Mitochondrial Functions and Estrogen Receptor-dependent Nuclear Translocation of Pleiotropic Human Prohibitin 2*

Proteins with multiple cellular functions provide biological diversity to eukaryotic cells. In the current studies, we identified the mitochondrial functions of human prohibitin 2 (PHB2), which was initially identified as a repressor of estrogen-dependent transcriptional activity. The mitochondrial complex of PHB2 consists of PHB1, voltage-dependent anion channel 2, adenine nucleotide translocator 2, and the anti-apoptotic Hax-1, which is a novel binding partner for PHB2. RNA interference-mediated knockdown of PHB2 in HeLa cells resulted in caspase-dependent apoptosis through down-regulation of Hax-1 and fragmentation of mitochondria. We also found that, although PHB2 is predominantly expressed in the mitochondria of HeLa cells, it translocates to nucleus in the presence of estrogen receptor α and estradiol. Here, we first demonstrated the roles of mammalian PHB2 in mitochondria and the molecular mechanism of its nuclear targeting and showed that PHB2 is a possible molecule directly coupling nuclear-mitochondrial interaction.

In eukaryotes, biological diversity is acquired from a limited number of genes due to the existence of proteins with multiple cellular functions. A key factor is the dynamic regulation of these diverse cellular functions. One such multifunctional protein, prohibitin (PHB) 2 1, was originally identified in mammals as a putative negative regulator of cell proliferation (1). PHB1 and PHB2 are closely related proteins and are highly conserved among yeast (2), plants (3), worms (4), flies (5), and mammals (6). Subcellular localization of PHBs has been confined to mitochondria in a variety of these species, although they also localize in nucleus in some mammalian cell lines (7,8). To date, it is known that PHBs are localized in the mitochondrial inner membrane where they form a large protein complex (9). In addition, variety functions of PHBs have been suggested, including a role in cell cycle regulation (1,10,11), transmembrane signal transduction (12,13), and control of life span (14).
In yeast, PHBs have been shown to exist in mitochondria and to function as chaperones that stabilize newly synthesized mitochondrial protein, possibly by negative regulation of the AAA protease (15,16). Even in Caenorhabditis elegans, PHBs are essential for embryonic viability and germ line differentiation, and deficiency of these proteins results in altered mitochondrial biogenesis (17).
Alternatively, in mammals, knowledge of human PHBs focused on their transcriptional regulatory functions has been accumulated. Human PHB1 has also been shown to interact with the retinoblastoma protein, which results in the inhibition of the transcriptional activity of E2F (10). In a B-cell lymphoma line, stable overexpression of PHB1 protects the cells from camptothecin-induced apoptosis, possibly via down-regulation of E2F activity (18). Furthermore, human PHB2, also known as a repressor of estrogen receptor (ER) activity, has been shown to interact with and inhibit the transcriptional activity of the ER (13). These findings suggest the involvement of mammalian PHB proteins in the regulation of transcription in the nucleus.
On the contrary, in fibroblasts, mammalian PHB proteins mainly localize in mitochondria, and their expression is up-regulated by mitochondrial stress and down-regulated during cellular senescence (19). Therefore, it is thought that mammalian PHB proteins also have crucial roles in the mitochondria. Recently, human PHB1 was reported to associate with mitochondrial complex I, suggesting that PHB1 plays a role in complex assembly (20). Thus, mammalian PHBs have been suggested to be involved not only in the regulation of transcription but also cellular senescence, apoptosis, and mitochondrial respiratory activity (9,18). The molecular functions of PHBs in mitochondria, especially PHB2, however, remain unclear.
