Tid1 Isoforms Are Mitochondrial DnaJ-like Chaperones with Unique Carboxyl Termini That Determine Cytosolic Fate*

Tid1 is a human homolog of bacterial DnaJ and the Drosophila tumor suppressor Tid56 that has two alternatively spliced isoforms, Tid1-long and -short (Tid1-L and -S), which differ only at their carboxyl termini. Although Tid1 proteins localize overwhelmingly to mitochondria, published data demonstrate principally nonmitochondrial protein interactions and activities. This study was undertaken to determine whether Tid1 proteins function as mitochondrial DnaJ-like chaperones and to resolve the paradox of how proteins targeted primarily to mitochondria function in nonmitochondrial pathways. Here we demonstrate that Tid1 isoforms exhibit a conserved mitochondrial DnaJ-like function substituting for the yeast mitochondrial DnaJ-like protein Mdj1p. Like Mdj1p, Tid1 localizes to human mitochondrial nucleoids, which are large protein complexes bound to mitochondrial DNA. Unlike other DnaJs, Tid1-L and -S form heterocomplexes; both unassembled and complexed Tid1 are observed in human cells. Results demonstrate that Tid1-L has a longer residency time in the cytosol prior to mitochondrial import as compared with Tid1-S; Tid1-L is also significantly more stable in the cytosol than Tid1-S, which is rapidly degraded. The longer cytosolic residency time and the half-life of Tid1-L are explained by its interaction with cytosolic Hsc70 and potential protein substrates such as the STAT1 and STAT3 transcription factors. We show that the unique carboxyl terminus of Tid1-L is required for interaction with Hsc70 and STAT1 and -3. We propose that the association of Tid1 with chaperones and/or protein substrates in the cytosol provides a mechanism for the alternate fates and functions of Tid1 in mitochondrial and nonmitochondrial pathways.

Tid1 is a human homolog of bacterial DnaJ and the Tid56 tumor suppressor of Drosophila melanogaster. DnaJ-like proteins function as co-chaperones with DnaK-like ATPases to promote the folding, translocation, and/or degradation of polypeptides (1)(2)(3). DnaJ-and DnaKlike proteins are highly conserved, and homologs are found in nearly every compartment of the eukaryotic cell and in several tumor viruses (4). In Drosophila, the absence of Tid56 results in abnormal differenti-ation and morphogenesis, giving rise to Tumorous imaginal discs that lead to lethality during early development (5). The mechanism underlying Tid56 function is unknown.
In humans, two alternatively spliced forms of Tid1 are expressed, Tid1-long (Tid1-L) and Tid1-short (Tid1-S), that differ only at their carboxyl-terminal tails (Fig. 1A) (6). Tid1-L contains 33 amino acids unique to its carboxyl terminus, whereas Tid1-S has 6 amino acids (Fig.  1A). Both Tid1-L and -S have a predicted amino-terminal mitochondrial targeting sequence as well as the signature J-domain carrying an HPD motif required for stimulating the ATPase activity of eukaryotic DnaK-like chaperones (1)(2)(3). Experiments in mice show that Tid1 is critical for early mammalian development (7). Mice specifically deficient in Tid1 in the heart develop dilated cardiomyopathy, progressive respiratory chain deficiency, and decreased copy number of mtDNA (8).
The vast majority of Tid1 localizes to mitochondria (6,9); however, its reported protein interactions and functions are primarily nonmitochondrial (10 -16, 18 -20). Tid1 has been shown to interact directly with cytosolic and nuclear proteins, which include the following: E7, an oncoprotein of human papillomavirus (21); Tax, a transcriptional activator from human T-cell leukemia virus type 1 (11,12); UL9, an origin binding protein from herpes simplex virus type 1 (13); the Ras GTPaseactivating protein (22); Jak2, a Janus kinase (10); Trk receptor tyrosine kinases (14); and hRFI, a novel protein expressed in esophageal cancers (23). The described functions of Tid1 include modulation of cell death and proliferation, in addition to participation in signaling pathways that are mediated, for example, by Jak2, NFB, Ras GTPase-activating protein, and Trk receptor tyrosine kinases.
These findings raise several fundamental questions that remain unanswered. Do Tid1 proteins function as mitochondrial DnaJ-like chaperones? Do Tid1-L and -S behave identically or differently, and do they interact with one another? What potential mechanism(s) determine whether Tid1 functions inside or outside mitochondria? In this study we demonstrate that Tid1 proteins exhibit a conserved mitochondrial DnaJ-like function, substituting for the yeast mitochondrial DnaJlike protein Mdj1p in a J-domain-dependent manner. Like Mdj1p, Tid1-L and -S localize to human mitochondrial nucleoids, which are large protein complexes associated with mtDNA. Co-immunoprecipitation experiments using isoform-specific antibodies show that unlike other DnaJ-like proteins described thus far, cellular Tid1-L and -S are present not only in an unassembled state but also as heterocomplexes. Pulse-chase analysis demonstrates that Tid1-L has a significantly longer residency time in the cytosol prior to mitochondrial import, in contrast to Tid1-S. Consistent with these findings, mutants of Tid1-L and -S lacking their amino-terminal mitochondrial targeting sequences (N-65L and N-65S, respectively) exhibit different fates in the cytosol. N-65L is relatively stable, whereas N-65S is rapidly degraded. These observations suggest that the increased cytosolic residency time and stability of Tid1-L may be explained by protein-protein interactions outside mitochondria. Results from two-step immunoprecipitation reactions show that unassembled Tid1-L interacts with cytosolic Hsc70, whereas Tid1-S does not. Unassembled Tid1-L also associates with the cytosolic STAT1 and STAT3. We show that the unique carboxyl terminus of Tid1-L is necessary for interactions with Hsc70 and the potential substrates STAT1 and STAT3. Collectively, these results support the model that the fate and function of Tid1 proteins both inside and outside of mitochondria are influenced by cytosolic protein interactions.
