Introduction

Mitochondria play key roles in cellular energy metabolism, various additional biochemical pathways, developmental processes, apoptosis and aging. These functions are intimately linked to the shape of the mitochondrial compartment. Mitochondria are highly dynamic organelles that constantly adapt their morphology to the requirements of cellular physiology by frequent fusion and fission1,2 and remodeling of their ultrastructure3. As many of the molecular components and cellular pathways have been conserved during evolution, mitochondrial structure and dynamics can be studied in baker’s yeast Saccharomyces cerevisiae as a model organism. Fusion and fission of mitochondria are mediated by large dynamin-related GTPases (DRPs). Fzo1 in yeast and the mitofusins, Mfn1 and Mfn2, in mammals are membrane-bound proteins that constitute the key components of the outer membrane fusion machinery. Mgm1 in yeast and Opa1 in mammals are intermembrane space proteins that mediate inner membrane fusion. Mitochondrial fission is mediated by Dnm1 in yeast and Drp1 in mammals, which are cytosolic proteins that are recruited to the mitochondrial surface by membrane-bound receptors and adaptor proteins where they assemble into helical structures that sever the mitochondrial outer membrane. It is currently unknown whether a separate machinery for inner membrane division exists, or whether both mitochondrial membranes are severed simultaneously by the Dnm1/Drp1 helices assembled on the outer membrane1,2.

Mitochondria also display a complex organization at the ultrastructural level. The mitochondrial inner membrane consists of two functionally and compositionally distinct subcompartments: cristae, which are inner membrane invaginations that accommodate the respiratory chain complexes and the inner boundary membrane, which is closely apposed to the outer membrane. These subcompartments are connected by narrow, tubular cristae junctions3. Several distinct protein complexes shape the mitochondrial inner membrane in a manner that is interdependent with the lipid composition of the organelle, however, their exact modes of action are not fully understood. The inner membrane fusion DRP, Mgm1/Opa1, is thought to play an independent role in maintenance of inner membrane structure and organization4,5. ATP synthase dimerization and higher order assembly are also essential for cristae formation and stabilization6. More recently, a large complex of inner membrane associated proteins, termed mitochondrial contact site and cristae organizing system (MICOS), has been implicated as a master organizing system of the inner membrane, which is thought to function as both a blueprint for cristae placement and cristae stabilization7,8,9.

The lipid composition of mitochondria also plays a critical role in mitochondrial shape and ultrastructure. Mitochondria of yeast cells lacking enzymes involved in cardiolipin (CL) biogenesis exhibit altered mitochondrial morphology, including extremely elongated cristae sheets that sometimes form inner membrane septae or onion-like structures10,11,12. Similarly, mitochondrial cristae defects are observed in Arabidopsis leaf cells, human lymphoblasts and mouse cardiomyocytes defective in cardiolipin biogenesis13,14,15. Furthermore, mitochondrial morphological and ultrastructural defects are observed in yeast and mammalian cells with reduced phosphatidylserine (PS) or phosphatidylethanolamine (PE) levels16,17.

The yeast protein, Mdm33, has been implicated in both mitochondrial dynamics and ultrastructure18. Mdm33 is an integral mitochondrial inner membrane protein with a majority of residues predicted to form coiled coil structures facing the matrix. Mutants lacking Mdm33 contain large, extended mitochondria frequently forming hollow spheres that enclose portions of the cytoplasm. Aberrant Δmdm33 mitochondria contain swollen parts filled with cristae and extended parts that are devoid of cristae. Overexpression of Mdm33 causes mitochondria to become fragmented and to contain septated or vesiculated inner membranes and leads to growth arrest18. However, it is not clear whether Mdm33 plays a primary role in the organization of the inner membrane, or inner membrane division and whether these functions are interconnected. To reveal the role of Mdm33 in mitochondrial shape we systematically analyzed its genetic and proteomic interactions. Our results suggest that MDM33 functions in a genetic network with genes important for inner membrane homeostasis and thereby supports mitochondrial division.

Results

MDM33 interacts with genes involved in phospholipid metabolism and mitochondrial inner membrane homeostasis

The MITO-MAP8 is based on 616,270 distinct pairwise genetic interactions of 1,482 genes and provides a comprehensive view of the connections between cellular pathways related to mitochondrial functions. Utilizing this resource we identified genes that positively or negatively interact with MDM33 in double deletions (Fig. 1A; Table S1). Positive genetic interactions frequently occur between genes acting in a common pathway, whereas negative interactions point to compensatory pathways. Strikingly, MDM33 shows strong positive interactions with the genes encoding enzymes required for synthesis of the major phospholipids, PS, PE and phosphatidylcholine (PC) and negative interactions with genes required for CL biosynthesis (Fig. 1B). Furthermore, MDM33 positively interacts with the prohibitin genes, PHB1 and PHB2, which encode an inner membrane complex modulating mitochondrial phospholipid homeostasis19 and with genes encoding subunits of the ER mitochondria encounter structure (ERMES) that participates in the exchange of lipids between mitochondria and the ER20. Two clusters of negative genetic interactors comprise components of mitochondrial distribution and ATP synthase subunits (Fig. 1A). These data suggest roles of Mdm33 in phospholipid metabolism and mitochondrial membrane homeostasis, processes that are also critical for mitochondrial distribution and function.

