Inner Mitochondrial Translocase Tim50 Interacts with 3β-Hydroxysteroid Dehydrogenase Type 2 to Regulate Adrenal and Gonadal Steroidogenesis*

Background: The role of mitochondrial translocases in steroid hormone synthesis is investigated. Results: Expression of inner mitochondrial translocase Tim50 is essential for DHEA and androstenedione synthesis. Conclusion: Tim50 interacts with 3βHSD2 for steroid hormone synthesis. Significance: During steroidogenesis, the flexibility of 3βHSD2 is essential to coordinate interaction between Tim50 and Tom22. In the adrenals, testes, and ovaries, 3β-hydroxysteroid dehydrogenase type 2 (3βHSD2) catalyzes the conversion of pregnenolone to progesterone and dehydroepiandrostenedione to androstenedione. Alterations in this pathway can have deleterious effects, including sexual development impairment, spontaneous abortion, and breast cancer. 3βHSD2, synthesized in the cytosol, is imported into the inner mitochondrial membrane (IMM) by translocases. Steroidogenesis requires that 3βHSD2 acts as both a dehydrogenase and isomerase. To achieve this dual functionality, 3βHSD2 must undergo a conformational change; however, what triggers that change remains unknown. We propose that 3βHSD2 associates with IMM or outer mitochondrial membrane translocases facing the intermembrane space (IMS) and that this interaction promotes the conformational change needed for full activity. Fractionation assays demonstrate that 3βHSD2 associated with the IMM but did not integrate into the membrane. Through mass spectrometry and Western blotting of mitochondrial complexes and density gradient ultracentrifugation, we show that that 3βHSD2 formed a transient association with the translocases Tim50 and Tom22 and with Tim23. This association occurred primarily through the interaction of Tim50 with the N terminus of 3βHSD2 and contributed to enzymatic activity. Tim50 knockdown inhibited catalysis of dehydroepiandrostenedione to androstenedione and pregnenolone to progesterone. Although Tim50 knockdown decreased 3βHSD2 expression, restoration of expression via proteasome and protease inhibition did not rescue activity. In addition, protein fingerprinting and CD spectroscopy reveal the flexibility of 3βHSD2, a necessary characteristic for forming multiple associations. In summary, Tim50 regulates 3βHSD2 expression and activity, representing a new role for translocases in steroidogenesis.

In the adrenals, testes, and ovaries, 3␤-hydroxysteroid dehydrogenase type 2 (3␤HSD2) catalyzes the conversion of pregnenolone to progesterone and dehydroepiandrostenedione to androstenedione. Alterations in this pathway can have deleterious effects, including sexual development impairment, spontaneous abortion, and breast cancer. 3␤HSD2, synthesized in the cytosol, is imported into the inner mitochondrial membrane (IMM) by translocases. Steroidogenesis requires that 3␤HSD2 acts as both a dehydrogenase and isomerase. To achieve this dual functionality, 3␤HSD2 must undergo a conformational change; however, what triggers that change remains unknown. We propose that 3␤HSD2 associates with IMM or outer mitochondrial membrane translocases facing the intermembrane space (IMS) and that this interaction promotes the conformational change needed for full activity. Fractionation assays demonstrate that 3␤HSD2 associated with the IMM but did not integrate into the membrane. Through mass spectrometry and Western blotting of mitochondrial complexes and density gradient ultracentrifugation, we show that that 3␤HSD2 formed a transient association with the translocases Tim50 and Tom22 and with Tim23. This association occurred primarily through the interaction of Tim50 with the N terminus of 3␤HSD2 and contributed to enzymatic activity. Tim50 knockdown inhibited catalysis of dehydroepiandrostenedione to androstenedione and pregnenolone to progesterone. Although Tim50 knockdown decreased 3␤HSD2 expression, restoration of expression via proteasome and protease inhibition did not rescue activity. In addition, protein fingerprinting and CD spectroscopy reveal the flexibility of 3␤HSD2, a necessary characteristic for forming multiple associations. In summary, Tim50 regulates 3␤HSD2 expression and activity, representing a new role for translocases in steroidogenesis.
Most mitochondrial proteins are encoded by nuclear genes, synthesized as precursors on cytosolic polysomes (1), and imported across the outer (OMM) 3 and inner mitochondrial membranes (IMM) (2)(3)(4)(5). Loosely associated protein complexes at the mitochondrial membranes, termed translocases, mediate protein import and sorting. Translocase complexes found at the OMM are called Tom (translocase, outer membrane), whereas those at the IMM are called Tim (translocase, inner membrane) (3,6,7). The complexes are not tightly linked to each other, making it difficult to accurately determine the precise mechanism of translocation (7,8). Most proteins enter the mitochondria via the Tom40 channel. After crossing the OMM, Tim50 of the Tim23 complex receives proteins that contain a presequence. Interaction of Tim50 with the preprotein promotes binding of the presequence to the IMS domain of Tom22 (9). Tim50, anchored in the IMS with one transmembrane helix, has a large C-terminal hydrophobic domain of ϳ350 amino acids exposed to the IMS (10) that coordinates function by maintaining the permeability barrier (11)(12)(13). Indeed, loss of Tim50 leads to a release of mitochondrial cytochrome c in cells as well as developmental defects in zebrafish embryos due to increased cell death (14). Tim21, of the Tim23 complex, then binds to Tom22, causing the release of the preprotein and its subsequent entry into the Tim23 channel. Preproteins are then sorted to the IMS or mitochondrial matrix (15,16).