In Vitro Mutagenesis and Plasmids-Tid1 variants were engineered by PCR-based mutagenesis. For expression in mammalian cells, Tid1 cDNAs were cloned into pCDNA3.1 or pIRES2-EGFP; for expression in yeast, Tid1 cDNAs were cloned into pYC2/CT. N-65 variants lacked the mitochondrial targeting sequence predicted by PSORT (amino acids 1-65). All mutations were verified by DNA sequencing.
Yeast-A yeast strain carrying the mdj1-5 allele that encodes a temperature-sensitive variant of Mdj1p (24) was transformed with plasmids for the expression of wild type or mutant Tid1. Suppression of the temperature-sensitive mdj1-5 growth defect was assayed by growing transformed strains to mid-log phase in selective galactose-containing medium, adjusting the liquid culture to the same cell density, spotting yeast onto medium containing the nonfermentable carbon sources ethanol and glycerol (YPEG), and incubating the yeast at the permissive (30°C) or nonpermissive (37°C) temperatures. mdj1-5 and ⌬mdj1 were provided by Dr. Elisabeth Schwarz (University of Halle, Germany).
Mitochondrial Nucleoids-Mitochondrial nucleoids were isolated from HEK293EBNA cells as described previously (25). Briefly, mitochondria were thawed on ice, resuspended in NE2 buffer (0.25 M sucrose, 20 mM Tris-HCl, pH 7.6, 2 mM EDTA, 7 mM ␤-mercaptoethanol), and diluted with an equal volume of 0.5ϫ NE2 buffer to a final concentration of 5-7 mg of mitochondrial protein/ml. Spermidine (1.0 M) was added to a final concentration of 3 mM, and mitochondria were lysed by adding 20% Nonidet P-40 to a final concentration of 0.5%. After gentle stirring for 15 min, the lysate was centrifuged at 12,000 ϫ g for 20 min resulting in a supernatant (S0) and pellet (P0) fraction. P0 was resuspended in NE2 buffer as above; S0 and P0 were layered on top of step gradients composed of 3.5 ml of 20%, 2.5 ml of 40%, 1.8 ml of 60%, and 0.9 ml of 75% sucrose in gradient buffer (20 mM Tris-HCl, pH 7.6, 1 mM EDTA, 1 mM spermidine, 7 mM ␤-mercaptoethanol, 1 mM phenylmethylsulfonyl fluoride) and centrifuged at 111,000 ϫ g for 75 min. The resulting gradient fractions (S1 and P1) were analyzed for distribution of mtDNA and protein. mtDNA containing fractions from P1 were collected, diluted with 2 volumes of ice-cold gradient buffer, treated again with 0.5% Nonidet P-40 for 15 min, and centrifuged through a second step gradient at 49,000 ϫ g for 3 h to yield S2 and P2. Mitochondrial nucleoids were isolated in P2.
Immunoblotting, Immunoprecipitation, and Metabolic Labeling-The methods used have been described previously (26). Cellular protein was extracted in lysis buffer (0.5% Triton X-100, 300 mM NaCl, 50 mM Tris, pH 7.4) containing protease inhibitor mixture (Roche Applied Science), NaF (1 mM), and Na 3 VO 4 (1 mM). Immunoblotting was carried out using the ECL detection system (Amersham Biosciences). For immunoprecipitation, cell extracts were incubated with antibody overnight at 4°C; protein A-agarose was added, incubated for 4 h at 4°C, and washed three times with wash buffer. Sequential two-step immunoprecipitations were performed as follows. In the first step, cell extracts or mitochondrial extracts were immunoprecipitated with isoform-specific antibodies recognizing the unique carboxyl terminus of Tid1-S (␣S) or -L (␣L) to determine whether the other isoform was co-precipitated. In the second step, the unbound extract was immunoprecipitated with ␣L or ␣S to determine whether unassembled isoforms were present. Metabolic labeling was performed by incubating cells in methionine-free DMEM for 30 min prior to replacement with methionine-free DMEM containing [ 35 S]methionine (100 Ci/ml; Trans 35 S-Label, ICN) for the time as indicated; cells were chased in complete DMEM. Radiolabeled protein was immunoprecipitated and visualized by autoradiography.
Fluorescence Microscopy-Fluorescence microscopy was performed as described previously (26). For indirect immunofluorescence, cells adhered to coverslips were fixed with 4% paraformaldehyde and permeabilized with either 0.1% Triton X-100 or acetone. Incubations with primary antibody were at room temperature followed by incubations with Cy-2-or Cy-3-conjugated secondary antibodies. Mitochondrial staining was carried out using MitoTracker Green or MitoTracker Red (500 nM; Molecular Probes).