Figure 1
figure 1

MDM33 interacts with genes encoding components of phospholipid biosynthesis.

(A) A genetic connection scatter plot of MDM33 was generated using data from the MITO-MAP8. Each point in the scatter plot represents one gene. The x-axis represents the cosine correlation between the Δmdm33 interaction scores and the interaction scores obtained for the represented gene. The cosine correlation coefficients are generated by comparing the genetic interaction profiles of Δmdm33 to the profiles of all other 1,481 genes included in the MITO-MAP8. It represents a measure of the similarity of two genes’ genetic interactomes and serves as a measure of their functional similarity. The y-axis indicates interaction score between Δmdm33 and the represented gene. Several genes acting in phospholipid metabolism, mitochondrial inner membrane homeostasis (prohibitins), mitochondrial ER contacts (ERMES), mitochondrial distribution and ATP synthase are highlighted. (B) Pathways of phospholipid biosynthesis. Positive (green), neutral (black) and negative (red) genetic interaction scores are indicated. CDP-DAG, cytidinediphosphate-diacylglycerol; CL, cardiolipin; MLCL, monolyso-cardiolipin; PA, phosphatidic acid; PC, phosphatidylcholine; PDME, phosphatidyldimethyl-ethanolamine; PE, phosphatidylethanolamine; PG, phosphatidylglycerol; PGP, phosphatidyl-glycerolphosphate; PI, phosphatidylinositol; PMME, phosphatidylmonomethyl-ethanolamine; PS, phosphatidylserine.

Full size image

To gain additional insight into Mdm33 function, we exploited the growth arrest phenotype of cells overexpressing MDM3318, reasoning that Mdm33 requires interaction partners to exert its function and that these interaction partners will also be required to induce the growth arrest. Thus, deletion of a gene encoding an interaction partner required for Mdm33 function should relieve the overexpression-induced growth defect. Therefore, we identified putative suppressors by quantifying the abundance of deletion mutants in cell pools by a microarray-based approach (Fig. S1A,B).

To induce MDM33 overexpression-dependent growth arrest we transformed a multicopy plasmid expressing MDM33 from the strong, inducible GAL1/10 promoter into a pool containing the 4,987 strains of the MATα haploid non-essential yeast deletion library21. We expected that only deletion mutants that are unable to express MDM33 properly or lack Mdm33 interaction partners would grow under inducing conditions. In a proof-of-principle experiment we plated a similar number of transformants of the isogenic wild type and the deletion mutant pool on repressing and inducing media. As expected, transformants of the wild type and pooled deletion mutants grew well on glucose-containing medium (SD) when MDM33 overexpression was repressed. Wild type transformants ceased to grow on galactose-containing medium (SGal) when MDM33 overexpression was induced. In contrast, some transformants of the deletion pool produced colonies under inducing conditions (Fig. 2A) suggesting that yeast cells are able to cope with high Mdm33 levels when critical interacting genes are lacking.

Figure 2
figure 2

Genome-wide screen for suppressors of MDM33 overexpression-induced growth arrest.

(A) Wild type cells or pooled deletion strains were transformed with pYX223-MDM33 (multicopy plasmid for expression of MDM33 under control of the GAL1/10 promoter) or the empty vector and plated on synthetic medium containing glucose (control (SD)) or galactose (induction (SGal)) as carbon source. Growth was observed after 3 days of incubation at 30 °C. (B) A pool containing the 4,987 strains of the MATα haploid non-essential yeast deletion library was transformed with pYX223-MDM33, plated on synthetic medium containing glucose (control) or galactose (induction) as carbon source and strain abundance was quantified by microarray hybridization. Shown are normalized and background corrected microarray fluorescence signal values for the barcodes of each deletion strain. Strains are in descending order according to the fluorescence intensity under inducing condition (blue circles, arbitrary units). Fluorescence intensity under control condition is indicated by orange circles. (C) Scatter plot showing the genetic interaction score between MDM33 and the represented gene according to the MITO-MAP (compare Fig. 1) and the normalized and background corrected microarray fluorescence signal taken from (B). (D) Strains were transformed with pYX223-MDM33 for expression of MDM33 under control of the GAL1/10 promoter or the corresponding empty vector. 10-fold serial dilutions were spotted on synthetic complete medium containing glucose or galactose as carbon source and incubated at 30 °C for 2 (glucose) or 8 (galactose) days. In this and the following figures all samples that are shown in the same figure received an identical treatment in the same experiment. (E) Strains were transformed with an empty vector or a multicopy plasmid for overexpression of SCO2 or YHM2 from the GAL1/10 promoter. 10-fold serial dilutions were spotted on synthetic complete medium containing glucose or galactose as carbon source. Growth was scored after incubation at 30 °C for 3 days.

