The two rotor components of yeast mitochondrial ATP synthase are mechanically coupled by subunit δ

Abstract

The mitochondrial ATP synthase is made of a membrane-integrated F₀ component that forms a proton-permeable pore through the inner membrane and a globular peripheral F₁ domain where ATP is synthesized. The catalytic mechanism is thought to involve the rotation of a 10-12 c subunit ring in the F₀ together with the γ subunit of F₁. An important and not yet resolved question is to define precisely how the γ subunit is connected with the c-ring. In this study, using a doxycycline-regulatable expression system, we provide direct evidence that the rest of the enzyme can assemble without the δ subunit of F₁, and we show that δ-less mitochondria are uncoupled because of an F₀-mediated proton leak. Based on these observations, and taking into account high-resolution structural models, we propose that subunit δ plays a key role in the mechanical coupling of the c-ring to subunit γ.

Most of the ATP in bacteria and in the mitochondria and chloroplasts of animals and plants is produced by an F₁·F₀-type ATP synthase. This enzyme utilizes a proton gradient across its host membrane to catalyze ATP synthesis from inorganic phosphate (P_i) and adenosine diphosphate (ADP) (1). It has two major structural domains, a membrane-integrated F₀ component that forms a proton-permeable pore through the membrane and a peripheral F₁ component where ATP is synthesized.

In the mitochondrial ATP synthase, the F₁ domain comprises five different subunits with the α₃β₃γ₁δ₁ε₁ stoichiometry. The three α subunits and the three β subunits form a crown where they alternate around a central α-helical coiled-coil in the γ subunit (2). The γ subunit extends from below the (αβ)₃ subcomplex where it interacts with subunits δ and ε to create a 40- to 45-Å-long ”central stalk,” which links F₁ to F₀ (3). Proton translocation within the F₀ is assumed to drive the rotation of a transmembrane ring of 10-12 c subunits. This ring is believed to be permanently fixed to the central stalk (4), with the result that subunit γ rotates with the c-ring when protons cross the F₀ (5). During its rotation, the γ subunit interacts sequentially with the three αβ pairs in a way that favors ATP synthesis in the catalytic sites, as required by the binding change mechanism (1). In this mechanism, a second, peripheral stalk also connecting F₁ to F₀ is thought to counter the tendency of the (αβ)₃ subcomplex to follow the rotation of the central stalk during catalysis (6).

The predicted rotation of the bacterial homologs of subunits γ and δ (referred to as ε in Escherichia coli) was visualized by attaching a fluorescently labeled actin filament to single-immobilized α₃β₃γ and α₃β₃γε subcomplexes (7, 8). By using similar methods, the c-ring was shown to rotate during ATP hydrolysis by the F₁·F₀ complex (9, 10). Cross-link experiments have demonstrated that the bacterial γ, ε (homolog of mitochondrial δ), and c subunits rotate together as a fixed ensemble during catalysis (11, 12).

An important question is to define precisely how the central stalk is connected to the c-ring. The binding must be strong enough to withstand the torque on the c-ring by the F₁ as the mechanical energy from rotation is converted into chemical energy in ATP. High-resolution structures have provided insights into the connections taking place at the F₁·c-ring interface (3, 4). It seems that there are only a few direct contacts between the c-ring and subunit γ whereas subunit δ shows extensive interactions with both the c-ring and subunit γ. The mitochondrial ε subunit is like a clip around the γ and δ subunits. Such an arrangement indicates that the δ and ε subunits may have a key role in connecting F₁ with the c-ring, with the result that the c-ring and subunit γ move in unison (5).

The function of subunits δ and ε within mitochondrial ATP synthase was investigated with mutants lacking the genes encoding these proteins in Saccharomyces cerevisiae, a facultative aerobe well suited for genetic analysis of the mitochondrial energy-transducing system. Both Δε (13) and Δδ (14, 15) mutants were found to produce many petites, i.e., cells with a large deletion (rho^-) or total loss (rho⁰) of the mitochondrial genome (mtDNA), 70% and 100%, respectively. This result hampered the analysis of the consequences of these mutations because three ATP synthase subunits are encoded by this DNA, including subunits a and c (also known as subunits 6 and 9, respectively, in yeast), which are essential components of the proton-translocating domain. Interestingly, the Δε mitochondria having retained mtDNA were found to be uncoupled because of an F₀-mediated proton leak. This observation is consistent with a role for subunit ε in coupling proton translocation to ATP synthesis. Whether such a leak also occurs when subunit δ is missing could not be assayed because of a total lack of functional mtDNA in the Δδ mutant.

