Proteins that are imported into the mitochondrial matrix must cross both the outer and inner mitochondrial membranes (Koehler et al., 1999; Bauer et al., 2000; Jensen and Johnson, 2001; Endo and Kohda, 2002; Pfanner and Chacinska, 2002). N-terminal targeting signals (presequences) direct the preproteins to mitochondria and through two protein import complexes, the TOM complex and the TIM23 complex. The TOM complex contains receptors for recognition of preproteins and a general import pore. The two central components of the TOM machinery are the multifunctional receptor Tom22 (van Wilpe et al., 1999) and the pore-forming protein Tom40 (Hill et al., 1998; Künkele et al., 1998). The presequence translocase contains three integral membrane proteins. The pore-forming subunit Tim23 is tightly associated with Tim17 (Dekker et al., 1997; Ryan et al., 1998; Truscott et al., 2001). Tim50 exposes a large domain to the intermembrane space (IMS) (Geissler et al., 2002; Yamamoto et al., 2002; Mokranjac et al., 2003). The import motor on the matrix side, including the mitochondrial heat shock protein 70 (mtHsp70), drives the completion of protein transport into the matrix.

While considerable information has been accumulated on the composition and function of the individual translocases, little is known about the cooperation of the TOM and TIM23 complexes in protein import. Preproteins in transit can span both mitochondrial membranes at so-called translocation contact sites (Schleyer and Neupert, 1985; Rassow et al., 1989, 1990; Wienhues et al., 1991; Jascur et al., 1992; Kanamori et al., 1997; Schülke et al., 1997). When matrix-targeted preproteins are arrested in translocation contact sites due to a tightly folded domain at the C-terminus, the TOM complex and TIM23 complex can be co-purified (Horst et al., 1995; Dekker et al., 1997; Sirrenberg et al., 1997; Schülke et al., 1999; Geissler et al., 2002). However, it remained open as to which components were required for generation and/or stabilization of the TOM–TIM–preprotein supercomplex. So far, two components that expose domains to the IMS have been suggested to play a role in the TOM–TIM connection. Yeast Tim23 consists of an inner membrane-integrated domain (residues 97–222) and a hydrophilic N-terminal domain that is exposed to the IMS. While the segment formed by amino acid residues 51–96 binds presequences as well as the IMS domain of Tim50 (Bauer et al., 1996; Komiya et al., 1998; Geissler et al., 2002; Yamamoto et al., 2002), a most interesting role has been reported for the N-terminal 50 residues. This segment of Tim23 spans the outer membrane and has been suggested to promote a cooperation of TOM and TIM23 complexes in translocation contact sites and the transfer of preproteins (Donzeau et al., 2000). The IMS domain of Tim50 interacts with preproteins while they are still in contact with the TOM complex and directs the preproteins to the TIM23 import channel (Geissler et al., 2002; Yamamoto et al., 2002; Mokranjac et al., 2003). A possible function of the IMS domain of Tom22 in formation of a TOM–TIM supercomplex has not been addressed as yet.

We report unexpected roles for the translocase subunits that expose domains to the IMS. The N-terminal outer membrane spanning segment of Tim23 is neither crucial for protein import nor for formation and stability of translocation contact sites. Tim50 is essential for generation but not for stabilization of the translocation contact site supercomplex, while Tom22–IMS forms a stabilizing element.

Fig. 1. The TOM–TIM supercomplex represents a productive translocation intermediate. (A) Formation of the supercomplex. Purified b2Δ-DHFR was incubated with wild-type yeast mitochondria for 15 min at 25°C in the (more ...)

