Functional F1-ATPase Essential in Maintaining Growth and Membrane Potential of Human Mitochondrial DNA-depleted ρ° Cells*

F1-ATPase assembly has been studied in human ρ° cells devoid of mitochondrial DNA (mtDNA). Since, in these cells, oxidative phosphorylation cannot provide ATP, their growth relies on glycolysis. Despite the absence of the mtDNA-coded F0 subunits 6 and 8, ρ° cells possessed normal levels of F1-ATPase α and β subunits. This F1-ATPase was functional and azide- or aurovertin-sensitive but oligomycin-insensitive. In addition, aurovertin decreased cell growth in ρ° cells and also reduced their mitochondrial membrane potential, as measured by rhodamine 123 fluorescence. Therefore, a functional F1-ATPase was important to maintain the mitochondrial membrane potential and the growth of these ρ° cells. Bongkrekic acid, a specific adenine nucleotide translocator (ANT) inhibitor, also reduced ρ° cell growth and mitochondrial membrane potential. In conclusion, ρ° cells need both a functional F1-ATPase and a functional ANT to maintain their mitochondrial membrane potential, which is necessary for their growth. ATP hydrolysis catalyzed by F1 must provide ADP3− at a sufficient rate to maintain a rapid exchange with the glycolytic ATP4− by ANT, this electrogenic exchange inducing a mitochondrial membrane potential efficient enough to sustain cell growth. However, since the effects of bongkrekic acid and of aurovertin were additive, other electrogenic pumps should cooperate with this pathway.

The biogenesis of mitochondrial proteins is controlled by both nuclear and mitochondrial genomes. The proteins coded by the mtDNA are subunits of enzyme complexes involved in oxidative phosphorylation. Since all these complexes also contain proteins coded by the nuclear genome, mechanisms regulating the coordinated expression and assembly of the subunits of nuclear and mitochondrial origin must exist. In yeast cells, nuclear DNA-coded components continue to be synthesized and imported into mitochondria, even when the synthesis of mtDNA-coded subunits is blocked (cf. for review, Refs. 1 and 2). The groups of Schatz and co-workers (3) and Neupert (4) have shown that a mitochondrial membrane potential is a key requirement for protein import into mitochondria (5). In normal cells, the mitochondrial membrane potential is mainly maintained by transmembrane proton pumping occurring during electron transfer or during ATP hydrolysis catalyzed by the ATPase-ATP synthase. In °cells depleted of mtDNA, these complexes cannot be functional since all complexes involved in proton pumping contain mtDNA-coded subunits (6). However, proteins of nuclear origin are imported into the °cell mitochondria (7). Therefore, the mitochondrial membrane potential must be maintained by other electrogenic pumps. In yeast, it has been suggested that the adenine nucleotide translocator (ANT) 1 mediates an exchange of ATP 4Ϫ synthesized in the cytosol during glycolysis for ADP 3Ϫ to maintain this mitochondrial membrane potential (8). However, the exact mechanism that maintains this potential has not been thoroughly investigated. Since the mtDNA-coded subunits are essential components for oxidative phosphorylation, °cells lacking mtDNA rely on glycolysis for their energy demand. The NADH produced during glycolysis is then reoxidized by the lactate dehydrogenase (9).
In addition, several reports indicate that some nuclearly coded subunits of the respiratory chain complexes are imported and partly assembled into the mitochondria of the °cells (10). The role (if any) of these partly assembled complexes is unknown.
The present study reports observations concerning the role and mechanism of assembly of the mitochondrial ATPase in human mtDNA-deficient °cells. The ATPase-ATP synthase (F0F1) is made of a hydrophilic portion F1 required for enzyme catalysis and connected by a stalk to the hydrophobic sector F0, which is involved in proton translocation occurring during ATP synthesis or ATP hydrolysis. (cf. Ref. 11, for review). In mammals, F1 contains five different subunits: ␣, ␤, ␥, ␦, and ⑀. F0 is made of the subunits a (or subunit 6), b, c (also named subunit 9 or DCCD-binding protein), d, e, f, g, and A6L (or subunit 8); the stalk contains OSCP, F6, and parts of F0 such as subunits b and d or of F1 such as ␥ (12). The subunits a (or 6) and A6L (or 8) that are coded by the mtDNA are absent in human °c ells. Here, we show that F1 is assembled in human °cells as an azide-or aurovertin-sensitive ATPase and that this ATPase activity is essential for the °cell growth because it is involved in the maintenance of their mitochondrial membrane potential.

