Mitochondrial protein import: molecular basis of the ATP-dependent interaction of MtHsp70 with Tim44.

Protein translocation across the mitochondrial inner membrane is driven by cycles of binding and release of mitochondrial heat shock protein 70 (mtHsp70) in the matrix. The peripheral inner membrane protein, Tim44, recruits mtHsp70 in an ATP-dependent manner to the import sites. We show that DnaK, the closely related Hsp70 of Escherichia coli, when targeted to the matrix of yeast mitochondria, interacts in a specific manner with Tim44. The interaction is, however, not regulated by ATP, and DnaK cannot support protein translocation. We used truncated mtHsp70s and chimeric proteins consisting of segments of mtHsp70 and DnaK to analyze which portions of mtHsp70 bind and functionally interact with Tim44. We show that Tim44 interacts with the beta-stranded core of the peptide binding domain of mtHsp70 and of DnaK. The alpha-helices A and B of the peptide binding domain of mtHsp70 are required to transmit the nucleotide state of the ATPase domain to the peptide binding domain. Tim44, by interacting in this way with the peptide binding domain, is proposed to coordinate the binding of mtHsp70 to the incoming preprotein and the subsequent release of the mtHsp70-preprotein complex from the TIM23 complex, the translocase of the inner membrane.

The major mtHsp70, 1 Ssc1p in yeast, mediates translocation of proteins across the inner membrane of mitochondria. Ssc1p cooperates with Tim44, a peripheral inner membrane protein associated on the matrix with the TIM23 complex, which constitutes the preprotein conducting import channel across the inner membrane (1)(2)(3)(4)(5)(6)(7). Tim44 is a dimer that recruits two molecules of Ssc1p to the import channel (6,8). It was proposed that Tim44, Ssc1p, and its co-chaperone Mge1p are all parts of an ATP-dependent molecular motor that drives translocation of preproteins by a "hand over hand" mode (6,7). Ssc1p binds to segments of a translocating preprotein as they emerge from the import channel. It is then released from the inner membrane, and the Ssc1p-preprotein complex can move further into the matrix. A central question regarding this pathway is how Tim44 can sense when to release mtHsp70 that has bound a preprotein. To address this question it is important to determine which portions of mtHsp70 interact with Tim44.
Molecular chaperones of the Hsp70 family are composed of two domains, an N-terminal ATPase domain of about 45 kDa that is connected by a short linker to a peptide binding domain of ϳ25 kDa (9). The first part of the peptide binding domain consists of eight anti-parallel ␤-strands that form the substrate binding pocket (10). The ␤-stranded core is followed by an ␣-helical portion consisting of five helices, ␣A-␣E, and a C terminus of unknown structure. ␣B consists of two halves connected by a flexible hinge. ␣A and the first half of ␣B are packed against the ␤-stranded core. The second half of ␣B and ␣C-␣E correspond to a compact ␣-helical subdomain (11). To efficiently close the substrate binding pocket, the ␣-helical subdomain folds like a lid over the ␤-stranded core. Hsp70s bind unfolded polypeptides in an ATP-dependent manner (12)(13)(14)(15)(16)(17). When Hsp70 is in the ATP form, the substrate binding pocket of Hsp70 is open; the chaperone can rapidly bind and release polypeptide substrates. In the ADP form the substrate binding pocket is closed, and polypeptide substrates can hardly bind to Hsp70, whereas bound polypeptides are tightly held and not released. A cycle driven by ATP hydrolysis allows Hsp70s to associate with an unfolded polypeptide in the ATP form and then hydrolyze ATP to tightly hold the substrate and finally release the bound polypeptide after exchange of ADP by ATP. To bind and release polypeptides on a physiologically relevant time scale, Hsp70s cooperate with co-chaperones that accelerate the nucleotide reaction cycle (15, 18 -20). Co-chaperones also tag the relevant substrates and thereby recruit the chaperones for specific tasks (21). Tim44 is such a co-chaperone, and it recruits mtHsp70 for protein import.
