Assembly-dependent translation of subunits 6 (Atp6) and 9 (Atp9) of ATP synthase in yeast mitochondria

Abstract The yeast mitochondrial ATP synthase is an assembly of 28 subunits of 17 types of which 3 (subunits 6, 8, and 9) are encoded by mitochondrial genes, while the 14 others have a nuclear genetic origin. Within the membrane domain (FO) of this enzyme, the subunit 6 and a ring of 10 identical subunits 9 transport protons across the mitochondrial inner membrane coupled to ATP synthesis in the extra-membrane structure (F1) of ATP synthase. As a result of their dual genetic origin, the ATP synthase subunits are synthesized in the cytosol and inside the mitochondrion. How they are produced in the proper stoichiometry from two different cellular compartments is still poorly understood. The experiments herein reported show that the rate of translation of the subunits 9 and 6 is enhanced in strains with mutations leading to specific defects in the assembly of these proteins. These translation modifications involve assembly intermediates interacting with subunits 6 and 9 within the final enzyme and cis-regulatory sequences that control gene expression in the organelle. In addition to enabling a balanced output of the ATP synthase subunits, these assembly-dependent feedback loops are presumably important to limit the accumulation of harmful assembly intermediates that have the potential to dissipate the mitochondrial membrane electrical potential and the main source of chemical energy of the cell.


Introduction
Mitochondria support aerobic respiration and produce the bulk of cellular ATP through the process of oxidative phosphorylation (OXPHOS) (Saraste 1999). Typically, OXPHOS involves 5 heterooligomeric protein complexes (numbered I-V) that are located in the mitochondrial inner membrane. Complexes I-IV transfer to oxygen electrons from carbohydrates and fatty acids coupled to the transport of protons from the mitochondrial matrix to the intermembrane space, and the resulting energy-rich transmembrane proton gradient is used by the Complex V (ATP synthase) to make ATP from ADP and inorganic phosphate.
The structural genes of the mitochondrial OXPHOS complexes are located in nuclear and mitochondrial DNA. As such, 2 separate translation machineries, one cytosolic and the other inside the mitochondrion, are utilized to synthesize their gene products (Ott and Herrmann 2010). A plethora of proteins that do not belong to the OXPHOS complexes assist translation and assembly of their subunits in Saccharomyces cerevisiae (Fox 1996(Fox , 2012Herrmann et al. 2013;Ott et al. 2016). A subset of these "helper" proteins are known to interact with the 5 0 -UTR (Untranslated Region) of a specific mitochondrial mRNA transcript, and in some instances, with the translation product as well, and these observations have led investigators to postulate the existence of regulatory feedback loops that couple translation and assembly of mitochondrial gene products (Fox 2012;Fontanesi 2013;Herrmann et al. 2013;Ott et al. 2016). The most thoroughly investigated example is Mss51, which facilitates Cox1 translation and assembly in complex IV (Soto et al. 2012). Studies have shown that Mss51 binds first to the 5 0 -UTR of the COX1 mRNA, where it activates translation, and then to the newly synthesized Cox1 protein until Cox1 is assembled with its partner subunits (Perez-Martinez et al. 2003; Barrientos et al. 2004;Pierrel et al. 2007;Mick et al. 2010). The posttranslational activity of Mss51 is similar to that described for another mitochondrial protein, which is a small complex composed of Cbp3 and Cbp6 polypeptides that maintains association with newly translated cytochrome b through its acquisition of heme cofactors and assembly with the nucleus-encoded subunits of complex III (Rodel 1986;Chen and Dieckmann 1994;Dieckmann and Staples 1994;Gruschke et al. 2011;Hildenbeutel et al. 2014). Similarly, the Sov1 protein assists in yeast the translational activation and assembly of the mtDNAencoded Var1 subunit of the mitochondrial ribosome (Seshadri et al. 2020).
Much less is known about the regulation of yeast ATP synthase biogenesis. This is an assembly of 28 subunits of 17 types that are encoded by 3 mitochondrial (ATP6, ATP8, and ATP9) and 14 nuclear genes. This enzyme, known as the F 1 F O complex, is organized into a largely hydrophobic domain (F O ) that transports protons through the membrane, and a hydrophilic domain (F 1 ) in the mitochondrial matrix where ATP is synthesized. The ATP6 and ATP9 genes encode subunits of the F O (6/a and 9/c) that form an integral proton channel made of one subunit 6 and an oligomeric ring of 10 subunits 9 (9 10 -ring). During proton translocation, the 9 10 -ring rotates, and this induces conformational changes in the F 1 that promote ATP synthesis. The ATP8 gene encodes a membrane-embedded protein (the subunit 8) that stabilizes the proton channel of ATP synthase. ATP9 is cotranscribed with tRNA ser and a mitochondrial ribosome subunit gene (VAR1), whereas ATP8 and ATP6 are co-transcribed with COX1 (Christianson and Rabinowitz 1983;Zassenhaus et al. 1984;Finnegan et al. 1991). Processing of the primary transcripts produces a major ATP9 mRNA with a 0.63 kb long 5 0 -UTR, and near equal amounts of 2 bicistronic ATP8,6 mRNAs that differ by the length of their 5 0 -UTR (Foury et al. 1998).
Among the proteins that are associated with the biogenesis pathway for subunits 6, 8, and 9, four (Nca2, Nca3, Aep3, and Nam1) contribute to ATP8/6 transcripts stability (Asher et al. 1989;Pelissier et al. 1992Pelissier et al. , 1995Ellis et al. 2004). A role for Aep3 in translation of subunit 8 has also been reported (Barros and Tzagoloff 2017). Other proteins include Atp22, which activates subunit 6 translation (Zeng et al. 2007a), and Smt1, which represses translation of both subunits 6 and 8 (Rak et al. 2016). The stability of the ATP9 transcript is dependent on a 35 kDa Cterminal cleavage fragment of Atp25 (Zeng et al. 2008) that is released in the matrix by the mitochondrial processing peptidase (Woellhaf et al. 2016). Two more proteins with activities linked to the ATP9 mRNA are Aep1 and Aep2 (Ackerman et al. 1991;Finnegan et al. 1991Finnegan et al. , 1995Payne et al. 1991Payne et al. , 1993Ziaja et al. 1993). Some studies have suggested a role for these proteins in ATP9 mRNA stability/processing Ziaja et al. 1993), while others have led investigators to propose that they activate translation (Payne et al. 1993;Ellis et al. 2004).