In this study, we further investigate the function of subunit δ in yeast ATP synthase by using an approach in which δ subunit expression can be regulated exogenously from a doxycyclinerepressible promoter (16). With this system, we were able to generate large cellular depletions in subunit δ with no substantial loss of mtDNA. We provide direct biochemical evidence that the rest of the F₁·F₀ complex can assemble without subunit δ, and we show that δ-less mitochondria are uncoupled because of a proton leak mediated by the ATP synthase proton channel. This finding provides strong functional evidence that the sole loss of subunit δ suffices to uncouple the ATP synthase. Based on the data reported in this study and taking into account current structural models, we propose that subunit δ plays a key role in the mechanical coupling of the c-ring with subunit γ.

Materials and Methods

Strains and Media. E. coli XL1-Blue strain (Stratagene) was used for the cloning and propagation of plasmids. Rich glucose (YPGA), galactose (YPGALA), or glycerol plus ethanol (YPEG) media, and synthetic complete medium used for growing yeast strains were prepared as described in ref. 17.

Genetic and Molecular Biology Methods. Previously described procedures were used for genetic experiments (18), yeast transformations (19), and DNA manipulations (20).

Construction of a Yeast Strain Expressing the δ Subunit Under the Control of a Doxycycline-Repressible Promoter. The coding sequence of the δ subunit gene (ATP16) was PCR-amplified from genomic DNA isolated from strain W303-1B with primers 5′-ata gga tcc ATG ACG ACT CGT TTG CTC CAA CTC ACT CGT CCT C-3′ and 5′-ata gcg gcc gcC TAA AGT TTT TTA CCC AAA ATG GAG GTA GC-3′ (capital letters refer to regions of ATP16 homology, with the start and stop codons underlined, respectively, and lowercase letters refer to restriction sites and filler DNA.) To insert the gene 3′ of the doxycycline-repressible promoter, the amplicon was cut with BamHI and NotI and ligated into plasmid pCM189 (16) cut with the same restriction enzymes to give plasmid pSDC13. The cloned gene was verified by DNA sequencing. Cells of strain SDC22 transformed with pSDC13 were selected on synthetic complete medium lacking uracil. A large fragment internal to the coding sequence (from position +51 to +430 after the initiator methionine codon) of chromosomal ATP16 gene was then deleted in SDC22/pSDC13 according to a described procedure (21). The KanMX4 module was PCR-amplified with primers 5′-ATG TTA CGT TCA ATT ATT GGA AAG AGT GCA TCA AGA TCA TTG AAT TTC gtc ata ggc cac tag tgg atc tg-3′ and 5′-TTT CAA TAC GGA TTG TAG GTT TTC TAA AAC TTC TAC TTG AAT TGC AGC TTc agc tga agc ttc gta cgc-3′ (capital letters refer to regions of homology to ATP16, and lowercase letters refer to regions of KanMX4 homology). Cells of strain SDC22 containing pSDC13 were transformed with the deletion cassette. The transformants were selected on YPGA supplemented with 200 μg/ml of geneticin and analyzed by PCR analysis. One clone called SDC6 carrying the expected deletion in chromosomal ATP16 gene, in a rho⁺ state and containing pSDC13, was retained for further analysis.

Biochemical Techniques. Mitochondria were prepared by the enzymatic method as described (22). Protein amounts were determined by the procedure of ref. 23. The specific ATPase activity was measured at pH 8.4 by using a described procedure (24). Oxygen consumption rates were measured with a Clark electrode in the respiration buffer (0.65 M mannitol/0.3 mM EGTA/3 mM Tris-phosphate/10 mM Tris-maleate, pH 6.75) as described (25). Variations in transmembrane potential (ΔΨ) were evaluated in the same buffer by measurement of fluorescence quenching of rhodamine 123 with a SAFAS (Monte-Carlo, Monaco) fluorescence spectrophotometer (26). SDS/PAGE was according to Laemmli (27). Western blot analyses were performed as described (28). Polyclonal antibodies raised against yeast ATP synthase subunits α and β were used after dilution 1:100,000 and 1:50,000, respectively; and 1:10,000 for subunits γ, δ, a, c, i, 4, and oligomycin sensitivity conferral protein (OSCP). Nitrocellulose membranes were incubated with peroxidase-labeled antibodies at a 1:10,000 dilution and revealed with the ECL reagent of Amersham Pharmacia.