MATERIALS AND METHODS
Materials-Cell culture reagents were from Life Technology Inc. except uridine, which was from Sigma. The °HeLa S3 cells devoid of * This work was supported by grants from the CNRS, the French Ministry of Education and Scientific Research (MERS), the Association Française contre les Myopathies (AFM), and the Région Rhône-Alpes. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 (13). The ϩ and °cells were kindly provided to us by Dr. Vayssière and Dr. Morais. The bongkrekic acid was a generous gift of Prof. P. V. Vignais. The peroxidase-labeled anti-mouse antibody was obtained from Biosys. The anti-F1-ATPase ␣ and ␤ subunits were prepared from the clones 20D6 and 14D5 (14), and the anti-cytochrome oxidase subunit II and subunit IV were prepared from the clones: 12C4-F12 and 10G8-C12-D12, respectively, kindly provided by Dr. Taanman (15). The fluorescent molecular probe NAO was obtained from Molecular Probes. The nitrocellulose membranes were from Schleicher and Schuell. All other biochemical reagents were from Boehringer Mannheim or Sigma.
Cell Culture-HeLa S3 cells were grown at 37°C in a humidified atmosphere with 5% CO 2 in Dulbecco's modified Eagle's medium, Ham's F-12 supplemented with 2% Ultroser G, antibiotics (200 units/ml penicillin and 200 g/ml streptomycin), and 0.5 mM pyruvate. Osteosarcoma 143B cells were grown in the same atmosphere in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum, antibiotics (200 units/ml penicillin plus 200 g/ml streptomycin), and 1 mM pyruvate. Uridine was added in cell culture media for °HeLa S3 cells (100 g/ml) and °143B cells (50 g/ml). Centrifugation of °143B cells, which are easily damaged, was avoided.
Verification of the Absence of Functional mtDNA by PCR and Reverse Transcription-PCR-ϩ and °cell pellets were treated for 1 h at 60°C with 10 g of proteinase K dissolved in PCR buffer, and quantitative PCR was directly performed on serial dilutions of cells to amplify mtDNA. Two couples of forward and reverse primers hybridized with the mtDNA sequence at positions 592 to 611 and 1344 to 1325 or 15437 to 15456 and 15777 to 15758, located at the level of the 12 S RNA and of the cytochrome b genes (6), respectively. The PCR was performed as in Boudizi et al. (16), using hybridization temperatures of 55 and 53°C, respectively. A control to check the amount of DNA in the various cell extracts was made by amplification of the nuclear gene coding for the 18 S RNA from positions 1196 to 1806 using 17-mer oligomers and a hybridization temperature of 62°C.
Cellular RNA extraction and reverse transcription were performed on ϩ and °cells as described previously (17). The PCR amplification of cDNAs corresponding to mitochondrial RNAs was made as above using additional forward and reverse primers hybridizing with the mtDNA sequence at positions 798 to 817 and 1344 to 1325 or 4875 to 4895 and 5117 to 5097, corresponding to 12 S RNA or ND2 fragments, respectively (6). The hybridization temperatures were 61 and 55°C, respectively.
