Two independent screens of the yeast deletion library have previously revealed two partially overlapping sets of pet genes. By plating the homozygous diploid yeast deletion library on media containing glycerol as a carbon source, Dimmer et al. 14 identified 341 deletion mutants that were unable to grow. In a very similar approach, Luban et al. 17 identified a set of 355 respiratory-deficient clones by screening the MATa yeast deletion library. While about two-thirds of the mutants in each screen were found to be respiratory-deficient also in the other screen, a surprisingly large number of mutants were isolated only once 17. It seems unlikely that this is due to differences in the genetic background, because both screens have been conducted in largely isogenic strains, BY4743 and BY4741 18. Here, we screened the MATα deletion library (BY4742 background) to obtain a third set of respiratory-deficient mutants. This was then compared with the data obtained by Dimmer et al. 14 and Luban et al. 17. The MATα deletion library contained 319 mutants that were unable to grow on glycerol-containing medium (Additional data file 1). Of these, 176 are common to all three sets of pet genes (Figure 1a). In the following we will refer to these genes as highly penetrant pet genes. 125 genes have been identified in two of three screens, and 237 genes have been identified only once (pet genes unique to this study are listed in Additional data file 2). Nineteen additional pet genes (not included in the set of 176 highly penetrant pet genes) were only covered by one or two libraries. Based on data from the Saccharomyces Genome Database 19 and manual annotation, we grouped all genes according to their frequency of occurrence in pet screens and the intracellular location and function of their gene products (Additional data file 3).

Strikingly, 129 out of the 176 pet genes found in all three screens encode proteins known to be located in mitochondria, corresponding to 73.3% (Figure 1b; Additional data file 3). The fraction of genes encoding mitochondrial proteins was reduced to 52.1% for pet genes found in two of three screens, and as low as 14.7% for pet genes that were found only once (Figure 1b; Additional data file 3). This demonstrates a clear correlation of the penetrance of pet phenotypes with mitochondrial functions of the affected gene products. The majority of the 176 pet genes found in all libraries encode proteins devoted to maintenance and expression of the mitochondrial genome and assembly of the respiratory chain (Figure 1c; Additional data file 3). Thirteen open reading frames (ORFs) are unlikely to encode proteins, because they overlap with other known genes (Additional data file 3), reducing the number of protein-coding genes to 163.

Differences of growth behavior of strains taken from different versions of the deletion libraries could either reflect inherent properties of the mutant strains, they could be due to technical differences between the various screens, or they could mean that a given deletion in one collection is wrong (as it has occasionally been observed by us and others; for example, strains not correct in the MATα library include Δrpo41 lacking the mitochondrial RNA polymerase). We reasoned that incorrect mutants will be enriched among strains that showed respiratory competence in one screen but were respiratory-deficient in the two other screens because it is more likely that a specific phenotype is obscured rather than generated by chance. To test this, we checked the genotypes of 29 mutants taken from the MATα library by PCR. Nineteen randomly chosen mutants were tested that were respiratory-competent in the MATα library, but respiratory-deficient in the MATa and homozygous diploid library. Of these, six mutants (Δyal012w, Δybl038w, Δydl202w, Δydr268w, Δyor205c, and Δypl029w) contained exclusively the wild-type allele, seven mutants (Δydr231c, Δydr332w, Δyil036w, Δyjr090c, Δymr066w, Δypr047w, and Δypr124w) contained a mixture of deletion and wild-type alleles, and six mutants were found to have the correct genotype (Δyal047c, Δybr163w, Δydr323c, Δykl148c, Δyml081c-a, and Δyml129c). In addition, we tested ten mutants that showed a pet phenotype only in the MATα library, but not in the MATa and homozygous diploid library, and ten mutants, that showed a pet phenotype in all three screens. All mutants of the two latter groups were found to have the correct genotype. This means whenever a wrong deletion was detected, a pet phenotype was obscured by the presence of the wild-type allele, whereas all respiratory-deficient mutants tested were found to have the correct genotype. We conclude that several discrepancies of growth phenotypes can be ascribed to wrong genotypes that are present in the deletion library. However, the fact that a relatively large number of mutants with confirmed correct genotypes show differences in their growth behavior points to a pronounced phenotypic plasticity of pet mutants. Furthermore, the correlation of the penetrance of pet phenotypes with mitochondrial localization of gene products (Figure 1b) is a clear indication that the phenotypic variability is not only due to wrong deletions present in the mutant libraries, but also reflects biological processes.

