Saccharomyces cerevisiae strains used in this study are shown in Table 1. Yeast strains were grown at 28°C yeast extract-peptone containing 2% dextrose (YPD), 2% galactose (YPGal) or 3% glycerol (YPGlyc) with or without the indicated supplementation of valinomycin. Synthetic dextrose (SD) or galactose (SGal) media contained 0.67% yeast nitrogen base, 50 mM succinic acid (pH 5.5) and 2% of the respective energy source. According to the auxotrophies of each strain, methionine (10 mg/l), histidine (10 mg/l), leucine (10 mg/l), or uracil (25 mg/l) were added. For the estimation of growth efficiencies, fresh overnight precultures were diluted in triplicate in multiwell plates to the same OD. Growth was then continuously monitored under the indicated conditions in a Bioscreen C system (Thermo) for the indicated times. For the degradation assays for specific respiratory complex subunits, cells expressing the indicated protein as a TAP fusion from the genomic locus were pregrown in SGal medium to exponential phase and then treated for the indicated times with 4 μM valinomycin.

We used plasmid pVT100U-mtGFP, which contains mitochondrially targeted GFP expressed from the ADH1 promoter, to visualize the complete mitochondrial network in yeast (Westermann and Neupert, 2000). For the microscopic survey of mitophagy we used plasmid pVT100-mtRosella, which contains the dual fluorescent indicator Rosella targeted to mitochondria expressed from the ADH1 promoter (Mijaljica et al., 2011). For constitutive overexpression of Atg11 fused with dsRed from the TDH3 promoter, we inserted the entire ATG11 gene into plasmid pAG415-GPD-dsRed (Alberti et al., 2007). For the constitutive expression of HA-tagged Atg11 expressed from the TDH3 promoter, we inserted the entire ATG11 gene into yeast expression plasmid pAG416-GPD-HA (Alberti et al., 2007). A yeast two-hybrid bait plasmid containing lexA-Atg11 expressed from the ADH1 promoter was constructed by insertion of the ATG11 gene into plasmid pBTM116.

Yeast whole cell extracts were prepared instantaneously by boiling the cell pellet from 10 ml of fresh culture in 2x Laemmli buffer (120 mM Tris/HCl pH 6.8; 3% SDS; 40 mM DTT; 4 mM EDTA; 12% saccharose; 0.1 mg/ml bromphenolblue) for 5 min. Proteins were separated by 10% SDS-PAGE and analyzed by immunoblotting on PVDF membranes using anti-peroxidase-anti-peroxidase (anti-PAP) antibody (Sigma; 1:10.000) and peroxidase labeled anti-rabbit (Amersham Biosciences; 1:10.000). In the coprecipitation experiments, we additionally applied an anti-HA antibody (12CA5 Roche; 1:10.000). The bands were visualized with ECL Plus (Amersham Biosciences) and quantified with a Fujifilm LAS3000 or ImageQuant LAS4000 system. DB71 staining of the membranes was used as a loading control (Hong et al., 2000).

Vacuoles were stained with CellTracker Blue CMAC (Life Technologies) in living cells. Cells were observed on a Leica confocal microscope LSM 780 using the following excitation and emission wavelengths: GFP/pHluorin (excitation 488 nm; emission 509 nm), dsRed (excitation 545 nm; emission 572 nm), CMAC (excitation 353 nm; emission 466 nm).

For the quantification of the respiration rate, yeast cells were exponentially grown in SGal medium, washed with water and finally resuspended in 40 mM NaPO₄ pH 7.4 with 1% glucose. Oxygen consumption was then quantified in intact cells using a Mitocell S200 Respiratory System (Strathkelvin Instruments) with a Clarke type oxygen electrode. Oxygen consumption rates were determined from at least three independent culture aliquots.

The entire ATG11 gene was cloned into plasmid pBTM116 to create a full length Atg11-lexA fusion used as the bait in the two hybrid screening. A genomic yeast library cloned in the pACT plasmid fused to the Gal4 transcriptional activation domain was co-transformed with the Atg11-lexA bait vector. Total transformants were collected from SD-trp-leu selective plates in TE buffer and then plated in different dilutions onto SD-trp-leu-his-ade plates containing 4 μM Valinomycin. Positive clones were verified by colorimetric β-gal assays on filter papers and quantitative β-gal assays using liquid yeast cultures.

Atg11-TAP expressing yeast cells were transformed with the Atg11-HA expression plasmid or the empty vector control. The transformants were grown to exponential phase under the indicated conditions and total protein extracts were obtained by glass bead lysis in buffer A (50 mM Tris HCl pH 7.5; 15 mM EDTA; 2 mM DTT; 150 mM NaCl; 1 mM PMSF; 0.1% Triton X-100). Equal amounts of total protein were incubated with Dynabeads Pan Mouse IgG (Invitrogen) for 2 h in the cold. Extensive washing with buffer A was followed by extract preparation in 2x Laemmli buffer and heating for 5 min at 95°C. Precipitated TAP proteins were analyzed by anti-PAP western blot, while coprecipitated HA-tagged proteins were analyzed by anti-HA western blot.