Immunodetection of F1-ATPase ␤ Subunit and Cytochrome Oxidase Subunit II and Subunit IV-Each cell type was cultured in an 80-cm 2 flask until 70% confluence. They were treated with trypsin and washed in phosphate saline buffer (10 mM sodium phosphate, 150 mM NaCl, pH 7.2). After centrifugation at 100 ϫ g for 5 min, the cell pellets, homogenized at 0 -4°C in 0.3 ml of buffer containing 10 mM Tris-HCl, 0.2 mM dithiothreitol, 1 mM EDTA, and 2 mM ⑀-aminocaproate, pH 7.5, were sonicated by five bursts of 5 s separated by 1 s of cooling using a Vibra cell™ 72405 equipped with a 0.12-inch probe. Protein concentrations of the homogenate were determined with the bicinchoninic acid assay according to manufacturer's instructions (Pierce).
Equivalent amounts of proteins of each cell type (50 g) were separated on a 15% polyacrylamide-SDS gel and transferred to a nitrocellulose membrane at a constant current of 0.8 mA/cm 2 for 1.5 h in a semi-dry apparatus (Hoefer) using 10 mM CAPS, 10% methanol, pH 11, as a transfer buffer. The nitrocellulose membranes were then saturated for 1 h in TBS containing 10% skimmed milk. To detect the different proteins, the membranes were first incubated for 1 h with a primary monoclonal antibody diluted in TBS containing 0.1% Tween 20 (v/v) and then for 1 h with peroxidase-labeled anti-mouse antibody (Biosys) diluted at 1:2000 in TBS containing 0.1% Tween 20 (v/v). The peroxidase activity was finally revealed with the ECL™ reagent according to the manufacturer's instructions (Amersham Pharmacia Biotech). Between each incubation, the membranes were washed once for 10 min and twice for 5 min in TBS containing 0.1% Tween 20 (v/v). The primary antibodies used were anti-F1-ATPase ␣ subunit, 20D6, (14) diluted 1:5000; anti-F1-ATPase ␤ subunit, 14D5, (14) diluted 1:4000; anti-COX II (15) diluted 1:5000, and anti-COX IV (15) diluted 1:2000.
Mitochondrial ATPase Activity Measurement-Cells were sonicated, and the protein concentrations were measured as described for immunological studies. The rate of ATP hydrolysis was measured at 37°C according to Pullman et al. (18) by adding the cell homogenate (25 to 100 g of proteins) to 660 l of reaction buffer containing 50 mM Tris-Hepes, pH 8.0, 3.3 mM MgSO 4 , 4 mM phosphoenolpyruvate, 0.33 mM NADH, 3.3 mM ATP, 10 g of lactate dehydrogenase, 50 g of pyruvate kinase, and 1 g of rotenone (to inhibit NADH oxidation by the mitochondrial NADH-ubiquinone oxidoreductase). The assay was performed in the presence or absence of one of the mitochondrial ATPase inhibitors, 3.5 M oligomycin, 60 M aurovertin B, or 2 mM sodium azide, to estimate the percentage of ATPase activity that was related to the F1 or F0F1 complex in the cell homogenate.
Cell Growth Assays-To determine the effects of aurovertin and bongkrekic acid on cellular growth, the cells were incubated into 24well culture plates at a density varying between 2.5 ϫ 10 3 to 10 4 cells/well. After 24 or 48 h, the culture medium was changed, and the tested drugs were added: 3 nM to 60 M aurovertin and/or 2-10 M bongkrekic acid or 0.1-0.5 M FCCP. The cells were grown for 5 days with a culture medium change after 3 days, when indicated. The cells released by trypsin treatment were counted by trypan blue exclusion.
Determination of Lactate Production in Culture Medium-The culture medium was collected after 5 days of cell treatment with bongkrekic acid. Proteins were precipitated with 7.5% trichloracetic acid. The assays were centrifuged for 7 min at 1,000 ϫ g. The supernatant fraction was extracted four times with an equal volume of diethyl ether. The lactate concentration in the deproteinized samples was estimated spectrophotometrically at 340 nm by using lactate dehydrogenase (19).