Blue Native (BN)/PAGE Analysis. BN/PAGE of mitochondrial extracts, solubilized with digitonin to protein ratio of 2 (g/g), was performed as described (29). As shown in Fig. 5, the BN/PAGE was loaded with 0.6 mg and 1.2 mg of extracted proteins, respectively, for SDC6 grown with no addition and SDC6 grown in the combined presence of doxycycline plus oligomycin. The protein complexes were separated in a 3-13% acrylamide continuous gradient gel, transferred to poly(vinylidene difluoride) membranes, and analyzed by Western blot.

Fig. 5.

SDS/PAGE and BN/PAGE analysis of δ-less mitochondria. Mitochondria were prepared from SDC6 cells grown in a rich glucose medium (YPGA) for eight to nine generations with no addition (-) or with 10 μg/ml doxycycline plus 6 μg/ml oligomycin (+). (A) SDS/PAGE and Western blot analysis of the mitochondria with antibodies against the indicated ATP synthase subunits. (B) Mitochondrial digitonin extracts were resolved by BN/PAGE (see Materials and Methods for details of the procedure), transferred to poly(vinylidene difluoride) membranes, and analyzed by Western blot with antibodies against the indicated ATP synthase subunits.

Epifluorescence Microscopy. Epifluorescence microscopy of 2-(4-dimethylaminostyryl)-1-methylpyridinium iodide (DASPMI)-stained cells was carried out with a Leica DMRXA microscope fitted with a ×100 immersion objective and a standard FITC filter.

Results

Doxycycline-Repression of δ Subunit Expression Is Rapidly Followed by Loss of Respiratory Growth. We created a modified yeast strain (SDC6) expressing subunit δ from a doxycycline-repressible (Tet-O) promoter (see Materials and Methods). In a nonfermentable medium (glycerol, ethanol) where the presence of a functional ATP synthase is absolutely required, SDC6 grew as well as wild-type yeast SDC22 (Fig. 1A). The growth of SDC22 was unaffected by 10 μg/ml doxycycline. By contrast, that of SDC6 was strongly affected by doxycycline. Indeed, after the addition of the drug, the cells grew for two additional generations (in ≈6 hr) and then stopped to multiply (Fig. 1A). When transferred back to a fresh medium devoid of doxycycline, most of the SDC6 cells (>98%) were able to resume respiratory growth, showing that the doxycycline-induced growth arrest was not due to a loss of functional mtDNA (not shown). Immunoblottings revealed that SDC6 was only partially depleted in subunit δ when it stopped growing, in comparison with SDC6 grown without doxycycline (Fig. 1B). SDC22 grown without doxycycline and SDC22 grown with doxycycline had a δ subunit content similar to that of SDC6 grown without doxycycline. This observation is consistent with the growth data, which showed a quite well adapted δ subunit expression from the unrepressed tet-O promoter, thereby precluding any deleterious effect of doxycycline on the growth of the wild-type yeast. Moreover, none of the other five ATP synthase subunits tested (α, β, γ, 4, and i) was in reduced amounts in SDC6 grown with doxycycline (data shown for subunit 4 in Fig. 1B), indicating that the partial depletion in δ subunit was specific.

Fig. 1.

A block in the expression of the δ subunit leads to a respiratory growth arrest. (A) Growth curves. Wild-type strain SDC22 expressing the δ subunit from its own promoter and SDC6 expressing the δ subunit from a doxycyclinerepressible promoter (tet-O) were pregrown in rich glycerol plus ethanol medium (YPEG). Fresh YPEG medium containing no or 10 μg/ml doxycycline was inoculated with these precultures (at the starting time of the growth curves shown). ⋄, SDC22 without doxycycline; ▪, SDC22 with doxycycline; •, SDC6 without doxycycline; ▾, SDC6 with doxycycline. (B) Western blot analysis. Mitochondria were prepared from the cultures shown in A at the time indicated by a dashed line (+DOX and -DOX refer to the presence or absence of doxycycline in the growth medium, respectively). The mitochondrial proteins were then analyzed by SDS/PAGE and Western blot with antibodies against the indicated ATP synthase subunits. Each lane was loaded with 10 μg of proteins.