Eight highly penetrant pet genes (YDR065w, YGR150c, YJL046w, YLL033w, YLR091w, YMR293c, YOR305w, and YPR116w) encode previously uncharacterized proteins, and earlier studies have revealed a respiratory-deficient phenotype for two additional ORFs of unknown function that were not covered by all three yeast deletion libraries, YNL213c 20 and YJL062w-a 21. We confirmed the identity of these mutant strains by PCR and named the genes RRG1 through RRG10 (for 'Required for respiratory growth') as the functional analysis described below proves that their products are novel factors required for respiratory growth.

We asked whether the respiratory-deficient phenotype observed for the 319 pet mutants isolated from the MATα deletion library is specific to glycerol metabolism or reflects a general lack of respiration competence. To test this, we plated the mutants also on complete media containing lactate or ethanol as sole carbon sources. The vast majority (305 strains, corresponding to 95.6% of the pet mutants) failed to grow on all non-fermentable carbon sources that were tested. Of the remainder, seven mutants showed a growth defect only on glycerol-containing medium, seven on glycerol or ethanol-containing media, and one mutant on glycerol or lactate containing media (Additional data file 4). As pet phenotypes are highly reproducible even on different carbon sources we conclude that our screen gives a largely accurate estimate of respiratory deficiencies in the MATα deletion library.

In order to define the genetic basis of respiratory deficiency, we subjected the complete set of 319 pet mutants isolated from the MATα deletion library to various functional tests (Figure 2). As a petite phenotype is often associated with the complete or partial loss of the mitochondrial genome 13, we first asked whether the pet mutants contain functional mtDNA. To test this, pet mutants were mated with a strain lacking the mtDNA polymerase Mip1. As the Δmip1 strain is [rho⁰] 22, resulting heterozygous diploid strains are able to grow on glycerol-containing medium only if functional mtDNA is provided by the pet mutant mating partner. We observed restoration of respiratory activity in 157 heterozygous diploid strains demonstrating that the parental pet strains possessed an intact mitochondrial genome. In contrast, 162 strains failed to grow on glycerol-containing medium after mating, suggesting that the parental pet mutants were [rho^-] or [rho⁰].

The complementation test with Δmip1 does not discern whether the protein encoded by the pet gene is obligatorily required for maintenance of mtDNA, or whether a functional mitochondrial genome had been spontaneously lost in the pet mutant during many generations of growth. To discriminate between these possibilities, we replenished cells with mitochondria containing a wild-type [rho⁺] mitochondrial genome by cytoduction. In brief, pet mutants were crossed with a [rho⁺] donor strain that carries a kar1 mutation to prevent karyogamy in the zygote. After counterselection against nuclear chromosomes of the donor strain, growth of the haploid progeny was assessed on glycerol-containing media. Restoration of respiratory activity after cytoduction was observed in 67 pet mutants, whereas 252 strains failed to grow on non-fermentable carbon sources.

Combining the results from the Δmip1 mating test and the cytoduction experiment allowed us to define four classes of pet mutants (Figure 3; Additional data file 5). Class I mutants were not rescued either by mating with Δmip1 or by cytoduction; class II mutants were rescued by mating with Δmip1 as well as by cytoduction; class III mutants were rescued only by mating with Δmip1, but not by cytoduction; and class IV mutants were rescued only by cytoduction, but not by mating with Δmip1. The basic properties of these classes are summarized in Table 1. In the following, the various classes of pet mutants are further examined (Figure 2).

The 118 class I mutants were [rho^-] or [rho⁰] and remained respiratory-deficient after introduction of functional mitochondria. This group of mutants is expected to include all components that are essential for maintenance of a [rho⁺] genome. In addition, we expected it to contain components deletion of which leads to a gradual loss of mtDNA and, at the same time, induces respiratory deficiency due to lack of functions not directly related to mtDNA maintenance. To discern between these possibilities, we subjected all class I mutants to various functional tests. First, we tested for the presence of mtDNA by DAPI (4',6-diamidino-2-phenylindole) staining immediately after cytoduction. Second, we tested growth on YPG medium after adaptation to the medium by pre-culture on YPG containing low amounts of glucose. And third, we tested mitochondrial protein translation activity by SDS-PAGE and autoradiography after labeling cycloheximide-treated cells with ³⁵S methionine.

