The Mitochondrial ADP/ATP Carrier Associates with the Inner Membrane Presequence Translocase in a Stoichiometric Manner*

Background: The TIM23 machinery cooperates with partner protein complexes of the inner mitochondrial membrane to drive preprotein import. Results: The mitochondrial ADP/ATP carrier is efficiently copurified with TIM23 complexes independently of the respiratory chain. Conclusion: The ADP/ATP carrier is an integral part of the TIM23 protein network and supports preprotein import. Significance: This study reveals an unexpected link between mitochondrial protein biogenesis and metabolite transport. The majority of mitochondrial proteins are synthesized with amino-terminal signal sequences. The presequence translocase of the inner membrane (TIM23 complex) mediates the import of these preproteins. The essential TIM23 core complex closely cooperates with partner protein complexes like the presequence translocase-associated import motor and the respiratory chain. The inner mitochondrial membrane also contains a large number of metabolite carriers, but their association with preprotein translocases has been controversial. We performed a comprehensive analysis of the TIM23 interactome based on stable isotope labeling with amino acids in cell culture. Subsequent biochemical studies on identified partner proteins showed that the mitochondrial ADP/ATP carrier associates with the membrane-embedded core of the TIM23 complex in a stoichiometric manner, revealing an unexpected connection of mitochondrial protein biogenesis to metabolite transport. Our data indicate that direct TIM23-AAC coupling may support preprotein import into mitochondria when respiratory activity is low.

Mitochondria are known as structurally and functionally versatile organelles that are the major sites of ATP synthesis in eukaryotic cells under aerobic conditions. They are surrounded by an outer membrane that mediates the communication of mitochondria with other parts of the cell and encompasses the highly folded, protein-rich inner membrane. The enzymatic machineries for ATP production from ADP and inorganic phosphate by oxidative phosphorylation are mainly found in the cristae domains of the inner mitochondrial membrane (1)(2)(3), where they associate physically to form different types of supercomplexes (4,5). Mitochondrial metabolite carrier proteins termed ADP/ATP carriers (AAC) 4 mediate the export of ATP synthesized within mitochondria (6 -10). Both ADP/ATP exchange across the inner mitochondrial membrane and import of phosphate into the matrix are crucial for oxidative phosphorylation to proceed (11).
The biogenesis and maintenance of healthy mitochondria is a complicated process because the vast majority of mitochondrial proteins is encoded by nuclear genes and synthesized in the cytosol. Nuclear-encoded mitochondrial proteins are produced as preproteins equipped with specific targeting information that guides them to their final destination. The most frequently found targeting signals are amino-terminal, cleavable presequences that direct preproteins to the inner mitochondrial membrane or the central matrix compartment (12,13). Presequence-carrying preproteins are imported by the general translocase of the outer membrane (TOM complex) and the presequence translocase of the inner membrane (TIM23 complex) in a membrane potential (⌬)-dependent fashion. The essential, membrane-bound core of the TIM23 complex (TIM23 CORE ) is composed of the Tim17, Tim23, and Tim50 subunits and constitutes the preprotein receptor domain and the protein-conducting channel (14 -18). TIM23 CORE associates dynamically with different energy-converting partner pro-tein complexes at the inner membrane. Via the adaptor proteins Tim21 and Mgr2, TIM23 CORE is coupled to respiratory chain supercomplexes composed of the cytochrome bc 1 complex (complex III) and cytochrome c oxidase (complex IV) (19 -26). This TIM23 complex form was termed TIM23 SORT because it is sufficient for the solely ⌬-driven lateral sorting of preproteins with an additional hydrophobic stop-transfer segment adjacent to the presequence into the inner membrane (27). For the ATP-dependent translocation of preproteins into the aqueous matrix, TIM23 CORE cooperates with the presequence translocase-associated import motor (PAM) that is composed of mitochondrial heat shock protein 70 (mtHsp70) and five cochaperones that differentially regulate the importdriving ATPase activity of mtHsp70 and the assembly status of PAM: Tim44, Pam16, Pam17, Pam18, and Mge1 (15)(16)(17)(18). The biogenesis of presequence-carrying polytopic inner membrane proteins has been shown to require a close cooperation of the TIM23/PAM machinery with the export translocase Oxa1 (28 -30). Therefore, the TIM23 complex acts as an intricate multipurpose molecular machine in which the coupling of distinct partner proteins to TIM23 CORE generates functional specificity.
