Mcp1 and Mcp2, two novel proteins involved in mitochondrial lipid homeostasis

Mitochondria are essential organelles of all eukaryotic cells that cannot form de novo and their biogenesis depends on different cellular processes. The vast majority of mitochondrial proteins have to be imported from the cytosol into the organelle. The proper inheritance of mitochondria into the daughter cell is facilitated by the tubular morphology of the organelle and its interaction with cytoskeletal structures. Furthermore, the intimate contact of mitochondria with the ER plays an important, although still ill-defined, role in mitochondrial function. A central function of the ER–mitochondria contact sites is probably lipid exchange between the two compartments (for reviews, see Daum and Vance, 1997; Rowland and Voeltz, 2012).

Mdm10 is a yeast mitochondrial protein proposed to be involved in many mitochondrial processes. It was first described in 1994 and shown to be an integral mitochondrial outer membrane (MOM) protein. Mutation and deletion of the gene results in altered mitochondrial morphology (condensed giant organelles) and distribution phenotype (Sogo and Yaffe, 1994). Deletion of the genes encoding the MOM proteins Mdm12, Mmm2/Mdm34 and the ER protein Mmm1 led to similar growth defects and mitochondrial inheritance and morphology phenotypes (Burgess et al., 1994; Berger et al., 1997; Dimmer et al., 2002; Youngman et al., 2004; Kornmann et al., 2009). Therefore these components were referred to as tubulation mediators (Okamoto and Shaw, 2005). Additionally it was reported that Mmm1 and Mdm10 are involved in the interaction of mitochondria with the actin cytoskeleton (Boldogh et al., 1998). Moreover, Mmm1, Mdm10 and Mdm12 have been identified as components of a high molecular mass complex essential for the efficient transmission of mtDNA to the daughter cell (Boldogh et al., 2003). Mdm10 is predicted to be a β-barrel protein of the outer membrane of mitochondria and depends on the topogenesis of the mitochondrial outer membrane β-barrel proteins (TOB) complex [also known as sorting and assembly machinery (SAM) complex] for proper insertion into the MOM (Paschen et al., 2003; Wiedemann et al., 2003). In addition, Mdm10 was reported to be a transient component of the TOB complex and to play a role in the assembly of β-barrel proteins (Meisinger et al., 2004; Wideman et al., 2010; Yamano et al., 2010b).

Recently, Mdm10 and the other tubulation mediators (Mdm12, Mmm1 and Mmm2) were shown to be components of the so called ER–mitochondria encounter structure (ERMES) complex (Kornmann et al., 2009). The MOM calcium-binding rho GTPase Gem1 was also suggested to associate with the ERMES complex (Kornmann et al., 2011; Stroud et al., 2011), although this notion was very recently debated (Nguyen et al., 2012). Accumulating evidence suggests that the ERMES complex and its component Mdm10 are involved in the lipid homeostasis of mitochondria. Mitochondria lacking Mdm10 have altered cardiolipin (CL) and phosphatidylethanolamine (PE) levels (Kornmann et al., 2009; Osman et al., 2009; Yamano et al., 2010b; Nguyen et al., 2012; Tamura et al., 2012a). Collectively, the aforementioned putative roles of Mdm10 leave open the question what the primary function of this protein is.

In the current study, we investigated the distinct role of Mdm10. To this end, we screened for multi-copy suppressors of the mdm10Δ growth phenotype and identified two proteins of so far unknown function, Mcp1 and Mcp2. Our results demonstrate that these novel mitochondrial proteins play a role in mitochondrial lipid homeostasis and further shed new light on the involvement of the ERMES complex in ergosterol trafficking.

The deletion of MDM10 leads to reduced growth of yeast cells under all conditions tested (Sogo and Yaffe, 1994). Most prominent is the growth defect on non-fermentable carbon sources (Fig. 1A). To better understand the molecular function of Mdm10 we searched for new components genetically interacting with Mdm10. Thus, we screened for high-copy suppressors of the mdm10Δ growth phenotype. To avoid indirect effects, we confirmed that the used haploid mdm10Δ strain still contained mtDNA (ρ⁺, data not shown). We transformed mdm10Δ cells with a yeast genomic library cloned in a high copy number (2 µ based) yeast shuttle vector. Transformants were initially selected on selective medium and then replica plated on glycerol-containing rich medium. Suppressor clones were identified by growth at elevated temperatures (37°C) and then analysed again at normal and reduced temperatures. We screened a total of 30,000 colonies and identified five clones that exhibited prominent rescue of the growth phenotype at 37°C.

Next, plasmids isolated from these clones, and a plasmid encoding MDM10 as a control, were transformed into the mdm10Δ cells and the growth of the transformed cells was monitored at both 30°C and 37°C on a non-fermentable carbon source. In addition, the inserted DNA fragments in the rescuing plasmids were sequenced. Two of the clones contained the MDM10 open reading frame (ORF) (Fig. 1A, #18, and data not shown). Two other plasmids contained a DNA fragment encoding ARS1523, YOR228C and WTM2 as annotated sequence features (Fig. 1A, #15, and data not shown). ARS1523 is an autonomously replicating sequence, whereas WTM2 encodes a transcriptional modulator without apparent link to mitochondrial function. YOR228C encodes an uncharacterized protein that was found in a proteomic analysis of the yeast mitochondrial outer membrane (Zahedi et al., 2006). The fifth clone contained a plasmid that comprised the annotated elements SYM1, YLR252W, YLR253W and NDL1 (Fig. 1A, #13). Interestingly, plasmid #13 has a very weak rescue capacity at 37°C, but rescues the growth defect quite well at 30°C (Fig. 1A). According to the Saccharomyces Genome Database (SGD) (www.yeastgenome.org), NDL1 has no obvious connection to mitochondrial function, and YLR252W is a dubious ORF. SYM1 encodes for a transporter of the inner mitochondrial membrane (Trott and Morano, 2004; Dallabona et al., 2010; Reinhold et al., 2012) with no obvious link to Mdm10 or any of its proposed functions. Of note, YLR253W encodes for a yet uncharacterized protein that was detected in the mitochondrial proteome (Reinders et al., 2006).

