A “Petite Obligate” Mutant of Saccharomyces cerevisiae

Within the mitochondrial F1F0-ATP synthase, the nucleus-encoded δ-F1 subunit plays a critical role in coupling the enzyme proton translocating and ATP synthesis activities. In Saccharomyces cerevisiae, deletion of the δ subunit gene (Δδ) was shown to result in a massive destabilization of the mitochondrial genome (mitochondrial DNA; mtDNA) in the form of 100% ρ–/ρ° petites (i.e. cells missing a large portion (>50%) of the mtDNA (ρ–) or totally devoid of mtDNA (ρ°)). Previous work has suggested that the absence of complete mtDNA (ρ+) in Δδ yeast is a consequence of an uncoupling of the ATP synthase in the form of a passive proton transport through the enzyme (i.e. not coupled to ATP synthesis). However, it was unclear why or how this ATP synthase defect destabilized the mtDNA. We investigated this question using a nonrespiratory gene (ARG8m) inserted into the mtDNA. We first show that retention of functional mtDNA is lethal to Δδ yeast. We further show that combined with a nuclear mutation (Δatp4) preventing the ATP synthase proton channel assembly, a lack of δ subunit fails to destabilize the mtDNA, and ρ+ Δδ cells become viable. We conclude that Δδ yeast cannot survive when it has the ability to synthesize the ATP synthase proton channel. Accordingly, the ρ–/ρ° mutation can be viewed as a rescuing event, because this mutation prevents the synthesis of the two mtDNA-encoded subunits (Atp6p and Atp9p) forming the core of this channel. This is the first report of what we have called a “petite obligate” mutant of S. cerevisiae.

The mitochondrial inner membrane contains the ATP synthase, which utilizes a transmembrane proton gradient to catalyze ATP synthesis from inorganic phosphate and ADP. The ATP synthase has two major structural domains, an F 0 component, which forms a protonpermeable pore across the membrane, and a peripheral, matrix-localized, F 1 component, where the ATP is synthesized (1).
The F 1 domain comprises five different subunits, all nucleus-encoded, with ␣ 3 ␤ 3 ␥ 1 ␦ 1 1 stoichiometry. The three ␣ subunits and the three ␤ subunits alternate in position within a hexamer that contains the adenine nucleotide processing sites (2)(3)(4). The ␥, ␦, and subunits form a subcomplex of F 1 , named the central stalk, linking the ␣ 3 ␤ 3 subcomplex to the ATP synthase proton channel (5). During catalysis, the central stalk rotates together with a transmembrane ring of 10 -12 c subunits in the F 0 (6 -12). In the course of this rotation, the ␥ subunit sequentially interacts with the three ␣␤ pairs in a way that favors ATP synthesis in the catalytic sites, as required by the binding change mechanism (1).
In the yeast Saccharomyces cerevisiae, null mutations of the ␦ or ␥ subunit massively destabilize the mitochondrial genome in the form of 100% cytoplasmic petites, which are cells with a large (Ͼ50%) deletion in mitochondrial DNA (mtDNA) 6 ( Ϫ ) or completely lacking mtDNA (°) (13,14). This is an intriguing observation, since many other mutations impairing mitochondrial oxidative phosphorylation, including a null mutation of the catalytic ␤ subunit of the ATP synthase, have no major effect on the stability of the mitochondrial genome (14).
Defects in the synthesis of the ␦ or ␥ subunit were shown to result in isolated mitochondria in a major uncoupling of mitochondrial respiration mediated by partial F 1 F 0 assemblies in which proton translocation is not coupled to ATP synthesis in the F 1 (15,16). In media containing oligomycin, a specific inhibitor of the ATP synthase assumed to block the enzyme proton channel, mtDNA maintenance could be observed in growing ⌬␦ or ⌬␥ yeast, providing in vivo evidence that an ATP synthase uncoupling may be the cause of the loss of mtDNA in cells unable to properly express the ␦ or ␥ subunit (16,17).
However, how the ATP synthase uncoupling caused by the absence of the ␦ or ␥ subunit destabilizes the mtDNA is still unknown. It has been hypothesized that the loss of the mtDNA allows ⌬␦ or ⌬␥ yeast to survive by eliminating the ATP synthase F 0 component, which is in part encoded by the mtDNA (16,17). The idea is that the F 0 -mediated mitochondrial uncoupling in ⌬␦ or ⌬␥ yeast would be lethal by preventing essential mitochondrial reactions from occurring. As a consequence, ⌬␦ or ⌬␥ yeast would need to inactivate the ATP synthase proton channel in order to survive.
However, in apparent contradiction with this idea, although a lack in the third central stalk subunit () also results in a major F 0 -mediated mitochondrial uncoupling, only partial effects were observed on the mtDNA in ⌬ mutant (18). Indeed, cultures of this mutant usually contain 30% ϩ cells, showing that retention of functional mtDNA in ⌬ yeast is not lethal. Furthermore, it has been reported that a lack in the ␦ subunit is viable in Kluyveromyces lactis, a "petite negative" yeast that cannot survive the Ϫ /°mutation (19). These observations raise the possibility that destabilization of the mtDNA in ⌬␦ and ⌬␥ S. cerevisiae strains results from an increase in the frequency of the Ϫ /°mutation rather than from the impossibility of these mutants surviving when they contain functional mtDNA.