Exponential cell culture aliquots upon the indicated growth conditions were incubated for 30 min with 2′, 7′-dichlorodihydrofluorescein diacetate (H₂DCFDA; Sigma) at a final concentration of 10 μM. H₂DCFDA is a chemically reduced form of fluorescein and used as a quantitative indicator of ROS levels in living cells. Inside the cell, it is cleaved by intracellular esterases, while further oxidation converts it to the fluorescent 2′, 7′-dichlorofluorescein. The cells were washed with water and resuspended in 1 ml of 50 mM Tris/HCl pH 7.5. After the addition of 10 μl of chloroform and 5 μl of 0.1% SDS, cell extracts were prepared by rigorous agitation (max speed for 30 s in a vortex homogenizor) with glass beads (0.5 mm diameter). Fluorescence was quantified in the supernatant in a microplate reader at 492 nm excitation and 525 nm emission wavelengths and normalized for the fluorescence of the same number of mock-treated cells.

Cycloheximide chase experiments were performed using yeast strains expressing different TAP-tagged subunits of the mitochondrial electron transport chain from their genomic loci. The experimental procedure is described in (Buchanan et al., 2016). Cycloheximide (Sigma) was used at a final concentration of 250 μg/ml in exponentially grown yeast culture aliquots.

We studied the abundance of individual subunits of the five respiratory complexes in yeast upon low respiratory rates during glucose growth or upon highly induced respiration during glycerol growth (Figure 1). We employed fully functional TAP-tagged versions of the different ETC subunits: Nde1 and Ndi1 (CI), Sdh1 and Sdh2 (CII), Cor1 and Qcr2 (CIII), Cox6 (CIV), and Atp1 and Atp5 (CV) (Figure 1A). Although budding yeast has single alternative oxidases (NADH:ubiquinone oxidoreductases Ndi1 and Nde1) replacing the function of the multisubunit ETC complex I present in higher eukaryotes, they are named as CI throughout the work to facilitate the presentation of the results. All proteins tested were detected in whole cell extracts as a major band corresponding to the full-length protein upon fermentative glucose growth. However, in cells actively growing in non-fermentative media, we observed the appearance of many degradation products in the case of the Nde1, Ndi1, Cor1 and Qcr2 proteins representing mitochondrial complexes CI and CIII (Figure 1B). In all cases, the abundance of multiple degradation bands was actually greater than the full length protein, which indicated that active respiration in yeast cells was accompanied by an important turnover of CI and CIII respiratory complexes.

We next addressed the question whether the increased degradation of certain respiratory complexes correlated with the respiration rate of the cells. The distribution of full length and degradation products of the Ndi1 and Qcr2 proteins was assayed in yeast cells grown in fully fermentative (glucose), partially respiratory (galactose) or fully respiratory (glycerol) conditions. As shown in Figure 1C, the percentage of degradation of the two proteins increased gradually with the increased dependence of the cells on respiratory metabolism.

Protein steady state levels are determined by the rates of their degradation and de novo synthesis. In order to determine the half-life of specific ETC subunits, we performed cycloheximide chase experiments upon fermentative and respiratory growth conditions. We found that the CV subunit Atp1 was stable during 1 h of impaired protein synthesis independently of the respiration rate (Figure 1D). The same result was obtained for the CI and CIII subunits Ndi1 and Cor1 upon fermentative conditions, which, however, were rapidly degraded upon respiratory growth (Figures 1D,E). These results suggested that specific ETC complexes, namely CI and CIII, are more rapidly degraded in yeast cells, which actively respire.

We next wanted to gain insights into the mechanisms underlying the respiration rate dependent degradation of mitochondrial complexes. We focused at the CIII subunit Qcr2 and analyzed its induced degradation in mitophagy mutants (atg11, atg32) and mutants with a lack of function of specific internal mitochondrial proteases (afg3, pim1, yme1). As shown in Figure 2, only mutations of the inner mitochondrial membrane i-AAA protease Yme1 and the Lon mitochondrial matrix protease Pim1 significantly decreased the percentage of Qcr2 degradation upon respiratory growth. These results indicated that the enhanced degradative turnover of respiratory complexes upon active respiration depends on the activity of specific mitochondrial proteases and not on mitophagy.