Full size image

To screen for genes interacting with MDM33, we grew the transformed yeast deletion pool under repressing and inducing conditions on agar plates. As each strain of the yeast deletion library is marked with a unique molecular barcode sequence21 (Fig. S1A) microarray hybridization can be used to quantify deletion mutants in pools (Fig. S1B). We isolated genomic DNA from the pooled transformants and determined the abundance of the deletion mutants under inducing vs. repressing conditions in two independent experiments. After high-density oligonucleotide array hybridization, approximately 80% of the deletion mutants produced a good signal after growth on glucose-containing medium, but were not detectable after growth on galactose-containing medium (Fig. 2B). As expected, Δgal3 and Δgal4, which lack factors important for induction of the GAL1/10 promoter, grew well under inducing conditions (Fig. S1C). Several hundred mutants lacking proteins of various cellular functions produced a significant signal under inducing conditions and thus represent putative interaction partners of MDM33 (Table S2).

To identify candidates that genetically interact with MDM33 upon both overexpression and deletion we merged the results from the suppressor screen with the MITO-MAP (interaction score less than -3 or more than 3). Six genes showed high scores in both data sets (Fig. 2C). ELP3 encodes a subunit of the RNA polymerase II holoenzyme and is responsible for transcriptional elongation22. This gene shows a very high number of genetic interactions and was excluded from further analysis as it likely interacts with MDM33 in an indirect manner. The remaining five genes encode mitochondrial proteins, namely the two subunits of the prohibitin complex, Phb1 and Phb219; a factor known to be associated with mitochondrial lipid metabolism, Fmp3023; subunit g of the ATP synthase, Atp2024; and a regulatory subunit of the mitochondrial protein import motor complex, Pam1725.

After transformation of the MDM33-overexpressing plasmid into deletion mutants, we confirmed that absence of Atp20, Fmp30, Pam17, Phb1, or Phb2 suppresses the growth defect of cells (Figs 2D and S2B). It has been observed that overexpression of several genes encoding mitochondrial inner membrane proteins is toxic in yeast26. To test whether our screen yielded suppressors specific for MDM33 we examined the effect of overexpressing two non-related genes encoding mitochondrial inner membrane proteins, SCO2 and YHM2, that have been reported to cause a growth arrest when expressed from the GAL1/10 promoter26. Sco2 acts in the delivery of copper to the cytochrome c oxidase27, whereas Yhm2 catalyzes mitochondrial citrate/oxoglutarate antiport28. Deletion of the prohibitin genes did not affect the growth of SCO2 or YHM2 overexpressing strains (Fig. 2E). We conclude that the suppression of overexpression-induced growth defects in our suppressor mutant strains is specific for MDM33. In addition, we confirmed that deletions of prohibitin genes and MDM33 display positive genetic interactions (Fig. S2A,B), as indicated by the MITO-MAP. Taken together, our genetic analysis suggests that Phb1, Phb2, Pam17, Fmp30, Atp20 and Mdm33 act in closely related cellular pathways. In the case of Pam17 we consider it possible that the Δpam17 deletion compromises the activity of the TIM23 protein import machinery and thereby reduces the levels of Mdm33 accumulating in mitochondria and renders the cells more resistant to MDM33 overexpression.

Next, we asked whether the overexpression of prohibitins would cause a growth arrest, similar to MDM33 overexpression. As Phb1 and Phb2 are functionally interdependent we overexpressed both proteins simultaneously from the same plasmid under control of the GAL1/10 promoter29. This overexpression resulted in a slight growth defect (Fig. S2C). However, simultaneous overexpression of Phb1 and Phb2 together with Mdm33 was almost lethal (Fig. S2C). This result confirms the genetic interactions and points to a common, non-redundant role of Mdm33 and prohibitins in the same cellular pathway.

In addition to its role in ATP synthesis the ATP synthase plays an important role in mitochondrial cristae biogenesis3. We reasoned that defects in ATP synthesis, but not in cristae formation might be phenocopied by growth on oligomycin-containing medium, an inhibitor of the ATP synthase. Addition of 5 μg/ml oligomycin to the medium prevented growth of wild type cells on non-fermentable carbon sources, suggesting that ATP synthase activity is effectively blocked under these conditions (data not shown). However, on fermentable medium the addition of oligomycin had no effect on the growth of wild type or Δmdm33 cells (Fig. S2D), suggesting that the interactions with genes encoding ATP synthase subunits are not caused by defects in ATP synthesis. Intriguingly, Atp20 is an ATP synthase subunit mediating dimerization of the ATP synthase complex, a process crucial for cristae biogenesis3,6. The identification of this gene both in the MITO-MAP and in the MDM33 overexpression screen suggests that the role of the ATP synthase in mitochondrial structure, rather than ATP synthesis, is important for the genetic interactions with MDM33.

