Biochemical Society Focused Meetings

The ERMES complex and ER–mitochondria connections

Agnès H. Michel, Benoît Kornmann


Cellular organelles need to communicate in order to co-ordinate homoeostasis of the compartmentalized eukaryotic cell. Such communication involves the formation of membrane contact sites between adjacent organelles, allowing privileged exchange of metabolites and information. Using a synthetic protein designed to artificially tether the ER (endoplasmic reticulum) to mitochondria, we have discovered a yeast protein complex naturally involved in establishing and maintaining contact sites between these two organelles. This protein complex is physiologically involved in a plethora of mitochondrial processes, suggesting that ER–mitochondria connections play a central co-ordinating role in the regulation of mitochondrial biology. Recent biochemical characterization of this protein complex led to the discovery that GTPases of the Miro family are part of ER–mitochondria connections. The yeast Miro GTPase Gem1 localizes to ER–mitochondria interface and influences the size and distribution of mitochondria. Thus Miro GTPases may serve as regulators of the ER–mitochondria connection.

  • cardiolipin
  • cytosol
  • endoplasmic reticulum (ER)
  • endoplasmic reticulum–mitochondria encounter structure (ERMES)
  • lipid binding
  • mitochondrion

ER (endoplasmic reticulum)–mitochondria contact sites

The eukaryotic cell is a highly complex entity that carries out many distinct functions that are interdependent and interconnected. Organelles are compartments dedicated to the realization of specific biochemical tasks that require an appropriate and controlled milieu. While the compartmentalization of the eukaryotic cell into organelles ensures that incompatible biochemical pathways remain separated, it also creates the need for communication routes that allow organelles to exchange the information and metabolites required to fulfil their function [1,2].

The ER and the mitochondria are two organelles that undergo this type of privileged communication. The membrane of both organelles can often be observed in close apposition by electron microscopy. Moreover, isolated subcellular mitochondrial fractions contain contaminating ER membranes [3]. These contaminating ER membranes have a molecular composition that is slightly different from that of the general ER, suggesting that these mitochondria-associated membranes represent a laterally differentiated subcompartment of the ER, which is physically attached to the mitochondria [4].

The physical connection between the ER and mitochondria might serve different physiological purposes.

The first one is lipid exchange. Mitochondria are enclosed by a double membrane, the OMM (outer mitochondrial membrane) and the IMM (inner mitochondrial membrane). Therefore the mitochondrion is a membrane-rich organelle. Most phospholipids are synthesized in the ER and distributed to other organelles via vesicular transport. Mitochondria, however, are not part of the endomembrane system and therefore do not receive phospholipids from the ER via vesicular trafficking. ER–mitochondria contact sites have been proposed to favour lipid exchange between both organelles [4]. Lipid exchange is not only necessary for the biogenesis of mitochondrial membranes but also for general lipid synthesis. The decarboxylation of PS (phosphatidylserine) to PE (phosphatidylethanolamine) is carried out by an enzyme, Psd1 (phosphatidylserine decarboxylase 1), that resides in the mitochondria, implying that the substrate and product of the reaction must respectively leave the ER to reach the mitochondria, and go back to the ER to be further distributed to other cellular membranes [5].

Another proposed role for the ER–mitochondria connection is to promote interorganelle calcium (Ca2+) exchange. When Ca2+ is released from the ER by the activation of the IP3 (inositol trisphosphate) receptor, an increase in Ca2+ concentration ([Ca2+]) can be observed both in the cytosol and in mitochondria [6]. This is surprising because mitochondria in isolation are only weakly able to take up Ca2+, even at the maximum concentrations that can be attained in the cytosol. This is inconsistent with the idea that the increase in cytosolic [Ca2+] is sufficient to cause an increase in mitochondrial [Ca2+]. A model to explain this discrepancy proposes that the close apposition of the ER and the mitochondria creates micro-domains in which [Ca2+] reaches much higher levels than in the rest of the cytosol, akin to the high neurotransmitter concentration that can be found at neurological synapses [6].

Many mitochondrial enzymes are Ca2+-dependent, therefore Ca2+ signalling within the mitochondria is critical for the generation of ATP. Moreover, mitochondrial Ca2+ participates in triggering apoptosis, indicating that mitochondrial Ca2+ must be tightly regulated [7].

