Intramembrane Proteolysis of Mgm1 by the Mitochondrial Rhomboid Protease Is Highly Promiscuous Regarding the Sequence of the Cleaved Hydrophobic Segment

doi:10.1016/j.jmb.2010.06.014 J. Mol. Biol. (2010) 401, 182–193

Available online at www.sciencedirect.com

Intramembrane Proteolysis of Mgm1 by theMitochondrialRhomboid Protease Is Highly Promiscuous Regardingthe Sequence of the Cleaved Hydrophobic Segment

Anja Schäfer1,2, Michael Zick2, Jan Kief1, Mirco Steger3,Heinrich Heide3, Stéphane Duvezin-Caubet2,4, Walter Neupert2

and Andreas S. Reichert1,2⁎

© 2010 Elsevier Ltd. All rights reserved.

1CEF MakromolekulareKomplexe, MitochondrialeBiologie, Fachbereich Medizin,Goethe-Universität Frankfurtam Main, Theodor-Stern-Kai 7,60590 Frankfurt am Main,Germany2Adolf-Butenandt-Institut fürPhysiologische Chemie,Ludwig-Maximilians-Universität München,Butenandtstr. 5, 81377München, Germany3Molekulare Bioenergetik,Fachbereich Medizin,Goethe-Universität Frankfurtam Main, Theodor-Stern-Kai 7,60590 Frankfurt am Main,Germany4Institut de Biochimie etGénétique Cellulaires,CNRS-Université Bordeaux 2,33077 Bordeaux, France

Received 6 January 2010;received in revised form28 May 2010;accepted 8 June 2010Available online15 June 2010

0022-2836/$ - see front matter © 2010 E

*Corresponding author. E-mail [email protected] used: RCR, rhomb

mtDNA, mitochondrial DNA; MPPprocessing peptidase; TM, transmemMgm1; Dld1, D-lactate dehydrogenafacing AAA; EDTA, ethylenediaminextracted ion current.

Rhomboids are a family of intramembrane serine proteases that areconserved in bacteria, archaea, and eukaryotes. They are required fornumerous fundamental cellular functions such as quorum sensing, cellsignaling, and mitochondrial dynamics. Mitochondrial rhomboids form anevolutionarily distinct class of rhomboids. It is largely unclear how theiractivity is controlled and which substrate determinants are responsible forrecognition and cleavage. We investigated these requirements for themitochondrial rhomboid protease Pcp1 and its substrate Mgm1. In contrastto several other rhomboid proteases, Pcp1 does not require helix-breakingamino acids in the cleaved hydrophobic region of Mgm1, termed ‘rhomboidcleavage region’ (RCR). Even transmembrane segments of inner membraneproteins that are normally not processed by Pcp1 become cleavable whenput in place of the authentic RCR ofMgm1.We further show that mutationalalterations of a highly negatively charged region located C-terminally to theRCR led to a strong processing defect. Moreover, we show that thedeterminants required for Mgm1 processing by mitochondrial rhomboidprotease are conserved during evolution, as PARL (the human ortholog ofPcp1) showed similar substrate requirements. These results suggest asurprising promiscuity of the mitochondrial rhomboid protease regardingthe sequence requirements of the cleaved hydrophobic segment.We proposea working hypothesis on how the mitochondrial rhomboid protease can,despite this promiscuity, achieve a high specificity in recognizing Mgm1.This hypothesis relates to the exceptional biogenesis pathway of Mgm1.

Keywords: rhomboid protease; substrate recognition; mitochondria; intra-membrane proteolysis; Mgm1

lsevier Ltd. All rights reserve

ress:

oid cleavage region;, mitochondrialbrane segment ofse 1; m-AAA, matrix-etetraacetic acid; XIC,

Introduction

Rhomboid proteases constitute a widely con-served family of intramembrane serine proteasespresent in bacteria, archaea, and eukaryotes.1

Members of this protease family are involved indiverse processes such as epidermal growth factor

d.

183Intramembrane Proteolysis of Mgm1

signaling in Drosophila melanogaster,2,3 quorumsensing in Providencia stuartii,4,5 and host cellinvasion by the protozoan parasite Toxoplasmagondii.6,7 Rhomboid proteases are membrane pro-teins that share a conserved core of six transmem-brane helices harboring the active site. In general,cleavage occurs within a single membrane-spanningsegment present in the substrate protein.8 Mito-chondrial rhomboid proteases belong to an evolu-tionarily distinct class,1 the PARL-like subfamily ofrhomboid proteases, which is characterized by anN-terminal mitochondrial targeting sequence andan additional seventh transmembrane segmentN-terminal to the conserved six-transmembrane core.1