We asked whether the supercomplex could be dissociated by chasing the accumulated preprotein to its fully imported form. However, when mitochondria containing the supercomplex were reisolated and washed, MTX still bound so tightly to b₂Δ-DHFR that a release of MTX and subsequent chase of the protein turned out to be inefficient (not shown). We therefore established a method to reversibly accumulate b₂Δ-DHFR by use of the DHFR substrates NADPH and dihydrofolate (DHF) (Figure 1B, lane 3). In a two-step import reaction, b₂Δ-DHFR was first accumulated in the supercomplex in the presence of NADPH and DHF (Figure 1C, lanes 3–8). The accumulated b₂Δ-DHFR was processed to the i-form by the matrix processing peptidase, yet remained accessible to externally added protease (Figure 1C, lanes 3 and 4). Upon reisolation of the mitochondria, a second incubation was performed in the absence of DHFR ligands (‘chase’), leading to the generation of protease-protected i-b₂Δ-DHFR (Figure 1C, lane 8). To monitor the presence of the supercomplex and the formation of protease-protected i-b₂Δ-DHFR in parallel, we performed a two-step import-chase assay and divided each sample in half (Figure 1D). In one set, the mitochondria were lyzed with digitonin and analyzed by BN–PAGE, while in the other set, the mitochondria were treated with proteinase K and analyzed by SDS–PAGE. The generation of protease-protected i-b₂Δ-DHFR correlated with the disappearence of the TOM–TIM supercomplex (Figure 1D, lanes 4–6). The chase reaction was not blocked by dissipation of the membrane potential Δψ (Figure 1D, lanes 7–9), confirming that the accumulated preprotein in the supercomplex had already passed the Δψ-dependent early import steps. When analyzed by different separation methods (see below), the size and the composition of the supercomplex were independent of which DHFR ligands were used (NADPH/DHFR or MTX). We conclude that the TOM–TIM supercomplex forms a chaseable translocation intermediate, i.e. represents a productive import intermediate.

To analyze the composition of the supercomplex, we used mitochondria derived from yeast strains either carrying a tagged Tim23 (protein A tag) or a tagged Tom22 (His₁₀ tag). The addition of these tags did not impair cellular growth or mitochondrial function (Meisinger et al., 2001; Geissler et al., 2002). Affinity-purification of Tim23 via IgG sepharose led to the co-purification of the other subunits of the TIM23 complex, including Tim50 and Tim44, independently of the presence or absence of the preprotein b₂Δ-DHFR (Figure 1E, lanes 5 and 6). The subunit e of the dimeric F₀F₁-ATPase, also termed Tim11, was used as a non-associated control protein (Arnold et al., 1998). Similarly, affinity-purification of Tom22 via Ni-nitrilotriacetic acid (Ni-NTA) agarose led to a co-purification of other Tom subunits, such as Tom40, independently of the presence of b₂Δ-DHFR (Figure 1F) (Geissler et al., 2002). However, subunits of the non-tagged translocase complex were only co-purified when b₂Δ-DHFR had been arrested in these mitochondria in the presence of ligand, i.e. co-purification of Tim proteins with tagged Tom22 (Figure 1F, lanes 2 and 3) (Geissler et al., 2002) and co-purification of Tom proteins with tagged Tim23 (Figure 1E, lane 5). Thus, both tagging methods allow for a specific purification of the TOM–TIM supercomplex and demonstrate that the stable association of TOM complex and TIM23 complex depends on the presence of the accumulated preprotein.

The N-terminal segment of Tim23 (residues 1–50) has been suggested to be important for formation of translocation contact sites (Donzeau et al., 2000). We constructed a yeast strain, tim23-3, that selectively lacked this segment. The strain grew like wild-type yeast under all conditions tested, including fermentable and non-fermentable media (Figure 2A). All marker proteins analyzed in tim23-3 mitochondria were present in wild-type amounts, including Tim17, Tim44 and Tim50, as well as Tom subunits and matrix chaperones (Figure 2B). As expected, Tim23 was no longer detected by an antibody generated against the N-terminal segment (Figure 2B and C). BN–PAGE using antibodies against Tim17 demonstrated that the TIM23 complexes migrated slightly faster than that of wild-type mitochondria, consistent with the lack of the N-terminal segment of Tim23 (Figure 2C, lanes 3 versus 4). The TOM complex was not altered (Figure 2C, lane 6).

Fig. 2. A yeast mutant lacking the N-terminal 50 residues of Tim23. (A) Growth of tim23-3 mutant cells is indistinguishable from wild-type cells. Wild-type (WT) and tim23-3 cells were subjected to consecutive 10-fold dilutions, spotted on fermentable (more ...)