Tid1-L and -S Are Conserved Mitochondrial DnaJ-like Proteins-
The function of mitochondrial DnaJ-like chaperones is best understood in Saccharomyces cerevisiae; the yeast mitochondrial DnaJ-like protein Mdj1p is essential for the quality control of matrix proteins and for the metabolism of mtDNA. A chaperone system that includes Mdj1p, mitochondrial Hsp70 (mtHsp70, also referred to as Ssc1p), and the mitochondrial nucleotide exchange factor GrpE (Mge1p) prevents protein aggregation and promotes the folding or degradation of non-native polypeptides. Mdj1p is also present in large protein-mtDNA complexes, called mitochondrial nucleoids, and is important for maintaining the mtDNA polymerase in an active conformation during cell stress (27,28).
As there are no functional assays for mitochondrial DnaJ-like activities in mammalian cells, we examined whether Tid1 proteins substituted for Mdj1p in yeast. Tid1-L and -S were expressed in a mutant strain mdj1-5 that produces a temperature-sensitive variant of Mdj1p. At 30°C mdj1-5 exhibits wild type growth, and at 37°C mdj1-5 fails to grow on the nonfermentable carbon sources ethanol and glycerol that require mitochondrial function. Expression of either Tid1-L or -S alone suppressed the growth defect of mdj1-5 at 37°C (Fig. 1B); thus Tid1 proteins functionally replace Mdj1p in the folding and degradation of mitochondrial proteins. Deletion of the amino-terminal mitochondrial targeting sequence of Tid1 from amino acids 1 to 65 as predicted by PSORT (N-65L or N-65S) abolished Mdj1p-like activity (Fig. 1B); Tid1 import into the mitochondrial matrix was therefore required for suppression of the mdj1-5 growth defect. Not all mammalian mitochondrial targeting sequences are functional in yeast; however, the native amino-terminal sequence of Tid1 was able to direct efficient import into yeast mitochondria. Mutational analyses of various DnaJ-like proteins demonstrate that the HPD motif within the J-domain is essential for stimulating the ATPase activity of a cognate DnaK-like protein (29). J-domain mutants of Tid1 were engineered in which the conserved histidine at position 121 was replaced by a glutamine (L H121Q or S H121Q ). When expressed in mdj1-5, L H121Q and S H121Q failed to suppress the respiration-dependent growth defect at 37°C (Fig. 1B), suggesting that Tid1-L and -S cooperate with mtHsp70. Successive truncation of the unique carboxyl terminus of Tid1-L did not interfere with suppression of the mdj1-5 growth defect (Fig. 1B), demonstrating that the unique carboxyl terminus of Tid1 proteins was not essential for Mdj1p-like function. The inability of the N-65 or J-domain mutants to substitute for Mdj1p was not attributed to a lack of protein expression. All mutant Tid1 proteins were expressed in mdj1-5 cells at similar levels with the exception of N-65L, which was consistently detected at higher levels for reasons that we do not understand (Fig. 1B).
Previous work has shown that Mdj1p is predominantly associated with the mitochondrial inner membrane (24,30) and bound to mtDNA (27). Experiments were performed to determine the submitochondrial distribution of Tid1 in human cells. Mitochondria were purified using the procedure of Spelbrink and co-workers (25). In mitochondrial extracts from HEK293EBNA cells using 0.5% Triton X-100, 4 -5-fold more Tid1-S than Tid1-L was observed; a similar ratio of Tid1-S to Tid1-L was detected in whole cell extracts (see Fig. 2, C and D). To determine the submitochondrial localization of Tid1, mitochondria were lysed with 0.5% Nonidet P-40, and the resulting lysates were centrifuged. Tid1 distributed to the pellet fraction (P0) and was not detected in the supernatant (S0) ( Fig. 2A, left panel). In a similar manner, the nucleoid marker mitochondrial TFAM distributed to the pellet fraction and was not detected in the supernatant. The S0 and P0 fractions were subjected to sucrose gradient centrifugation (1°gradient) resulting in supernatant (S1) and pellet (P1) fractions. Immunoblotting results showed that after the 1°gradient, Tid1, TFAM, and the mitochondrial inner membrane marker Cox2 (cytochrome c oxidase II) distributed to the P1 fractions (Fig. 2, A and B), which contained mtDNA (data not shown). The P1 fractions were pooled and treated for a second time with Nonidet P-40 and then subjected to a second sucrose gradient centrifugation (2°gradient), which was required to separate inner membrane proteins such as Cox2 from the nucleoid components TFAM and mtDNA. After the 2°gradient, Tid1-L and -S co-purified with the nucleoid marker TFAM in the pellet fractions (P2, f19, and f20) ( Fig. 2A) and were not found with the inner membrane marker Cox2 in the supernatant (S2) (Fig. 2B). Tid1 proteins like Mdj1p are localized to the inner membrane and co-purify with mitochondrial nucleoids. Taken together, these results demonstrate that Tid1-L and -S exhibit a conserved chaperone activity in yeast and behave as mitochondrial DnaJ-like proteins in cultured human cells, which co-purify with mitochondrial nucleoids.
