Mitophagy in yeast is independent of mitochondrial fission and requires the stress response gene WHI2

Over the past two decades, mitochondria have been recognized as a highly dynamic network constantly undergoing fusion and fission (Bereiter-Hahn and Voth, 1994; Nunnari et al., 1997). Several diseases, such as autosomal dominant optic atrophy type I (Alexander et al., 2000; Delettre et al., 2000), Charcot–Marie–Tooth neuropathy type 2A and 4A (Niemann et al., 2005; Zuchner et al., 2004) and Parkinson's disease (Banerjee et al., 2009; Exner et al., 2007; Yang et al., 2008), are associated with alterations in mitochondrial dynamics. Studies using yeast as a model organism have identified key players required for fusion and fission of mitochondria. Three proteins have been shown to be essential for mitochondrial fission, Dnm1, Fis1 and Mdv1, and these form a protein complex (Cerveny et al., 2001; Mozdy et al., 2000; Naylor et al., 2006; Otsuga et al., 1998; Sesaki and Jensen, 1999; Tieu and Nunnari, 2000). Loss of either corresponding gene results in mitochondria that display a net-like hyperfused morphology. A fourth factor, Caf4, has been reported to influence the localization of Dnm1 clusters on the mitochondrial surface, although it is not essential for fission of mitochondria (Griffin et al., 2005; Schauss et al., 2006). The mitochondrial fusion process requires Fzo1, Ugo1 and Mgm1 (Fritz et al., 2001; Hermann et al., 1998; Rapaport et al., 1998; Sesaki and Jensen, 2001; Sesaki and Jensen, 2004; Sesaki et al., 2003; Wong et al., 2000; Wong et al., 2003). Mgm1 is a large GTPase that mediates inner membrane fusion and its function has been shown to depend on mitochondria being in an intact bioenergetic state (Herlan et al., 2004; Herlan et al., 2003). In general, there is considerable knowledge on the molecular players involved in fusion and fission. However, the regulation and the functional significance of mitochondrial dynamics are still elusive.

One hypothesis is that fusion and fission dynamics act as a quality checkpoint for mitochondrial functionality (Duvezin-Caubet et al., 2006; Herlan et al., 2004; Kim et al., 2007; Lyamzaev et al., 2004; Twig et al., 2008). Our observations in yeast and mammalian cells, using several models of mitochondrial dysfunction, have shown that fusion competence is reduced when mitochondrial function is impaired (Duvezin-Caubet et al., 2006; Herlan et al., 2004). This causes dysfunctional mitochondria to become isolated from the intact network, hence providing a possible way to distinguish functional from dysfunctional mitochondria on a morphological basis; it has been hypothesized that this process could target damaged mitochondria for degradation by autophagy. Indeed, Twig and colleagues (Twig et al., 2008) have shown that, in mammalian cells, autophagy of mitochondria is impaired when fission of mitochondria is blocked and that mitochondrial depolarization precedes autophagy of mitochondria. Moreover, when autophagy is blocked, damaged mitochondria appear to accumulate (Krebiehl et al., 2010; Mortensen et al., 2010; Twig et al., 2008).

Autophagy is the process of delivering dispensable material to the lysosome or vacuole for degradation and is controlled by TOR (target of rapamycin) signaling (Carrera, 2004; Noda and Ohsumi, 1998; Yang and Klionsky, 2009). It comprises several pathways that differ in the selected cargo, as well as in the delivery pathway to the vacuole. Macroautophagy is characterized by degradation of cytosol in bulk and is therefore considered nonselective. Selective types of autophagy include the cytoplasm-to-vacuole targeting (Cvt) pathway, which delivers specific proteins to their final destination in the vacuole, as well as mitophagy and pexophagy, which, for example, remove superfluous organelles upon a shift of carbon source. Studies on the molecular mechanism of autophagy in yeast have identified several proteins involved in the process, termed the autophagy-related (Atg) proteins (Huang and Klionsky, 2007; Klionsky et al., 2003). Most of these proteins are required for all autophagy pathways; for example, Atg1 is required for autophagy induction and Atg8 for the formation of the cargo transport vesicle (the autophagosome). Selective types of autophagy are further characterized by recognition and targeting of the cargo, which is mediated by Atg11, Atg20 and Atg24 (Nice et al., 2002; Yorimitsu and Klionsky, 2005).

Less is known about the factors that influence the selective degradation of mitochondria, known as mitophagy. Recent reports confirm that mitochondria can be selectively targeted for autophagy (Kanki and Klionsky, 2008; Kim et al., 2007). In mammalian systems, PINK1 and PARKIN have been shown to be involved in targeting depolarized mitochondria to autophagosomes (Geisler et al., 2010; Narendra et al., 2010). Furthermore, NIX (BNIP3L) has been identified as an essential requirement for elimination of mitochondria during erythrocyte differentiation (Sandoval et al., 2008; Schweers et al., 2007) and has been shown recently to act as a selective receptor for mitophagy, interacting physically with LC3A (MAP1LC3A) and all GABARAP proteins (GABARAP, GABARAPL1, GABARAPL2) (Novak et al., 2010). In yeast, mitophagy has been suggested to depend on Uth1 (Kissova et al., 2004), Aup1 (Tal et al., 2007) and proteins required for selective forms of autophagy, namely Atg11, Atg20 and Atg24 (Kanki and Klionsky, 2008). Recently, Atg32 has been identified in two independent screens for yeast strains deficient in mitophagy (Kanki et al., 2009b; Okamoto et al., 2009). Furthermore, it has been shown that this factor is located at the mitochondrial outer membrane and thus it has been proposed to act as a receptor for targeting mitochondria to the autophagosomal membrane. In addition, Atg33 has been identified in one of the screens as being required for mitophagy (Kanki et al., 2009a). So far, mitophagy in yeast has been observed upon mitochondrial dysfunction (Nowikovsky et al., 2007; Priault et al., 2005), in stationary phase (Okamoto et al., 2009; Tal et al., 2007), and upon nitrogen starvation (Kanki and Klionsky, 2008; Kissova et al., 2007) and rapamycin treatment (Kissova et al., 2004). The latter two conditions are known to inhibit TOR signaling, which also leads to a general induction of autophagy (Carrera, 2004; Noda and Ohsumi, 1998).