Genes essential for maintenance of mtDNA were defined by the following criteria: At least 95% of the cells observed by DAPI staining after cytoduction were devoid of mtDNA and the remainder contained less than five mtDNA nucleoids per cell. This phenotype was observed after cytoduction in the Δmip1 mutant lacking the mtDNA polymerase and, therefore, is indicative of instantaneous loss of mtDNA. In addition, cells lacking genes essential for maintenance of mtDNA were expected to be unable to grow on YPG after adaptation to the carbon source, and they were unable to produce even trace amounts of mitochondria-encoded proteins. Sixteen mutants were identified that matched these criteria (Table 2). We propose that the gene products lacking in these mutants are particularly important for maintenance of mtDNA. As expected, this group includes several components known to be involved in mtDNA metabolism: the mtDNA polymerase Mip1 22; mtDNA helicases Hmi1 23 and Pif1 24; Apn1, a DNA repair protein active in the nucleus and mitochondria 25; and aconitase, Aco1, an enzyme of the citric acid cycle that has an additional role in mtDNA maintenance 26.

It has been observed that a block of mitochondrial protein synthesis leads to a rapid and quantitative loss of mtDNA 27. However, the reasons for this phenomenon are still unknown. Here, we observed instantaneous loss of mtDNA in cells lacking Mrpl37, Mtf1, Mtg2, Rsm24, and Slm5, which are all required for mitochondrial transcription or translation, and in a deletion mutant lacking the dubious ORF YKL091w, which overlaps with the MRPL38 gene encoding a mitochondrial ribosomal protein. Loss of mtDNA at a relatively high rate was also observed in several other class I mutants lacking components of the mitochondrial protein synthesis machinery. These findings underscore the importance of mitochondrial protein synthesis for maintenance of mtDNA. Moreover, rapid loss of mtDNA was observed in the Δatp4 mutant lacking ATPase subunit b. This is consistent with earlier observations 28; however, the molecular reasons are not understood 29. Also, Δpet100 mutants lacking a factor required for cytochrome c oxidase assembly showed rapid loss of mtDNA. As loss of mtDNA in Δatp4, Δmrpl37, Δmtf1, Δmtg2, Δpet100, Δrsm24, and Δslm5 occurs instantaneously (as rapid as in Δmip1) we consider it likely that replication and/or inheritance of mtDNA is actively suppressed in these strains. These results point to an active role of Atp4, Mrpl37, Mtf1, Mtg2, Pet100, Rsm24, and Slm5 in regulating mtDNA abundance in yeast mitochondria.

Other factors required for mtDNA inheritance are Mgm1, Doc1 and the newly identified protein Rrg5. Mgm1 is a dynamin-related protein required for mitochondrial genome maintenance by mediating mitochondrial fusion 303132. Doc1 is involved in cyclin proteolysis as a processivity factor required for the ubiquitination activity of the anaphase promoting complex (APC) 33. Intriguingly, Doc1 has been found in the mitochondrial proteome 634. Thus, it is tempting to speculate that it links mtDNA replication and/or inheritance to the cell cycle. The RRG5 gene (YLR091w) encodes a protein of unknown function. Its sequence does not show similarities to any characterized protein. As the Rrg5 protein has been localized to mitochondria 63435, we propose that it is a novel factor essential for maintenance of mtDNA.

In addition to class I mutants, 44 pet mutants were identified that were not complemented by mating with Δmip1 but could be rescued by cytoduction. These strains are able to maintain newly re-introduced mtDNA when they are grown on non-fermentable carbon sources (class IV; Additional data file 5). It is conceivable that these mutants have a tendency to spontaneously lose their mitochondrial genome when they are grown on fermentable carbon sources for longer times. This has been observed previously for Δmdm31 and Δmdm32 mutants that showed a pet phenotype in the screen performed by Dimmer et al. 14, but not in the screens performed by Luban et al. 17 and in the screen reported here. Freshly made Δmdm31 and Δmdm32 deletion mutants have been found to be able to maintain [rho⁺] mtDNA 36. However, mtDNA is not stably inherited and is gradually lost after several generations of growth in glucose-containing medium 36. To test this systematically for all class IV mutants, we replenished mtDNA by cytoduction and then passaged the strains in liquid YPD medium for 10 days to allow for loss of mtDNA. Presence or absence of mtDNA was assayed by DAPI staining immediately after cytoduction and after 10 days of replicative growth. In all strains, at least 90% of the cells contained mtDNA directly after cytoduction. Continued growth in glucose-containing medium led to increased loss of mtDNA in many mutants (Additional data file 6), suggesting that gradual loss of mtDNA accounts for the pet phenotype in many class IV mutants. Only few mutants maintained mtDNA as stably as the wild type (Additional data file 6). We consider it possible that these mutants require more generation times or special growth conditions to induce loss of mtDNA, or that these mutants rapidly accumulate mtDNA point mutations or deletions rendering the mitochondrial genome non-functional over time. Interestingly, 77% of the class IV mutants have not been found in the screens by Dimmer et al. 14 and Luban et al. 17, suggesting that many of the affected genes are only indirectly related to maintenance of respiratory activity.

