Structure, functioning, and assembly of the ATP synthase in cells from patients with the T8993G mitochondrial DNA mutation. Comparison with the enzyme in Rho(0) cells completely lacking mtdna.

The structure and functioning of the ATP synthase of human fibroblast cell lines with 91 and 100%, respectively, of the T8993G mutation have been studied, with MRC5 human fibroblasts and Rho(0) cells derived from this cell line as controls. ATP hydrolysis was normal but ATP synthesis was reduced by 60% in the 100% mutants. Both activities were highly oligomycin-sensitive. The levels of F(1)F(0) were close to normal, and the enzyme was stable. It is concluded that the loss of ATP synthesis is because of disruption of the proton translocation step within the F(0) part. This is supported by membrane potential measurements using the dye JC-1. Cells with a 91% mutation load grew well and showed only a 25% loss in ATP synthesis. This much reduced effect for only a 9% difference in mutation load mirrors the reduced pathogenicity in patients. F(1)F(0) has been purified for the first time from human cell lines. A partial complex was obtained from Rho(0) cells containing the F(1) subunits associated with several stalk, as well as F(0) subunits, including oligomycin sensitivity conferring protein, b, and c subunits. This partial complex no longer binds inhibitor protein.

An F 1 F 0 type ATP synthase is involved in oxidative and photosynthetic ATP synthesis in prokaryotes and eukaryotes. Recent electron microscopy (1), crystallography (2), fluorescence (3), cross-linking (4,5), and single molecule video microscopy studies (6) show that the F 1 F 0 complex is a molecular motor. In the emerging view of the structure and function of the enzyme, it is thought that the well-known central stalk connecting F 1 to F 0 is a rotor that moves in relation to a stator holding the catalytic ␣/␤ pairs (reviewed in Refs. 7 and 8). The stator is probably the second stalk observed recently at the periphery of the F 1 F 0 complex (9). Key subunits in the rotor are the ␥ and ⑀ subunits of F 1 and the c 12 subunit ring of F 0 (10). The stator contains the a and b 2 subunits of F 0 and the ␦ subunit (Escherichia coli nomenclature) attached at the top of the three ␣/␤ pairs (11,12). Eukaryotic F 1 F 0 -ATPases have up to eight additional subunits when compared with the bacterial enzyme, most of which contribute to the stalks or F 0 parts (13,14).
The biogenesis and assembly of eukaryotic F 1 F 0 is compli-cated because of the presence of polypeptides encoded in two separate genomes with synthesis of these occurring in two compartments, the cytosol and the mitochondrion. In mammals, subunit 6 (or a in E. coli) and A6L are encoded in the mitochondrial genome. In yeast, subunit 9 (also called subunit c) is mitochondrially encoded (15,16). Assembly of the F 1 part occurs in the absence of some or all of the F 0 subunits, as revealed by studies of Rho 0 cells in Saccharomyces cerevisiae (17), Kluyveromyces lactis (18), and human cell lines (19). A growing number of human diseases are now known that are because of a deficiency of mitochondrial function, and many are the result of mutations of mtDNA. A particularly interesting mutation of subunit 6 is the human T8993G mutation, by which Leu-156 is changed to Arg. Depending on the percentage of mtDNA molecules with this mutation, patients present with neurogenic muscle weakness, atraxia, and retinitis pigmentosa or Leigh's Syndrome (above 95% mutation) (20).
Studies of the functioning of the ATP synthase in cell lines from patients with the T8993G mutation have established that ATP synthesis is decreased (21)(22)(23)(24), but an unambiguous distinction between effects on assembly and on enzyme turnover has not been made. Here, we have examined the ATP synthase in mitochondria from fibroblasts of patients with 91 and 100%, respectively, of the T8993G mutation. Activity measurements are reported and the levels of assembled enzyme in mutant and Rho 0 ATP synthase have been measured. The ATP synthase has been isolated from cell cultures for the first time. Finally, the role of the inhibitor protein in determining the properties of mutant and Rho 0 enzymes is described.