It has been suggested that also AAC proteins may interact with the TIM23 complex, but this association was considered unspecific or mainly indirect via respiratory chain complexes (24,31,32). We performed a comprehensive analysis of the TIM23 CORE interactome in the yeast Saccharomyces cerevisiae on the basis of stable isotope labeling with amino acids in cell culture (SILAC) combined with mass spectrometry and identified the abundant metabolite carrier AAC as a major interaction partner. Surprisingly, we found that purified TIM23 complexes contain AAC and the channel-forming core subunit Tim23 in a 1:1 ratio. Formation of the TIM23-AAC supercomplex is independent of respiratory chain complexes and may facilitate preprotein import primarily under conditions where respiration is compromised.
The inward-directed force generated by the import motor PAM was analyzed using the model preprotein b 2 (220)-DHFR (19). The preprotein was preincubated for 5 min with 5 M methotrexate in import buffer supplemented with 3% (w/v) BSA, 4 mM NADH, 10 mM creatine phosphate, and 0.2 mg/ml creatine kinase. Mitochondria were added, and reaction mixtures were incubated for 20 min at 25°C. Thereafter, ⌬ was uncoupled by the addition of 10 M valinomycin. Reactions were further incubated for different time periods, stopped by transfer to ice, and treated with 50 g/ml proteinase K where indicated. To assess the import-driving force, the proteinase K-resistant intermediate (i) form of b 2 (220)-DHFR was quantified. Values were normalized to the amount of i-form generated during the first incubation in the presence of ⌬ without proteinase K treatment.
Quantitative Analysis of Mitochondrial Protein Complexes-TIM23 complexes were purified from isolated mitochondria via protein A-tagged Tim23. The RTS 100 wheat germ CECF kit (5 Prime, Hilden, Germany) was used for cell-free protein synthesis according to the instructions of the supplier. Proteins were produced with C-terminal hexahistidine tags for subsequent purification. Wheat germ lysates containing the proteins of interest were diluted 12.5ϫ in buffer A (50 mM sodium phosphate (pH 8.0), 300 mM NaCl, 0.5% (w/v) SDS, 2 mM PMSF, and 1ϫ EDTA-free protease inhibitor cocktail (Roche)) supplemented with 5 mM imidazole and incubated with pre-equilibrated PerfectPro Ni 2ϩ -NTA agarose beads (5 Prime) for 1 h at room temperature. Protein-loaded Ni 2ϩ -NTA beads were washed extensively with buffer A containing 10 mM imidazole, and bound proteins were eluted with buffer B (50 mM sodium phosphate (pH 7.0), 300 mM NaCl, 300 mM imidazole, 0.5% (w/v) SDS, 2 mM PMSF, and 1ϫ EDTA-free protease inhibitor cocktail (Roche)). Different volumes of the elution fractions were analyzed by SDS-PAGE and colloidal Coomassie staining together with multiple defined amounts of a protein standard (low molecular weight calibration kit (GE Healthcare)). Stained protein bands were quantified using MultiGauge software (Fujifilm, Tokyo, Japan). The concentrations of purified proteins in the Ni 2ϩ -NTA elution fractions were calculated from the given protein amounts in the protein standard bands of comparable molecular weight. Subsequently, different amounts of the elution fractions of TIM23 complex isolations from mitochondria and the respective in vitro-synthesized purified proteins were analyzed by 6 -16.5% Tris-Tricine SDS-PAGE and Western blotting using antisera against Tim23, AAC, Yhm2, or Pam18. The amounts of the different proteins recovered in the elution fractions of TIM23 complex isolations were calculated and normalized to the amount of the bait protein Tim23. The total amount of Yhm2 in isolated mitochondria was calculated in a similar way using purified Yhm2 as a standard (25).
