Hsp90 Inhibition Decreases Mitochondrial Protein Turnover

Background Cells treated with hsp90 inhibitors exhibit pleiotropic changes, including an expansion of the mitochondrial compartment, accompanied by mitochondrial fragmentation and condensed mitochondrial morphology, with ultimate compromise of mitochondrial integrity and apoptosis. Findings We identified several mitochondrial oxidative phosphorylation complex subunits, including several encoded by mtDNA, that are upregulated by hsp90 inhibitors, without corresponding changes in mRNA abundance. Post-transcriptional accumulation of mitochondrial proteins observed with hsp90 inhibitors is also seen in cells treated with proteasome inhibitors. Detailed studies of the OSCP subunit of mitochondrial F1F0-ATPase revealed the presence of mono- and polyubiquitinated OSCP in mitochondrial fractions. We demonstrate that processed OSCP undergoes retrotranslocation to a trypsin-sensitive form associated with the outer mitochondrial membrane. Inhibition of proteasome or hsp90 function results in accumulation of both correctly targeted and retrotranslocated mitochondrial OSCP. Conclusions Cytosolic turnover of mitochondrial proteins demonstrates a novel connection between mitochondrial and cytosolic compartments through the ubiquitin-proteasome system. Analogous to defective protein folding in the endoplasmic reticulum, a mitochondrial unfolded protein response may play a role in the apoptotic effects of hsp90 and proteasome inhibitors.


INTRODUCTION
Hsp90 is an abundant cytosolic chaperone (1-2% of cytosolic protein) involved in the turnover, trafficking and activity of a large number and variety of client proteins. These include membraneassociated and soluble protein kinases (many recognized as oncogenes) and transcription factors [1,2]. Small molecule hsp90 inhibitors that bind to the N-terminal ATP binding pocket inhibit chaperone function. Consequently, many client proteins are targeted for degradation via the ubiquitin-proteasome pathway. In response to hsp90 inhibition, cancer cell lines exhibit several types of response, including reversal of transformation, differentiation and apoptosis [3][4][5]. Selective cytotoxicity for cancer cells is associated with expression of hsp90 client proteins, such as Bcr-Abl, FLT3 bearing an internal tandem duplication, and ErbB2, and has been suggested to reflect a survival requirement for multiple signaling pathways that depend on hsp90 chaperone function [6][7][8][9]. Apoptotic cell death in hsp90-inhibited cells involves mitochondrial pathways with cytosolic accumulation of cytochrome c and SMAC/Diablo [10].
As an early event in cells destined for apoptosis, hsp90 inhibitors induce mitochondrial proliferation, resulting in the accumulation of a fragmented network of mitochondria filled with homogeneous, electron-dense material obscuring the internal cristae [11,12]. Expansion of the mitochondrial compartment is associated with reduced mitochondrial membrane potential (DY M ), pointing to a possible defect in mitochondrial biogenesis. Furthermore, mitochondria isolated from cells treated with hsp90 inhibitors stimulate nuclear condensation in cell-free apoptosis assays.
Unlike most of the known cytoplasmic clients of hsp90, we demonstrate that mitochondrial proteins accumulate in cancer cells treated with hsp90 inhibitors. Mitochondrial protein upregulation occurs at a post-transcriptional level, and similar changes in mitochondrial ultrastructure and mitochondrial protein expression are exhibited in cancer cells treated with proteasome inhibitors. Furthermore, we demonstrate that the F1F0-ATPase OSCP subunit undergoes hsp90-dependent ubiquitination and has an increased protein half-life in hsp90-inhibited cells. This process of organellar protein turnover is reminiscent of endoplasmic reticulum associated degradation (ERAD), which functions both as a quality control and a regulatory pathway for ER proteins [13,14].
Hsp90-regulated degradation of mitochondrial proteins represents a novel and unexpected pathway in overall cellular economy. Accumulation of mitochondrial proteins leading to intra-organellar protein folding stress may be an early and important event in cells treated with hsp90 or proteasome inhibitors for triggering mitochondrial apoptosis pathways.