For a long time, proteinaceous factors involved in tethering the ER and the mitochondria resisted identification. In recent years, however, many proteins have been shown to be involved in ER–mitochondria tethering. These include the ER-resident Ca2+ channel IP3 receptor, the mitochondrial voltage-dependent anion channel [8], the chaperones grp75 and sigma-1R [9], the sorting protein PACS-2 [10], the fission factor Fis1, the ER protein Bap31 [11] and the mitofusin Mfn2 [12].

A screen to identify the ER–mitochondria protein junction

We turned to the simple budding yeast Saccharomyces cerevisiae and designed a screen aimed at discovering the bridges that connect the ER to the mitochondria. Inspired by a study in mammalian cells that used engineered chimaeric proteins to artificially tether the ER to the mitochondria [13], we reasoned that artificial ER–mitochondria tethering might allow the growth of mutants in which endogenous tethering components were mutated. We created an artificial ER–mitochondria tether and screened for mutants that could not grow in the absence of this construct. Two mutations displayed the expected behaviour, both of them affecting the MDM12 gene [14].

Mdm (mitochondrial distribution and morphology) 12 is a peripheral OMM protein identified more than 10 years ago in a screen for mitochondrial distribution- and morphology-deficient mutants [15]. Mutants defective for Mdm12 grow poorly on fermentable medium and are totally incapable of growing on a non-fermentable carbon source [14,15]. These phenotypes are rescued by the expression of the artificial tether. Mdm12 is embedded in a complex containing two mitochondrial proteins (Mdm10 and Mdm34) and a fourth protein called Mmm1 (mitochondrial morphology maintenance 1) (Figure 1A). This protein was long thought to be inserted in the OMM. We were able to show that, despite deceptive appearances, Mmm1 is an ER-integral protein [14]. All four proteins form stable complexes at the interface of the ER and the mitochondria, thus zippering the two organelles. We named this complex ERMES (ER–mitochondria encounter structures). ERMES forms one to five discrete focal structures per cell [16] at the ER–mitochondria interface [14] (Figure 1B).

Figure 1 Topology of the ERMES complex at the ER–mitochondria connection

(A) Mmm1 is an ER protein, while Mdm10 is an OMM β-barrel. Mdm12 is a cytosolic protein. Mdm34 is associated with the OMM through an unclear mechanism. (B) The ERMES complex accumulates in a few discrete foci per cell. Here a green fluorescent protein (GFP)-tagged version of Mdm34 (green) is imaged together with a marker for the mitochondria (red). The outlines of the cells are shown as broken lines.

Physiology of the ER–mitochondria connection: lipid exchange

As stated above, one role of ER–mitochondria connection is to facilitate lipid exchange between both organelles [5]. Two pieces of evidence from the analysis of ERMES corroborate this model.

The first piece of evidence stems from an unbiased genome-scale screen that utilizes genetic interactions to compare thousands of mutants according to the similarity of their genetic interaction pattern. Hierarchical clustering allows the sorting of genes according to their proximity in this analysis. In the case of ERMES, all four components of the complex tightly segregated together, demonstrating the power of the method at detecting functional relationships [14]. In addition, the cluster contained the GTPase Gem1 (see below) and Psd1. This latter enzyme is central in the biosynthesis pathway of PE. While all other PE biosynthesis steps are carried out in the endomembrane system, Psd1 is an IMM protein, meaning that both its substrate and products have to come from and go back to the ER respectively [5], indicating a functional link between ERMES and ER–mitochondrial lipid exchange.

The second piece of evidence arises from bioinformatic analyses. Mdm12 is the founding member of the SMP (synaptotagmin-like, mitochondrial and lipid-binding protein) family of membrane proteins. Both Mdm34 and Mmm1 also contain SMP domains, together with a handful of conserved eukaryotic proteins [17,18]. Homology searches using secondary structure-based hidden Markov models detected a remote resemblance between SMP domains and a class of lipid-binding proteins [18]. This family of lipid-binding proteins is characterized by its tubular shape and is called TULIP (tubular lipid-binding). Lipid binding in the TULIP domain involves an elongated hydrophobic groove that buries the hydrophobic moieties of bound lipids, whereas the polar head remains usually mainly accessible to the aqueous environment (Figure 2B). The similarity between SMP and TULIP domains suggests a very tempting model where the SMP domains found in Mdm12, Mdm34 and Mmm1 directly catalyse lipid exchange between adjacent membranes at the ER–mitochondria interface [18].