MATERIALS AND METHODS
MRC5 Fibroblasts-MRC5 fibroblasts were obtained from the American Type Culture Collection. The population doubling of the cells was in the range of 30 -45 before harvesting to isolated mitochondria. To grow Rho 0 cells, MRC5 cells (population doubling ϭ 28 -30) were cultured continuously for a further 14 -20 population doublings in the presence of 50 ng/ml ethidium bromide. The T8993G mutant-containing fibroblast cell lines were from two patients. Both were diagnosed with Leigh's Disease in the clinic of the Hospital for Sick Children, Toronto, Canada. The levels of mutation were established by standard protocols (20). All cells were grown as described before (25) in high glucose Dulbecco's modified Eagle's medium, supplemented with 10% bovine calf serum, 50 g/ml uridine, 110 g/ml pyruvate, and 10 mM HEPES buffer to maximize growth rates.
Isolation of Mitochondria from Fibroblast Cell Lines-Mitochondria were isolated from cells according to standard procedures (26) with the modifications described below. Cells (ϳ5 ϫ 10 7 ) were washed three times with 40 ml of Ca 2ϩ -Mg 2ϩ -free phosphate-buffered saline and then suspended in 5 ml of 0.25 M sucrose, 1 mM EGTA, 10 mM HEPES/NaOH, 0.5% bovine serum albumin, (pH 7.4), containing 0.5 l/mg pepstatin, 0.5 g/ml leupeptin, and 1 mM phenylmethylsulfonyl fluoride. This suspension was homogenized with 15 up and down strokes in a rotating glass Teflon homogenizer and then centrifuged at 1500 ϫ g for 10 min at 4°C. This homogenization procedure was repeated three times, and the supernatants were pooled and centrifuged at 1500 ϫ g to remove * 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 U.S.C. Section 1734 solely to indicate this fact. any pellet. A crude mitochondrial pellet was collected by centrifugation at 10,000 ϫ g for 12 min at 4°C. This pellet was washed once in 5 ml of washing buffer (10 mM Tris/HCl, 1 mM EDTA, 0.25 M sucrose, pH 7.5) containing protease inhibitors and then resuspended in 200 -500 l of the same buffer for storage at Ϫ80°C in small aliquots to avoid freezethawing cycles.
Isolation of F 1 F 0 -IF 1 Complex from Human Mitochondria-The affinity column procedure of Vá zquez-Contreras et al. (27) was adapted to isolate F 1 F 0 from fibroblast cell lines as follows. A crude mitoplast fraction was used as starting material to avoid release of F 1 by the sonication step used to make submitochondrial particles in the previous procedure. These were prepared by treating the isolated mitochondria with digitonin (2.5 mg/mg protein) in 0.15 M sucrose, 10 mM Tris/HCl, pH 7.5. This level of detergent was found to release outer membrane proteins (e.g. porin) but not F 1 in initial checks by Western blotting with protein-specific monoclonal antibodies. Mitoplast fractions were diluted 40-fold in distilled water containing 1 mM EDTA in addition to protease inhibitors as above. This osmotic shock treatment released mitochondrial matrix proteins, leaving a crude fraction of inner membrane, that was collected at 10,000 ϫ g for 15 min at 25°C. The membrane pellets were dissolved in ADP medium (150 mM sucrose, 20 mM MES-Tris, 1 1 mM ADP, pH 6.8), and then diluted 1:1 with dodecyl-␤-D-maltoside (3 mg/mg protein final concentration). When [ 14 C]DCCD was used to label the c subunit of F 1 F 0 , it was added at a concentration of 80 M and incubated for 4 h before the addition of the detergent. A precipitate was removed by centrifugation at 170,000 ϫ g for 20 min, and the supernatant was applied to the Sepharose-EAH column (1.5 ϫ 1 cm) preequilibrated in "column buffer" (20 mM MES-Tris, 2 mM EDTA, 10 mM sodium cholate, 5 mM ADP, 3 mg/ml asolectin, pH 6.8). The column was washed with 10 -15 ml of the column buffer, and enzyme eluted with 1 M KCl were added to the same buffer. Peak fractions were collected, and the protein precipitated with 45% ammonium sulfate at 4°C for 1 h.