3␤HSD2 has both dehydrogenase and isomerase activities. We hypothesize that the bifunctional activity of 3␤HSD2 requires the protein to undergo a conformational change. We propose that 3␤HSD2 associates with the IMM or OMM translocases facing the IMS and that this interaction promotes a conformational change leading to the two different enzymatic activities. We found that 3␤HSD2 associated with the IMM but did not integrate into it. This likely was the result of the interaction between 3␤HSD2 and Tim50, resulting in a transient complex formation of 3␤HSD2, Tom22, Tim23, and Tim50. The study demonstrates the flexibility of 3␤HSD2, which allowed the protein to interact with the membrane bound domains of Tim50. Cytochrome P450c17, expressed in the endoplasmic reticulum (22)(23)(24), catalyzes the conversion of 17-OH pregnenolone to DHEA (Fig. 1A) in adrenal and testicular tissues. We show that the absence of Tim50 inhibited the catalysis of DHEA to androstenedione and pregnenolone to progesterone but did not affect the expression of endoplasmic reticulum resident cytochrome P450c17 (22)(23)(24). Moreover, restoration of 3␤HSD2 expression in the Tim50 knockdown cells via proteasome and protease inhibition did not rescue steroidogenic activity. Thus, the association of 3␤HSD2 with the translocase appears to play a key role in the regulation of enzymatic activity.

EXPERIMENTAL PROCEDURES
Isolation and Purification of Mitochondria-Mitochondria were obtained from cell lines or, as a control, from pig adrenal glands. The pig adrenals were excised immediately after animal sacrifice at the Department of Animal Sciences, University of Florida. The adrenal tissues were diced in mitochondrial isolation buffer (250 mM sucrose, 10 mM HEPES, 1 mM EGTA, pH 7.4), whereas the cell lines were washed with PBS two times and then incubated with 1 mM HEPES, pH 7.4, for 30 min at 4°C. Next, tissues and cells were homogenized in a hand-held allglass Dounce homogenizer with 40 gentle up and down strokes, and the cell debris was removed by spinning at 3500 ϫ g for 10 min. The supernatant containing the mitochondrial fraction was purified by differential centrifugation following a previously reported procedure (25,26), and the pellet was washed and then resuspended in an energy regeneration buffer (125 mM sucrose, 80 mM KCl, 5 mM MgCl 2 , 10 mM NaH 2 PO 4 , 10 mM isocitrate, 1.0 mM ATP, 1.0 mM NADP, 0.1 mM ADP, and 25 mM HEPES, pH 7.4) before storage at either Ϫ86°C or in liquid nitrogen.
Protein Import-To synthesize [ 35 S]methionine labeled proteins, the cDNAs for 3␤HSD2 and the COX IV dihydrofolate reductase fusion were subcloned into the SP6 vector, and proteins were synthesized in a cell-free system in the presence of [ 35 S]Met using TNT-rabbit reticulocyte (Promega). Ribosomes and associated incompletely translated polypeptide chains were removed by centrifugation at 150,000 ϫ g for 15 min at 4°C (27). For all protein import experiments, a 100-l mixture of isolated mitochondria (100 g) and synthesized protein was incubated in a 26°C water bath with or without proteinase K, and the import reaction was terminated by the addition of 1 mM mCCCP and an equal volume of boiling 2ϫ SDS sample buffer. The import reactions were analyzed by electrophoresis through SDS-polyacrylamide gels, fixing the gels in methanol/acetic acid (40:10), and then drying and exposing gels to a phosphorimaging screen.
Native Polyacrylamide Gel Electrophoresis (PAGE)-To analyze complex formation, 35 S-labeled 3␤HSD2 was incubated with mitochondria isolated from MA-10 cells, and the reaction was stopped by placing the mix on ice. Membranes were then lysed with buffer containing digitonin, and the samples were electrophoresed through 3-16% gradient native gels. In some cases gels were fixed, dried, and then exposed to phosphorimaging In other cases protein complexes were transferred to a membrane and then probed with antibodies against 3␤HSD2, Tom22, Tim50, or Tim23.
Mitochondrial Fractionation-The mitochondrial compartments were individually purified following a standard procedure with minor modifications. In brief, the OMM fraction was extracted with 1.2% digitonin, and the remaining mixture of matrix and IMM was purified through 0.5% Lubrol, where the matrix fraction remained in solution after centrifugation, and the IMM fraction formed the pellet (28). Proteins with a nonspecific association or loose association with the membranes were removed by incubation with proteinase K. Fractions were then processed for Western blotting using the indicated antibodies.
Proteolytic Digestion-Limited proteolytic digestion experiments were performed at room temperature or 4°C using 5 g of 3␤HSD protein and different concentrations of trypsin (sequencing grade, Promega) for either 15 or 45 min. The reactions were terminated by adding of an equal volume of SDSsample buffer containing 2 mM PMSF and then incubating in a boiling water bath. After electrophoresis, the samples were stained with Coomassie Brilliant Blue or silver nitrate.
Protein identification using the generated MS/MS spectra was performed by searching the NCBI non-redundant data base using Mascot MS/MS Ion Search at Matrix Science with consideration for the carbamidomethylation of cysteine and the oxidation of methionine.
