import into mitochondria from wild-type, tom70⌬, tom20⌬, or tom22⌬ yeast cells. The import assays were carried out with or without the addition of purified Hsp70 and Hsp90. The details of the import reaction and treatment with Pronase and protease inhibitor were described in the legend for Fig. 2. Proteins (250 g each) were subjected to SDS-PAGE and fluorography. p, precursor; m, mature protein. FIGURE 7. Mitochondrial targeting of chimeric signal-containing CYPs in tom70⌬, tom20⌬, and tom22⌬ yeast cells. Wild-type and tom null mutant strains were transformed with CYPϩ5/1A1, ϩ33/1A1, 2E1, or su9-DHFR. A, TOM40, TOM70, TOM20, and TOM22 levels of various deletion strains. Immunoblots were carried out using 40 g of protein each. B, mitochondrial and microsomal fractions of cells expressing different CYPs or Su9-DHFR were subjected to immunoblot analysis (30 g each) using CYP-specific antibodies as indicated. C, mitochondria from tom70, tom22, tom70, or tom22 null mutant cells expressing different CYPs or Su9-DHFR were subjected to trypsin treatment, and proteins were subjected to immunoblot analysis as described in A. D, microsomal fractions from tom70⌬, tom70⌬ parent, tom22⌬, and tom22⌬ parent strains expressing ϩ5/1A1 and 2E1 were treated with or without trypsin and subjected to immunoblot analysis using CYP-specific antibodies. ual) were analyzed for assessment of the type of association of Hsp90 and ϩ33/1A1 with TOM40. Fig. 8E shows that both the ϩ33/1A1 and Hsp90 bound to WT TOM40, and most of the bound Hsp90 protein was eluted with 40 mM NaCl (Fig. 8, E and F), whereas elution of ϩ33/1A1 required 80 mM NaCl, suggesting a more stable interaction of the latter with TOM40 (Fig. 8, E and G). Additionally, Hsp90 and ϩ33/1A1 proteins bound at markedly reduced levels with ⌬143 TOM40 (Fig. 8, E-G), suggesting that the binding requires the intact cytosol exposed Pro-rich domain of TOM40. These results provide insight into the mechanism of the bypass pathway, which involves direct binding of Hsp90 for delivery of the client protein to TOM40.

DISCUSSION
Mitochondrial targeting signals are heterogeneous, ranging from ␣-helical, ␤-sheet, or unstructured (13). The most common type of mitochondrial signals are pre-sequences, which consist of 15-40 N-terminal amino acids that form amphiphilic helices with positively charged residues lining one side of each helix. These signals are commonly observed in mitochondrial matrix-and inner membrane-targeted proteins and cleaved by matrix metalloprotease(s) after the protein is translocated into the mitochondrial compartment (13). The pre-sequences are critically important for the binding of precursor proteins to mitochondrial import receptors, TOMs, and translocases of inner mitochondrial membrane (50). The second category of mitochondria-targeting signals is internal; these reside deep inside the protein (51). A third type of signal is a variation of the internal signal, where a cryptic mitochondrial signal follows immediately downstream of an N-terminal transmembrane domain. This type of targeting signal is known to occur in a number of membrane-anchored proteins such as Bcs1, Tim14, Mdj2, D-AKAP1, and others (13,(51)(52)(53)(54)(55)(56). We have described a  35 S-labeled ϩ33/1A1 protein (40,000 cpm) was incubated with lipid vesicles (5 mg each) containing TOM40 or that lacking TOM40 for 20 min at 30°C. The vesicles were recovered by pelleting through 0.4 M sucrose as described in C and subjected to SDS-PAGE and fluorography. C, 35 S-labeled client proteins, DHFR, CcOVb, and ϩ5/1A1 (left panel) and ϩ33/1A1 (right panel) (50,000 cpm each) in intact RRL were used for binding to lipid vesicles containing 2 g of WT TOM40 protein for 20 min at 30°C. Liposomes were recovered by layering on 0.4 M sucrose followed by centrifugation at 125,000 ϫ g for 1 h and then washed twice with 20 mM Tris-HCl (pH 7.4) and 10 mM NaCl; equal portions were analyzed on two companion polyacrylamide gels. One gel was subjected to fluorography (upper panels), and the second gel was subjected to immunoblot analysis using TOM40 antibody (lower panels). D, effect of Hsp90 and Hsp70 on binding of ϩ33/1A1 to TOM40 vesicles. ϩ33/1A1 protein was translated in Hsp90-and Hsp70-depleted RRLs, and 50,000 cpm of each were incubated with liposomes containing WT TOM40 in the presence or absence of various amounts of purified Hsp90 and Hsp70 as indicated. In E, Hsp90 and ϩ33/1A1 proteins were co-translated in Hsp90-depleted RRL, and 50,000 cpm of each were incubated with liposomes containing WT and ⌬143 TOM40 proteins (2 g protein each) as described in C. Liposomes were recovered and sequentially washed with Tris-HCl buffer containing 10, 40, and 80 mM NaCl. The washes and also residual proteins (RES) in the vesicle were analyzed by SDS-PAGE and fluorography. F and G, quantification of Hsp 90 (F) and ϩ33/1A1 (G) protein bands from autoradiogram in E. The Input lanes represent 20% of total counts used for binding. NS, non-specific. fourth type of mitochondria-targeting signal, designated as a chimeric signal, based on its ability to direct the same primary translation product to two different subcellular compartments, namely, microsomes and mitochondria or cytosol and mitochondria (16,18,21,22). The chimeric signals we defined are positionally similar to the Bcs1-type internal signal and follow immediately downstream of an N-terminal transmembrane domain. However, in contrast to the transmembrane localization of proteins containing internal signals (category 3), proteins containing chimeric signals are targeted to the mitochondrial matrix (15). The reason for this difference is probably because the stop transfer signals of CYPs, which enable the proteins to associate with the ER membranes through a single transmembrane anchor (54,55), are incapable of eliciting a similar stop transfer event in the mitochondrial outer or inner membranes. Similar to the Bcs1-type signals, chimeric N-terminal or C-terminal signals are not cleaved by matrix proteases (17,52).