Here we show that DnaK, the Hsp70 of Escherichia coli, when targeted to the matrix of yeast mitochondria, is also recruited by Tim44 to the inner membrane import sites. The interaction with Tim44, however, is not modulated by nucleotides, and DnaK does not drive protein translocation. We constructed a series of chimera between Ssc1p and DnaK to identify the segments in Ssc1p that are necessary for a functional interaction with Tim44. We show that Tim44 interacts with the ␤-stranded core of the peptide binding domain of Ssc1p and of DnaK but not via the substrate binding pocket. The interaction is independent of nucleotides. The ␣-helices A and B of the peptide binding domain of Ssc1p, but not of DnaK, modulate the interaction between the ␤-stranded core and Tim44 in a nucleotide-dependent fashion. Implications of the mechanism of protein translocation are discussed.
Protein Translocation into Isolated Mitochondria-Mitochondria (1 mg/ml protein) were suspended in import buffer (600 mM sorbitol, 3 mg/ml bovine serum albumin, 50 mM HEPES-KOH, pH 7.2, 80 mM KCl, 10 mM MgCl 2 ) containing 2.5 mM ATP, 5 mM NADH, 10 mM phosphocreatine, and 0.1 mg/ml creatine kinase. Import reactions were started by the addition of 35 S-labeled precursor proteins synthesized by in vitro transcription/translation in reticulocyte lysate. After a 5-min incubation at 25°C, samples were transferred on ice and split into two aliquots. One aliquot remained untreated, and the other was treated with 50 g/ml PK or trypsin for 25 min on ice. Protease digestion was stopped by the addition of 1 mM phenylmethylsulfonyl fluoride or 0.5 mg/ml soybean trypsin inhibitor, respectively, and mitochondria were reisolated by centrifugation. Proteins were resolved by SDS-PAGE, transferred to nitrocellulose membrane, and detected by autoradiography.
Immunoprecipitation with Tim44 Antibodies-Mitochondria (1 mg/ ml) were solubilized for 30 min at 4°C in 20 mM Tris-HCl, pH 7.4, 100 mM sodium acetate, 0.2% Triton X-100, 1 mM phenylmethylsulfonyl fluoride containing either 4 mM EDTA (-ATP) or 2 mM MgCl 2 and 2 mM ATP (ϩATP). After a clarifying spin, the detergent extracts (250 g) were incubated with gentle agitation for 1 h at 4°C with affinitypurified antibodies against Tim44 bound to protein A-Sepharose beads (6). Immunoprecipitates were washed three times for 5 min with 1 ml of lysis buffer and analyzed by SDS-PAGE. Ssc1p and Ssc1-3p were detected after Western blotting with antibodies against Ssc1p. DnaK and DnaK/Ssc1p chimera were detected with antibodies against DnaK.
Coprecipitation of Translocation Intermediates-Radiolabeled pSu9-(1-69)DHFR was incubated for 5 min at 25°C in import buffer containing 3 M MTX. Then, import buffer was supplemented with 2.5 mM ATP and 5 mM NADH, and mitochondria were added and incubated for 15 min at 25°C. Mitochondria were reisolated by centrifugation and solubilized in 20 mM Tris-HCl, pH 7.4, 150 mM sodium acetate, 0.4% Triton X-100, 15 mM imidazole, and 1 mM phenylmethylsulfonyl fluoride. Samples were subjected to a clarifying spin, and the supernatants were incubated with Ni-NTA beads. The beads were washed three times with 1 ml of solubilization buffer.

Expression of E. coli DnaK in Yeast and Targeting to Mitochondria-To target E. coli
DnaK into the matrix of yeast mitochondria, the open reading frame was fused to the presequence of the ATPase subunit 9 of N. crassa (pSu9) (Fig. 1A). A His tag was introduced between the presequence and the N terminus of DnaK. This construct was expressed under control of a galactose-inducible promoter in the temperature-sensitive yeast strain ssc1-3, which carries a mutation in the major mtHsp70 (23). DnaK was processed to the mature form and localized in the mitochondrial matrix (Fig. 1B). The amount of DnaK that could be isolated by Ni-NTA chromatography was comparable with the amount of Ssc1-3p in control mitochondria as estimated from Coomassie Blue-stained gels (data not shown). This finding indicates that DnaK was expressed at a physiologically relevant level. The mitochondria-targeted DnaK bound to ATP agarose (Fig. 1C), suggesting that it was properly folded in the matrix of the mitochondria. However, when yeast cells were shifted to nonpermissive temperature, DnaK in the mitochondrial matrix did not rescue the temperature-sensitive phenotype of the ssc1-3 strain (data not shown).
