The MDM31 and MDM32 genes are required for normal distribution and morphology of mitochondria in the yeast Saccharomyces cerevisiae. They encode two related proteins located in distinct protein complexes in the mitochondrial inner membrane. Cells lacking Mdm31 and Mdm32 harbor giant spherical mitochondria with highly aberrant internal structure. Mitochondrial DNA (mtDNA) is instable in the mutants, mtDNA nucleoids are disorganized, and their association with Mmm1-containing complexes in the outer membrane is abolished. Mutant mitochondria are largely immotile, resulting in a mitochondrial inheritance defect. Deletion of either one of the MDM31 and MDM32 genes is synthetically lethal with deletion of either one of the MMM1, MMM2, MDM10, and MDM12 genes, which encode outer membrane proteins involved in mitochondrial morphogenesis and mtDNA inheritance. We propose that Mdm31 and Mdm32 cooperate with Mmm1, Mmm2, Mdm10, and Mdm12 in maintenance of mitochondrial morphology and mtDNA.
Mitochondria are ubiquitous and essential organelles of eukaryotic cells. Because they cannot be generated de novo, they have to be inherited during cell division (Warren and Wickner, 1996). Inheritance of mitochondria involves active transport of the organelles along cytoskeletal tracks, concomitant with frequent membrane division and fusion events (Yaffe, 1999). The mitochondrial genome, which encodes a small subset of mitochondrial proteins, has to be partitioned to the daughter cell in an active and ordered manner (Azpiroz and Butow, 1993; Okamoto et al., 1998; Berger and Yaffe, 2000; Garrido et al., 2003; Meeusen and Nunnari, 2003). As mitochondria are double membrane-bounded organelles, transport processes occurring at the mitochondrial surface and partitioning events of matrix components must be coordinated across two membranes. For example, mitochondrial DNA (mtDNA) is located in protein-containing complexes, termed nucleoids, in the matrix. It has been suggested that inheritance of these nucleoids requires a segregation machinery in the cytosol (Berger and Yaffe, 2000; Aiken Hobbs et al., 2001; Boldogh et al., 2003; Meeusen and Nunnari, 2003). Moreover, it is conceivable that maintenance of the structure of the inner membrane depends on an intimate coordination with the behavior of the outer membrane, involving interactions of proteins in both membranes. However, the molecular processes coordinating the behavior of the double membranes during mitochondrial inheritance are not well understood.
Mitochondria form highly dynamic interconnected networks in many cell types from yeast to man (Bereiter-Hahn, 1990; Nunnari et al., 1997; Jakobs et al., 2003). In recent years a growing number of proteins controlling mitochondrial motility and behavior have been identified, mainly in the baker's yeast Saccharomyces cerevisiae (Hermann and Shaw, 1998; Jensen et al., 2000; Scott et al., 2003). In yeast, establishment, maintenance, and motility of the branched mitochondrial network depend on the actin cytoskeleton (Boldogh et al., 2001). Some mitochondrial outer membrane proteins have been suggested to play a role in microfilament-dependent inheritance of mitochondria and mtDNA. Yeast mutants lacking Mdm10, Mdm12, or Mmm1 have giant spherical mitochondria (Burgess et al., 1994; Sogo and Yaffe, 1994; Berger et al., 1997), which show severely compromised intracellular motility (Boldogh et al., 1998, 2003). As these proteins are often localized next to mtDNA nucleoids, and as mtDNA nucleoids are disorganized in mutants, it has been proposed that Mdm10, Mdm12, and Mmm1 are parts of a cytoskeleton-dependent double membrane-spanning transport machinery required for inheritance of mitochondria and mtDNA (Aiken Hobbs et al., 2001; Boldogh et al., 2003; Meeusen and Nunnari, 2003). Mmm2 (alternative name Mdm34) has been identified as another protein that participates in this process (Dimmer et al., 2002; Youngman et al., 2004). Mmm2 is located in a separate complex in the outer membrane, and mutants lacking Mmm2 harbor aberrant mitochondria and disorganized mtDNA nucleoids (Youngman et al., 2004).
It can be predicted that there must be partners in the inner membrane that physically and/or functionally interact with the outer membrane proteins Mmm1, Mmm2, Mdm10, and Mdm12 in mediating the inheritance of mitochondrial membranes and mtDNA nucleoids. It has been suggested that Mmm1 in yeast spans both mitochondrial membranes and exposes a small NH2-terminal segment to the matrix (Kondo-Okamoto et al., 2003). However, the NH2-terminal extension is absent in other homologous proteins, such as MMM1 in Neurospora crassa (Prokisch et al., 2000), and it is not required for maintenance of normal tubular networks and mtDNA nucleoids in yeast (Kondo-Okamoto et al., 2003). Thus, there must be other, yet unknown, inner membrane proteins participating in these processes. By screening a comprehensive yeast gene deletion library, we recently isolated several novel genes important for mitochondrial distribution and morphology, MDM (Dimmer et al., 2002). Here, we show that MDM31 and MDM32 encode novel components of the mitochondrial inner membrane. We propose that Mdm31 and Mdm32 functionally cooperate with the outer membrane machinery mediating maintenance of mitochondrial morphology and inheritance of mtDNA.