In this study, to clarify the dynamic regulatory mechanism of the pleiotropic PHB2 in mammalian cells, we examined mitochondrial function of human PHB2 and then investigated its targeting mechanism to the nucleus. First, we identified the binding partners for PHB2 in the mitochondria by immunoprecipitation and mass spectrometric analysis. Hax-1, which was initially identified as an HS1-binding protein (21) and as an inhibitor of apoptosis (22), was found to directly associate with PHB2 in mitochondria. RNA interference (RNAi)-mediated knockdown of PHB2 resulted in a reduction of the level of Hax-1 protein. In addition, caspase-mediated apoptotic cell death was enhanced in PHB2 knockdown cells. This induction of cell death was likely due to the down-regulation of Hax-1, because its knockdown also caused the same manner of cell death without a reduction of PHB2 protein level. Furthermore, the knockdown of PHB2 caused the fragmentation of mitochondria by a mechanism independent of Hax-1.
Secondly, we found that, although PHB2 is predominantly expressed in the mitochondria of HeLa cells, it translocates to the nucleus in the presence of ER␣ and estradiol (E2). We further found that human PHB2 contains both an uncleavable mitochondrial targeting sequence (MTS) at its N terminus and an ER␣-dependent nuclear localization sequence at its C terminus, suggesting that it is shuttled from the nucleus to mitochondria.
Taken together, our results show that human PHB2 has pleiotropic functions in the mitochondria, including inhibition of apoptosis via the PHB2⅐Hax-1 complex and regulation of the mitochondrial morphology. We also demonstrated that, in the presence of ER␣ and E2, PHB2 is translocated into the nucleus where it functions as a repressor of transcription. PHB2 is a versatile molecule that couples transcription in the nucleus and regulation of mitochondrial function in the mitochondria, suggesting that it plays a role in communication between these two organelles.
Cell Culture and Transfection-HeLa and MCF7 cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, 100 units/ml penicillin, and 100 g/ml streptomycin at 37°C in an atmosphere containing 5% CO 2 . Transfection was performed using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. Mitochondria were stained with rhodamine 123 (0.25 g/ml) or MitoTracker Red CM-H 2 XRos (250 nM, Molecular Probes, Eugene, OR) for 30 min at 37°C. Cells expressing pSilencer 3.1-H1 Puro constructs were selected with 2 g/ml puromycin (Sigma). Caspase activity was inhibited with 50 M Z-VAD-FMK (Calbiochem, San Diego, CA). Cells expressing the ERexpression plasmid were cultured in Opti-MEM I without phenol red (Invitrogen) supplemented with 10 nM ICI 182,780 (Tocris), and treated with 1 M of E2 for 2 h. For inhibition of cytoplasmic translation, the cells were simultaneously cultured with 50 or 100 g/ml cycloheximide (CHX).
Immunocytochemistry-HeLa cells were plated on 35-mm poly-L-lysine-coated glass-bottomed dishes (Matsunami Glass Ind.) and fixed for 20 min at room temperature with 4% paraformaldehyde and 0.4% Triton X-100 in PBS. The cells were incubated with antibodies against FLAG (Sigma, rabbit polyclonal), cytochrome c (BD Pharmingen, mouse monoclonal), ER␣ (Upstate Biotechnology, rabbit polyclonal), or Myc (BD Biosciences, mouse monoclonal) in PBST (PBS with 0.05% Tween 20) containing 2% horse serum. After washing three times with PBS, the cells were incubated with Alexa 488-conjugated anti-rabbit IgG (Molecular Probes) and Cy3-conjugated anti-mouse IgG (Amersham Biosciences) in PBST containing 2% horse serum for 1 h at room temperature. Fluorescent images were captured and analyzed with a Radiance TM Laser Scanning Confocal Microscope System (Bio-Rad).
Preparation and Fractionation of Mitochondria-Mitochondria were prepared from HeLa cells as previously described (24). To assess membrane association, mitochondria suspended in sucrose solution (0.25 M sucrose supplemented with 0.2 mM EDTA) were sonicated on ice. Intact mitochondria were removed by centrifugation at 4°C for 10 min at 10,000 ϫ g, and the supernatant containing sonicated mitochondria was further centrifuged at 4°C for 30 min at 100,000 ϫ g. The pellets were collected as the mitochondrial membrane fraction. The mitochondrial membrane fraction was then treated for 30 min on ice with 0.1 M Na 2 CO 3 in sucrose solution. The solution was centrifuged at 4°C for 30 min at 100,000 ϫ g to separate the soluble proteins from the membranes.