Mdm33 is in proximity of Phb1, Phb2 and Atp20

To test whether the genetic interactions reflect physical proximity of the proteins in cells, we searched for Mdm33 interaction partners using a proteomic analysis. Specifically, we chemically cross-linked lysates of cells expressing a functional GFP-Mdm33 fusion using DSP and purified Mdm33 using anti-GFP antibodies. Interacting proteins were identified by mass spectrometry (LC-MS/MS). Analysis of total spectral counts indicated that the most robust interacting proteins were the prohibitins, which were detected proportionally to Mdm33 in purifications from cells expressing GFP-Mdm33 at endogenous levels from an allele integrated at the MDM33 locus, a low-copy, or a multi-copy plasmid. In addition, Atp1 and Atp2, the alpha and beta subunits of the F1 sector of the ATP synthase, were found to interact with GFP-Mdm33 (Table 1). Thus, proteomic analysis suggests proximity between the Mdm33 complex, prohibitins and the ATP synthase. These results are in good agreement with the genetic interactions of MDM33 with PHB1, PHB2 and ATP20.

Table 1 Mdm33 is in proximity of prohibitins and ATP synthase.
Full size table

Phospholipid homeostasis rather than fusion-fission dynamics is required for the formation of aberrant mitochondria in Δmdm33 cells

Cells lacking MDM33 have a distinct mitochondrial morphological defect that includes formation of characteristic lariat-shaped structures and hollow spheres18. Therefore, we systematically examined mitochondrial morphology in cells with combined deletions of the MDM33 gene and genes identified in our genetic and proteomic approaches and also genes encoding proteins directly involved in mitochondrial structure or dynamics (Table 2). Δmdm33 was found to be epistatic to most of the other deletion alleles, i.e. the Δmdm33 mitochondrial morphology phenotype was dominant, substantiating the central role of Mdm33 in mitochondrial morphogenesis. However, deletion of FMP30, GEM1, MDM10, MDM12, MDM31, MDM34, or MMM1 resulted in a mitochondrial morphology phenotype that was epistatic to Δmdm33, i.e. the distinct characteristic mitochondrial morphological defect of Δmdm33 was largely lost in these double mutants (Fig. 3). Δmdm33 Δphb1 and Δmdm33 Δphb2 double mutants showed an intermediate phenotype with largely tubular mitochondria and only few, small lariat-like structures (Fig. 3). Intriguingly, all of these genes are implicated in mitochondrial lipid metabolism: (i) Mdm10, Mdm12, Mdm34 and Mmm1 are core components of the ERMES complex that forms contact sites connecting the ER with mitochondria and participates in the import of lipids from the ER into mitochondria20,30,31. Gem1 is a regulator of ERMES32,33. (ii) Both Fmp30 and prohibitins are known to show strong genetic interactions with genes involved in mitochondrial CL and PE biosynthesis and Fmp30 is required for maintenance of CL in the absence of mitochondrial PE synthesis23,34,35. (iii) Mdm31 is known to play an important role in CL biosynthesis in mitochondria and overexpression of MDM31 can partially compensate for the loss of ERMES36. The requirement of mitochondrial phospholipid biosynthesis factors for the manifestation of the characteristic Δmdm33 mitochondrial phenotype further supports a potential role of Mdm33 in this pathway.

Table 2 Screen for genes that are required for the formation of aberrant mitochondrial structures in the Δmdm33 mutant.
Full size table
Figure 3
figure 3

Formation of aberrant mitochondrial structures in Δmdm33 cells depends on components of inner membrane homeostasis, but not fusion and fission pathways.

Cells expressing mitochondrial matrix targeted GFP (mtGFP) were grown to logarithmic growth phase in YPD and analyzed by fluorescence microscopy. Cell outlines are indicated by a white line. The arrow indicates smaller spherical mitochondria in Δmgm1 Δmdm33 double mutant cells. Bar, 5 μm. See also Table 2.

Full size image

In contrast, double mutants lacking the MDM33 gene and genes encoding components of the mitochondrial fission machinery, DNM1, FIS1, or MDV1, resembled that of Δmdm33 single mutants. Lariat-shaped mitochondria characteristic for Δmdm33 could be found in double mutants lacking mitochondrial fusion factors Fzo1, Mdm30, or Mgm1. Mitochondria of these double mutants were more fragmented than that of Δmdm33 single mutants due to the fusion defect (Table 2). Triple mutants lacking Mdm33 and the outer membrane fusion and fission factors Fzo1 and Dnm1 appeared like the Δmdm33 single mutant (Fig. 3). Thus, mitochondrial fusion-fission dynamics is not required for formation of lariat-shaped mitochondria in cells lacking Mdm33. These results suggest that Mdm33 acts upstream of mitochondrial fusion and fission.