MDM31 and MDM32 encode two members of a novel protein family
The MDM31 (systematic name YHR194W) and MDM32 (systematic name YOR147W) genes encode two related proteins of 66.7 and 75.6 kD, respectively. Both proteins share 16.4% amino acid identity with each other. Related genes encoding homologous proteins can be found in the genomes of Candida albicans, Schizosaccharomyces pombe, N. crassa, and other ascomycetes fungi (Fig. 1 A; for an alignment see online supplemental material). Remarkably, these more distantly related fungi have only one homologous gene, which is more closely related to MDM31 (between 27.8% amino acid identity for S. pombe and 52.3% for C. albicans). Other species of the family Saccharomycetaceae have two related isoforms (Cliften et al., 2003; Kellis et al., 2003). Thus, the second isoform apparently has arisen by a relatively recent gene duplication event.
All members of the Mdm31 protein family have a similar domain structure (Fig. 1 B). The NH2 termini have the characteristics of typical mitochondrial presequences. They are rich in positively charged residues, lack acidic charges, and have a high content of hydroxylated residues. Computational prediction of mitochondrial presequences by the MitoProt II program (Claros and Vincens, 1996) gives very high probabilities for mitochondrial targeting (between 0.9518 for S. pombe Mdm31 and 0.9989 for Mdm32). Hydropathy analysis (Hofmann and Stoffel, 1993) predicts two transmembrane segments, one close to the NH2 terminus of the matured protein and another one at the very COOH terminus. The predicted domain structure is very similar for all five family members with the exception of the S. pombe protein that lacks a hydrophobic segment at its COOH terminus.
Mdm31 and Mdm32 are located in the mitochondrial inner membrane
To determine the intracellular location of Mdm31 and Mdm32, wild-type yeast cells were fractionated into mitochondria, microsomes, and cytosol. Cell fractions were analyzed by Western blotting using specific antisera against Mdm31 and Mdm32. Both proteins cofractionated with the mitochondrial ADP/ATP carrier (AAC; Fig. 2 A), demonstrating a mitochondrial location. To determine the intramitochondrial location, isolated mitochondria were subfractionated. When intact mitochondria were treated with proteinase K (PK), both Mdm31 and Mdm32 were protected against proteolytic degradation (Fig. 2 B, lane 2), indicating that they are located in the interior of the organelle. When the outer membrane was selectively opened by hypotonic swelling, Mdm31 and Mdm32 were accessible to PK (Fig. 2 B, lane 3), indicating that a major domain is exposed to the intermembrane space. When the mitochondrial membranes were lysed with detergent, the proteins were completely degraded by PK (Fig. 2 B, lane 4). Upon carbonate extraction, all of Mdm31 and about half of Mdm32 cofractionated with mitochondrial membranes (Fig. 2 C), demonstrating that they are integral membrane proteins. It should be noted that partial extraction by carbonate has been observed also for other mitochondrial membrane proteins (Mokranjac et al., 2003). We conclude that Mdm31 and Mdm32 are located in the mitochondrial inner membrane. Protected fragments in protease-treated mitoplasts could never be observed in immunoblots of endogenous protein or after in vitro import of radiolabeled protein (unpublished data). We suggest that major parts of Mdm31 and Mdm32 are located in the intermembrane space, and the short NH2 termini are exposed to the matrix.
Cells lacking Mdm31 and Mdm32 show severe defects in mitochondrial distribution and morphology
To examine the role of Mdm31 and Mdm32 in mitochondrial distribution and morphology, Δmdm31 and Δmdm32 deletion mutants and a Δmdm31/Δmdm32 double mutant were constructed. All mutants were viable, both on fermentable and nonfermentable carbon sources (see section Mdm31 and Mdm32 are required for organization of mtDNA nucleoids) and showed identical phenotypes. Examination of mutant strains expressing mitochondria-targeted GFP (mtGFP) by confocal microscopy revealed highly aberrant mitochondrial structures (Fig. 3 A). Most of the cells harbored one or few giant spherical mitochondria (Fig. 3 A, b, c, e, and f). Often, the organelles contained one or few small hollow inclusions (Fig. 3 A, e and f). Some cells contained several, relatively small mitochondria (Fig. 3 A, d). Branched tubular mitochondrial networks resembling the wild type (Fig. 3 A, a) were not observed in the mutants. A quantification of mitochondrial phenotypes is given in Table I.