Fluorescence of NAO and R123 in ϩ and °Cells-The fluorescence of NAO, which is reputed to be proportional to the amount of cardiolipin and independent of the mitochondrial membrane potential, was tested to compare the amount of mitochondrial membranes in °and ϩ cells (20). R123 fluorescence was used to estimate the mitochondrial membrane potential (21). Each cell type was distributed into 96-well culture plates at a density of 1 to 5 ϫ 10 4 cells/well and incubated 24 to 48 h before fluorescence measurement. Triplicates of each cell dilution were treated with or without 30 M aurovertin and/or 10 M bongkrekic acid for 30 min to 24 h before R123 fluorescence measurement. FCCP (0.1 M) was used in some experiments to determine the residual fluorescence intensity when the mitochondrial membrane potential was collapsed. The cells were washed once with Hanks' balanced salt solution and incubated for 30 min at 37°C (without CO 2 ) with 0.1 ml of 6.3 M NAO or at 37°C in a humidified atmosphere containing 5% CO 2 with 1 or 10 M R123. The cells were then washed twice with Hanks' balanced salt solution, and the medium was removed. NAO and R123 fluorescence were rapidly measured in a microplate fluorescence reader (Victor) with excitation at 485 nm and emission at 535 nm (21). It was checked that, under the tested conditions, R123 fluorescence was not modified by the presence of aurovertin, bongkrekic acid, or FCCP. A linear correlation made between the number of cells counted by trypan blue exclusion and crystal violet staining (22) permitted an estimate of the number of cells in all wells.

RESULTS
Characterization of °Cells-°HeLa S3 cell growth was dependent on the presence of pyruvate and uridine, as shown previously for 143B cells (13). The doubling times for °and ϩ HeLa S3 cells were 51 and 25 h, respectively, and that of °and ϩ 143B cells, 29 and 20 h. In the °cells, no full-size mtDNA could be detected by Southern blotting. However, although the cytochrome b fragment could not be amplified by PCR, a fragment corresponding to the 12 S RNA could be amplified in the two °cell types. This came from sequences integrated into the nuclear DNA, as shown previously (23). The °cells were devoid of functional mtDNA since no mitochondrial mRNA could be revealed by reverse transcription-PCR (data not shown). The immunoblot obtained after SDS-polyacrylamide gel electrophoresis of cellular proteins transferred to nitrocellulose, incubated with antibodies, and stained (Fig. 1A) demonstrates that, as expected, the mitochondrially encoded COX II was absent from the two types of °cells and present in ϩ cells. On the contrary, the nuclearly encoded COX IV and F1-ATPase ␤ subunits were expressed both in ϩ and °cells. The expression of COX IV was slightly lower in °than in ϩ cells, whereas that of the F1-ATPase ␤ subunit was similar in both types of ϩ and °cells. In a parallel experiment, it was shown that the F1-ATPase ␣ subunit was also expressed similarly in all cell types (data not shown).
ATPase Activity in °Cells-The ATPase activity was tested in °and ϩ cell homogenates. To differentiate the part of this activity that was due to the mitochondria from that originated from other cellular ATPases, inhibitors specific to the mitochondrial ATPase were added. The difference between the total activity and that obtained in the presence of the inhibitors corresponds to the mitochondrial F0F1 activity. Fig. 1B shows that oligomycin, which binds to F0 (24), did not inhibit the ATPase activity of °cells devoid of the mitochondrially coded F0 subunits 6 and 8. However, oligomycin inhibited the ATPase activity of ϩ HeLa S3 cells by 35% and that of ϩ 143B cells by 50%. On the contrary, inhibitors such as azide (25) and aurovertin (26), which bind to F1, decreased the rate of ATP hydrolysis both in °and ϩ cell extracts. Aurovertin was the most efficient inhibitor, decreasing ATP hydrolysis by 55% in ϩ and °HeLa S3 cells and by 65% in ϩ and °143B cells. In addition, to test whether this aurovertin-sensitive F1-ATPase activity depended on soluble F1 or on partly assembled F0F1 complex, its cold sensitivity was studied. Fig. 1, C and D, show that the °cell ATPase activity decreased much more rapidly at 4°C than that of ϩ cells and than that studied after incubation of °or ϩ cells at 30°C. Within 1 h at 4°C, 50 or 60% of this activity was lost for the °HeLa S3 or °143B cells, respectively, whereas at 30°C, it decreased by about 10% for the °H eLa S3 cells and was almost stable for the °143B cells. The ATPase activity was also stable for the ϩ 143B cells at both temperatures, whereas it decreased for the first h by 25 and 10% in ϩ HeLa S3 cells at 4 and 30°C, respectively.
