Mitochondrial morphology depends on balanced fusion and fission events. A central component of the mitochondrial fusion apparatus is the conserved GTPase Fzo1 in the outer membrane of mitochondria. Mdm30, an F-box protein required for mitochondrial fusion in vegetatively growing cells, affects the cellular Fzo1 concentration in an unknown manner. We demonstrate that mitochondrial fusion requires a tight control of Fzo1 levels, which is ensured by Fzo1 turnover. Mdm30 binds to Fzo1 and, dependent on its F-box, mediates proteolysis of Fzo1. Unexpectedly, degradation occurs along a novel proteolytic pathway not involving ubiquitylation, Skp1–Cdc53–F-box (SCF) E3 ubiquitin ligase complexes, or 26S proteasomes, indicating a novel function of an F-box protein. This contrasts to the ubiquitin- and proteasome-dependent turnover of Fzo1 in α-factor–arrested yeast cells. Our results therefore reveal not only a critical role of Fzo1 degradation for mitochondrial fusion in vegetatively growing cells but also the existence of two distinct proteolytic pathways for the turnover of mitochondrial outer membrane proteins.
Mitochondria are essential organelles whose structure and function adapt to different cellular conditions through continuous fusion and fission events (Okamoto and Shaw, 2005). Mitochondrial dynamics exert essential physiological and developmental roles, regulate apoptotic processes, and affect energy production within mitochondria (Chen and Chan, 2005). Two neuropathies, Charcot-Marie-Tooth type 2A and autosomal dominant optic atrophy, are caused by mutations in essential fusion components, namely, mitofusin 2 or OPA1 (Chen and Chan, 2005). Although several components involved in fusion have been identified, many of them in yeast, molecular mechanisms underlying mitochondrial fusion events are poorly understood. A central role has been assigned to yeast Fzo1 in the outer membrane of mitochondria (or to its mammalian homologues mitofusin 1 and 2) as part of a fusion complex that also contains Ugo1 and Mgm1 in yeast (Meeusen and Nunnari, 2005).
Less clear is the role of the F-box protein Mdm30, whose loss leads to the accumulation of aggregated and fragmented mitochondria (Fritz et al., 2003). Among the 21 annotated proteins of Saccharomyces cerevisiae with an F-box motif (Willems et al., 2004), some assemble in Skp1–Cdc53–F-box (SCF) E3 ubiquitin ligase complexes, which mediate proteasomal proteolysis of specific substrates (Petroski and Deshaies, 2005). Indeed, Mdm30 has been linked to the turnover of the transcription factor Gal4 in the nucleus (Muratani et al., 2005). Mdm30 was not only localized to the cytosolic fraction but also found in association with mitochondria (Fritz et al., 2003). Therefore, in analogy to the degradation of resident ER proteins, Mdm30 may affect mitochondrial dynamics by coupling the mitochondrial fusion machinery to the ubiquitin–proteasome system (UPS) in the cytosol. Consistently, accumulation of Fzo1 has been observed in cells lacking Mdm30 (Fritz et al., 2003). However, it remained unclear whether this indeed reflects Mdm30-dependent proteolysis of Fzo1 and whether the UPS is involved. Notably, 26S proteasomes have been linked to the degradation of Fzo1 in α-factor–arrested yeast cells (Neutzner and Youle, 2005). However, degradation does not depend on Mdm30 under these conditions (Neutzner and Youle, 2005). Moreover, the involvement of 26S proteasomes remained controversial, as only proteasome inhibitors have been used, which are known to be effective only in yeast cells with increased membrane permeability or lacking drug-efflux pumps (Lee and Goldberg, 1996).
We have analyzed the role of Mdm30 for the regulation of mitochondrial dynamics and demonstrate for the first time Mdm30-dependent proteolysis of Fzo1 in vegetatively growing yeast cells. Mdm30 is part of a novel proteolytic pathway that does not involve SCF complexes and 26S proteasomes and thus is strikingly different to the proteasomal degradation of Fzo1 in α-factor–arrested yeast cells.