Hetero-oligomeric as Well as Unassembled Tid1-L and -S Are Observed in Cultured Cells-To investigate the assembly state of Tid1-L and -S, two-step immunoprecipitations were performed. To determine the lev-els of complexed and unassembled Tid1-L, cell extracts were first immunoprecipitated with isoform-specific antibodies recognizing the unique carboxyl terminus of Tid1-S (␣S) to determine whether the long isoform was co-precipitated; the unbound extract was then subjected to a second sequential immunoprecipitation using antibodies recognizing the unique carboxyl terminus of Tid1-L (␣L) to determine whether unassembled Tid1-L was present. Using the same protocol, complexed and unassembled Tid1-S were also determined by sequential immunoprecipitations except that ␣L was used in the first step and ␣S in the second step (Fig. 2, C-E). The isoform specificity of these antibodies was demonstrated as shown in supplemental Fig. 1. In whole cell extracts as well as in mitochondrial extracts from HEK293EBNA cells, Tid1-L and FIGURE 2. Tid1 submitochondrial localization and oligomerization. A, left panel, isolated mitochondria from HEK293EBNA cells were lysed with Nonidet P-40 (0.5% final concentration) and centrifuged, resulting in a supernatant (S0) and pellet (P0) fraction that were immunoblotted with antibodies recognizing Tid1-S and the nucleoid marker mitochondrial TFAM. A, right panel, and B, S0 and P0 were centrifuged through a sucrose step gradient (1°grad). The resulting gradient fractions (S1 and P1) were collected and immunoblotted for Tid1, TFAM, and the inner membrane protein Cox2. P1 fractions containing mtDNA were subjected to a second detergent treatment and sucrose gradient centrifugation (2°grad) resulting in supernatant (S2) and pellet fractions (P2). Mitochondrial nucleoids were isolated in the P2 fractions (generally f18 -f20). Tid1 distributes with TFAM in P2 fractions (A) and not with Cox2, which is in S2 fractions (B). B, right panel, longer exposure of immunoblot shown in left panel, demonstrating that Tid1-L as well as Tid1-S co-purify with mitochondrial nucleoids. C-E, sequential two-step immunoprecipitations demonstrate the presence of heterocomplexes and unassembled Tid1. Lanes 1, total whole cell or mitochondrial extract as indicated; lanes 2 and 4, both complexed and unassembled Tid1; and lanes 3 and 5, unassembled Tid1-L or -S. C and D, HEK293EBNA cells: lanes 1, whole cell extract (400 g) or mitochondrial extract (30 g); lanes 2, cell extracts (2 mg) or mitochondrial extracts (150 g) were first immunoprecipitated with isoform-specific antibodies recognizing the unique carboxyl terminus of Tid1-S (␣S) to determine whether the long isoform was co-precipitated; lanes 3, the unbound extract was then subjected to a second immunoprecipitation using antibodies recognizing the unique carboxyl terminus of Tid1-L (␣L) to determine whether unassembled Tid1-L was present; lanes 4, cell extracts (2 mg) or mitochondrial extracts (150 g) were first immunoprecipitated with ␣L; lanes 5, unbound extract was immunoprecipitated with ␣S. E, U2OS cells: lanes 1-5 are the same as in C and D except that in lane 1 whole cell extract (20 g) was assayed; lanes 2-5, whole cell extract (1 mg) was used for sequential immunoprecipitations. Immunoblotting was performed using ␣Tid1 antibodies that recognize both isoforms.
-S were found to co-immunoprecipitate with one another in the first step (Fig. 2, C and D, lanes 2 and 4). Densitometric analysis suggested that the stoichiometry of long and short isoforms in Tid1 heterocomplexes was 1:1 in HEK293EBNA whole cell and mitochondrial extracts (Fig. 2, C and D, compare lanes 2 and 4). Sequential second step immunoprecipitation demonstrated that unassembled Tid1-L and -S were also present in both whole cell and mitochondrial extracts (Fig. 2, C and  D, lanes 3 and 5). In HEK293EBNA cells, the level of Tid1-S was 4 -5fold greater than Tid1-L in both whole cell extracts and in mitochondria (Fig. 2, C and D, lanes 1). We observed that the relative levels of Tid1-L and -S vary depending on the cell line assayed. For example, in U2OS cells, the ratio of Tid1-L and -S is essentially 1:1 (Fig. 2E, lane 1). Twostep immunoprecipitations using U2OS cell extracts demonstrated that Tid1-L and -S were also found in both a hetero-oligomeric complex as well as in an unassembled state (Fig. 2E, lanes 2-5). In U2OS cells, densitometric analysis suggested that the stoichiometry of long and short isoforms in Tid1 heterocomplexes in this cell line was between 2:1 and 1:1 (Fig. 2E, compare lanes 2 and 4).
Tid1-L Has a Longer Cytosolic Residency Time Prior to Mitochondrial Import and Greater Cytosolic Stability as Compared with Tid1-S-The biosynthesis and fate of Tid1-L and -S were examined by metabolic labeling pulse-chase analysis and fluorescence microscopy. Indirect immunofluorescence demonstrated that endogenous as well as overexpressed Tid1 were predominantly localized to mitochondria (Fig. 3, A  and B). In U2OS cells, endogenous Tid1-L and -S co-localized with the mitochondrial dye Mitotracker Green (Fig. 3A). Similarly, in COS-7 cells, overexpressed Tid1-L and -S co-localized with Mitotracker ( Fig.  3B and supplemental Fig. 2). Pulse-chase analysis demonstrated that Tid1-L and -S were synthesized as full-length precursor proteins that were post-translationally processed to the mature mitochondrial form (Fig. 3D). COS-7 cells overexpressing Tid1-L or -S were metabolically labeled with [ 35 S]methionine for 30 min and then chased for up to 12 h. Immunoprecipitation of Tid1 showed that the precursor of Tid1-S was efficiently imported into mitochondria; in the pulse, 55% of newly synthesized Tid1-S was already processed to the mature mitochondrial form. By contrast, the precursor of Tid1-L was processed more slowly; in the pulse 84% of the newly synthesized Tid1-L was present as the precursor form, and even after a 12-h chase the precursor protein was still detectable. Once imported into mitochondria, both Tid1-L and -S were for the most part stable.