SILAC, MS, and Data Analysis-Wild-type yeast cells were grown in the presence of [ 13 C 6 ]lysine and [ 13 C 6 ]arginine, whereas cells expressing protein A-tagged Tim23 (Tim23 ProtA ) were grown in the presence of standard 12 C-containing amino acids (25). Mitochondria were isolated from both strains and mixed in equal amounts (total protein content). Subsequently, mitochondria were lysed in digitonin buffer, and detergent extracts were subjected to IgG affinity chromatography. Purified TIM23 complexes were acetone-precipitated, followed by proteolytic digestion with trypsin. Nano-HPLC/ESI-MS/MS analyses of the resulting peptide mixtures of three independent replicates were performed using an UltiMate TM 3000 HPLC system (Dionex LC Packings/Thermo Fisher Scientific, Idstein, Germany) coupled directly to an LTQ-Orbitrap XL instrument (Thermo Fisher Scientific). Peptides were first loaded onto a C18 micro-precolumn (0.3 mm inner diameter ϫ 5 mm; PepMap, Dionex LC Packings/Thermo Fisher Scientific), pre-concentrated and washed in 0.1% (v/v) trifluoroacetic acid (flow rate, 30 l/min), and then separated on a C18 reverse-phase nano-LC column (75 m inner diameter ϫ 150 mm; PepMap, Dionex LC Packings/Thermo Fisher Scientific). For peptide elution, a 150-min linear gradient with increasing concentrations of acetonitrile (4 -34% (v/v)) in 0.1% (v/v) formic acid was applied at a flow rate of 300 nl/min. The LTQ-Orbitrap was equipped with a nano-ESI source (Thermo Fisher Scientific) and was operated in a data-dependent mode acquiring highresolution MS precursor scans in the Orbitrap (m/z 300 -1500; resolution, 60,000 at m/z 400; automatic gain control, 5 ϫ 10 5 ; maximum fill time, 500 ms). Concomitantly, up to six of the most intense multiply charged ions were further fragmented by low-energy, collision-induced dissociation in the linear ion trap (automatic gain control, 1 ϫ 10 4 ; maximum fill time, 100 ms). Collision-induced dissociation was performed at a normalized collision energy of 35%, applying a Mathieu stability parameter q value of 0.25 and an activation time of 30 ms. The dynamic exclusion time was set to 45 s. Mass spectrometric raw data were processed using MaxQuant (version 1.0.13.13) in combination with the search engine Mascot (version 2.2). For peptide and protein identification, MS/MS data were searched against a decoy database derived from the Saccharomyces Genome Database using the following parameters: mass tolerances for precursor and fragment ions, 7 ppm and 0.5 Da, respectively; methionine oxidation as variable modification; proteolytic enzyme, trypsin; maximum of missed cleavages, 2; and at least one unique peptide with a minimum of six amino acids. A false discovery rate of Ͻ1% was applied on both the peptide and protein levels. Relative protein quantification was performed on the basis of differentially [ 12 C 6 ]/[ 13 C 6 ]arginine-and [ 12 C 6 ]/ [ 13 C 6 ]lysine-labeled unique peptides and a minimum ratio count of one. Low-scoring variants of identified peptides were not considered for quantification. For data visualization, the mean of log 10 -transformed SILAC (light/heavy) ratios were calculated and plotted against the p value (Ϫlog 10 ) determined for each protein using a one-sided Student's t test.