Next, we asked which genes are required for mitochondrial protein synthesis. Mutants defective in this process are expected to be found in either class I or class III. Class I contains mutants that have lost their mtDNA as a consequence of blocked mitochondrial translation activity, whereas class III contains mutants that are defective in translation but maintain an intact mitochondrial genome. In order to be able to test for mitochondrial protein synthesis activity, we replenished wild-type mtDNA in class I mutants by cytoduction. After this treatment, mtDNA could be visualized by DAPI staining in 102 mutants, whereas 16 mutants lacking genes essential for maintenance of mtDNA immediately became [rho⁰] (see above; Table 2). For class III mutants, we reasoned that some strains might be unable to grow on medium containing glycerol as the sole carbon source because of synergistic effects of compromised mitochondrial function in combination with catabolite repression, which reduces the expression of genes required for respiration 8. Therefore, we first relieved catabolite repression in all class III mutants by growth on glycerol-containing medium supplemented with limiting amounts of fermentable carbon source (3% glycerol/0.1% glucose) before replicating the strains on glycerol-containing medium. After this treatment, 77 strains were able to grow on plates containing glycerol as the sole carbon source (Additional data file 7). We conclude that the gene products lacking in these mutants are dispensable for respiration.

Then, we tested mitochondrial translation in a total number of 159 deletion mutants (102 class I mutants with replenished mtDNA and 57 class III mutants unable to grow on glycerol-containing medium after adaptation to the carbon source). Strains were grown to logarithmic growth phase in medium containing fermentable carbon sources, before cytosolic translation was stopped by the addition of cycloheximide. Newly synthesized mitochondrial proteins were labeled with ³⁵S methionine, and cell extracts were analyzed by SDS-PAGE and autoradiography.

Mitochondrial translation products could not be detected in 88 mutants (Table 3). We conclude that these genes are required for mitochondrial protein synthesis. Encoded proteins include 39 subunits of the mitochondrial ribosome and several additional components required for mitochondrial transcription, translation or assembly of the respiratory chain 37. In addition, mitochondrial translation activity was absent in several mutants lacking proteins known to be required for mtDNA inheritance, such as Fzo1 3839, Mhr1 40, Msh1 41, or Mgm101 42. Supposedly, in these strains - and likely also in other class I mutants - the mitochondrial genome had been largely lost or damaged during growth of the strains in the time between cytoduction and the labeling reaction. It should be noted that strain-dependent effects might also play a role, because, for example, Δpet309 was observed to be completely translation-inactive here, whereas mitochondrial translation products could be observed when this mutant was constructed in the W303 genetic background 43. Five genes (RRG1, YGR102c, RRG2, RRG6, and RRG8) encode uncharacterized proteins, and two dubious ORFs (YDR114c and YNL184c) overlap with genes encoding mitochondrial ribosomal proteins. A possible role of Rrg1, Rrg2, Rrg6, and Rrg8 as novel components required for mitochondrial protein synthesis is discussed below.

Specific alterations of the pattern of newly translated mitochondrial proteins were observed in ten mutants (Figure 4 and Table 4). A role in the expression of specific mitochondria-synthesized proteins has already been described for Aep2 44, Cbs2 45, Mrs1 46, Mss51 47, Pet54 4849, and Pet494 50. We observed that the pattern of mitochondrial translation products was also altered in Δcoq3, Δcyc3, Δrrg10, and Δvma8 mutants. Coq3 is required for the biosynthesis of ubiquinone (coenzyme Q) in mitochondria 51. We observed that mutant cells show a strong reduction of Cox1 (Figure 4, lane 11). Cyc3 is the mitochondrial cytochrome c heme lyase that attaches the heme cofactor to apo-cytochrome c in the intermembrane space 52. Strikingly, mutant mitochondria show a strong reduction of Cox1 and cytochrome b and generate an additional protein band above Cox3 (Figure 4, lane 2), pointing to a role of Cyc3 also in the biogenesis of other mitochondrial proteins. Rrg10 is an uncharacterized mitochondrial protein that might play a specific role in the expression of the mitochondrial COX1 gene (Figure 4, lane 7), as discussed below. Cox1 and Atp6 are also reduced in the Δvma8 mutant lacking a subunit of the vacuolar H⁺ ATPase 53, suggesting that expression of these proteins is particularly sensitive to changes in cell metabolism (Figure 4, lane 4).