Mdm33 affects phospholipid homeostasis in mitochondria

Our genetic and proteomic analyses suggested that MDM33 participates in the organization of the inner membrane and mitochondrial lipid homeostasis. Therefore, we analyzed the phospholipid composition of mitochondria isolated from wild type, Δmdm33 cells, or MDM33-overexpressing cells by mass spectrometry. Although deletion of MDM33 had no measurable effect on mitochondrial phospholipid composition, phospholipids that are synthesized in mitochondria, PE and CL, were strongly reduced in mitochondria from cells overexpressing MDM33 (Fig. 4A), suggesting that Mdm33 affects mitochondrial phospholipid homeostasis. The combined deletion of genes required for production of CL and PE is synthetic lethal in yeast, indicating that CL and PE are partially redundant and that a sufficient level of either of them is required for cell survival37. Thus, disturbance of mitochondrial phospholipid biosynthesis upon overexpression of MDM33 might contribute to the observed growth defect. Furthermore, it is known that changes in mitochondrial lipid composition cause septation of the mitochondrial inner membrane11. Therefore, we consider it likely that alterations of the PE and CL content in mitochondrial membranes contribute to the mitochondrial ultrastructure defect observed upon overexpression of MDM3318 (Fig. 5F).

Figure 4
figure 4

Mdm33 affects mitochondrial phospholipid homeostasis.

(A) Phospholipidome of mitochondria isolated from cells transformed with pYX223-MDM33 for expression of MDM33 under control of the GAL1/10 promoter (blue) or the corresponding empty vector (orange) and grown in synthetic complete medium containing galactose as carbon source. Phospholipid analysis was done by quantitative mass spectrometry. Bars represent the mean values of two independent mitochondrial preparations. The mean values of two technical replicates for each of the mitochondrial preparations are indicated by black dots. (B) Wild type cells expressing mitochondrial matrix targeted GFP (mtGFP) and ERFP-tagged Mmm1 (pRS316-MMM1-ERFP) were transformed with pYX223-MDM33 or the empty vector, grown to the logarithmic growth phase in synthetic complete medium containing galactose as carbon source and analyzed by fluorescence microscopy. Bar, 5 μm. (C) Western blot analysis of whole cell extracts of wild type cells carrying an empty vector or pYX223-MDM33. Cells were grown overnight in synthetic complete medium containing galactose as carbon source and diluted to logarithmic growth phase. Protein was extracted from the cells by boiling in sample buffer after alkaline treatment. (D) In vitro Psd1 activity assay. Mitochondria were isolated from wild type cells carrying an empty vector or overexpressing Mdm33 (pYX223-MDM33) or a FLAG-tagged version of Mdm33 (pYX223-F-MDM33) and incubated with liposomes containing NBD-PS for 30 minutes at 30 °C. Total lipids were isolated and separated by TLC. Shown is the NBD-fluorescence. The same samples were analyzed by Western blotting. (E) Quantifications are the ratio of the NBD-PE and NBD-PS signals normalized to the wild type-ratio. Shown are mean values and standard deviation obtained from three independent experiments.

Full size image
Figure 5
figure 5

Mdm33 is required for mitochondrial fission.

(A) Cells expressing mitochondrial matrix targeted GFP (mtGFP) were grown to logarithmic growth phase in YPD, either mock-treated or incubated for 40 min with 0.5 mM sodium azide, fixed in 3.7% formaldehyde and analyzed by fluorescence microscopy. For each strain cells containing fragmented mitochondria were scored. Error bars indicate standard deviations of 3 independent experiments with 150 cells per experiment. Images are merges of DIC and GFP fluorescence. Bar, 5 μm. (B) Time lapse series of cells grown to logarithmic growth phase in synthetic complete medium and expressing mitochondrial matrix targeted ERFP (mtERFP) and Dnm1-GFP. Images represent maximum intensity projections of fluorescence image z stacks. Arrows highlight mitochondrial division events. Bar, 5 μm. (C) Cells expressing Dnm1-GFP, Mmm1-ERFP and mitochondria targeted BFP (mtBFP) were grown to logarithmic growth phase in synthetic complete medium and analyzed by fluorescence microscopy. For both strains cells were scored for association of Mmm1-ERFP with Dnm1-GFP patches. Bar, 5 μm. (D) Strains were transformed with a multicopy plasmid for overexpression of MDM33 from the inducible GAL1/10 promoter (pYX223-MDM33) or an empty vector. 10-fold serial dilutions were spotted on synthetic complete medium containing glucose (repression of the GAL promoter) or galactose (induction of the GAL promoter) as carbon source and incubated at 30 °C for 2-4 days. (E) Control cells (empty vector) and cells overexpressing MDM33 from the GAL1/10 promoter (pYX223-MDM33) and mtGFP were grown overnight in synthetic complete medium containing galactose as carbon source, diluted to logarithmic growth phase and analyzed by fluorescence microscopy. Bar, 5 μm. (F) Electron micrographs of ultrathin sections of cells grown as in (E). Bars, 200 nm.