Fox and co-workers (20) have developed elegant approaches for the study of mitochondria based on the insertion into the mtDNA of nonrespiratory genes like ARG8 m . This gene is a mitochondrial version of the nuclear ARG8 gene encoding a mitochondrial protein involved in arginine biosynthesis. In this study, we have used ARG8 m to better understand why and how a lack in ␦ subunit destabilizes the mtDNA. To this end, we have inserted ARG8 m into a noncoding region of the mitochondrial genome (i.e. keeping intact all of the genes that normally reside in yeast mitochondria). Thus, in a ⌬arg8 nucleus and in the absence of external arginine, this mtDNA is required not only for respiration but also for arginine biosynthesis. With this system, we clearly show that the ATP synthase proton translocating activity is lethal to yeast cells missing the ␦ subunit and demonstrate that the Ϫ /°mutation is a suppressor allowing survival of ⌬␦ yeast. This is the first report of what we have called a "petite obligate" mutant. We discuss the results in relation with potential mechanisms regulating the assembly of the ATP synthase.

MATERIALS AND METHODS
Strains, Media, and Genetic Techniques-The S. cerevisiae strains used are listed in Table 1. Escherichia coli XL1-Blue strain (Stratagene) was used for the cloning and propagation of plasmids. Complete glucose (YPGA), galactose (YPGALA), or glycerol (N3) and minimal media for growing yeast were prepared as described in Ref. 22. The yeast sporulation medium and the procedure for converting yeast into °strains by ethidium bromide treatment have been described (27). Yeast transformation (28), crosses, and tetrad dissection (29) were performed as described previously.
Construction of a COX2-ARG8 m Transcriptional Unit and Its Insertion Upstream of COX2 in the mtDNA-To create an ARG8 m construct that could direct integration into intergenic regions of the mtDNA, we took advantage of the EcoRI site engineered on plasmid pPT24, 285 bases upstream of the COX2 start codon (i.e. upstream of the promoter) (23). Using standard cloning procedures and two-step PCR strategies (24) with pPT24 (containing COX2) and pDS24 (containing ARG8 m ) (20) as templates, we constructed an EcoRI cassette containing the ARG8 m open reading frame precisely flanked on the 5Ј side by 143 bp of the COX2 promoter region and on the 3Ј side by 119 bp of the COX2 terminator. This ARG8 m cassette was inserted into the EcoRI site of pPT24 in the forward orientation to yield plasmid pSDC10. The initial COX2 locus of pPT24 (from 485 bases upstream to 2015 bases downstream of the start codon) is therefore separated in two parts in pSDC10, the region Ϫ485 to Ϫ285 before the first EcoRI site and the region Ϫ285 to ϩ2015 after the ARG8 m cassette and the second EcoRI site (see Fig.  1A). The plasmid pSDC10 was introduced by co-transformation with the nuclear selectable LEU2 plasmid pFL46 into the °strain DFS160 by microprojectile bombardment using a biolistic PDS-1000/He particle delivery system (Bio-Rad) as described (25). Mitochondrial transformants were identified among the Leu ϩ nuclear transformants by their ability to produce both respiring and arginine-prototrophic diploid clones when mated to the nonrespiring and arginine auxothrophic NB40-3C strain, bearing deletions in both the mitochondrial COX2 gene (cox2-62) and in the nuclear ARG8 gene. The DNA recombination events leading to the simultaneous integration of ARG8 m and COX2 in NB40-3C mtDNA are illustrated in Fig. 1A. One respiratory growthcompetent and arginine-prototrophic clone (called SDC12) was retained for further analyses, and the ARG8 m integration was verified molecularly by Southern analysis and sequencing. A control diploid strain, SDC15, isogenic to SDC12 but carrying the wild-type mtDNA was constructed by crossing DFS160 with NB80, the COX2 equivalent of NB40-3C (26).
Biochemistry and Cell Biology Techniques-Mitochondrial protein synthesis analysis was as described in Ref. 48. Oxygen consumption rates were measured with a Clark electrode in the growth medium as described previously (31). SDS-PAGE was done according to Laemmli (32). Western blot analyses were performed as described previously (33). Polyclonal antibodies raised against yeast Atp4p, subunit ␦, Atp9p, and Atp6p were used after 1:10,000 dilutions, and those against Arg8p were used after a 1:2000 dilution. Nitrocellulose membranes were incubated with peroxidase-labeled antibodies at a 1:10,000 dilution and revealed with the ECL reagent of Amersham Biosciences. Epifluorescence microscopy of DASPMI-stained cells was carried out with a Leica DMRXA microscope fitted with a ϫ100 immersion objective and a standard fluorescein isothiocyanate filter.