For the protease protection assay, mitochondria were treated for 20 min at room temperature with 0.25 mg/ml trypsin (Sigma) in sucrose solution containing the indicated concentrations of digitonin (Sigma) or 1% Triton X-100. The reaction was stopped by adding trichloroacetic acid.
The submitochondrial fraction was prepared as follows: sonicated mitochondria in 0.45 M sucrose were layered onto a linear sucrose gradient (11 ml, 0.6 -1.6 M sucrose in 10 mM HEPES-KOH, pH 7.4, and 0.2 mM EDTA) and centrifuged at 4°C for 16 h at 100,000 ϫ g. The gradient was collected in 25 0.5-ml fractions and then analyzed by Western blotting.
Cell Counting-The living cell number in a 6-well plate was determined by trypan blue staining. Cells were trypsinized and collected by centrifugation. The cell pellets were suspended in 0.4% trypan blue solution (Invitrogen), and living cells were counted.
Mass Spectrometry-Protein bands on acrylamide gels were stained with Coomassie Brilliant Blue R-250 and cut out. In-gel digestion with trypsin was carried out as described by Shevchenko et al. (25). Proteins were identified using a Finnigan LTQ liquid chromatography-tandem mass spectrometry system (Thermo Electron Corp.).
Western Blotting-Samples were separated by electrophoresis on SDS-polyacrylamide gels (10% or 12% acrylamide) and then electrophoretically transferred to nitrocellulose membranes (Hybond ECL, Amersham Biosciences). The membranes were probed with primary and horseradish peroxidaseconjugated secondary antibodies, and immunoreactive bands were visualized with enhanced chemiluminescence reagents  (24)), and anti-Mfn1/ hFzo1 (1:100, rabbit anti-sera against the peptide fragment, CVQLENELENFTKQFLPSSNEES, corresponding to the Cterminal portion of the protein).
Immunoprecipitation-The mitochondrial pellet from cells expressing PHB2-FLAG or Hax-1-FLAG was extracted with radioimmune precipitation assay buffer (20 mM Tris-HCl, pH 8, 150 mM NaCl, 1% sodium deoxycholate, and 1% Triton X-100). After sonication for 4 min on ice, the solution was centrifuged at 4°C for 15 min at 10,000 ϫ g. Immunoprecipitation was carried out by incubation of the supernatant with 5 g of anti-FLAG antibody (Sigma, mouse monoclonal) and protein G-Sepharose (Amersham Biosciences) at 4°C for overnight.
RT-PCR-Total RNA was isolated from HeLa cells using TRIzol (Invitrogen) according to the manufacturer's instructions. Two micrograms of the total RNA was subjected to RT-PCR (SuperScript II, Invitrogen) using random hexamer primers for the RT reaction.

RESULTS
Mammalian PHB Proteins Mainly Localize in the Mitochondria of HeLa Cells-In fibroblasts, mammalian PHB proteins are found mainly in the mitochondria (14), whereas they are also localized in the nucleus in MCF7 cells (7,8). We expressed human PHB proteins with a FLAG tag at their C termini in HeLa cells and found that they are mainly present in the mitochondria where they colocalize with mitochondrial cytochrome c (Fig. 1A, left panel). When we expressed GFP fusion proteins of PHBs in MCF-7 or HeLa cells, like the FLAG-tagged proteins, PHB1-GFP and PHB2-GFP were expressed in the mitochondria of HeLa cells (Fig. 1A, right panel) along with cytochrome c (data not shown). We found that a portion of PHB2-GFP was also present in the nucleus of MCF7 cells, whereas PHB1-GFP was only present in the mitochondria in these cells (Fig. 1A, right panel). These results demonstrate that, in HeLa cells, PHB proteins mainly localize in the mitochondria.