Another subgroup of yeast nuclear gene products associated with ATP synthase biogenesis assists oligomerization of its subunits. Two such proteins, Atp11 and Atp12, are essential for creating the [ab] 3 hexamer, which houses the catalytic sites for ATP synthesis and contributes $85% of the mass in F 1 (Ackerman and Tzagoloff 1990b;Wang et al. , 2001Ackerman 2002;Lefebvre-Legendre et al. 2005). A third protein that is part of this process, Fmc1, is necessary for Atp12 folding/stability at elevated temperature (36 C) (Lefebvre-Legendre et al. 2001. Other work has shown that a complex of 2 proteins, Ina22 and Ina17, facilitates assembly of ATP synthase peripheral stalk, which is a structure protruding from the membrane and that contacts the surface of the F 1 to prevent rotation of the [ab] 3 hexamer during catalysis (Lytovchenko et al. 2014). In an early model of F 1 F O assembly, the 9 10 -ring was proposed to assemble spontaneously (Arechaga et al. 2002). This was challenged by later studies suggesting that oligomerization of the 9 10 -ring is chaperoned by a 32 kDa N-terminal fragment of Atp25 (Zeng et al. 2008). This fragment is homologous to the bacterial ribosomesilencing factor and was found associated with the mitochondrial ribosome (Woellhaf et al. 2016). Two proteins (Atp10 and Atp23) have been assigned roles in procuring the assembly/stability of subunit 6 (Paul et al. 2000;Tzagoloff et al. 2004;Osman et al. 2007;Zeng et al. 2007c). Atp10 was shown to associate in a physical complex with newly translated subunit 6 and promote a favorable interaction with the 9 10 -ring . Atp23 was identified to be a protease that cleaves the first 10 amino acid residues of the nascent subunit 6 polypeptide and has been linked also to folding the processed protein (Osman et al. 2007;Zeng et al. 2007bZeng et al. , 2007c. Finally, the interaction of subunit 6 with the 9 10 -ring has been reported to depend on Oxa1 (Jia et al. 2007), which is a protein translocase involved in the insertion of certain mitochondrial and nuclear gene products into the inner membrane (Altamura et al. 1996;Herrmann et al. 1997;Hell et al. 2001).
Herein we report that translation of subunit 6 and 9 is enhanced in mutant strains with specific defects in the assembly of these proteins. These translation modifications involve assembly intermediates interacting with these proteins within the final ATP synthase and cis-regulatory sequences that control gene expression in the organelle. These findings suggest that the subunit 6 is part of an assembly-dependent feedback loop that is different from a previously reported regulatory model for this protein (Rak et al. , 2011. Additionally, they contradict the generally accepted view that the 9 10 -ring forms separately, independently of any other ATP synthase component.

Construction and isolation of yeast strains
The genotypes of strains used in this study are listed in Table 1. All the mutant strains are derivatives of wild-type strain MR6 (Rak et al. 2007b). Deletion of AEP1, AEP2, and ATP25 in wild-type strain MR6 (Rak et al. 2007b) was done using KANMX4 DNA-cassettes [amplified with Low and Up primers (Table 2)] and genomic DNA was isolated from aep1D, aep2D, and atp25D deletion strains from Euroscarf. The deletions of AEP1, AEP2, and ATP25 in MR6 were PCR-verified with Ver-KanMx primers (Table 2). Prior to deleting AEP1, AEP2, or ATP25, the MR6 strain was transformed with the pAM19 plasmid, which encodes the PaAtp9-5 gene of Podospora anserina (here referred to as ATP9-nuc) (Bietenhader et al. 2012). In the presence of pAM19, aep1D, aep2D, and atp25D strains This study have a much less reduced propensity to produce q À /q 0 cells lacking functional mtDNA. The ectopic integration of ATP6 in mtDNA, with and without the atp6-L 173 P mutation, was done using a previously described procedure (Steele et al. 1996). As a first step, the WT and mutated ATP6 genes flanked by 75 bp of the 5 0 -UTR and 118 bp of the 3 0 -UTR of COX2 were PCR-amplified using GeneArt Invitrogen Gene Synthesis technology and cloned at the EcoRI site of pPT24 plasmid (Thorsness and Fox 1993), giving plasmids pRK49 and pRK77, respectively. These plasmids were introduced into mitochondria of the q 0 strain DFS160 by biolistic transformation using the Particle Delivery Systems PDS-1000/He (BIO-RAD) as previously described (Bonnefoy and Fox 2001), giving the petite synthetic qstrains RKY89-2 and RKY116-2. These strains were crossed with the strain RKY83 (Bader et al. 2020) in which the entire coding sequence of ATP6 is replaced with ARG8 m (atp6::ARG8 m ) and where the gene COX2 is partially deleted (cox2-62). The RKY89-2 Â RKY83 and RKY116-2 Â RKY83 crosses produced clones (RKY112 and RKY117, respectively) that were arginine auxotrophic and able to grow in rich glycerol (YPGlyA). The ectopic integration of the WT or the mutated ATP6 gene in strains RKY112 and RKY117 in the intergenic region of COX2 was PCRverified with primers oAtp6-2, oAtp6-4, o5 0 UTR2 and o5 0 UTR1 primers ( Table 2). The modified mitochondrial genetic loci in strains RKY112 and RKY117 are schematically represented in Fig. 4a.
Construction of plasmids encoding Aep1, Aep22, Atp25, and Atp25-Cter The AEP1 and AEP2 genes were amplified using primer pairs Aep1-KL/Aep1-Low and Aep2-KL/Aep2-Low, respectively (Table 2) and cloned into pGEM-T easy vector (Promega). The AEP1 and AEP2 sequences were cut off from the resulting plasmids with BamHI/PstI and XbaI/SalI, respectively, and cloned into Yep351 (Hill et al. 1986) under control of the MET25 promoter, yielding plasmids pRK54 and pRK57, respectively (the plasmids used in this study are listed in Table 3). The plasmid pRK56 was constructed by cloning in Yep351 the SacI-HindIII fragment encoding Atp25 in the previously described plasmid pG29ST18 (Zeng et al. 2008). The plasmid pAK13 was constructed by cloning in pRS424 (Brachmann et al. 1998) the HindIII-BamHI fragment encoding Atp25-Cter in the previously described plasmid pG29ST23 (Zeng 2008). The pAK13 was digested by NotI and BamHI and ligated with mitochondrial targeting sequence present in the previously described plasmid pYX232 (Westermann and Neupert 2000), yielding plasmid pAK19.