Effects of Aurovertin on °and ϩ Cell Growth-To determine whether the functional F1-ATPase activity could play a role in the survival of °cells, the effects of 3 nM-30 M aurovertin B were tested on the growth of °and ϩ cells. The percentage of cell growth after 5 days of aurovertin treatment was compared with control cells. The sensitivity to aurovertin was higher in ϩ than in °cells, since 50% of cell survival was obtained at about 50 nM aurovertin for ϩ HeLa S3 cells and about 30 M for °HeLa S3 cells or between 5 and 10 M for ϩ 143B cells and above 30 M for °143B cells (not shown). Therefore, the aurovertin-induced inhibition of the mitochondrial F1-ATPase activity decreased cell growth in °cells as well as in the parental ϩ cells, although the aurovertin concentration exhibiting the same effect was higher in °than in ϩ cells. In addition, the aurovertin concentration necessary to inhibit the 143B cell growth was higher than that inhibiting the HeLa S3 cells.
Mitochondrial Membrane Potential in °and ϩ Cells-The green fluorescence of R123 was measured on both °and ϩ cell types to compare their mitochondrial membrane potential. The green fluorescence of NAO was measured in parallel assays to estimate the amount of mitochondrial membranes. Fig. 2 shows that the NAO fluorescence was similar in °and ϩ HeLa S3 cells ( Fig. 2A) and was slightly lower in °143B than in ϩ 143B cells (Fig. 2B). R123 fluorescence intensity shows that the mitochondrial membrane potential was of the same order of magnitude in °HeLa S3 and ϩ HeLa S3 cells ( Fig.  2A). In °143B cells, it was much lower than in ϩ 143B cells, but the difference was much more important than that for NAO (Fig. 2B). In the case of HeLa S3 cells, a 30-min treatment with 0.1 M FCCP strongly decreased the mitochondrial membrane potential of both °and ϩ cells to a similar basal level. In the case of 143B cells, FCCP strongly decreased ϩ mitochondrial membrane potential, but its effect was less important with °1 43B cells. However, the R123 fluorescence observed in the presence of FCCP reached a similar low level in °and ϩ 143B cells. Therefore, in the absence of mitochondrial membrane potential, the fluorescence intensity obtained with the uncoupler FCCP seems to correspond to the basal fluorescence of R123 in the cells. Increasing the R123 concentration from 1 to 10 M did not change the relative fluorescence intensity observed in any cell type. In conclusion, the mitochondrial membrane potential, as estimated by R123 fluorescence, is similar in °and ϩ HeLa cells, whereas that observed in °143B cells is only 10 to 20% that of the ϩ 143B cells. Because of the low mitochondrial membrane potential of °143B cells, the following studies involving mitochondrial membrane potential were conducted with HeLa S3 cells.
Effect of Aurovertin on °and ϩ HeLa S3 Cell Mitochondrial Membrane Potential-The effects of 30 M aurovertin were tested on the mitochondrial membrane potential of °and ϩ HeLa S3 cells, as measured by R123 fluorescence. Fig. 3 shows that in ϩ and °HeLa S3, a significant decrease of mitochondrial membrane potential was observed after a 30-min treat-ment. Similar results were obtained with a 6 h-treatment (not shown). Since the cell number was not modified after a 30-min or a 6-h aurovertin treatment, the aurovertin effect on mitochondrial membrane potential is not due to a lower cell number.