DnaK interacted with Mge1p, the mitochondrial homologue of the bacterial co-chaperone GrpE, in an ATP-dependent manner ( Fig. 2A). Using an antibody against Tim44, DnaK coimmunoprecipitated with Tim44 in the presence and absence of ATP (Fig. 2B). Apparently, the interaction between DnaK and Tim44 was not regulated by ATP. Furthermore, DnaK did not support import of pSu9-(1-69)DHFR into mitochondria that were preincubated at 37°C in order to inactivate the mutant Ssc1-3p (data not shown).
DnaK/Ssc1p Chimeric Proteins-What are the functional elements that allow Ssc1p, but not DnaK, to interact with Tim44 in an ATP-dependent manner and to drive protein translocation? DnaK and Ssc1p are the two members of the Hsp70 protein family that share the highest degree of sequence similarity. The N-terminal ATPase domain and the ␤-stranded portion of the peptide binding domain are highly conserved between DnaK and Ssc1p (65 and 68% amino acid sequence identity, respectively). The ␣-helical portion, which comprises the lid and the C-terminal domain, is significantly less conserved (22% identity). Based on the sequence alignment with Ssc1p and on the crystal structure of the peptide binding domain of DnaK (10) (Fig. 3A and B), we divided the chaperones into four structural elements: (i) the ATPase domain; (ii) the ␤-stranded core constituting the substrate binding pocket (␤1-␤8); (iii) helix A plus helix B (␣A-␣B), the latter containing the flexible hinge; and (iv) the C-terminal lid domain (␣C-␣E plus the C terminus). Combining these four structural elements, we constructed chimeric Hsp70s (Fig. 3C). The elements originating from Ssc1p and DnaK are indicated by "C" and "K," respectively, in the appropriate position of a four-letter acronym.
When the chimeric chaperones were expressed in the ssc1-3 strain, none of the chimeric Hsp70s was able to rescue the temperature-sensitive growth phenotype of the ssc1-3 strain (data not shown). Similarly, the chimeric Hsp70s did not complement a disruption of the SSC1 gene (data not shown). As estimated from Coomassie Blue-stained gels, the expression levels of the chimera were significantly lower (ϳ25%) than those of Ssc1p in mitochondria from WT cells yet were comparable with those of Ssc1-3p (data not shown). Thus, it is not clear whether the failure of the chimera to substitute for Ssc1p is because of the lower expression levels or reflects functional deficiencies of the chimera.
For purposes of functional characterization, mitochondria from ssc1-3 cells harboring DnaK or one of the chimeric Hsp70s were lysed with Triton X-100 in the absence or the presence of ATP, and the His-tagged chimera were adsorbed to Ni-NTA beads. WT mitochondria containing a C-terminally His-tagged version of Ssc1p were used as a control. Like DnaK and Ssc1p, the chimeric Hsp70s interacted with Mge1p in a nucleotide-dependent manner (data not shown), demonstrating that all Hsp70s underwent physiologically relevant nucleotide-dependent conformational changes.
To study the interaction of the chimeric chaperones with Tim44, immunoprecipitation with affinity-purified antibodies against Tim44 was performed in the presence and absence of ATP (Fig. 3D). In the absence of ATP, Ssc1p, DnaK, and each of the chimeric Hsp70s coprecipitated in a complex with Tim44. In the presence of ATP, the complexes of Tim44 with DnaK, CKKK, and KKKC were stable. In contrast, the complexes of Tim44 with KKCC, KCCC, and Ssc1p dissociated in the presence of ATP, indicating a nucleotide-dependent interaction of these Hsp70s with Tim44. Thus, the ␣-helices A and B of the peptide binding domain of Ssc1p appear to be necessary to transmit the nucleotide-induced conformational changes from the ATPase domain to Tim44 and thereby regulate the interaction of the chaperones with Tim44 in an ATP-dependent manner.