Conversion of S. cerevisiae into a Conditional Petite Negative Yeast with an Intact Respiratory Capacity by Inserting ARG8 m into a Noncoding Region of the mtDNA
Acetylornithine aminotransferase (Arg8p) is a nuclear encoded mitochondrial protein involved in ornithine and arginine biosynthesis (34,35). Although it is normally synthesized in the cytosol and then imported into mitochondria, Fox and co-workers (20, 36 -38) have shown that Arg8p can be synthesized as well directly inside the mitochondrion from a recoded gene, ARG8 m , inserted into the mtDNA. In the present study, we have inserted ARG8 m in the intergenic region upstream of the COX2 gene, using a protocol described by Mireau et al. (39). The ARG8 m open reading frame was flanked with the 5Ј-and 3Ј-untranslated regions of COX2 and then integrated into the mtDNA by homologous recombination (see "Materials and Methods" and Fig. 1A).
We have isolated in this way a diploid strain (SDC12) with the expected integration of ARG8 m in mtDNA and homozygous for a null mutation in ARG8 (⌬arg8/⌬arg8 [ ϩ ARG8 m ]). SDC12 accumulated normal levels of Arg8p (Fig. 1B), grew well on media devoid of arginine, and was respiratory competent (Fig. 1C ). As expected, arginine was required for SDC12 to grow in the presence of ethidium bromide, an intercalating agent inducing the loss of mtDNA (Fig. 1C ).
The mtDNA in SDC12 showed a good stability. Indeed, after about 15 generations in complete 10% glucose (i.e. conditions where arginine biosynthesis and oxidative phosphorylation are not required for growth), over 98% of the cells were still respiratory competent and arginine-prototrophic. A good mtDNA stability was also observed in SDC12 meiotic segregants.
In SDC12, the ARG8 m gene is under control of the COX2 5Ј-untranslated region. Therefore, ARG8 m expression should depend on Pet111p, a nucleus-encoded Cox2p translational activator (40,41). To make sure FIGURE 1. Integration of ARG8 m upstream of COX2 into mtDNA allows wild-type accumulation of Arg8p and arginine prototrophy while retaining respiratory competence. A, a plasmid (pSDC10) containing ARG8 m flanked by COX2 expression sequences and inserted into the engineered EcoRI site upstream of the wild-type COX2 gene (hatched box, upstream promoter; box with vertical stripes, downstream terminator) was constructed and introduced by biolistic transformation into °mitochondria. The resultant synthetic Ϫ was crossed with the ϩ strain NB40-3C, which bears a deletion in COX2 (represented by a dotted line) to yield the diploid strain SDC12, in which ARG8 m is inserted upstream of COX2 in the mtDNA. B, mitochondria were prepared from the wild type yeast strain MC1 (ARG8), strain SDC15 lacking ARG8 and containing wild type mtDNA (⌬arg8), and strain SDC12 (⌬arg8 ARG8 m ). The mitochondrial proteins (20 g) were electrophoresed in SDS-polyacrylamide gel, transferred onto nitrocellulose and probed with antibodies against Arg8p. C, cells from SDC12 (⌬arg8 ARG8 m ) and SDC15 (⌬arg8) were spotted onto the indicated media and incubated 5 days at 28°C. Functional mtDNA Is Lethal in the Absence of ␦-F 1 Subunit that this was the case, an SDC12 meiotic segregant (SDC12-4B) was crossed with a °⌬pet111 strain (NB151-3C ⌬arg8 pet111::LEU2), to give SDC3 (⌬arg8/⌬arg8 ⌬pet111/ϩ [ ϩ ARG8 m ]). All the spores from SDC3 are necessarily ⌬arg8 and should therefore not grow in media lacking arginine if they cannot express ARG8 m . As expected, the ascii usually contained two spores, both respiratory growth-deficient and leucine-prototrophic, as a result of the pet111 deletion (not shown). These spores were also arginine-auxotrophic. Following their transformation with a plasmid-borne wild type PET111 gene, respiratory competence and arginine prototrophy were recovered, demonstrating the requirement for Pet111p to express ARG8 m .

Loss of Regulation of the ATP Synthase Caused by Lack of the ␦ Subunit Is Lethal
Elimination of the ␦ subunit gene in S. cerevisiae always results in 100% Ϫ /°populations, and evidence suggests that this conversion into petites is caused by an uncoupling of the ATP synthase (13)(14)(15)(16)(17). Below, we describe experiments we performed with the ARG8 m system to determine how this ATP synthase defect impairs mtDNA maintenance.