Density Gradient Ultracentrifugation-Complexes were resolved by a sucrose density gradient (top 10% to bottom 30%) with a cushion of 200 l of 68% (2.0 M) sucrose in a final volume of 2 ml. Approximately 100 g of protein in a volume of 100 l was layered on the top and centrifuged at 4°C in a Beckman TLA55 rotor at 55,000 rpm for 1 or 4 h. After centrifugation, fractions of 125 l were collected. Gradient fractions were analyzed by native PAGE to show that the complexes were not degraded during centrifugation through the sucrose density gradient. Fractions were subjected to immunoprecipitation to determine specific interactions. An aliquot of 20 l of each fraction was electrophoresed on a 15% SDS-polyacrylamide gel, and the components were analyzed by Western blot.
Co-immunoprecipitations-Specific antibodies were preincubated with protein A-Sepharose CL-4B (0.5 g/l, GE Healthcare) in 100 l of 1ϫ co-immunoprecipitations buffer (1% Triton X-100, 200 mM NaCl, and 0.5% sodium deoxycholate) for 2 h at 4°C. To remove unbound antibody, beads were washed with 1ϫ co-immunoprecipitations buffer and then incubated with a rabbit IgG control antibody (Sigma) for 1 h to block unbound beads and then washed. The freshly isolated mitochondrial pellet (25 mg for each sample) was resuspended with ice-cold lysis buffer (20 mM Tris HCl pH 8.0, 137 mM NaCl, 10% glycerol, 1% Triton X-100, 2 mM EDTA), and unsolubilized material was removed by ultracentrifugation (30 min at 100,000 ϫ g). The supernatants were incubated with antibody-bound protein A-Sepharose beads overnight at 4°C. After washing, the protein A-Sepharose pellets were resuspended in 100 mM glycine, pH 3.0, for 10 s, and then a pretitrated volume of 1.0 M Tris, pH 9.5, was added to adjust the pH to 7.4. Immune complexes were then analyzed by Western blotting.
Metabolic Conversion Assays-Isolated mitochondria (300 g) were incubated in potassium phosphate buffer for the conversion of [ 3 H]pregnenolone to progesterone. Three million counts of [ 3 H]pregnenolone were used for each reaction and chased with 30 g of cold unlabeled progesterone. The reaction was initiated by the addition of NAD and incubated at 37°C for 4 h in a shaking water bath. To measure the conversion of DHEA to androstenedione, mitochondria were incubated with 80,000 counts of [ 14 C]DHEA, and the reaction was initiated by the addition of NAD and incubated as above. For some of the experiments, protease or proteasome inhibitors were included. For both reactions, the steroids were extracted with ether/acetone (9:1 v/v), and equal amounts of a cold pregnenolone-progesterone (50:50, Sigma) mixture in CH 2 Cl 2 were added as a carrier. The extracts were concentrated under a stream of nitrogen or air and then separated by TLC (Whatman) using a chloroform/ethyl acetate (3:1) mobile phase. The radioactivity was enhanced using [ 3 H]Enhancer, and the amounts of each steroid were determined using phosphorimaging.
Gas Chromatography-Mass Spectrometry (GCMS)-The spots extracted from TLC plates were subjected to GCMS analysis on an Agilent 7890 GC with 5975C mass spectrometer. The column used was an Agilent HP-5 with dimensions 30-m ϫ 0.25-mm inner diameter with a 0.25-m film thickness. Samples were dissolved in 50 l of dichloromethane, and 1 l was injected onto the column using a pulsed splitless injection. The temperature program was as follows: ramp at 10°C/min from 70 to 310°C and hold for 6 min. Spectra were collected in full scan mode over the mass range m/z 50 -500 to facilitate comparison of MS spectra with the NIST/EPA/NIH NIST08 mass spectral library.

RESULTS
3␤HSD2 Is an IMM-associated Protein-The specific location of 3␤HSD2 within the mitochondria is not yet known. Thus, we did fractionation studies using isolated mitochondria from mouse Leydig MA-10 cells and found 3␤HSD2 localized to the IMM (Fig. 1B). The size of the 3␤HSD2 associated with the IMM fraction appeared similar to that of the control 3␤HSD2 purified from baculovirus, indicating that the protein did not undergo signal sequence cleavage. We then evaluated if 3␤HSD2 integrates into the IMM or if it merely associates with the membrane by performing import assays with cell-free-synthesized 35 S-labeled 3␤HSD2 and isolated mitochondria from MA-10 cells (Fig. 1, C-E). We separated the non-imported (S) from the imported fraction (P) by centrifugation and analyzed membrane integration by extracting the pellet with Na 2 CO 3 . Sodium carbonate breaks protein-protein interactions but not lipid-protein interactions (3). Although the identical size of 35 Slabeled 3␤HSD2 before and after import suggests an uncleaved N terminus, a common characteristic for IMM proteins, our observation that 3␤HSD2 returned to the supernatant after carbonate extraction indicates that the protein was not integrated into the IMM. A brief incubation with a protease results in the digestion of proteins with a nonspecific membrane association as the enzyme cannot enter the membrane. Thus, the fact that incubation of fractions with proteinase K only ablated signal in the non-imported fractions suggests that a specific association between 3␤HSD2 and the IMM. As a control, we evaluated each fraction for the presence of COX IV, an integrated IMM protein facing the matrix. The Western blots validated the mitochondrial integrity during fractionation, extraction, and proteolysis (Fig. 1C). In summary, the results confirm that 3␤HSD2 was imported like a typical inner membrane protein and suggest that 3␤HSD2 may specifically associate with the IMM without integration into the membrane.