Full size image

PE biosynthesis occurs via multiple pathways38 (Fig. S3). In yeast the main pathway starts in the ER where PS is synthesized. PS is then transported to the mitochondrial inner membrane where the PS decarboxylase, Psd1, converts it to PE39,40. Contact sites of ER and mitochondria play a major role in the import of lipids into mitochondria20,41,42,43. Therefore, we checked whether the formation of ERMES, which was suggested to play a role in phospholipid exchange between both organelles20, is impaired by overexpression of MDM33. Impairment of ERMES is expected to change the localization of Mmm1, an ER-resident ERMES subunit, from patch-like assemblies to a diffuse ER signal20. Analysis of Mmm1-ERFP expressing strains by fluorescence microscopy revealed that ERMES localization is not affected by MDM33 overexpression (Fig. 4B), suggesting that at least ERMES-dependent ER-mitochondria contact sites persist under these conditions.

Psd1 is required for processing of Mgm116, which exists in two alternatively processed forms, both of which are required for mitochondrial inner membrane fusion44. As mitochondria are similarly fragmented in cells lacking Mgm145 and in cells overexpressing MDM3318, we considered the possibility that altered lipid composition of mitochondria containing excess Mdm33 results in a defect in processing of Mgm1. However, Western blot analysis of total cell extracts revealed that this is not the case (Fig. 4C).

Next, we examined the conversion of PS to PE by addition of liposomes containing fluorescently labeled PS (NBD-PS) to isolated mitochondria and subsequent thin layer chromatography of mitochondrial lipids46. As expected, conversion of PS to PE was completely blocked when mitochondria were isolated from strains lacking phosphatidylserine decarboxylase activity (Fig. S4A). In contrast, mitochondria isolated from Δmdm33 cells showed wild type-like PS to PE conversion activity (Fig. S4B). Strikingly, mitochondria containing high Mdm33 levels showed only about 50-70% PS to PE conversion activity compared to the wild type (Fig. 4D,E). A Western blot analysis revealed that the protein level of Psd1 was not changed by MDM33 overexpression (Fig. 4D). Taken together, our results suggest that the presence of excess Mdm33 has an impact on PE synthesis in mitochondria.

Mdm33 contributes to mitochondrial division

Δmdm33 cells harbor large mitochondria that are often interconnected and overexpression of MDM33 results in mitochondrial fragmentation. These observations are suggestive of a role of Mdm33 in mitochondrial division18. To test a function of Mdm33 in this process more directly, we induced mitochondrial fragmentation in wild type, Δdnm1 and Δmdm33 cells by treatment with sodium azide and observed mitochondrial morphology by fluorescence microscopy (Fig. 5A). Treatment of wild type cells led to rapid mitochondrial fragmentation, while Δdnm1 mitochondria remained interconnected, confirming that fragmentation is dependent on the mitochondrial fission machinery. In contrast, Δmdm33 mutants retained considerable fission activity. However, the number of cells with fragmented mitochondria was strongly reduced compared to the wild type suggesting that Mdm33 contributes to mitochondrial fission (Fig. 5A). Next, we analyzed cells expressing Dnm1-GFP by time-resolved live cell fluorescence microscopy. We could observe Dnm1-GFP-dependent matrix constriction and mitochondrial division in Δmdm33 cells (Fig. 5B). However, these events were restricted to a rather small tubular portion of the mitochondria and never occurred in the large ring-like structures. We conclude that Mdm33 contributes to efficient mitochondrial division, albeit it does not appear to constitute an essential component of the mitochondrial division machinery.

ER tubules wrap around mitochondria and mediate mitochondrial constriction prior to Dnm1 assembly47. This is spatially and functionally linked to ERMES, which tethers the ER and mitochondria20,30,48. The Miro GTPase Gem1 is required to disintegrate these contacts after the division event, thereby generating motile mitochondrial tips48. We asked whether the association of ERMES with sites of mitochondrial division might be disturbed in cells lacking Mdm33. We observed that the ER-resident ERMES subunit Mmm1 colocalizes with Dnm1-GFP in wild type and Δmdm33 cells, suggesting that this step of mitochondrial division does not require Mdm33 (Fig. 5C).

To test whether mitochondrial fragmentation and growth arrest upon MDM33 overexpression depend on the outer membrane fission machinery we overexpressed MDM33 in Δdnm1 strains. We observed a strong growth defect in cells lacking Dnm1, although we did not observe mitochondrial fragmentation (Fig. 5D,E). Instead we found parts of the mitochondrial network to be swollen. Electron microscopy revealed that overexpression of MDM33 in the absence of Dnm1 causes swelling of the mitochondria and inner membrane septae formation (Fig. 5F). This indicates that mitochondrial fragmentation upon MDM33 overexpression is Dnm1-dependent, whereas growth arrest and inner membrane remodeling are independent of Dnm1. Taken together, our results are consistent with the idea that Mdm33 is important to keep mitochondria in a fission-competent shape.