Many mutant cells generated buds that were devoid of mitochondria. Occasionally, a giant mitochondrion was positioned at the bud neck, which it could not pass (Fig. 3 A, b). To quantify these effects, logarithmically growing cultures of mtGFP-expressing cells were analyzed by fluorescence microscopy. Cells were counted that showed mitochondria both in the mother and the daughter cell, mitochondria stuck at the bud neck, or mitochondria-free buds. For comparison, we included in this analysis a Δmmm1 strain, which has giant spherical mitochondria (Burgess et al., 1994) similar to the Δmdm31 and Δmdm32 mutants. Only 50–60% of the buds contained mitochondria in Δmdm31, Δmdm32, Δmdm31/Δmdm32, and Δmmm1 cells. In most of the remaining cells, mitochondria were somewhere deposited in the mother cell. A few percent of the cells showed mitochondria positioned directly at the bud neck (Table II).
To exclude the possibility that the observed phenotypes were caused by defects in the organization of the actin cytoskeleton, or that deletion of the MDM31 and MDM32 genes has pleiotropic effects on the structure of several cell organelles, we stained filamentous actin, the ER, and vacuoles. All these structures appeared normal in the mutants (Fig. 3 B). We conclude that Mdm31 and Mdm32 play an important and specific role in controlling mitochondrial distribution and morphology.
It has been speculated that Mdm31 and Mdm32 might be novel components of the mitochondrial membrane fusion machinery (Mozdy and Shaw, 2003). To test this possibility, we monitored fusion of mitochondria in vivo by mating of mutant cells preloaded with different fluorescent mitochondrial markers (Nunnari et al., 1997). Mixing of the markers could be observed in zygotes lacking Mdm31 or Mdm32, as well as in zygotes lacking both proteins (Fig. 3 C). This demonstrates that Mdm31 and Mdm32 do not play an essential role in mitochondrial fusion. Interestingly, several zygotes were found in which the fluorescently labeled mitochondria of both parental cells remained separate (Fig. 3 C). However, these nonfused mitochondria were never seen close together. This observation suggests that in the latter cases fusion did not occur because the mitochondria did not approach each other. Heterologous crosses of Δmdm31 and Δmdm32 single deletion mutants showed complementation in zygotes, i.e., mitochondria looked like wild type and fused in an efficient manner (Fig. 3 C). We suggest that the function of Mdm31 and Mdm32 is required for efficient fusion in cells, even though these proteins are not integral components of the mitochondrial fusion machinery.
Mitochondria lacking Mdm31 and Mdm32 show dramatically altered internal structure
As Mdm31 and Mdm32 are inner membrane proteins, we considered it likely that also the internal structure of mutant mitochondria is altered. To examine this possibility, Δmdm31, Δmdm32, and Δmdm31/Δmdm32 cells were examined by electron microscopy and compared with the wild type. Electron micrographs of wild-type cells grown on glucose-containing medium showed characteristic cross sections of tubular mitochondria containing cristae as invaginations of the inner membrane (Fig. 4 A). In contrast, the ultrastructure of Δmdm31 (Fig. 4 C), Δmdm32 (Fig. 4, B and D), and Δmdm31/Δmdm32 (Fig. 4, E–H) mutant cells was dramatically altered. The organelles were generally very large. These giant organelles were largely devoid of cristae. Only in some organelles a few small cristae were found (Fig. 4 E, arrows). Frequently, circular-shaped double membrane structures were seen inside the organelles (Fig. 4, B–G). These structures were of varying sizes, but the spacing between the membranes was remarkably constant and was identical to the size of the intermembrane space. This finding suggests that the double membranes were derived from the mitochondrial outer and inner membranes, and that the compartment surrounded by the circular membranes topologically corresponds to the exterior of the organelle. Consistently, these structures appeared as holes in sections obtained by confocal microscopy (Fig. 3 A). Occasionally, an internal membrane was connected with the inner membrane surrounding the organelle (Fig. 4, D, G, and H). In these cases, the intermembrane space was continuous with the space between the membranes of the circular inclusion (Fig. 4 G, arrows; enlarged image in Fig. 4 H). We conclude that deletion of the MDM31 and MDM32 genes has dramatic consequences on the organization of the mitochondrial membranes and the global structure of the organelle.
Mdm31 and Mdm32 are required for normal mitochondrial motility
Two lines of evidence suggested that motility of mitochondria is compromised in cells lacking Mdm31 and Mdm32. First, mutant cells often carried buds devoid of mitochondria (Fig. 3 A and Table II), and second, in many mutant zygotes, mitochondria did not fuse because they did not approach each other (Fig. 3 C). To examine mitochondrial movement directly, the behavior of mitochondria was followed over time by confocal time-lapse microscopy of mtGFP-expressing cells. Cells grown logarithmically in glucose-containing medium were transferred to a microscope chamber that was continuously flushed with fresh medium. Wild-type mitochondria are highly dynamic under these experimental conditions (Jakobs et al., 2003). Their shapes and positions were observed to change completely within a few minutes (Fig. 5 A). In contrast, mutant mitochondria of Δmdm31, Δmdm32, and Δmdm31/Δmdm32 strains were almost immotile. They hardly changed their positions within time periods of 15 to 30 min (Fig. 5, B–D). However, sometimes subtle shape changes were observed in the mutant mitochondria. Generally, these shape changes started with the occurrence of small protrusions (Fig. 5, B–D, arrows), probably by a force pulling on the organelle. In most cases, these protrusions were retracted soon afterwards (Fig. 5, C and D). Together, these results demonstrate that mitochondrial motility is severely compromised in Δmdm31, Δmdm32, and Δmdm31/Δmdm32 mutants.