Effects of Bongkrekic Acid on Cell Growth and Mitochondrial Membrane Potential-Since the respiratory chain is not functional in °cells, the membrane potential must be set up by a mechanism independent of this chain. The ANT, which exchanges ADP 3Ϫ against ATP 4Ϫ between the two faces of the inner membrane, is likely to participate in the maintenance of this potential, as suggested for °yeast cells (8). To determine whether the role played by the F1-ATPase could be mediated via the ANT, the effect of bongkrekic acid was tested on the growth and the mitochondrial membrane potential of both °a nd ϩ cells. Fig. 4A shows that bongkrekic acid modified neither the ϩ 143B cells growth nor that of the °143B cells after 5 days of treatment even at the highest tested concentration (10 M). On the contrary, an inhibition of about 50% °H ela S3 cell growth was observed after 5 days of treatment with 10 M bongkrekic acid. Moreover, the inhibition could already be detected after a 3-day treatment (data not shown). In the case of the ϩ HeLa S3 cells, no effect was observed before 5 days. At that time, 10 M bongkrekic acid inhibited the cell growth by about 30%, but a culture medium acidification was observed. If this medium was replaced after 3 days, the growth inhibition was limited to 15%. Since, in the presence of bongkrekic acid, the ATP produced by oxidative phosphorylation cannot reach the cytoplasm, the ϩ cell must then reconstitute its ATP pool via glycolysis. In this case, the NADH produced during glycolysis must be re-oxidized, and lactate should be accumulated. To verify this hypothesis, the lactate concentration was measured in ϩ and °HeLa S3 cells (Fig.  4B). In the absence of bongkrekic acid, the lactate concentration (calculated per cell) was 6 times higher in °than in ϩ cells, as expected since in °cells the ATP production completely relies on glycolysis. The lactate concentration strongly increased in ϩ HeLa S3 cells treated with 10 M bongkrekic acid, even if the number of cells was reduced. On the contrary, the lactate concentration decreased in °HeLa S3 cells treated with 10 M bongkrekic acid. However, the number of °cells was simultaneously reduced. If the lactate amount was compared with the number of cells, it increased about 300% in ϩ cells and only 30% in °cells.
Bongkrekic acid (10 M) did not modify the mitochondrial membrane potential of ϩ HeLa S3 cells either after a 30-min treatment (Fig. 5) or after a 22-h treatment (data not shown). On the contrary, in °HeLa S3 cells, the mitochondrial mem- brane potential was reduced after a 30-min treatment as well as after a 22-h treatment.

Cooperative Effects of Aurovertin and Bongkrekic Acid Treatments on °HeLa S3 Growth and Mitochondrial Membrane
Potential-To verify whether F1-ATPase and ANT act together or in two independent pathways to maintain the °HeLa S3 cell growth and mitochondrial membrane potential, the effects of aurovertin and bongkrekic acid were tested simultaneously (Fig. 6). The presence of either 30 M aurovertin or 10 M bongkrekic acid decreased the cell growth by about 50% after a 5-day treatment, whereas the simultaneous addition of bongkrekic acid and aurovertin decreased it by up to 85% that of the control value. The °cell growth decrease observed with aurovertin or bongkrekic acid was less extensive than that observed with 0.1 or 0.5 M FCCP (decrease of 92% of the control value). Similar to the effects observed on cell growth, the mitochondrial membrane potential was decreased more extensively in the presence of aurovertin and bongkrekic acid than when each drug was tested separately. When aurovertin concentration was increased up to 60 M, no additional effect was observed either on cell growth or on mitochondrial membrane potential.