A Block in δ Subunit Expression Results in a Massive Proton Leak Across the Mitochondrial Inner Membrane Mediated by the ATP Synthase Proton Channel. From these data (see above and Fig. 1), it seemed that a partial depletion in subunit δ of ≈50% was sufficient to abolish respiratory growth at 28-30°C. This is an interesting observation because lowering the ATP synthase level by up to 80% has only minor effects on yeast respiratory growth above 20°C (30). We therefore suspected that, in addition to a reduction in mitochondrial ATP synthesis, a block in subunit δ synthesis resulted in other deleterious effects. This possibility was investigated as described hereafter by the analysis of isolated mitochondria prepared after growing SDC6 and SDC22 for 6 hr in glycerol + ethanol-rich medium containing or not containing 10 μg/ml doxycycline.

Oxygen Consumption Measurements. The two SDC22 mitochondrial preparations exhibited similar respiratory activities, showing that doxycycline at the concentration used had no significant incidence on wild-type yeast respiration (Table 1). Mitochondria from SDC6 grown without doxycycline behaved like wild-type SDC22 mitochondria. With NADH, they showed a basal respiration (state 4) corresponding to 12% of the maximal respiration rate measured in the presence of the membrane uncoupler carbonyl cyanide m-chlorophenylhydrazone (CCCP; Fig. 2A and Table 1). After subsequent addition of ADP, a 4-fold increase in oxygen consumption up to 45% of the maximal respiration rate was observed (state 3 respiration corresponding to classical phosphorylating conditions). As expected, the addition of oligomycin, a specific inhibitor of the ATP synthase proton channel, had no effect on state 4 (Fig. 2A and Table 1) but inhibited state 3 (not shown).

Table 1. Respiratory and ATP hydrolytic activities of mitochondria.

	Respiration rate, %					ATPase SA, μmol Pi·min-1·mg-1
Condition	State 4	+Oligo.	State 3	+CCCP	RCR	—Oligo.	+Oligo.	% of inhibition
SDC22	12 ± 2	14 ± 2	51 ± 2	100	4.3	11.5 ± 1.1	0.6 ± 0.2	95
SDC22 + Dox	12 ± 2	12 ± 2	48 ± 3	100	3.9	11.2 ± 0.2	0.6 ± 0.1	95
SDC6	12 ± 4	17 ± 3	43 ± 5	100	3.6	11.3 ± 1.7	0.7 ± 0.4	95
SDC6 + Dox	89 ± 5	16 ± 3	N	100	N	7.7 ± 1	0.8 ± 0.2	90

Mitochondria were isolated from SDC22 and SDC6 grown in rich glycerol plus ethanol medium (YPEG) with or without doxycycline as described in the legend of Fig. 1. Additions were 0.3 mg/ml proteins, 4 mM NADH (state 4), 400 μM ADP (state 3), 6 μg/ml oligomycin (+Oligo.), and 3 μM + CCCP. Respiratory control ratio (RCR) is expressed as state 3 to state 4 respiration rates. N, not applicable. The results shown are representative of duplicate experiments for SDC22 and quadruplicate experiments for SDC6. Respiratory activities are expressed as percentages of the maximal respiration rates assayed with CCCP (2,289 ± 329 nmol O·min^-1·mg^-1 for SDC22, 2,734 ± 272 nmol O·min^-1·mg^-1 for SDC22 plus Dox, 1,582 ± 368 nmol O·min^-1·mg^-1 for SDC6, and 1,778 ± 418 nmol O·min^-1·mg^-1 for SDC6 plus Dox). SA, specific activity.

Fig. 2.

Oxygen consumption of mitochondria. Mitochondria prepared from SDC6 grown without (A) or with (B) doxycycline (as described in the legend of Fig. 1) were assayed for oxygen consumption. The additions were 0.3 mg/ml mitochondrial proteins (M), 4 mM NADH, 400 μM ADP, 6 μg/ml oligomycin, and 3 μM CCCP.