In vivo labeling of mitochondrial translation products A previously described protocol (Barrientos et al. 2002) was used to assess neo-synthesis of the mtDNA-encoded proteins in yeast cells grown to early exponential phase (10 7 cells/mL) in 10 mL of rich (YPGalA) or complete synthetic (CSM) galactose media. After centrifugation, the cells were washed twice with a low sulfate medium containing 2% galactose, supplemented with histidine, tryptophan, leucine, uracil, adenine, and arginine (50 mg/L each), and then incubated for 2 h at 28 C in 500 mL of the same medium (to starve the cells for endogenous cysteine and methionine). After inhibiting cytosolic protein synthesis for 5 min with 1 mM cycloheximide, 50 mCi of [ 35 S]-methionine þ [ 35 S]-cysteine (Amersham Biosciences) was added, and the incubation was prolonged for 20 min. Total protein extracts were prepared and samples with the same level of incorporated radioactivity were separated by SDS-PAGE (Laemmli 1970) in 17.5% acrylamide gels (to separate Atp8 and Atp9) and in 12% acrylamide containing 4M urea and 25% glycerol (to separate Atp6, Cox3, Cox2, and cytochrome b). After electrophoretic migration, the gels were dried and the radioactive proteins were visualized by autoradiography using a PhosphorImager after 1-week of exposure.

BN-and SDS-PAGE
Mitochondria were isolated from yeast cells grown until middle exponential phase (2-3 Â 10 7 cells/mL) in rich galactose medium (YPGalA) at 28 C using the previously described enzymatic method (Gu erin et al. 1979). Protein concentration was determined according to (Lowry et al. 1951) in the presence of 5% SDS. Blue native polyacrylamide gel electrophoresis (BN-PAGE) was done as described in (Schagger and von Jagow 1991). Briefly, a sample of mitochondria containing 200 mg of proteins was suspended in 100 mL of 30 mM HEPES, 150 mM potassium acetate, 12% glycerol, 2 mM 6-aminocaproic acid, 1 mM EGTA, 2% digitonin (from Sigma) buffered at pH 7.4 and supplemented with an inhibitor cocktail tablet (from Roche). After 30 min of incubation on ice, the extract was cleared by centrifugation (18,600 g, 4 C, 30 min), supplemented

Name
Sequence with 4.5 mL of loading dye (5% Serva Blue G-250, 750 mM 6-aminocaproic acid), and run in NativePAGE TM 3-12% Bis-Tris Gel (Invitrogen). The proteins were transferred onto a polyvinylidene difluoride membrane and probed with 1:5,000 diluted polyclonal antibodies against Atp6/subunit a, Atp1/subunit a, and Atp9/subunit c (a gift from J. Velours). The tested proteins were revealed using peroxidase-labeled antibodies (Promega) at a 1:5,000 dilution and the ECL reagent of Pierce ECL Western Blotting Substrate (Thermo Scientific). SDS-PAGE analysis and immunological detection of proteins from isolated mitochondria or whole cells was performed as previously described (Schagger and von Jagow 1991). The open source program Image J (http://rsbweb.nih.gov/ij/) was used to quantify the immunological signals and statistical significance between samples was evaluated using Student's t-test.

Regulation of ATP6
Translation of assembly-defective subunit 6 variants is upregulated In previous work, we observed that an assembly-defective mutant version of subunit 6 harboring a leucine-to-proline change at position 173 (L 173 P) in mature subunit 6 (which is produced after cleavage of the 10 N-terminal residues present in the precursor protein) was translated at a rate that was 2-3-fold faster than the wild-type protein [ (Kucharczyk et al. 2009a); see below]. To ascertain that the L 173 P change was responsible for this effect, mitochondrial translation was herein investigated in 8 genetically independent recombinant 6-L 173 P clones issued from crosses between RKY12 (a synthetic qwith the 6-L 173 P mutation, see a) The influence of the subunit 6 (Atp6) L 173 P variant on mitochondrial translation was investigated in 8 genetically independent recombinant 6-L 173 P clones issued from crosses between RKY12 (a synthetic qwith the 6-L 173 P mutation) and MR10 [a q þ strain in which the coding sequence of ATP6 is replaced with ARG8 m (atp6::ARG8 m )], in comparison to 8 genetically independent WT recombinant clones issued from crosses between SDC30 (a synthetic qcontaining the wild-type ATP6 gene) and MR10 (see Table 1 for strain genotypes). b) Mitochondrial translation in 5 other subunit 6 mutants. Labeling of the mitochondrial translation products was performed in galactose grown cells during 20 min in the presence of [ 35 S]-methionine and [ 35 S]-cysteine and cycloheximide to block cytoplasmic translation. Total cellular protein extracts were then prepared and separated by SDS/PAGE in a 12% polyacrylamide gel containing 4 M urea and 25% glycerol. The gels were dried and the radiolabeled proteins were visualized using a PhosphorImager after 1-week exposure. The subunit 6 (Atp6) and Cox3 signals were quantified and compared to each other within each sample. The Atp6/Cox3 ratio was set to 1 for the WT. Table 1) and MR10 [a q þ strain in which the coding sequence of ATP6 is replaced with ARG8 m (atp6::ARG8 m ), see Table 1], in comparison to 8 genetically independent WT recombinant clones issued from crosses between SDC30 (a synthetic qcontaining the wild-type ATP6 gene, see Table 1) and MR10. Owing to the presence in RKY12 and SDC30 of the nuclear karyogamy delaying kar1-1 mutation, it was possible to isolate the mitochondrial DNA recombinants in MR10 haploid background (which is the one of WT strain MR6). As a result, the 16 clones are isogenic except for the L 173 P mutation. Mitochondrial translation in these 16 clones was conducted with whole cells in the presence of cycloheximide (to block cytosolic translation) during a rather short time (20 min) after the addition of a mixture of radioactively labeled [ 35 S]-methionine and [ 35 S]-cysteine, a procedure widely used to evaluate the rate of synthesis of the mitochondrial translation products. As shown in Fig. 1a, subunit 6 was upregulated in the 8 L 173 P clones vs. the 8 reconstructed WT clones, which demonstrates that no other change in nuclear and mitochondrial DNA was Fig. 2. Upregulation of the subunit 6-L 173 P variant is F 1 -dependent. a) ATP synthase in BN-gels. Mitochondria were isolated from the indicated strains grown in rich galactose at the indicated temperature. Proteins were extracted with digitonin (2 gr/gr protein) and separated by BN-PAGE on a 3-10% polyacrylamide gel (50 mg/lane). The proteins were transferred to a PVDF membrane and probed with antibodies against subunit 9. The F 1 -F O monomers and dimers are identified on the right. b) Cells from WT and fmc1D yeasts with or without the subunit 6 mutation L 173 P were grown in rich galactose at 28 C or 36 C, as indicated, and then incubated for 20 min with [ 35 S]-methionine and [ 35 S]-cysteine in the presence of cycloheximide to block cytoplasmic translation. Total cellular protein extracts were then prepared and separated by SDS/PAGE in a 12% polyacrylamide gel containing 4 M urea and 25% glycerol (with a 30:0.8 ratio of acrylamide and bis-acrylamide). The gels were dried and the radiolabeled proteins were visualized using a PhosphorImager. The subunit 6 (Atp6) and Cox3 signals were quantified and compared to each other within each sample. The Atp6/Cox3 ratio was set to 1.0 for the WT. The identity of the band above Cox3 (designated by a hash sign) in the samples containing the subunit 6 variant L 173 P is unknown. (2) 17.5% of polyacrylamide (to resolve Atp8 and Atp9). The 2 gels shown in b) were loaded with equal amounts of radioactivity from the same experiment. After drying of the gels under vacuum, the radiolabeled proteins were visualized using a PhosphorImager after one-week of exposure. The subunit 6 (Atp6) and Cox3 signals were quantified and compared to each other within each sample. The Atp6/Cox3 ratio was set to 1.0 for the WT. The identity of the band above Cox3 (designated by a hash sign) in the samples containing the 6-L 173 P variant is unknown.