In sum, 61 respiratory-deficient mutants showed a wild-type-like mitochondrial translation pattern (Additional data file 8). We conclude that these genes are not essential for mitochondrial genome maintenance or mitochondrial protein synthesis. This group contains 32 genes encoding known mitochondrial proteins, many of which are required for assembly of the respiratory chain. Eighteen genes encode known extra-mitochondrial proteins, and 11 ORFs are uncharacterized. Five of the uncharacterized ORFs are unlikely to encode proteins because they overlap with known protein-coding genes, whereas six ORFs (YDL129w, YDL133w, YDL033w/RRG4, YNL213c/RRG9, YOL071w, and YOL083w) might encode novel proteins involved in maintenance of respiratory activity. Possible roles of Rrg4 and Rrg9 in this process are discussed below.

Half of the pet genes encoding extra-mitochondrial proteins are associated with vacuolar functions (Additional data file 8). Moreover, a surprisingly large number of genes encoding V-ATPase subunits are highly penetrant pet genes (Figure 1c; Additional data file 3). What might be the function of the vacuole in maintenance of respiratory activity in yeast? We suggest three possibilities. First, vacuolar functions in metabolite storage or in cytosolic ion and pH homeostasis 5455 might interfere with mitochondrial metabolism. Second, loss of V-ATPase activity has been reported to render cells hypersensitive to oxidative stress 565758, which might have an impact on mitochondrial functions as well. And third, the vacuole is the terminal compartment receiving cellular components destined for degradation by autophagic pathways. As also mitochondria are degraded by autophagy in yeast 59, it is possible that the vacuole plays an important role in mitochondrial quality control and turn-over. The high number of pet mutants lacking V-ATPase subunits clearly demonstrates that there is an important - as yet not fully understood - functional relationship between the vacuole and mitochondria.

The respiratory-deficient phenotype of 23 pet mutants was rescued by mating with Δmip1 as well as by cytoduction (class II; Additional data file 9). These mutants contained a [rho⁺] mitochondrial genome, as indicated by the mating experiment. In addition, three independently performed cytoduction experiments suggest that replenishment of cytoplasmic material reproducibly restores and maintains respiratory growth, at least for a few generations. These observations point to the possibility that respiratory competence may involve acquired properties that are not strictly linked to the nuclear or mitochondrial genotype. In order to corroborate this assumption, we tested whether cytoduction with a [rho⁰] donor strain would also restore respiratory growth. Rescue was observed in 11 strains (Additional data file 9), suggesting that, at least in some cases, cytoplasmic components other than mtDNA are able to improve respiratory functions. We hypothesize that respiratory deficiency may be an acquired phenotype that does not exclusively depend on the genotype.

Among ten class II mutants lacking known mitochondrial proteins (Δcoq5, Δcoq10, Δcox10, Δcox16, Δcox19, Δmct1, Δmss2, Δnfu1, Δslm3, and Δsom1) are four mutants that are specifically defective in the assembly of the cytochrome c oxidase (COX complex). Cox10 is required for the synthesis of the heme A cofactor 6061, Cox19 is a metallochaperone that delivers copper to the COX complex 62, Mss2 is required for the membrane translocation of the carboxyl terminus of the mitochondria-encoded Cox2 protein 63, and Cox16 contributes to assembly of the COX complex by an as yet unknown mechanism 64. Intriguingly, all four of these proteins are required for assembly of COX subunits at a post-translational stage. While respiratory-deficiency in Δcox10, Δcox16, Δcox19, and Δmss2 mutants has been documented before 60626364, we asked whether acquired properties might contribute to the loss of respiratory activity in these mutants. To exclude effects due to differences in mtDNA copy number, we first quantified the abundance of the mitochondrial COX3 gene by RT-PCR. We found that mtDNA is stably maintained in Δcox10, Δcox16, Δcox19, and Δmss2 mutants at a level very similar to wild-type cells (Figure 5a).

Next, we tried to rescue the deletion mutants with plasmids encoding wild-type copies of the respective genes under control of their endogenous promoters. Remarkably, after growth on selective medium a substantial number of transformants remained respiratory-defective after complementation with the respective wild-type gene (Figure 5b). The occurrence of respiratory-deficient clones was not induced by the transformation procedure per se because transformation of wild-type cells with the same plasmids yielded 100% respiration-competent clones (not shown). In order to test whether Δcox10, Δcox16, Δcox19, and Δmss2 clones lose properties required for respiratory competence over time, we subjected the deletion mutants to chronological aging 65, that is, continued incubation of stationary phase cultures. Mutant cells were incubated on glucose-containing medium for several days at room temperature before transformation with the complementing plasmids. Under these conditions, the fraction of clones that could not be rescued increased to 60 to 81% for mutant cells, whereas only 6% of aged wild-type clones were observed to be respiratory-deficient after transformation (Figure 5b). This suggests that mitochondria in Δcox10, Δcox16, Δcox19, and Δmss2 cells become irreversibly damaged over time, producing a respiratory-deficient phenotype that cannot be rescued any more. Apparently, this damage is already induced during vegetative growth and is markedly enhanced during aging.