⌬␦/ϩ Yeast Is Genetically Unstable-We first aimed to analyze the segregation of the ␦ subunit deletion (atp16::KanMX4) and ARG8 m markers upon sporulation of SDC13, a diploid heterozygous for the ␦ deletion, homozygous for ⌬arg8, and containing the ARG8 m mtDNA. It has been reported that diploid strains with a heterozygous mutation in the ␦ subunit gene (i) have an increased propensity to produce Ϫ /°p etites (30 -40%), (ii) grow slowly on respiratory substrates, and (iii) exhibit partially uncoupled mitochondria, indicating that the reduced ␦ subunit gene dosage from 2 to 1 has semidominant negative effects (15). To see whether such defects occurred also in our ⌬␦/ϩ SDC13 strain, SDC13 cells were plated for single colonies and incubated for 6 days on glucose plates (Fig. 2). Very small size colonies, probably corresponding to Ϫ /°cells were observed, but in a rather limited proportion (less than 10%). Two other distinct types of colonies were observed. One (named SDC13-G), representing about 20% of the population, consisted of grande colonies like those formed by the corresponding wild type (SDC12). The other type (SDC13-I), which was the most frequent (70%), consisted of colonies with an intermediate size, suggesting a reduced but not complete loss of respiratory capacity. In addition, these colonies had a scalloped shape. This trait is typical of strains with an increased propensity to produce Ϫ /°petites (47), but it could also reflect a reduced cell viability. When SDC13-I subclones were plated again for single colonies, a similar colonial heterogeneity was observed, whereas SDC13-G subclones gave essentially grande colonies. Similar observations were made with other, genetically independent, SDC13 isolates (all checked by Southern blot).
Sporulation and Tetrad Analysis of ⌬␦/ϩ ARG8 m Yeast-One SDC13-I subclone and one SDC13-G subclone were sporulated, and tetrads were dissected on rich glucose plates. In the case of SDC13-I, a rather large proportion of the tetrads (42%, n ϭ 40; see Table 2) were incomplete with only two or three viable spores. A lower proportion of incomplete tetrads (20%, n ϭ 20, Table 2) was obtained from SDC13-G. A number of complete tetrads showing a correct segregation of the mating types was analyzed, 11 for SDC13-G and nine for SDC13-I. Geneticin resistance segregated 2:2 in all SDC13-G tetrads. However, only two tetrads showed the expected co-segregation of Geneticin resistance with respiratory growth deficiency. In the nine other SDC13-G tetrads, all spores were respiratory competent. In the case of SDC13-I, four tetrads showed the expected 2:2 co-segregation of Geneticin resistance and respiratory deficiency. In three other tetrads, all spores grew on glycerol and were Geneticin-sensitive. In the remaining two tetrads, Geneticin resistance segregated 2:2, but all spores were respiratory competent.
All spores from SDC13 are necessarily ⌬arg8 and depend therefore on their capacity to remain ϩ and to express ARG8 m in order to proliferate in media lacking arginine. With no exception, all of the spores that were at the same time respiratory deficient and Geneticin-resistant, thus presumably deleted for the ␦ subunit, were arginine-auxotrophic. Test crosses with a °tester revealed that the early progenies of these spores were entirely composed of Ϫ /°cells. Not surprisingly, all of the spores that were respiratory competent were able to grow in media lacking arginine.
Tetrads from SDC13-I were also dissected on a glucose medium lacking arginine. Many were incomplete, but this was seen as well, although to a lesser extent, with ascii from the wild type SDC12 strain. It might be that germination on minimal medium is less efficient. However, significantly, none of the SDC13-I spores that germinated in the absence of external arginine was ⌬␦.
Altogether, the results of this first set of experiments indicated that ⌬␦ cells were strictly unable to propagate/express the mtDNA or that they were not viable when they contained functional mtDNA. The

TABLE 2
Tetrad analyses of heterozygous ⌬␦/؉ yeast SDC13-G and SCD13-I are subclones of SDC13, a diploid heterozygous for deletion of the ␦ subunit gene by the KanMX4 gene conferring resistance to Geneticin (G418). These strains and the corresponding wild type diploid strain (SDC12) were sporulated, and ascii were dissected on rich glucose plates. Gly ϩ , the ability of the meiotic segregants to grow on glycerol medium; Gly Ϫ , respiratory growth deficiency; G418 r , Geneticin resistance; G418 s , geneticin sensitivity.

Strain
Phenotypes in complete tetrads Number of analyzed tetrads 4 Gly ؉ G418 s 2 Gly ؊ G418 r 2 Gly ؉ G418 s 2 Gly ؉ G418 r 2 Gly ؉ G418 s Gly ؉ G418 r erratic behavior of ⌬␦/ϩ yeast and the poor recovery of ⌬␦ spores after germination will be discussed under "Discussion." In the Absence of Atp4p, a Lack of ␦ Subunit Fails to Destabilize the Mitochondrial Genome-The primary effect of a lack in ␦ subunit is an uncoupling of the ATP synthase (15,16). The absence of ATP synthesis in the F 1 is probably not the cause of the instability of the mtDNA in ⌬␦ yeast, since many other mutations impairing oxidative phosphorylation do not have major effects on the mtDNA. More reasonably, the uncoupled F 0 is considered responsible for the loss of the mtDNA. We explored this possibility by assaying whether a nuclear mutation that inactivates the ATP synthase proton channel can suppress the deleterious effects on the mtDNA caused by elimination of the ␦ subunit gene. Although the two subunits (Atp6p and Atp9p) forming the core of the ATP synthase proton channel are encoded by the mtDNA, a number of nuclear mutations are known to inactivate this channel. This is the case with a null mutation of the nuclear ATP4 gene, which encodes a subunit of the ATP synthase peripheral stalk (42). This element connects, via its periphery, the ␣ 3 ␤ 3 subcomplex of F 1 to the membrane domain of the ATP synthase (Fig. 4A). The peripheral stalk is believed to function like a stator, preventing the ␣ 3 ␤ 3 particle from rotating together with the central stalk during catalysis. In the absence of Atp4p, Atp6p can still be synthesized but fails to accumulate, whereas the ATP synthase F 1 component is unaffected (42) (see also Fig. 3). Presumably, Atp4p is required for the insertion of Atp6p in the ATP synthase, and unassembled Atp6p proteins are rapidly degraded by mitochondrial proteases.