Hsp90 inhibition triggers mitochondrial protein accumulation
COLO 205 colon cancer cells treated with hsp90 inhibitors, including herbimycin A (HA), 17-allylamino-17-demethoxygeldanamycin (17-AAG) and radicicol, undergo terminal differentiation and apoptotic cell death within 72-96 h, preceded by dramatic changes to the mitochondrial compartment or chondriome [11] ( Fig. 1 A and data not shown). Both the number of mitochondrial profiles and mitochondrial mass per cell, assessed by flow cytometry of cells stained with nonyl acridine orange, increased at 24-72 h of treatment with hsp90 inhibitors (Fig. 1 B-D and data not shown). Electron micrographs (EM) demonstrate that, in hsp90 inhibitor-treated cells, the mitochondrial matrix becomes progressively opacified with electron-dense material, and by 48 h the internal cristal membranes are completely obscured ( Fig. 2A, B).
Increased mitochondrial matrix density can be due to a loss of water and ions (condensed conformation), but also may represent increased matrix protein content [15]. Mild pronase treatment of EM sections removed the electron-dense matrix material, consistent with its proteinaceous nature (Fig. 2C, D). Mitochondrial fractions were prepared for protein quantitation at various times after addition of hsp90 inhibitor. Total mitochondrial protein (normalized to cell number) increased two-fold after 48 h of treatment with HA or radicicol (Fig. 2E). Mitochondrial protein normalized to mitochondrial volume per cell (estimated by the summed area of EM mitochondrial profiles) also appeared to increase after HA treatment (Fig. 2F).
We next determined whether expression of individual mitochondrial proteins changed in cells treated with hsp90 inhibitors (Fig. 3A). Immunoblots demonstrated that subunits of cytochrome c oxidase (COX I, III, IV, Vb), F1F0-ATPase (Complex V a, d, OSCP), and Complex III (Core 2), as well as cytochrome c, were upregulated with HA treatment. Mitochondrial proteins targeted to the outer mitochondrial membrane (VDAC) and matrix (pyruvate dehydrogenase E1a) also accumulated in HA-treated cells. Several electron transport chain subunits (Complex I 39 and 30 kD, Complex II 70 kD, and Complex V b) were not induced, indicating that hsp90 inhibition produced selective changes to the mitochondrial proteome (Fig. 3A). Protein expression levels of three non-mitochondrial proteins, proliferating cellular nuclear antigen (PCNA), glyceraldehyde-3-phosphate dehydrogenase (GAPDH) and tubulin, did not change in HA-treated cells. Similar increases in mitochondrial protein expression were observed with another hsp90 inhibitor, 17-AAG (Fig. 3B) and a second cell line, 143B osteosarcoma cells (Fig. 3C).
To investigate the level of regulation of mitochondrial protein expression in hsp90 inhibitor-treated cells, we performed Northern blots for mRNAs representing nuclear and mtDNA-encoded mitochondrial proteins (Fig. 4A). Neither mRNAs for chromosomal genes (COX IV (COX4), ATPase a (ATP5A), and OSCP (ATP5O)) nor the mtDNA gene COX I (COI) encoding proteins upregulated by hsp90 inhibitors were induced by HA treatment (Fig. 4A, B). In addition, mtDNA copy number did not increase with HA treatment (Fig. 4C). Thus, hsp90 inhibitors increase expression of several mitochondrial proteins via post-transcriptional regulation.