As mitochondrial metabolism and aging are linked to the generation of potentially harmful reactive oxygen species (ROS) 66 we asked whether ROS accumulate in COX assembly mutants. High levels of ROS generated in yeast cells convert the non-fluorescent compound dihydrorhodamine 123 (DHR) to the oxidized fluorescent chromophore rhodamine 123 67. Upon incubation of young wild-type Δcox10, Δcox16, Δcox19, and Δmss2 cultures with DHR (8 h in liquid YPD medium) only very few cells showed significant staining (Figure 5c). After continued incubation (32 h), about 60% of wild-type cells and 90 to 98% of mutant cells showed significant rhodamine staining (Figure 5c). Very similar results were obtained when aging was allowed for up to 80 h (not shown). Furthermore, we noticed that rhodamine staining in wild-type cells was relatively faint and often restricted to tubular structures (presumably representing the mitochondrial network), whereas the signal was much stronger and dispersed throughout the cytosol in mutant cells (Figure 5c). We conclude that Δcox10, Δcox16, Δcox19, and Δmss2 cells produce elevated ROS levels during chronological aging. Presumably, ROS induce irreversible damage to mitochondrial proteins, lipids and/or mtDNA, thereby preventing rescue of the mutant phenotype by transformation with complementing plasmids. On the other hand, replenishment of fresh mitochondria by cytoduction might improve respiratory performance, at least for a limited time. It remains to be shown whether accumulation of ROS-induced damage is a general feature of class II mutants.

All previously uncharacterized RRG genes analyzed herein can be clearly related to mitochondrial functions. Proteins Rrg1, Rrg2, and Rrg5 through Rrg10 have been localized to mitochondria by high-throughput green fluorescent protein (GFP) fusion protein localization 35 and/or mitochondrial proteome analysis 634. The Rrg3 protein carries a putative mitochondrial presequence, whereas the intracellular location of Rrg4 remains unknown. Functional properties of RRG genes are summarized in Table 5.

Δrrg1, Δrrg2, Δrrg4, Δrrg5, Δrrg6, Δrrg8, and Δrrg9 are class I pet mutants lacking a functional mitochondrial genome. DAPI staining revealed defects in the organization of mtDNA that emerged early after introduction of wild-type mitochondrial genomes by cytoduction in Δrrg1, Δrrg2, Δrrg6, Δrrg8, and Δrrg9 mutants. Nucleoids appeared larger compared to the wild type, the number of nucleoids per cell was reduced, and several cells were completely devoid of mtDNA (not shown). These observations suggest that Rrg1, Rrg2, Rrg6, Rrg8, and Rrg9 play an important role in maintenance of mtDNA. Immediate and complete loss of mtDNA after cytoduction in the Δrrg5 mutant indicates an essential role of Rrg5 for maintenance of mtDNA (see above).

Interestingly, Rrg2 contains a pentatricopeptide (PPR) motif. PPR protein-encoding genes can be found in virtually all sequenced eukaryotic genomes, but are particularly abundant in plants. PPR proteins are localized in plastids and mitochondria where they are involved in the control of various stages of gene expression 68. Lack of mitochondrial translation activity and early loss of mtDNA observed here are consistent with a role of Rrg2 in control of mitochondrial gene expression.

Δrrg3 is a class III pet mutant able to maintain a [rho⁺] genome and wild-type-like mitochondrial protein translation activity. Although a mitochondrial location of Rrg3 has not been shown experimentally, the Mitoprot program 69 predicts the presence of a mitochondrial presequence with a high probability (0.9484). Mutants lacking Rrg3 (alternative name Aim22) show an increased petite frequency 70. The protein has high homology to lipoate-protein ligase A family members 71. Thus, it is conceivable that Rrg3 mediates the attachment of the lipoic acid cofactor to mitochondrial multienzyme complexes, such as pyruvate dehydrogenase, α-ketoglutarate dehydrogenase, glycine decarboxylase or others. Intriguingly, it has recently been reported that lipoate-protein ligase activity is important for maturation of RNase P, an enzyme that processes mitochondrial precursor tRNAs 72. It will be interesting to determine whether Rrg3 plays a specific role in this process.