We concentrated on the two non-characterized ORFs YOR228C and YLR253W as well as SYM1 and cloned them into a centromeric yeast expression vector. Overexpression of both former proteins rescued the growth phenotype of mdm10Δ cells under almost all tested conditions while YOR228C was under all conditions the more effective suppressor (Fig. 1B; supplementary material Fig. S1). We therefore named the proteins Mcp1 (Mdm10 complementing protein 1, Yor228c) and Mcp2 (Mdm10 complementing protein 2, Ylr253w). In contrast, overexpression of Sym1 could not restore the growth phenotype of mdm10Δ cells (supplementary material Fig. S2). Of note, overexpression of both Mcp1 and Mcp2 failed to rescue the mdm10Δ growth defect at extreme conditions such as high temperature and non-fermentable carbon source (37°C and YPG medium; supplementary material Fig. S1A).

To test whether the rescue effects of Mcp1 and Mcp2 are additive we co-overexpressed both proteins in the mdm10Δ strain. However, overexpression of both proteins did not result in any additional rescue effects beyond those observed with Mcp1 alone (supplementary material Fig. S3).

Mdm10 is part of the ERMES complex that contains also the mitochondrial proteins Mdm12 and Mmm2 as well as the ER protein Mmm1. In contrast to the suppression capacity of Mcp1 and Mcp2, overexpression of the other ERMES components Mmm1, Mmm2 or Mdm12 did not rescue the growth phenotype of mdm10Δ cells (supplementary material Fig. S4). These results suggest that the primary function of Mcp1 and Mcp2 differs from those of the ERMES components.

We were next interested in the subcellular localization of Mcp1 and Mcp2. Previous high-throughput studies suggest that both are mitochondrial proteins (Huh et al., 2003; Reinders et al., 2006; Zahedi et al., 2006). However, we wanted to study the localization and topology of both proteins on a single gene level. To this end we expressed C-terminally HA-tagged plasmid-borne constructs of both proteins in the corresponding deletion strain as well as a genomically FLAG-tagged version of Mcp2. The HA-tagged proteins were confirmed to be functional as their overexpression complemented the mdm10Δ growth phenotype (supplementary material Fig. S5).

Cells expressing either Mcp1–HA or Mcp2–HA were fractionated by differential centrifugation into mitochondrial, microsomal and cytosolic fractions. Both Mcp1–HA and Mcp2–HA were present in the mitochondrial fraction (Fig. 2A,B). Similar behaviour was observed with a genomically FLAG-tagged version of Mcp2 (supplementary material Fig. S6A). Mcp1 is a protein of 303 amino acid residues, with five predicted α-helical transmembrane domains (TMDs) (http://www.cbs.dtu.dk/services/TMHMM/). Homologue proteins of Mcp1 could be found only in fungi (http://blast.ncbi.nlm.nih.gov/Blast.cgi). The primary sequence of Mcp2 comprises 569 amino acids, and in contrast to Mcp1, a homologue could be found in all higher eukaryotes. The human orthologue is called aarF domain containing kinase 1 (ADCK-1). Comparison of the sequences of Mcp2 and ADCK-1 shows 33% identity with coverage of 92% (http://blast.ncbi.nlm.nih.gov/Blast.cgi). Several prediction programs suggest at least one TMD within the primary sequence of Mcp2 (Biegert et al., 2006).

To study, whether Mcp1 and Mcp2 are indeed integral membrane proteins, we subjected mitochondria containing Mcp1–HA or Mcp2–HA to an alkaline extraction treatment. Under these conditions, soluble and peripheral membrane proteins can be separated by centrifugation from integral membrane proteins. Mcp1–HA (Fig. 2C), Mcp2–HA (Fig. 2D), and the genomically tagged Mcp2–FLAG (supplementary material Fig. S6B) were found, like the integral outer membrane protein Tom70 or Porin, in the pellet fraction, suggesting that both Mcps are integral membrane proteins.