Figure 2 ERMES complexes may exchange lipids between the ER and the mitochondria

(A) Bioinformatic comparison of the structures of ERMES members show that three of four canonical ERMES components harbour an SMP domain, either alone (Mdm12), or in combination with a transmembrane domain (TM; Mmm1) or a DUF (domain of unknown function; Mdm34). Mdm10 harbours no SMP domain and instead bears characteristics of MOM (mitochondrial outer membrane) β-barrels. (B) The SMP domain is related to the TULIP domain. The crystal structure of a TULIP protein, the human bactericidal permeability-increasing (BPI) protein [35], shows two phospholipid molecules bound in two pseudo-symmetrical lipid-binding domains. The lipids have their acyl chain buried within the protein and their polar head exposed to the solvent (upper panel). Lower panel displays a 90° rotation of the structure, showing that the lipid-binding groove extends along the length of the protein.

Physiology of the ER–mitochondria connection: a plethora of other processes

The physiological role of ERMES does not appear limited to lipid exchange. In fact, ERMES components have been under heavy scrutiny, long before our discovery of their involvement in ER–mitochondria connections. These studies point to a much broader physiological role of ERMES in mitochondrial biology.

ERMES localizes adjacent to the mtDNA (mitochondrial DNA), which is arranged in the mitochondria as nucleoids [19]. Strikingly, ERMES co-localizes specifically with actively replicating nucleoids, indicating that ERMES may be involved in the regulation of mtDNA replication [20]. A physical link between ERMES and mtDNA replication proteins has even been shown [20], and ERMES mutants frequently lose mtDNA, indicating a defect in replication or segregation of the mitochondrial genome [19].

ERMES complexes are also indirectly connected to mitochondrial protein import. Mdm10 is a central component of ERMES but a fraction of it can be found as a part of another complex, the SAM (sorting and assembly machinery), that assembles β-barrels in the OMM [21]. Mdm10 is itself a β-barrel protein and competes with other β-barrels for the substrate-binding pocket in the SAM [22]. Mdm10 presence within SAM is especially important for the assembly of TOM40 (translocase of the mitochondrial outer membrane 40) [21]. Too little Mdm10 causes TOM40 to remain associated with the SAM complex after its full assembly. Conversely, too much Mdm10 causes premature displacement of TOM40 from the SAM complex [22]. Mutations in other components of ERMES cause the same defects in β-barrel assembly as does Mdm10 overexpression [22,23], suggesting that both complexes compete for a limiting amount of Mdm10. Although the link between ER–mitochondria connection and mitochondrial protein import is unclear, this competition suggests that SAM and ERMES may regulate each other [24].

Thus it appears that ERMES is found at an important crossroad of mitochondrial biology, potentially extending its influence on processes as diverse as membrane biogenesis, genome replication and protein import. This suggests that ERMES may be a co-ordinator that ensures that these processes are performed in harmony for the proper function of mitochondria [25].

The Miro GTPase Gem1 as a regulator of ERMES

If ERMES is indeed such a co-ordinator, then it is expected to sense cellular cues and propagate appropriate responses. This immediately raises the following question: what regulates ERMES? We obtained a preliminary answer by characterizing the ERMES complex biochemically.

Using affinity purification of functional Tap-tagged versions of Mmm1 and Mdm34, we and another group purified intact ERMES complexes containing all four known components [26,27]. In addition, we identified the Miro GTPase Gem1 as a novel ERMES component. Gem1 co-localizes with ERMES core components in a few foci per cells [26,27] (Figure 3B).

Figure 3 The Miro GTPase Gem1 is a component of the ERMES complex

(A) Architecture of Miro GTPases: Two GTPase domains flank two Ca2+-binding domains. The C-terminus anchors the protein to the OMM. C, C-terminus; Cyt, cytosol; IMS, inter-membrane space; N, N-terminus. (B) Green fluorescent protein (GFP)-tagged Gem1 is found in focal structures together with mCherry-tagged Mdm34 at the ER–mitochondria interface. The broken lines represent the outlines of the cells.