Results And Discussion
Mdm30-dependent and -independent degradation of Fzo1
The accumulation of Fzo1 in Δmdm30 cells suggests that the F-box protein Mdm30 is involved in the degradation of Fzo1 (Fig. 1 A; Fritz et al., 2003). We therefore assessed the stability of Fzo1 in wild-type and Δmdm30 cells after inhibition of protein synthesis with cycloheximide. This analysis revealed that Fzo1 is constitutively degraded in wild-type cells, whereas it remained stable in the absence of Mdm30 (Fig. 1 A). Similar experiments were performed in yeast cells in the exponential or post–diauxic-shift phase, but a significant dependence of Fzo1 stability on the growth phase was not observed (Fig. 1 A). To distinguish between complete Fzo1 turnover and processing, we followed Fzo1 degradation using antibodies directed against either the NH2-terminal GTPase domain or a peptide located at the COOH-terminal segment of Fzo1. Moreover, the stability of an Fzo1 variant carrying an NH2-terminal HA tag was examined. As no proteolytic fragments of Fzo1 were detected, we conclude that Fzo1 is completely degraded in an Mdm30-dependent manner.
Mdm30 might affect the localization and/or complex assembly of Fzo1 and thereby affect Fzo1 stability and mitochondrial fusion. Therefore, mitochondrial and postmitochondrial supernatant fractions of wild-type and Δmdm30 cells were analyzed for the presence of Fzo1 (Fig. 1 B). Regardless of the presence of Mdm30, Fzo1 was recovered from the mitochondrial fraction, which also contained the mitochondrial outer membrane protein Tom40 but was devoid of the soluble cytosolic protein Tpi1 (Fig. 1 B). Fzo1 assembles into an 850-kD complex in wild-type cells (Rapaport et al., 1998) and was found to be present in a complex of the same size in Δmdm30 cells (Fig. 1 C). Thus, Mdm30 does not affect the assembly of Fzo1 in the outer membrane of mitochondria.
Fzo1 was recently reported to be degraded upon mating factor–induced cell cycle arrest (Neutzner and Youle, 2005; Fig. 1 D). Intriguingly, under these conditions, proteolysis occurs independent of Mdm30 (Neutzner and Youle, 2005; Fig. 1 D) or any other nonessential yeast gene encoding an F-box protein (unpublished data). We therefore conclude that Fzo1 is degraded along two physiologically and mechanistically distinct pathways, one prevalent in vegetatively growing cells and the other in α-factor–arrested cells.
Mdm30 binds to Fzo1 and mediates proteolysis in an F-box–dependent manner
To assess the requirement of the Mdm30 F-box motif for Fzo1 degradation, four conserved amino acid residues within the motif were replaced, namely, L19P, P20A, E22Q, and I23A (Fig. 2 A). In a second construct, the entire F-box was deleted (amino acids 1–58; Fig. 2 A, ΔF-box). Wild-type or mutant forms of Mdm30 were found in equal association with mitochondria upon subcellular fractionation (unpublished data). After expression in wild-type and Δmdm30 strains, the steady-state level of Fzo1 was monitored by immunoblotting. In wild-type cells, the steady-state concentration of Fzo1 was unaffected by the presence of the mutant Mdm30 forms (Fig. 2 B). In Δmdm30 cells, however, only wild-type Mdm30 but not Mdm30 variants with a defective F-box were able to restore degradation and low Fzo1 steady-state concentrations (Fig. 2 B). These results demonstrate that proteolysis of Fzo1 depends on the integrity of the Mdm30 F-box.
To examine whether Mdm30 interacts directly with Fzo1, we performed coimmunoprecipitation experiments using Fzo1-specific antibodies in cells expressing Mdm30 or F-box variants thereof. Mdm30 was precipitated with Fzo1-specific antibodies, demonstrating complex formation of both proteins in wild-type cells (Fig. 2 C). Thus, in agreement with studies on other F-box proteins (Skowyra et al., 1997), Fzo1 recognition by Mdm30 does not depend on its F-box (Fig. 2 C). It should be noted that after solubization of mitochondrial membranes with digitonin, Mdm30 or mutant variants did not coelute with Fzo1 upon sizing chromatography, indicating a rather dynamic interaction (unpublished data).