RESULTS
Interactome of the TIM23 Core Complex-To obtain a compendium of TIM23-interacting proteins, we differentially labeled yeast cells that expressed either an unmodified (wildtype) or protein A-tagged version of the presequence translocase core component Tim23 from the chromosome. Isolated mitochondria of both strains were mixed and solubilized under mild conditions with the non-ionic detergent digitonin. The resulting detergent extracts were subjected to IgG affinity chromatography. Mass spectrometric analysis of the elution fractions was performed, and light-over-heavy peptide ratios were assigned to the identified proteins. In this way, we obtained the first comprehensive view of the interactome of the TIM23 complex ( Fig. 1A; for a complete list of identified proteins, see supplemental Table S1). Among the most enriched proteins were the known core components Tim17, Tim23, and Tim50, the TIM23 SORT -specific subunits Tim21 and Mgr2, components of the import motor PAM, and subunits of respiratory chain supercomplexes, indicating that both major forms of the TIM23 complex were recovered. The group of further identi-fied proteins enriched with Tim23 ProtA contained subunits of known partner protein complexes like the TOM complex (Tom40, Tom22, and Tom20) and the oxidase assembly (OXA) machinery (Oxa1) that transiently cooperate with TIM23 (15-18, 29, 30) as well as several respiratory chain assembly factors that may link import and sorting of respiratory chain proteins to their incorporation into mature complexes. A dual role as TIM23-associated component and complex IV assembly factor has been demonstrated for human TIM21 (36). Among the most enriched proteins in our TIM23 interactome dataset was a so far uncharacterized protein termed Aim19. Deletion of the AIM19 gene leads to a mild respiratory growth defect (37), and a proteomic analysis identified Aim19 as a potential interaction partner of respiratory chain supercomplexes (38), which may explain its strong enrichment with tagged Tim23. Moreover, subunits of the mitochondrial contact site and cristae organizing system (MICOS complex) (39) that is required for the formation of crista junctions were found to be enriched with Tim23 (Mic10, Mic12, Mic27, and Mic60), albeit with lower efficiency. A recent study has shown that active preprotein import sites containing TOM and TIM23 complexes tethered by a preprotein in transit form distinct clusters that are preferentially located in the vicinity of crista junctions (40), pointing to a possible link between the mitochondrial contact site and cristae organizing system and the preprotein import machinery.
Interestingly, we found several mitochondrial metabolite carriers in the TIM23 interactome. Two members of this abundant protein family, the major ADP/ATP carrier isoform Aac2 and the citrate/oxoglutarate carrier Yhm2 (41), were among the most enriched proteins, together with PAM and respiratory chain subunits (Fig. 1A). These proteins were also detected in our recent proteomic analysis of only the TIM23 SORT form that led to the identification of the Mgr2 subunit (25). Because metabolite carrier proteins are imported via the carrier translocase of the inner membrane (TIM22 complex) (42), a copurification of AAC and Yhm2 as client proteins seems highly unlikely. Western blot analysis of label-free TIM23 complex purifications via Tim23 ProtA confirmed the efficient coisolation of the TIM23 CORE and PAM components as well as respiratory chain subunits (Qcr6 and Rip1 of complex III and Cox4 of complex IV) with Tim23 under our experimental conditions (Fig.  1B). Both AAC and Yhm2 were similarly detected in the elution fraction, whereas subunits of other highly abundant protein complexes of the inner mitochondrial membrane, like Atp19 of the F 1 F o -ATP synthase, were not recovered.
AAC Associates with TIM23 Complexes in a Stoichiometric Manner-Our previous analysis of total mitochondrial protein levels had shown that AAC is ϳ18 times more abundant than Tim23 (25). We determined the total amount of Yhm2 in isolated yeast mitochondria using the same biochemical approach. Yhm2 was present at ϳ16 pmol/mg of mitochondrial protein, which is virtually identical to the amount of Tim23 (25). Therefore, a comparison of the Western blot signals obtained for TIM23 core subunits and Yhm2 in the load and elution fractions of Tim23 ProtA isolations (Fig. 1B) suggests that Yhm2 is associated with TIM23 in a substoichiometric manner. However, in this type of analysis, the large difference in the total amounts of AAC and Tim23 made it difficult to estimate how much AAC is bound to TIM23 complexes. Divergent views have been reported before on the significance and specificity of a putative interaction between AAC and TIM23. AAC recovered with protein A-tagged Tim23 in an earlier