The molecular mechanisms and the regulation of mitophagy remain largely unclear. Here, we focus on deciphering the role of mitochondrial dynamics in mitophagy in yeast. In particular, the role of mitochondrial fission is debated; whereas in mammalian cells fission has been shown to be required for mitophagy (Twig et al., 2008), two screens performed in yeast revealed apparently conflicting results in this regard. One study reported reduced induction of mitophagy upon nitrogen starvation in a dnm1 deletion strain, whereas deletion strains for the other mitochondrial fission factors, Mdv1, Fis1 and Caf4, did not show such a defect (Kanki et al., 2009a). A different study was unable to show that Dnm1, Mdv1, Caf4 or Fis1 has any effect on mitophagy (Okamoto et al., 2009). By contrast, induction of mitophagy, as observed by loss of Mdm38 function, has been proposed to depend upon Dnm1 (Nowikovsky et al., 2007). It is thus not resolved whether mitochondrial fission is required for mitophagy in yeast.

Our present data show no direct correlation between mitochondrial morphology and mitophagy. Fragmentation of mitochondria was not sufficient to induce mitophagy. In strains defective in mitochondrial fission, induction of mitophagy was as efficient as in wild-type strains. Reduced mitophagy activities were initially observed in a fis1 deletion strain (Δfis1) but not in any other fission mutant or upon expression of various dominant-negative variants of Dnm1. However, we found that the effect observed in Δfis1 was due to a secondary mutation in the locus of the stress response gene WHI2 (Whiskey 2). This genetic suppression had been reported previously to arise in Δfis1 and is responsible for an overgrowth phenotype in stationary phase (Cheng et al., 2008). Whi2 was suggested to regulate cell cycle control, as Δwhi2 cells fail to enter G1 and arrest randomly within the cell cycle in stationary phase (Radcliffe et al., 1997; Sudbery et al., 1980). Moreover, Δwhi2 mutants fail to accumulate storage glycogen and are sensitive to environmental stresses, such as heat shock. A later study provided the first biochemical evidence for its function in the regulation of Msn2 phosphorylation, thereby implicating a function for Whi2 in the general stress response (Kaida et al., 2002). Recently, it has been suggested that Whi2 influences the subcellular localization of Ras2, linking it to the Ras-protein kinase A (PKA) signaling pathway (Leadsham et al., 2009). Our present data demonstrate that Whi2 is required for an efficient induction of mitophagy, which could point to a link between TOR, Ras-PKA signaling, autophagy and the general stress response pathway. In summary, we suggest that fragmentation is neither the sole signal of mitochondrial dysfunction nor a prerequisite for mitochondria to be targeted for degradation in yeast.

In order to determine the extent of mitophagy, as well as autophagy, in a quantitative manner, we employed the modified alkaline phosphatase (ALP) assay described previously (Campbell and Thorsness, 1998; Noda et al., 1995). In a strain lacking the endogenous alkaline phosphatase Pho8, we expressed an inactive proenzyme of Pho8 that was targeted to either the mitochondrial matrix (mtPho8) or the cytosol (cytPho8). We confirmed the correct localization of both mtPho8 and cytPho8 by cell fractionation and proteinase K treatment (supplementary material Fig. S1B,C) and that neither deletion of PHO8 nor expression of these constructs impaired the growth of yeast cells (supplementary material Fig. S1A). It is known that proenzyme Pho8 is activated upon proteolytic processing by proteinase A (PEP4) within the vacuole. We found that the specific activities of mtPho8 and cytPho8 increased significantly when the respective strains were grown under nitrogen starvation or after addition of rapamycin (supplementary material Fig. S2A,B). ALP activities in the mock-treated cells and in cells lacking proteinase A (Δpep4) remained at background levels, confirming that mtPho8 and cytPho8 are exclusively activated in the vacuole and that background activities of other proteases and phosphatases do not impair the assay. We conclude that the increase in mtPho8 and cytPho8 activity reflects induction of mitophagy and autophagy, respectively.