Mdm30-mediated Fzo1 turnover is independent of the SCF complex or of ubiquitylation
Mdm30 was found associated with both Skp1 and Cdc53 in high throughput protein–protein interaction studies (Uetz et al., 2000) and in vitro (Kus et al., 2004). Its requirement for Fzo1 proteolysis therefore suggests that degradation occurs through ubiquitylation by SCF complexes (Petroski and Deshaies, 2005). Protein shut-off experiments were performed in cells deficient for core components of SCF complexes (skp1-11, skp1-12, and cdc53-1) or for the E2 enzyme Cdc34 (cdc34-2). Fzo1 was degraded in these SCF-deficient cells with kinetics similar to those of wild-type cells (Fig. 3 A). This was in contrast to Grr1, a known substrate of the SCF complex (Galan and Peter, 1999), which stably accumulated in these mutants (Fig. 3 B). We also analyzed the stability of Fzo1 in cells that allow tetracycline-sensitive expression of the essential genes CDC53 and CDC34 (Mnaimneh et al., 2004) but did not observe a stabilization of Fzo1 upon repression of CDC53 or CDC34 (unpublished data). Thus, Mdm30-dependent proteolysis of Fzo1 does not require SCF ubiquitin ligase complexes. These findings are reminiscent of the F-box proteins Ctf13 and Rcy1, which do not assemble in SCF complexes (Russell et al., 1999; Galan et al., 2001), and substantiate that F-box proteins may exert cellular functions other than ubiquitin-dependent protein degradation. It should be emphasized, however, that Mdm30 could mediate the degradation of other substrate proteins by interaction with SCF complexes. This hypothesis is supported by the recently observed Mdm30-dependent ubiquitylation of Gal4 in the nucleus (Muratani et al., 2005).
To directly examine a possible ubiquitylation of Fzo1 by other E3 ubiquitin ligases, cellular ubiquitin pools were limited either by using a mutant E1 enzyme (uba1ts) or through the overexpression of mutant ubiquitin forms, unable to form ubiquitin chains (K29R, K48R, K63R, or RRR). Neither approach caused a stabilization of Fzo1 (Fig. 3 C). We conclude that Mdm30-dependent constitutive Fzo1 degradation is independent of SCF and ubiquitin.
We also examined the role of ubiquitylation for Fzo1 proteolysis upon mating factor–induced cell cycle arrest in uba1ts cells and found it to be impaired (Fig. 3 D). Thus, in striking contrast to its constitutive and Mdm30-dependent degradation, the proteolytic breakdown of Fzo1 in arrested cells is ubiquitin dependent (Fig. 3 D).
Proteasome independence and ATP dependence of Fzo1 degradation
We next examined the stability of Fzo1 in yeast cells with deficient proteasomal activity. Proteolysis of Fzo1 was not impaired in pre1-1 or cim5-1 cells carrying mutations in a proteolytic subunit of the 20S core particle or in an ATPase subunit of the 19S regulatory complex, respectively (Fig. 4 A). Similarly, deletion of the proteasomal assembly factor Ump1 did not interfere with the proteolytic breakdown of Fzo1 in vegetatively growing cells (Fig. 4 A). This is in contrast to Fzo1 proteolysis in cells arrested in the presence of mating factor (Fig. 4 B). In agreement with previous inhibitor studies (Neutzner and Youle, 2005), Fzo1 remained stable after treating pre1-1 or cim5-1 cells with α-factor (Fig. 4 B). We conclude that the constitutive Mdm30-dependent turnover of Fzo1 occurs independently of the 26S proteasome, whereas degradation of Fzo1 upon α-factor arrest involves ubiquitylation and 26S proteasomes.
To further characterize the Mdm30-mediated Fzo1 proteolysis, we examined a potential ATP dependence of this proteolytic pathway. Two chemical inhibitors of ATP production in mitochondria were used: NaN3, an inhibitor of the cytochrome c oxidase, and carbonylcyanide m-chlorophenylhydrazone (CCCP), an uncoupler dissipating the membrane potential across the mitochondrial inner membrane. Both treatments completely blocked degradation of Fzo1, which appears to be ATP dependent (Fig. 4 C).
We therefore analyzed the stability of Fzo1 in mitochondria lacking the ATP-dependent i-AAA protease subunit Yme1, which is active in the intermembrane space of mitochondria (Leonhard et al., 1996), but did not observe any effect on Fzo1 degradation (Fig. 4 A). Moreover, we examined the involvement of vacuolar degradation, which is responsible for the general removal of mitochondrial proteins under rapamycin-induced autophagic conditions (Kissova et al., 2004). However, Fzo1 was not stabilized in pep4 cells lacking active vacuolar peptidases (Fig. 4 A). We conclude that Fzo1 is constitutively degraded by a yet unknown proteolytic system whose activity is apparently ATP dependent.