Sometimes, the initial deformation resulted in a successful translocation event of the entire organelle (Fig. 5 B). To test whether or not the residual translocation activity might be mediated by the actin-dependent transport machinery of mitochondria, we examined the binding of mitochondria to actin filaments in vitro. Isolated mitochondria of wild type, Δmdm31, and Δmdm32 strains were incubated with filamentous actin in the presence or absence of ATP. Then, mitochondria were sedimented by centrifugation through a sucrose cushion, and bound actin was detected by immunoblotting. Mitochondria lacking Mdm31 or Mdm32 were able to interact with actin filaments in an ATP-dependent manner, similar to wild-type mitochondria (Fig. 5 E). This suggests that mitochondrial motility defects in the mutants are due to structural aberrations of the organelle rather than defects of the machinery mediating interactions with the cytoskeleton.
mdm31 and Δ mdm32 mutations are epistatic to Δ fzo1, Δ dnm1, and Δ mdm33 mutations
To investigate functional relationships of MDM31 and MDM32 with other genes encoding components important for mitochondrial structure and behavior, we constructed a series of double mutants. Δmdm31 and Δmdm32 strains were crossed with the following deletion strains: Δfzo1, a mutant defective in mitochondrial fusion (Hermann et al., 1998; Rapaport et al., 1998); Δdnm1, a mutant defective in outer membrane division (Otsuga et al., 1998); and Δmdm33, a mutant defective in inner membrane division (Messerschmitt et al., 2003). Resulting diploids were subjected to tetrad dissection, and mitochondrial morphology of haploid progeny was analyzed by fluorescence microscopy. In all cases, the parental mutants had clearly distinguishable phenotypes. Double mutants obtained from all crosses displayed mitochondria indistinguishable from their Δmdm31 and Δmdm32 parents (Table I). This finding indicates that the Δmdm31 and Δmdm32 mutations are epistatic to Δfzo1, Δdnm1, and Δmdm33 mutations; i.e., in the absence of Mdm31 or Mdm32, mitochondrial morphology does not depend on Fzo1, Dnm1, or Mdm33. We propose that the function of Mdm31 and Mdm32 is superior to mitochondrial fusion and division.
mdm31 and Δ mdm32 mutations are synthetically lethal with Δ mmm1, Δ mmm2, Δ mdm10, and Δ mdm12 mutations
We asked if MDM31 and MDM32 have overlapping functions with MMM1, MMM2, MDM10, and MDM12, because mutants lacking these genes have very similar phenotypes (Burgess et al., 1994; Sogo and Yaffe, 1994; Berger et al., 1997; Boldogh et al., 1998, 2003; Aiken Hobbs et al., 2001; Youngman et al., 2004). Δmdm31 and Δmdm32 mutants were crossed with Δmmm1, Δmmm2, Δmdm10, and Δmdm12 strains. Upon tetrad dissection, we observed in all crosses a 1:1:4 segregation into parental ditype tetrads, nonparental ditype tetrads, and tetratype tetrads (Table III). Spores containing both deleted alleles were not viable; i.e., Δmmm1, Δmmm2, Δmdm10, and Δmdm12 mutations are synthetically lethal with Δmdm31 and Δmdm32 mutations. Synthetic lethality of two mutations in different genes often indicates that the gene products are required for the same cellular processes (Guarente, 1993; Hartman et al., 2001). The synthetic lethal phenotype was confirmed in a plasmid shuffling experiment using the Δmdm32/Δmmm1 double mutant (unpublished data). These results show that the function of Mdm31 and Mdm32 is essential for cell viability in the absence of Mmm1, Mmm2, Mdm10, and Mdm12.
Mdm31 and Mdm32 are required for organization of mtDNA nucleoids and localization of mtDNA adjacent to Mmm1 foci
Using strains obtained from the yeast gene deletion collection, we reported previously that Δmdm31 and Δmdm32 mutants are respiratory-deficient (Dimmer et al., 2002). Here, we observed that it is possible to grow newly made Δmdm31, Δmdm32, and Δmdm31/Δmdm32 mutants on nonfermentable carbon sources. Serial dilutions of wild-type and mutant cultures were spotted onto plates containing either glucose or glycerol as carbon source and incubated at 30 or 37°C. Mutant strains showed a moderate growth defect under most conditions, and the Δmdm31/Δmdm32 double mutant showed a more severe growth defect on nonfermentable carbon sources at elevated temperature (Fig. 6 A). As inheritance of mtDNA depends on the integrity of the mitochondrial compartment (Berger and Yaffe, 2000), we reasoned that our initial observation of a petite phenotype in the mutants may be due to the gradual loss of the mitochondrial genome over several generations. To test this idea, Δmdm31, Δmdm32, and Δmdm31/Δmdm32 strains were grown in liquid medium containing glucose to allow for loss of mtDNA. Cultures were maintained in the logarithmic growth phase at 30°C. At different time points, aliquots were taken and plated at an appropriate dilution onto glucose-containing medium. Subsequently, colonies were replica-plated onto glycerol-containing medium, and the percentage of colonies able to grow was determined as a measure of the fraction of respiratory-competent cells in the initial culture. After 3 d in glucose-containing medium, only ∼50% of the mutant cells were respiratory-competent (very similar numbers were obtained for all three mutant strains). We conclude that Mdm31 and Mdm32 are required for normal inheritance of mtDNA.