The cytosolic stability of Tid1-L and -S was addressed by analyzing the half-life of cytosolic Tid1 mutants N-65L and N-65S lacking their respective mitochondrial targeting sequences. Indirect immunofluorescence demonstrated that little if any overexpressed N-65L and N-65S co-localization with Mitotracker ( Fig. 3C and supplemental Fig. 2); rather, these proteins were distributed throughout the cytosol co-localizing with cytosolic green fluorescent protein (data not shown). Pulsechase experiments showed that N-65S was rapidly turned over; after a 30-min pulse and a 2-h chase, little if any N-65S was observed (Fig. 3E,  left panel). By contrast, N-65L exhibited a significantly longer half-life; after a 12-h chase, N-65L was still present (Fig. 3E, right panel). The unique carboxyl terminus of Tid1-L thus confers increased stability outside mitochondria. Taken together, these data suggested the possibility that Tid1-L may interact with cytosolic proteins resulting in protease resistance, either by blocking degradation or promoting the folded state. The interaction of Tid1-L with cytosolic proteins may also explain the slower mitochondrial import kinetics of Tid1-L as compared with Tid1-S as shown in Fig. 3D.
Hsc70 Interacts with Unassembled Tid1-L-In bacteria and eukaryotes, DnaJ-like proteins have been shown to physically interact with DnaKlike (Hsp70-like) proteins, which function together as a chaperone system in the folding, translocation, and degradation of unfolded polypeptides (1,(31)(32)(33). The ability of Tid1-L and -S to substitute for yeast Mdj1p in a J-domain-dependent manner suggests that Tid1 cooperates with Hsp70-like proteins. In addition, it has been shown that some mitochondrial precursor proteins synthesized in the cytosol interact with cytosolic Hsc70, which facilitates the mitochondrial import process (34 -36). Co-immunoprecipitations were performed to determine the extent to which Tid1-L and -S interact with mtHsp70 and Hsc70. Endogenous Tid1-L and -S from 293-T cell extracts were immunoprecipitated with isoform-specific antibodies, and immunoblotting was used to detect co-precipitated mtHsp70 or Hsc70. Each Tid1 isoform co-precipitated similar amounts of mtHsp70 as predicted; by contrast, Tid1-L co-precipitated a higher level of cytosolic Hsc70 as compared with Tid1-S (Fig. 4A). Tid1 isoforms also co-precipitated one another (Fig. 4A), consistent with results shown in Fig. 2, C, D, and E. To determine more precisely whether Tid1-L or -S bound to Hsc70, two-step immunoprecipitations were carried out in U2OS cell extracts. First step immunoprecipitations with ␣S antibodies showed that no detectable Hsc70 was present (Fig. 4B, lane 1). Second step immunoprecipitation with ␣L antibodies demonstrated that unassembled Tid1-L was associated with Hsc70 (Fig. 4B, lane 2). When ␣L antibodies were used in the first step immunoprecipitation, Hsc70 was co-precipitated (Fig. 4B, lane  3); second step immunoprecipitations with ␣S antibodies demonstrated that unassembled Tid1-S did not associate with Hsc70 (Fig. 4B, lane 4). Collectively, these results demonstrated that Hsc70 was associated with unassembled Tid1-L and that significantly less or no Hsc70 interacted with unassembled Tid1-S or with Tid1 heterocomplexes. Densitometric analysis of immunoblotting results from U2OS cells showed that Tid1-L co-precipitated ϳ0.6% of the total cellular Hsc70, which is an abundant cellular protein. Co-immunoprecipitation experiments using isolated mitochondria demonstrated that mitochondrial Tid1-L or -S did not interact with Hsc70 (supplemental Fig. 3), supporting the notion that Tid1-L interacts with Hsc70 in the cytosol.