To characterize the interaction of Ssc1p with Tim44 in more detail, we constructed C-terminally truncated versions of Ssc1p (Fig. 4A, left). The deleted structural elements are indicated by  4). The proteaseresistant fragment generated by trypsin is indicated by an asterisk. C, DnaK expressed in mitochondria binds ATP. DnaK was expressed in ssc1-3 cells, and mitochondria were isolated and exposed to 37°C for 10 min. Mitochondria were solubilized with Triton X-100, and the clarified lysates were incubated with ATP-agarose beads. Mitochondria from ssc1-3 cells without DnaK were used for control.

FIG. 2. Interaction of DnaK with Mge1p and Tim44.
A, interaction of DnaK with Mge1p. Mitochondria from ssc1-3 cells harboring DnaK (ssc1-3 ϩ DnaK) were lysed with Triton X-100 in the absence or presence of 2 mM ATP. After a clarifying spin, the lysate was incubated with Ni-NTA beads. Mitochondria containing a His-tagged Ssc1p were used for control. Precipitated proteins were resolved by SDS-PAGE and analyzed using antibodies against Mge1p. B, DnaK forms a complex with Tim44 that is not dissociated by ATP. Mitochondria were pretreated at 4 or 37°C to inactivate Ssc1-3p. Then the mitochondria were solubilized with Triton X-100 in the absence or the presence of ATP, and the lysates were subjected to immunoprecipitation with affinity-purified antibodies against Tim44. Immunoprecipitates (IP) were resolved by SDS-PAGE. Ssc1p and DnaK were detected by immunostaining with the corresponding antibodies.
a "0" in the corresponding position of the acronym. CC00 lacks the ␣-helical portion and C-terminal domain, and C000 consists only of the ATPase domain. In addition, we constructed DHFR-0CCC and DHFR-0KKK, fusion proteins between the chimeric preprotein pSu9-(1-69)DHFR and the entire peptide binding domain of Ssc1p and DnaK, respectively. To study the interaction of the chimera with Tim44, radiolabeled precursors were synthesized in vitro and imported into isolated mitochondria. The radiolabeled precursors of Ssc1p and pSu9-(1-69)DHFR were imported for control. After the import reactions the samples were treated with PK to remove precursors associated with the surface of mitochondria. All precursors were efficiently imported into the matrix and processed to their mature forms (data not shown). Subsequently the mitochondria were lysed with Triton X-100, and immunoprecipitations with antibodies against Tim44 were performed in the presence or absence of ATP (Fig. 4A, right). Radiolabeled Ssc1p coprecipitated with Tim44 in the absence but not in the presence of ATP, indicating that the imported chaperone behaved like the endogenous Ssc1p. CC00 coprecipitated in a complex with Tim44 in the absence but also in the presence of ATP. This supports the conclusion (see above) that the ␣-helices A and B are required for the ATP dependence of the Tim44-Ssc1p interaction. C000 (ATPase domain) was not coprecipitated with Tim44, indicating that it either did not bind stably to Tim44 or that it could not efficiently compete with the endogenous Ssc1p. This result contradicts earlier findings by Krimmer et al. (24). Possible explanations for the apparent discrepancy are discussed.
Interestingly, DHFR-0CCC and DHFR-0KKK interacted with Tim44, whereas imported DHFR did not coprecipitate Tim44 (Fig. 4B). This indicates that the peptide binding domains of Ssc1p and DnaK interact with Tim44. The interaction of both chimera with Tim44 was not modulated by ATP. In particular, the ␤-stranded portion of the domain is required for the interaction with Tim44, whereas the helices ␣A and ␣B are not required for binding per se but modulate the interaction in response to nucleotides.