Discussion

Here, we report several lines of evidence for an interaction of Mdm33 with molecular components and pathways of mitochondrial inner membrane homeostasis. The Δmdm33 deletion shows negative genetic interactions with genes required for synthesis of CL and positive genetic interactions with genes required for synthesis of the major phospholipids PS, PE and PC and with the prohibitin genes, PHB1 and PHB2. The growth arrest upon MDM33 overexpression is relieved by deletion of genes with a known role in mitochondrial lipid homeostasis, PHB1, PHB2 and FMP30. Taken together, our genetic analyses revealed a highly interconnected genetic interaction network also including ERMES, the ERMES regulatory subunit Gem1 and genes of the mitochondrial PE and CL biosynthesis pathways, namely PSD1 and CRD1 (Fig. 6). Furthermore, Mdm33 was found in crosslinks with Phb1 and Phb2. Thus, genetic data coupled with the proteomic analysis point to an involvement of Mdm33 in inner membrane phospholipid homeostasis.

Figure 6
figure 6

MDM33 is part of a genetic network modulating mitochondrial phospholipid biosynthesis.

Nodes in the genetic interaction network of MDM33 represent deleted or overexpressed genes; connecting lines represent genetic interactions. The network is based on the scores obtained in the MITO-MAP (blue and grey) and in the suppressor screen (blue). Green and red lines indicate genetic interactions between deletion mutants (MITO-MAP); black lines indicate genetic interactions of MDM33 overexpression with deletion mutants (suppressor screen). The genetic interaction network was created using the Cytoscape software60.

Full size image

Prohibitins are known modulators of mitochondrial inner membrane homeostasis and it is well established that they have an intimate functional relationship with the lipid composition of mitochondrial membranes, especially with the levels of PE and CL19. We previously proposed that ring-like prohibitin complexes might serve as membrane organizers modulating the distribution of PE and CL within the membrane35. Strikingly, Mdm33 overexpression lowers the levels of the same phospholipids that are also affected by prohibitins, namely PE and CL. An interaction of Mdm33 with prohibitins is further substantiated by three independent lines of evidence: a positive genetic interaction of prohibitin deletions with Δmdm33, suppression of MDM33 overexpression-induced growth arrest by prohibitin deletions and a physical interaction of prohibitins with Mdm33 in cross-linked cell extracts. It is therefore conceivable that Mdm33 cooperates with prohibitins to modulate the proper organization of the inner membrane and the associated lipid biosynthetic pathways.

Accumulating evidence suggests that modulation of lipid composition is an important aspect of mitochondrial membrane fusion and fission38,49. For example, the biosynthesis of PE and CL is intimately linked to mitochondrial fusion in yeast16,50,51, CL plays a major role in mitochondrial division in Arabidopsis thaliana52 and CL stimulates the assembly of Mgm1/Opa1 and Drp151,53,54. We show here that excess Mdm33 has an impact on the abundance of PE and CL. These are both non-bilayer forming lipids with small headgroups and therefore confer negative curvature to mitochondrial membranes, a feature that is particularly important during fusion or fission55. Thus, it is conceivable that Mdm33 activity affects the dynamic properties of the inner membrane.

Our previous characterization of the Δmdm33 mutant provided evidence for a role of Mdm33 in mitochondrial inner membrane division. Mutant cells have a unique phenotype with extremely extended and often ring-shaped giant mitochondria. We proposed that these aberrant organelles can form only in the absence of frequent division events18. Here, we show that deletion of the MDM33 gene impedes azide-induced mitochondrial fragmentation. Furthermore, time-resolved fluorescence microscopy revealed that Dnm1 and ERMES-dependent mitochondrial fission events do occur in the absence of Mdm33, albeit at strongly reduced frequency. This indicates that Mdm33 plays an auxiliary role in mitochondrial fission, rather than being an essential part of the division machinery. Lariat-shaped mitochondria characteristic of mutants lacking MDM33 are prevalent in double and triple mutants lacking fusion and fission factors. Thus, Δmdm33 is epistatic to the deletion of genes encoding core components of the fusion and fission machineries. Furthermore, Mdm33 overexpression-induced inner membrane septae form independently of Dnm1. These observations suggest that Mdm33 acts upstream of fusion and fission.

Taken together, our observations suggest that Mdm33 is in an ideal position to link mitochondrial inner membrane lipid homeostasis and structure to mitochondrial dynamics. We consider it possible that the activity of Mdm33 is required to prepare the inner membrane for division by local modulation of its phospholipid composition and biophysical properties. For example, Mdm33 might contribute to the formation of microdomains that are particularly enriched in PE and CL to generate curvature. Formation of these microdomains then renders the membrane competent for fission (i.e. fusion of the inner membrane from the matrix side). Future work will be required to test this idea and reveal whether Mdm33 is directly involved in this process.

The exact molecular role of Mdm33 remains unknown. The reduction of mitochondrial PE and CL levels upon Mdm33 overexpression points to an involvement in phospholipid metabolism. On the other hand, we did not detect an altered lipid composition of Δmdm33 mitochondria. Thus, it is possible that Mdm33 is not the limiting factor in wild type mitochondria, or that the activity of Mdm33 modulates the local distribution, rather than the absolute level, of mitochondrial phospholipids. It is equally well possible that Mdm33 indirectly affects membrane phospholipid homeostasis, e.g. by acting as a regulator of the prohibitin ring complex, or MICOS assembly, or ATP synthase dimerization. We are confident that the genetic and proteomic interaction network reported here will help to reveal the molecular function of Mdm33 in the future.