Mdm30 controls mitochondrial fusion exclusively through Fzo1 proteolysis
Deletion of MDM30 impairs mitochondrial fusion and results in the formation of fragmented mitochondria accompanied by the accumulation of Fzo1 (Fig. 5 A; Fritz et al., 2003). To examine whether Fzo1 accumulation in Δmdm30 cells is a general consequence of impaired mitochondrial fusion, the steady-state concentrations of Fzo1 were determined in Δugo1, Δmgm1, and Δpcp1 mutant cells. Fzo1 did not accumulate in either of these mutant cells, demonstrating that there is no direct link between the cellular level of Fzo1 and ongoing mitochondrial fusion (Fig. 5 B). To further substantiate the specific effect of Mdm30 on Fzo1 stability and mitochondrial fusion, we analyzed the effect on the steady-state level of Fzo1 of the 131 nonessential deletion strains and in the 16 tet-off essential genes known to have fragmented mitochondria (Dimmer et al., 2002; Altmann and Westermann, 2005). However, apart from Δmdm30 cells, we did not observe altered Fzo1 levels (unpublished data). It therefore seems likely that Mdm30 affects mitochondrial dynamics by specifically regulating the stability of Fzo1.
To substantiate the specific effect of Mdm30 on Fzo1, we assessed mitochondrial morphology upon overexpression of Fzo1 in wild-type cells. Strikingly, mitochondria in cells overexpressing Fzo1 in the presence of an MDM30 wild-type allele resembled the fragmented mitochondria seen in Δmdm30 cells (Fig. 5 A). In agreement with this observation, proteolysis of Fzo1 was only observed when Fzo1 was expressed from a centromeric low-copy plasmid but not when expressed from a multicopy plasmid (Fig. 5 C). These findings were further supported by modulating Fzo1 expression in wild-type and Δmdm30 cells using the inducible GAL1 promoter. Growing yeast cells in the presence of glucose, raffinose, or galactose results in repression, nonrepression, or induction of Fzo1 synthesis. Importantly, tubular mitochondria were observed in Δmdm30 cells grown under nonrepressed conditions in the presence of raffinose, i.e., at intermediate expression levels of Fzo1 (Fig. 5 D and Table I). This confirms that mitochondrial morphology can be modulated by changing Fzo1 expression levels. Thus, mitochondrial fusion requires a tight control of the steady-state concentration of Fzo1. We conclude that lowering Fzo1 protein levels allows bypassing of Mdm30. Therefore, Mdm30 promotes mitochondrial fusion exclusively by controlling the steady-state level of Fzo1.
Overexpressing Fzo1 in wild-type cells prevents its degradation, suggesting that Mdm30 could be the limiting factor. Therefore, Mdm30 was overexpressed in cells containing a galactose-inducible FZO1 allele. Surprisingly, we found that the simultaneous overexpression of Fzo1 and Mdm30 is deleterious for cellular growth and leads to abnormal cellular and mitochondrial morphologies (Fig. 5 E). It is therefore conceivable that the Fzo1–Mdm30 complex sequesters an as-yet-unknown factor that is essential for growth.
Our experiments demonstrate a crucial role of Fzo1 proteolysis for the maintenance of mitochondrial morphology. The stability of Fzo1 is controlled by a novel, apparently ATP-dependent proteolytic system. This system involves the F-box protein Mdm30 but not ubiquitylation and SCF complexes, opening new perspectives for cellular activities of F-box proteins. Our results also demonstrate that proteins in the outer membrane of mitochondria can be degraded in at least two ways, only one of which is dependent on the UPS. At the same time, they suggest a mechanism that regulates organellar morphology by the use of two different proteolytic systems during different physiological states of a eukaryotic cell.
Materials And Methods
Yeast strains and growth media
Yeast strains were grown according to standard procedures on complete or synthetic media supplemented with 2% (wt/vol) glucose or, when indicated, 2% (wt/vol) galactose or raffinose and 0.02% (wt/vol) glucose. 100 μg/ml cycloheximide, 10 μM α-factor (or 0.1 μM in the case of Δbar1 cells), 2 mM NaN3, or 1 mM CCCP was added when indicated. The cell cycle arrest in the presence of α-factor was verified microscopically.