Ubiquitination and proteasomal degradation of mitochondrial proteins
We next examined mitochondrial protein synthesis and turnover in hsp90 inhibitor-treated cells. 35 S-methionine/cysteine pulse-chase labeling followed by immunoprecipitation of the nuclear-encoded F1F0-ATPase OSCP subunit revealed an increase in protein half-life from 36-48 h in untreated cells to greater than 48 h in HA-treated cells (Fig. 5A, B). No differences were observed in initial 35 S-labeling of OSCP between untreated and HA-treated cells. Therefore, the increase in mitochondrial OSCP protein expression following hsp90 inhibition is due to reduced protein turnover.
Hsp90 co-purifies with mitochondria in COLO 205 cells, but dissociates after treatment with hsp90 inhibitors (Fig. 5C). In addition to a direct role in stability of client proteins, hsp90 also functions in protein turnover mediated by the chaperone-dependent ubiquitin E3 ligase, CHIP [16]. To determine whether ubiquitin-dependent protein turnover affects the mitochondrial compartment, we treated COLO 205 cells with proteasome inhibitors. Condensed mitochondrial morphology and increased mitochondrial mass developed within 24 h of treatment with MG132, comparable to that observed with hsp90 inhibitors (Fig. 6). Immunoblot analysis showed increased expression of the same mitochondrial proteins (COX I, III, IV, OSCP, Complex V-alpha) with two different proteasomal inhibitors, MG132 and MG232, as seen with 17-AAG treatment (Fig. 7A). In contrast, Raf-1, a known hsp90 client protein, is degraded in cells treated with 17-AAG but accumulated in cells treated with proteasome inhibitors.
Hsp90 and proteasome inhibitors can cause ER protein folding stress and modulate the unfolded protein response (UPR) [17,18]. To test whether mitochondrial protein accumulation is secondary to ER protein folding stress, we treated OSCP-GFP expressing 143B cells with tunicamycin, an inducer of the ER UPR. OSCP-GFP protein levels did not increase in cells treated wtih tunicamycin, while in combination with 17-AAG, tunicamycin reduced mitochondrial protein accumulation compared with 17-AAG alone ( Figure S1). We also tested inhibitors of autophagy for effects on mitochondrial protein expression in COLO 205 cells. Treatment with two different lysosomal inhibitors, monensin and chloroquine, or a macroautophagy inhibitor, 3-methyladenine (3-MA) did not increase mitochondrial protein expression at 24 h ( Fig. 7 A, B).
We next investigated whether proteasomal functions were directly involved in turnover of mitochondrial proteins. Purified mitochondria were prepared from a 10,0006g pellet fraction of COLO 205 cell lysates by fractionation on a continuous iodixanol gradient (19-27%). Anti-ubiquitin immunoblotting recognized a high molecular weight smear in mitochondrial extracts from untreated cells, consistent with polyubiquitinated species (Fig. 8A). Mitochondria prepared from cells treated with HA or MG132 had reduced amounts of ubiquitinated proteins. Ubiquitin immuno-reactive bands were sensitive to trypsin digestion of intact mitochondria, but mitochondrial association was retained in 0.1 M sodium carbonate (pH 11.5), suggesting that ubiquitinated proteins are either tightly bound to the mitochondrial surface or inserted in the outer mitochondrial membrane (OMM) (Fig. 8B). Digitonin extracts of OMM also contained high molecular mass ubiquitinated bands, although a greater proportion remained associated with the mitoplast fraction (Fig. 8C).