Donzeau et al. (2000) had reported that mutant mitochondria lacking the N-terminal segment of Tim23 were strongly impaired in the import of one presequence-containing preprotein, a fusion protein between the presequence of F₀-ATPase subunit 9 and DHFR (Su9-DHFR). We performed a systematic analysis of processing and transport to a protease-protected location of a number of preproteins by tim23-3 mitochondria, including the matrix-targeted protein F₁-ATPase subunit β, the inner membrane protein cytochrome c₁, as well as the matrix-targeted DHFR fusion proteins Su9-DHFR and b₂Δ-DHFR. Surprisingly, we observed only mild import defects for all preproteins analyzed (Figure 3A and B). This mild import defect of tim23-3 mitochondria was similarly observed with small, radiochemical amounts of preprotein (Figure 3B, left panel) and large, saturating amounts of preprotein (Figure 3B, right panel). Why did Donzeau et al. (2000) observe a strong inhibition of protein import into mitochondria lacking the N-terminal Tim23 segment? A likely explanation is that in their import experiments, they used a yeast strain that overexpressed the truncated Tim23 together with a C-terminal His₁₂ tag (Bauer et al., 1996), while the yeast strain tim23-3 used here expressed the N-terminally truncated Tim23 from its own promoter without any further modifications. We conclude that the N-terminal segment of Tim23 plays only a supporting role in dynamic protein import.

Fig. 3. A connection of mitochondrial outer and inner membranes via Tim23 is not critical for formation of the TOM–TIM supercomplex. (A) Import of preproteins into tim23-3 mitochondria. The 35S-labeled precursors of F1-ATPase subunit β (more ...)

We then addressed the question of whether the N-terminal segment of Tim23 was required for formation of translocation contact sites by analyzing the TOM–TIM supercomplex with BN–PAGE. The tim23-3 mutant mitochondria, however, were able to generate the supercomplex in the presence of accumulated b₂Δ-DHFR and MTX (Figure 3C, lanes 6–8). A quantitation revealed that the efficiency of supercomplex formation was only reduced by ~10% in tim23-3 mitochondria compared with wild-type mitochondria (Figure 3C). Similarly, in a two-step import reaction, the generation of the supercomplex in the presence of NADPH/DHF and the subsequent chase of b₂Δ-DHFR into the matrix were only mildly affected (not shown). Thus, the N-terminal segment of Tim23 that connects the TIM23 complexes to the outer membrane is essential neither for dynamic import of preproteins nor for formation of the TOM–TIM supercomplex.

Fig. 4. The IMS domain of Tom22 is required for stabilization of the TOM–TIM supercomplex. (A) Protein composition of tom22-2 mitochondria. Wild-type (WT) and tom22-2 mitochondria (odd-numbered and even-numbered samples, 10 and 20 µg (more ...)

Three possibilities were conceivable to explain this strong effect on the TOM–TIM–preprotein supercomplex in tom22-2 mitochondria: (i) a reduced stability of the individual translocase complexes, TOM and TIM23; (ii) a reduced efficiency of preprotein import into the mutant mitochondria; or (iii) a role of the IMS domain of Tom22 in stabilization of the TOM–TIM–preprotein connection. (i) The size of the TOM complex of tom22-2 mitochondria analyzed by BN–PAGE was slightly smaller than that of wild-type mitochondria, as expected due to the shortening of Tom22. The majority of the TOM complex migrated in the high molecular weight range ~400 kDa, while only a small fraction (<10%) was dissociated to a smaller form of ~100 kDa (Figure 4D, lanes 1 and 2) (van Wilpe et al., 1999). The TIM23 complexes were indistinguishable between wild-type and tom22-2 mitochondria (Figure 4D, lanes 3 and 4). Thus, the lack of the IMS domain of Tom22 only mildly impairs the stability of the TOM complex. (ii) Moczko et al. (1997) reported a moderate reduction (~30%) of import of radiochemical amounts of presequence-containing preproteins into tom22-2 mitochondria compared with wild-type mitochondria. However, the import of large amounts of preprotein had not been studied. We therefore analyzed the import of saturating amounts of b₂Δ-DHFR into tom22-2 mitochondria under conditions similar to those used for generation of the supercomplex. The efficiency of import, when analyzing either processing to the i-form or transport to a protease-protected location, was reduced by ~30% in comparison to wild-type mitochondria (Figure 5A). Thus, a lack of the IMS domain of Tom22 moderately impairs the import of preproteins, independently of whether they are added in small, radiochemical amounts or in large, saturating amounts. Taken together, these results demonstrate that the dramatic reduction of supercomplex yield in tom22-2 mitochondria can neither be explained by a reduced stability of the individual TOM and TIM23 complexes nor by a general reduction of protein import. We suggest that the IMS domain of Tom22 is important for stabilization of the TOM–TIM–preprotein connection.