Mitochondria from SDC6 grown in the presence of doxycycline, i.e., partially depleted in subunit δ, exhibited a very peculiar behavior in oxygen consumption experiments. After adding NADH, they exhibited a high respiration rate that was almost insensitive to subsequent ADP or CCCP addition (Fig. 2B and Table 1). Interestingly, in the presence of oligomycin, the rate of oxygen consumption was decreased ≈5-fold down to values seen normally in state 4 (Fig. 2B and Table 1), and the mitochondria recovered a normal sensitivity to membrane proton ionophores. These data pointed to a large uncoupling in the mitochondria partially depleted in subunit δ due to a proton leak through the F₀ component of the ATP synthase.

ATPase Activities. Mitochondria from SDC22 grown with or without doxycycline had a similar oligomycin-sensitive ATPase activity (≈11 μmol P_i·min^-1·mg^-1, see Table 1), showing that doxycycline had no significant incidence on ATP synthase accumulation and functioning in a wild-type genetic context. A similar ATPase activity was detected in SDC6 mitochondria from cells grown without doxycycline, showing that properly assembled and functional δ subunits were produced in sufficient amounts from the uninhibited Tet-O promoter. SDC6 grown with doxycycline had a reduced but still large oligomycin-sensitive ATPase activity, representing 60-70% of that of the control mitochondria not depleted in δ subunit (Table 1). A BN/PAGE analysis of the uncoupled mitochondria revealed the presence of intact F₁·F₀ complexes in amounts consistent with the ATPase activity measurements (not shown). These data established that the respiratory growth arrest induced by a block in δ subunit expression occurred when the cells still contained large amounts of functional ATP synthases.

Mitochondrial Membrane Potential Analysis. The results of the respiration and ATPase activity measurements indicated that a specific block in δ subunit expression was quite rapidly followed by the establishment of an F₀-mediated proton leak across the mitochondrial inner membrane. We then determined the influence of this proton leak on the mitochondrial membrane potential (ΔΨ) by monitoring the fluorescence quenching of rhodamine 123 on isolated mitochondria (Fig. 3). The mitochondria from SDC6 grown without doxycycline were properly energized by the respiratory complexes. Indeed, the addition of an electron donor (ethanol) to the mitochondrial suspension produced a large fluorescent quenching of the dye, which collapsed after subsequent addition of potassium cyanide (KCN; Fig. 3A). By contrast, mitochondria from SDC6 grown with doxycycline were very poorly energized by ethanol (Fig. 3B). However, after subsequent oligomycin addition, a stable KCN-sensitive increase in ΔΨ was seen.

Fig. 3.

Energization of mitochondria. Energization of the mitochondrial inner membrane was monitored by fluorescence quenching of rhodamine 123 with intact mitochondria isolated from SDC6 grown in the absence (A) or presence (B) of doxycycline (as described in the legend of Fig. 1). The additions were 0.5 μg/ml rhodamine 123, 0.3 mg/ml mitochondrial proteins (M), 10 μl of ethanol (EtOH), 6 μg/ml oligomycin, 0.2 mM potassium cyanide (KCN), and 3 μM CCCP.

We then probed in vivo mitochondrial energization by staining whole cells with DASPMI, a fluorescent dye sensitive to mitochondrial ΔΨ (Fig. 4). The mitochondrial network was clearly stained in SDC6 grown without doxycycline, reflecting a proper energization of their inner membrane (Fig. 4A). By contrast, when SDC6 was grown in the presence of doxycycline, no mitochondrial network could be detected (Fig. 4B). This lack of staining was indistinguishable from that of SDC6 grown without doxycycline when mitochondrial ΔΨ was collapsed with CCCP (Fig. 4C).

Fig. 4.

Fluorescence microscopy of whole cells with a mitochondrial membrane potential probe. SDC6 was grown without (A) or with (B)10μg/ml doxycycline for 6 hr in rich galactose medium (YPGALA). The cells were then incubated for 30 min with 5 μM DASPMI and then examined by fluorescence microscopy. SDC6 cells grown without doxycycline were also incubated with 10 μM CCCP (C) to demonstrate complete dissipation of mitochondrial membrane potential. (Upper) Nomarski views. (Lower) Mitochondrial stainings. (Bar = 3 μm.)

These data confirmed that the uncoupling of mitochondria harboring a partial depletion in subunit δ was due to an F₀-mediated proton leak, and that this leak prevented the maintenance of a sufficient mitochondrial ΔΨ. Such a failure in mitochondrial energization was likely the primary cause of the respiratory growth arrest seen after the block in δ subunit expression.