DISCUSSION
First, our data demonstrate that °cells devoid of the F0F1-ATPase subunits 6 and 8 contain an active F1-ATPase essential for their growth. The size and amount of F1 ␣ and ␤ subunits were similar in °and ϩ cells. Therefore, the ␣ and ␤ subunit precursors were imported in °cells, and their signal sequences were normally processed in the mitochondria. Previous studies have already shown that the °143B cells contained the same amount of F1 ␤ subunit mRNA as that of the parent ϩ cells, whereas the relative abundance of some other transcripts encoding mitochondrial proteins such as cytochrome c, cytochrome oxidase subunit IV and VIaL, and ANT 2 and ANT 3 were slightly more abundant (27). Doxycline-induced inhibition of mitochondrial protein synthesis in human leukemia cells also decreased the contents of complex III or complex IV subunits of nuclear origin without modifying that of the F1-ATPase ␣ and ␤ subunits during two culture generations (10). The authors suggest that respiratory chain complex subunits of nuclear origin were rapidly degraded when they were not assembled. In the case of complex III subunits, the partial assembly of sub-complexes has been demonstrated in yeast mitochondria (28). In the two human °cell lines studied here, the cytochrome oxidase subunit IV amount was only slightly decreased after many culture generations. This suggests that at least some of the cytochrome oxidase subunits of nuclear origin can be imported without being degraded. In the case of the ATPase, the normal amount of the ␣ and ␤ subunits suggested that F1 could be assembled and remain stable.
The ATPase activity measured in the human °cells was due to this mitochondrial F1-ATPase. Indeed, it was aurovertinsensitive (26). On the contrary, since oligomycin sensitivity (24) depends on mtDNA-encoded subunit 6 (29), only ϩ cells were inhibited by oligomycin. Therefore, F1 was assembled and active in human °cells despite the absence of subunits 6 and 8. It was, however, not efficiently associated with the subunits constituting the stalk linking F1 to F0 since the ATPase activity was cold-labile. Indeed, it has been shown in beef heart F1 that the binding of isolated F1 to OSCP (30,12) or to other subunits constituting the stalk such as subunits b, F6, and d (12) decreases F1 sensitivity to cold exposure. Therefore, the cold-sensitivity of the F1-ATPase in human °cells indicates that at least some of the stalk subunits are deficient. However, F1 might be bound to the membrane. Indeed, a binding site for F1 on F0 in the absence of F6 and OSCP have been disclosed in the beef enzyme (31).
A similar assembly of the F1 subunits in a cold-labile and oligomycin-insensitive H ϩ -ATPase has been reported in yeast °cells. In addition, this F1 H ϩ -ATPase activity was modulated by the insaturation degree of the mitochondrial membrane phospholipids, and therefore F1 was membrane-bound (32). In yeast, the mtDNA codes not only for the subunits 6 and 8 as in human but also for the DCCD-binding protein. Therefore, a fortiori, if F1 was membrane-bound in yeast, this must also be true in human °cells that might contain the DCCD-binding protein. Although the pathway of the F0F1 assembly has not been extensively studied in mammals, many experiments have been performed using yeast mutants lacking one or several F0F1 subunits. It was reported that, in yeast, subunit 6 was not essential for the binding of F1 subunits to components of the F0 factor (33). The assembly of F0F1 subunits involved the sequential addition of subunits 9 (DCCD-binding protein), 8, and 6 to a membrane-bound F1, and two other proteins, subunits b and OSCP, were not found in the complex when the F0 sector was not properly assembled (34,35). However, the b subunit could be bound to a subunit 6-deficient mutant (36), and a tight binding of F1 to the membrane required the pres-ence of the d subunit (37). If the same assembly pathway is true for human °cells, F1 should bind to the membrane after insertion of the d subunit, and the presence of F1 would permit the insertion of the DCCD-binding protein and eventually that of the b subunit. OSCP, if present, might not be efficiently associated with the membrane since the subunit 6 is absent. This would explain the cold sensitivity of the human °cell ATPase activity.