Since mitochondria have two distinct membranes, we wanted to know in which one Mcp1 and Mcp2 are located. To that end, we treated mitochondria containing HA-tagged Mcps with proteinase K (PK) under different conditions. Mcp1–HA full length protein was no longer detected after the proteolytic treatment of intact organelles and instead, two fragments of around 15 kDa were observed (Fig. 2E). As expected for a MOM protein, the amount of these fragments was reduced if mitochondria were swollen and the outer membrane was ruptured before protease treatment. Furthermore, solubilisation of the membrane by the detergent Triton X-100 exposed Mcp1 to complete degradation (Fig. 2E). Hence, we propose that Mcp1 is a MOM protein, with its C-terminus facing the intermembrane space (IMS). This proposal is in agreement with previous identification of Mcp1 in the MOM proteome (Zahedi et al., 2006). In contrast to Mcp1, Mcp2–HA was protected from the protease as long as the MOM was intact. If the outer membrane was ruptured by hypo-osmolar treatment, the protein was partially degraded by the protease (Fig. 2F). These results suggest that Mcp2 is a mitochondrial membrane protein exposed to the IMS located either in the inner or outer membrane. To discriminate between these two possibilities, we separated inner and outer membrane mitochondrial vesicles by density gradient centrifugation. Analysis of the different fractions of the gradient showed that Mcp2–HA migrated similar to the inner membrane protein Cox2, whereas as expected, the integral MOM protein Tom20 was found in fractions with lower density (Fig. 2G). These findings suggest that Mcp2 is an integral protein of the mitochondrial inner membrane.

To better understand the importance of the two novel proteins for cellular functions, we investigated whether the deletion of Mcp1, Mcp2 or both results in growth phenotypes of yeast cells. All three deletion mutants grew similarly to wild-type cells under all tested conditions (Fig. 3A and data not shown). We next tested the outcome of overexpression of either Mcp1 or Mcp2 and did not observe any effect on the growth behaviour of wild-type yeast cells (Fig. 3B and data not shown). Taken together it appears that both proteins are not essential for optimal cellular growth under the investigated conditions.

To decipher the molecular function of Mcp1 and Mcp2 we first analysed mitochondrial morphology upon their deletion. To visualize the mitochondria, we transformed the cells of interest with a plasmid encoding matrix targeted GFP (mtGFP) (Westermann and Neupert, 2000). Cells lacking Mcp1 or Mcp2 did not show alterations in mitochondrial morphology (Fig. 4A,B). We further asked whether overexpression of Mcp1 or Mcp2 could restore the mdm10Δ morphology phenotype. To this end we co-transformed the mdm10Δ strain with an mtRFP plasmid together with overexpression plasmids encoding either Mcp1 or Mcp2. Overexpression of Mcp1 in the mdm10Δ strain resulted in a rescue of the morphology phenotype to the same extent as re-introducing Mdm10 itself (Fig. 4C,D). In contrast, only about 20% of the mdm10Δ cells overexpressing Mcp2 harboured restored tubular networks and about 40% of the cells exhibited a variable morphology phenotype ranging from elongated condensed structures to short tubules (Fig. 4C,D). To better understand the role of Mcp1 and Mcp2 in mitochondrial morphology we overexpressed the proteins also in wild-type cells. Increased levels of Mcp1 had no effect on the organelle's morphology, yet overexpression of Mcp2 resulted in shortened, condensed or aggregated mitochondria in some of the cells (Fig. 4E,F). Collectively, it appears that elevated levels of Mcp2 affect the morphogenesis of mitochondria.

It was reported that mitochondria lacking Mdm10 show also a mild reduction in the levels of subunit Tom22 of the translocase of the outer membrane (TOM) complex (Meisinger et al., 2004). We wanted to know whether the observed growth rescue by overexpression of Mcp1 or Mcp2 could be due to restored Tom22 levels or due to elevated levels of other mitochondrial proteins. To address this question, we isolated mitochondria from wild-type or mdm10Δ cells bearing an empty plasmid or from mdm10Δ cells overexpressing Mcp1, Mcp2 or Mdm10 itself. In contrast to the report above we were not able to detect reduced levels of Tom22 in mitochondria lacking Mdm10. Unaffected Tom22 levels were observed in both crude mitochondria (Fig. 5A) and mitochondria purified by a Percoll gradient (data not shown). A potential explanation for this discrepancy is the different genetic background of the deletion strains. Whereas we used the wild-type background W303, Meisinger et al. (Meisinger et al., 2004) utilized the BY4741/2/3 background. Indeed, it was recently reported that steady state levels of TOM and TOB components vary between mdm10Δ mutants of different genetic backgrounds (Yamano et al., 2010b). Nonetheless, we did not observe substantial changes in the steady state levels of the other TOM components Tom20, Tom40 or Tom70 as well as of the TOB subunits Mas37 and Tob55 upon overexpression of Mcp1 or Mcp2 in mdm10Δ cells or in cells lacking Mdm10 (Fig. 5A and data not shown). Taken together, it seems that the suppression capacity of Mcp1 and Mcp2 is not related to alteration in the levels of import components of the MOM.

A major consequence of loss of Mdm10 is an altered assembly of TOM and TOB complexes as reflected by blue native PAGE (BN-PAGE) analysis (Meisinger et al., 2004; Yamano et al., 2010b). Absence of Mdm10 leads to reduced levels of the assembled TOM complex and to an additional unassembled species of Tom40 of about 100 kDa (Fig. 5B, second lane) (Meisinger et al., 2004; Yamano et al., 2010b). Upon re-introduction of Mdm10 in mdm10Δ cells this lower molecular mass complex is no longer observable. In contrast, upon overexpression of Mcp1 or Mcp2 the unassembled species is still visible and TOM complex levels are mildly reduced (Fig. 5B). We therefore conclude that the rescue by Mcp1 and Mcp2 is not derived by an effect on the stability of the TOM complex.