Fig. 5. Protein import into tom22-2 mitochondria. (A) Import of purified b2Δ-DHFR into wild-type (WT) or tom22-2 mitochondria (80 mM KCl in import buffer). Where indicated, the mitochondria were treated with proteinase K after (more ...)

In order to address under which import conditions into energized mitochondria the stabilizing role of Tom22–IMS was critical for the import of b₂Δ-DHFR, we made use of a previous observation for the cytosolic receptor domains of Tom20 and Tom22. Experiments performed at differing ionic strengths revealed a salt-sensitivity in the interaction between preprotein and the cytosolic receptor domain of Tom22, indicating the involvement of ionic interactions, yet a salt-enhanced interaction was observed between preprotein and Tom20, indicating the involvement of hydrophobic-type interactions (Brix et al., 1997). The hydrophobic nature of the Tom20–preprotein interaction was subsequently directly demonstrated by the high resolution structure of the receptor domain (Abe et al., 2000). We imported b₂Δ-DHFR at increased ionic strength and found a strong import defect into tom22-2 mitochondria as compared with wild-type mitochondria when 250 mM KCl instead of the usual 80 mM KCl were included (Figure 5B). The import efficiency of b₂Δ-DHFR into tom22-2 mitochondria was <20% of that into wild-type mitochondria at higher ionic strength (Figure 5C). Thus the IMS domain of Tom22 is of particular importance for protein import into energized mitochondria under conditions where ionic interactions are disfavored and hydrophobic interactions are favored, resembling the characteristics of Tom20, but not that of the cytosolic domain of Tom22 (Brix et al., 1997; Kanamori et al., 1999; Abe et al., 2000).

Fig. 6. The TOM–TIM supercomplex is not dissociated by a release of Tim50. (A) Import of b2Δ-DHFR is inhibited by depletion of Tim50. Purified b2Δ-DHFR was imported into Tim50-depleted or wild-type (WT) mitochondria. The (more ...)

A surprising observation was made, however, when the composition of the supercomplex of wild-type mitochondria was analyzed by BN–PAGE. While Tim23 and a fraction of Tom40 were shifted to the high molecular weight supercomplex size by the accumulation of b₂Δ-DHFR in the presence of MTX, no Tim50 could be detected in the supercomplex (Figure 6C). In contrast, the affinity-purification of the supercomplex via tagged Tom22 led to an efficient co-purification of Tim50 (Geissler et al., 2002) (see below; Figure 6E, second panel). We therefore established sucrose gradient centrifugation as a third independent means to analyze the supercomplex. Wild-type mitochondria were incubated with b₂Δ-DHFR, lyzed with digitonin and analyzed by centrifugation on a sucrose gradient. In the presence of MTX, i.e. accumulation of b₂Δ-DHFR in translocation contact sites, the preprotein and the bulk of Tim23 and Tim17 were shifted to the high molecular weight range above 600 kDa (Figure 6D, right panel). Only a fraction of Tom40 was shifted, as expected, since the TOM complexes are more abundant than the TIM23 complexes and therefore only a fraction of TOM complexes carry the two-membrane-spanning preprotein when the TIM23 complexes are saturated with preprotein (Dekker et al., 1997; Sirrenberg et al., 1997). Only a relatively small fraction of Tim50, however, was found in the molecular weight range of the supercomplex, comparable to the situation observed for Tim44 (Figure 6D, right panel). As a control, we used subunit e of the F₀F₁-ATPase, which was not affected by the presence or absence of the two-membrane-spanning preprotein. Thus, two subunits of the TIM23 complex, Tim23 and Tim17, are efficiently shifted to the high molecular weight range of the supercomplex, while the bulk of the other two Tim subunits, Tim50 and Tim44, are not present in the supercomplex after the sucrose gradient separation.