The Δrrg4 mutant has recently been identified as one of 86 gene deletion mutants that show an increased assembly of Rad52, a central protein of the homologous recombination machinery, in subnuclear foci reflecting DNA repair centers. Therefore, the gene has been named IRC19 (for 'increased recombination centers') 73. Interestingly, several other genes related to mitochondrial function were also isolated in this screen, including CBT1, COX16, MRP17, MRPL1, MRPS16, and YMR31. It has been suggested that an increase of oxidative damage due to impaired respiratory chain functions might stimulate spontaneous DNA lesions in the nucleus and, therefore, constitutes a functional link between mitochondrial respiration and DNA repair processes in the nucleus 73. As a Rrg4-GFP fusion protein can not be visualized in cells 35, the intracellular location of Rrg4 remains unknown.

The RRG6 gene has recently been found in a screen for components involved in remodelling of the endoplasmic reticulum (ER). It has been named HER2 (Hmg2-induced ER remodelling); however, its molecular role in shaping the ER membrane remained unknown 74. As the Rrg6 protein has been localized to mitochondria by both GFP tagging and proteome analysis 63435, we propose that its primary function is related to maintenance of respiratory activity. The protein is highly homologous to bacterial glutamyl-tRNA amidotransferases, and a role in mitochondrial protein synthesis is consistent with our observation that mitochondrial translation is blocked in the Δrrg6 mutant (Table 3). Recently, RRG5 (alternative name GEP5) and RRG6 (alternative names GEP6 or HER2) have been shown to genetically interact with genes encoding prohibitin ring complexes in the mitochondrial inner membrane 75; however, the functional significance of this interaction is not yet understood.

Δrrg7 is a class II pet mutant presumably acquiring respiratory deficiency independent of its genotype. The RRG7 gene encodes a mitochondrial protein 35 that has homologs in fungi and other lower eukaryotes. Its function in mitochondrial biogenesis is currently unknown; however, the deletion mutant has been reported to exhibit increased sensitivity to the synthetic tripeptide arsenical 4-(N-(S-glutathionylacetyl)amino) phenylarsenoxide that targets mitochondria by inactivating the adenine nucleotide translocator. This drug inhibits proliferation of actively dividing endothelial cells and is an inhibitor of angiogenesis during tumor formation 76.

Δrrg10 is a class III pet mutant able to maintain a [rho⁺] genome. The RRG10 gene encodes a small mitochondrial protein 3435 of only 85 amino acid residues. Analysis of the mitochondrial translation pattern revealed a reduction of Cox1, suggesting that Rrg10 plays a specific role in transcription or maturation of mitochondrial mRNAs and/or translation or assembly of mitochondrial gene products.

Yeast strains used in this study were isogenic to BY4741, BY4742 and BY4743 18, with the exception of strain J1361 80, which was used for cytoduction. The [rho⁰] cytoduction donor strain was generated by growth of J1361 overnight in YPD medium supplemented with 50 μg/ml ethidium bromide. Complete loss of mtDNA was controlled by DAPI staining. The MATα gene deletion library 16 and its supplement covering newly assigned small ORFs 21 was obtained from BioCat (Heidelberg, Germany), and MATa single mutant Δmip1 was obtained from EUROSCARF (Frankfurt, Germany). Plasmid pRS416/MSS2 was constructed by PCR amplification of the MSS2 gene using primers 5' AAA GGA TCC GAT TTT ATG TGT GGA ATG CTA ACG ATG AAC and 5' AAA CTC GAG CTC TAA CAG TAT TTC CTA ATT ATT TCA TAG GTA AC and subcloning into the BamHI and XhoI sites of vector pRS416 81. Plasmid pRS416/COX16 was constructed by PCR amplification of the COX16 gene using primers 5' AAA GGA TCC AAT ATT ACC GTG AAT ATC GCG AGC TAC and 5' AAA CTC GAG AGG TAT TTA CAA TCA TTT CCT AGA CAT TCT and subcloning into the BamHI and XhoI sites of vector pRS416. For complementation tests, yeast strains were transformed with plasmids pG12/T4 60 expressing COX10, pRS416/COX16, pG188/T1 62 expressing COX19, or pRS416/MSS2.

PCR analyses to confirm the identity of deletion mutants were performed in a way that one primer was homologous to a sequence within the coding region or within the deletion marker cassette, respectively, whereas the other primer was homologous to a sequence outside the deleted part of the gene. Thus, a PCR product can be generated only if the correct allele corresponding to the primer combination is present. Primers used to confirm the identity of yeast deletion mutants are listed in Additional data file 10.