Assay of ATP Hydrolysis and ATP Synthesis-ATPase activities were measured with an ATP regenerating system (28) at 37°C in the presence of rotenone (2 M) and potassium cyanide (5 mM) to inhibit enzymatic oxidation of NADH. Sodium sulfite (50 mM) was also included to avoid the formation of the ADP-Mg 2ϩ inhibited state of the enzyme (29). Sensitivity to 5 mM sodium azide was tested by adding the inhibitor directly to the cuvette. Oligomycin (10 g/mg protein) sensitivity was measured by preincubating the samples for 10 -15 min with oligomycin, at room temperature before taking the activity measurement. Activation of ATPase, i.e. release of control by the inhibitor protein (IF 1 ), was achieved by diluting the samples 4-fold before incubating for 2 h at 37°C in a medium containing 250 mM KCl, 30 mM Tris-sulfate, 2 mM EDTA, 75 mM sucrose, pH 8. 0.
ATP synthase activity was measured as follows: 100 g of mitochondria were incubated in 20 mM Tris/HCl, 0.15 M sucrose, 1 mM ADP, 20 mM P i , 5 mM MgCl 2 , 20 mM Tris/HCl, 100 M diadenosine pentaphosphate to inhibit adenylate kinase, 10 mM glucose, 1 mg/ml hexokinase (Sigma, type F-300), pH 7.5, including protease inhibitors. Samples (0.45 ml) were incubated at 37°C with vigorous stirring in 25-ml Erlenmeyer flasks for maximal oxygenation. Reactions were started with 50 mM succinate and arrested at desired times with 25 mM EDTA, followed by transfer to ice-cold water. Samples were transferred to Eppendorf tubes, boiled for 5 min, and centrifuged (10,000 ϫ g, 20 min) to remove denatured protein. In supernatants, the synthesized glucose 6-phosphate, which is heat-stable, was oxidized by NADP in the presence of 30 units of glucose-6-phosphate dehydrogenase. NADPH formation was monitored at 340 nm until completion. ATP formation was calculated after correction for trapping efficiency of hexokinase (75-95%), which was determined by adding known amounts of ATP to the reaction mixture in the absence of succinate. The rates of ATP synthesis reported are at early reaction times (up to 30 min), i.e. to avoid substrate depletion.
Other Methods-Protein was measured using the BCA assay of Pierce. Protein was first trichloroacetic acid-precipitated according to a modified Lowry method (30) to prevent interference of any component of the samples with the BCA assay. Gel electrophoresis was made according to Laemmli (31) in a 10 -22% gradient. For Western blotting, gels were transferred to polyvinylidene difluoride-P SQ membranes (0.2-M pore size, Millipore) for 1 h at 100 V. Reactive bands were detected with the Femto lucent kit of Chemicon, according to the specifications of the product. Anti-␣-monoclonal antibodies were prepared and character-ized in this laboratory and will be described elsewhere. Rat liver inhibitor protein was expressed and purified from E. coli. as described by Lebowitz and Pedersen (32).

RESULTS
Functional Parameters of Fibroblasts with the T8993G Mutation-Two different fibroblast cell lines containing the T8993G mutation were examined in this study, one with 91% of mutant mtDNA and the second with 100%, i.e. homoplasmic for the mutation. MRC5 fibroblasts were used as one control, whereas Rho 0 cells from this MRC5 cell line were used to examine the effect of a complete absence of the ATP synthase subunit 6 and A6L. All cell lines were grown in high glucose and uridine. Under these conditions, the population doubling times of the MRC5 control fibroblasts was 4 days, that of the 91% T8993G mutant was 5 days, Rho 0 MRC5 cells were 6 -8 days, whereas that of the 100% T8993G mutant was 14 -16 days.