To determine the compartment specific localization of 3␤HSD2 at the IMM, we performed Western analysis on mitochondrial fractions from steroidogenic MA-10 cells and  35 S-labeled 3␤HSD2 is shown. After incubation of 35 S-labeled 3␤HSD2 with isolated mitochondria, the imported (P) fraction was separated from non-imported fraction (S) by washing. In some cases fractions were extracted with Na 2 CO 3 . Fractions were briefly treated with (middle panel) or without protein kinase (PK, top panel). After electrophoresis, gels were fixed and dried. Gels were also processed for Western blotting using an antibody directed against COX IV (bottom panel). D and E, quantitation of bands was done using phosphorimaging. Data are presented as the mean Ϯ S.E. of three independent experiments. Although not an integral protein, 3␤HSD2 specifically associates with IMM; WASH, washing; CARB, carbonate. F-H, shown are Western blots of MA-10 mitochondrial compartments using antibodies directed against Tim50 (F), Tim23 and 3␤HSD2 (G), and VDAC2 and 3␤HSD2 (H). NOVEMBER 11, 2011 • VOLUME 286 • NUMBER 45 included antibodies directed against proteins known to localize in the IMM or OMM. We found that the IMM fraction contained both Tim50 and Tim23 (Fig. 1, F and G). To confirm that we successfully separated the OMM from the IMM, we probed fractions for the presence of voltage dependent anion channel (VDAC2), an OMM protein, and saw that VDAC2 primarily localized to the OMM (Fig. 1H). 3␤HSD2 was seen in the IMM fractions ( Fig. 1, G and H). Similarly, 3␤HSD2 also localized to the IMM in human breast cancer MCF-7 cells and human adrenal NCI H295 cells (data not shown).

Tim50 Regulates Steroidogenesis
3␤HSD2 Is Associated with Tim50 and Tom22-From the above studies we found that 3␤HSD2 specifically associated with the IMM but was not integrated into the membrane (Fig. 1, B-E). This association with the IMM led us to evaluate what proteins 3␤HSD2 may bind to in the mitochondria. We did import assays using cell-free-synthesized 35 S-labeled 3␤HSD2 and mitochondria isolated from steroidogenic MA-10 cells. Samples were treated with proteinase K to proteolyze non-specifically associated proteins. Digitonin lysis of the imported 35 Slabeled 3␤HSD2 and subsequent native gradient gel electrophoresis resulted in the appearance of three 3␤HSD2containing complexes, 750, 490, and 390 kDa in size ( Fig. 2A). The two smaller complexes may represent a breakdown from the larger complex, but resistance to proteinase K suggests tightly associated protein complexes. LC-MS/MS analysis of the major 750-kDa complex revealed several mitochondrial proteins (supplemental Table ST1), including outer membrane translocase Tom22, inner membrane translocase Tim50, and 3␤HSD2. These results suggest that the outer and inner mitochondrial translocases may coordinate their functions through 3␤HSD2.
Tranlocases help ensure the proper localization of mitochondrial proteins. Most mitochondrial proteins, synthesized as precursor proteins in the cytosol (3,5,6,32), enter the mitochondria through Tom40, a translocase complex at the OMM (3,6). Proteins then reach their final destination within the mitochondria through the action of additional translocase complexes. For example, Tim23 at the IMM directs proteins with a presequence to either the IMM or matrix. Each translocase complex consists of multiple proteins that can transiently associate with proteins in other complexes (7,8). Tim50, a component of the Tim23 complex, binds to proteins emerging from the Tom40 channel and stimulates interaction between the presequence and Tom22 (9). Tim21, another component of Tim23 (11), competes with the presequence for binding to Tom22 and thereby causes the release of the protein and its subsequent translocation through the Tim23 channel.
To validate that 3␤HSD2 associates with translocases, we performed Western analysis of the native complexes using antibodies directed against 3␤HSD2, Tom22, Tim50, and Tim23 (Fig. 2, B-E). We included Tim23 antibodies, as others have shown that IMM resident Tim50 (supplemental Fig. S2e) asso-  35 S-labeled 3␤HSD2 was imported into isolated mitochondria, and the digitonin (Dig)-solubilized fraction was analyzed through native gradient gel electrophoresis. 3␤HSD2-containing complexes were identified at 750, 490, and 390 kDa. PK, protein kinase. B-E, shown are Western blots of the native-page complexes probed with antibodies directed against 3␤HSD2, Tom22, Tim50, and Tim23. F and G, sucrose density gradient analysis of the digitonin-lysed mitochondrial complex after 1 (F) and 4 (G) h of ultracentrifugation is shown. Each fraction was probed with the indicated antibodies, and the distribution of the proteins was graphed. Data presented are the mean Ϯ S.E. of three independent experiments. The graphs indicate that 3␤HSD2 was associated for a limited time of 1 h with all the translocases. H, shown are co-immunoprecipitation using isolated mitochondria from MA-10 cells and the indicated antibodies (ab) followed by Western blotting with the 3␤HSD2 antibody. 3␤HSD2 directly interacted with Tim23 and Tim50.

Tim50 Regulates Steroidogenesis
ciates with flexible Tim23 (10) to coordinate the function of mitochondrial protein by maintaining a permeability barrier (11)(12)(13). We indeed observed all four proteins in the 390-kDa complex, whereas the 490-kDa complex contained only Tim50 and Tom22, and the 750-kDa complex contained only 3␤HSD2 and Tom22 (Fig. 2, B-E). It is possible that Tim23 was not identified in the MS analysis because of the higher abundance of other proteins. The observation that Tom22 localized to all three complexes may be because its C terminus faces the IMS side of the IMM, whereas the majority of the protein is associated with the OMM. The minor association of Tom22 with the 750-kDa complex is possibly due to a loose association with 3␤HSD2 (Fig. 2C). The lack of 3␤HSD2 detection in the 490-kDa complex suggests an association of the Tim50 emerging at the trans side of the Tom channel and promoting binding of the presequence to the Tom22 IMS domain, resulting in a direct interaction between 3␤HSD2 and Tom22. Alternatively, the lack of 3␤HSD2 in the 490-kDa complex may indicate that it was buried within the native complex. As a result, the association between Tim50 and 3␤HSD2 was likely to be transient and, thus, flexible enough to coordinate interaction with Tom22. In summary, the different sizes and components of the native complexes support the notion that 3␤HSD2 participates in the fluid and transient associations that form between the different translocases (7,8).