label-free study was considered contamination (31). Another study, in which a hexahistidine tag on the Aac2 protein was used, came to the conclusion that TIM23 mainly interacts with large supercomplexes composed of respiratory chain complexes and AAC (24). However, a proteomic analysis of the AAC interactome on the basis of affinity purification experiments with tagged Aac2 identified respiratory chain complexes, but not TIM23 components, as candidate partner proteins (43). Notably, it has also been reported that affinity tags on Aac2 might considerably influence its interaction with partner proteins (32). Therefore, we decided to directly determine the amount of AAC associated with TIM23 complexes in the presence of unmodified AAC by a quantitative Western blotting approach that we established earlier to determine the amount of Mgr2 associated with TIM23 SORT (25). In brief, mitochondrial proteins were synthesized in a cell-free system and purified. Different defined concentrations of the purified proteins were then compared directly with the amounts of the respective proteins recovered in the elution fractions of TIM23 complex isolations via Tim23 ProtA . The surprising result of these experiments was that AAC is coisolated with Tim23 in a molar ratio of ϳ1:1 (Fig.  2). In contrast to AAC, Yhm2 was ten times less abundant in the elution fractions of TIM23 complex isolations than the tagged bait protein Tim23 (Fig. 2). It should be noted here that the light-over-heavy SILAC ratios reflect the specificity of coisolation for a given protein (relative enrichment with the tagged bait over the untagged wild-type control sample) but not the absolute protein amount in the elution fractions. Our quantitative Western blot analysis revealed that, despite similar SILAC ratios, the absolute protein amounts of AAC and Yhm2 and their molar stoichiometry to the bait protein Tim23 in the elution fractions differ by one order of magnitude. Moreover, we quantified the amount of the import motor component Pam18 isolated with tagged Tim23. It is assumed that Pam18 and Tim23 are present in equimolar amounts in the coupled TIM23/PAM machinery (17) but that not all TIM23 CORE complexes in mitochondria are associated with PAM (18). In agreement with these models, our quantitative analysis of TIM23 complex isolations revealed that a substoichiometric amount of Pam18 is coisolated with tagged Tim23 (ϳ0.6 pmol Pam18/1 pmol of Tim23 in the elution fraction) (Fig. 2).
AAC Binds to TIM23 Independently of the Respiratory Chain-Our finding that AAC is coisolated with tagged Tim23 in a 1:1 ratio is difficult to reconcile with the hypothesis that AAC mainly associates with TIM23 complexes indirectly via respiratory chain supercomplexes (24) because the latter have been shown before to interact selectively with TIM23 SORT complexes (21,22). To directly assess the mutual dependence of respiratory chain and AAC interactions with TIM23, we generated a yeast strain that expressed protein A-tagged Tim23 in the presence of unmodified AAC but lacked mitochondrial DNA and, therefore, assembled respiratory chain complexes (Tim23 ProtA rhoϪ). We isolated mitochondria from Tim23 ProtA rhoϪ and, for comparison, Tim23 ProtA rhoϩ cells, both grown in fermentable medium (YPGal), solubilized them under mild conditions as described above, and performed affinity purification experiments. For the TIM23 CORE subunit Tim17, both the steady state protein levels in the detergent extracts of mitochondria (Fig. 3A, lanes 1 and 2) and the coisolation efficiency with tagged Tim23 (Fig. 3A, lanes 3 and 4) were similar in rhoϩ and rhoϪ mitochondria. Therefore, the loss of mitochondrial DNA did not considerably affect the amount and assembly state of TIM23 CORE . The protein levels of nuclear-encoded respiratory chain complex subunits, like Cox4, were strongly reduced in rhoϪ mitochondria as expected (Fig.  3A, lanes 1 and 2). Unassembled Cox4 was not detectable in the elution fractions of Tim23 ProtA rhoϪ isolations, indicating that fully assembled respiratory chain complexes, but not individual subunits, associate with TIM23 complexes (Fig. 3A, lanes 3 and  4) (21,27). For AAC, we obtained a differential result. Although the total protein levels in detergent extracts of mitochondria were reduced strongly in rhoϪ compared with rhoϩ mitochondria (Fig. 3, A, lanes 1 and 2, and B), the amount of AAC coisolated with tagged Tim23 was only reduced moderately (Fig. 3,  A, lanes 3 and 4, and B). Therefore, AAC associates efficiently with TIM23 complexes in the absence of assembled respiratory chain complexes. Because the coisolation of AAC with tagged Tim23 does not depend on respiratory chain complexes, we conclude that the major fraction of AAC recovered with the purified TIM23 machinery is directly associated with TIM23 CORE complexes.