S. cerevisiae was cultivated and manipulated according to standard procedures 82. For screening for respiratory-deficient mutants, yeast deletion strains were manually transferred with a sterile pinning tool from 96-well plates to rich media plates with either 2% glucose as fermentable (YPD) or 3% glycerol as non-fermentable (YPG) carbon source. The screening of the entire library on YPG was performed once. The screening procedure was as similar as possible to the screen performed earlier by us in the Dimmer et al. study 14. Respiratory-deficient mutants were screened in addition on media containing 3% ethanol or 3% lactate (pH adjusted to 7.0 with NaOH) as non-fermentable carbon sources. The growth behavior was evaluated by visual inspection after 3 days (YPD) or 6 days (YPG and other non-fermentable media) of incubation at 30°C.

For high-throughput complementation tests with Δmip1, yeast deletion strains were transferred with a sterile pinning tool from 96-well plates to a lawn of MATa Δmip1 cells on YPD plates. After incubation overnight to allow for mating, cells were replica-plated two times on plates containing minimal SD medium selective for diploid cells. Then, growth on YPG plates was determined as above. Cytoduction was performed as described 80. Cytoduction experiments were repeated at least three times. To adapt yeast deletion strains to non-fermentable carbon sources, cells were transferred from YPD plates to YPG plates containing 0.1% glucose, replica-plated once on YPG/0.1% glucose and then replica-plated on YPG.

Labeling of mitochondrial translation products in vivo was performed essentially as described 83 with the following minor modifications: cytosolic translation was stopped with 0.3 mg/ml cycloheximide, the labeling reaction was performed for 30 minutes, and the chase reaction was performed for 15 minutes. Mitochondrial translation products were analyzed by SDS-PAGE, transfer to nitrocellulose and autoradiography.

Using sterile toothpicks, cells were taken from glycerol stocks and spread on YPD plates as patches of about 2 cm² size. Plates were incubated for 14 to 28 days at room temperature to allow chronological aging. Then, a small amount of aged cells (or young cells taken directly from glycerol stocks as a control) was spread on a fresh YPD plate, incubated overnight at 30°C, and transformed according to the rapid transformation protocol described by Truong and Gietz 84. Using sterile toothpicks, at least 100 transformants were transferred as short streaks (approximately 1 cm) to fresh SD plates selective for the marker of the transformed plasmids. Plates were incubated for 1 day at 30°C and then replica-plated on YPD and YPG using sterile velvet. Carbon source-dependent growth of transformants was visually scored after 2 to 3 days at 30°C.

Staining of mtDNA with DAPI in methanol-fixed cells was as described 30. For the analysis of ROS production, 1 μl DHR (2.5 mg/ml in DMSO) was added to 500 μl cell suspension and incubated for 2 h at 30°C. Cells were harvested by centrifugation, washed in phosphate-buffered saline, resuspended in phosphate-buffered saline and analyzed by microscopy. Epifluorescence microscopy was performed using a Zeiss Axioplan 2 microscope equipped with a HBO 100 mercury lamp, Zeiss filter sets 01 and 09 and a Plan-Neofluar 100× 1.30 NA Ph3 oil objective (Carl Zeiss Lichtmikroskopie, Göttingen, Germany). Images were recorded with an Evolution VF Mono Cooled monochrome camera (Intas, Göttingen, Germany) and processed with Image Pro Plus 5.0 and Scope Pro 4.5 software (MediaCybernetics, Silver Springs, MD, USA).

Yeast strains were grown overnight in liquid YPD medium and DNA was extracted using the YeaStar™ Genomic DNA Kit (Zymo Research, Orange, USA) according to the manufacturer's instructions. PCR reactions were performed in 20 μl volume in 96-well plates using Maxima™ SYBR Green qPCR Master Mix (2×) (Fermentas, St Leon-Rot, Germany) according to the manufacturer's instructions in an ABI PRISM 7000 Sequence Detection System (Applied Biosystems, Foster City, CA, USA). The following primers were used: GAL4 forward, 5' TTT CTC CTG GCT CAG TAG GGC; GAL4 reverse, 5' AGT TAC GAG AGG GTG GAC GGT; COX3 forward, 5' ATT GAA GCT GTA CAA CCT ACC GAA TT; COX3 reverse, 5' CCT GCG ATT AAG GCA TGA TGA. Data were analyzed with Sequence Detection Software Version 1.2.3 7000 System SDS software Core Application (Applied Biosystems) and calculated according to the 2^-ΔΔC_T-method 85.