involved in the increased rate of subunit 6 synthesis in the L 173 P clones. Two other subunit 6 variants, L 173 R (Rak et al. 2007a) and W 126 R (Kucharczyk et al. 2013), which assemble normally but are compromised functionally, were translated at the normal rate. In an effort to explain these findings, we investigated by pulse labeling the translation efficiency for 5 additional subunit 6 variants, harboring either single-or double-point mutations. We observed accelerated translation for 1 subunit 6 variant, while the 4 others were synthesized at the normal rate (Fig. 1b). Remarkably, in related work, we found that the strain showing enhanced subunit 6 synthesis was the only subunit 6 variant from this group of 5 that showed an assembly defect (Kucharczyk et al. 2019). Hence, it would appear that yeast accelerates the rate of subunit 6 translation when assembly of the ATP synthase stalls at the point where subunit 6 should be incorporated to complete the enzyme's proton channel.
What is the mechanism behind accelerated synthesis of assembly-defective subunit 6?
There is solid evidence in the literature that supports a model for ATP synthase biogenesis in which subunit 6 does not get incorporated until late in the process Rak et al. 2007b). Hence, it was deemed logical to investigate some of the structural elements that would already be in place in the partially assembled enzyme for actions that promote enhanced expression of assembly-defective subunit 6 variants. This investigation was conducted with strains where the subunit 6 L 173 P mutation was combined to genetic modifications leading to defects in the assembly of the F 1 or the 9 10 -ring, and we also tested the requirement for the 5 0 -UTR of ATP6 by ectopically expressing the subunit 6 variant L 173 P from the 5 0 -UTR of COX2.

The F 1
The importance of the F 1 in the upregulation of the assemblydefective subunit 6 variant L 173 P was tested using the conditional strain fmc1D. This strain has a diminished capacity to assemble the F 1 especially when grown at 36 C (5-10% vs. WT), whereas quite good levels of F 1 accumulated in fmc1D cells grown at 28 C [(Lefebvre-Legendre et al. 2001), Fig. 2a]. At both temperatures, the strain fmc1D normally synthesizes the subunit 6 as well as the other proteins encoded by the mitochondrial DNA (Fig. 2b). The synthesis of the subunit 6 variant L 173 P was still stimulated in strain fmc1D grown at 28 C, Fig. 4. Pulse labeling of Atp6 and Cox3 and assembly of ATP synthase in strains expressing subunit 6 with and without the L 173 P mutation from the 5 0 UTR of COX2. a) Schema of mitochondrial genetic loci. As represented, the ectopic ATP6 gene, with or without the L 173 P mutation, is located upstream of the COX2 gene under control of the 5 0 -UTR and 3 0 -UTR of COX2, in a mitochondrial genome where the coding sequence of the native ATP6 gene is replaced with ARG8 m . The corresponding abbreviated genotypes (WT, 5 0 -UTR COX2 -ATP6 WT and 5 0 -UTR COX2 -ATP6 L173P ) are indicated on the right. b) Fresh glucose cultures of the strains with the indicated genotypes were serially diluted and spotted on rich glucose and rich glycerol media. The glucose and glycerol plates were scanned after 4 and 6 days of incubation at 28 C, respectively. c) In vivo labeling of Atp6 and Cox3. Pulse labeling of Atp6 and Cox3 was performed for 20 min with [ 35 S]-methionine and [ 35 S]-cysteine in the presence of cycloheximide to block cytoplasmic translation, in cells freshly grown in rich galactose medium. Total protein extracts were then prepared and separated by SDS/PAGE in a 12% polyacrylamide gel containing 4 M urea and 25% glycerol. After drying of the gel under vacuum, the radiolabeled proteins were visualized using a PhosphorImager after 1-week exposure. The subunit 6 (Atp6) and Cox3 signals were quantified and compared to each other within each sample. The Atp6/Cox3 ratio was set to 1.0 for the WT. d) Steady state levels of subunit 6. Total cellular proteins samples (20 mg) extracted from the 4 analyzed strains grown in rich galactose medium were separated by SDS-PAGE in a 12% polyacrylamide gel. The proteins were transferred to a nitrocellulose membrane and probed with antibodies against subunit 6 (Atp6) and porin. The levels of subunit 6 relative to porin are expressed in % of WT. Standard deviation and statistical significance between the 2 strains are indicated (* corresponds to a P-value <0.05) e) Protein complexes were extracted from isolated mitochondria with digitonin (2 gr/gr protein) and separated by BN-PAGE in a 3-10% polyacrylamide gel. After their transfer to a PVDF membrane, they were probed with antibodies against a-F 1 . The F 1 F O dimers and monomers, and free F 1 are identified in the left-hand margin. f) In vivo labeling of Arg8 in strains RKY112 and RKY116. The rate of Arg8 synthesis in RKY112 and RKY116 from the 5 0 UTR of ATP6 (atp6::ARG8 m ) relative to Var1 was probed by pulse labeling in whole galactose grown cells for 20 min with [ 35 S]-methionine and [ 35 S]-cysteine in the presence of cycloheximide to block cytoplasmic translation, with WT and RKY20 strains as controls that do not encode Arg8. Total protein extracts were then prepared and separated by SDS/PAGE in a 12% polyacrylamide gel containing 4 M urea and 25% glycerol. After drying of the gel under vacuum, the radiolabeled proteins were visualized using a PhosphorImager.
although less efficiently than in a WT nuclear background, whereas the upregulation of subunit 6 was totally lost at 36 C. Hence, it appears that the mechanism responsible for the translation stimulation of the assembly-defective 6-L 173 P variant is dependent on F 1 . The gel of Fig. 2b shows a band of unknown identity (designated by a hash sign) above Cox3 in samples containing the subunit 6 variant L 173 P. As described below, it was detected in other experiments in which this mutant was used, and with a yeast strain lacking the ATP6 gene (atp6D), indicating that this extra polypeptide was not some aberrant translation product derived from the modified atp6 loci.