These experiments clearly indicate that 3␤HSD2 and the three translocases form a network that connects the OMM and IMM. To further demonstrate that these proteins remain closely localized in the mitochondria, we purified digitoninlysed mitochondria through a sucrose density gradient and then examined the fractions by Western blotting (Fig. 2, F and  G). During a longer separation (4 h), Tim23 and Tom22 appeared closely associated as they co-localized in fractions 8 -12, whereas Tim50 and 3␤HSD2 were more distantly separated (Fig. 2G), suggesting a weak interaction between 3␤HSD2 and the translocases. However, during a shorter separation (1 h), 3␤HSD2 showed an overlapping distribution with Tim23, Tim50, and Tom22 (Fig. 2F), suggesting the interaction of all four proteins. The interactions may not have been of equal strength, however, as the distribution of 3␤HSD2 more closely mirrored that of Tim50 than Tom22 (Fig. 2F). Indeed, co-immunoprecipitation with the translocase antibodies followed by Western blotting for 3␤HSD2 demonstrated that 3␤HSD2 interacted strongly with Tim50, weakly with Tim23, but not with Tom22 (Fig. 2H). Taken together, it seems that 3␤HSD2 and Tim50 specifically interact, and through this interaction 3␤HSD2 associates with the other translocases transiently.
Tim50 regulates progesterone and androstenedione synthesis-We next evaluated if the interaction between Tim50 and 3␤HSD2 had a biological effect by knocking down Tim50 (⌬Tim50) in steroidogenic MA-10 cells and measuring the metabolic conversion of pregnenolone to progesterone. siRNA reduced 3␤HSD2 expression by 90% (Fig. 3A, supplemental Fig.  S2a) and mitochondria from these cells produced 65 ng/ml progesterone as compared with 370 ng/ml from control cells (Fig.  3B, supplemental Fig. S2b). We also performed rescue experiments by transfecting ⌬Tim50 cells with expression vectors for either wild-type or mutant Tim50. The IMS domain of Tim50 lies between amino acids 133-476 and is indispensible for Tim50 interaction with Tim23; thus, we generated two Tim50 mutants in this domain, L279S and L288S. The endo group reported that these mutants had reduced interactions with Tim23 (11). We also generated the mutant L134S to evaluate the effect of a conformational change at the beginning of the IMS domain. Wild-type and mutant Tim50 vectors expressed approximately the same level of Tim50 protein (Fig. 3C). However, only wild-type Tim50 and L134S rescued progesterone synthesis (Fig. 3D, supplemental Fig. S2c). Studies have shown that point mutations in Tim50 reduce activity that is likely related to the degree of conformation change of the protein (11). Mutation in steroidogenic proteins can be lethal for human survival (17,33). Thus, Tim50 mutants may be misfolded and unable to interact with 3␤HSD2, resulting in the loss of 3␤HSD2 activity.
Two current models (11,12) suggest that the Tom22 C terminus faces Tim50, so Tom22 may interact transiently with 3␤HSD2. Moreover, 3␤HSD2 charge segments may provide binding sites for presequence-carrying preproteins traversing the OMM and IMM (34). Thus, we also evaluated the effect of Tom22 knockdown. In ⌬Tom22 cells, 3␤HSD2 expression was reduced 80% as compared with wild-type cells (supplemental Fig. S3b). Progesterone synthesis was also reduced, from 503 to 190 ng/ml (Fig. 4B), indicating an indispensible role for Tom22 in 3␤HSD2 activity. Perhaps the 62% reduction in progesterone synthesis resulted from the loss of the weak interaction between 3␤HSD2 and Tom22 because Tom22 binds preproteins through both its cytosolic domain and intermembrane space domain, which is stably associated with the channel protein Tom40 (35) that has a significant C-terminal domain at the IMS (36,37).
3␤HSD2 also converts DHEA to androstenedione; thus, we explored the effect of translocase knockdown on this activity. Rodents do not convert DHEA directly to androstenedione, so for this experiment we used human adrenal NCI H295 cells and found that knockdown of Tom22, Tim23, and Tim50 resulted in a similar level of androstenedione synthesis as treatment of cells with the 3␤HSD2 inhibitor, trilostane (Fig. 4C). Again, in ⌬Tim50 cells, wild-type Tim50 restored androstenedione synthesis (Fig. 4D); however, none of the Tim50 mutants, including L134S, restored activity (Fig. 4D, supplemental Fig. S3c).