Immunoprecipitation of OSCP from cellular extracts recovered a broad smear at ,80-180 kD size and a 30 kD band in addition to the expected 21 kD band representing mature OSCP. Subsequent immunoblotting with an anti-ubiquitin antibody recognized the 80-180 kD and 30 kD species, but not the 21 kD band, consistent with poly-and mono-ubiquitinated OSCP (Fig. 8D). OSCP immunoprecipitates from cells treated with HA exhibited only a 21 kD band, with no reactivity on anti-ubiquitin immunoblots. In view of the effect of proteasome inhibitors on OSCP protein expression, these results are consistent with ubiquitin-mediated OSCP turnover. Furthermore, hsp90 appears to be required for the attachment of ubiquitin to OSCP.   Post-import proteasomal turnover of mitochondrial proteins OSCP is a component of the F1F0-ATP synthase complex localized to the mitochondrial matrix, physically separated from cytoplasmic ubiquitin ligases. To address whether imported mitochondrial OSCP was susceptible to proteasomal degradation, we stably expressed an OSCP-GFP fusion protein in 143B cells. OSCP-GFP is efficiently imported to the mitochondrial matrix [19]. Proteasome and hsp90 inhibitors increased expression of OSCP-GFP, as well as the endogenous OSCP protein (Fig. 9A).
Monitoring of GFP fluorescence intensity by flow cytometry provided results similar to immunoblot analysis ( Fig. 9B-top panel). To isolate effects on imported mitochondrial OSCP-GFP, we treated cells with the uncoupling agent CCCP to prevent new DY M -dependent import of matrix-targeted proteins prior to adding proteasome or hsp90 inhibitors. CCCP treatment of 143B-OSCP-GFP cells decreased GFP fluorescence, as expected [20]. Even so, treatment with MG132 or HA increased OSCP-GFP expression in cells pre-treated with CCCP, as shown by immunoblot and flow cytometry, with a mitochondrial localization determined by fluorescence microscopy ( Fig. 9A-C). These results    are consistent with hsp90-and proteasome-dependent degradation of the OSCP subunit after mitochondrial import.
In an established model of organellar protein turnover by ubiquitin-mediated proteasomal degradation, endoplasmic reticulum-associated degradation (ERAD), endoplasmic reticulum proteins are exported to the cytoplasm in a process termed retro-translocation. We questioned whether a similar process could operate for mitochondrial proteins. ERAD retro-translocation intermediates can be trapped by inhibition of downstream proteasomal processing [21][22][23][24]. We analyzed the localization of OSCP in mitochondrial membrane fractions before and after treatment with proteasome inhibitors. Sub-mitochondrial fractionation demonstrated the expected distribution for OSCP in mitoplasts (inner mitochondrial membrane and matrix) isolated from control cells (Fig. 10A). The abundant outer mitochondrial membrane protein VDAC is incompletely extracted by digitonin, presumably due to its association with contact sites between inner and outer mitochondrial membranes [25]. However, in cells treated with HA or MG132, a substantial proportion of the mature 21 kD OSCP is associated with the OMM. Notably, removal of the OSCP transit peptide occurs in the mitochondrial matrix, implying that export is required for the appearance of the 21 kD OSCP protein in the OMM fraction. A smaller amount of COX II appears in the OMM fraction from treated cells, suggesting that mtDNA-encoded proteins may also be exported.
In order to detect dynamic retro-translocation, we pulse-labeled cells with 35 S-methionine/cysteine and examined mitochondrial import of OSCP by immunoprecipitation from mitochondrial fractions. In parallel, mitochondria were treated with trypsin to degrade OSCP associated with the OMM. The percentage of trypsin-resistant 35 S-labeled mitochondrial OSCP increased during the first 4 h of chase, reflecting mitochondrial import of newly synthesized protein (Fig. 10B). However, after an additional 4 h chase period, the percentage of OSCP that was trypsin-sensitive increased in mitochondrial fractions, consistent with retro-translocation of previously imported protein.