To obtain further evidence that Tim44 interacts with the peptide binding domain of Ssc1p in a specific manner, we constructed DHFR-0CCC1-2, a fusion protein between pSu9-  (10). The hydrophobic substrate binding pocket, the hinge, and the lid domain are indicated (11). C, schematic outline of DnaK/Ssc1p chimeric constructs. CKKK contains the presequence of Ssc1p (pre). KKKC, KKCC, and KCCC were constructed with the pSu9 presequence followed by a His tag (see Fig. 1A). The fusion points are indicated according to the numbering of amino acid residues in DnaK. Portions of DnaK and Ssc1p are indicated as white and black boxes, respectively. D, interaction of chimeric Hsp70s with Tim44. Mitochondria from ssc1-3 cells containing the chimeric DnaK proteins were exposed for 37°C for 10 min and lysed with Triton X-100 in the absence or presence of ATP. The clarified lysates were subjected to immunoprecipitation with affinity-purified antibodies against Tim44. DnaK and chimeric Hsp70s were detected by immunostaining of Western blots with antibodies against DnaK. Ssc1p was detected with Ssc1p antibodies.
(1-69)DHFR and the peptide binding domain of Ssc1-2p (Fig.  4B, left). The conditional mutant protein Ssc1-2p contains a mutation in the ␤-stranded core of peptide binding domain (Pro to Ser in position 442 of the precursor) (23). At nonpermissive temperature, Ssc1-2p interacts with substrate proteins in mitochondria but no longer interacts with Tim44 (3,25). Purified Ssc1-2p protein in the ADP form displays a reduced on-rate for peptide (26), suggesting that the flexibility of the closed peptide binding domain is compromised because of the mutation. DHFR-0CCC1-2 and, for control, DHFR-0CCC were imported into mitochondria, and immunoprecipitation with antibodies against Tim44 was performed (Fig. 4B, right). To induce the temperature-sensitive phenotype associated with the ssc1-2 mutation, the immunoprecipitates were incubated at 37°C or at 4°C for control. DHFR-0CCC was detected in the immunoprecipitate at 4 and at 37°C. In contrast, the complex between Tim44 and DHFR-0CCC1-2 was stable at 4°C but dissociated at 37°C. This indicates that the mutation in Ssc1-2p affects the conformation of the peptide binding domain independently of the ATPase domain. Because Ssc1-2p binds substrate proteins at permissive and nonpermissive temperature, the observation demonstrates that Tim44 is not bound like a substrate by the chimera DHFR-0CCC1-2.
Protein Import Facilitated by Chimeric Hsp70s-To investigate whether the chimeric Hsp70s facilitate preprotein translocation, mitochondria were first preincubated at 37°C to inactivate endogenous Ssc1-3p and then incubated at 25°C with the radiolabeled precursor pSu9-(1-69)DHFR (Fig. 5A). pSu9-(1-69)DHFR was not imported into the matrix of mitochondria from the ssc1-3 strain, demonstrating that the mutant chaperone was efficiently inactivated by the preincubation. Rather, as observed previously, low amounts of the intermediate form, which was processed only once by the matrix-processing peptidase, accumulated as a dead-end translocation intermediate in the intermembrane space (data not shown). Likewise, no import above background was observed with mitochondria harboring the chimera CKKK, KKKC, or KKCC. In contrast, pSu9-(1-69)DHFR was imported into mitochondria harboring KCCC where it was efficiently processed to the mature form. The import efficiency was lower than that of mitochondria harboring Ssc1p, indicating that the chimera KCCC was not fully functional and/or not present at the same level as WT Ssc1p.
For further analysis mitochondria were preincubated at 37°C, and then the kinetics of import were measured at either 25 or 37°C (Fig. 5B). At 25°C, mitochondria harboring KCCC imported the precursor efficiently, yet at a slower rate than WT mitochondria. This difference was even more pronounced when import was carried out at 37°C; import of pSu9-(1-69)DHFR into WT mitochondria was faster at 37°C than at 25°C, whereas import into mitochondria containing KCCC was slower at 37°C than at 25°C, suggesting that the chimera KCCC is temperature-sensitive.