Substrate recognition by rhomboid proteases, ingeneral, was addressed in a number of studies, anddifferent mechanisms have been proposed. Proces-sing of the epidermal growth factor receptor ligandSpitz by Drosophila rhomboid-1 was shown todepend on a substrate motif that contains helix-breaking residues such as glycine and proline at theN-terminal part of the cleaved transmembranesegment.9 A variety of rhomboids from eukaryoticand prokaryotic species were shown to have asimilar dependency on helix-breaking residues suchas rhomboid-1.9,10 Cleavage of Escherichia colilactose permease by the rhomboid protease GlpGwas shown to be more efficient when the residuelocated N-terminally to the cleaved peptide bond(P1 position) was small.11 Moreover, substrates ofother nonrhomboid intramembrane proteases suchas presenilins, site 2 protease, and signal peptidepeptidase need to be processed by another proteaseprior to intramembrane cleavage,12–14 but such arequirement is commonly not observed for sub-strates of rhomboid proteases. Recently, a sequence-specific recognition motif present in a number ofprokaryotic and eukaryotic rhomboid substrates hasbeen identified.15 It surrounds the cleavage site andhas been reported to be more relevant to cleavagethan helix-destabilizing residues. However, mito-chondrial rhomboid substrates do not contain such arecognition motif,15 suggesting that other recogni-tion mechanisms apply to the PARL subfamily ofrhomboid proteases.In Saccharomyces cerevisiae, the mitochondrial

rhomboid protease Pcp1 is essential for maintenance

Fig. 1. Intramembrane proteolysis of Mgm1 occurs irrespresidues in the RCR. (a) Schematic representation of the alteMgm1 is imported via the TIM23 complex (TIM23). MPP reminserted into the inner membrane (IM) and l-Mgm1 is formedthe IM. This depends on the hydrophobicity of the TM, sufficProcessing by Pcp1 leads to the formation of s-Mgm1 and occIMS, intermembrane space; p-Mgm1, precursor protein of Mg(boldface) and flanking regions of the putative Mgm1 homologthe RCR with reference to S. cerevisiae Mgm1 are highlighted iRCR are highlighted in dark gray; and charged residuesrepresentation of the N-terminal region of Mgm1 and respectiPcp1 and MPP cleavage sites are indicated by arrowheadsmitochondrial targeting signal. Bottom: Analysis of the intramacid exchanges of helix-breaking residues within the RCR. TΔmgm1Δpcp1 (−Pcp1) strains expressing the indicated Mgantibodies raised against Mgm1. WT, wild-type Mgm1; l, l-M

of mitochondrial DNA (mtDNA) and normalmitochondrial morphology.16–19 In wild-type cells,mitochondria continuously fuse and divide in abalanced manner.20 When fusion of mitochondria isblocked, the tubular mitochondrial networkbecomes fragmented due to ongoing fission events.As a consequence, mtDNA is not inherited correctly,and yeast cells lose their capacity for respiratorygrowth. One of the components required formitochondrial fusion is the dynamin-like GTPaseMgm1. Mgm1 exists as a long isoform (l-Mgm1) andas a short isoform (s-Mgm1). The balanced forma-tion of both isoforms is required for normalmitochondrial morphology and maintenance ofmtDNA.17,21 Biogenesis of these two isoforms wasshown to occur by a pathway termed ‘alternativetopogenesis’ (Fig. 1a).22 During import into mito-chondria, the Mgm1 precursor takes two alternativeroutes: either the N-terminal targeting sequence isremoved by the mitochondrial processing peptidase(MPP) and the protein is integrated into themitochondrial inner membrane with the transmem-brane segment of Mgm1 (TM) located close to theN-terminus (amino acid residues 94–111), giving riseto l-Mgm1; or the protein is not inserted via the TMinto the inner membrane but is translocated furtheruntil a second hydrophobic segment (residues 156–169) reaches the inner membrane; processing by therhomboid protease Pcp1 then generates s-Mgm1.22

The cleavage of Mgm1 occurs between residues 160and 161 or between residues 162 and 16317,22 withinthis second hydrophobic segment, which we termedrhomboid cleavage region (RCR). The proportion ofMgm1 precursor that is inserted into the innermembrane with its RCR and cleaved by Pcp1depends on the hydrophobicity of the TM, the ATPlevel, and the presence of a functional protein importmotor. These studies suggested that formation of l-Mgm1 and formation of s-Mgm1 are competingprocesses.22

Altogether, it is still unclear which structuralelements in mitochondrial substrate proteins deter-mine the recognition of and subsequent cleavageby the mitochondrial rhomboid protease. Weaddressed this question by studying the determi-nants within Mgm1 for processing by Pcp1. We alsoanalyzed to what degree our conclusions can be

ective of the presence or the absence of helix-breakingrnative topogenesis of Mgm1, as proposed previously.22

oves the mitochondrial targeting signal. Either the TM is, or Mgm1 gets translocated further until the RCR reachesient levels of matrix ATP, and a functional import motor.urs when the cleavage site within the RCR reaches the IM.m1. (b) Alignment of sequences corresponding to the RCRs of various fungal organisms. Conserved residues withinn light gray; negatively charged residues C-terminal to theN-terminal to RCR are underlined. (c) Top: Schematicve modifications. Numbers indicate amino acid positions.. Amino acid exchanges are marked in boldface. MTS,embrane proteolysis of Mgm1 variants containing aminootal yeast cell extracts from Δmgm1/Δmgm1 (+Pcp1) orm1 variants were analyzed by immunoblotting usinggm1; s, s-Mgm1; f, Mgm1 fragments.