Association of PHB2 with Hax-1 in Mitochondria-To clarify the function of human PHB2 in mitochondria, we next performed immunoprecipitation studies using a mitochondrial extract from HeLa cells expressing PHB2-FLAG. Immunopre-cipitation with an anti-FLAG antibody (FLAG-IP) revealed that PHB2-FLAG specifically and reproducibly coprecipitates four proteins with molecular masses of 30 -37 kDa (Fig. 1B), which were not detected in untransfected control extract (data not shown). Mass spectrometric analysis identified these proteins as PHB1, HS-associated protein X-1 (Hax-1) (21), voltage-dependent anion channel 2 (VDAC2), and adenine nucleotide translocator 2 (ANT2), respectively (Fig. 1B). We expected that PHB1 would coprecipitate with PHB2, because they form stable complex in mitochondria (9). ANT and VDAC are components of the permeability transition pore (PTP), and the association of PHB2 with ANT or VDAC has been reported as a possible mitochondrial nucleoid complex in Xenopus (26). Therefore, Hax-1 appears to be a novel binding partner of PHB2. The interaction between PHB2-FLAG and Hax-1 in the mitochondrial extract was confirmed by Western blotting with an anti-Hax-1 antibody following to FLAG-IP (Fig. 1C, upper panel). We further confirmed that Hax-1-FLAG coimmunoprecipitates with PHB2, PHB1, and VDAC (Fig. 1C, lower panel).
Hax-1 Is an Integral Protein of the Outer Mitochondrial Membrane-Hax-1, originally isolated as an HS-1 binding protein, is known to localize mainly in mitochondria (21). We confirmed that FLAG-tagged Hax-1 is mainly localized in the mitochondria of HeLa cells ( Fig. 2A).
To clarify the submitochondrial localization of Hax-1, we initially examined whether Hax-1 is a membrane-integrated protein. Following sonication of HeLa cell mitochondria, most of the Hax-1 protein remained in the mitochondrial membrane pellets as did the integral inner membrane protein PHB1 and the inner membrane-associated protein cytochrome c (Fig. 2B, lane 3). As expected, alkali treatment of the mitochondrial membrane pellets released cytochrome c, an inner membraneassociated protein, from the membrane pellets (Fig. 2B, lanes 4  and 5). In contrast, Hax-1 as well as PHB1 and PHB2 remained in alkali-washed membrane pellets, indicating that they are integral mitochondrial membrane proteins (Fig. 2B, lane 5).
We further examined whether Hax-1 is integrated in the outer or inner mitochondrial membrane. Submitochondrial particles were fractionated by sucrose density gradient centrifugation, and the outer membrane and inner membrane fractions were collected. The effectiveness of the separation of these two membrane types was confirmed by Western blotting with antibodies to the outer membrane protein VDAC and the inner membrane protein PHB1. We found that Hax-1 was present in the outer membrane fraction as along with VDAC (Fig.  2C), indicating that Hax-1 is integrated in the outer mitochondrial membrane.
Protease protection assay was performed to further define the submitochondrial localization of Hax-1. Hax-1 as well as the mitochondrial matrix protein pyruvate dehydrogenase E2 subunit and the intermembrane space protein cytochrome c or optic atrophy 1 (OPA1) remained in intact mitochondria after trypsin treatment (Fig. 2D, lane 2). Selective disruption of the outer membrane with digitonin decreased Hax-1, cytochrome c, and OPA1 levels following trypsin treatment but not effect the level of pyruvate dehydrogenase E2 (Fig. 2D, lane 4), dem-onstrating that, like cytochrome c and OPA1, Hax-1 localizes in the intermembrane space. PHB proteins have been reported to be mitochondrial inner membrane proteins, and it has been suggested that their C termini are exposed to the intermembrane space in yeast (15). However, PHB proteins are known to be resistant to protease treatment after disruption of the outer membrane (19,27), which may be due to their tight folding. Our experiments are consistent with these previous studies that PHB proteins are resistant to proteolytic degradation following disruption of the outer membrane.