In order to determine whether a lack of Atp4p enables ⌬␦ yeast to remain ϩ and viable, we constructed a diploid strain (SDC17) heterozygous for ⌬␦ and ⌬atp4, homozygous for ⌬arg8, and containing the ARG8 m mtDNA (⌬␦/ϩ ⌬atp4/ϩ ⌬arg8/⌬arg8 [ ϩ ARG8 m ]). All spores from SDC17 are necessarily ⌬arg8 and therefore depend on their capacity to express ARG8 m to grow on a glucose medium lacking arginine. The spores carrying both the ⌬atp4 and ⌬␦ mutations were found to be arginine-prototrophic. On the contrary, as expected, the ⌬␦ spores that were ATP4 ϩ could grow only in the presence of arginine because of the inability of ⌬␦ yeast to grow in a ϩ state (see above).
To ascertain that inactivation of ATP4 was actually responsible for the ability of the ⌬atp4 ⌬␦ meiotic segregants to remain ϩ and viable, we performed a complementation test with a plasmid-borne wild type ATP4 gene. A ⌬␦ ⌬atp4 clone was transformed with this plasmid or with the corresponding empty vector (pRS313) and then plated for single colonies on plasmid-selective (histidine minus) glucose media containing or not containing arginine. The arginine supplement was not required to grow ⌬atp4 ⌬␦ cells transformed with the empty vector. On the contrary, ⌬atp4 ⌬␦ cells transformed with ATP4 could produce colonies only in the presence of arginine (Fig. 4B), and these proved to be exclusively composed of Ϫ /°cells as revealed by crossings with a wild type °strain (Fig. 4C). These data demonstrated that elimination of Atp4p suffices to allow ⌬␦ yeast to remain ϩ and viable.
A Lack in Atp4p Rescues ⌬␦ Yeast Viability by Preventing an F 0 -mediated Collapsing of the Mitochondrial ⌬⌿-A lack in Atp4p is assumed to allow ⌬␦ yeast to remain ϩ and viable by preventing the assembly of  (16). On the right is a hypothetical assembly when both subunit ␦ and Atp4p are missing. A lack in Atp4p results in proteolytic degradation of Atp6p (42) (see also Fig. 3), whereas Atp8p, Atp9p, and F 1 are still assembled (42); protons can no longer be transported through the complex. B, a ⌬arg8 ⌬␦ ⌬atp4 [ ϩ ARG8 m ] strain (SDC17-31B) was transformed with a plasmid-borne ATP4 gene or the corresponding empty plasmid (pRS313). The transformation mixtures were plated equally onto glucose medium containing (ϩArg) or not containing (ϪArg) arginine and onto a glucose medium devoid of arginine but supplemented with oligomycin (ϪArg ϩ Oligo) and incubated 5 days at 28°C. All of the media were devoid of histidine (ϪHis) for plasmid selection. C, SDC17-31B cells transformed either with the ATP4-containing or the corresponding empty plasmid were plated for single colonies onto a glucose medium containing arginine (ϩArg) or a glucose medium lacking arginine but supplemented with oligomycin (ϪArg ϩ Oligo), as in B. The colonies were replica-mated with a °tester strain (KL14-4A/60) and then replicated onto rich glycerol (N3) plates. Photographs were taken after a 4-day incubation at 28°C. the ATP synthase proton channel (as discussed above). Direct evidence for this assumption was sought with a regulatable strain (SDC33) in which ␦ subunit and Atp4p are expressed, from a doxycycline-repressible (tetO) and a galactose-inducible (GAL10) promoter, respectively (Fig. 5A).
We first analyzed SDC33 by fluorescence microscopy with DASPMI, a mitochondrial ⌬⌿-sensitive probe (Fig. 5D). As expected, the mitochondria were normally stained as a continuous reticulate network in SDC33 cells grown in a galactose minus doxycycline medium (i.e. conditions in which both ␦ subunit and Atp4p, and hence the whole ATP synthase, can be synthesized). The mitochondria were also stained when both subunits were repressed (glucose ϩ doxycycline). In these conditions, however, the mitochondria appeared fragmented presumably because of the change in carbon source and/or a weaker energization of the inner membrane due to the loss of mitochondrial oxidative phosphorylation. The SDC33 cells that were grown in glucose plus doxycycline were transferred and incubated for 3 h in galactose ϩ doxycycline, conditions in which Atp4p can be synthesized while the ␦ subunit gene remains repressed. In these conditions, the mitochondrial network could not be visualized, indicating a collapsing of the mitochondrial ⌬⌿. That collapsing was not observed when oligomycin was added prior to the transfer of SDC33 to galactose plus doxycycline. Thus, the lack of mitochondrial staining seen in the absence of oligomycin was apparently caused by an F 0 -mediated proton transport.