We asked whether or not mtDNA is normally organized and distributed in cells lacking Mdm31 and Mdm32. mtDNA was stained in living cells by expressing an Abf2-GFP fusion protein, which binds to mtDNA nucleoids (Okamoto et al., 1998), or with a DNA-specific dye, DAPI. Wild-type cells contained normal mtDNA nucleoids, seen as 10–20 small fluorescent dots (Fig. 6 B). In contrast, Δmdm31 and Δmdm32 mutants contained only one or two misshapen mtDNA-containing structures per cell. The staining pattern of these structures was diffuse, and they were generally rather large (Fig. 6 B). Very similar results were obtained after DAPI staining of methanol-fixed cells (unpublished data). We conclude that Mdm31 and Mdm32 are required for establishment and/or maintenance of mtDNA nucleoid structure.
It has been reported that Mmm1 is located in distinct foci on the mitochondrial outer membrane. These foci are often found next to mtDNA nucleoids (Aiken Hobbs et al., 2001; Boldogh et al., 2003; Meeusen and Nunnari, 2003), and their formation depends on the presence of the outer membrane protein Mmm2 (Youngman et al., 2004). It is thought that Mmm1-containing foci (in cooperation with yet unknown inner membrane proteins) contribute to the structural organization and inheritance of mtDNA nucleoids (Aiken Hobbs et al., 2001; Boldogh et al., 2003; Meeusen and Nunnari, 2003; Youngman et al., 2004). The aberrant mtDNA nucleoids seen in Δmdm31 and Δmdm32 mutants and the genetic interactions with Δmmm1 prompted us to investigate whether the formation of Mmm1 foci and/or their localization next to mtDNA depends on the presence of Mdm31 and Mdm32. First, we tested whether or not the steady-state level of Mmm1 is altered in mitochondria of cells lacking Mdm31 and Mdm32. Immunoblot analysis showed that Mmm1 was present in similar amounts in mitochondria isolated from wild-type, Δmdm31, Δmdm32, and Δmdm31/Δmdm32 cells (Fig. 6 C). The same result was obtained for Mmm2 and Mdm10 (Fig. 6 C). Mdm12 is required for localization of Mmm1 to mitochondria (Boldogh et al., 2003). As the level of Mmm1 was not changed in Δmdm31, Δmdm32, and Δmdm31/Δmdm32 mutant mitochondria, we conclude that also Mdm12 must be present in sufficient amounts. Thus, synthesis, mitochondrial targeting, and stability of Mmm1, Mmm2, Mdm10, and Mdm12 are not compromised in Δmdm31, Δmdm32, and Δmdm31/Δmdm32 mutants.
Next, we analyzed the intracellular distribution of Mmm1 and mtDNA by fluorescence microscopy. Consistent with previous reports (Aiken Hobbs et al., 2001; Meeusen and Nunnari, 2003; Youngman et al., 2004), wild-type cells expressing an Mmm1-DsRed fusion protein showed a punctate staining pattern. The majority of Mmm1 foci was located next to mtDNA nucleoids stained by Abf2-GFP or DAPI (Fig. 6 B). Mmm1 punctae were seen also in Δmdm31 and Δmdm32 cells, demonstrating that Mdm31 and Mdm32 are not required for Mmm1 foci formation (Fig. 6 B). However, mtDNA nucleoids were disorganized and Mmm1 foci were only rarely seen in the vicinity of mtDNA. Most Mmm1-DsRed–expressing mutant cells showed Mmm1 foci distantly located from diffusely organized mtDNA (Fig. 6 B). We conclude that Mdm31 and Mdm32 are required for localization of Mmm1 foci next to mtDNA.
Mdm31 and Mdm32 are present in distinct complexes in the mitochondrial membranes
We determined whether or not Mdm31 and Mdm32 are subunits of larger protein complexes. Isolated wild-type mitochondria were solubilized with a mild detergent, digitonin. Protein complexes were separated by gel filtration and analyzed by Western blotting. Interestingly, Mdm31 and Mdm32 reside in separate complexes. Mdm31 was eluted at ∼600 kD, which was clearly larger than the size of the Mdm32 complex at ∼175 kD (Fig. 7 A). The size of the Mdm31 complex was not changed in the absence of Mdm32, and vice versa (Fig. 7 B).