Yeast strains used were derivatives of W303 and BY4741. The strains Δmdm30, Δfzo1, Δugo1, Δpcp1, or Δmgm1 and the isogenic wild-type strain BY4741 were obtained from Euroscarf. For expression of Fzo1 in wild-type and Δmdm30 cells under the control of the GAL1 promoter, previously described strains were used (Fritz et al., 2003). MDM30 and PEP4 were deleted by PCR-based homologous recombination in W303. To increase the efficiency of the α-factor arrest, BAR1 was deleted by PCR-based homologous recombination in wild-type, cim5-1, and pre1-1 cells.
FZO1, MDM30, and mdm30 mutant variants (F-box*, with the L19P, P20A, E22Q, and I23A mutations; ΔF-box, without the residues 1–58) encoding an NH2-terminal FLAG tag were cloned into the multicopy vector pJDCEX2 (2μ; LEU2 and CUP1 promoter). Fzo1 harboring an NH2-terminal HA tag was expressed under the control of the endogenous promoter using the centromeric plasmid pRS315 or the multicopy plasmid YEplac181. For visualizing mitochondria, the plasmid pVT100U-mtGFP (mitochondria-targeted GFP) was used.
Protein synthesis shut off and α-factor cell cycle arrest
To monitor constitutive proteolysis of Fzo1, cycloheximide was added to logarithmically growing yeast cultures at a final concentration of 100 μg/ml. Thermosensitive strains were incubated for 1 h at 37°C before adding cycloheximide. To examine Fzo1 degradation upon α-factor–induced cell cycle arrest, early log phase cells grown on YPD were treated with 20 mM sodium citrate and 10 μM α-factor (or 0.1 μM in Δbar1 strains). At the indicated time points, cells corresponding to 3 OD units were collected and lysed at alkaline pH (Tatsuta and Langer, 2006), and protein extracts were analyzed by SDS-PAGE and immunoblotting. Mean values of at least three independent experiments are shown.
Coimmunoprecipitation of Fzo1 and Mdm30
Wild-type cells expressing Mdm30 and variants thereof from a copper-responsive promoter were grown on synthetic media without an additional copper supplement. Crude mitochondria obtained from 60 OD units of these cells were solubilized for 15 min in 1% (wt/vol) digitonin in immunoprecipitation buffer (150 mM KAc, 4 mM MgAc, 30 mM Tris/HCl, pH 7.4, 1 mM PMSF, and 1 mM ATP). After a clarifying spin for 15 min at 45,000 g, supernatants were incubated for 1 h at 4°C under constant mixing with protein A–Sepharose and polyclonal antibodies against the GTPase domain of Fzo1. After two washing steps in immunoprecipitation buffer and one washing step in 10 mM Tris/HCl, pH 7.4, bound material was eluted with SDS sample buffer and used for Western blot analysis.
Antisera against the COOH terminus of Fzo1 (Rapaport et al., 1998), the GTPase domain of Fzo1 (provided by J. Nunnari, University of California, Davis, Davis, CA), the HA epitope (3F10; Roche), the FLAG epitope (M2-F3165; Sigma-Aldrich), or the myc epitope (9B11; Cell Signaling Technology) were used.
Overnight cultures were analyzed by epifluorescence microscopy on a microscope (Axioplan; Carl Zeiss MicroImaging, Inc.) using a 100× oil-immersion objective. GFP fluorescence was visualized in living cells. Images were acquired with a camera (Quantix; Photometrix) and processed with MetaMorph 4.5 software (Universal Imaging Corp.).
We are grateful to J. Dohmen for the ump1ts and uba1ts strains, the plasmid pJDCEX2, the Tpi1 antibodies, and his helpful advice; R. Deshaies for the skp1-11 and skp1-12 strains; J.M. Galan for yeast strains expressing Grr1-Myc; C. Mann for the cim5-1 strain; K. Nasmyth for the cdc34-2 strain; J. Nunnari for the GTPase Fzo1 antibody; M. Tyers for cdc53-1; and D. Wolf for the strains pre1-1 and cim3-1. The expert technical assistance of D. Tils and G. Zimmer and helpful advice of M. Collart and B. Daignan-Fornier are gratefully acknowledged.
This work was supported by a European Molecular Biology Organization fellowship to M. Escobar-Henriques (ALTF 586-2002) and funds of the Deutsche Forschungsgemeinschaft to T. Langer (SFB635, P18) and B. Westermann (WE 2174/2-4).
Abbreviations used in this paper: CCCP, carbonylcyanide m-chlorophenylhydrazone; mtGFP, mitochondria-targeted GFP; SCF, Skp1–Cdc53–F-box; UPS, ubiquitin–proteasome system.