Subunit 9
Yeast cells lacking the subunit 9 have a high propensity to lose the mitochondrial genome (Bietenhader et al. 2012), which makes it difficult to evaluate the importance of this protein in the upregulation of assembly-defective subunit 6. To overcome this difficulty, we used aep1D and aep2D strains unable to synthesize subunit 9 and a previously characterized nuclear ATP9 gene present in P. anserina [PaAtp9-5 (D equard-Chablat et al. 2011), here referred to as ATP9-nuc]. Although the protein product of ATP9-nuc (Atp9-nuc) shows a number of primary sequence differences with yeast subunit 9, a yeast strain lacking the native mitochondrial ATP9 gene (atp9D) had the ability to grow slowly on nonfermentable carbon sources upon transformation with a plasmid harboring ATP9-nuc (Bietenhader et al. 2012), showing that Atp9-nuc can assemble and function within the yeast ATP synthase albeit much less efficiently than subunit 9. That Atp9-nuc only partially compensates for the absence of subunit 9 is not very surprising considering that S. cerevisiae is not optimized for ATP synthase biogenesis with a nucleus-encoded version of this protein that must be imported into mitochondria posttranslationally, routed to the inner membrane, and assembled with the other yeast ATP synthase subunits. As expected, Atp9-nuc enabled aep1D and aep2D strains to grow slowly from respiratory carbon sources and when expression of Atp9-nuc was inhibited with doxycycline, they did not show any respiratory growth (Fig. 3a). The subunit 9 was not detected by pulse labeling in these 2 strains while the other proteins encoded by the mitochondrial DNA were all effectively synthesized (Fig. 3b). Importantly, although the F O assembles poorly with Atp9-nuc, the F 1 forms and accumulates efficiently (Bietenhader et al. 2012), which permitted us to evaluate the importance of the membrane motor of ATP synthase in the upregulation of the subunit 6 variant L 173 P without any interference due to a lack in F 1 . As shown in Fig. 3c, the L 173 P substitution did not stimulate subunit 6 synthesis in aep1D and aep2D mutants expressing Atp9nuc indicating that subunit 9, like the F 1 , is required for upregulating (1) 12% polyacrylamide containing 4 M urea and 25% glycerol (to resolve Var1, Cox1, Cox2, Cytochrome b, Cox3, and Atp6); (2) 17.5% polyacrylamide (to resolve Atp8 and Atp9). The 2 gels were loaded with equal amounts of radioactivity from the same experiment. After drying of the gels under vacuum, the radiolabeled proteins were visualized using a PhosphorImager. The subunit 9 (Atp6) and Cox3 signals were quantified and compared to each other within each sample. The Atp9/Cox3 ratio was set to 1.0 for the WT. Standard deviation and statistical significance of the differences between strains is indicated (* corresponds to a P-value <0.05). b) Evaluation of subunit 9 synthesis in WT yeast after blocking expression of ATP9-nuc with 5 mg/mL doxycycline (DOX), using the same procedure as in a). c) Northern blot analyses. Total RNAs isolated from cells with the indicated genotypes (q 0 is a derivative of the WT strain totally devoid of mtDNA) freshly grown in rich galactose medium were separated in a 1% agarose gel. After their transfer to a Nytran membrane, the RNAs were hybridized with 32 P labeled DNA probes specific to ATP9 and 21S RNA. The ATP9 mRNA signals were normalized to those corresponding to 21S RNA. d-f) Steady state levels of proteins in cells. Total proteins extracted from cells with the indicated genotypes grown in rich galactose medium were separated by SDS-PAGE (20 mg/lane), transferred to a nitrocellulose membrane, and probed with antibodies against the indicated proteins. The levels of subunits 6 (Atp6), 9 (Atp9), and Arg8 (expressed from atp9ϻARG8 m ) are normalized to porin. Ade13 is used as a cytosolic protein marker. The symbol "*" indicates a statistically significant difference between samples (P-value <0.05).
subunit 6 in response to mutations in this protein that compromise its assembly.
The 5 0 -UTR of the ATP6 mRNA Expression of each yeast mitochondrial gene involves a peculiar 5 0 -UTR and specific translational activator protein(s) (Fox 2012;Herrmann et al. 2013;Ott et al. 2016). To test the implication of the 5 0 -UTR of ATP6 in the upregulation of the assembly-defective subunit 6 variant L 173 P, we ectopically expressed ATP6 with or without the L 173 P mutation from the 5 0 -UTR of COX2 in a strain where the native ATP6 gene is in-frame replaced with ARG8m (Rak et al. 2007b) (Fig. 4a). In the resulting strains (5 0 UTR COX2 -ATP6 WT and 5 0 UTR COX2 -atp6 L173P ), subunit 6 translation is activated by Pet111 (the translation activator of COX2) instead of Atp22 (the translation activator of ATP6). Both strains grew well on glycerol (Fig. 4b) and synthesized subunit 6 as effectively as in WT yeast (Fig. 4c), which indicated that the 5 0 -UTR of the ATP6 mRNA is needed for stimulating translation of the subunit 6 variant L 173 P. Interestingly, despite a good rate of synthesis the ectopically expressed subunit 6 without the L 173 P mutation (5 0 UTR COX2 -ATP6 WT ) was about 2 times less abundant at the steady-state than in WT cells expressing subunit 6 from its own 5 0 -UTR (Fig. 4d). While BN-PAGE analyses revealed only fully assembled ATP synthase dimers and monomers in mitochondria from the WT, those from the 5 0 UTR COX2 -ATP6 WT strain additionally contain free F 1 particles (Fig. 4e). The presence of free F 1 is typically observed in strains where subunit 6 fails to assemble properly (Rak et al. 2007b), and it can therefore be inferred that the assembly of subunit 6 was less efficient when it was expressed from the 5 0 -UTR of COX2 instead of its native 5 0 -UTR. Not surprisingly, a further decrease in the levels of subunit 6 was observed with the ectopic atp6-L 173 P gene (Fig. 4e), reflecting the deleterious consequences by its own of the leucine-to-proline change on the incorporation of subunit 6 into ATP synthase.