Co-isolation with Tim23
ProtA FIGURE 2. AAC associates with TIM23 complexes in a stoichiometric manner. TIM23 complexes were purified from digitonin-solubilized mitochondria via protein A-tagged Tim23. The mitochondrial proteins Tim23, AAC (major isoform Aac2), Yhm2, and Pam18 were synthesized in a cell-free system and purified. The concentrations of purified proteins were determined as described under "Experimental Procedures." Different amounts of isolated TIM23 complexes (lanes 1-3) and in vitro-synthesized, purified proteins (lanes 4 -10) were analyzed by Tris-Tricine SDS-PAGE and immunoblotting with the indicated antisera. Because of the presence of a hexahistidine tag for purification, the in vitro-synthesized proteins show a slightly slower migration in the gel than their mitochondrial counterparts. Bottom panel, the molar amounts of AAC, Yhm2, and Pam18 recovered in the elution fractions of TIM23 complex isolations were calculated and normalized to the amount of the tagged bait protein Tim23. Data are mean Ϯ S.E. (n ϭ 3).  Furthermore, we observed an enhanced association of import motor components, like Pam16, Pam18, and Tim44, with TIM23 CORE (Fig. 3A, lanes 3 and 4) in rhoϪ mitochondria, suggesting that the equilibrium of the different TIM23 forms (19,27) was shifted toward the PAM-bound form. We asked how this enhanced PAM coupling to TIM23 affects the inwarddirected force generated by the import motor on a preprotein at the expense of ATP hydrolysis (17,19). To address this question, we used the model preprotein b 2 (220)-DHFR, which consists of the N-terminal domain of the nuclear-encoded mitochondrial protein cytochrome b 2 fused to DHFR (19,44). In the presence of methotrexate, the folded state of the DHFR moiety is stabilized so that the cytochrome b 2 portion of the preprotein is imported into mitochondria, whereas the folded DHFR moiety remains on the mitochondrial surface, leading to the formation of a two-membrane-spanning translocation intermediate. The matrix-targeting signal of b 2 (220)-DHFR is removed by the mitochondrial processing peptidase, leading to the formation of an i-form. Upon accumulation of b 2 (220)-DHFR in import sites, ⌬ was dissipated to eliminate its contribution to the inward-directed force on the preprotein (45). After different time periods, samples of the import reaction were treated with proteinase K. In rhoϩ mitochondria the activity of the ATP-driven import motor efficiently "pulled" the folded DHFR moiety of the preprotein against the outer mitochondrial membrane so that a substantial fraction of the accumulated i-form remained protected against digestion with proteinase K (Fig. 3C, lanes 2-4) (19). In rhoϪ mitochondria, the amounts of the proteinase K-resistant i-form were considerably lower compared with rhoϩ mitochondria (Fig.  3C, lanes 6 -8; right panel, quantification). Therefore, despite the enhanced coupling of PAM subunits to TIM23 in rhoϪ mitochondria, the import-driving activity of the motor was reduced. These data suggest that ATP levels in the mitochondrial matrix are limiting for PAM activity in rhoϪ mitochondria that do not synthesize ATP by oxidative phosphorylation.
When rhoϩ yeast cells are grown in a fermentable medium like YPGal, mitochondrial ATP synthesis rates are lower than in cells grown in respiratory medium such as YPG. Cellular ATP is mainly synthesized by substrate-level phosphorylation in the cytosol under fermentative growth conditions (46). In mitochondria isolated from Tim23 ProtA rhoϩ cells grown in YPGal, we detected reduced overall levels of AAC compared with mitochondria from cells grown in YPG (Fig. 3D, lanes 1 and 2; right panel, quantification). However, the amount of AAC coisolated with tagged Tim23 was similar regardless of using YPGal or YPG for growth of yeast cells (Fig. 3D, lanes 3 and 4). Therefore, despite the lower overall levels of AAC in cells that import ATP from the cytosol into mitochondria (rhoϪ cells, fermentative growth), the TIM23-bound AAC pool is largely retained.