Tim50 Coordinates in Specific Steps of Steroidogenesis-We next evaluated the expression of P450c17 in the knockdown cells. This enzyme is responsible for converting pregnenolone and progesterone to their 17-hydroxy forms (Fig. 1A) and then converting these to DHEA and androstenedione. P450c17 plays a crucial role in DHEA expression as the 17,20 lyase activity is about 50 times more efficient for the conversion of 17␣-hydroxypregnenolone to DHEA than for the conversion of 17␣hydroxyprogesterone to androstenedione (Fig. 4E, middle  panel). Fig. 4E shows that the expression level of P450c17 (middle panel) remained the same between control and knockdown cells. However, DHEA showed a marked decrease in ⌬Tim50 cells (top panel), whereas 3␤HSD2 expression was nearly ablated in all the knockdown cells (bottom panel). These results suggest that defects in DHEA conversion were due solely to 3␤HSD2, although the role of Tim50 in DHEA synthesis requires further investigation.
The possibility exists that the decrease in 3␤HSD2 expression seen in cells with ⌬Tim23 and ⌬Tom22 knockdown was due to the misfolding and subsequent proteolysis by IMS resident proteases of unused 3␤HSD2. To test this possibility, we incubated ⌬Tim50 cells with increasing concentrations of protease and proteasomal inhibitors and determined 3␤HSD2 expression (supplemental Fig. S3d). The expression of 3␤HSD2 was restored with the cysteine protease inhibitor, E64, and proteasomal inhibitor lactacystin (Fig. 5A). However, restoration of 3␤HSD2 expression did not rescue the conversion of pregnenolone to progesterone (Fig. 5B) or DHEA to androstenedione (Fig. 5C), confirming that the presence of Tim50 is mandatory for 3␤HSD2 activity.
Flexibility of 3␤HSD2 Helps in the IMM Network-Given the multiple interactions of 3␤HSD2 with the different translocases, the protein is likely flexible. To address this possibility, we performed protein fingerprinting of baculovirus-expressed 3␤HSD protein. We incubated the protein with trypsin for 15 or 45 min at two temperatures to identify domains that are differentially protected from proteolysis. For flexible proteins, lysine and arginine residues will be exposed to trypsin, generating proteolysis to smaller fragments. Trypsin treatment of 3␤HSD for 45 min at 4°C resulted in one protected band (Fig.  ⌬Tom22 (B). As compared with control cells, ⌬Tom22 mitochondria produced 63% less progesterone whereas⌬Tim23 mitochondria produced 60% less. C, shown is metabolic conversion of DHEA to androstenedione after knockdown of Tim50, Tim23, and Tom22 in human adrenal NCI H295 cells. DHEA to androstenedione was significantly reduced in ⌬Tim50 cells as compared with ⌬Tim23 and ⌬Tom22 cells. D, shown is measurement of DHEA to androstenedione in ⌬Tim50 cells transfected with wild-type or mutant Tim50 cDNA. Tim50 knockdown ablated the conversion of DHEA to androstenedione, and only expression of wild-type Tim50 could rescue the enzymatic activity. Plasmid Bluescript KS is an internal negative control. E, shown is expression of DHEA, P450c17, and 3␤HSD2 in control NCI H295, ⌬Tim50, ⌬Tim23, and ⌬Tom22 cells as determined by Western blot. DHEA expression was decreased in ⌬Tim50 cells without affecting its precursor, P450c17.

Tim50 Regulates Steroidogenesis
6A). LC-MS/MS analysis of this band identified sequence matching the N-terminal 18 -362 amino acids of the wild-type protein (supplemental Tables ST2 and ST3), indicating no protection of bands in 3␤HSD2 (Fig. 6A, right-hand panel  schematic).
We next used CD spectroscopy to further evaluate the flexibility of 3␤HSD2. This technique can determine conformational changes with minor variations and distinguish secondary structural components with a minima at 198 -204 nm indicating random coils, at 208 nm and 222 nm indicating ␣-helices, and 218 nm indicating ␤-sheets (38,39). Computational analysis suggests 3␤HSD2 is a mixture of ␣ and ␤: 32% ␣-helix, 18% ␤-sheet, and 50% turn (Fig. 6B). In the presence of the low CMC detergent, Cymal-5, 3␤HSD2 exhibited a ␣/␤ conformation that changed to a more flexible ␤-turn (Fig. 6B, black square) in the presence of the high CMC detergent, Igepal CO-720. Smaller proteins or shorter peptides with more ␤-turn have higher substrate activity as compared with proteins with less or no ␤-turn (40). Although we observed a direct relationship between 3␤HSD2 activity and protein concentration, we also noted that the addition of detergent with low CMC increased activity (supplemental Fig. S3g). The activity of 3␤HSD2 was 5-fold higher with Cymal-5, (detergent with lower CMC) (Figs. 6C and supplemental Fig. S3g) than with Igepal-CO-720 (detergent with higher CMC). This experiment strongly suggests that the lower CMC detergent created conditions similar to those in the inner mitochondrial space, and thus 3␤HSD2 reaches the appropriate conformation while associated with the translocases, supporting the likelihood that 3␤HSD2 is a flexible protein.