Both ectopic strains (5 0 UTR COX2 -ATP6 WT and 5 0 UTR COX2 -atp6 L173P ) express Arg8 from the 5 0 UTR of ATP6 (Fig. 4a). As they are both deficient in the incorporation of subunit 6 into ATP Fig. 6. Upregulation of the mitochondrial ATP9 locus by ATP9-nuc and/or a lack in Atp6 is F 1 -dependent. a) Cells with the indicated genotypes were grown in rich galactose medium and their mitochondrially encoded proteins were radioactively labeled for 20 min with [ 35 S]-methionine and [ 35 S]cysteine in the presence of cycloheximide to inhibit cytosolic translation. Total protein extracts were then prepared and resolved by SDS/PAGE in 2 types of gel loaded with the same amount of radioactivity (as in Fig. 2). After drying of the gels, the proteins were using a PhosphoImager. b) Quantification of subunit 9 and Arg8 expressed from the ATP9 locus (atp9ϻARG8 m ). The levels of subunit 9 and Cox3 in each sample were quantified and compared to each other. Those in lanes 1-4 are compared to each other after setting to 1.0 the Atp9/Cox3 ratio lane 1. Those in lanes 7-10 are compared to each other after setting to 1.0 the Atp9/Cox3 ratio in lane 7. The levels of Arg8 and Var1 were quantified in lanes 5, 6, 11, 12. The Arg8/Var1 ratio in lane 5 was set to 1.0 and compared to the one in lane 6. The Arg8/Var1 ratio in lane 11 was set to 1.0 and compared to the one in lane 12.
synthase it was expected that Arg8 should be synthesized similarly in the 2 strains. This was indeed observed (Fig. 4f).

Regulation of ATP9
Translation of subunit 9 is stimulated by ATP9-nuc and a lack in subunit 6 Because of the primary sequence differences between the fungal Atp9-nuc and yeast subunit 9 proteins (D equard-Chablat et al. 2011) and the poor capacity of Atp9-nuc to support the assembly of a functional ATP synthase in yeast cells lacking subunit 9 (Bietenhader et al. 2012), we reasoned that expressing Atp9-nuc in wild-type yeast could influence translation and/or assembly of the yeast subunit 9 protein. In support of this hypothesis, co-expressing Atp9-nuc and subunit 9 resulted in an accelerated rate of subunit 9 synthesis, whereas that of the other mitochondrial gene products was unaffected (Fig. 5a). This effect was lost when expression of the plasmid-borne fungal gene was inhibited with doxycycline (Fig. 5b), which demonstrated that Atp9-nuc was truly responsible for the increased rate of subunit 9 synthesis. The steady state levels of ATP9 mRNA and its protein product Fig. 7. Influence of a lack in Atp12 or subunit 6 on the steady state levels and assembly of subunit 9. a) Steady states levels of subunit 9 in atp12D and atp6D yeasts. Total protein extracts from WT, atp6D, and atp12D strains grown in rich galactose medium were prepared and separated by SDS-PAGE in a 12% polyacrylamide gel. The proteins were transferred to a nitrocellulose membrane, and probed with antibody against subunit 6 (Atp6), subunit 9 (Atp9), and porin (Por1). The reported values (expressed as %WT) were calculated from 3 independent experiments. Statistical significance of the differences between samples is indicated. b) Assembly of subunit 9 in atp12D and atp6D yeasts. Mitochondria were isolated from WT, atp6D, and atp12D strains grown in rich galactose medium. Digitonin extracts containing the same quantity of proteins (50 mg) were prepared and separated in 3-10% polyacrylamide BN gel, transferred to a PVDF membrane, and probed with antibodies against subunit 9. The F 1 -F O monomers and dimers, and subunit 9-ring are identified in the left-hand margin. Fig. 8. Model of assembly-dependent translation of subunits 6 and 9 of yeast ATP synthase. a) Previously reported model (Rak et al. , 2011. 1: The subunits of F 1 (a, b, c, d, e) assemble with the help of Atp11, Atp12, and Fmc1. 2 and 3: The subunit 9 is produced and assembled with the help of Aep1, Aep2, Atp25-C, and Atp25-N independently of any other ATP synthase component. 4: The F 1 alone is involved in the activation of subunit 6 translation by Atp22. 5: The F 1 and the 9 10 -ring associate to each other. 5: The subunit 6 with the help of Atp10 and Atp23 is incorporated into ATP synthase. b) Model based on the results reported in this study. 1: The subunits of F 1 (a, b, c, d, e) assemble with the help of Atp11, Atp12 and Fmc1. 2, 3: The synthesis and assembly of subunit 9 with the help of Aep1, Aep2, Atp25-C, and Atp25-N is F 1 -dependent. 4: The F 1 -9 10 intermediate stimulates translation of subunit 6 by Atp22. 5: The subunit 6 with the help of Atp10 and Atp23 is incorporated into ATP synthase. The peripheral stalk subunits 8, 4, i, j, f, d, and OSCP of the ATP synthase monomer are not represented.
were identical in untransformed WT and WT cells transformed with the Atp9-nuc gene ( Fig. 5c and d). Cumulatively, these results showed that effect from Atp9-nuc was on translation, not transcription, and that the accelerated translation of subunit 9 did not lead to a stable accumulation of the overexpressed protein. One could hypothesize that the enhanced labeling of subunit 9 in WT expressing ATP9-nuc resulted from interactions with Atp9-nuc that stabilize subunit 9 and not from a higher rate of translation. At odds with this, the expression of Arg8 from the atp9ϻARG8 m locus in atp9D yeast (in which the subunit 9 is thus absent) was also considerably stimulated by Atp9-nuc (Fig. 5e), which definitely demonstrates that the higher labeling of subunit 9 in WT yeast expressing Atp9-nuc was caused by a faster rate of synthesis of this protein.