The ATPase activities of mitochondria isolated from different cell lines are summarized in Fig. 1A. The 91% mutant showed somewhat higher and the 100% mutant slightly lower rates of ATP hydrolysis than control MRC5 cells. This activity in both mutant cell lines was highly oligomycin-sensitive. In contrast, the ATPase activity of Rho 0 cells was nearly 2-fold higher than the control and there was essentially no oligomycin sensitivity FIG. 1. ATPase and ATP synthase activities of control, T8993G, and Rho 0 human mitochondria. A, ATPase activities of mitochondrial membranes were measured as described under "Materials and Methods" after exposing the mitochondria to an osmotic shock and collecting membranes. B, ATP synthase activities were measured with whole mitochondria as described under "Materials and Methods." An average of three to six different determinations is shown together with its standard deviation. Control, T8993G 91%, and T8993G 100% activities correspond to white, gray, and black bars, respectively. The data for Rho 0 cells are at the right of the figure. of the enzyme. ATP synthesis rates of mitochondria from each of the four cell lines are presented in Fig. 1B. The T8993G mutation in 91% of mtDNA copies gave a drop of around 25% in the rate of ATP synthesis. At 100% of the mutation, ATP synthesis rates were only 40% of controls. There was negligible ATP synthesis of mitochondria from Rho 0 cells. ATP synthesis in the T8993G mutants was highly oligomycin-sensitive.
In one set of experiments, the morphology of the different fibroblast cell lines and the ability of their mitochondria to generate a membrane potential was monitored by fluorescence microscopy after adding the dye JC-1 (Fig. 2). This dye is accumulated in mitochondria in response to the membrane potential of the organelle. Regions of high membrane potential show up as yellow/orange, whereas areas of lower membrane potential are green (see Ref. 33 for the properties of JC-1). It can be seen that the morphology of cells homoplasmic for the T8993G mutation is different from that of control MRC5 cells or either the 91% mutant or Rho 0 cells (not shown). The 100% cells were much larger; they were also flatter and stuck to the cell culture dishes much more strongly than more normal fibroblast cell lines. In MRC5 cells, there were regions of high membrane potential revealed by the regions of yellow/orange fluorescence. The addition of oligomycin greatly enhanced the number and total area of these regions, as expected if blocking ATP synthesis slowed the dissipation of pH gradient and associated ⌬ generated by the respiratory chain. In the 100% mutant (but not in the 91% mutant), there were significantly more areas of high fluorescence, and oligomycin did not greatly increase these levels. The mitochondria of Rho 0 cells also showed regions of high membrane potential, which were more punctuate than in wild type or T8993G mutants. The extent of the regions of high ⌬ was much lower than in the 100% mutant and did not increase upon the addition of oligomycin (result not shown).
Assembly and Stability of F 1 F 0 from Western Blotting Studies-The similarity in levels of oligomycin-sensitive ATPase activity between the T8993G cells and control MRC5 cells is good evidence that a functional F 1 F 0 is assembled in near normal amounts in the mutant cell lines. To confirm this, the amount of the ␣ subunit was compared with that of the sub-units of succinate dehydrogenase and cytochrome-c oxidase for each cell line over a range of protein concentrations (1-5 g) using Western blotting. As shown in Fig. 3, A and B, the levels of the F 1 subunits in the T8993G 91% cell lines were similar to those of controls, whereas the levels in the 100% mutant and Rho 0 cell lines were, if anything, somewhat greater than those found in the control MRC5 cells. To test the stability of the assembled enzyme, release of F 1 was measured by the presence of ␣ subunit in the supernatant of the osmotic shock treatment. There was no significant release of ␣ subunit in either the 91 or 100% mutants. In contrast, F 1 was released in large amounts from Rho 0 cells by this treatment (Fig. 3C). These results show that F 1 is stably bound to F 0 in the T8993G mutant cells but partially dissociated from Rho 0 membranes. This was confirmed in purification studies.