The results of these mitochondrial ⌬⌿ analyzes were corroborated by oxygen consumption assays performed in whole cells (Fig. 5D). In the absence of both ␦ subunit and Atp4p (Fig. 5C, Glucose ϩ Doxycycline), the respiration rate could be stimulated more than 3-fold by the protonionophore CCCP, indicating a normal permeability of protons across the inner membrane. When the cells were transferred into galactose ϩ doxycycline (i.e. to induce Atp4p while keeping the ␦ subunit repressed), the respiration became over the time gradually less sensitive to CCCP to FIGURE 5. A lack in Atp4p prevents an F 0 -mediated collapsing of the mitochondrial ⌬⌿ when the ␦ subunit is missing. SDC33 expresses Atp4p and subunit ␦, respectively, from a galactose-inducible promoter and a doxycycline-repressible promoter, respectively. SDC22 is a wild type strain expressing both genes from their own promoter. A, SDS-PAGE and Western blot analysis with antibodies against Atp4p and subunit ␦ of total cellular proteins extracted from SDC22 and SDC33 grown in the indicated conditions. B and C, SDC33 cells grown in glucose ϩ doxycycline were shifted to galactose ϩ doxycycline. At the indicated time after the shift, the cells were assayed for Atp4p accumulation by Western blot (B) and for oxygen consumption in the growth medium (C ). During the course of oxygen consumption measurements, 3 M CCCP was added to the 10 7 cells/ml suspension. For each experiment, the oxygen consumption rates are indicated as percentages of the maximal value obtained after the addition of CCCP. D, whole cells of strain SDC33 grown under the indicated conditions were incubated for 15 min with a 5 mM concentration of the mitochondrial membrane potential probe DASPMI and examined by fluorescence microscopy (Dox, doxycycline; Oligo, oligomycin). reach a nearly CCCP-insensitive state 3 h after the galactose switch. Immunoblottings showed that this progressive uncoupling of mitochondrial respiration was paralleled by a gradual accumulation of Atp4p in the cells (Fig. 5C). These experiments showed that a lack in Atp4p prevents an F 0 -mediated collapsing of the mitochondrial ⌬⌿ when ␦ subunit is missing. Consistent with this, ⌬␦ cells, reconstituted by transforming ⌬␦ ⌬atp4 [ ϩ ARG8 m ] cells with ATP4, remained ϩ and viable in the absence of external arginine when the growth medium was supplemented with oligomycin (Fig. 4B).

DISCUSSION
Previous work has shown that yeast lacking the ATP synthase ␦ subunit gene (⌬␦) produced populations containing Ϫ /°cells exclusively (13,14). Some evidence already suggested that ⌬␦ yeast is unable to maintain intact ϩ mtDNA because of the loss of regulation of the ATP synthase in the form of a passive proton transport through the F 0 (i.e. not coupled to ATP synthesis in the F 1 ) (15)(16)(17). However, it was unclear why and how this ATP synthase defect compromised mtDNA maintenance.
To further investigate this question, we have introduced into the yeast mtDNA a mitochondrial version (ARG8 m ) of the nuclear ARG8 gene encoding a mitochondrial protein involved in arginine biosynthesis (20). As a consequence, the mtDNA became required not only for respiration but also for arginine prototrophy (in a ⌬arg8 nuclear context). We used this nuclear-to-mitochondria gene relocation to gain a better understanding of why a lack in ␦ subunit destabilizes the mtDNA, according to the following rationale.
The mitochondrial genome is rather unstable in S. cerevisiae, even in wild type strains, which always produce a few percent Ϫ /°cells. This may account for the relative ease with which many different types of mutations can increase the production of Ϫ /°cells, up to 100%, including mutations of components having no direct connection to the mitochondrial genome (21). Indeed, when oxidative phosphorylation is impossible, whether there is or is not functional mtDNA, as in ⌬␦ yeast, ϩ and Ϫ /°cells might have a similar multiplication rate by using glycolytic ATP. Thus, because of the irreversibility of the Ϫ /°mutation, a modest increase in the frequency of this mutation might suffice for a complete disappearance of ϩ cells in a few generations only. In the ARG8 m system and in the absence of external arginine, Ϫ /°cells cannot divide and are therefore expected to accumulate much less rapidly, provided that the systems in charge of mtDNA propagation/expression are still active and that retention of mtDNA is not detrimental to cellular viability.