We next wondered whether Mcp1 and Mcp2 play a role in the efficiency of TOM complex assembly. Previous reports showed that in the absence of Mdm10 the typical assembly pathway of radiolabeled newly imported Tom40 via the 250 kDa intermediate I and the 100 kDa intermediate II is hampered (Model et al., 2001; Meisinger et al., 2004; Yamano et al., 2010b). We made the same observation in mdm10Δ mitochondria isolated from the W303 genetic background (Fig. 5C). In the absence of Mdm10 only a minor portion of the radiolabeled Tom40 molecules became assembled after 90 minutes of in vitro import (Fig. 5C). To test whether Mcp1 or Mcp2 could restore this defect, we used mitochondria isolated from an mdm10Δ strain overexpressing Mcp1 or Mcp2. Although we could see a slight increase in the amount of assembly intermediates I and II upon overexpression of either Mcp1 or Mcp2, this overexpression did not affect the amount of assembled TOM complex (Fig. 5C). We therefore propose that Mcp1 and Mcp2 do not take over the role of Mdm10 in the assembly of functional TOM holo-complex. The increase in assembly intermediates could be a secondary effect of the not completely restored mitochondrial lipid composition (see below).

Mitochondria lacking Mdm10 show also a defect in the in vitro import of Tom22 and a 400 kDa assembly intermediate is observed under these conditions (Meisinger et al., 2004). Mitochondria lacking Mdm10 with elevated levels of either Mcp1 or Mcp2 exhibited only a slight restoration, if at all, in this assembly defect (Fig. 5D). Of note, we observed an increase in the overall amount of newly synthesized Tom22 molecules incorporated into mitochondria lacking Mdm10 and overexpressing Mcp2. However, we could not find any support for a direct role of Mcp2 in the import of Tom22. These results substantiate our aforementioned findings that Mcp1 and Mcp2 have no direct function in the assembly of the TOM complex.

It is well documented that alterations in Mdm10 levels affect the amounts of various forms of the TOB complex (Thornton et al., 2010; Yamano et al., 2010b; Yamano et al., 2010a; Becker et al., 2011). In agreement with these reports, we observed three different TOB-complex species (Fig. 5E). In wild-type cells we observed the TOB core complex (containing Tob55, Tob38 and Mas37) as the main form. Upon loss of Mdm10, a larger complex became visible (Fig. 5E, second lane, asterisk). This complex resembles the reported TOB–Tom5/Tom40 species (Thornton et al., 2010). Overexpression of Mdm10 led to an additional TOB complex form that migrates like the previously reported TOB–Mdm10 holo-complex (Meisinger et al., 2004; Yamano et al., 2010b). Since upon overexpression of Mcp1 and Mcp2 we still observed the same TOB complex pattern as for mdm10Δ cells, we propose that Mcp1 and Mcp2 are not involved in the biogenesis and/or assembly of the TOB complex.

We next asked whether Mcp1 or Mcp2 are functionally related to the other subunits of the ERMES complex. To study the genetic interaction of Mcp1 or Mcp2 with other ERMES components we overexpressed the Mdm10 supressors in mmm1Δ, mmm2Δ or mdm12Δ cells. Mcp1 and Mcp2 could rescue (completely or partially) the loss of each of the ERMES subunits (see representative clones in Fig. 6A–C). To substantiate these results, we tested at least three different clones of each rescue transformation and compared them to two transformants harbouring the empty plasmid (supplementary material Fig. S7 and data not shown). Interestingly Mcp2 is a better suppressor for loss of Mdm12 whereas Mcp1 is more efficient in suppressing the mmm1Δ and mmm2Δ phenotypes. These results show that Mcp1 and Mcp2 have distinct functions and both can compensate for the loss of functional ER–mitochondria tethering complex ERMES.

The respiratory chain complexes form supercomplexes that can be assessed by BN-PAGE and Coomassie staining and/or immunodecoration (Schägger, 2001). We realized that the supercomplex pattern of mitochondria lacking Mdm10 is altered in comparison to wild-type organelles and this assembly phenotype can be reversed by overexpression of Mdm10 itself, Mcp1 or Mcp2 (Fig. 7A). To verify the identity of these supercomplexes we immunodecorated mitochondrial proteins separated by BN-PAGE with antibodies specific for subunits of complex III (Cor1) and complex IV (Cox2) of the oxidative phosphorylation (OXPHOS) pathway. Indeed, we observed a dramatic change of OXPHOS complex and supercomplex patterns in mdm10Δ cells (Fig. 7B,C). Upon immunodecoration with an antibody specific for Cor1, we observed an increase in the amount of the complex III dimer (III₂) and a strong decrease of the complex III/IV hetero-tetramer and trimer (III₂IV₂, III₂IV). In addition, we detected an additional band of around 750 kDa of unknown composition (Fig. 7B, arrowhead). Overexpression of Mcp1, similarly to the re-introduction of Mdm10 itself, lead to a complete rescue of the migration pattern (Fig. 7B). Overexpression of Mcp2 resulted in a partial complementation of the defect in the formation of III/IV supercomplexes. Although the upper band of the III₂IV hetero-trimer and the tetramer III₂IV₂ were restored, the lower trimer band was missing and the additional band at 750 kD was still visible (Fig. 7B, last lane). Similar complementation capacities were observed upon immunodecoration with an antibody against the complex IV subunit Cox2. There is an additional band at 750 kDa that remains upon Mcp2 overexpression (Fig. 7C arrowhead). Yet the trimer and the tetramer are restored in case of the overexpression of Mcp1, Mcp2 and as a control Mdm10 (Fig. 7C). Of note the steady-state levels of the mitochondrially encoded Cox2 are slightly decreased in mdm10Δ cells (Fig. 7D). This is probably due to either the reported alterations in mitochondrial DNA and nucleoid phenotype in mdm10Δ cells (Sogo and Yaffe, 1994) or the reduced stability of the protein in the membrane. This reduction in the steady-state levels could be restored by overexpression of Mdm10 and its two suppressors (Fig. 7D). Taken together, overexpression of Mcp1 or Mcp2 rescued defects in respiratory chain protein complex assembly derived by the loss of Mdm10.