Purification of the F 1 F 0 from Cell Lines-For comparison of subunit structure, F 1 F 0 , from different cell lines, was purified by an affinity column method first devised by Tuena de Gómez-Puyou and Gómez-Puyou (34). 4 -6 mg of purified mitochondria was required for the purification procedure and, therefore, the slow growth and consequent low yields of the 100% mutant cell line precluded isolation of the enzyme from this source. Mitochondria of each of the cell lines were converted to mitoplasts prior to detergent solubilization of membranes. The yield of F 1 F 0 was 200 -220 g from the control and T8993G 91% mutant, but only 120 -150 g from Rho 0 mitochondria. his was sufficient for studies of both subunit composition and activity effects. Fig. 4 shows a comparison of the subunit compositions of F 1 F 0 isolated from beef heart, MRC5 cells, the 91% mutant, and Rho 0 cell lines. Enzyme from the mutant had the same polypeptide composition as that isolated from control MRC5 fibroblasts, as expected given the functional properties of the enzyme. In contrast, there was only a partial assembly of the F 1 F 0 in the Rho 0 cells, again consistent with the functional assays. Based on the subunit profile (Fig. 4, lane 3), subunits ␣, ␤, and ␥ are present in Rho 0 mitochondria along with OSCP but at lower levels than in the control or mutant cell lines. There are bands on the gel at the position of subunits b and d, but both are in lowered amounts relative to the above mentioned subunits. As expected, subunit a was absent. The presence of OSCP in the F 1 F 0 preparations could be confirmed by Western blotting (Fig. 5A). To examine the presence of subunit c, the isolation of F 1 F 0 was carried out using control and Rho 0 mitochondria after first labeling with [ 14 C]DCCD as described under "Materials and Methods." The enzyme was treated with performic acid before SDS gel electrophoresis to dissolve c subunit aggregates as described by Glaser et al. (35). With this protocol, bovine F 1 F 0 labeled with [ 14 C]DCCD contained several peaks of radioactivity along the gel, but they converged into a single band in the c subunit region after performic acid treatment (Fig. 5, B and C). The same general picture was obtained with control (not shown) and Rho 0 -F 1 F 0 , although the dissolution of c subunit aggregates was only partial in the human samples. These results confirm the presence of subunit c in Rho 0 -F 1 F 0 .
Inhibitor Protein Functioning in Mutant Cell Lines-In eukaryotes, the relative ATP hydrolysis and ATP synthesis rates of F 1 F 0 are under the control of a small polypeptide, the inhibitor protein, or IF 1 (for a review see Ref. 36). The functioning of this protein can be monitored by comparing the ATPase activity of the enzyme at 25°C where IF 1 exerts an effect with that after a 1-h incubation in 250 mM KCl at 40°C, a treatment which disrupts the binding of IF 1 , favoring its release. With the control and both the 91 and 100% mutant mitochondria, there was consistent and reproducible 2-fold activation after IF 1 release (Fig. 6), indicating that this inhibitor affected ATPase activity in the mutant as well as in MRC5 mitochondria. In contrast, the ATPase activity of Rho 0 mitochondria was only modestly increased (1.3-fold). These same effects were observed when purified F 1 F 0 was examined but, in this case, the ATPase activity of Rho 0 -F 1 F 0 was lower after the heat plus KCl treatment than at 25°C, presumably because of partial disassembly of the F 1 and/or denaturation of protein (Fig. 6). Significantly, the ATPase activity of Rho 0 mitochondria at 25°C was similar to that of control mitochondria that had been treated to remove the effect of IF 1 . This suggests that the inhibitor was no longer bound to the partly assembled F 1 F 0 of Rho 0 cells. This idea was tested directly using antisera raised against bovine heart IF 1 and reactive against the human protein.
In line with the activity measurements, it was found that after osmotic shock, control as well as both T8993G 91 and 100% membranes showed similar levels of IF 1 when detected by Western blotting. However, Rho 0 membranes showed very little, or undetectable, levels of the 10 -12-kDa band corresponding to IF 1 (Fig. 5D). The same result was obtained with purified Rho 0 -F 1 F 0 but with lower signal/noise ratio (not shown). Together, these data indicate that the interaction and inhibition by IF 1 are preserved in the T8993G mutant but not in Rho 0 -F 1 F 0 .