N-terminal Amino Acid Sequence Is Responsible for Interaction with Translocases-To understand the mechanism of interaction between 3␤HSD2 and the amino acids connecting Tim50 and Tom22, we designed experiments to determine those specific regions of 3␤HSD2 responsible for interaction with the translocases using native gradient gel electrophoresis. We made deletion constructs from the N terminus and the C terminus by deleting amino acids in increments of three at a time but keeping the first N-terminal methionine. Import of the cell-free-synthesized deletion mutants into isolated mitochondria followed by digitonin lysis and electrophoresis in native gradient PAGE showed that 3␤HSD2 lacking the first three amino acids (from amino acids 2-5 generating ⌬2-5 3␤HSD2) did not form a 750-kDa complex (Fig. 7A), suggesting that the first three amino acids are essential for import into mitochondria. However, mitochondrial import, fractionation, and FIGURE 5. Presence of Tim50 is essential for steroidogenic activity. A, a Western blot of ⌬Tim50 MA-10 cells shows restored 3␤HSD2 expression to a greater extent with E64 than lactacystin, MG132, and 4-(2-aminoethyl)benzenesulfonyl fluoride (AEBSF). B and C, measurement of pregnenolone to progesterone and DHEA to androstenedione conversion using mitochondria from ⌬Tim50 cells that were treated with the indicated inhibitors is shown. Although lactacystin and E64 restored 3␤HSD2 expression, the inhibitors did not restore 3␤HSD2 activity in ⌬Tim50 cells. Trilo, trilostane.  NOVEMBER 11, 2011 • VOLUME 286 • NUMBER 45 extraction with sodium carbonate of N⌬2-5 3␤HSD2 demonstrated that the deletion mutant (Fig. 7B, bottom panel) was imported with similar efficiency to wild-type 3␤HSD2 (Fig. 7B,  top panel). The fact that N⌬2-5 3␤HSD2 was imported into the mitochondria but did not form a complex implies that import and translocase interaction are not the same. We next determined the minimum number of amino acids required for the generation of the 3␤HSD2 complex by generating mutants with internal deletions of 15 amino acids (Fig. 7B). We found that the complex was restored to that of the wild type only when the mutant retained the first 15 amino acids (N⌬15-30). Coimmunoprecipitation of the N⌬15-30 3␤HSD2 complex with the 3␤HSD2 antibody did not show any interaction with the Tim50 or Tim23 translocases. To determine the biological activity of the different deletion mutants, we cotransfected COS-1 cells with 3␤HSD2 and F2 (fusion of cytochrome P450scc-ferrodoxin-ferrodoxin reductase). As nonsteroidogenic COS-1 cells do not have any P450scc or ferrodoxin and ferrodoxin reductase, we provided these as a single fused protein, F2 (41,42). Metabolic conversion of the transfected cells showed full activity with the wild-type 3␤HSD2 but no activity with any of the deletion mutants, suggesting the N terminus association with the translocases is crucial for 3␤HSD2 activity.

Tim50 Regulates Steroidogenesis
To investigate the role of the C terminus for interaction with Tim50, we deleted 3, 10, 20, 30, and 40 amino acids from the C terminus generating C-3, C-10, C-20, C-30, and C-40 3␤HSD2 mutants. Import of the C-20 and C-40 mutants into isolated mitochondria, and isolation of the complex through native gradient PAGE after digitonin lysis showed that the complex containing C-20 3␤HSD2 was similar in size to that of wild-type 3␤HSD2. The complex was absent with the C-40 3␤HSD2 mutant, suggesting that the C terminus plays a lesser role in the interaction with the associated proteins at the IMM. However, import of the C-20 and C-40 into mitochondria and extraction with sodium carbonate showed that both C-terminal deletion mutants were imported in a fashion similar to wild-type 3␤HSD2 protein (Fig. 7D). Based on these results, it appears that N terminus and C terminus contribute differently to the interaction with IMM resident proteins.

DISCUSSION
For decades researchers have sought to identify the mechanisms of action of 3␤HSD2 as well as its receptor. In this study we show that 3␤HSD2 interacted with the IMM translocase, Tim50, from the C-terminal to the IMS side of the mitochondria and that this interaction helped to mediate 3␤HSD2 metabolic activity. In the yeast model, Tim50 interacts with Tim23, facilitating the efficient transfer of proteins from the Tom40 complex to the Tim23 complex and facilitating the transport of proteins across the IMM by promoting the motor function of Hsp70 (11). Point mutations in the IMS domain of Tim50 reduce these activities, likely related to the degree of conformation change of the protein and the subsequent inability to interact with Tim23 (11,43). The observed defects in motor functions are indirect consequence of growth defects (11), and this is due to the absence of steroids required for sexual development. Sugiyama et al. (44) found that Drosophila with mutations in the yeast Tim50 ortholog exhibit reduced growth and mitochondrial membrane potential. Similarly, in the human cell line 293T, Tim50 and Tim23 interact, and knockdown of Tim50 results in accelerated cytochrome c release and subsequent apoptosis. Developmental defects stemming from increased cell death are also seen in zebrafish with Tim50 knockdown (14). Given the interaction of Tim50 and Tim23 in the IMS, it was not surprisingly that 3␤HSD2 also formed a transient association with Tim23. We also found an association with Tom22, a component of the Tom40 complex. Knockdown of Tim50 expression reduced the expression of 3␤HSD2 by 80% as well as decreased 3␤HSD2 activity. Even with the restoration of 3␤HSD2 expression, 3␤HSD2 remained inactive in Tim50 knockdown cells. Expression of an unrelated protein under identical conditions did not restore steroidogenic activity, sug- FIGURE 7. 3␤HSD2 acts at the IMM directionally. A, different N-terminal 35 S-labeled 3␤HSD2 deletion mutants, as indicated, were imported into isolated mitochondria (Mito), and the digitonin-solubilized fraction was analyzed through native gradient gel electrophoresis. N-terminal 15 amino acids are essential for interaction with the membrane. B, import and fractionation of wild-type (top panel) and N⌬2-5 35 S-labeled 3␤HSD2 (bottom panel) after import into the isolated mitochondria are shown. After incubation of 35 S-labeled 3␤HSD2 with isolated mitochondria, the imported (P) fraction was separated from nonimported fraction (S) by washing (WASH) and also by extraction with Na 2 CO 3 (CARB). The result shows most of the fraction was remained in the supernatant (S) than in pellet (P) fraction after extraction with sodium carbonate. Thus, both the wild-type and mutant was processed into the mitochondria in a similar fashion. C, the indicated C-terminal 35 S-labeled 3␤HSD2 mutants were imported into isolated mitochondria, and the digitonin-solubilized fraction was analyzed through native gradient gel electrophoresis in a similar fashion as in panel A. Association with the mitochondrial membrane was ablated when more than 20 amino acids were deleted. D, import and fractionation of C-20 (top panel) and C-40 35 S-labeled 3␤HSD2 (bottom panel) after import into the isolated mitochondria is shown. After incubation with isolated mitochondria, the imported (P) fraction was separated from non-imported fraction (S) by washing and also by extraction with Na 2 CO 3 . The result shows C-20 and C-40 are processed in the same fashion of wild-type 35  gesting that the orientation of 3␤HSD2 with Tim50 is essential. These results suggest that the presence of Tim50 is mandatory for 3␤HSD2 docking at the IMS and that this promotes 3␤HSD2 function. Patients with 3␤HSD2 mutations had similar phenotypic expression, supporting a role for Tim50 in 3␤HSD2 function. Introduction of wild-type Tim50 into the Tim50 knockdown cells restored 3␤HSD2 activity; however, introduction of Tim50 mutants (11) that could not interact with Tim23 did not restore activity. Moreover, knockdown of either Tim23 or Tom22 reduced 3␤HSD2 metabolic activity, supporting the notion that complex formation contributes to regulation of 3␤HSD2 activity.