Miro GTPases are characterized by their unusual architecture. Two GTPase domains flank two Ca2+-binding EF hands. The C-terminus of Miro GTPases constitutes a hydrophobic tail, which anchors the protein to the OMM [28] (Figure 3A). This architecture strongly suggests that Gem1 could play a regulatory role in ERMES biology. Indeed, contrary to deletions of the canonical ERMES members, deletion of Gem1 does not compromise complex formation. However, ERMES foci in Gem1-deficient cells appear larger and in reduced number compared with wild-type counterparts, suggesting that Gem1 influences ERMES organization [26]. The recruitment of Gem1 to ERMES appears to be itself a regulated process. Mutations that kill the GTPase activity of the first domain abrogate Gem1 localization to ERMES foci, while mutations in the second domain do not. The same holds true for mutations of the first and second EF hands.

gem1Δ cells display phenotypes related to ERMES deficiencies. As noted above, Gem1 co-segregated with other ERMES proteins in an unbiased genetic interaction map, indicating that gem1Δ and ERMES-deficient cells suffer from a common problem [14].

Among the analysed genetic interactions are a series of synthetic phenotypes of enzymes of the CL (cardiolipin) (diphosphatidylglycerol) biosynthesis pathway. Like ERMES core components, Gem1 is synthetically lethal with enzymes of the CL biosynthesis pathway [26,29], consistent with a role for Gem1 in regulating ERMES action on lipid metabolism.

Accordingly, expressing a mutant form of Gem1 incapable of localizing to ERMES does not cure this synthetic lethality. Interestingly, although correctly localized to ERMES, mutants of the second GTPase domain are incapable of rescuing the synthetic lethality [26,30].

This suggests a model in which both GTPase domains perform distinct functions. The first domain localizes Gem1 to ERMES, and, once localized, Gem1 acts in the lipid biosynthesis pathway via its second GTPase domain. The cues to which Gem1 responds as well as Gem1's action on ERMES remain to be identified.

Miro GTPases have conserved roles at the ER–mitochondria interface

Gem1 stands out among the ERMES components, because it is highly conserved [28]. Miro GTPases are found in most eukaryotes, with members in distant phyla such as metazoans, fungi [31], plants [32] and many protists. This contrasts with the remaining components of the ERMES complex, which can be unambiguously identified only in fungi [17].

Mammals actually have two isoforms of Miro, Miro-1 and Miro-2, that are both important for Ca2+-regulated mitochondrial movement along microtubules [33,34]. Using a monoclonal antibody against Miro-1 we showed that Miro-1 is found in few foci on the mitochondrial surface, which are highly reminiscent of ERMES foci. Moreover, those foci are systematically found at sites where the ER and mitochondria overlap, suggesting that Miro-1 is preferentially found at ER–mitochondria connections, probably embedded in a tethering complex [26].


The molecular mechanisms that mediate ER–mitochondria communication are still poorly understood. Recent years have seen major advances, and common themes in organelle tethering are emerging. The search for an ER–mitochondria tethering complex was originally motivated by the supposed central role of the ER–mitochondria connection in lipid and Ca2+ exchange between the two organelles. The discovery of ERMES now extends the physiological scope of ER–mitochondria connection to additional functions such as mtDNA replication and mitochondrial protein import. Although the connection between those processes is unclear, their remarkable diversity suggests that ER–mitochondria contact sites represent a major site of communication between mitochondria and the rest of the cell [25,26]. Future research will probably provide an integrated view of these communication routes. Important questions include how lipids are extracted from some membranes and delivered to others; how are ERMES complexes physically connected to the mitochondrial genome in a large assembly spanning the ER, the OMM and the IMM; last, but not least, what is the molecular nature of the tethering complex(es), if any, to which Miro proteins are associated in metazoan cells.


The BK laboratory is supported by the Swiss National Science Foundation [grant number PPOOP3-133651].


  • Cellular Traffic of Lipids and Calcium at Membrane Contact Sites: A Biochemical Society held at the Snowbird Ski and Summer Resort, Snowbird, UT, U.S.A., 6–9 October 2011. Organized and Edited by Tim Levine (Institute of Ophthalmology, London, U.K.) and William Prinz (National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health, Bethesda, MD, U.S.A.).

Abbreviations: CL, cardiolipin; ER, endoplasmic reticulum; ERMES, ER–mitochondria encounter structures; IMM, inner mitochondrial membrane; IP3, inositol trisphosphate; Mdm, mitochondrial distribution and morphology; Mmm1, mitochondrial morphology maintenance 1; mtDNA, mitochondrial DNA; OMM, outer mitochondrial membrane; PE, phosphatidylethanolamine; Psd1, phosphatidylserine decarboxylase 1; SAM, sorting and assembly machinery; SMP, synaptotagmin-like, mitochondrial and lipid-binding protein; TOM, translocase of the mitochondrial outer membrane; TULIP, tubular lipid-binding


View Abstract