We considered the possibility that Mmm1 forms a double membrane-spanning protein complex together with Mdm31 or Mdm32. To test this possibility, we asked if Mmm1 cofractionates with Mdm31 or Mdm32 in gel filtration. Consistent with a previous study (Youngman et al., 2004), Mmm1 was found in a complex slightly larger than 600 kD, the size of which was similar to the Mdm31 complex (Fig. 7 A). If Mdm31 and Mmm1 were subunits of the same complex, it could be expected that the size of this complex would change in the absence of one of the subunits. Therefore, we performed gel filtration experiments with mitochondria isolated from Δmdm31, Δmdm32, Δmdm31/Δmdm32, and Δmmm1 mutant cells. Neither deletion had a significant effect on the size of the other complexes (Fig. 7 B). We also noticed that the peak fractions of Mdm31 and Mmm1 were sometimes shifted by one fraction, with the Mdm31 complex being slightly larger (Fig. 7 B). Furthermore, we could not detect a direct interaction of Mdm31 or Mdm32 with Mmm1 in coimmunoprecipitation and cross-linking experiments. Thus, Mdm31, Mdm32, and Mmm1 are subunits of separate complexes in the mitochondrial membranes.
The structural and functional similarities of Mdm31 and Mdm32 point to a close collaboration of these proteins. Even though they assemble into separate complexes, they might still interact in a weak or transient manner. To test this possibility, we imported radiolabeled Mdm32 into mitochondria and performed coimmunoprecipitation experiments with specific antibodies directed against endogenous Mdm31. Upon translation in vitro of Mdm32 in the presence of [35S]methionine, SDS-PAGE, and autoradiography, a single band corresponding to the size of the precursor protein was observed (Fig. 7 C, lane 1). Upon incubation with isolated mitochondria, a slightly smaller form was generated by processing of the presequence by the matrix processing peptidase (Fig. 7 C, lanes 2 and 6). After import into wild-type mitochondria, a fraction of matured Mdm32 could be coimmunoprecipitated with Mdm31 antibodies (Fig. 7 C, lane 3). No precursor protein was found associated with Mdm31, demonstrating that the reaction was specific for the imported protein. Furthermore, no signal was obtained with mitochondria lacking Mdm31 (Fig. 7 C, lane 5) when preimmune serum was used (Fig. 7 C, lane 7) or when nonrelated inner membrane proteins were imported (not depicted). Thus, Mdm31 and Mdm32 interact with each other in a specific manner. The observation that only a small fraction of imported Mdm32 was coimmunoprecipitated with Mdm31 is consistent with a rather weak or transient interaction. We propose that Mdm31 and Mdm32 are subunits of two distinct protein complexes in the inner membrane that cooperate in establishing mitochondrial distribution and morphology.
Mutants lacking the inner membrane proteins Mdm31 and Mdm32 display phenotypes that are strikingly similar to mutants lacking either one of the outer membrane proteins Mmm1, Mmm2, Mdm10, and Mdm12 (Burgess et al., 1994; Sogo and Yaffe, 1994; Berger et al., 1997; Boldogh et al., 1998, 2003; Aiken Hobbs et al., 2001; Youngman et al., 2004). First, mutant cells harbor giant spherical mitochondria; second, aberrant mitochondria are largely immotile; third, the internal structure of mitochondria is dramatically altered; fourth, mtDNA is unstable; and fifth, mtDNA nucleoids are disorganized. Deletion of either one of the MDM31 and MDM32 genes is synthetically lethal with deletion of either one of the MMM1, MMM2, MDM10, and MDM12 genes, suggesting that the gene products are required for the same cellular processes. We propose that Mdm31 and Mdm32 cooperate with Mmm1, Mmm2, Mdm10, and Mdm12 in maintaining mitochondrial morphology.
What might be the role of Mdm31 and Mdm32 in mitochondrial biogenesis? It has been proposed that Mmm1, Mmm2, Mdm10, and Mdm12 are involved in the attachment of mtDNA to the mitochondrial membranes and provide a link to a segregation machinery on the cytosolic side of the organelle. This hypothesis is based mainly on two findings. First, disordered nucleoids are seen in Δmmm1, Δmmm2, Δmdm10, and Δmdm12 mutants (Boldogh et al., 2003; Youngman et al., 2004). Similar structures are also found in the Δabf2 mutant, which lacks a mitochondrial member of the nonhistone high mobility group protein family (Newman et al., 1996). Thus, disordered nucleoids are indicative of a defect of mtDNA packaging and/or attachment to the membrane. Second, GFP fusion proteins of Mmm1, Mmm2, Mdm10, and Mdm12 form foci, a subset of which is located next to a subset of mtDNA nucleoids (Aiken Hobbs et al., 2001; Boldogh et al., 2003; Meeusen and Nunnari, 2003; Youngman et al., 2004). However, the identity of inner membrane proteins that might link matrix-localized nucleoids to the putative segregation machinery in the outer membrane remained obscure. Here, we show that steady-state levels of Mmm1 in mitochondria, Mmm1 foci formation, and assembly of Mmm1 into a high molecular weight complex are not affected in mutants lacking Mdm31 and Mdm32. However, Mmm1-containing complexes lose their ability to interact with mtDNA nucleoids in Δmdm31 and Δmdm32 mutants. We propose that Mdm31 and Mdm32 are required to link mtDNA nucleoids to an Mmm1-containing segregation machinery in the mitochondrial outer membrane.