It was previously shown that the lipid composition of mitochondria and especially cardiolipin levels affect the biogenesis and function of protein complexes in mitochondrial membranes (Jiang et al., 2000; Schägger, 2002; Zhang et al., 2002; Pfeiffer et al., 2003; Zhong et al., 2004; Nury et al., 2005; Zhang et al., 2005; van der Laan et al., 2007; Claypool et al., 2008; Kutik et al., 2008). Furthermore, there are several reports that loss of Mdm10 and defects in the ERMES complex result in an altered phospholipid composition of mitochondria (Kornmann et al., 2009; Osman et al., 2009; Yamano et al., 2010b; Nguyen et al., 2012; Tamura et al., 2012a). We wondered whether the rescue by Mcp1 and Mcp2 overexpression can be explained by changes in the lipid contents of mitochondria. To this end we analysed by mass spectrometry the lipid composition of mitochondria isolated from the strains of interest.

As reported before, the loss of Mdm10 led to a change in phospholipid composition of mitochondria. Especially phosphatidylethanolamine (PE) and cardiolipin (CL) levels were reduced, whereas the relative amounts of phosphatidylserine (PS) were increased (Fig. 8A; supplementary material Table S1) (Kornmann et al., 2009; Osman et al., 2009; Yamano et al., 2010b; Nguyen et al., 2012; Tamura et al., 2012a). We also analysed the different species of the single phospholipids upon deletion of MDM10 and found an increase of the 34∶2 species of PC, PE, PS and PI (supplementary material Table S2). Mcp1 overexpression could compensate for reduction of PE and increase in PS levels in mdm10Δ cells. Yet CL levels were only slightly restored. The effect of Mcp2 overexpression was less pronounced, still PE levels were back to wild-type values, CL levels were moderately increased and PS levels decreased in comparison to mdm10Δ cells (Fig. 8A; supplementary material Table S1). The fact that overexpression of Mcp1 leads to a better restoration of regular lipid composition is in agreement with the higher rescue capacity of this suppressor.

The deletion of MCP1 or MCP2 led only to minor changes in phospholipid composition of mitochondria. Most obvious is a minor statistically insignificant reduction of phosphatidylcholine (PC) in the case of mcp1Δ. Furthermore a very minor reduction in phosphatidylinositol (PI) amounts for mcp2Δ is visible (Fig. 8A; supplementary material Table S1).

Taken together, overexpression of Mcp1 and Mcp2 in mdm10Δ cells results in partial restoration of the altered phospholipid levels of these cells.

Changes in mitochondrial phospholipid composition connected to the loss of ERMES function were reported before (Kornmann et al., 2009; Osman et al., 2009; Yamano et al., 2010b; Tamura et al., 2012a). Yet, an open question is whether Mdm10 is also involved in the regulation of ergosterol trafficking into and out of mitochondria. In this context we analysed the levels of ergosterol in mitochondria isolated from cells with altered levels of Mdm10, Mcp1 and Mcp2. The deletion of MDM10 resulted in a twofold increase of the mitochondrial ergosterol levels. Of note, this increment was reversed by the overexpression of either Mdm10 itself or Mcp1 but not by overexpression of Mcp2. The deletion of each MCP did not alter the amounts of ergosterol (Fig. 8B; supplementary material Table S1). We excluded the possibility that the restored ergosterol levels upon overexpression of Mcp1 are related to lower ER contamination in the purified mitochondria samples by immunodecoration of these samples with an antibody against the ER marker, Erv2 (supplementary material Fig. S8).

We next tested whether other ERMES components are also involved in ergosterol homeostasis. To this end we compared the ergosterol content of mitochondria isolated from mdm10Δ, mdm12Δ and mmm1Δ cells grown on rich medium to that in organelles from wild-type cells. As shown in Fig. 8C, loss of Mdm10, Mdm12 or Mmm1 leads to an ∼50% increase of ergosterol in comparison to wild-type mitochondria. Hence, these results indicate an important novel role for Mdm10 and the ERMES complex in the homeostasis of ergosterol in mitochondria, and might suggest a role for Mcp1 in this process.

The aforementioned functional links of the Mcp proteins to ERMES components lead us to investigate whether the deletion of MCP1 or MCP2 shows genetic interactions with deletions of the ERMES components. To this end we mated haploid W303 cells lacking Mcp1 or Mcp2 (mcp1Δ or mcp2Δ) with cells lacking one of the ERMES components (mmm1Δ, mmm2Δ, mdm10Δ or mdm12Δ) of the opposing mating type. After selection for the heterologous diploid double deletion strains W303a/α and sporulation we performed tetrad dissection. The summary of the results is shown in supplementary material Fig. S9A; mcp1Δ is synthetic lethal with mdm10Δ, mmm2Δ and the deletion of the ER protein encoding MMM1. Similarly, deletion of MCP2 together with MMM1 leads to inviable spores. supplementary material Fig. S9B illustrates the results of genetic lethal interactions. The arrowheads point to barely visible microcolonies identified by marker analysis of the other spores that did not grow further upon re-streaking on a novel plate. In those cases where the double deletion strains were viable (deletion of MCP1 together with MDM12 and MCP2 with MDM10, MDM12 or MMM2) we observed a severe negative synthetic growth defect most pronounced at higher temperatures (supplementary material Fig. S9C). Taken together, these results show a strong negative genetic interaction between either MCP1 or MCP2 with any of the ERMES components.