Coupling of AAC to TIM23 Supports Mitochondrial Protein Import When Respiration Is Compromised-It has been shown that the transport activity of AAC is not required for preprotein import via TIM23 into actively respiring mitochondria (47,48). However, when the respiratory activity of mitochondria is very low or absent, like in rhoϪ mitochondria, AAC activity is not only important to maintain sufficient matrix ATP levels. In addition, the electrogenic import of ATP (four negative charges) from the cytosol into the mitochondrial matrix in exchange for ADP (three negative charges) generates a ⌬ of physiological polarity (i.e. negative on the matrix side) (11). It seemed possible that local coupling of TIM23 complexes to AAC as a ⌬-generating device may support ⌬-dependent preprotein import under such conditions. We aimed to establish an experimental setup to test this idea using the model preprotein Su9-DHFR, which contains the N-terminal mitochondrial targeting sequence of Neurospora crassa ATP synthase subunit 9 and the passenger protein DHFR (49). When ⌬ generation via the respiratory chain and reverse action of F 1 F o -ATP synthase was inhibited by the addition of antimycin A and oligomycin, import of Su9-DHFR was largely blocked (Fig. 4A,  lane 1). Import activity was strongly increased when an ATPregenerating system (creatine phosphate and creatine kinase) was added to the reaction buffer (Fig. 4A, lane 2). Import of Su9-DHFR driven by high ATP levels in the presence of antimycin A and oligomycin was blocked when the AAC inhibitor carboxyatractyloside was added (Fig. 4A, lane 3), demonstrating that the observed import activity depended on AAC-mediated nucleotide exchange. We then used this experimental setup to screen several mutant mitochondria available in our laboratories for specific import defects under such conditions. The mutant yeast strain tim17-11 (see "Experimental Procedures") was selected initially on the basis of its temperaturesensitive growth phenotype. Mitochondria isolated from tim17-11 cells grown at permissive temperature imported Su9-DHFR with comparable efficiency as wild-type mitochondria in the absence of inhibitors, demonstrating that the TIM23/PAM machinery was not compromised (Fig. 4B). (No considerable differences were observed in the steady-state levels of all mitochondrial proteins tested, including TIM23 and PAM subunits as well as AAC in tim17-11 and wild-type mitochondria (data not shown).) However, AAC-dependent import of Su9-DHFR in the presence of antimycin A, oligomycin, and an ATP-regenerating system was reduced significantly in tim17-11 mitochondria (Fig. 4C). This import defect was not caused by a reduced ⌬ or an increased overall sensitivity of the TIM23 complex toward low ⌬ because the response of Su9-DHFR import efficiency to a gradual uncoupling of ⌬ (independent of its source) using carbonyl cyanide m-chlorophenylhydrazone (21,25) was indistinguishable in WT and tim17-11 mitochondria (Fig. 4D). Assessment of the import-driving activity of PAM exerted on b 2 (220)-DHFR accumulated in import sites demonstrated that the import motor is fully active in tim17-11 mutant mitochondria (Fig. 4E). We then introduced a protein A tag on Tim23 in tim17-11 and the corresponding wild-type strain. Cells were grown at a permissive temperature, and mitochondria were isolated and solubilized in digitonin buffer. Detergent extracts were used to purify TIM23 complexes by affinity chromatography. Coisolation of AAC with TIM23 CORE complexes was reduced considerably in tim17-11 mutant mitochondria, whereas the amounts of the essential import motor subunit Pam18 in the elution fractions of wild-type and tim17-11 mitochondria were identical (Fig. 4F). Taken together, these data suggest that the selective defect in AACdependent preprotein import observed in tim17-11 mutant mitochondria is mainly caused by an impaired association of TIM23 and AAC.