We asked whether the chimera KCCC interacts directly with translocation intermediates that are arrested in a membranespanning fashion at the import site. To generate translocation intermediates, mitochondria were pretreated at 37°C and subsequently incubated with pSu9-(1-69)DHFR in the presence of MTX to stabilize the folded DHFR moiety and thereby prevent unfolding and complete import of the precursor (27). Mitochondria were then lysed with Triton X-100 in the absence of ATP, and the His-tagged chimeric Hsp70s were precipitated with Ni-NTA beads (Fig. 5C). The arrested translocation intermediate coprecipitated with KCCC, whereas no interaction of pSu9-(1-69)DHFR with DnaK or KKCC was observed.
Finally, we asked whether KCCC was able to tightly hold an arrested precursor in the import channel. Mitoplasts were prepared and incubated with pSu9-(1-69)DHFR in the presence of MTX (Fig. 5D). In mitoplasts the arrested translocation intermediate spans the inner membrane and is processed to the intermediate form, iSu9-(1-69)DHFR (5), which readily allows identification of arrested species. In mitoplasts generated from WT mitochondria iSu9-(1-69)DHFR was resistant to treatment with PK. This indicates that the MTX-stabilized DHFR moiety was tightly apposed to the outer face of the inner membrane because of interaction of Ssc1p with segments of the precursor exposed into the matrix. In mitoplasts from the ssc1-3 mutant (after preincubation at 37°C) and in ssc1-3 mitoplasts harboring DnaK, little iSu9-(1-69)DHFR was generated, and its resistance to protease treatment was very low (Fig. 5D). Thus, Ssc1-3p and DnaK did not interact with the membrane-spanning intermediate. In mitoplasts containing KCCC iSu9-(1-69)DHFR was generated, but it was largely sensitive to added PK (Fig. 5D). This indicates that KCCC was FIG. 4. Tim44 interacts with the ␤-stranded core of the peptide binding domains of Ssc1p and DnaK. A, the precursors of Ssc1p, CC00, and C000 and the precursors of DHFR-0CCC, DHFR-0KKK, and DHFR, which carry the pSu9 presequence, are outlined schematically. The DHFR moiety is indicated by a dotted box. Portions of DnaK and Ssc1p are indicated as white and black boxes, respectively. The radiolabeled chimeric preproteins were imported for 15 min at 25°C into WT mitochondria. Samples were treated with 1 M valinomycin and 50 g/ml trypsin. Mitochondria were reisolated, washed twice in import buffer, and then lysed with Triton X-100 in the absence or presence of 2 mM ATP. The lysates were subjected to immunoprecipitation with antibodies against Tim44. Immunoprecipitates (IP) were analyzed by SDS-PAGE and autoradiography. pSu9-(69)DHFR was used as a control. B, interaction of Tim44 with the peptide binding domain of Ssc1-2p is temperature-sensitive. The precursors DHFR-0CCC1-2, which contains the peptide binding domain of Ssc1-2p, is shown in comparison with DHFR-0CCC derived from WT Ssc1p. The position of the amino acid exchange in the peptide binding domain of Ssc1-2p is indicated (P442S with respect to the Ssc1-2p precursor). The radiolabeled precursors were imported into WT mitochondria. For immunoprecipitations, antibodies against Tim44 that were bound to protein A-Sepharose beads were used. Incubation with the beads was for 15 min at 4 or at 37°C, and then they were washed twice at 4°C. The immunoprecipitates were resolved by SDS-PAGE and analyzed by autoradiography. not fully competent to hold iSu9-(1-69)DHFR tightly apposed to the inner membrane and to efficiently prevent retrograde movement of the preprotein through the import channel.