The accelerated degradation of CI and CIII subunits of the mitochondrial electron transport chain could be the result of the accumulation of damage at these specific complexes upon heavy respiration. The proteolytic removal of CI and CIII could therefore be necessary to maintain the overall function of the mitochondrial electron transport chain. We wanted to go a step further and test how the turnover of specific ETCs responds to mitochondrial damage. We chose the antibiotic valinomycin to provoke mitochondrial dysfunction, because it locally disrupts the mitochondrial membrane potential independently of a specific ETC by the formation of K⁺ selective pores. We first confirmed the suitability of valinomycin treatment of yeast cells as a model to study the adaptation to mitochondrial damage. As shown in Figure 3, valinomycin inhibits yeast growth in the low μM range specifically upon respiratory conditions and reduces the O₂ consumption rate up to 80%. We next performed kinetic quantitative western analysis to measure the effect of valinomycin treatment on the turnover of subunits of the individual ETCs. We found an increased degradation over time specifically at the CI and CIII subunits Ndi1 and Qcr2 and for cytochrome c, but not at CII, CIV or CV (Figures 3C,D). These results indicated that also externally induced mitochondrial dysfunction induces ETC specific protein degradation in yeast.

We next tested whether mitophagy or the function of the mitochondrial fusion and fission machinery was implied in the observed selective degradation of mitochondrial proteins upon valinomycin treatment. We quantified the abundance of the Ndi1 and Qcr2 proteins during valinomycin exposure in the mitophagy mutants atg11 and atg32 and in the fusion/fission mutants fzo1 and fis1 (Figures 4A,B). We found that the mitophagy adaptor Atg11 was essential to decrease Ndi1 and Qcr2 protein levels upon damage. However, the Atg32 mitophagy receptor was dispensable in this process. Furthermore, we confirmed that mitochondrial dynamics, both fusion and fission, were important for efficient Qcr2 degradation, while for Ndi1 degradation we only observed a dependence on mitochondrial fusion. These data suggested that valinomycin-induced mitochondrial damage triggered a mitophagic response dependent on mitochondrial dynamics to degrade specific ETCs from the organelles network. However, specific respiratory complexes might depend on mitochondrial fusion and fission processes to a different degree.

We wanted to monitor a selective removal of specific ETC subunits from yeast mitochondria upon damage by alternative approaches. We therefore quantified the abundance of functional GFP-fusion proteins of CI (Ndi1), CIII (Qcr2) and CV (Atp5) by confocal fluorescence microscopy and compared it to whole mitochondria visualized by mt-GFP. We acquired the integrated GFP intensities in each case over a cell population upon fermentative growth (glucose), partially respiratory growth (galactose) and after valinomycin treatment (Figure 5A). Expectedly, all mitochondrial ETC proteins increased during the shift from glucose to galactose growth similarly to the observed extension of the total mitochondrial network. Treatment of galactose grown cells with valinomycin reduced the total mitochondrial signal approximately to the level of glucose repressed cells, and a similar decrease was observed for the CV subunit Atp5. However, Ndi1- and Qcr2-GFP levels decreased significantly more upon external mitochondrial damage. These data indicated again that CI and CIII subunits are removed from the mitochondrial network upon damage in a selective manner.

In order to confirm our genetic data which indicated that a selective and Atg32 independent mitophagy occurred after mitochondrial damage in yeast, we employed the Rosella indicator to visualize the extent of autophagic digestion of mitochondrial parts in the vacuole upon valinomycin-induced damage. Rosella is directed to mitochondria by a specific targeting sequence and contains a pH-sensitive GFP fused to red fluorescent protein (dsRed). Upon normal growth, red and green light emission overlaps completely in mitochondria, whereas during mitophagy mtRosella enters the vacuole where the green signal is lost due to the acidic pH. As shown in Figure 5B, glucose or galactose cultures did not contain mitophagy-positive cells. However, treatment of galactose grown cells with valinomycin rapidly induced the percentage of cells, which engaged in mitophagy. Interestingly, at this short time point, the deprivation of a N-source, which is the standard treatment to induce autophagic degradation of mitochondria in yeast cells, did not produce mitophagy positive cells. However, vacuolar degradation of mitochondria was observed after prolonged N-starvation (Figure 5B). Taken together, we confirmed that valinomycin induced damage leads to the rapid removal of mitochondrial parts in the vacuole in a more instantaneous manner as compared to starvation conditions. We next tested the implication of the known mitophagy mediators Atg11 and Atg32. While Atg11 was absolutely necessary to induce mitochondrial degradation at the vacuole upon valinomycin damage, in atg32 mutant cells an induction of mitophagy-positive cells was still observed, which was only slightly reduced as compared to wild type (Figure 5B).