Interestingly, this ATPase activity was essential to maintain the human °cell growth. Indeed, aurovertin-induced inhibition of the F1-ATPase activity strongly decreased °cell growth. The effect was, however, lesser for the 143B cell lines than for the HeLa S3 cell lines, whereas the inhibition of the ATPase activity by aurovertin was of the same order of magnitude in both cell lines. This is probably related to metabolic adaptation of these two types of cancer cells. It cannot be excluded that, in the 143B cells, some multidrug resistance could have been developed (cf. Ref. 38, for review). This could reduce the intracellular steady-state levels of various drugs added to the extracellular medium. Indeed, the cell growth sensitivity to aurovertin or to bongkrekic acid was lower in 143B than in HeLa S3 cells. However, aurovertin decreased cell growth in both °cell lines. The sensitivity to aurovertin was higher in ϩ than in °cells. This is probably due to the fact that, during the ethidium bromide treatment used to transform the ϩ into °cells, the cells must have induced an alternative pathway to improve their ATP production via the glycolysis. This is demonstrated by the higher lactate production observed in °than in ϩ cells in the absence of inhibitor. In °cells, oxidative phosphorylation cannot produce ATP to sustain cell growth, and therefore, the glycolytic flux must be increased to provide the ATP necessary to meet the cell energy demands. Simultaneously, the NADH produced during glycolysis must be reoxidized via, for example, the lactate dehydrogenase, which increases lactate concentration.
In °HeLa S3 cells, the mitochondrial membrane potential, as measured by R123 fluorescence, reached the same level as that of ϩ HeLa S3 cells. The existence of a normal mitochondrial membrane potential in other °cells has already been observed previously. Rat hepatoma °cells (9) as well as human °HeLa cells (39) took up R123 to the same extent as the parent cells. On the contrary, °143B cells show R123 fluorescence intensity, which is only 10 to 20% that of ϩ 143B cells. This effect could be related to an adaptive difference between °H eLa S3 cells and °143B cells. It is possible that the °143B cells effectively possess only a low mitochondrial membrane potential, which would, however, be sufficient for their growth. In such a case, a mitochondrial membrane potential corresponding to only 10 to 20% of the normal mitochondrial membrane potential would be large enough to maintain cell viability and, hence, to sustain protein import into the mitochondria. The low mitochondrial membrane potential could explain why the °143B cells are easily damaged; for example, contrary to the °HeLa S3 cells, which are as resistant as the ϩ cells, °1 43B cells cannot survive a mild centrifugation. One cannot however exclude that the 143B cells developed a secondary resistance to drugs during the ethidium bromide treatment used to produce the °cells from the ϩ ones. It would mean that the mitochondrial membrane potential measured in these °143B cells excluding R123 could be artificially lower than it is in reality.
Aurovertin simultaneously reduced the °HeLa S3 cell growth rate and their mitochondrial membrane potential estimated by R123 fluorescence, as it does for ϩ HeLa S3 cells. The mitochondrial membrane potential, as well as the growth of °HeLa S3 cells was also reduced by bongkrekic acid. On the contrary, bongkrekic acid did not modify the ϩ HeLa S3 mitochondrial membrane potential and barely decreased their growth under the tested conditions. However, bongkrekic acid efficiently inhibited ANT in the ϩ cells since bongkrekic acid induced lactate overproduction and therefore shifted energy production from oxidative phosphorylation to glycolysis. The simplest explanation for the simultaneous sensitivity of the °H eLa S3 to bongkrekic acid and aurovertin is that, in °cells, the mitochondrial membrane potential was generated by the ANT exchanging ATP 4Ϫ for ADP 3Ϫ produced by an active F1. The presence of aurovertin reduced the rate of ADP 3Ϫ production in the mitochondrial matrix, leading to a decrease in the nucleotide exchange rate, which became insufficient to maintain the high mitochondrial membrane potential necessary for an efficient cell growth. However, the inhibition of one or the other partner could only produce a partial inhibition. When F1 was inhibited, ATP could be hydrolyzed by other mitochondrial ATP-consuming processes, as for example, those involved in the import of mitochondrial proteins of nuclear origin (3). When the ANT was inhibited, other electrogenic pumps, such as that involving PPi (40), could in part maintain this membrane potential whether or not calcium movements also participate in this mechanism (41). Our observations are consistent with those reported by Giraud and Velours (42) showing that Ϫ Saccharomyces cerevisiae cells lacking the F1 ␦ subunit exhibit a slow growth phenotype and a membrane potential decrease comparable with that described here for aurovertin-treated °c ells. In conclusion, an active F1-ATPase is mandatory for the maintenance of the mitochondrial membrane potential essential for human °cell growth.