First, we asked whether induction of mitophagy by rapamycin is a selective type of autophagy or whether it instead represents a random sequestration of mitochondria upon induction of bulk autophagy. Therefore, we investigated the induction of mitophagy and autophagy in mutants known to be deficient in selective types of autophagy, such as mitophagy, pexophagy and the Cvt pathway. Deletion of atg11 completely prevented induction of mitophagy by rapamycin but did not grossly impair bulk autophagy (Fig. 1A,B). Furthermore, mitophagy but not autophagy was reduced to ~40% in the atg20 and atg24 deletion strains (Fig. 1A,B), which is consistent with the known partial effect on mitophagy of these mutants (Kanki and Klionsky, 2008). Neither mitophagy nor autophagy was induced in a strain lacking the essential autophagy component Atg8 (Fig. 1A,B), further confirming the specificity of the ALP assay. To corroborate these results, we analyzed cell lysates of the aforementioned strains by SDS-PAGE and western blotting (Fig. 1C,D). The extent of the processing of Pho8 in the individual strains was consistent with the measured alkaline phosphatase activities, confirming that rapamycin-induced mitophagy depends upon Atg11, Atg20 and Atg24. In addition, we monitored the total levels of Tim23, a subunit of the mitochondrial TIM23 translocase, Aac2, the major ADP/ATP carrier of the inner membrane, and MPP, the mitochondrial processing peptidase. The levels of these markers decreased upon addition of rapamycin in the control strain but only marginally decreased in the strains deleted for either atg8 or atg11, demonstrating further that the degradation of mitochondria occurs in a selective manner. In the Δatg20 and Δatg24 strains, the levels of Tim23, Aac2 and MPP were diminished, to some extent, upon addition of rapamycin, which is consistent with the partial induction of mitophagy observed by the ALP assay. In general, we noted that monitoring mitophagy by western blotting of mitochondrial markers gave considerably more variable results. We attribute this to varying activities of vacuolar and mitochondrial proteases, either pre- or post-lysis, which are likely to affect the extent of degradation of individual marker proteins in an unequal and not entirely reproducible manner. By contrast, the ALP assay is very reproducible and specifically monitors the activation of the inactive proenzyme Pho8 within the vacuole without showing any Pep4-independent activation (e.g. by cytosolic or intramitochondrial proteolysis).

In addition, to further corroborate the biochemical results, we performed a visual assay for mitophagy. Cells expressing GFP–Atg8 and mtDsRed were analyzed by confocal fluorescence microscopy after rapamycin treatment. In mock-treated cells, GFP–Atg8 was evenly distributed in the cytosol and mtDsRed labeled a typical mitochondrial network. Upon treatment with rapamycin, accumulation of GFP in the vacuole was observed in nearly all cells, indicating that GFP–Atg8 was efficiently shuttled to the vacuole and that there was autophagy (Fig. 1E). Upon rapamycin treatment, the mitochondrial mtDsRed signal was unambiguously detectable in the vacuole in ~10% of the control cells but in none of the cells lacking Atg11 (Fig. 1E,F). Thus, according to this visual assay, as well as the biochemical assays, induction of mitophagy by rapamycin depended upon Atg11. In summary, we conclude that, upon rapamycin treatment, mitochondria are degraded through a selective form of autophagy and that the modified ALP assay used provides a reliable, quantitative and sensitive way of monitoring mitophagy and autophagy in yeast.

In order to assess whether fragmentation of mitochondria is sufficient to trigger mitophagy, we used the temperature-sensitive mgm1-5 strain, which shows defective inner membrane fusion at the non-permissive temperature of 37°C (Wong et al., 2000). The mitochondrial morphology of the mgm1-5 strain and a corresponding wild-type strain, both expressing mtPho8 and mtDsRed, was determined at 24°C and 37°C. At the non-permissive temperature, the mgm1-5 strain showed a complete fragmentation of mitochondria after 2 hours of incubation and an aggregation of mitochondria after 16 hours of incubation (Fig. 2A), consistent with published data for this strain (Wong et al., 2000). The wild-type control maintained the typical tubular network of mitochondria at both temperatures. Next, we analyzed mitophagy by the ALP assay at different time points of incubation at the non-permissive temperature. ALP activity remained at background levels for both strains after 5 or 9 hours of incubation, indicating that mitophagy was not induced by fragmentation of mitochondria (Fig. 2B). Next, we tested whether temporary fragmentation of mitochondria for 5 or 26 hours at the non-permissive temperature, followed by a shift back to the permissive temperature for 5 hours, would trigger mitophagy. However, shifting back to the permissive temperature also did not result in increased ALP activity (Fig. 2C). We conclude that fragmentation of mitochondria alone is not sufficient to trigger mitophagy. To exclude the possibility that a shift to 37°C, or impairing Mgm1 function itself, blocks mitophagy, we confirmed that rapamycin was still able to induce mitophagy efficiently at the permissive and the non-permissive temperature in both strains (data not shown). Furthermore, we observed that rapamycin treatment alone did not result in mitochondrial fragmentation, showing that fragmentation does not precede mitophagy in yeast (Fig. 1E; supplementary material Fig. S3). Rapamycin did not have a general impact upon the ability of mitochondria to generate an electrochemical gradient across the inner membrane, as mitochondria were still efficiently labeled by the membrane-potential-sensitive dye Rhodamine B hexyl ester (supplementary material Fig. S3). However, we cannot exclude a minor effect of rapamycin on the mitochondrial membrane potential, as there was a slight reduction in signal intensity. Nevertheless, fusion and fission of mitochondria proceeded in a normal manner after rapamycin treatment, in accordance with the overall wild-type-like mitochondrial morphology (Fig. 1E; supplementary material Fig. S3) and time-lapse fluorescence microscopy (data not shown). Thus, rapamycin itself does not appear to alter mitochondrial dynamics or cause major mitochondrial dysfunction.