Tid1 isoforms differ only at their extreme carboxyl termini suggesting that this region mediates association with Hsc70. The mitochondrial targeting sequence of Tid1 is another potential region of interaction with Hsc70. Carboxyl-terminal truncations within full-length mitochondrial Tid1-L, lacking the last 3, 13, 21, or 32 amino acids (L-3, L-13, L-21, and L-32, respectively), were engineered for mammalian cell expression ( Fig. 4C and see Fig. 1). These mutants were overexpressed in 293-T cells at levels similar to overexpressed wild type Tid1 (data not shown), consistent with the expression levels observed in yeast (Fig. 1B). Deletion of up to 21 amino acids within the carboxyl terminus of Tid1-L (L-3, L-13, and L-21) did not affect the levels of co-precipitated Hsc70; by contrast, complete removal of the unique carboxyl-terminal tail (L-32) substantially reduced Hsc70 interaction. These immunoprecipitation experiments required the use of antibodies that recognize endogenous Tid1 as well as overexpressed mutant Tid1; thus, the presence of endogenous Tid1-Hsc70 complexes in all of the immunoprecipitated samples complicated interpretation of the results. Therefore, a second set of carboxyl-terminal truncation mutants were analyzed in which the amino-terminal targeting sequence was replaced by the Myc epitope (mycN-65L or mycN-65S). Introduction of the Myc epitope permitted us to distinguish the Tid1 mutants from endogenous Tid1. Immunoprecipitation of mycN-65 Tid1 proteins showed that Hsc70 co-precipitated only with N-65L; little if any Hsc70 was associated with N-65S. These results are consistent with data showing that higher levels of Hsc70 associate with unassembled Tid1-L as compared with unassembled Tid1-S. These results also show that the mitochondrial targeting sequence of Tid1 was not necessary for Hsc70 association. However, removal of three or more amino acids from the carboxyl terminus of N-65L interfered with Hsc70 association. Taken together, these results suggest that the unique carboxyl terminus of Tid1-L either contains a discrete peptide sequence that FIGURE 4. Unassembled Tid1-L interacts with Hsc70 but Tid1-S does not. A, endogenous Tid1-L or -S from 293-T cell extracts (800 g) were immunoprecipitated (IP) with ␣S or ␣L and immunoblotted with antibodies recognizing mtHsp70, Hsc70, or Tid1. B, sequential two-step immunoprecipitation from U2OS cell extracts (1 mg) followed by immunoblotting with ␣Hsc70 or ␣Tid1. Lane 1, extracts were first immunoprecipitated with ␣S; lane 2, unbound extract was subjected to a second immunoprecipitation with ␣L. Lane 3, extracts were first immunoprecipitated with ␣L; lane 4, unbound extract was subjected to a second immunoprecipitation with ␣S. mediates Hsc70 association or influences the conformational state of Tid1-L that promotes Hsc70 interaction.
STAT1 and STAT3 Interact with Unassembled Tid1-L-Tid1-L and Hsc70 may cooperate with one another to form a cytosolic chaperone system in the biogenesis and function of protein substrates. A first step in understanding the physiological function of such a chaperone system requires the identification of protein substrates. Previous work has shown that overexpressed Tid1-L or -S leads to the inhibition of IFN-␥stimulated transcription (10). Overexpression of Tid1-L or -S in COS-1 cells inhibits IFN-␥-stimulated expression of a luciferase reporter gene whose expression is controlled by IFN-␥-activated sequence elements. Data from that study also showed that Tid1 proteins co-immunoprecipitate with Jak2 as well as with the IFN-␥ receptor R2 subunit.
Based on these findings we investigated whether endogenous Tid1-L or -S interact similarly or differently with components of the IFN-␥ pathway. The interaction between Tid1 and STAT1 and -3 was assayed in U2OS cells, as their expression levels permitted examination of endogenous associations. Tid1-L co-immunoprecipitated STAT1 as well as STAT3; by contrast, Tid1-S showed little interaction with these STATs (Fig. 5A). When cells were treated with IFN-␥, Tid1-L remained associated with active phosphorylated STAT1 (Fig. 5B). The small amount of STAT1 observed in the anti-Tid1-S immunoprecipitate was perhaps mediated by Tid1-L present in heterocomplexes with Tid1-S. Support for this interpretation was provided by sequential immunoprecipitation experiments, which demonstrated that unassembled Tid1-L associated with STAT1 and phospho-STAT1, whereas unassembled Tid1-S did not (Fig. 5C). As IFN-␥ does not activate STAT3 in U2OS cells, IL-6 treatment was used instead. Tid1-L co-immunoprecipitated STAT3 in the presence and absence of IL-6, whereas Tid1-S did not (Fig. 5D). We were not able to confirm the association of Tid1-L with phospho-STAT3 because the two commercially available antibodies tested did not detect phospho-STAT3 by immunoblotting even in cell extracts. Results of two-step immunoprecipitation experiments demonstrated that unassembled Tid1-L associated with STAT3, whereas Tid1-S associated with much lower levels of STAT3 (Fig. 5E). To determine whether the intact carboxyl terminus of Tid1-L was required for STAT association, the N-65L truncation mutants were assayed. Results showed that deletion of three or more amino acids from the carboxylterminal tail of N-65 L interfered with STAT1 and -3 interactions (Fig.  5F). Taken together, these results demonstrate that the unique carboxyl-terminal tail of Tid1-L is necessary for interactions not only with Hsc70 but also with the potential cytosolic protein substrates STAT1 and STAT3.