Respiratory functions of mitochondria are dispensable in S. cerevisiae when cells are grown on fermentable carbon sources. Hence, a defect in mtDNA inheritance is not sufficient to explain the observed synthetic lethal phenotypes of mdm mutants. Besides their role in respiration, mitochondria execute a variety of different metabolic functions, including biogenesis of iron sulfur clusters, which are essential for life (Lill and Kispal, 2000). Thus, the inheritance of the organelle is an essential process. We observed that mitochondria lacking Mdm31 and Mdm32 are almost immotile, similar to mitochondria lacking Mmm1, Mdm10, and Mdm12. Consequently, compromised mitochondrial motility leads to the appearance of mitochondria-free buds in the mutants. It is conceivable that a combination of the defects in double mutants lacking Mmm1 (or Mmm2, or Mdm10, or Mdm12) and Mdm31 (or Mdm32) results in a complete block of mitochondrial transport, and thus causes inviability of daughter cells.
We consider it unlikely that deletion of MDM31 and MDM32 directly influences the ability of mitochondria to bind to the actin-dependent transport machinery because mitochondria lacking Mdm31 and Mdm32 were found to be able to interact with actin in an ATP-dependent manner in vitro. It has been suggested that Mmm1 is required for coupling of mitochondria to the actin cytoskeleton (Boldogh et al., 1998). However, several lines of evidence suggest that Mmm1 is not directly acting as a receptor for actin-dependent motility factors. The function of Mmm1 has been conserved in the filamentous fungus N. crassa, which uses microtubules for mitochondrial transport (Prokisch et al., 2000; Westermann and Prokisch, 2002). Mitochondria isolated from loss-of-function mutants in N. crassa are still able to bind to the cytoskeleton (Fuchs et al., 2002), and the N. crassa protein complements the yeast mutant (Kondo-Okamoto et al., 2003). Similarly, homologues of Mdm31 are found in organisms that rely on microtubules for mitochondrial transport, such as N. crassa and S. pombe. This finding suggests that the main function of Mdm31, Mdm32, and Mmm1 is independent of the cytoskeletal system used by the cell for mitochondrial motility.
Mitochondria lacking Mdm31 and Mdm32 show dramatic changes in the organization of their internal membranes. This is not merely due to a defect in cristae formation, because some cristae are formed in glucose-grown cells (Fig. 4 E), and cristae are quite numerous in glycerol-grown cells (unpublished data). Interestingly, Aiken Hobbs et al. (2001) reported a similar phenotype for Δmmm1 mitochondria. Also in this mutant, cristae were lost and large extended or ring-shaped membrane inclusions were seen. These authors suggested that Mmm1 may be part of an internal scaffold-like structure required for normal mitochondrial shape and attachment of mtDNA. Our observations support a model in which Mdm31 and Mdm32 perform a similar function in the inner membrane. They may cooperate with Mmm1 in maintaining this scaffold-like structure and coordinate the behavior of the outer and inner membrane and provide anchoring sites for mtDNA nucleoids. When this function is lost, the internal structure of the organelle becomes disorganized, mitochondria lose their elongated shape, mtDNA nucleoids are destabilized, and mitochondrial motility is compromised as a consequence of aberrant mitochondrial shape.
Based on their genetic interactions and biochemical data, we can now propose at least three distinct functional entities involved in mitochondrial inheritance, the action of which is superior to the machineries of fusion and fission (summarized in Fig. 8). Mmm1, Mdm10, and Mdm12 have been proposed to be subunits of the same complex in the outer membrane (Boldogh et al., 2003). As combined deletion of their genes does not produce synthetic phenotypes (Berger et al., 1997; Hanekamp et al., 2002), these components likely share the same function. Mmm2 is a subunit of a separate complex in the outer membrane. Even though Δmmm1/Δmmm2 double mutants are viable on fermentable carbon sources, conditional mmm1 and mmm2 alleles produce a synthetic lethal phenotype on nonfermentable carbon sources (Youngman et al., 2004). Hence, Mmm1 and Mmm2 act in functionally separable parallel pathways. Mdm31 and Mdm32 are functionally interdependent subunits of two novel complexes in the inner membrane, which might interact in a transient and dynamic manner. They are the first known inner membrane proteins that cooperate with the outer membrane proteins in inheritance of mitochondria and mtDNA. The functional characterization of novel components involved in these processes that were reported by Youngman et al. (2004) and herein revealed an unanticipated complexity of the machinery controlling mitochondrial behavior. It is a challenge for the future to reveal the precise molecular interactions of these complexes with components in the matrix and on the cytosolic face of the organelle that contribute to the complex process of mitochondrial inheritance.