Regulation vs quality control
Ubiquitin-mediated proteasomal degradation of ER proteins provides a mechanism for regulating protein expression as well as quality control of protein folding. To examine the effect of hsp90-regulated mitochondrial protein turnover on assembly of a multi-protein electron transport complex in COLO 205 cells, we solubilized mitochondrial membranes in dodecyl maltoside detergent and analyzed complex assembly by centrifugation on a sucrose density gradient [26]. Immunoblots were performed on collected fractions for individual cytochrome c oxidase subunits (Fig. 11). In untreated COLO 205 cells, COX I and COX II exhibit a broad peak centered at fraction 5, while COX IV has two peaks, one in fraction 5 coincident with COX I/II, and a second peak at lower density in fraction 8. In cells treated with hsp90 inhibitor, a unimodal distribution of COX IV is observed, with disappearance of the low density peak. These results suggest that inhibition of hsp90-regulated turnover of mitochondrial proteins can promote assembly of respiratory chain complexes, consistent with a regulatory function. Although oxygen consumption in COLO-205 cells treated with HA is only ,35% of control cells, maximal respiration rates in the presence of the uncoupler CCCP, are identical for treated and untreated cells, suggesting that the total respiratory capacity is maintained for a significant period, in agreement with the complex assembly data ( Figure S2).
In its quality control function, ERAD extracts microsomal proteins as part of the UPR. We observed a consistent reduction in the amount of mitochondrial protein extractable with nonionic detergents in cells treated with hsp90 inhibitors (Fig. 12A). Taking into account the increase in total (SDS-extractable) mitochondrial protein in HA-treated cells, the amount of Triton X-100 insoluble mitochondrial protein increased more than three-fold after 48 h of HA treatment.
We surmised that the dense mitochondrial matrix observed in electron micrographs of cells treated with proteasome or hsp90 inhibitors could represent aggregated or unfolded protein. In order to examine mitochondrial protein solubility in situ, we lightly fixed cells prior to extraction in low concentrations of non-ionic detergent, followed by processing for electron microscopy. Control COLO 205 cells exhibited empty mitochondrial profiles consistent with complete extraction of mitochondrial matrix proteins with up to 1 h of pre-fixation (Fig. 12B). In contrast, mitochondrial matrix proteins in HA-treated cells were completely resistant to Triton X-100 extraction with as little as 15 min of pre-fixation ( Fig 12C). Thus, one consequence of mitochondrial protein accumulation in cells treated with hsp90 inhibitors is a change in protein solubility, with potentially negative consequences for mitochondrial function.