DISCUSSION
Different views have been reported on the relationship between the ADP/ATP carrier and the protein import machinery of the mitochondrial inner membrane. Our detailed quan-titative proteomic and biochemical analyses revealed that isolated TIM23 complexes contain AAC in a ϳ1:1 ratio compared with the central Tim23 subunit. This is the first demonstration of a stoichiometric association between a metabolite carrier   (n ϭ 3). E, the import-driving force generated by the import motor PAM in wild-type and tim17-11 mitochondria was analyzed and quantified as in Fig. 3C. F, WT and tim17-11 mitochondria containing protein A-tagged Tim23 were solubilized with digitonin. Detergent extracts were used for affinity purification of TIM23 complexes. Load and elution fractions were subjected to SDS-PAGE und immunoblotting. Signals were quantified using Fiji software. The amounts of AAC and Pam18 coisolated with tagged Tim23 were normalized to the amounts of the core TIM23 subunit Tim17 detected in the elution fractions. Data are mean Ϯ S.E. (AAC, n ϭ 5; Pam18, n ϭ 4). and a mitochondrial protein translocase. The TIM23-bound pool of AAC, however, only represents a minor fraction of the total amount of the protein in mitochondria because AAC is present in vast excess over TIM23 CORE complexes. This may explain why, in earlier studies, the amount of TIM23-bound AAC was underestimated or even appeared below the detection level of the antibodies used. We found that the efficient association of AAC with TIM23 is mostly retained, even when total AAC levels in mitochondria are strongly reduced. This finding underscores the specificity of the TIM23-AAC interaction. The observed decrease of total AAC levels in mitochondria from cells grown under fermentative conditions is in agreement with results of a recent study on the regulation of mitochondrial protein import by reversible phosphorylation. Under fermentative growth conditions, the TOM complex receptor Tom70 becomes phosphorylated by protein kinase A, leading to an inhibition of receptor activity and decreased import of mitochondrial metabolite carriers like AAC (50). Dienhart and Stuart (24) suggested that the TIM23 machinery mainly associates with a large supercomplex of respiratory chain and AAC. They considered unlikely that AAC is an integral part of the TIM23 CORE population because the mobility of TIM23 complexes did not change in the absence of the major AAC isoform Aac2 upon separation of solubilized mitochondrial protein complexes by blue native electrophoresis. Our data, however, show that the association of AAC with TIM23 does not depend on the presence of respiratory chain complexes. We speculate that AAC dissociates from TIM23 CORE complexes in wild-type mitochondria during blue native electrophoresis. A similar behavior has been reported for the essential TIM23 CORE subunit Tim50 (51). This might explain why the migration of TIM23 CORE complexes, as detected in native gels, appeared unaltered when aac2⌬ mutant mitochondria were analyzed (24).
Our finding that AAC directly associates with TIM23 complexes in a stoichiometric manner amends the current view of preprotein import into mitochondria and highlights the striking potential of SILAC-/MS-based analysis of membrane protein complexes to identify previously unrecognized subunits and specific interaction partners. Moreover, our biochemical studies shed light on a possible function of the newly identified TIM23-AAC supercomplex. To achieve efficient import and sorting of presequence-carrying preproteins to different mitochondrial compartments, the TIM23 complex relies on the coupling to different energy-converting machineries, like the ATP-driven motor PAM and respiratory chain complexes. Under conditions of low respiratory activity, AAC mediates the import of ATP into mitochondria from the cytosol in exchange with ADP. This process is necessary to maintain matrix ATP levels that are sufficient to fuel ATP-dependent molecular machines like PAM and, additionally, contributes to the generation of ⌬. This becomes most obvious in rhoϪ mitochondria, where AAC activity is required for preprotein import (52). We propose that both direct ATP supply to the PAM machinery and the efficient use of ⌬ for preprotein import may be facilitated by the direct physical coupling of AAC to TIM23 under conditions of net ATP import into mitochondria from the cytosol (rhoϪ, fermentative growth, and oxygen limitation). This view is supported by our analysis of tim17-11 mutant mitochondria that show reduced TIM23-AAC association. Preprotein import into isolated tim17-11 mitochondria is selectively defective under conditions that render the import reaction sensitive to inhibition of AAC activity. A similar mechanism of local energetic coupling has been demonstrated for the association of TIM23 SORT with ⌬-generating respiratory chain supercomplexes in respiring cells that supports the ⌬-dependent step of preprotein insertion into the inner membrane (21).
In summary, this study identifies the mitochondrial ADP/ ATP carrier as a partner protein of the presequence translocase of the inner mitochondrial membrane. Our findings reinforce the view that the TIM23 complex is embedded into an intricate protein network in the inner mitochondrial membrane. The partner proteins of TIM23 support the preprotein import reaction at different stages and under specific physiological conditions.