In summary, DnaK and the chimeric Hsp70s specifically interacted with Tim44, indicating that Tim44 recruits the Hsp70s via a conserved interface. Two of the chimera interacted with Tim44 in an ATP-dependent manner, but only KCCC, a construct that contained the ATPase domain of DnaK and the whole peptide binding domain of Ssc1p, interacted with a preprotein in transit and supported import in vitro. Yet, KCCC was not fully functional; it supported import at reduced rates and in a temperature-sensitive manner, and it could not efficiently hold a translocation intermediate in the import channel. The hybrid protein was temperature-sensitive in function; expression levels in vivo were significantly lower than those of Ssc1p. Thus, despite the striking similarity between DnaK and Ssc1p, KCCC could not substitute for Ssc1p in vivo. Fine tuning of the interaction between the ATPase and the peptide binding domain appears to be necessary to drive translocation at rates sufficiently high to support biogenesis of mitochondria in vivo.

DISCUSSION
What are the determinants in mtHsp70 that enable it to interact with Tim44 and to drive ATP-dependent protein translocation into the matrix of mitochondria? To answer this question we targeted E. coli DnaK to the mitochondrial matrix of yeast cells. This DnaK could not substitute for Ssc1p, the authentic mtHsp70, in vivo. In particular, DnaK in the matrix was properly folded but did not facilitate mitochondrial protein translocation, a process essential for the viability of yeast under all growth conditions. What are the determinants that enable Ssc1p but not DnaK to drive protein translocation? DnaK in the matrix interacted with Mge1p in an ATP-dependent manner and was assembled into a specific complex with Tim44. However, the interaction between Tim44 and DnaK was not regulated by ATP. Accordingly, the defect in protein import may be due to the deficiency of DnaK to interact with Tim44 in an ATP-dependent manner.
We used this situation to study which domains of Ssc1p interact with Tim44 and what the elements are that regulate the interaction in response to ATP. To this end we analyzed chimeric Hsp70s in which portions of DnaK were replaced by corresponding segments of Ssc1p. The result of these domainswapping experiments is that the helices ␣A and ␣B of the peptide binding domain of Ssc1p are required for a nucleotidedependent interaction of Hsp70 with Tim44. This finding is in agreement with the observation that a C-terminally truncated version of Ssc1p lacking the helices ␣D and ␣E was reported to interact with Tim44 in an ATP-dependent manner (24). We showed that Tim44 interacted with a C-terminally truncated version of Ssc1p that lacked the entire ␣-helical subdomain in a nucleotide-independent manner. Thus, the helices ␣A and ␣B are required for the nucleotide-dependent regulation of the interaction but not for the interaction per se.
The chimeric fusion proteins in which the ATPase domains of Ssc1p or DnaK were replaced by DHFR were isolated in a stable complex with Tim44. This clearly demonstrates that FIG. 5. Protein translocation is supported by chimeric DnaK/ Ssc1p proteins. A, import of pSu9-(1-69)DHFR. Mitochondria harboring Ssc1-3p or in addition to Ssc1-3p the indicated chimeric Hsp70 were pretreated for 10 min at 37°C. ATP (2 mM) and NADH (3 mM) were added, and radiolabeled pSu9-(1-69)DHFR was imported at 25°C for 5 min. Samples were treated with PK (50 g/ml) and analyzed by SDS-PAGE and autoradiography. p, precursor; i, intermediate form; m, mature protein. B, kinetics of the import of pSu9-(1-69)DHFR into mitochondria containing the chimeric protein KCCC. Radiolabeled pSu9-(1-69)DHFR was imported at 25°C (upper panel) or 37°C (lower panel) for the indicated time periods. Samples were treated with 50 g/ml PK and analyzed by SDS-PAGE, and imported mSu9(1-69)DHFR was quantified using a phosphorimaging system. Typical experiments were quantified. C, the chimera KCCC interacts directly with translocation intermediates. Mitochondria from ssc1-3 cells containing the various forms of Hsp70 were preincubated at 37°C for 10 min. Radiolabeled pSu9-(1-69)DHFR was added in the presence of MTX, and the mitochondria were incubated for 10 min at 25°C. Mitochondria were then reisolated, lysed with Triton X-100, and incubated with Ni-NTA beads. Samples were analyzed by SDS-PAGE. Radiolabeled pSu9-(1-69)DHFR, which coprecipitated with the His-tagged Hsp70s, was detected by autoradiography. D, protease resistance of translocation intermediates. Mitoplasts were isolated from ssc1-3 cells containing indicated forms of Hsp70 and preincubated for 10 min at 37°C. pSu9-(1-69)DHFR was added in the presence of MTX and incubated at 25°C for 10 min. Samples were transferred to ice, treated with 50 g/ml PK, and analyzed by SDS-PAGE. The intermediate form protected against PK treatment was quantified using a phosphorimaging system. The columns are the means of two measurements. The amount of iSu9-(1-69)DHFR in WT mitochondria was set to 100%.