DISCUSSION
Hsp90 inhibitors induce apoptosis, with selectivity for cancer cells, via mitochondrial pathways. Analysis of the mitochondrial compartment, using mitochondrially targeted fluorescent probes and electron microscopy, revealed early increases in mitochondrial mass, associated with distinctive ultrastructural changes, in cells treated with hsp90 inhibitors. To investigate the basis of these changes, we analyzed mitochondrial protein expression. We identified an increase in total mitochondrial protein, normalized to cell number, as well as multiple individual mitochondrial proteins in COLO 205 and 143B cells treated with hsp90 inhibitors, including nuclear-encoded (COX IV, Vb, F1F0-ATPase a, d, OSCP, cytochrome c, VDAC1, PDH E1a) and mtDNA-encoded respiratory complex subunits (COX I, III). Notably, several other respiratory complex subunits examined were unaffected. We excluded transcriptional regulation for several of these proteins by the stable steady state levels of cognate cytosolic and mitochondrial mRNAs and mtDNA copy number. Unexpectedly, ATP5A and ATP5O mRNAs declined after 48 and 72 h of HA treatment (Fig. 4). This may represent a retrograde response (mitochondria-to-nuclear signaling) to the increased mitochondrial mass.
Pulse-chase studies demonstrated that the mitochondrial F1F0-ATPase OSCP subunit is stabilized in hsp90-inhibited cells. Treatment with proteasome inhibitors MG132 and MG262, but not lysosomal (monensin, chloroquine) or autophagy (3-methyladenine) inhibitors, produced a similar pattern of mitochondrial protein upregulation in COLO 205 cells, suggesting that proteasomal protein degradation pathways are involved in mitochondrial protein turnover. We next demonstrated that immunoprecipitation of OSCP from purified mitochondrial lysates isolated several slower migrating bands recognized by anti-ubiquitin immunoblots and consistent with mono-and poly-ubiquitinated OSCP, in addition to the mature 21 kD OSCP. This result is consistent with direct proteasomal degradation of a mitochondrial protein, mediated via poly-ubiquitin conjugation. Furthermore, hsp90 inhibition suppressed the recovery of ubiquitin-modified OSCP in mitochondrial lysates, suggesting hsp90 is required for effective interaction of mitochondrial OSCP with one or more ubiquitin ligases.
The ubiquitin-proteasome system is not previously known to participate in the turnover of mitochondrial proteins. However, several proteomic efforts to catalogue ubiquitin-modified proteins have identified mitochondrial proteins including respiratory chain subunits residing in the inner mitochondrial membrane and matrix (see Table 1). A recent analysis of the mouse heart ubiquitinated proteome shows that proteins localized in mitochondria represent a major fraction (38 %) and expands the list of the ubiquitin conjugated mitochondrial proteins [33]. Proteasomal turnover of the mitochondrial heme synthetic enzyme 5aminolevulinic acid synthase 2 was demonstrated [34]. The prominent 30 kD band consistent with monoubiquitinated OSCP, could represent an intermediate product in polyubiquitination, or carry out a separate role in protein trafficking, lysosomal degradation, or protein interactions [35].
Proteasome and hsp90 inhibitors increased the expression and mitochondrial fluorescence of an OSCP-GFP fusion protein, even  Voltage-dependent anion selective channel protein 2 POR2_HUMAN X X mitochondrial transport [28], [29] 44 Translocase of the outer mitochondrial membrane TOM5 X mitochondrial translocation [27] 45 Translocase of the outer mitochondrial membrane TOM7 X mitochondrial translocation [27] 46 Translocase of outer mitochondrial membrane, 70 kDa OM70_HUMAN X X mitochondrial translocation [27] 47 Mitochondrial import receptor subunit Tom22 homolog OM22_HUMAN X mitochondrial translocation [28] 47 Translocase of the inner membrane TIM50 X mitochondrial translocation [27] . in the presence of a protonophore causing collapse of mitochondrial membrane potential, DY M . Mitochondrial import of matrixtargeted proteins, including OSCP, is DY M -dependent, implying that these inhibitors affect the turnover of previously imported mitochondrial OSCP. Ubiquitin-conjugated proteins in mitochondrial fractions are tightly associated with the outer mitochondrial membrane as shown by digitonin extraction, protease susceptibility, and resistance to alkali extraction. These results are analogous to protein ubiquitination in endoplasmic reticulum associated degradation (ERAD), where both soluble and membrane inserted ER proteins undergo retro-translocation before ubiquitin conjugation at the cytosolic face of the ER [13,14]. Mitochondrial subfractionation demonstrated a shift in localization of mature OSCP to the OMM in cells treated with hsp90 or proteasome inhibitors. This likely represents a retro-translocation intermediate, as observed for several ERAD substrates [21][22][23][24], and suggests that hsp90 function and proteasome activity are required, at least in part, for complete retro-translocation. Hsp90 co-purified with COLO 205 mitochondria, although the association was diminished in HA-and MG132-treated cells. TRAP1, a mitochondrial paralog of hsp90, is also inhibited by the tested hsp90 inhibitors [36,37]. Knockdown of TRAP1 expresssion using TRAP1 short-hairpin RNAs did not affect responses to 17-AAG, although basal expression levels for several mitochondrial proteins were diminished consistent with the postulated role of TRAP1 as an intramitochondrial chaperone ( Figure S3).
Hsp90 is reported to function in protein degradation, in addition to its role in protein folding, via its association with CHIP, an E3 ubiquitin ligase [16]. Reduction of CHIP protein expression by 60-90% by shRNA-mediated RNA interference did not result in increased mitochondrial protein expression or suppress responses to 17-AAG ( Figure S3), raising the possibility of alternative E3-ligases, perhaps recognizing processed mitochondrial proteins as N-end rule substrates, as hsp90 co-factors. Hsp90 may also facilitate transport of ubiquitin-conjugated OSCP to the proteasome. Hsp90 inhibition leads to an increase in insoluble ubiquitinated cytosolic proteins [38].
Mitochondrial proteins can also be degraded in a nonselective manner via autophagy, or more specifically, mitophagy. Autophagic degradation of organelles and bulk cytosolic components is physiologically activated during nutrient starvation to maintain an adequate supply of nutrients for cell survival. Inhibition of autophagy with 3-MA or inhibition of lysosomal activity with monensin or chloroquine did not lead to increased mitochondrial protein expression in COLO 205 cells grown in standard media. In contrast, 3-MA increased OSCP-GFP expression in cells grown under nutrient starvation conditions [39] (Figure S4). Our data suggests that hsp90 and the 26S proteasome are components of an alternative pathway for selective degradation of mitochondrial proteins.
Increased mitochondrial protein expression in cells treated with hsp90 inhibitors was associated with improved cytochrome oxidase assembly, but ultimately led to an increase in detergent-insoluble mitochondrial proteins, and mitochondrial dysfunction [11]. Analogous to defective protein folding in the endoplasmic reticulum, a mitochondrial unfolded protein response may play a role in the apoptotic effects of hsp90 and proteasome inhibitors [40].