Methods

Yeast strain constructions

Plasmids and cloning procedures are described in the supplementary methods. Standard procedures were used for manipulation of yeast. Yeast deletion mutants were taken from the yeast deletion collection21. Double deletion mutants were constructed by mating and tetrad dissection. Yeast strains were isogenic to BY4741, BY4742 and BY474356, if not indicated otherwise. For rapid generation of double mutants the synthetic genetic array (SGA) technology was used57. ORF replacement of MDM33 was achieved by homologous recombination. The URA3 marker used for this purpose was amplified using the plasmid pYES-mtGFP58 as template and oligonucleotides 5′-GAT CAT TGG GGT CTT TTT CGT TGT GAA ATT GTA ACG GGT GAA CTC AGT GAT TGA ATC TTA GAT CAC ACT GCC TTT G and 5′-TGT ATT TAT GAT TTT ATT ATG TAC AAG GAT AAA GGA TGA AAA AAA TGC ATG CGT GTT ACC GCT GTT GAG ATC CAG TTC.

Microscopy

Epifluorescence microscopy was performed using an Axiophot or an Axioplan 2 microscope (Carl Zeiss Lichtmikroskopie, Göttingen, Germany) equipped with a Leica DCF360FX Camera with Leica LAF AF Version 2.2.1 Software (Leica Microsystems, Wetzlar, Germany) or an Evolution VF Mono Cooled monochrome camera (Intas, Göttingen, Germany) with Image ProPlus 5.0 and Scope Pro4.5 software (Media Cybernetics, Silver Spring, MD), respectively. For time-resolved live cell microscopy in Fig. 5B cells were observed with a Leica DMI 6000 wide field fluorescence microscope equipped with a Leica DFC360FX camera and Leica LAS AF Software Version 2.1.0. Image manipulations other than minor adjustments of brightness and contrast were not performed. For electron microscopy, cells were grown to log phase and prepared essentially as described59. Ultrathin 50 nm sections were poststained for 20 min with 2% uranyl acetate and for 3 min in lead citrate. Samples were examined in a Zeiss CEM 902 (Carl Zeiss, Oberkochen, Germany) transmission electron microscope operated at 80 kV. Micrographs were taken using a 1350 × 1050 pixel Erlangshen ES500W CCD camera (Gatan, Peasanton, CA) and Digital Micrograph software (version 1.70.16).

Microarray design and hybridization, immunoprecipitation and LC MS/MS analysis

The rationale and procedure for microarray design and hybridization is outlined in the supplementary methods. Immunoprecipitation was performed basically as described previously8 with adjustments, as outlined in the supplementary methods.

Lipid profiling by mass spectrometry and in vitro Psd1 activity assay

Mitochondria for lipid profiling and measurement of Psd1 activity were isolated from yeast cells by differential centrifugation and further purified by sucrose gradient centrifugation. Mass spectrometric phospholipid analysis of isolated mitochondria was performed as described11. The preparation of liposomes and the in vitro assay for Psd1 activity were performed essentially as described previously46. Lipids in stock solutions in chloroform were mixed at the desired molar ratio and the solvent was evaporated under a flow of dry nitrogen. The dried lipids were hydrated in 800 μl of reaction buffer (300 mM sucrose, 150 mM KCl, 10 mM Tris-HCl, pH 7.5, 1 mM DTT) by repeated cycles of incubation at 30 °C, vortexing and freeze-thawing. After 1 h hydration the liposomes were prepared by extruding 30 times at 30 °C using an Avanti Mini-Extruder with 0.2 μm polycarbonate membranes according to the manufacturer’s instructions. Liposomes were stored at 4 °C and used within 5 days.

For thin layer chromatography, phospholipids were extracted from mitochondria by vortexing in 500 μl of 2:1 chloroform/methanol for 15 min at room temperature using a Disruptor Genie (Scientific Industries, Bohemia, NY). 100 μl of water were added and the samples were vortexed for additional 5 min. The organic phase was separated by centrifugation at 400 x g for 5 min and dried under a constant flow of nitrogen. The samples were resuspended in 60 μl of chloroform and 20 μl of each sample was subjected to TLC analysis. Silica gel plates (Fluka Analytical, Sigma-Aldrich, St. Louis, MO) were developed with chloroform/methanol/acetone/water/acetic acid (50:10:20:5:15, vol/vol/vol/vol/vol). NBD fluorescence was imaged with an ImageQuant LAS 4000 gel documentation system (GE Healthcare Europe GmbH, Freiburg, Germany) using excitation and detection wavelengths for GFP.

Additional Information

Accession codes: Microarray and sample annotation data were deposited at NCBI’s Gene Expression Omnibus (GEO) data repository under accession no. GSE49580.

How to cite this article: Klecker, T. et al. Interaction of MDM33 with mitochondrial inner membrane homeostasis pathways in yeast. Sci. Rep. 5, 18344; doi: 10.1038/srep18344 (2015).