It has been reported, for mammalian cells, that mitophagy and autophagy are induced by drugs that impair mitochondrial oxidative phosphorylation (Chen et al., 2007; Ding et al., 2010; Novak et al., 2010). To test whether this is also the case in yeast, we treated cells for 5 hours with carbonyl cyanide m-chlorophenylhydrazone (CCCP) to dissipate the mitochondrial membrane potential. We did not observe any induction of mitophagy or autophagy (Fig. 2D,E). Addition of CCCP efficiently dissipated the membrane potential, as virtually no mitochondrial labeling, but instead only weak cytosolic labeling, was observed using the membrane-potential-sensitive dye Rhodamine B hexyl ester (supplementary material Fig. S3). Furthermore, this treatment resulted in a fragmented mitochondrial morphology, consistent with the fact that mitochondrial fusion is inhibited in the absence of a membrane potential (Legros et al., 2002; Meeusen et al., 2004). In addition, a longer incubation (up to 24 hours) with CCCP did not lead to any induction of mitophagy or autophagy (data not shown). Thus, even CCCP-induced fragmentation of mitochondria is not sufficient to induce mitophagy. Induction of mitophagy or autophagy was also not observed when the F₁F_o-ATPase inhibitor oligomycin was used (Fig. 2D,E). Moreover, neither CCCP nor oligomycin treatment impaired rapamycin-induced mitophagy or autophagy, demonstrating that the ALP assay was not inhibited per se. In addition, these drugs did not exert a stimulatory effect on rapamycin-induced mitophagy. Taken together, these results show that induction of mitophagy is not triggered by mitochondrial fragmentation or inhibition of oxidative phosphorylation and apparently is mechanistically distinct between mammalian and yeast cells.

Next, we asked whether blocking mitochondrial fission inhibits mitophagy in yeast. We expressed Dnm1 variants with a dominant-negative effect on mitochondrial fission (Otsuga et al., 1998), and quantified the induction of mitophagy and autophagy upon rapamycin treatment. Expression of Dnm1^K41A and Dnm1^S42N resulted in a fishnet-like hyperfused mitochondrial morphology, confirming that these variants indeed exerted a dominant-negative effect (data not shown). Upon rapamycin treatment, ALP activities for mtPho8 and cytPho8 increased to a similar extent in the strains expressing the dominant-negative variants of Dnm1 to that in the control strains (Fig. 3A,B), indicating that neither mitophagy nor autophagy are affected when mitochondrial fission is impaired. We further confirmed that induction of mitophagy still depended upon Atg11, excluding the possibility that a non-selective type of mitophagy was induced upon expression of wild-type or dominant-negative variants of Dnm1 (supplementary material Fig. S4A). Western blotting revealed that processing of the proenzymes of mtPho8 or cytPho8 was not impaired by expression of the dominant-negative variants of Dnm1, consistent with the ALP activities determined above (Fig. 3C,D). In addition, rapamycin-induced degradation of the mitochondrial markers Tim23, Aac2 and MPP was not impaired in the presence of these Dnm1 variants, thereby confirming that inhibiting mitochondrial fission does not block mitophagy (Fig. 3C). The level of the fusion factor Mgm1 was affected in a similar way upon rapamycin addition but did not show any altered expression under conditions without rapamycin (Fig. 3C). In summary, these data suggest that mitophagy in yeast is independent of mitochondrial fission.

It cannot be ruled out that, upon expression of dominant-negative variants of Dnm1, a residual mitochondrial fission activity, sufficient for mitophagy, remained. To address this in more detail, we analyzed induction of mitophagy in strains lacking one of the fission factors Dnm1, Mdv1, Caf4 or Fis1. Consistent with the mitochondrial morphology reported earlier for these strains (Cerveny et al., 2001; Griffin et al., 2005; Mozdy et al., 2000; Naylor et al., 2006; Otsuga et al., 1998; Schauss et al., 2006; Sesaki and Jensen, 1999; Tieu and Nunnari, 2000), we observed a fishnet-like hyperfused mitochondrial morphology for the Δdnm1, the Δmdv1 and the Δfis1 strains, and a more wild-type-like tubular morphology for the Δcaf4 strain (Fig. 4E; supplementary material Fig. S4B; data not shown). We did not observe an alteration of mitochondrial appearance upon addition of rapamycin in any of these strains (Fig. 4E; supplementary material Fig. S4B; data not shown). The correct localization of mtPho8 in these mutant strains was confirmed by cellular subfractionation and western blotting (supplementary material Fig. S1D). Consistent with results obtained using dominant-negative Dnm1 variants, the Δdnm1 strain showed a normal increase in mtPho8 or cytPho8 activity upon rapamycin treatment, confirming that Dnm1-dependent fission is not required for mitophagy or autophagy (Fig. 4A,B). The same result was obtained for the Δmdv1 and Δcaf4 strains, further corroborating that fission is not a prerequisite for mitophagy or autophagy. In line with these results, we did not observe a physical interaction of these fission components with Atg8 using a yeast two-hybrid approach (data not shown). We also noted that CCCP and oligomycin did not induce mitophagy or autophagy, nor did they lead to an inhibition of rapamycin-induced mitophagy or autophagy in the Δdnm1 strain (supplementary material Fig. S4C,D). Thus, under hyperfused conditions, the ability of mitochondria to generate ATP or a membrane potential is also not required for these processes. However, the Δfis1 strain showed ~50% less induction of mitophagy and ~30% less induction of autophagy, when treated with rapamycin (Fig. 4A,B). These results were corroborated by western blotting (Fig. 4C,D) and the visual assay for mitophagy (Fig. 4E,F). Notably, we often observed a more intense GFP–Atg8 signal intensity in the Δfis1 strain compared with that in control cells (data not shown); this could be due to a decreased turnover or an increased expression of GFP–Atg8 in this strain.