Knockdown of PHB2 Induces Reduction of Hax-1 Protein and Apoptosis-To clarify the function of human PHB2 in mitochondria and the significance of its interaction with Hax-1, we performed RNAi-mediated knockdown studies. Small interfering RNA (siRNA) for PHB1 or PHB2 based on a short hairpin RNA expression vector containing a puromycin-resistant gene was expressed in HeLa cells. Semiquantitative RT-PCR showed that the expression levels of PHB1 and PHB2 mRNAs were specifically reduced by siRNAs for PHB1 (siPHB1) and PHB2 (siPHB2), respectively (Fig. 3A). Western blotting confirmed that siPHB1 and siPHB2 decreased the levels of PHB1 and PHB2 but did not affect the expression of other mitochondrial proteins (cytochrome c, VDAC, and pyruvate dehydrogenase E2 subunit) or cytosolic glyceraldehyde-3-phosphate dehydrogenase (GAPDH) (Fig. 3B). Although the siR-NAs were specific for each PHB protein at the mRNA level (Fig. 3A), the protein levels of PHB1 and PHB2 were reduced by siPHB2 and siPHB1, respectively. This is not surprising because the two PHB proteins are interdependent in yeast and C. elegans so that deletion of one decreases the protein level of the other (17,27).
We next examined the effect of PHB2 knockdown on the expression of Hax-1. We found that knockdown of PHB2 reduced the level of Hax-1 protein (Fig. 3B), although the siRNA did not change the level of Hax-1 mRNA (Fig. 3A). On the contrary, knockdown of Hax-1 specifically reduced Hax-1 expression with no effect on the expression of PHB proteins (Fig. 3C). Thus, it was indicated that PHB2 regulates Hax-1 expression at the protein level.
Hax-1 is known to inhibit apoptosis, and its expression is decreased during apoptosis possibly due to cleavage by the Omi/HtrA2 protease (22,28,29). We suspected that knockdown of PHB proteins would induce apoptosis in HeLa cells, because it down-regulated the level of the anti-apoptotic protein Hax-1. Indeed, 4 days after transfection, a subpopulation of PHB or Hax-1 knockdown cells was not stained with rhodamine 123 (Fig. 4A), which indicates decrease of mitochondrial membrane potential, and, 5 days after transfection, also showed release of cytochrome c from mitochondria (Fig. 4B).
In addition, 5 days after transfection, the cell number of PHBs or Hax-1 knockdown cells was reduced by the siRNAs compared with the empty siRNA vector (Fig. 4C). The reduction of cell number was likely to be caused by cell death, because it was inhibited by the caspase inhibitor Z-VAD-FMK (Fig. 4C), indicating that caspase-dependent apoptosis occurs in these cells. Therefore, it appears that knockdown of PHB2 induces apoptotic cell death through down-regulation of Hax-1 expression. Treatment with Z-VAD-FMK, however, did not prevent the down-regulation of Hax-1 (Fig. 4D), suggesting that Hax-1 functions in the apoptotic cascade upstream from caspase activation.