Mdm10 is a β-barrel protein that was shown to be involved in mitochondrial core functions like morphogenesis, inheritance, cytoskeleton interaction, biogenesis of β-barrel proteins, mitochondria–ER contact, and lipid transport. However, many of the aforementioned processes are inter-connected. For example, a decreased assembly of MOM import complexes can lead to reduced import of mitochondrial morphology proteins, which in turn would lead to a morphology phenotype. Along the same line, a compromised mitochondria–ER contact can result in hampered lipid flow between the involved compartments, changes of the lipid composition of their membranes and subsequently alteration in assembly of membrane-embedded complexes and/or in fusion of mitochondrial membranes. An altered mitochondrial lipid composition in the absence of Mdm10 is observed in many studies (Kornmann et al., 2009; Osman et al., 2009; Yamano et al., 2010b; Osman et al., 2011; Nguyen et al., 2012). Accordingly, Mdm10 was suggested to be involved in lipid homeostasis of mitochondria. Considering all these possibilities, the identification of the primary function of Mdm10 remains a central challenge.

In the present work we identified two novel mitochondrial proteins, which we named Mcp1 and Mcp2, as high-copy suppressors of the growth phenotype of mdm10Δ cells. Both proteins also suppress phenotypes that are related to altered lipid homeostasis in mdm10Δ cells but only marginally affect import defects in these cells. Thus, our findings suggest that the prominent function of Mdm10 is its contribution to lipid transport. Mcp1 and Mcp2 rescue OXPHOS supercomplex formation that is disturbed in mdm10Δ mitochondria and partially restore the lipid profile of the organelle. Supercomplex formation is restored almost completely, yet cardiolipin levels are only moderately rescued upon overexpression of Mcp1. On the other hand, PE quantities are even elevated in these mitochondria in comparison to organelles from wild-type cells. We propose that a certain threshold amount of non-bilayer forming lipids (PE and CL) is required for the proper formation of the OXPHOS supercomplexes. These two lipids can compensate for one another as also indicated by the observation that yeast cells cannot tolerate simultaneous reduction of CL and PE (Osman et al., 2011). Such a threshold is apparently not achieved in cells lacking Mdm10. The overexpression of Mcp1 and to a lower extent of Mcp2 restores the levels of CL and PE to a point above this critical amount, therefore restoring supercomplex formation. The recent identification of Mdm31 as another suppressor of the growth phenotype of mdm10Δ mutation supports this conclusion, as overexpression of Mdm31 in cells lacking Mdm10 restores the cardiolipin levels in these cells (Tamura et al., 2012a). In contrast, overexpression of components of the TOB or TOM complexes did not result in a rescue of the growth phenotype of mdm10Δ cells (data not shown). Of note, Mdm10 appears to have a unique role within the ERMES complex as our results indicate that overexpression of none of the other subunits of this complex could compensate for the absence of Mdm10.

How do Mcp1 and Mcp2 influence the lipid composition of mitochondria? Deletion of both genes has only minor influence on lipid levels. The importance of the two suppressors becomes obvious in the absence of a functional ERMES complex. Mcp1 and Mcp2 overexpression partially rescue the absence of each component of this ER–mitochondria connecting structure. Of note, despite various pull-down assays we were not able to obtain any evidence of a physical interaction of Mcp1 or Mcp2 with any of the ERMES complex subunits (data not shown). Supporting our findings are two recent studies set out to investigate the composition of the ERMES complex that have not identified other components of this complex besides Mdm10, Mdm12, Mmm1, Mmm2 and Gem1 (Kornmann et al., 2011; Stroud et al., 2011). Hence, we suggest that the MCP proteins are not associated with the ERMES complex but rather can provide an alternative pathway for lipid flow that ‘bypasses’ the requirement for the ERMES complex. The location of Mcp1 in the MOM positions it in an optimal place to facilitate such an exchange. A tempting yet speculative model would suggest an additional mitochondria–ER contact that functions as a back-up pathway for the known ERMES mediated inter-organellar lipid exchange. It is known that in higher eukaryotes apparently other mediators (like MFN2 and VDAC) contribute to ER–mitochondria association (for reviews, see de Brito and Scorrano, 2010; Patergnani et al., 2011; Toulmay and Prinz, 2011). Indeed it was reported that although a defective ERMES complex impairs phospholipid exchange between ER and mitochondria it does not abolish it (Kornmann et al., 2009). Recent evidence even suggests that phosphatidylserine transport from the ER to mitochondria is not directly influenced by the ERMES complex (Nguyen et al., 2012), ruling out ERMES as the exclusive mediator of lipid exchange.