We next undertook experiments to determine the cause of the apparent discrepancy regarding the effect on mitophagy in the different fission mutants. It had been reported that a secondary mutation in the WHI2 locus was present in a Δfis1 strain obtained from the EUROSCARF clone collection (Cheng et al., 2008). As it will become evident below, the Δfis1 strain from EUROSCARF we used also contained this secondary mutation. We thus refer to this strain as Δfis1*. In summary, although mitophagy can occur independently from fission, it seems to be reduced in the Δfis1* deletion strain.

To address whether Fis1 plays a mechanistic role, possibly distinct from fission, in mitophagy or whether the observed impairment in mitophagy was due to a whi2 mutation, we checked the Δfis1* strain for a potential whi2 mutation. Sequencing of a PCR product amplified from the WHI2 locus identified the same mutation (T206G) in Δfis1* as reported by Cheng and colleagues (Cheng et al., 2008); this mutation introduces a premature stop codon after 68 amino acids (Fig. 5A). As wild-type Whi2 consists of 486 amino acids, it is probable that the early stop codon results in a complete loss of function. Consequently, we analyzed Whi2 for its role in mitophagy and autophagy. Indeed, induction of mitophagy was significantly reduced in the Δwhi2 strain (Fig. 5B), resembling the effect in the Δfis1* strain. Induction of autophagy was decreased by ~10% in the Δwhi2 strain, although this decrease was not statistically significant (Fig. 5C). The visual mitophagy assay using confocal fluorescence microscopy revealed that mtDsRed accumulation in the vacuole was markedly reduced in Δwhi2 compared with that in the control strain, confirming the result obtained by the ALP assay (Fig. 5D,E). These data suggest that the mitophagy phenotype in the Δfis1* could indeed result from the loss of function of Whi2. Given these observations, we asked whether loss of Whi2 was specific to mitophagy and decided to test whether it could also affect the Cvt pathway. We found that Ape1 processing, and the increased Ape1 processing induced by rapamycin, was dependent on Atg1, as reported previously (Kim et al., 2001; Kim et al., 2002), but was not dependent on Whi2 (Fig. 5F). We conclude that Whi2 plays no general role in selective types of autophagy. In addition, we noted that overexpression of Whi2 in a strain expressing GFP–Atg8 led to an increased processing of the GFP moiety compared with that in the control strain lacking Whi2 (Fig. 5G). This suggests that Whi2 is functionally linked to Atg8 and appears to positively modulate autophagic flux in yeast cells. Taken together, we propose that Whi2 is a factor predominantly affecting mitophagy and, to a lesser extent, autophagy.

To address further the role of Fis1 and Whi2 in mitophagy, we expressed Whi2 and Fis1 from a low-copy-number plasmid under their respective endogenous promoters in the Δfis1* and the Δwhi2 strains. The ALP assay showed a recovery of mitophagy upon rapamycin treatment only when Whi2 was expressed but not when Fis1 was expressed (Fig. 6A). In addition, efficient induction of autophagy was only observed when wild-type Whi2 was reintroduced (Fig. 6B). These results were further confirmed by western blotting, as mtPho8 processing was more pronounced when Whi2 was expressed compared with that in the situations when functional Whi2 was lacking (Fig. 6C,D). In addition, rapamycin-induced reductions in the levels of the mitochondrial marker proteins Tim23, Aac2 and MPP were prominent when Whi2 was expressed, but were less pronounced in the absence of functional Whi2 (Fig. 6C,D). However, some Whi2-independent degradation of mitochondrial markers was detectable; we attribute this to the fact that mitophagy is not fully blocked in the absence of Whi2 and that, according to our own experience, western blotting is of limited reliability to monitor mitophagy. The efficient recovery of mitophagy upon expression of Whi2, but not Fis1, was also observed using the visual mitophagy assay (Fig. 6E). Using this assay, it appeared that the expression of Fis1 slightly promoted mitophagy. However, this minor increase was not significant and can be attributed to the rather high variation observed in the strains using this assay. Finally, we also generated a Δfis1 strain in a w303 background and confirmed by DNA sequencing that no suppressor mutation in the WHI2 gene was present. This strain did not show any impairment in the induction of mitophagy or autophagy by rapamycin (supplementary material Fig. S5A–C). Overall, several lines of evidence demonstrate that Whi2, but not Fis1, is required to fully induce mitophagy and autophagy after rapamycin treatment.

In this study, we demonstrate that mitophagy induced by inhibition of TOR signaling is independent of mitochondrial fission. Neither expression of dominant-negative variants of Dnm1 nor deletion of components of the mitochondrial fission machinery resulted in impaired mitophagy. Furthermore, we found that mitochondrial fragmentation due to a block in fusion is not sufficient to trigger mitophagy. These results might help to shed light on the conflicting results published previously. In particular, apparently opposing results were reported for the role of the fission component Dnm1 in mitophagy. In a previous study, downregulation of Mdm38 led to an induction of mitophagy, whereas mitophagy was not observed in a double deletion strain lacking Mdm38 and Dnm1 (Nowikovsky et al., 2007). This result led to the proposal that mitophagy depends on Dnm1 (Nowikovsky et al., 2007). However, a direct role for Dnm1 in this particular case still remains unclear – the extent of mitophagy in the single deletion strain lacking Mdm38 and the effect of impairing Dnm1 function upon downregulation of Mdm38 were not addressed. Another study has shown that mitophagy is reduced by 20–40% in a strain lacking Dnm1, suggesting that mitochondrial fission is crucial for mitophagy in yeast (Kanki et al., 2009a). However, despite the reduction observed under these conditions the data rather confirm that mitophagy is, in principle, efficiently induced in this strain. Furthermore, the same study did not show an impairment of mitophagy upon the lack of any other fission factor, which is inconsistent with the claim that mitochondrial fission impairs mitophagy. In line with our results, a different screen revealed that none of the four components involved in mitochondrial fission affected mitophagy (Okamoto et al., 2009). Another possibility, explaining some of the reported discrepancies discussed above, is that undetected secondary mutations in other genes have occurred. This is not unlikely in light of our results, as we also initially found a specific effect for Fis1, but not for the other fission factors, which we finally demonstrated was caused by a secondary mutation in WHI2. In rescue experiments, we demonstrated that it was not the lack of Fis1 that was responsible for the initial defect in mitophagy but rather a suppressor mutation in WHI2. In summary, several lines of evidence demonstrate that mitochondrial fission is not strictly required for mitophagy.