Knockdown of PHB Protein Causes the Fragmentation of Mitochondria-We next investigated the RNAi effect on the morphology of mitochondria by staining living PHB knockdown cells with MitoTracker Red. This analysis revealed that ϳ30% of the PHB knockdown cells contained fragmented mito- and separated into supernatants (sup.) and membrane-pellets (ppt.) by centrifugation at 100,000 ϫ g. The membrane pellets were further treated with Na 2 CO 3 (alkali) and separated into "sup." and "ppt." by centrifugation at 100,000 ϫ g. Samples were analyzed by Western blotting using antibodies (␣) against PHB1, PHB2, Hax-1, and cytochrome c (cyt. c). C, the outer membrane (OM) and the inner membrane (IM) fractions were prepared by sucrose density gradient centrifugation of submitochondrial particles. Proteins were analyzed by Western blotting with indicated antibodies. The Western blot analysis revealed that Hax-1 and the outer membrane protein VDAC are present in the outer membrane fraction. D, mitochondria were treated with various concentrations of trypsin, digitonin, and Triton X-100 and then analyzed by Western blotting with indicated antibodies. The protease protection assay showed that Hax-1 and cytochrome c are localized in the intermembrane space. chondria (Fig. 5, A and B), whereas cells expressing the empty siRNA vector or Hax-1 knockdown cells showed tubular mitochondria with almost no morphological changes. Therefore, it appears that PHB proteins are involved in the regulation of mitochondrial morphology independently of Hax-1 function. On the other hand, overexpression of each PHB protein did not activate mitochondrial fusion and show long tubular mitochondria ( Fig. 1A and data not shown). These results indicate that PHB proteins are not directly involved in the mitochondrial fusion machinery as OPA1 and Mfn/Fzo (30). To address the mechanism of involvement of PHB proteins in the mitochondrial morphology, we examined whether PHB proteins regulate the protein levels of such mitochondrial fusiogenic proteins. Protein blot analysis shows knockdown of PHB proteins induces significant reduction of OPA1 protein, but not of Mfn1/hFzo1, whereas that of Hax-1 induces no effect (Fig. 5C). Because the fragmented mitochondria were induced by knockdown of OPA1, it is likely that PHB proteins regulate the mitochondrial morphology by stabilizing OPA1 protein.
ER␣-dependent Nuclear Translocation of PHB2-In this study, we showed that PHB2 mainly localizes and functions in the mitochondria of HeLa cells. Human PHB2 is also known as a repressor of ER activity, however, it remains unclear how PHB2 is delivered to the nucleus. Because MCF7 expresses the PHB2 ligand, ER␣, but HeLa cells do not, we hypothesized that the nuclear distribution of PHB2 is driven by ER␣. To determine the validity of this hypothesis, we expressed PHB2-GFP along with ER␣ in HeLa cells and observed its subcellular distribution. In the absence of E2, PHB2-GFP mainly localized in the mitochondria (Fig. 6A, upper panel, ϪE2). Surprisingly, PHB2-GFP translocated into the nucleus in the presence of ER␣ and E2 (Fig. 6A, upper panel, ϩE2). This ER␣-dependent nuclear translocation was specific for PHB2, because the distri-  . c). Arrowheads indicate cells that release cytochrome c from mitochondria. FI, fluorescence image; TI, transmission image. Scale bars, 10 m. C, 3 days after transfection with indicated siR-NAs, 2 ϫ 10 5 cells were seeded on 6-well plates, and selected by puromycin. From 4 days after transfection, the cells were treated with (ϩ) or without (Ϫ) the caspase inhibitor Z-VAD-FMK, and the cell number was determined on the 5th day after transfection. D, cells were treated as described in C and examined by Western blotting with indicated antibodies. bution of PHB1-GFP was not changed by ER␣ and E2 (Fig. 6A,  lower panel).
In addition, the translocation of mitochondrial PHB2 into the nucleus still occurred in the presence of translation inhibitor cycloheximide (CHX) (Fig. 6B). Furthermore, a chase assay in the same cell directly indicates that the translocation of PHB2-GFP from the mitochondria to nucleus occurred in the presence of CHX (Fig. 6C). Therefore, it appears that PHB2 in mitochondria, not newly synthesized in cytoplasm, translocates into the nucleus.
The C-terminal Domain Is Sufficient for ER␣-dependent Nuclear Translocation of PHB2-In yeast and C. elegans, PHB proteins have been reported to possess a non-cleavable leader peptide at their N termini (17,31). Structural prediction also indicates that human PHB2 likely possesses an N-terminal MTS, but PHB1 does not (Fig. 7A). We found that the N-terminal domains of human PHB proteins (amino acids 1-50) acted as MTS when expressed as FLAG-tagged proteins (Fig. 7B).