Further supporting such a scenario are the synthetic lethality or severe synthetic growth defects of double deletion strains of mcp1Δ and mcp2Δ in combination with a deletion of any of the four classical ERMES complex subunits. These genetic interactions substantiate the hypothesis that Mcp1 and Mcp2 are part of a ‘bypass route’ for lipids between the two compartments that can partially compensate for the major ERMES-facilitated pathway. The loss of both lipid exchange pathways would lead to a complete block of lipid exchange and thus to a lethal outcome. Such an alternative lipid transport pathway does not necessarily have to be between ER and mitochondria. For instance, it has been reported that a functional connection between mitochondria and the vacuole is mediated by the cardiolipin biosynthesis pathway (Chen et al., 2008), and a contact between mitochondria and lipid particles was also reported (Beller et al., 2010).

Because Mcp1 and Mcp2 play a role in lipid homeostasis of mitochondria, we wondered whether Mcp1 and Mcp2 are also suppressors of other genes known to be involved in lipid composition of the mitochondrial membranes. Ups1 and Ups2 are two proteins that were recently shown to be involved in the intra-mitochondrial lipid transport (Connerth et al., 2012; Tamura et al., 2012a). We overexpressed Mcp1 and Mcp2 in cells lacking Ups1 or Ups2 and did not observe any rescue effect (data not shown). These results support a role for the Mcps in global mitochondrial lipid transport rather than in intra-mitochondrial membrane lipid exchange.

Very recently, Mdm31 was identified as another high copy suppressor of mdm10Δ (Tamura et al., 2012a). Mdm31 was previously shown to be involved in mitochondrial morphogenesis, inheritance (Dimmer et al., 2005) and ion homeostasis (Kucejova et al., 2005). However, the precise molecular function of this protein is currently unknown. Although we screened about 30,000 colonies (about five times the number of genes of S. cerevisiae), Mdm31 was not amongst our candidates. Since it was previously reported that various phenotypes of ERMES mutants vary according to the genetic background (Yamano et al., 2010b), we suppose that our usage of another strain as compared with the one employed by Tamura et al. (Tamura et al., 2012a) can explain why Mdm31 was not identified in the present study.

Our findings provide for the first time evidence that, apart from the phospholipid composition of mitochondrial membranes, also ergosterol levels are affected by the ERMES complex. This observation has important implications for the potential influence of the ERMES complex on the biophysical properties of mitochondrial membranes. A higher ergosterol level, as observed in the mdm10Δ, mmm1Δ and mdm12Δ cells, results in more rigid membranes which would behave differently in respect to organelle morphology and membrane protein insertion. Similarly, it has been shown that downregulation of genes involved in ergosterol biosynthesis leads to a severe mitochondrial morphology phenotype (Altmann and Westermann, 2005), suggesting an important role of ergosterol in overall organelle morphogenesis. Furthermore, the proper membrane integration of some single-span mitochondrial MOM proteins requires a low ergosterol content of the membrane (Kemper et al., 2008; Krumpe et al., 2012; Merklinger et al., 2012). Moreover, it is tempting to assume that the function of protein translocases and especially insertases depends on the biophysical properties of the membrane they are embedded in and therefore on the lipid composition of these membranes. Indeed it has been shown, that the assembly of protein import complexes TOM, TIM22 and TIM23 depend on cardiolipin (Kutik et al., 2008; Gebert et al., 2009). Collectively, our observation that deletion of Mdm10 causes alteration in ergosterol content can help to explain part of the phenotypes of the mdm10Δ mutant.

Taken together, this study provides new insights into the regulation of the lipid composition of mitochondrial membranes. We show that Mdm10 plays a crucial role in ergosterol homeostasis of the organelle and two novel proteins Mcp1 and Mcp2 that are involved in mitochondrial lipid homeostasis were identified. These two proteins might be involved in an ERMES-independent pathway of lipid exchange to and from mitochondria.

Yeast strains were grown in standard rich medium with either glucose (YPD) or glycerol (YPG), synthetic medium with either glucose (SD) or glycerol (SG), or lactate medium. For drop dilution assays, cells were cultured to an OD₆₀₀ of 1.0 and diluted in fivefold increment followed by spotting 5 µl of each cell suspension on the corresponding solid medium.

Standard genetic techniques were applied for the growth and manipulation of Saccharomyces cerevisiae cells. The strains used were isogenic to W303a and W303α. To delete complete open reading frames by homologous recombination the HIS3MX6 cassette was amplified from the plasmid pFA6a-HIS3MX6 (Wach et al., 1997) and the KanMX4 cassette from pFA6a-KanMX4 (Wach et al., 1994) with gene-specific primers. Primer sequences are listed in supplementary material Table S3. Double-deletion strains were obtained by tetrad dissection. All deletion strains were confirmed by genome based PCR with gene specific primers. Transformation of yeast strains was performed by the lithium acetate method.

The open reading frames of MDM10, MMM1, MMM2, MDM12, MCP1 and MCP2 genes were amplified by PCR from yeast genomic DNA and cloned into the plasmid pYX142 that encodes an additional HA tag sequence. By omission of the stop codon of the ORFs, a plasmid-borne HA tag was introduced C-terminally to the coding sequence. All constructs were analysed by sequencing. Primer sequences are listed in supplementary material Table S3.

Cells lacking Mdm10 (mdm10Δ) were transformed with a high-copy-number yeast genomic library constructed by F. Lacroute in the URA3 2 µ plasmid pFL44L (Stettler et al., 1993). Clones were initially selected at 30°C on synthetic medium lacking uracil. Then clones were replica plated on glycerol-containing full medium and further selected at 37°C. Plasmid DNA was extracted from the clones grown at 37°C and transformed into E. coli cells. The DNA inserts of plasmids isolated from the transformed E. coli cells were analysed by sequencing.