This unexpected result seems to put the hypothesis of fission and fusion dynamics acting as a quality control checkpoint for mitochondria into question. However, it is still far from understood how fusion and fission, as well as mitophagy, are initiated or regulated. We cannot rule out the possibility that, under normal growth conditions, basal levels of mitophagy in yeast are triggered in coordination with or as a consequence of mitochondrial fission, similar to the situation in mammalian cells (Twig et al., 2008). Such a low level of steady-state mitophagy might be too little to be detected with the biochemical methods used here. Moreover, we cannot exclude that under a specific condition in which mitochondrial dysfunction is induced, fission promotes degradation of damaged mitochondria. Such an induction of mitophagy might possibly require one or more additional triggers besides fragmentation. In mammalian cells, the induction of mitophagy through recruitment of PARKIN to mitochondria has been shown to depend on the dissipation of the membrane potential (Narendra et al., 2009), suggesting that mitochondrial dysfunction is involved in generating specific signals for mitophagy. In yeast, the induction of mitophagy by mitochondrial dysfunction has only been observed in a few studies (Nowikovsky et al., 2007; Priault et al., 2005). However, drug-induced dissipation of the membrane potential and inhibition of the F₁F_o-ATPase does not lead to an induction of mitophagy in yeast. Furthermore, neither a mitochondrial membrane potential nor the ATP-generating ability of mitochondria is required for rapamycin-induced mitophagy. In future, this and other models of mitochondrial dysfunction need to be investigated in greater detail. In summary, our study does not rule out the possibility that mitochondrial morphology is important for the selection of dysfunctional parts of mitochondria in proliferating cells or that it participates in signaling of a mitochondrial dysfunction to induce mitophagy. However, it shows that mitophagy in yeast does not necessarily require mitochondrial fission, and alternative pathways for mitochondrial quality control by mitophagy need to be considered.

As fission is not required for mitophagy in yeast, how can the sequestration of mitochondrial parts by autophagosomes be accomplished? Kim and colleagues (Kim et al., 2007) reported that mitochondrial fission and autophagosome formation are coordinated events. This means that, during that process, mitochondrial fission might be mediated by the autophagic machinery itself. Recently, organelle-specific mitophagy receptors in yeast and human have been identified, namely Atg32 and NIX, respectively (Kanki et al., 2009b; Novak et al., 2010; Okamoto et al., 2009). The interaction of those proteins with components of the autophagy machinery indicates their involvement in the selection of mitochondria for degradation. It is certainly possible that those and other factors also function in the fission process during autophagosome formation. Future studies will have to dissect whether this is indeed the case and whether mammalian and yeast cells differ in their capacity to separate parts from the mitochondrial network by autophagosome formation. Another possibility as to why mammalian cells but not yeast cells have evolved a dependency on the regular fission machinery could be linked to the size of the mitochondrial tubular network itself. In yeast, the dimensions of mitochondrial tubules are possibly small enough to allow them to be surrounded by autophagosomes. In addition, repeated cell divisions might be sufficient to generate small enough entities. By contrast, in mammalian cells, the larger cell size and the lower frequency of cell divisions (or its absence in post-mitotic tissues) might have led to the dependency on the mitochondrial fission machinery. Moreover, regulating mitochondrial dynamics would create an additional level of regulation for the selective degradation of mitochondria in mammalian cells, which is likely to be of particular importance for post-mitotic cells.