In a first set of experiments, we analyzed the segregation of the ␦ subunit deletion and ARG8 m markers upon sporulation of a heterozygous ⌬␦/ϩ yeast strain (SDC13). Consistent with a previous study (15), we found that this strain had a reduced respiratory capacity (also observed in Ref. 50), presumably because of a partial mitochondrial uncoupling. In addition, we noticed the accumulation (20%) in SDC13 glucose cultures of "revertants" (called SDC13-G) with a good respiratory capacity. Upon sporulation, these revertants segregated the ␦ subunit deletion gene marker 2:2, but most (Ͼ90%) of the spores were respiratory competent. SDC13 subclones with a reduced respiratory capacity (SDC13-I) produced many aberrant tetrads as well, with two or three viable spores and a large excess of respiratory competent spores. Similar observations have been made by J. Velours and M.-F. Giraud. 7 A possible explanation is that the partial mitochondrial uncoupling resulting from the reduced dosage of the ␦ subunit gene from 2 to 1 is so deleterious to the cell that it favors a rapid selection of mutations restor-ing the integrity of mitochondria, maybe by duplications of the remaining ␦ subunit gene or correction of the disrupted ␦ subunit gene by gene conversion.
Whatever the explanation for the genetic drift and aberrant meiotic segregation properties of ⌬␦/ϩ yeast, a clear result emerged from these experiments. Indeed, in all tetrads exhibiting the expected segregation of the ␦ subunit deletion marker and respiratory deficiency, the ⌬␦ spores were arginine-auxotrophic, and their early progenies were composed exclusively of Ϫ /°cells. In addition, we could never observe germination of ⌬␦ spores on media lacking arginine. These data indicate that ⌬␦ yeast is strictly unable to propagate, or survive the presence of, functional mtDNA.
In a second set of experiments, we found that mtDNA maintenance was possible in ⌬␦ yeast when the ATP synthase F 0 component was inactivated either chemically by oligomycin or genetically by a nuclear mutation (⌬atp4) that prevents insertion of Atp6p in the ATP synthase. One possible explanation (model 1) is that the proton translocating activity of the F 0 is lethal to yeast lacking the ATP synthase ␦ subunit. A reasonable view is that the very strong weakening of the mitochondrial electrical potential (⌬⌿) observed after a block in the synthesis of the ␦ subunit (16) may result in a general defect in mitochondrial biogenesis. Consequently, essential reactions taking place in mitochondria can no longer proceed, leading to the death of the cell. Consistent with this idea, a recent study in yeast has shown that a collapsing of the mitochondrial ⌬⌿ is followed by the onset of mitophagy and reduced cell viability (43). Accordingly, as previously discussed (16,17), the Ϫ /°mutation can be viewed as a rescuing event allowing survival of ⌬␦ yeast. Indeed, after the loss of the mtDNA, the ATP synthase proton channel cannot be synthesized, and consequently a lack of ␦ subunit can no longer dissipate the mitochondrial ⌬⌿. The inner membrane is believed to be sufficiently energized in Ϫ /°⌬␦ yeast by the electrogenic activity of the ADP/ATP translocase (13). Thus, according to model 1, conversion of ⌬␦ yeast into petites would preferentially proceed by selection of preexisting Ϫ /°cells undergoing the loss of the ␦ subunit gene.
Alternatively (model 2), the F 0 activity would not kill ⌬␦ cells but would induce petites at high frequencies (e.g. by preventing a sufficient transport into mitochondria of the systems needed for mtDNA propagation. Such a mechanism seems to occur in strains harboring subunit deletions, which induce many petites (70%), although ϩ cells lacking the subunit are viable (18).
The experiments performed with ATP4 are consistent with a petite selection scheme. Indeed, not a single ϩ cell was recovered by crossing with a °tester hundreds of colonies formed (in the presence of arginine) by ⌬␦ ϩ ⌬atp4 cells transformed with ATP4. This observation argues against model 2, since in that case, a number of the initial ϩ cells would probably have been recovered in the test crosses. Probably, the reconstituted ⌬␦ colonies arose from ⌬␦ ϩ ⌬atp4 cells that were already Ϫ /°before the transformation with ATP4 (Fig. 4B shows that about 70% of the plated transformants were not ϩ ).
Normally, ϩ mtDNA segregates extremely efficiently into haploid spores. Assuming that it is also the case with ⌬␦/ϩ cells, according to model 1, ␦ deletion spores are expected to be killed. Even if they were largely underrepresented, viable ⌬␦ spores could be recovered. However, because of the rapid genetic drift of ⌬␦/ϩ yeast (see above), it is difficult from these experiments to appreciate to what degree petite induction contributes to the conversion of ⌬␦ yeast into populations lacking functional mtDNA.
Mutations in the ␣-F 1 or ␤-F 1 subunit or in Atp11p, an F 1 assembly factor, have also been shown to suppress the formation of Ϫ /°cells in strains deleted for the ␦ subunit (14). These genetic observations are in accordance with our conclusion that that the loss of mtDNA in ⌬␦ yeast is caused by an uncoupling of the ATP synthase. Indeed, as we have shown, the F 0 is absent in a yeast fmc1 mutant unable to assemble the ␣ 3 ␤ 3 subcomplex of F 1 (48) and in a mutant lacking the ␤ subunit (49). Thus, in cells lacking both the ␦ subunit and the catalytic unit of F 1 , mtDNA maintenance is tolerated because it cannot result in a collapsing of the mitochondrial ⌬⌿. Direct evidence has been provided earlier supporting this explanation (15).