While subunit 6 was synthesized normally in WT cells expressing the fungal gene for Atp9-nuc (Fig. 5a), the steady state levels of subunit 6 were diminished by about 50% relative to untransformed WT (Fig. 5f). In view of the fact that unassembled subunit 6 is highly susceptible to proteolytic degradation (Lefebvre-Legendre et al. 2001;Kucharczyk et al. 2009b;Rak et al. 2009;Godard et al. 2011;Kabala et al. 2014;Su et al. 2020), the lack of subunit 6 induced by Atp9-nuc suggested that Atp9-nuc results in ATP synthase formation defects (see below for a possible explanation). Possibly, the lack in subunit 6 was responsible for, or contributed to the stimulation of subunit 9 translation. Indeed, subunit 9 translation was found stimulated in a strain lacking the ATP6 gene (atp6D) [(Rak et al. 2007b), see also Fig. 6a, lane 3  and b). However, transforming atp6D yeast with Atp9-nuc resulted in a much higher (5.6-fold) stimulation of subunit 9 synthesis (Fig. 6a, lane 4), indicating that the partial lack in subunit 6 in WT cells expressing Atp9-nuc was not alone responsible for the enhanced rate of subunit 9 translation.
Upregulation of ATP9 induced by ATP9-nuc and/or a lack in subunit 6 is F 1 -dependent The rate of subunit 9 synthesis was found unaffected in strains lacking the a or b subunit of F 1 or unable to assemble these proteins due to the absence of Atp11 or Atp12, indicating that the F 1 does not influence subunit 9 synthesis (Rak et al. , 2011. However, in the absence of Atp12, subunit 9 synthesis was no longer upregulated in strains expressing ATP9-nuc and/or lacking subunit 6 (Fig. 6a, lanes 2, 3, 4 vs. 8, 9, 10, respectively, and b), indicating that expression of subunit 9 is not F 1 -independent. The F 1 -dependency of the ATP9 locus expression is further supported by the enhanced labeling of Arg8 from the atp9ϻARG8 m allele (atp9D) in the presence of ATP9-nuc (Figs. 5e and 6a, lanes 5 vs. 6, and b) and the loss of this effect when the ATP12 gene was deleted (Fig. 6a, lanes 11 vs. 12, and b). Another interesting observation is the much higher labeling of Arg8 in atp9D vs. atp6D yeasts (Fig. 6a, lanes 3 vs. 5), which is not very surprising due to the presence of 10 subunits 9 for 1 subunit 6 in ATP synthase.
The poor labeling of Cox1 in lanes 3-12 results from the lack of functional ATP synthase. This effect was observed in yeast ATP synthase defective mutants as long as there are no proton leaks through the F O Rak et al. 2007aRak et al. , 2007bKucharczyk et al. 2009aKucharczyk et al. , 2009bKucharczyk et al. , 2010, which is the case in all the mutant strains with a reduced Cox1 labeling in Fig. 6. Based on these observations, it has been argued that the biogenesis of complex IV is modulated by the activity of F O (Su et al.

2019).
Subunit 9 is more susceptible to degradation and fails to assemble in the absence of F 1 Despite a good rate of synthesis (Fig. 6a, lanes 1 vs. 7), the steady levels of subunit 9 were diminished by about 80% in atp12D vs. WT (Fig. 7a) and the residual subunit 9 proteins were not detected as oligomeric rings in BN gels as they were in atp6D yeast (Fig. 7b). These observations suggest that formation/stability of the subunit 9-ring is F 1 -dependent (see below).

Discussion
There is solid evidence in the literature supporting a model in which ATP synthase formation starts with the assembly of F 1 followed by association with the subunit 9-ring (9 10 -ring) and peripheral stalk subunits, and finally incorporation of subunit 6, in a process that involves a number of helper proteins with specific actions Ackerman and Tzagoloff 2005;Rak et al. 2007b;Kucharczyk et al. 2008;Rak et al. 2009). In contrast, very few studies have so far been devoted to the mechanisms allowing the ATP synthase subunits to be produced in the good stoichiometry from 2 cell compartments (the cytosol and the mitochondrial matrix), an issue we have addressed in the present study.
It has been reported that translation of subunits 6 and 8 is largely inhibited in strains with a virtual absence of F 1 due to genetic ablation of a-F 1 or b-F 1 or 1 of 2 proteins (Atp11 and Atp12) that associate them to each other while the synthesis of subunit 9 is unaffected by the loss of F 1 (Rak et al. , 2011. On this basis, it was suggested that the F 1 is the sole component of ATP synthase involved in the activation/stimulation of subunits 6 and 8 synthesis (by Atp22), and that the 9 10 -ring forms separately, independently of any other ATP synthase component (Fig. 8a). This view is challenged by the present study.
We provide evidence that both the F 1 and the 9 10 -ring influence the synthesis of subunit 6. Indeed, point mutations in this protein that partially compromise its assembly result in the acceleration of the rate of subunit 6 synthesis ( Fig. 1), and this effect was lost in strains with a diminished capacity to assemble the F 1 (Fig. 2) or the 9 10 -ring (Fig. 3C), and in a strain expressing subunit 6 from the 5 0 UTR of the COX2 gene (Fig. 4). These data suggest the existence of an assembly-dependent regulatory loop in which functional interaction of Atp22 [the translational activator of subunit 6 (Zeng et al. 2007a)] with the 5 0 -UTR of the ATP6 mRNA involves both the F 1 and the 9 10 -ring rather than F 1 alone as was previously proposed (Rak et al. , 2011 (Fig. 8B).
Interestingly, while wild-type subunit 6 was efficiently synthesized from the 5 0 -UTR of COX2 (Fig. 4c), there was at the steady state about 50% less subunit 6 than in wild-type yeast (Fig. 4d) and defects in ATP synthase were observed as evidenced by the presence of free F 1 in mitochondrial samples resolved by BN-PAGE (Fig. 4e). Possibly, when not produced from its native 5 0 -UTR, nascent subunit 6 interacts less efficiently with Atp10 and Atp23 [2 proteins specifically required to incorporate subunit 6 into ATP synthase (Ackerman and Tzagoloff 1990a;Osman et al. 2007;Zeng et al. 2007c), and this reduces the yield in assembly. This conclusion is in line with previous studies that have revealed the importance of 5 0 -UTRs in post-translational biogenesis of other yeast mtDNA-encoded proteins (Sanchirico et al. 1998;Gruschke and Ott 2010;Gruschke et al. 2012).