Our data further indicate that regulation of mitophagy in yeast is connected to the stress response and the Ras-PKA signaling pathway. We initially observed an effect on mitophagy in a Δfis1 deletion strain, which we demonstrated was caused by a secondary mutation in WHI2. This suppressor mutation has been described previously as arising in fis1 deletion strains and results in an overgrowth phenotype (Cheng et al., 2008). Although in that previous study, the EUROSCARF Δcaf4 strain was also reported to contain a mutation in WHI2, we did not observe a defect in mitophagy in the Δcaf4 strain used herein. Consistent with this result, our Δcaf4 strain did not contain a mutation in WHI2. So far, very little is known about the cellular function of Whi2. On the basis of previous studies, loss of WHI2 results in a decreased response to certain stresses; for example, nutrient limitation upon entering into the stationary phase (Radcliffe et al., 1997; Sudbery et al., 1980). This is in line with our observation that loss of Whi2 function leads to a decreased response after inhibition of TOR, which is known to partially mimic nutrient depletion. We found a direct dependency of mitophagy, and to a lesser extent of autophagy, on Whi2, demonstrating for the first time a role of Whi2 in both processes. Our data further indicate that Whi2 positively modulates autophagic flux. In a previous study, Whi2 was shown to interact physically with the phosphatases Psr1 and Psr2, and the transcription factor Msn2, which regulate the general stress response (Kaida et al., 2002). It is therefore possible that the induction of autophagy and mitophagy is directly or indirectly regulated by the general stress response. In line with this theory, it has been demonstrated that the Msn2-dependent stress response is activated upon inhibition of TOR and nutrient depletion (Beck and Hall, 1999; Gorner et al., 2002). Moreover, it is known that Msn2 is regulated by the Ras-PKA signaling pathway (Gorner et al., 1998). A recent study by Leadsham and colleagues (Leadsham et al., 2009) supports a role for Whi2 in the Ras-PKA signaling pathway, which in turn is known to influence autophagy (Budovskaya et al., 2004; Schmelzle et al., 2004). Interestingly, mitochondria in Δwhi2 cells were reported to exhibit a reduced membrane potential, to be fragmented and to produce high amounts of reactive oxygen species (ROS) during diauxic shift (Leadsham et al., 2009). The data presented here suggest that these effects could be due to a reduced elimination of damaged mitochondria through mitophagy. In general, our study indicates that TOR signaling, Ras-PKA signaling and the general stress response pathway are linked to mitophagy. It is noteworthy that loss of Whi2 function has a stronger impact on mitophagy than upon autophagy, suggesting that the induction of the two processes might be regulated independently. Future studies will be necessary to decipher the precise roles and the interplay of these signaling pathways on mitophagy and autophagy, and will help to improve our understanding of how mitochondrial quality control is regulated.

The yeast strains and plasmids used are listed in supplementary material Tables S1 and S2. Single deletion strains were obtained from the EUROSCARF clone collection (BY4742 Mat α; his3Δ1; leu2Δ0; lys2Δ0; ura3Δ0); gene deletion was confirmed by PCR. The mutant mgm1-5 strain was a gift from Jodi Nunnari (Molecular and Cellular Biology, University of California, Davis, CA). PHO8 was replaced with a HIS cassette amplified from pFA6aHISMX6 using the following primers: 5′-AGACCACAGGGTAGTCAACAGCAGCGGCAACAACGACAAAATGCGCGTCGTACGCTGCAGGTCGC-3′ and 5′-CGAGATTTCACTTCTTCACGCTATAGAATGCACCTAAAACTCGCCATCGATGAATTCGAGCTC-3′. Deletions of PHO8 were confirmed by western blotting and analysis of the activity of Pho8. The pYX242 cytPho8 and pVT100U cytPho8 plasmids were constructed by amplification of PHO8 (nucleotide position 181–1701), including an additional start codon, followed by ligation into the AvrII and MluI sites of pYX242 and the Pst1 and BamHI sites of pVT100U. For construction of pVT100U mtPho8, the PHO8 fragment was ligated into the KpnI and SacI sites of pVT100U containing Su9 (nucleotide position 1–207) from Neurospora crassa (Westermann and Neupert, 2000). The resulting Su9–pho8 fragment was amplified and cloned into the MluI and SacI sites of pYX242, resulting in pYX242 mtPho8. For endogenous FIS1 expression, the gene was amplified from genomic DNA, including 515 bp of the 5′–UTR (untranslated region) and 238 bp of the 3′-UTR of FIS1. YPLac33 WHI2 and pRS313 mtdSRed.T4 were gifts from Marie J. Hardwick (Molecular Microbiology and Immunology, Johns Hopkins University, Baltimore, MD) and Stefan Jakobs (Department of NanoBiophotonics, Max Planck Institute for Biophysical Chemistry, Göttingen, Germany), respectively. GFP–Atg8 including the endogenous promoter of Atg8, kindly provided by Hagai Abeliovich (Institute of Biochemistry, The Hebrew University of Jerusalem, Israel), was cloned into pRS316. The plasmid encoding Whi2–His₉ was constructed by amplifying WHI2 from genomic DNA (BY4742) and inserting the coding sequence of a His₉ tag, followed by ligation into the PvuII and BamHI sites of pVT100U. Culturing of yeast strains was performed using standard methods (Sherman, 2002) at 30°C. Strains containing plasmids were grown on selective liquid SG medium [2% (v/v) glycerol, 0.1% (w/v) glucose, 0.17% yeast nitrogen base (YNB) supplemented with 5 g/l (NH₄)₂SO₄ and amino acids] unless otherwise stated.

For monitoring autophagy or mitophagy, strains deleted for PHO8 were transformed with cytPho8- or mtPho8-expressing plasmids, respectively. The cultures were kept in the exponential growth phase for at least six generations and were then treated with 1 μM rapamycin (LC Laboratories) in DMSO or shifted to liquid SG-N medium [SG without (NH₄)₂SO₄]. For determination of the Pho8 enzyme activity, the yeast culture [with an attenuance at 600 nm of ~4] was harvested, washed with ice-cold 0.2 mM PMSF in H₂O and resuspended in 200 μl of lysis buffer (10 mM MgCl₂, 0.1 M KAc, 50 mM KCl, 10 nM ZnSO₄, 20 mM PIPES pH 7.0, 0.5% Triton X-100 and 1 mM PMSF). The cell suspension was mixed with an equal volume of 0.5-mm glass beads (Roth) and lysed for 10 seconds at 5000 rev./minute in a Precellys 24 cell homogenizer (Bertin Technologies). After mixing with 100 μl of additional lysis buffer, cell lysates were centrifuged at 13,000 g and the supernatants transferred into a 96-well plate. The protein concentration was determined with the BCA assay (Pierce) according to the manufacturer's instructions at three different dilutions. Pho8 activity was determined by incubation with reaction buffer (40 mM MgCl₂, 0.5% Triton X-100, 0.25 M Tris-HCl pH 8.5 and 4 mM p-nitrophenylphosphate) for 15 minutes at 37°C for at least two different concentrations. The reaction was stopped with 2 M glycine-NaOH pH 11, and absorption was determined at 405 nm using a Thermo Scientific plate reader. Specific activities are expressed as a percentage normalized to that in the control strain treated with rapamycin for 24 hours. Cell lysates were subjected to SDS-PAGE and western blotting.