Tim44 interacted with the peptide binding domains of Ssc1p and DnaK. A corresponding construct was not investigated by Krimmer et al. (24). We did not detect a complex of Tim44 with a truncated Ssc1p consisting only of the ATPase domain. In contrast, Krimmer et al. (24) could coprecipitate newly imported ATPase domain with Tim44. In both approaches immunoprecipitation with Tim44 antibodies was used to assess binding. In addition to the strong interaction of Tim44 with the peptide binding domain demonstrated here, an additional weak interaction of Tim44 with the ATPase domain cannot be ruled out on principle grounds. However, characterization of the interaction of Tim44 with the conditional mutant protein Ssc1-2p provides further evidence supporting that Tim44 does not or only weakly interact with the ATPase domain. Ssc1-2p carries a single mutation in the ␤-stranded core of the peptide binding domain. When the mutant phenotype is induced, Ssc1-2p remains soluble and interacts with nucleotides in a similar manner as Ssc1p (26). Yet, despite the presence of a functional ATPase domain, Ssc1-2p does not interact with Tim44 at nonpermissive temperature (3,25,28,29). On the other hand, the peptide binding domain of Ssc1-2p, when fused to DHFR, forms a complex with Tim44 at permissive but not at nonpermissive temperatures. Furthermore, all intragenic suppressor mutations of Ssc1-2p that restore the binding to Tim44 have been mapped in the ␤-stranded core of the peptide binding domain (25,26,30). This finding supports the notion that the peptide binding domain rather than the ATPase domain of Ssc1p mediates the interaction with Tim44. Finally, because the truncated version of Ssc1p consisting of the ATPase domain and the ␤-stranded substrate binding pocket interacted with Tim44 independently of ATP, a putative interaction of the ATPase domain with Tim44 must also be independent of ATP. Yet, the coprecipitation of the ATPase domain with Tim44 reported by Krimmer et al. (24) was not observed in the presence of ATP. Interestingly, the ATPase domain of Hsp70s is structurally related to actin (31), and Ssc1p was shown to have a tendency to form dimers and higher oligomeric assemblies in the absence but not in the presence of ATP (32). Thus, the small fraction of the ATPase domain that coprecipitated by Tim44 antibodies in the absence of nucleotides (24) could be mediated via dimerization with a full-size Ssc1p in complex with Tim44. Such a complex would be indirect and dissociated by ATP.
Our data demonstrate that Tim44 interacts with the ␤-stranded portion of the peptide binding domain. Tim44 in this way appears to recruit Ssc1p and precisely position the substrate binding pocket at the outlet of the import channel. This would facilitate the binding of Ssc1p to a preprotein as soon as it emerges from the import channel and trigger ATP hydrolysis in order to close the peptide binding domain and efficiently trap the incoming polypeptide chain. The conformational change in the peptide binding domain associated with the binding to the incoming preprotein may then destabilize the interaction between Ssc1p and Tim44. This may facilitate the release of the chaperone-preprotein complex from the import site, a step required for further inward movement of the translocating preprotein. Inward movement could be driven by Brownian motion (7). Additionally, Tim44 may serve as a fulcrum at the inner membrane, and Ssc1p may actively pull on a translocating polypeptide to facilitate unfolding of the folded precursor (25,29). However, as Tim44 interacts directly with the peptide binding domain of Ssc1p, there would be no lever arm available that could translate a conformational change into a substantial movement of the peptide binding domain perpendicular to the membrane.