We next addressed the physiological relevance of the mitochondrial protein turnover by mitophagy upon valinomycin damage described above. The K⁺-ionophore valinomycin is known to induce an overproduction of mitochondrial ROS (Andrukhiv et al., 2006; Selivanov et al., 2008). As shown in Figure 6A, loss of function of the autophagic core proteins Atg8, Atg9 or Atg11 rendered yeast cells hypersensitive to valinomycin stress. By contrast, deletion of Atg32 did not produce a detectable phenotype. The treatment with the antioxidant glutathione partially rescued the sensitivity of the atg8, 9 and 11 mutants. Additionally we quantified the intracellular ROS accumulation before and after valinomycin treatment in wild type and the same autophagy mutants. As depicted in Figure 6B, the atg8, atg9, and atg11 mutants showed increased ROS levels upon mitochondrial damage, which was statistically significant in the atg8 and atg11 strains. Taken together, our results indicated that the core autophagic machinery including the Atg11 adaptor, but not Atg32, played an essential role in the damage-induced selective mitophagy, which is a physiologically important function for the removal of dysfunctional parts of the organelle.

In order to reveal mechanistic insights into the selective removal of damaged mitochondria in yeast, we studied the function of the Atg11 adaptor. Atg11 is present in yeast cells at a very low copy number (Ghaemmaghami et al., 2003) and we have confirmed that Atg11-GFP expressed at endogenous levels is practically undetectable in normally growing cells. We therefore studied the effects of a gain of Atg11 function and constitutively overexpressed the protein as a fusion with dsRed. Unexpectedly we found that overexpressed Atg11 localizes to the mitochondria, as shown in Figure 7A. Moreover, mitochondrial targeting of Atg11 was independent of Atg32, a known interactor for Atg11 at the mitochondrial outer membrane upon starvation. We then asked whether overexpressed Atg11 at mitochondria provoked changes in the turnover of CI and CIII subunits of the electron transport chain. We measured the percentage of degradation, which occurred at the Ndi1 and Qcr2 proteins upon the shift from fermentation to respiration in the absence and presence of Atg11 overexpression. As shown in Figure 7B, both subunits suffered an enhanced degradation already upon fermentative growth in the Atg11 gain of function strain. This degradation rate was comparable to the rate observed upon full induction of respiration. We next analyzed the valinomycin-induced mitophagy in the context of Atg11 overexpression with the Rosella indicator. Mitophagy-positive cells were exclusively detected after the induction of mitochondrial damage even during the constitutive overexpression of Atg11 (Figure 7C). Of note, Atg11 overexpression increased the number of cells committed to mitophagy slightly in the wild type and notably in the atg32 mutant. These date indicated that Atg11 association with mitochondria is an important step in the induction of mitophagy. However, mitochondrially targeted Atg11 alone is not sufficient to trigger mitophagy, which remains completely dependent on mitochondrial damage.

We next asked whether mitochondrial damage caused changes in the Atg11 protein. Previously, a starvation-induced interaction of Atg11 with Atg32 at the mitochondrial surface has been shown by bimolecular fluorescence complementation assays (Mao et al., 2013). Since the valinomycin-induced selective mitophagy described here is Atg11 dependent but Atg32 independent, we investigated the intracellular distribution of endogenously expressed Atg11-GFP during the diauxic shift and mitochondrial damage. As shown in Figure 8A, Atg11 cannot be detected in yeast cells actively growing in glucose or galactose. After a short induction of mitochondrial damage with valinomycin, we observed the appearance of Atg11 dots close to the mitochondrial network in approximately 35% of the cells. These results suggested that Atg11 concentrates in autophagosomal bodies at the mitochondria rapidly after the induction of damage. Interestingly, the same occurs initially during the diauxic shift, here provoked by the switch from glucose to galactose medium for 2 h. Nitrogen starvation induced mitochondrial Atg11 foci in a smaller number of cells (∼17%). We conclude that valinomycin-mediated mitochondrial damage rapidly induces the formation of Atg11 containing autophagosomal structures at the organelles surface.

Since it is unknown how Atg11 associates with yeast mitochondria upon damage, we performed a two-hybrid screen with Atg11 as the bait upon valinomycin treatment. Among the Atg11 interactors we could not identify a mitochondrial protein candidate, however, we found several clones corresponding to the C-terminal domain of Atg11. We confirmed that a small C-terminal domain (969-1179) was sufficient to interact with the full length Atg11 protein (Figure 8B). We then investigated whether the Atg11 dimerization or oligomerization was regulated during the diauxic shift or mitochondrial damage. We therefore performed in vivo immunoprecipitation assays using differentially tagged full length Atg11 proteins. As shown in Figure 8C, the Atg11-Atg11 interaction was not observed upon glucose or galactose growth, during the diauxic shift or upon short N-starvation. Exclusively after the exposure to valinomycin we observed a clear dimerization of Atg11. These data indicated that Atg11 dimer and/or oligomer formation in the yeast cell is inducible by mitochondrial damage and therefore might represent an early signaling event in the induction of selective mitophagy described here.