To identify the domain responsible for ER␣-dependent nuclear targeting of PHB2, we examined the localization of a deletion mutant of PHB2 that lacks the C-terminal domain, which contains the previously reported ER␣-binding site (32). The deletion mutant (amino acids 1-100) was located in the mitochondria in a manner unaffected by the presence of ER␣ and E2, whereas the full-length PHB2 was translocated to nucleus in the presence of ER␣ and E2 (Fig. 7C). Therefore, the C-terminal domain is necessary for the ER␣-dependent nuclear targeting of PHB2.
We further examined whether the C-terminal domain of PHB2 is sufficient for the ER␣-dependent nuclear translocation. N-terminal deletion mutants of PHB proteins (PHB1-(45-272)-GFP and PHB2-(51-299)-GFP, respectively) were not expressed in the mitochondria (Fig. 7D), demonstrating that the N-terminal portions of PHB proteins are the sole determinant of mitochondrial targeting. Using the deletion mutants, we found that the C-terminal domain of PHB2 (amino acids 51-299), which uniformly distributes in the nucleus and cytoplasm, accumulates in the nucleus in the presence of ER␣ and E2 (Fig. 7D). On the other hand, the C-terminal domain of PHB1 (amino acids 45-272) did not specifically accumulate in the nucleus under these conditions. Therefore, the C-terminal domain of PHB2 appears to be sufficient for ER␣-dependent nuclear translocation in the presence of E2.
Thus, human PHB2 localizes in the mitochondria through MTS at its N termini and in the nucleus through its C-terminal domain, suggesting the shuttling of PHB2 between the two organelles. Supporting this idea, the MTS of human PHB2 seemed to be non-cleavable as like as that in yeast and C. elegans, because PHB2-FLAG expressed in HeLa cells and in vitro-translated PHB2-FLAG had the same molecular masses (Fig. 7E). The shuttling of PHB2 was further supported by a chase assay, in which PHB2-GFP accumulated in the nucleus was reduced after removal of E2 (Fig. 7F). In the cell, mitochondrial signal of PHB2-GFP around the nucleus seemed to be increased (Fig. 7F), convincing the translocation of PHB2 from the nucleus to mitochondria.

DISCUSSION
The molecular function of mammalian PHB proteins, especially PHB2, in mitochondria has not been well understood. In addition, although mammalian PHB proteins have been reported to interact with transcription factors and function as transcriptional repressors, how they are targeted to the nucleus has not been determined. In this study, we analyzed the mitochondrial PHB2 protein complex and identified the anti-apoptotic protein Hax-1 as a novel binding partner for PHB2. We also clarified the pleiotropic functions of PHB2 on mitochondria, including regulation of apoptosis by forming a complex with Hax-1 and maintenance of mitochondrial morphology. Furthermore, we identified the molecular mechanism by which PHB2 is targeted to the nucleus.

PHB2 Protein Complex in Mitochondria
Our immunoprecipitation analyses showed that PHB2 forms a complex with known mitochondrial proteins, PHB1, Hax-1, ANT2, and VDAC2 in mitochondria. An in vitro binding assay showed us that PHB2 directly interacts with Hax-1. Unexpectedly, we did not detect a direct interaction between PHB2 and PHB1, suggesting that PHB2 has a weaker physical interaction with PHB1 than with Hax-1. The effective association between PHB1 and PHB2 might require an additional cellular factor, because it has only been demonstrated by immunoprecipitation in whole cell lysates (14). It is unlikely that PHB1 interacted alter mitochondrial morphology in aged cells (34). Further, in C. elegans, knockdown of PHB proteins induces abnormal mitochondrial morphology (17). Therefore, the maintenance of mitochondrial morphology by PHB proteins might be conserved from yeast to mammals.
Mitochondrial fragmentation is associated with a deficiency in the mitochondrial fusion activity. Considering mitochondrial function of PHB proteins as a stabilizer of proteins, it is possible that they maintain the expression of proteins involved in mitochondrial morphology. OPA1 might be one of such candidates, because it localizes in the intermembrane space and regulates mitochondrial fusion (35). In fact, the protein level of OPA1 is significantly reduced by knockdown of PHB proteins but not by that of Hax-1 protein. Because knockdown of Hax-1 induces apoptotic cell death but maintains normal mitochondrial morphology, it is likely that mitochondrial fragmentation and apoptosis caused by PHB2 knockdown occur by different pathways.