DISCUSSION
This study examines the effect of the T8993G mutation on the structure and functioning of the ATP synthase using fibroblast cell lines with 91 and 100%, respectively, of the mutation. MRC5 fibroblasts were used as one control. The properties of the ATP synthase in the mutant were also compared with those of enzyme in Rho 0 cells. Rho 0 cells lack mtDNA and, therefore, do not synthesize subunits 6 and A6L. As a consequence, a   FIG. 3. Assembly of F 1 ␣ subunit in control, T8993G, and Rho 0 mitochondrial membranes as evidenced by Western blotting. Mitochondria were exposed to an osmotic shock as described under "Materials and Methods," membranes were collected by centrifugation, and released soluble proteins were concentrated in Centricon tubes. Membrane proteins (A) were blotted against succinate dehydrogenase (SD70), F 1 ␣, and cytochrome oxidase CoxIV monoclonal antibodies. Lanes 1-4 correspond to 1 g of control, T8993G 91%, T8993G 100%, and Rho 0 membranes. The same order of samples is repeated in lanes 5-8 and 9 -12 but with 2 and 5 g of total protein/lane, respectively. Relative amounts of a/SD70 and a/CoxIV were quantitated as described under "Materials and Methods" (B). C, ␣ subunit in the supernatant after osmotic shock treatment. functional ATP synthase is not assembled. Rho 0 cells are useful models of the state of the ATP synthase in cells showing mitochondrial depletion syndrome (25), a recently described pathology in humans.
Previous studies of cell lines and tissues carrying the T8993G mutation have found that the introduction of Arg for Leu at position 156 in the sequence of subunit 6 diminishes ATP synthesis by around 50% (20 -23). ATP hydrolysis was examined in very few of these studies, and the question of whether the loss of functioning was because of reduced and/or altered assembly versus impaired enzyme turnover/catalysis has been addressed only once. Houstek et al. (37) studied several tissues of a patient homoplasmic for the T8993G mutation (99% load) and found that the ATP synthase was unstable and disassembled based on native blue gel electrophoresis methods. They concluded that abnormal assembly was the cause of the pathogenicity of the mutation.
The T8993G Mutation Alters Catalysis rather than Enzyme Assembly or Stability-Here, we have studied a fibroblast cell line in which the T8993G mutation is 100%. In this mutant cell line, ATP synthesis was 40% of a control fibroblast line, whereas ATP hydrolysis was essentially normal. Both ATP synthesis and hydrolysis were highly oligomycin sensitive. These results are consistent with previous studies made with fibroblast mitochondria (22,23). To determine the levels of assembly of ATP synthase in the mutant, Rho 0 , and control mitochondria, Western blotting was carried out with an anti-␣ monoclonal antibody. Consistent with the activity measurements (Fig. 1), similar levels of membrane-associated F 1 were found in control, T8993G mutant, and Rho 0 mitochondria (Fig.  3). Therefore, the specific activities of the F 1 F 0 in ATP hydrolysis are very similar in the mutants and control. An active F 1 was still assembled in Rho 0 cells that lack subunit 6 and cannot form a functional ATP synthase. However, in Rho 0 cells, this ATPase activity was not oligomycin-sensitive.
For the above reasons, we conclude that a functional F 1 F 0 is assembled in normal amounts in the T8993G mutant. Moreover, this assembled enzyme is stable. Thus, F 1 is not released by osmotic shock of mitochondria from the T8993G mutant. This is in contrast to release of F 1 by similar treatment of Rho 0 mitochondria where the F 1 F 0 is not fully assembled. Also, an intact F 1 F 0 could be isolated from the 91% mutant by detergent solubilization of mitochondria using affinity chromatography, where binding to the resin was via the F 1 part. A labile enzyme would be expected to fall apart during such purification. In summary, our data indicate that cells with the T8993G mutation contain normal amounts of a stable F 1 F 0 , arguing that the lower ATP synthesis rates observed are caused by deficient catalytic functioning of the enzyme. An altered assembly, as found by Houstek et al. (37) in the postmortem biopsies from different tissues, might have resulted from enzyme degradation rather than incomplete assembly or instability induced by the mutation.