Somewhat intriguingly, while translation of subunits 6 and 8 is largely inhibited in strains with a virtual absence of F 1 (Rak et al. , 2011, it is fully preserved with a 90% drop in F 1 in yeast cells lacking a protein (Fmc1) that stabilizes Atp12 at 36 C [(Lefebvre-Legendre et al. 2001), Fig. 2]. The present study provides an explanation for these observations. Indeed, Atp22 was barely detectable in atp11D and atp12D mitochondria, whereas it accumulates well in those from fmc1D yeast grown at 36 C ( Supplementary Fig. 1). The block in subunit 6 synthesis in strains totally lacking the F 1 is thus probably due to the absence of Atp22 in mitochondria. Consistently, subunit 6 synthesis was fully recovered in these strains upon transformation with a high copy number plasmid harboring ATP22 (Rak and Tzagoloff 2009). The lack of Atp22 in atp11D and atp12D mitochondria might result from the extremely detrimental consequences of a total or virtual lack in F 1 . Indeed, these mitochondria have a very poor content in complexes III and IV (Ebner and Schatz 1973), are very unstable genetically (Ebner and Schatz 1973), energize very poorly the inner membrane (Lefebvre-Legendre et al. 2003), and show severe defects in protein import (Yuan and Douglas 1992), whereas mitochondria from fmc1D cells grown at 36 C are much less affected (Lefebvre-Legendre et al. 2001) and do not show defects in protein import (Aiyar et al. 2014). Thus, major mitochondrial damage rather than disruption of a specific F 1 -dependent regulatory mechanism seems responsible for the lack of subunits 6 and 8 synthesis in yeast cells totally devoid of F 1 .
In revealing that subunit 6 synthesis is modulated by both the F 1 and the 9 10 -ring, our study describes a novel paradigm for the regulation of subunit 6 in which the F 1 -9 10 subcomplex rather than the F 1 alone modulates the synthesis of this protein. This model is in line with the concept of assembly-dependent regulation loop developed for cytochrome b and Cox1 according to which the synthesis of these proteins is stimulated by their direct protein partners (the F 1 is not a direct partner of subunit 6). As the subunit 6 is incorporated after formation of the F 1 -9 10 intermediate, in all the models thus far proposed Ackerman and Tzagoloff 2005;Rak et al. 2007bRak et al. , 2009Kucharczyk et al. 2008), it makes more sense that this intermediate rather than the F 1 alone signals the need for producing subunit 6 as a means to couple the synthesis of this protein to its assembly.
Unlike previous studies (Rak et al. , 2011, we conclude that the subunit 9 is also part of an assembly-dependent regulatory loop. A first clue was the higher pulse labeling of subunit 9 without any change in the levels of the ATP9 mRNA in WT yeast transformed with a nuclear ATP9 (PaAtp9-5) gene from P. anserina (D equard-Chablat et al. 2011), here referred to as ATP9-nuc, and the loss of this effect after genetic ablation of F 1 (Figs. 5 and 6). As ATP9-nuc also stimulated expression of Arg8 in a strain where the coding sequence of ATP9 was replaced with ARG8m (Figs. 5e and 6a, lanes 3 and 5, and Fig. 6b), it is evident that ATP9-nuc stimulated the rate of translation at the ATP9 locus whether the subunit 9 was or not present.
The ability of Atp9-nuc to partially restore the capacity of atp9D cells to grow in respiratory media shows that it can interact and function within yeast ATP synthase, albeit less efficiently than subunit 9 (Bietenhader et al. 2012). In the absence of Atp25-Nter, a polypeptide required for assembling the 9 10 -ring (Zeng et al. 2008), the compensation of atp9D yeast by Atp9-nuc is less efficient ( Supplementary Fig. 2), indicating that Atp9-nuc can functionally interact with Atp25-Nter. As a result of the interactions between Atp9-nuc and Atp25-Nter, there is less Atp25-Nter available for the assembly of the 9 10 -ring. This may be responsible for the reduced content of subunit 6 in WT expressing Atp9nuc (50% vs. WT) despite a normal rate of synthesis of this protein. Indeed, when it cannot interact properly with its protein partners, the subunit 6 becomes highly susceptible to proteolytic degradation (Ackerman and Tzagoloff 1990a;Zeng et al. 2007bZeng et al. , 2008Kucharczyk et al. 2009b;Godard et al. 2011). There is thus no doubt that expressing Atp9-nuc in WT yeast results in defects in the assembly of ATP synthase, and this is what possibly causes the upregulation of subunit 9. Indeed, it is a reasonable assumption that the 9 10 -ring forms progressively with the addition one by one of subunit 9 monomers and newly synthesized subunit 9 is provided until completion of the ring. In the presence of Atp9nuc, incomplete subunit 9 oligomers will be more abundant and in response, subunit 9 is synthesized more rapidly. The enhanced rate of subunit 9 translation observed in the absence of subunit 6 [3-4Â in atp6D yeast; 5.6Â in atp6D yeast expressing Atp9-nuc ( Fig. 6)] may be explained similarly. Indeed, when not wedged between the F 1 and subunit 6, the 9 10 -ring is certainly less stable and dissociates more frequently.
Another interesting finding here reported is the high susceptibly to degradation of unassembled subunit 9, either when it is produced in excess (in WT þ Atp9-nuc, Fig. 5) or when the F 1 is absent (in atp12D yeast, Fig. 7a). The residual subunit 9 polypeptides in atp12D yeast (20% vs. WT) could not be detected as oligomers in BN-gels (Fig. 7b), and newly assembled 9 10 -ring could not be detected after pulse labeling in mitochondria isolated from a strain lacking the F 1 (Rak et al. 2011). These observations indicate that the subunit 9 cannot assemble in the absence of F 1 . It has been proposed that the entire process of yeast mitochondrial gene expression occurs within large platforms called MIOREX that interact with the mitochondrial ribosome and proteins that help the assembly of the OXPHOS complexes (including those involved in subunit 9 biogenesis) (Kehrein et al. 2015a(Kehrein et al. , 2015b. It is tempting from the data here reported to speculate that newly formed F 1 interacts with this platform to facilitate cotranslational insertion of subunit 9 within ATP synthase (Fig. 8b).
The present study reveals that the synthesis of subunits 6 and 9 is coupled to the incorporation of these proteins into ATP synthase by mechanisms similar to those involved in the regulation of complexes III and IV biogenesis. In addition to increasing the yield in assembly of these 2 proteins, these regulations are likely important to prevent accumulation of free F O and F 1 particles that can dissipate the mitochondrial membrane electrical potential and the main source of chemical energy of the cell.

Data availability
Strains and plasmids are available upon request. The authors affirm that all data necessary for confirming the conclusions of the article are present within the article, figures, and tables.
Supplemental material is available at GENETICS online.

Conflicts of interest
None declared.