DISCUSSION
The physiological importance of Tid1 proteins in a variety of cellular processes, including the immune response, apoptosis, senescence, and development, has been established (7, 8, 14 -20, 37, 38). However, the mechanism underlying Tid1 function either inside or outside mitochondria is poorly understood. Results presented here demonstrate that Tid1-L and -S exhibit a conserved DnaJ-like chaperone activity within the mitochondrial matrix; when expressed individually, each isoform is able to replace the yeast mitochondrial DnaJ-like protein Mdj1p (Fig.  1B). The ability of Tid1 proteins to substitute for Mdj1p requires an intact J-domain that is required for stimulating the ATPase activity of mtHsp70 and an amino-terminal mitochondrial targeting sequence. These results suggest that Tid1 cooperates with mtHsp70 and functions as a chaperone system in yeast and mammalian mitochondria. Like . Unassembled Tid1-L interacts with STAT1 and -3 but Tid1-S does not. A, endogenous Tid1-L and -S were immunoprecipitated (IP) from U2OS cell extracts (1 mg) using ␣L or ␣S. These samples and the whole cell extract (WCE, 4 g) were immunoblotted with antibodies recognizing Tid1, STAT1, or STAT3. B, endogenous Tid1-L and -S were immunoprecipitated with ␣L or ␣S from protein extracts derived from U2OS cells treated with or without IFN-␥ (1,000 units/ml) for 15 min at 37°C; samples were immunoblotted with antibodies recognizing Tid1, STAT1, or tyrosine-phosphorylated STAT1 (P-STAT1). C, sequential two-step immunoprecipitation from U2OS cell extracts of endogenous Tid1 followed by immunoblotting with antibodies recognizing Hsc70 or Tid1. Lane 1, extracts were first immunoprecipitated with ␣S; lane 2, unbound extract was subjected to a second immunoprecipitation with ␣L. Lane 3, extracts were first immunoprecipitated with ␣L; lane 4, unbound extract was subjected to a second immunoprecipitation with ␣S. Tid1 heterocomplexes are immunoprecipitated in lanes 1 and 3; unassembled Tid1-L or -S are immunoprecipitated in lanes 2 and 4, respectively. D, endogenous Tid1-L or -S from U2OS cells treated with or without IL-6 (20 ng/ml) for 30 min at 37°C were immunoprecipitated using ␣L or ␣S and immunoblotted using antibodies recognizing Tid1 and STAT3. E, sequential two-step immunoprecipitation as in C, except immunoprecipitated samples were immunoblotted with antibodies recognizing Tid1 and STAT3. F, Myc epitope-tagged N-65 mutants of Tid1 were transiently expressed in U2OS cells and were immunoprecipitated with ␣Myc and then immunoblotted for STAT1, STAT3, or Tid1.
Mdj1p, Tid1-L and -S co-purify with mitochondrial nucleoids isolated from human cells. Previous experiments show that Mdj1p is required for the maintenance of mtDNA and is important in maintaining mtDNA polymerase in an active conformation during cell stress (27,28). Taken together, these data support the conserved function of Tid1 in the folding and degradation of mitochondrial proteins as well as in the maintenance of mtDNA.
Unlike other DnaJ-like proteins, Tid1-L and -S are present as both hetero-oligomers and unassembled proteins in isolated mitochondria and whole cell extracts (Fig. 2, C and D, Fig. 4, A and B, and Fig. 5, A-E). Previous studies show that the eukaryotic DnaJ-like proteins Sis1p and Ydj1p function as homodimers (39,40). It is likely that Tid1 isoforms function as both homodimers and heterodimers. However, further experiments are required to determine the precise stoichiometry of Tid1 complexes and the extent to which hetero-and homo-complexes occur in mitochondria or other cellular compartments. The relative ratio of the two isoforms in mitochondria and whole cell extracts depends upon the cell type examined. For example, in HEK293EBNA or 293-T cells there is 4 -5-fold more Tid1-S than -L, whereas in U2OS cells, the level of Tid1 isoforms is virtually the same (Fig. 2, C-E). One can envisage that the different relative levels and assembly states of Tid1 isoforms provide a means for diversifying and regulating the cellular function of Tid1.
Tid1-L and -S have significantly different rates of protein import into mitochondria and different residency times (i.e. half-lives) in the cytosol (Fig. 3). Pulse-chase experiments show that the newly synthesized cytosolic precursor of Tid1-S is short lived; during a 30-min pulse, the majority of full-length Tid1-S was processed to the mature mitochondrial form. By contrast, the precursor of Tid1-L was relatively long lived; during the pulse, full-length Tid1-L was the predominant form observed and was present even after a 12-h chase. At least two factors contribute to these differences as follows: 1) Tid1-L and -S have different intrinsic rates of protein import into mitochondria as suggested by pulse-chase experiments; and 2) Tid1-L and -S have different sensitivities to proteolysis in the cytosol. The cytosolic stability of Tid1-L and -S was addressed by analyzing the half-lives of Tid1 mutants lacking their respective mitochondrial targeting sequences (N-65S and N-65L). Results show that N-65S is more sensitive to protein turnover; after a 2-h chase little metabolically labeled protein is detected. By contrast, N-65L is significantly more stable; even after 12-h chase 40% of the labeled protein is still present. Indirect immunofluorescence shows that N-65S and N-65L are both nonmitochondrial and diffusely distributed throughout the cytosol. N-65L also appears to distribute to the plasma membrane and to punctate structures throughout the cytosol; however, these localization patterns may result from its increased stability and higher cytosolic protein levels as compared with N-65S.
The different sensitivity to protein degradation and rate of mitochondrial protein import of Tid1-L and -S correlate with differences in their association with Hsc70 and cytosolic protein substrates (Figs. 4 and 5). Tid1-L interacts with Hsc70 and substrate proteins such as STAT1 and -3, potentially explaining its cytosolic stability and slower rate of mitochondrial import. By contrast, little if any Tid1-S associates with Hsc70 or STAT1 or -3; in the absence of such protein interactions Tid1-S is unstable and is not retained in the cytosol. Results of co-immunoprecipitation experiments show that primarily unassembled Tid1-L interacts with the Hsc70 and STAT1 and -3. When Tid1-L is complexed with Tid1-S, little or no association with Hsc70 is detected (Fig. 4, A and B,  lane 1). A direct interaction between unassembled Tid1-L and Hsc70 is likely, in light of previous work demonstrating the direct interaction between purified DnaJ-and DnaK-like proteins (1,41). In addition, unassembled Tid1-L co-immunoprecipitates the transcription factors STAT1 and STAT3; by contrast, little if any unassembled Tid1-S coimmunoprecipitates with these STATs (Fig. 5). Previous studies demonstrate that STAT1 and STAT3 are present in the cytosol as latent monomers or dimers; upon cytokine activation, STAT1 and -3 are phosphorylated, and the active homo-or heterodimers translocate to the nucleus where they promote gene expression (42)(43)(44)(45)(46)(47). Our results show that cellular levels of Tid1-L influence STAT1-dependent IFN-␥ signaling; depletion of Tid1-L using RNA interference results in an upregulation of IFN-␥-activated transcription and an increase in STAT1 phosphorylation. 4 Further experiments are required to determine whether Tid1-L and Hsc70 function together as a chaperone system that modulates STAT1 and -3 or other nonmitochondrial proteins.