We directly determined the quantitative presence of Tim50 in the supercomplex with the three methods used. Affinity purification via tagged Tom22 led to a recovery of ~80% of Tim50 molecules in relation to the channel protein Tim23 in the supercomplex (Figure 6E, second panel, columns 2 and 4). After sucrose gradient separation, only ~10% of Tim50 molecules, but ~70% of Tim23 molecules, were found in the supercomplex (Figure 6E, third panel, columns 2 and 4). After BN–PAGE, no Tim50 was observed in the supercomplex, i.e. the level was below the detection limit of <1%, while ~70% of Tim23 molecules were present in the supercomplex (Figure 6E, bottom panel, columns 2 and 4). Since the accumulation of b₂Δ-DHFR in translocation contact sites and lysis of mitochondria with digitonin were performed in the identical manner for all three experimental approaches, only the method of separation was different. Thus, the bulk of Tim50 was present in the supercomplex in mitochondria, yet is released by BN–PAGE or sucrose gradient centrifugation, similar to the release of Tim50 from the TIM23 complex that has been observed by various methods (Geissler et al., 2002; Yamamoto et al., 2002; Mokranjac et al., 2003). Therefore, the TOM–TIM–preprotein connection remains stable under all three methods of separation, independently of whether Tim50 is present or not. Moreover, we did not obtain evidence for a direct interaction between Tom22–IMS and Tim50–IMS (not shown), yet the formation of the supercomplex strictly depended on the presence of the two-membrane-spanning preprotein. Our results indicate that Tim50 is crucial for generation of the supercomplex, but is not an essential structural component once the complex has been formed.

The two mitochondrial translocases for matrix-targeted preproteins can function independently of each other. However, during protein import, the TOM complex and TIM23 complex can be stably connected via a precursor polypeptide chain in transit. We show that an accumulated preprotein can be efficiently chased into the matrix, concomitantly with a dissociation of the TOM and TIM23 complexes. Thus the TOM–TIM–preprotein supercomplex represents a productive import intermediate, i.e. is a functional translocation contact site complex. We have characterized the role of three translocase subunits that expose domains to the IMS and separated dynamic and stabilizing elements.

The N-terminal segment of Tim23 spans the outer membrane, tethering the TIM23 complex of the inner membrane to the outer membrane. This two-membrane-spanning topology of Tim23 was proposed to facilitate the formation of translocation contact sites and the transfer of preproteins from the TOM complex to the TIM23 complex (Donzeau et al., 2000). Unexpectedly, we found that deletion of this N-terminal segment of Tim23 only mildly impaired dynamic protein import, as well as formation of the TOM–TIM supercomplex. Since the morphology of the mitochondrial inner membrane is altered in mitochondria lacking the Tim23 segment, including a disappearance of cristae (Donzeau et al., 2000), the two-membrane-spanning topology of Tim23 is apparently crucial for the structural arrangement of the mitochondrial membranes, but not for protein translocation. We conclude that tethering of TIM23 complexes to the outer membrane by the N-terminal segment of Tim23 mildly stimulates protein import, probably by supporting an enrichment of TIM23 complexes in areas of the inner membrane that are close to the outer membrane (Donzeau et al., 2000). However, the formation of translocation contact site supercomplexes can occur in the absence of this N-terminal segment of Tim23.

Tim50 is essential for directing preprotein segments, which have passed through the Tom40 channel of the outer membrane, to the Tim23 channel of the inner membrane (Geissler et al., 2002; Yamamoto et al., 2002; Mokranjac et al., 2003). The generation of the TOM–TIM supercomplex thus depends strictly on the presence of Tim50. Surprisingly, however, Tim50 can be quantitatively released from the supercomplex, yet the stable interaction between the TOM complex, TIM23 complex and preprotein is not disturbed. When the supercomplex is isolated by a mild affinity purification, Tim50 is present in nearly stoichiometric amounts to Tim23, similar to the situation in the TIM23 complex itself. In a sucrose gradient centrifugation, a major fraction of Tim50 is released and, by BN–PAGE, virtually all Tim50 molecules are released from the supercomplex as well as from the TIM23 complex. These findings also explain the seemingly contradictory results that have been reported on the stoichiometry of Tim50 in the TIM23 complex (Geissler et al., 2002; Yamamoto et al., 2002; Mokranjac et al., 2003). Thus, Tim50 is an essential, but loosely associated, subunit of the TIM23 complex that plays an important dynamic role in protein import, but is not a critical structural part.