For mitochondrial morphology analysis, strains were transformed with pHS12 mtDsRedT.3 and the cultures were kept in the exponential growth phase for at least six generations, before being imaged using a Zeiss Axioplan microscope. Pictures were taken with a CCD camera. To visualize mitophagy and autophagy by confocal fluorescence microscopy, strains expressing mtDsRed (pRS313 mtDsRed.T4) and GFP–Atg8 were kept in the exponential growth phase for at least six generations and were then treated with DMSO or 1 μM rapamycin for 7.5 hours. To analyze the mitochondrial membrane potential, cells were stained with 100 nM Rhodamine B hexyl ester perchlorate (Invitrogen) for 15 minutes. Yeast cells were harvested and fixed with 2% low-melting-point agarose (Sigma) before microscopy. Images were taken with a laser-scanning confocal microscope (ECLIPSE TE 2000-E, Nikon), with a 1.49 NA oil immersion lens (APO TIRF 60×, Nikon) and a microscope-integrated intermediate magnification of 1.5-fold. GFP and DsRed were successively excited with an argon laser and a helium-neon laser (GFP: 488 nm excitation, 515 nm detection, 30 nm bandwidth; DsRed: 543 nm excitation, 605 nm detection, 75 nm bandwidth). Background reduction was performed with appropriate saturation levels using software EZ-C1 v. 3.70 (Nikon) and ImageJ. Image stacks for analysis of mitochondrial morphology and quantification of mitophagy were acquired with steps of 0.25 to 0.4 μm per plane in the z-direction and a total thickness of 5–7 μm. The acquired stacks were rendered with EZ-C1 software. For quantification of mitophagy, the percentage of cells showing vacuolar accumulation of mtDsRed was calculated. At least three independent experiments were performed for each strain.

Cells were harvested and treated with zymolyase to generate spheroplasts, essentially as described previously (Daum et al., 1982). Spheroplasts were homogenized in a glass–teflon homogenizer. After this step, samples representing the total cell extracts were taken. Total cell extracts were centrifuged at 3000 g for 5 minutes to remove cell debris. For crude separation of mitochondria and cytosol, the cell extract was centrifuged at 11,000 g for 10 minutes. The mitochondrial pellet was washed twice in HS buffer (20 mM HEPES-KOH pH 7.4, 0.6 M sorbitol and 1 mM PMSF) and stored in liquid nitrogen. To obtain the cytosolic fraction, the supernatant was clarified by centrifugation at 90,000 g for 30 minutes. Mitochondria were purified on a 60%–55%–44% (w/v) sucrose step gradient at 126,444 g for 2.30 hours, harvested from the 44%–55% interphase, centrifuged at 11,000 g for 10 minutes and resuspended in HS buffer. To obtain the vacuolar fraction, the cell extract was separated on a four-step Ficoll gradient and vacuoles were harvested from the 0%–4% (w/v) Ficoll interphase, as described previously (Haas, 1995). Protein concentrations of each fraction were determined by the Bradford assay (Bio-Rad), according to the manufacturer's instructions, and were then mixed with SDS sample buffer and subjected to SDS-PAGE and western blotting using polyclonal antibodies against the indicated organellar marker proteins.

Mitochondria were diluted to 0.2 μg/μl in HS or HST buffer (HS plus 1% Triton X-100) and treated with 0.1 μg/μl proteinase K for 30 minutes on ice. Proteinase K was inhibited by adding PMSF to a final concentration of 1 mM. Mitochondria were centrifuged at 11,000 g, washed with HS buffer and precipitated with 12% trichloric acid. The samples were centrifuged, washed with acetone, and the dry pellets were resuspended in SDS loading buffer. Samples including an untreated control were subjected to SDS-PAGE and western blotting.

Samples with equal amounts of protein were mixed with SDS loading buffer (2% SDS, 10% glycerin, 60 mM Tris-HCl pH 6.8, 0.005% Bromphenol Blue and 5% 2-mercaptoethanol), separated by SDS-PAGE, transferred onto nitrocellulose membranes and detected by western blotting using rabbit polyclonal antibodies and goat anti-(rabbit Ig) secondary antibodies (Bio-Rad).

We are grateful to Stephane Duvezin-Caubet, Michael Zick, Kai Hell and Walter Neupert for inspiring scientific discussions, and Christiane Kotthoff and Christian Bach for excellent technical assistance. We thank J. Nunnari, H. Abeliovich, S. Jakobs, K. Okamoto, and J. M. Hardwick for kindly providing strains, plasmids and antibodies. This work was supported by the Deutsche Forschungsgemeinschaft DFG project no. RE-1575/1-1 (to A.R. and A.O.), Cluster of Excellence ‘Macromolecular Complexes’ at the Goethe University Frankfurt DFG project EXC 115 (to A.R.) and the SFB 594 (to A.R. and N.M.).