The following model is proposed to resolve the apparent paradox of how proteins targeted primarily to mitochondrial function in nonmitochondrial pathways (Fig. 6). Newly synthesized Tid1 that is produced in the cytosol as a full-length precursor protein has alternate fates. Like other mitochondrial matrix proteins encoded by nuclear DNA and synthesized in the cytosol, Tid1 carries an amino-terminal targeting sequence that directs its translocation into the mitochondrial matrix (Fig. 6a). Upon import the targeting signal is cleaved off thus generating the mature active mitochondrial protein, which functions in mitochondrial biogenesis and quality control. Tid1 also has an alternate fate in the cytosol. In this study we show that Tid1-L is associated at steady state with the cytosolic proteins Hsc70 and STAT1 and -3 (Figs. 4 and 5). Two potential mechanisms for the nonmitochondrial localization of Tid1 are shown (Fig. 6, b and c). In one pathway, newly synthesized Tid1 precursor proteins associate with cytosolic Hsc70 and/or substrate proteins, which prevent protein import and processing (Fig. 6b). Another pathway resulting in the nonmitochondrial localization of Tid1 involves the partial translocation of Tid1 across the outer and inner membranes of the mitochondria, followed by cleavage of the targeting sequence and retrograde movement of the processed protein back into the cytosol 4 B. Lu, S. Kotenko, and C. K. Suzuki, unpublished data. FIGURE 6. Model for the alternate cellular fates of Tid1. Tid1-L and -S are encoded by nuclear genes and synthesized as full-length precursor proteins in the cytosol, which carry an amino-targeting presequence that directs their post-translational import into the mitochondrial matrix. a, the full-length precursor is targeted to mitochondria and translocated into the matrix whereupon its targeting sequence is cleaved off by the mitochondrial processing peptidase (MPP, scissors) resulting in the processed protein (*). Matrix localized Tid1 functions in the folding and degradation of mitochondrial proteins and mtDNA maintenance. b, the full-length Tid1 precursor (**) interacts with nonmitochondrial proteins in the cytosol such as Hsc70, STAT1, and -3 and is not targeted to mitochondria. c, the full-length Tid1 precursor is partially translocated across the mitochondrial outer and inner membranes and cleaved by MPP; the processed Tid1 (*) moves back into the cytosol. Interaction of processed Tid1 with cytosolic proteins may occur either before or after partial translocation across mitochondrial membranes. The association of Tid1 with proteins in the cytosol functions in diverse processes including cell growth and signaling. (Fig. 6c). Yeast fumarase has been shown to distribute in this manner to the cytosol as well as the mitochondria. Pines and co-workers (48) demonstrated that full-length fumarase is partially translocated across the mitochondrial outer and inner membranes whereupon its targeting sequence is cleaved off (48). The majority of processed fumarase (70%) moves back into the cytosol, and the remaining fraction is fully translocated into the matrix. In the results presented here, endogenous Hsc70 was shown to co-immunoprecipitate only with endogenous Tid1-L that is processed (42 kDa) and not the precursor form (50 kDa). We were curious to know whether STAT1 or STAT3 co-precipitated the processed or full-length form of Tid1-L; however, commercially available antibodies recognizing STAT1 or -3 were not suitable for immunoprecipitation experiments in U2OS or 293-T cells. By using another approach, we examined the association of Tid1-L with the Janus kinase Jak2, which phosphorylates STAT1 and STAT3. Published work has shown that Tid1 and Jak2 interact with one another directly in the yeast two-hybrid assay and in cultured mammalian cells (10). When overexpressed in 293-T cells, Jak2 was found to co-immunoprecipitate both the full-length and processed form of overexpressed Tid1-L; in the absence of overexpressed Tid1-L, Jak2 co-precipitated only the fulllength form of endogenous Tid1-L (supplemental Fig. 4). Taken together, these results suggest that full-length unprocessed Tid1-L may also associate with nonmitochondrial proteins such as Jak2, Hsc70, STAT1, or STAT3 in the cytosol.
The behavior and function(s) of Tid1-L and -S likely vary, depending on cell type, because of differences in the relative expression of Tid1 and Hsp70 chaperones, as well as differences in the expression of substrate proteins. When the expression level of cytosolic Hsp70 and substrate is relatively high, full-length Tid1 may be more efficiently retained in the cytosol. In addition, it is conceivable that in certain cell types the unique carboxyl terminus of Tid1-S, like Tid1-L, also mediates interactions with a subset of chaperones and protein substrates in the cytosol, thereby promoting its cytosolic stability, retention, and function. Such interactions are predicted to be required for the nonmitochondrial localization and function of Tid1. Further experiments are required to determine the mechanism(s) underlying the alternate fates and functions of Tid1 and whether cytosolic retention of Tid1 affects mitochondrial function.