Assembly of ATP Synthase in the Absence of δ Subunit. An important question was to determine the physical nature of the protein assemblies responsible for the proton leak that occurred in the absence of the δ subunit. As shown above, on a nonfermentable substrate, SDC6 cells stopped growing after the addition of doxycycline in two generations only, well before a complete depletion in δ subunit and hence in intact F₁·F₀ complexes. The resulting mix in intact ATP synthases and δ-less enzymes hampered the analysis of the δ-less assemblies responsible for the observed mitochondrial uncoupling. To obtain a larger depletion in δ subunit, it was necessary to allow the cells to divide more, to dilute/degrade the δ subunits present at the time at which δ subunit expression was blocked by doxycycline. This result was possible in a fermentable medium where SDC6 could grow indefinitely in the presence of doxycycline. However, the cells then lost their mtDNA quite rapidly, and hence the three F₀ subunits encoded by this DNA. After 15 generations in such conditions, <10^-4 cells were still rho⁺ vs. >99% when SDC6 was grown without doxycycline. Interestingly, mtDNA was lost much less rapidly when SDC6 was grown in the combined presence of doxycycline and oligomycin. We observed that the retention of mtDNA was more efficient when the cells were grown with glucose rather than galactose. In these conditions, 60% of the cells were still rho⁺ after 15 generations. This result indicated, as previously discussed (31), that the strong mtDNA instability in a δ-less context is a direct consequence of the F₀-mediated proton leak generated after blocking the expression of δ subunit (see Discussion).

Thus, by growing SDC6 in glucose in the combined presence of doxycycline and oligomycin, it was possible to deplete the cells in subunit δ extensively without major mtDNA loss. In such conditions, we found that eight to nine generations were sufficient for a virtual depletion in δ subunit (see Fig. 5A) with only 25% of rho^-/rho⁰ cells. In a first step to determine the influence of a lack in subunit δ on the assembly of ATP synthase, we used immunoblotting to analyze the δ-less mitochondria for their content in subunit α, β, γ, OSCP, 4, and a. These were in normal (α, β, OSCP) or only partially diminished (γ, 4, a) quantities, in comparison with SDC6 mitochondria not depleted in subunit δ (see Fig. 5A). This is an interesting observation because disturbances in ATP synthase assembly usually lead to degradation of the unassembled subunits. For example, deleterious mutations in ATP synthase subunit α, β, 4, c, or 8 do not affect synthesis but all prevent accumulation of subunit a, presumably because of a higher susceptibility to proteases (ref. 32 and references therein, and ref. 33). Thus, given the quite good detection of subunit a in the δ-less mitochondria, we could already suspect that a specific lack in subunit δ would have no major incidence on the assembly of the other subunits in the F₁·F₀ complex.

This possibility was confirmed by a BN/PAGE analysis of the δ-less mitochondria. We used conditions where the ATP synthase is normally isolated as dimers and monomers (see Materials and Methods). After their electrophoretic separation, the mitochondrial protein complexes were transferred to poly(vinylidene difluoride) and probed with antibodies against several ATP synthase subunits. As expected, SDC6 mitochondria not depleted in subunit δ gave two major signals with all tested antibodies, corresponding to dimeric and monomeric ATP synthase [Fig. 5B, lanes labeled (-) at positions 1 and 2, respectively]. Complexes of nearly the same size were detected with the δ-less protein extracts, with antibodies against subunits α, β, γ [shown in Fig. 5B, lanes (+)], i, a, c, 4, and OSCP (not shown). These data show that the rest of the ATP synthase can assemble without subunit δ.

Additional ATP synthase subcomplexes never seen with wild-type mitochondria were detected in the δ-less mitochondria: (i) between dimeric and monomeric ATP synthases [Fig. 5B,lanes(+) at position 3] and (ii) small ones [Fig. 5B, lanes (+) at position 4]. Those at position 3 reacted with all ATP synthase antibodies tested, except δ, whereas those at position 4 gave a signal only with antibodies against subunit α and β. As discussed below, we believe that these additional assemblies did not exist in the intact δ-less mitochondria but were artificially created by damage to the δ-less F₁·F₀ complexes during their extraction with detergent and/or in the course of their electrophoresis.