The loss of the ␦ subunit gene has been reported to be viable in Kluyveromyces lactis, a "petite negative" yeast, which, despite a good fermenting capacity, does not survive the Ϫ /°mutation (19). Thus, contrary to S. cerevisiae, the mitochondrial genome is tolerated in K. lactis when the ␦ subunit is missing. This is a very interesting observation, indicating that elimination of the ␦ subunit does not have the same consequences in K. lactis than in S. cerevisiae. Indeed, as we have shown, the ␦ subunit is not required in S. cerevisiae for the assembly of the remaining ATP synthase subunits, but the resulting partial F 1 F 0 complexes can still transport protons and dissipate the mitochondrial ⌬⌿ (16). As a result, in normal conditions, the synthesis of the ␦ subunit must be tightly regulated in S. cerevisiae to avoid any deficit in this protein. The ability of a ⌬␦ mutant of K. lactis to remain ϩ and viable suggests that a lack in the ␦ subunit does not prevent the maintenance of a sufficient mitochondrial ⌬⌿. It might be that in K. lactis, the ATP synthase proton channel cannot assemble or function when the ␦ subunit is missing. Thus, it appears that different mechanisms may exist among species to avoid accumulation of aberrant ATP synthase assemblies that can uncouple the mitochondrion.
Within the ATP synthase, the ␦ subunit interacts with the subunit to couple the ␣ 3 ␤ 3 ␥ 1 subcomplex of F 1 to the enzyme proton channel (4,5). In S. cerevisiae, a null mutation of the ␥ subunit always results in 100% petites (14), and there is evidence that a failure in the expression of this subunit is followed by an F 0 -mediated proton leak across the inner mitochondrial membrane (15). Thus, the Ϫ /°mutation might be for ⌬␥ yeast, as it is for ⌬␦ yeast, a rescuing event allowing cell survival. Interestingly, cultures of ⌬ yeast usually contain 30% ϩ cells showing that mtDNA maintenance is not lethal to ⌬ yeast (18). Yet, mitochondria isolated from a ⌬ culture show defects in ⌬⌿ maintenance identical to those seen when the ␦ subunit is lacking. Experiments using a regulatable subunit gene are ongoing to clarify this apparent contradiction.
Mutations of subunits belonging to ATP synthase components other than the central stalk were also found to have strong negative effects on the stability of the mtDNA (see Ref. 21 for a review). For example, a null mutation of the peripheral stalk Atp4p subunit usually resulted in 70% Ϫ /°cells (42). No F 0 -mediated proton leak was detected in ⌬atp4 mitochondria, which is not very surprising, since the Atp6p subunit cannot insert when Atp4p is missing (42) (this study). Thus, something else than an ATP synthase proton leak must be responsible for mtDNA destabilization in the ⌬atp4 mutant.
A substantial loss of mtDNA was also observed for mutations of the ATP synthase having no major incidence on its proton translocating and ATP synthesis activities, such as null mutations of subunit e or g (44,45). Each of these mutations results in about 40% Ϫ /°cells after growth in a lactate medium, and probably more would accumulate after growth on a fermentable carbon source. Abnormal mitochondrial morphologies characterized by numerous digitations and onion-like structures presumably corresponding to uncontrolled biogenesis and/or folding of the inner membrane were found in cells lacking subunit e or g (45). Such morphological defects may alter the association of the mtDNA to the inner membrane and/or transmission of the mtDNA to daughter cells. The importance of the mitochondrial structure for mtDNA maintenance is well illustrated by studies of mutants defective in systems directly controlling mitochondrial structure such as Fzo1p, whose absence leads to fragmentation of the mitochondrial network followed by a massive conversion into cells lacking mtDNA (46).
The present study, together with a report from Mueller and co-workers (14) provides a clear demonstration of a mutation causing a massive loss of mtDNA with no direct effect on the components in charge of the maintenance and inheritance of the mitochondrial genome. This is an interesting observation in view of the need for mtDNA maintenance of a huge number of mitochondrial proteins (Ͼ200) with no obvious link to the mtDNA (21). The ϩ mtDNA nutritional base selection system we used in this study may help to clarify the connections between the immediate effects of mutations in these various proteins and mtDNA instability.
The observations reported in this study provide the basis for a genetic screen targeted on the F 0 component of the ATP synthase. Indeed, as we have shown, in order to survive, ⌬␦ yeast needs to inactivate the ATP synthase proton channel. Thus, in a ⌬␦ context, F 0 -inactivating mutations can be positively selected. In the ARG8 m system, the most frequent rescuing event, the Ϫ /°mutation, can be eliminated from the screen, by selecting the isolates in an arginine-lacking medium. We thus now have in hand a very powerful genetic system, which may help to better define the structural determinants essential for the activity of the ATP synthase proton channel and facilitate the study of the biogenesis of this channel, which is still largely unknown.