Materials And Methods
Plasmid and yeast strain constructions
Standard methods were used for cloning of DNA and growth and manipulation of yeast strains. Cloning procedures and strain constructions are described in detail in the online supplemental Materials and methods.
Mitochondria were labeled with mtGFP (Westermann and Neupert, 2000) or mitochondria-targeted DsRed (Mozdy et al., 2000). Staining of the actin cytoskeleton with rhodamine-phalloidin (Amberg, 1998) and DAPI staining of mtDNA in living cells (Aiken Hobbs et al., 2001) was performed according to published procedures. Staining of the vacuole with 5-(and-6)-carboxy-2',7'-dichlorofluorescein diacetate (Molecular Probes) was performed according to the manufacturer's instructions. The ER was visualized with ER-targeted GFP (Prinz et al., 2000). Abf2-containing structures were labeled with a chimeric protein consisting of Abf2 and a GFP moiety derived from mtGFP. Mmm1-containing structures were stained with Mmm1 fused to DsRed.T4 (Bevis and Glick, 2002).
Epifluorescence microscopy was performed using a microscope (model Axioplan 2; Carl Zeiss MicroImaging, Inc.) equipped with a Plan-Neofluar 100×/1.30 Ph3 oil objective (Carl Zeiss MicroImaging, Inc.). Images were recorded either with a SPOT cooled color camera (Diagnostic Instruments) and processed with Lite Meta-Morph imaging software (Universal Imaging Corp.) or with an Evolution VF Mono Cooled monochrome camera (Intas) and processed with Image Pro Plus 5.0 and ScopePro 4.5 software (MediaCybernetics). Confocal images were taken with a confocal microscope (model TCS SP1; Leica) equipped with a 1.2 NA 63× water immersion lens (Leica; 63×, Planapo). For imaging, living cells were embedded in 1% low melting point agarose and observed at RT. Quantification of mitochondrial morphology defects was performed without prior reference to strain identity.
Analysis of mitochondria–actin interactions in vitro
Actin filaments were prepared by polymerizing nonmuscular human actin (tebu-bio GmbH) according to the manufacturer's instructions. Binding of filamentous actin (at a concentration of 100 μg/ml) to isolated mitochondria and cosedimentation of actin with mitochondria were performed as described previously (Lazzarino et al., 1994). Actin was detected by immunoblotting with monoclonal anti-actin antibodies (c4d6; Lessard, 1988).
Gel filtration analysis
Isolated mitochondria (1 mg) were pelleted by centrifugation for 10 min at 10,000 g and resuspended in 200 μl buffer A (1% digitonin, 150 mM NaCl, 5 mM EDTA, 1 mM PMSF, and 10 mM Tris-HCl, pH 7.4). After incubation for 1 h at 4°C under agitation, mitochondrial extracts were centrifuged for 30 min at 90,000 g in a rotor (model TLA45; Beckman Coulter) at 4°C. The supernatant was loaded on a Superose 6 gel filtration column (25-ml column volume; Amersham Biosciences) and chromatographed in buffer A with 0.05% digitonin (flow rate 0.5 ml/min). 0.5-ml fractions were collected, and proteins were precipitated with TCA and analyzed by SDS-PAGE and Western blotting. Calibration standards were as follows: thyroglobulin, 670 kD; apoferritin, 440 kD; alcohol dehydrogenase, 150 kD; carboanhydrase, 29 kD.
Antigens were expressed using the pQE system (QIAGEN) according to the manufacturer's instructions. Antisera were generated by injection of inclusion bodies into rabbits. Subfractionation of yeast cells and isolation, purification, and subfractionation of mitochondria were performed as described previously (Rowley et al., 1994). Mitochondrial fusion was examined according to published procedures (Nunnari et al., 1997; Fritz et al., 2003). Import of radiolabeled Mdm32 and coimmunoprecipitation was performed as described previously (Messerschmitt et al., 2003).
Online supplemental material
An alignment of Mdm31 protein family members is available as Fig. S1. Cloning procedures and yeast strain constructions are described in supplemental Materials and methods.
We thank Jutta Dierolf, Gabi Ludwig, and Margit Vogel for technical assistance; Miriam Hammermeister for her contributions to some experiments; Walter Neupert and Stefan Hell for continuous support; and Hannes Herrmann for many stimulating discussions and critical comments on the manuscript.
This work was supported by the Deutsche Forschungsgemeinschaft through grants SFB 413/B3 and We 2174/3-1.
Abbreviations used in this paper: AAC, ADP/ATP carrier; mtDNA, mitochondrial DNA; mtGFP, mitochondria-targeted GFP; PK, proteinase K.