Our studies with xenobiotic-inducible CYPs, GST, and APP led to the discovery of a new family of mitochondria-targeting signals that are referred to as chimeric signals (14 -18, 21, 22). In the case of CYPs and APP, chimeric signals are located at the N terminus, whereas in GSTs these signals are located at the C terminus (14 -18, 21, 22). The characteristic feature of these signals is that the first half (20 -30 amino acids) of the protein is hydrophobic and the second half (10 -12 amino acids) of the protein consists of sequences rich in positively charged amino acids followed by a Pro-rich domain. Studies have also shown that chimeric signals are not substrates for mitochondrial matrix proteases (14). Unlike the canonical signals that target proteins exclusively to mitochondria, the chimeric signals are dynamic in the sense that they direct the targeting of the protein to mitochondria as well as non-mitochondrial compartments, depending on the physiological demand of the cell.
In this study, we tested four different chimeric signals, including two N-terminally truncated forms to mimic endoproteolytic cleavage in the cytosol, as in the case of CYP1A1 (ϩ5/1A1 and ϩ33/1A1), and two that are targeted to mitochondria as intact uncleaved signals, namely CYP2B1 and CYP2E1 (15,16,18,21,23), for their ability to bind different TOM receptor proteins under both in vitro and in vivo conditions. Mitochondrial targeting of both CYP2B1 and 2E1 was enhanced markedly by PKA-mediated phosphorylation at the unique target sites, Ser-128 and Ser-129, respectively (16,18). The results presented here show that the client protein binding to Hsp70 and Hsp90 is enhanced by internal phosphorylation, which may be the reason for increased mitochondrial import. Our results also suggest that the chimeric signals fall into at least two distinct classes: one class, i.e. ϩ33/1A1 and CYP2B1, which bind to TOM40 directly in the absence of other TOM proteins, and the second class, i.e. ϩ5/1A1 and CYP2E1, which require other peripheral TOMs for binding to reconstituted TOM40 (Fig. 8). In this respect, CcOVb, which also binds to TOM40 in the absence of other TOMs, resembles the first category of chimeric signals.
As shown in a model (Fig. 9), the chimeric signals of CYP 2B1, 2E1, ϩ5/1A1, and ϩ33/1A1 are capable of targeting proteins to the channel-forming TOM40 protein in the absence of some or all peripheral TOM receptors (TOM70, TOM20, and TOM22 in the case of CYP 2B1 and ϩ33/1A1, and TOM20 and TOM22 in the case of CYP 2E1 and ϩ5/1A1). It was shown that mitochondria-destined proteins with pre-sequences interact first with TOM20 and then TOM22, whereas proteins with internal signals interact first with TOM70 followed by sequential interaction with TOM20 and TOM22 (57). Interestingly, the chimeric signals of CYP 2B1 and ϩ33/1A1 interact first with TOM20 and TOM22 in the presence of Hsp70 (see Fig. 9A), thus mimicking the characteristic feature of canonical pre-sequences. In the presence of Hsp70 and Hsp90, CYP2B1 and ϩ33/1A1 are delivered directly to TOM40, thus bypassing all peripheral TOMs. The signals of CYP2E1 and ϩ5/1A1, on the other hand, mimic the features of internal targeting signals by first interacting with TOM70 (Fig. 9B). Here again, in the presence of Hsp70 alone, these proteins require all three peripheral TOMs, whereas in the presence of Hsp70 and Hsp90, they can bypass TOM20 and TOM22 but require TOM70. These results clearly show the divergent nature of chimeric signals in addition to the functional evolution of the TOM complex, which is capable of interacting with diverse types of mitochondria-targeting signals.
Hsp family cytoplasmic chaperones are known to facilitate the mitochondrial import of client proteins partly by preventing misfolding or protein aggregation (21, 58 -61). More recent studies by Hartl and co-workers provided evidence for the direct physical and functional interactions between cytoplasmic chaperones, Hsp70 and Hsp90 and mitochondrial import machinery (48,60). Hsp70 and Hsp90 dock onto the TOM complex through the specialized tetratricopeptide (TRP) domain of TOM70 and deliver client proteins to the TOM complex for import (60). In yeast, Hsp70 alone appears to be sufficient for pre-protein import, whereas in mammalian cells, both Hsp70 and Hsp90 are involved in pre-protein delivery and FIGURE 9. A proposed TOM bypass mechanism for the mitochondrial targeting of chimeric signal-containing proteins. A, in the presence of Hsp70 alone, ϩ33/1A1 and CYP2B1 proteins associate with TOM20 and TOM22 on their way to TOM40 without requiring TOM70. In the presence of Hsp70 and Hsp90, CYP2B1 and ϩ331A1 bypass all peripheral TOMs and associate directly with TOM40. B, in the presence of Hsp70 only, ϩ51A1 and 2E1 proteins associate with TOM70, TOM20, and TOM22 and are delivered to TOM40. In the presence of Hsp70 and Hsp90, these proteins require only TOM70 and bypass TOM20 and TOM22 for association with TOM40.