Tim50 also affected DHEA expression. P450c17 converts pregnenolone and progesterone to their 17 hydroxy forms and then converts those metabolites to DHEA and androstenedione. The 17,20-lyase activity of P450c17 is ϳ50 times more efficient for conversion of 17 OH pregnenolone to DHEA than for conversion of 17 OH progesterone to androstenedione; thus, P450c17 plays a crucial role in DHEA expression. Knockdown of translocases did not alter expression of P450c17, but knockdown of Tim50 reduced expression of DHEA (Fig. 4E). Thus, the reduced conversion of DHEA to androstenedione seen in Tim50 knockdown cells likely resulted from the reduced expression of 3␤HSD2. Our results confirm, 3␤HSD2 alone cannot catalyze the conversion of 17-hydroxy progesterone to DHEA. The absence of Tim23 or Tom22 did not inhibit DHEA expression, but minimally affected conversion, possibly due to the absence of 3␤HSD2. In summary, DHEA expression requires the presence of Tim50 for the catalytic reaction. However, at this time we cannot predict how these proteins interact.
The Tom complex in the OMM and the Tim50 complex in the IMM are not permanently linked and can interact only in the presence of a translocating precursor protein (45)(46)(47). Therefore, a precursor pre-protein such as 3␤HSD2, translocating through the Tom complex, must engage the Tim50 complex in the boundary inner membrane through the transient contact sites (48). The IMS domains of Tim23-Tim50 form a minimal binding interface required for the interaction of these two proteins. However, it is questionable whether Tim23 and Tim50 interact solely through their soluble domains given the lower K d for the interaction (49). The transmembrane domain of Tim50 may stabilize flexible Tim23, resulting in a stable Tim23-Tim50 interaction (50). It is intriguing that the isolated IMS domains of Tim23 and Tim50 interact with each other despite the fact Tim23 is largely unfolded (49). The flexible 3␤HSD2 was active as the protein retained similar binding capacity with its ligand trilostane (51). The flexibility of Tim23 seems essential, as it allows for docking Tim50 for stabilization and can thus connect transiently with Tom22. Indeed, proteolysis with trypsin confirms that 3␤HSD2 has no protease-resistant domain. However, the interaction between the translocases and the 3␤HSD2 is tightly regulated. This is possibly a mandatory condition for 3␤HSD2 to adopt a conformation needed for interaction in a proper orientation with the translocases.
Mitochondrial-targeted protein import generally starts at the N terminus. After import, the N-terminal sequence, which contributes to sorting the protein to a specific location, may be cleaved. In this study 3␤HSD2 is imported starting from the N terminus, but its N-terminal sequence is not cleaved, which also occurs for the IMM resident small Tim proteins (3). Deletion of three amino acids (N⌬2-5 3␤HSD2) from the N terminus did not ablate the import but did inhibit its association with the mitochondrial translocases. To associate with the IMM resident proteins, 3␤HSD2 required its first 15 amino acids (⌬15-30 3␤HSD2) at the N terminus. However, this complex was unable to metabolize pregnenolone to progesterone. This might be the sole reason that patients with a change in the N-terminal amino acids of 3␤HSD2 develop congenital adrenal hyperplasia (17). We believe that deletion of the small section of the 3␤HSD2, especially from the N terminus, ablated activity due to the absence of interaction with the Tim50. Alternatively, the orientation of 3␤HSD2 with Tim50 and Tom22 may be inhibited. The C terminus deletion of 10 amino acids had no major effect. Mutations of more than 30 (C-30 3␤HSD2) amino acids only show severe phenotypic expression of sexual differentiation (21). The possibility exists that the C-terminal hydrophobic amino acids of 3␤HSD2 remain within the proteins core (52) and do not participate in the interaction with the mitochondrial translocases. In summary, 3␤HSD2 associates with the inner membrane translocases in a fashion that requires tight regulation, and any change in the expression of the inner mitochondrial translocase, Tim50, ablates steroidogenesis.