Reagents and Antibodies
All chemicals were purchased from Sigma-Aldrich with the exception of MG262 and Tunicamycin (Calbiochem). Herbimycin A, 17-AAG and radicicol (Sigma) were dissolved in DMSO at 1 mg/ml and stored in aliquots at 270uC. MG132 and MG262 were prepared as 1 mM stocks in DMSO and stored at 270uC.

Flow cytometry
For nonyl acridine orange (NAO) staining, cells were resuspended in phosphate-buffered saline (PBS) at 1610 6 cells/ml and incubated with 1 nM NAO for 10 min at room temperature in the dark. Cells were analyzed on a FACScan cytometer (Becton Dickinson). Annexin V binding and propidium iodide (PI) staining was performed according to manufacturer's protocol (Annexin V:FITC Apoptosis Detection Kit II, BD Biosciences).

Electron microscopy
Transmission electron microscopy was performed as described [41]. To address the nature of the dense mitochondrial matrix material, ultrathin sections of cell pellets embedded in watersoluble media were incubated with 1 mg/ml pronase at 37uC for 1 h. Cells were pre-fixed for various times (15 min, 30 min or 1 h) in 0.25% glutaraldehyde, exposed to 0.25% Triton X-100 in PBS for 15 min and post-fixed in half strength Karnovsky's fixative. Transmission electron micrographs were taken using a Jeol 1010 transmission electron microscope operating at 80 kV. The 2D area of mitochondrial profiles was calculated from transmission electron micrographs using Image J (NIH). Profiles were traced in sets of 10 cells (6,000x magnification) for control and HAtreatment. The mitochondrial area was summed for each cell to derive an average total mitochondrial area per cell, calculated in mm 2 after correction by the magnification factor.

Western blotting
Whole cell lysates were prepared from 5610 6

Mitochondrial preparations
Cells were disrupted either by nitrogen cavitation (500 psi for 10 min) or 10-15 passages through a 25G needle in homogenization buffer (0.25 M sucrose, 1 mM EGTA, 10 mM HEPES, 0.5 % bovine serum albumin, pH 7.4, with protease inhibitors). Mitochondrial pellets were obtained from post-nuclear supernatant by centrifugation at 10,0006g. For some studies, mitochondria were purified by centrifugation on a continuous 19-27% iodixanol gradient (Optiprep) for 2 h at 70,0006g. Alkali extraction was performed with 0.1 M Na 2 CO 3 for 30 min on ice; trypsin digestion (50 mg/ml) was performed at 37uC. Outer mitochondrial membrane (OMM) preparations were obtained by digitonin (1 mg/mg protein) extraction for 20 min. Sucrose gradient fractionation of respiratory chain complexes was performed as described [26].

Immunoprecipitation
The F1F0-ATPase OSCP subunit was immunoprecipitated from cells lysed in TENT buffer (0.5% Triton X-100, 50 mM Tris-HCL, 150 mM NaCL, 5 mM EDTA, pH 7.4 with freshly added protease inhibitors) at 4uC. After centrifugation at 12,0006g for 15 min to remove insoluble material, the extracts were incubated overnight at 4C with anti-OSCP antibody cross-linked to Protein G Sepharose beads (Amersham Biosciences), washed three times with TENT buffer, and eluted with 100 mM glycine, pH 2.8. Eluted proteins were neutralized with 1 M Tris pH 8.0, denatured at 65uC for 10 min in 1X Laemmli buffer with 10 mM fresh DTT, and separated by 4-15% gradient SDS-PAGE for Western blotting.
For pulse-chase studies, cells were plated 24 h before labeling with 50 mCi/ml [ 35 S]-methionine/cysteine (Trans label, ICN) in met/cys-free medium. For trypsin sensitivity of mitochondrial OSCP, cells were pulse-labeled for 2 h and then chased in the presence of cycloheximide 100 mg/ml. Mitochondrial suspensions were prepared at each time point (from 8610 6 cells). One-half of each sample was incubated with trypsin 50 mg/ml for 20 min at 37uC, followed by immunoprecipitation of OSCP from TENT extracts of trypsin-treated and untreated mitochondria. Gels were fixed for 30 min at room temperature in 40% methanol, 10% acetic acid, and 5% glycerol, followed by incubation in Amplify Fluorographic Reagent (GE Healthcare), gel drying and exposure to phosphor storage screens (Molecular Dynamics). Autoradiograph images (Typhoon 8600, GE Healthcare) were analyzed using ImageQuant software.

Differential solubility of mitochondrial proteins
Mitochondrial fractions (50 mg protein) were extracted in TENT buffer for 30 min at 4C. Soluble proteins were recovered by centrifugation for 30 min at 20,0006g. The insoluble pellet was dissolved in CHAPS-Urea (8M urea, 2% CHAPS, 150 mM NaCl, 5 mM EDTA, 10 mM Tris pH 7.4) at room temperature. Protein concentration of TENT-soluble and insoluble fractions was measured by Bradford assay (Bio-Rad).

Fluorescence microscopy
143B cells expressing OSCP-GFP fusion protein were grown on glass coverslips and treated with 0.5 mM HA, 20 mM CCCP, or both for 24 h. Cells were fixed in 4% paraformaldehyde and mounted on slides for fluorescence imaging (Nikon Eclipse E800, 100X Plan Fluor objective). Photomicrographs of cells subjected to different treatments were acquired during the same experimental session with identical exposure times.

Statistics
Statistical analysis was done using Student's paired t-test. The data are expressed as means6standard error and statistical significance was indicated by P,0.05. Figure