Mitochondrial PHB2 Localizes to the Nucleus in an ER␣-dependent Manner
We found that, in HeLa cells, mitochondrial PHB2 translocates to the nucleus in the presence of ER␣ and E2. Therefore, E2-dependent binding of ER␣ to PHB2 is thought to be essential for its nuclear targeting. In fact, we found that the C-terminal domain of PHB2 that contains the previously reported ER␣-binding site (32) was necessary and sufficient for the ER␣-dependent nuclear translocation. The ER␣-dependent nuclear translocation occurred with PHB2 but not PHB1, demonstrating a highly specific interaction between PHB2 and ER␣.
PHB proteins contained an atypical MTS at their N termini that was sufficient for their mitochondrial targeting. Because the N-terminal region of PHB2 did not respond to ER␣ and E2, it appears that the nuclear targeting and mitochondrial targeting of PHB2 occurs independently. Considering the existence of separate mitochondrial targeting and ER␣-dependent nuclear localization sequences intramolecularly, we suspect that PHB2 shuttles between the mitochondria and nucleus. By the chase assay of PHB2-GFP in the same cell, it was further supported that PHB2 shuttles between the two organelles.
Several proteins have been reported to localize both in the mitochondria and nucleus. Among these, however, there are few examples of conditional translocation between the two organelles. Two such proteins, apoptosis-inducing factor and endonuclease G, are localized in mitochondria and translocate to the nucleus following stimulation with inducers of apoptosis (36). In these cases, unlike PHB2, the proteins are thought to possess MTS that are processed after their import into the mitochondria (37,38), suggesting that, once released from the mitochondria, they cannot re-enter. Therefore, PHB2 likely has a distinct mode of translocation from these molecules in that it potentially shuttles between the organelles via non-cleavable targeting sequence. Fig. 8 shows a summary of the location and functions of PHB2. Through an unconventional N-terminal MTS, PHB2 is imported into mitochondria where it plays multiple roles, such as inhibition of apoptosis through PHB2⅐Hax-1 complex and maintenance of mitochondrial morphology via OPA1. On the other hand, binding of the ER␣⅐E2 complex causes mitochondrial PHB2 to translocate to the nucleus where it inhibits ER␣dependent transcription. Then, how PHB2 in the mitochondria detects ER␣ and E2 in the first place? Recently, ER␣ has also been shown to localize in the mitochondria (39). Therefore, it is not surprising that the first interaction between PHB2 and ER␣ occurs in the mitochondria. If so, it is possible that PHB2 is delivered from the mitochondria to nucleus by ER␣. Alternatively, if PHB2 might shuttle from the mitochondria to cytoplasm, because it possesses an unprocessed MTS, the possibility should not be excluded that PHB2 firstly binds to ER␣ in the cytoplasm.
Recently, a PHB2 knock-out mouse was produced, although the homozygous animals did not develop and died in the embryonic stage (40). An essential role of PHB2 in early development agrees well with findings in C. elegans (17). Thus, both the transcriptional function and mitochondrial functions of PHB2 may be required for development. Alternatively, it was recently reported that ERs are present and function in the mitochondria (39). E2-dependent interaction between ER␣ and PHB2 might occur in the mitochondria and be important for the mitochondrial targeting and functions of ER␣.
PHB2 is a versatile molecule that regulates transcription in the nucleus as well as regulation of mitochondrial functions in the mitochondria. PHB2 also possesses a non-cleavable N-terminal MTS, indicating a possibility of shuttling between the two organelles. Considering above, PHB2 would be a candidate that couples nuclear-mitochondria interaction. Further experiments to prove the physiological significance of the nuclearmitochondrial interaction by PHB2 should be required.