In contrast, the IMS domain of Tom22 is of particular importance for the stabilization of the supercomplex. Tom22 is the central receptor protein that is tightly associated with the Tom40 channel and is thus a structural part of the TOM complex as well as the supercomplex. Upon deletion of the IMS domain of Tom22, the stability of the TOM complex is only mildly affected and the TIM23 complex is unaffected. However, the stability of the TOM–TIM–preprotein supercomplex is strongly reduced. The import of preproteins is moderately impaired in the mutant mitochondria, but not blocked (Court et al., 1996; Moczko et al., 1997). Thus, the transfer of preproteins from TOM to TIM23 is still possible as long as all other import conditions are optimal. When further import impairments occur, such as an increased ionic strength or the removal of cytosolic receptor domains, the lack of the IMS domain of Tom22 causes a strong import defect (Moczko et al., 1997; Kanamori et al., 1999; this study). We conclude that the IMS domain of Tom22 promotes an optimal cooperation of the TOM and TIM23 complexes in translocation contact sites by functioning as a stabilizing, structural element of the TOM–TIM–preprotein supercomplex.

In conclusion, mitochondrial translocation contact sites can be isolated as a TOM–TIM23–preprotein supercomplex. In addition to the accumulated preprotein and the channels Tom40 and Tim23 (with the tightly associated Tim17), we found that the IMS domain of Tom22 is a structural element that stabilizes the supercomplex. In contrast, Tim50 is crucial for dynamic protein import and generation of the supercomplex, but does not represent an essential structural element of the supercomplex.

Table I. Yeast strains

For the two-step import assay (chase experiment), the b₂Δ-DHFR precursor was incubated in the presence of 30 µM NADPH and 30 µM DHF in import buffer for 10 min at 25°C prior to addition of mitochondria. The import reaction was performed for 12 min at 25°C. Mitochondria were reisolated and resuspended in import buffer. The chase reaction was performed at 25°C and stopped on ice. The samples were analyzed by immunodecoration following BN–PAGE or SDS–PAGE.

For purification of the supercomplex, Tim23_ProtA or Tom22_His10 mitochondria (250 µg protein) were incubated with purified b₂Δ-DHFR in the presence of 5 µM MTX for 15 min at 25°C. For control reactions, the preprotein was omitted in the import reaction. The mitochondria were reisolated and solubilized with digitonin. The mitochondrial extract (250 µl) was subjected to a clarifying spin at 14 000 g for 15 min. For isolation via Tim23_ProtA, the supernatant was incubated for 2 h at 4°C with IgG–Sepharose and then washed with 30 vol. of wash buffer (20 mM Tris–HCl pH 7.4, 1% digitonin, 0.1 mM EDTA, 200 mM NaCl, 10% glycerol and 1 mM PMSF). Proteins were eluted with SDS sample buffer (Geissler et al., 2002). For isolation via Tom22_His10, the supernatant was incubated with 0.5 ml Ni-NTA agarose for 30 min on ice in the presence of 5 mM imidazole. After washing with increasing amounts of imidazole (20–50 mM) in 30 vol. wash buffer (20 mM Tris–HCl pH 7.4, 0.2% digitonin, 0.1 mM EDTA, 200 mM NaCl, 10% glycerol and 1 mM PMSF), bound proteins were eluted twice with 200 mM imidazole (Meisinger et al., 2001; Geissler et al., 2002). The samples were separated on SDS–PAGE, blotted onto polyvinylidene fluoride (PVDF) membranes and immunodecorated with various antisera.

For the two-dimensional analysis, mitochondria (100 µg protein) with accumulated b₂Δ-DHFR and control mitochondria were lyzed in 70 µl 1.5% digitonin-containing buffer and separated on 6–13% gradient gel. Gel lanes were excised and applied on top of an SDS-polyacrylamide gel. After separation, proteins were blotted onto PVDF membranes and detected by immunodecoration.

We thank Dr Rob Jensen for anti-Tim17 antiserum, and Drs Andreas Geissler, Thomas Krimmer and Michiel Meijer for discussion and experimental advice. This work was supported by the Deutsche Forschungsgemeinschaft, Sonderforschungsbereich 388, Max Planck Research Award/Alexander von Humboldt Foundation and the Fonds der Chemischen Industrie/BMBF. A.C. was a recipient of a FEBS long-term fellowship.