According to recent topology studies of the E. coli. subunit a (39 -41), the mutation T8993G replaces a Leu in the penultimate transmembrane ␣ helix of subunit 6 with an Arg. Crosslinking experiments in E. coli (42)(43)(44) indicate that the region of this replacement is close to the interaction face between subunit 6 and the ring of c subunits and, therefore, in a region important for proton pumping function. The mutation could disrupt proton pumping efficiency by making the channel leaky to protons, i.e. by uncoupling or by slowing the rate of the proton translocation step (for example, by slowing the rotation of the c ring relative to subunit 6). The membrane potential measurements in Fig. 2 argue for, at most, very limited leakage through the F 0 because of the mutation. The extent of regions of high membrane potential along the mitochondrial reticulum in the 100% mutant cell line is higher than that of control MRC5 cells, implying lower rather than higher rates of dissipation of ⌬pH and ⌬ in mutant cells. In the T8993G cells, the proportion of the mitochondrial reticulum with high ⌬ approaches that of control cells in which dissipation of pH via ATP synthesis is blocked by oligomycin. Thus, the L156R mutation of mammalian subunit 6 seems to reduce rather than increase proton conduction through F 0 . It is interesting to note that Vá zquez-Memije et al. (22,23) have recently shown that F 1 F 0 from the T8993G mutant has enhanced sensitivity to oligomycin compared with control cell lines. They used this inhibitor to decrease the mutant load of heteroplasmic cultured cells carrying the T8993G mutation (38).
There is an alternative, previously overlooked explanation of how the T8993G mutation could alter ATP synthesis. In eukaryotes, ATP synthesis is under tight control by one or more inhibitor proteins. Altered binding of the human IF 1 could slow ATP synthesis without significantly affecting the rate of ATP hydrolysis. There is precedent to believe that mutation of subunit 6 could alter inhibitor functioning. Gardner and Cain (45) have recently shown that mutations of subunit a in the E. coli F 1 F 0 can alter the interaction of the intrinsic inhibitor polypeptide (subunit ⑀) with the F 1 complex. Here, we have addressed the role of IF 1 in the phenotype of the T8993G mutation for the first time. The functioning of IF 1 , as measured by inhibition of ATPase activity, was found to be the same in the mutant as in the control MRC5 cells, ruling out that binding of this polypeptide has been altered to favor hydrolysis and limit ATP synthesis. Studies of Rho 0 cells are again a useful control. In the absence of subunit 6 (and A6L) in these cells, there is no effect of the IF 1 on ATPase activity. It appears that IF 1 no longer binds to the enzyme without the mitochondrially encoded subunits present. This loss of inhibitory effect may be functionally FIG. 4. SDS gel electrophoresis of control, T8993G, and Rho 0 F 1 F 0 isolated from human mitochondria. F 1 F 0 complex from human mitochondria was isolated as described under "Materials and Methods," and 30 g of protein were loaded on a 10 -22% SDS gel and silver-stained. Lane 1, bovine heart F 1 F 0 ; lane 2, control human F 1 F 0 ; lane 3, Rho 0 F 1 F 0 ; lane 4, T8993G 91% F 1 F 0 . Lines show the positions of subunits ␣, ␤, ␥, ␦, b, OSCP, and d. Low molecular weight subunits were not resolved. Asterisks show the position of creatine kinase below the ␤ subunit and adenine nucleotide translocase below ␥; these are some impurities frequently associated with F 1 F 0 .
significant. Rho 0 cells must survive by glycolysis as the exclusive mode of ATP production. At the same time, cell viability requires continued stability of mitochondria which, in turn, requires a ⌬ to facilitate protein transport into the organelle. Recent studies suggest that this ⌬ is generated by reversal of the ATP/ADP translocase, with an active ATPase (the F 1 ) providing ADP from imported ATP to maintain the exchange re-action (46,47). Any inhibition of the F 1 by IF 1 would reduce this ADP production.