Discussion

Using a doxycycline-regulatable system, we show here that a specific block in δ subunit synthesis leads to a massive proton leak across the mitochondrial inner membrane mediated by the ATP synthase proton channel. The main question was to determine the physical nature of the δ-less assemblies responsible for the proton leak. In this respect, it is important to note that membranes stripped of F₁ are leaky for protons and again become impermeable to protons after a subsequent F₀ inhibitor addition (34). Thus, it was important to know whether δ-less mitochondria were leaky for protons because they can assemble F₀ but not F₁ or because of the sole absence of subunit δ from the F₁·F₀ complex.

For this investigation, it was necessary to obtain cells unable to express the δ subunit but having retained the capacity to synthesize all other ATP synthase subunits. Unfortunately, strains lacking the δ subunit gene (Δδ) very rapidly lose their mtDNA (rho^-/rho⁰) and hence the capacity to synthesize the three F₀ subunits encoded by this DNA (including subunits a and c) (14). Our data provide a reasonable explanation for this mtDNA instability. We indeed demonstrate that a block in δ subunit expression leads to a major failure in mitochondrial energization because of a proton leak through the F₀ component of ATP synthase. This result should be a lethal situation, even in a facultative aerobe like S. cerevisiae where the ATP can be made by fermentation (17, 35). However, such a situation can be overcome in S. cerevisiae by the loss of its mtDNA. Indeed, after the loss of this DNA [a nonlethal and quite frequent (10^-2) event in S .cerevisae], the F₀ can no longer be synthesized, so the absence of subunit δ ceases to be deleterious. Consistent with this interpretation, we observed that the integrity and expression of the mtDNA could be well preserved in δ-less cells when these were grown in a medium containing an F₀ inhibitor. Thus, the doxycycline-regulatable system allowed us to investigate the influence of a lack in subunit δ on the assembly of the other ATP synthase components.

The loss of a single ATP synthase subunit often leads to posttranslational degradation of several other enzyme subunits, presumably because a proper quaternary structure is required to protect the enzyme from proteolytic breakdown (ref. 32 and references therein, and ref. 33). The proton channel subunits are known to be particularly sensitive to such effects, especially subunit a, which fails to accumulate in most ATP synthase mutants. Such pleiotropic effects were not seen in our δ-less mitochondria, which a priori was already a strong indication that, without subunit δ, the other enzyme subunits can associate properly with each other. This finding was confirmed by BN/PAGE analysis of the δ-less mitochondria, which revealed the presence of assemblies with no subunit δ but having nearly the same size as intact F₁·F₀ complexes. However, these assemblies were detected in relatively small amounts, and additional ATP synthase subcomplexes were found (at positions 3 and 4 in Fig. 5B). In this respect, it is important to note that most of the direct contacts between the F₁ and the proton channel seem to be mediated by subunit δ (4); i.e., with no subunit δ, the F₁ may be connected to the rest of the complex by the second stalk only. Such complexes may be damaged when extracted from the membrane with detergents and/or in the course of their electrophoretic migration. This finding could explain the low abundance in which the δ-less assemblies were detected, as well as the presence of additional ATP synthase subcomplexes. According to their size and immunological reactivity, the latter could have been produced by a partial dissociation of α and β subunits after opening the δ-less mitochondria. There are known cases where other supramolecular mitochondrial assemblies were made fragile by the loss of one protein subunit and are disrupted during BN/PAGE analysis (36).

Alterations in ε or γ subunit expression have also been shown to be responsible for a proton leak through the F₀ (13, 37). Thus, like subunit δ, subunits ε and γ are not needed for the assembly of the ATP synthase proton channel. By contrast, as we have previously shown, the proton channel is absent in cells unable to assemble the (αβ)₃ subcomplex, e.g., after genetic deletion of subunit β or loss of a chaperone specifically involved in α/β hexamerization (33). Moreover, when the second stalk is altered, e.g., after genetic deletion of subunit 4 (homolog of bacterial b), F₁ forms, although the proton translocating domain does not (38). Finally, it is known that F₁ forms as a soluble ATPase in the absence of mtDNA, i.e., when F₀ cannot be synthesized (35). Thus, both the F₁ and the second stalk are needed for proper assembly/stability of the proton translocating domain whereas neither the second stalk nor the proton-translocating domain are required for stable accumulation of F₁ oligomers. These data indicate an ordered sequence starting with formation of F₁, followed by association of F₁ with the second stalk, and finally insertion of the proton-translocating domain. Our data show that this sequence can operate as well when subunit δ is missing. However, subunit δ must certainly be inserted in an early event, or at least at the same time as the assembly of F₀, to avoid deleterious proton leaks across the mitochondrial inner membrane.