Subcellular fractionation was performed according to published procedures (Walther et al., 2009). Mitochondria of yeast cells were isolated by differential centrifugation as described previously (Daum et al., 1982). Further purification of mitochondria was achieved via a self-generated percoll or sucrose step gradients (Graham, 2001). For swelling experiments, mitochondria were incubated in hypotonic buffer (20 mM Hepes, pH 7.2) for 30 minutes on ice. Proteins were retrieved by trichloroacetic acid (TCA) precipitation and solubilized in sample buffer. Mitochondrial MOM and MIM vesicles were generated and separated as described (Tamura et al., 2012b). After resuspension of the vesicles, sucrose concentration was adjusted to 0.85 M. A sucrose step gradient (0.9/1.0/1.1/1.2/1.3 M) was applied for separation and centrifugation was performed for 16 hours (200,000 g, 4°C) in a Beckman SW60 rotor. Protein samples were analysed by SDS-PAGE and immunoblotting using the ECL system.

Mitochondria were lysed in 40 µl buffer containing either Triton X-100 or digitonin [0.5% Triton X-100 or 1–1.5% digitonin, 20 mM Tris-HCl, 0.1 mM EDTA, 50 mM NaCl, 10% (v/v) glycerol, 1 mM PMSF, pH 7.2]. After incubation on ice for 15 min and clarifying spin (30,000 g, 15 minutes, 2°C), 5 µl sample buffer [5% (w/v) Coomassie blue G, 500 mM 6-amino-N-caproic acid, 100 mM Bis-Tris, pH 7.0] was added, and the mixture was analysed by electrophoresis in a 6–13% or 4–8% gradient blue native gel (Schägger, 2002). Gels were blotted on polyvinylidene fluoride membranes and proteins were further analysed by autoradiography or immunodecoration.

Import experiments were carried out with radiolabeled precursor proteins and isolated mitochondria in import buffer (250 mM sucrose, 0.25 mg/ml BSA, 80 mM KCl, 5 mM MgCl₂, 10 mM MOPS-KOH, 2 mM NADH, and 2 mM ATP, pH 7.2). Radiolabeled precursor proteins were synthesized in the presence of [³⁵S]methionine in rabbit reticulocyte lysate (Promega, Mannheim, Germany) according to the manufacturer's protocol. Import reactions were performed for various time periods at 25°C. At the end of the import reaction, mitochondria were re-isolated by centrifugation (13,200 g, 10 minutes, 2°C) and then analysed by SDS- or BN-PAGE.

For visualization of mitochondria, yeast cells were transformed with an expression vector harbouring the mitochondrial presequence of subunit 9 of Fo-ATPase of Neurospora crassa fused to GFP or RFP (Mozdy et al., 2000; Westermann and Neupert, 2000). Microscopy images were acquired with an Axioskop20 fluorescence microscope equipped with an Axiocam MRm camera using the 43 Cy3 and 38 Endow GFP filter sets and the AxioVision software (Zeiss, Göttingen, Germany). For statistical analysis, yeast mitochondria were also visualized by staining with the mitochondria-specific dye rhodamine B hexyl ester (Life Technologies GmbH, Darmstadt, Germany).

Cells at the log phase (OD₆₀₀<2.0) grown on selective medium containing galactose were harvested and mitochondria were isolated first by differential centrifugation and then crude organelles were purified further by a self-generated percoll gradient or sucrose step gradient (see above). Mass spectrometry analysis was performed in positive ion mode on a triple quadrupole-linear ion trap hybrid mass spectrometer (QTRAP 5500, AB Siex), except for analysis of cardiolipin and ergosterol, which was done in negative ion mode on a quadrupole time-of-flight mass spectrometer (QStar Elite, AB Siex). Lipid extractions (total lipid amount of 1.5–2.5 nmol) were performed in the presence of internal lipid standards [PC, PE, PS with species 14∶1/14∶20∶1/20∶1 and 22∶1/22∶1), PI (37∶4), CL 56∶0] (Lorizate et al., 2013). Dried lipids were dissolved in 10 mM ammonium acetate.

Quantification of PC was performed by precursor ion scanning, selecting for a fragment ion of 184 Da. PE, PS and PI were quantified by neutral loss scanning, selecting for a neutral loss of 141, 185 and 277 Da, respectively. For CL quantification, 10 µl of lipid extracts was diluted 1∶2 with 0.1% piperidine in methanol and automatically infused (Triversa Nanomate, Advion Biosciences) (Osman et al., 2009). CLs were detected as single charged molecules. CL species were analysed by targeted product ion scanning. The peak areas of CL-derived fatty acid fragments were extracted from the respective product ion spectra via the ‘Extract Fragments’ script (Analyst QS 2.0). Isotope correction for M+2 ions was done manually. Ergosterol quantification was performed as described (Ejsing et al., 2009). Lipid species identification and data evaluation were performed with the LipidView software (ABSciex).

We thank K. Rehn and E. Kracker for technical support, N. Bonnefoy for the yeast genomic library, D. Papić and F. Essmann for critically reading the manuscript as well as C. Ernst and A. Peschel for help with the lipid analysis. B.B. is an investigator of the Cell Networks Cluster of Excellence (EXC81).