The Heteroplasmy of mtDNA Mutations May Not Reflect the Levels of Mutant Protein-At 91% of T8993G mutation, cells grew essentially normally; there was a high ATPase activity and ATP synthesis was reduced only 25%. At 100% of mutation, the cells grew very poorly and ATP synthesis was reduced 60%. This big difference in growth and activity effects, in relation to only a small difference in percent of mutant mtDNAs, parallels the phenotype of the T8993G mutation in patients. At 100%, the effect is a debilitating and usually fatal condition known as Leigh's Syndrome. Below 95%, patients have the milder symptoms collectively called neurogenic muscle weakness, atraxia, and retinitis pigmentosa (20). The implication is that tissues tolerate a 25% loss of ATP production but are greatly impaired by a 60% loss. Such threshold effects have been seen for many other mutations in mtDNA. The quantitative activity studies described here show that there is considerably more ATP synthesis activity in the 91% mutant than can be accounted for by 91% of mutant enzyme molecules working at 40% efficiency and 9% molecules functioning at "wild-type" rates. Based on the data in Fig. 1, this would give an overall ATP synthesis rate of 11 compared with 19 nmol of ATP synthesized/min/mg observed. We conclude that there must be more than 9% of fully functional enzyme. This could occur if the levels of subunit 6 synthesized (both normal and mutant) are not limiting for assembly of the F 1 F 0 and if the association of wild-type subunit 6 is preferred over the mutant by the assembling enzyme complex. An alternative explanation is that the mutant subunit 6 is less well expressed or more readily degraded, thereby leading to a higher incorporation of wild-type subunit than expected for the ratio of mutant to wild-type mtDNA molecules. , and bovine heart (lane 3) purified F 1 F 0 were blotted against an OSCP polyclonal antibody. B, bovine heart mitochondria were labeled with [ 14 C]DCCD and the F 1 F 0 complex was purified as described under "Materials and Methods." The enzyme (30 g) was subjected to 10 -22% SDS electrophoresis, and radioactivity was counted before (q) and after (E) performic acid treatment. C, the same procedure as in B was applied to human Rho 0 F 1 F 0 mitochondria to detect the [ 14 6) were loaded/lane. The identity of the band was confirmed in parallel blots against rat liver IF 1 expressed and purified in E. coli as described under "Materials and Methods." FIG. 6. Activation of ATPase activities of control, T8993G, and Rho 0 mitochondria by heat and KCl. Mitochondria (100 g) were exposed to an osmotic shock, and the collected membranes were activated by heat and KCl as described under "Materials and Methods." Controls were incubated at room temperature without KCl. The ratio of the activated/nonactivated hydrolysis is shown for control, T8993G 91%, T8993G 100%, and Rho 0 samples. The same activation procedure was applied to purified F 1 F 0 from control and Rho 0 enzymes (last two lanes). The results show the average of three to six different determinations, and the standard deviation was 15% or lower (not shown).
The F 1 Part Is Assembled on a Subset of F 0 Subunits in Rho 0 Cells-Another interesting finding of the present study relates to the assembly of F 1 F 0 . It had been reported previously that human cells grown in the presence of doxicycline, to block mitochondrial protein synthesis, still assembled a functional F 1 , which was weakly associated to the membrane through subunit c (19). Here we show that in the absence of both subunits 6 and A6L, the rotor subunits, appear to be assembled, including c subunits, and the components of the second stalk, subunits b and OSCP, are present. The interactions among these subunits keep the F 1 attached to the mitochondrial inner membrane, although more weakly than in control cell lines. This is in line with findings in E. coli and yeast where subunit a is not essential for assembly of subunits b and c or for attachment of F 1 to the membrane (17, 48 -50). However, subunits b and c and subunit 8 of yeast (or A6L) are essential for assembly of subunit a (50 -53). Thus, subunit a must be added at the last steps in assembly of F 1 F 0 . This would prevent uncontrolled proton conduction through the F 0 during the assembly process. Interestingly, our results suggest that assembly of the inhibitor protein into F 1 F 0 requires subunits a or A6L for proper control of the ATPase activity, as IF 1 is missing when subunits a and A6L are not present. The inhibitor protein might also add further stability to the F 1 F 0 complex (54). If IF 1 is added late in assembly, other factors must prevent uncoupled ATPase activity of the nascent F 1 .