Recent cross-linking experiments have indicated that the F₀ is a rotary proton channel, meaning that, if there is no rotation of the c-ring, there is no proton flow through the F₀, even when the whole F₁ has been detached from the membrane-embedded F₀ complexes (34). It is therefore a reasonable assumption that the F₀-mediated proton leak we observed when subunit δ is missing involves the rotation of the c-ring. In conclusion, we show that, with no δ subunit, the rest of the enzyme can assemble and that, in the δ-less F₁·F₀ complex, the proton channel is likely to be functional whereas the c-ring rotation is unable to drive ATP synthesis in the F₁. This conclusion is perfectly consistent with high-resolution structures indicating that subunit δ interacts extensively with both the c-ring and subunit γ whereas little direct contact exists between the c-ring and subunit γ (3, 4). Such an arrangement, together with our functional data, suggests that, without subunit δ, the c-ring should be unable to drag subunit γ and consequently should rotate freely. Interestingly, a partial depletion in subunit δ (around 50%) is sufficient to uncouple the mitochondria almost totally. This finding indicates that protons are transported more rapidly by δ-less F₁·F₀ complexes than they are in intact ATP synthases, which is consistent with the notion of a motor running without load.

There is evidence that the bacterial and chloroplastic homologs of mitochondrial δ (referred to as subunit ε) have a similar action in the coupling of F₁ to F₀. Indeed, elimination of the E. coli ε subunit gene (uncC) results in growth defects that are much more severe than those seen with single deletions in the other unc genes, and membranes from the uncC mutant strain are more leaky for protons than the wild type (39). Similarly, thylakoid membranes stripped of CF₁ can bind CF₁ particles devoid of subunit ε, but photophosphorylation was not reconstituted and fluorescence quenching of quinacrine was induced by light only after the addition of N,N′-dicyclohexylcarbodiimide (DCCD; ref. 40). This finding indicates that, despite major differences among species in the central stalk component of the ATP synthase (see ref. 5 for review), the mechanisms ensuring the mechanical coupling of the two rotors of the enzyme have been conserved throughout evolution from bacteria to higher eukaryotes.

We show, by using the doxycycline-regulatable system, that a partial depletion in δ subunit around 50% suffices to uncouple the mitochondrion almost totally. By contrast, diploid cells lacking one of the two δ subunit genes (Δδ/+), which may in theory have a 50% depletion in subunit δ, exhibited only a partial mitochondrial uncoupling that was not sufficient to prevent respiratory growth (37). However, the authors did not investigate to what extent Δδ/+ cells were actually depleted in subunit δ. In our hands, we reproducibly observed that >80% of ascii arising from Δδ/+ cells gave an aberrant pattern of germination, with three to four respiratory growth competent spores instead of the expected 2:2 segregation (not shown). A possible explanation is that Δδ/+ cells may alleviate the uncoupling effects of a lack in δ subunit by duplicating their single δ subunit gene, by gene conversion, or by duplication of whole or part of the chromosome where it is located. Such genetic rearrangements do not have the time to occur with the exogenous regulatable expression system. Indeed, this system makes it possible to create large populations with a substantial depletion in subunit δ in quite a short time (a few hours) whereas dozens of generations, after the initial mutation giving a Δδ/+ cell, will be needed before one can investigate the properties of such cells. We therefore believe that the doxycycline-regulatable expression system gives a much better picture of the effects due to the absence of subunit δ than Δδ/+ cells.

In this work, we report the use of an exogenously regulated expression system in the study of the mitochondrial energy-transducing system. It is especially appropriate for studying proteins directly involved in energy transducing but whose loss or modifications indirectly lead to disturbances in other mitochondrial functions, e.g., maintenance of mtDNA, which is dramatically compromised in strains with an F₀-mediated proton leak. As such, it will help to better understand the structure-function relationships of the central stalk component of the F₁·F₀ complex and to unravel the mechanisms preventing accumulation in the inner membrane of uncoupled ATP synthases.