Fig. 1 (legend on previous page)

184 Intramembrane Proteolysis of Mgm1

185Intramembrane Proteolysis of Mgm1

extended to the human mitochondrial rhomboidprotease PARL. Our results suggest a mechanism ofMgm1 processing that involves the action of theTIM23 translocase bywhich the RCR is positioned insuch a way that it becomes accessible to Pcp1. Thishypothetical mechanism can explain the surprisingpromiscuity of Pcp1 with respect to the amino acidsequence of the RCR and can ascribe an importantrole to regions flanking the RCR in presenting theMgm1 substrate to the protease.

Fig. 2. Distinct hydrophobic segments become cleav-able by Pcp1 when put in the sequence context of Mgm1.(a) Top: Schematic representation of the replacement ofthe RCR by the TM and Dld1. Bottom: Total yeast cellextracts of Δmgm1/Δmgm1 (+Pcp1) or Δmgm1Δpcp1(−Pcp1) strains expressing the indicated Mgm1 variantswere analyzed by immunoblotting using antibodies raisedagainst Mgm1. (b) Top: Indicated Mgm1 variants wereimmunoprecipitated, and l-Mgm1 and s-Mgm1 isoformswere visualized by SDS-PAGE and silver staining. Bandscorresponding to s-Mgm1 were excised and subjected tonanoscale liquid chromatography/electrospray ionizationtandem mass spectrometry. Bottom: Schematic represen-tation of identified N-terminal peptides of the indicated s-Mgm1 variants. Black triangles indicate observed cleavagesites. Mutated amino acid residues are shown in boldface.WT, wild-type Mgm1; l, l-Mgm1; s, s-Mgm1; f, Mgm1fragments.

Results

Helix-breaking residues in the RCR of Mgm1 arenot required for intramembrane proteolysis byPcp1

To determine which residues within the RCR ofMgm1 are important in making it a substrate for therhomboid protease Pcp1, we performed a sequencecomparison of the RCR of Mgm1 with corres-ponding sequences from other fungal species(Fig. 1b). Despite a similar overall hydrophobicity,there is no strong sequence conservation of thissegment. However, in all organisms analyzed, thisregion contains several helix-breaking residues suchas glycine and proline (Fig. 1b). To test whetherhelix-breaking residues are important for Mgm1cleavage, we exchanged glycine 156, proline 159, orboth residues with either alanine or valine (Fig. 1c).The Mgm1 variants G156A, G156V, P159A, P159V,and G156A/P159A were expressed in a strainlacking endogenous Mgm1 (Δmgm1). This back-groundwas used in all of the following experiments,unless indicated otherwise. Total cell extracts wereprepared, and formation of Mgm1 isoforms wasanalyzed by SDS-PAGE and immunoblotting. Inall mutants tested, the ratio of Mgm1 isoformswas identical with that in wild type (Fig. 1c).The formation of the short isoform was dependenton Pcp1, as s-Mgm1 was not observed in aΔmgm1Δpcp1 background. Taken together, helix-breaking residues in the RCR do not play a role inthe processing of Mgm1 by Pcp1.

Pcp1 is capable of cleaving transmembranehelices distinct from the RCR

To investigate whether other elements within theRCR have a role in substrate recognition, we alteredseveral residues flanking the Pcp1 cleavage site,exchanged or deleted a conserved aspartate residue(D155) N-terminal to the RCR, or introduced fourhydrophobic residues in the RCR. However, noneof these mutations impaired the Pcp1-dependentformation of s-Mgm1 (Fig. S1).Next, we tested whether the RCR at all contains

substrate recognition information by replacing thewhole RCR with transmembrane segments thatnormally are not subject to cleavage by Pcp1. Wechose the TM or D-lactate dehydrogenase 1 (Dld1), a

protein of the intermembrane space anchored to theinner membrane by a single transmembranesegment.23 Both constructs were efficiently pro-cessed in a Pcp1-dependent manner (Fig. 2a).Thus, two segments that are not cleaved by Pcp1in their native context become a substrate for Pcp1when put in place of the authentic RCR segment. Weconclude that the RCR of Mgm1 does not contain anapparent sequence element required for cleavage byPcp1, demonstrating that recognition must occur viadeterminants outside of this segment.

Fig. 3. Deletion of a stretch of negatively chargedresidues located C-terminally to the RCR impairscleavage of Mgm1 by Pcp1. (a) Top: Schematic represen-tation of the N-terminal region of Mgm1 variants withindicated deletions (Δ). Bottom and (b): Total yeast cellextracts from Δmgm1/Δmgm1 (+Pcp1) or Δmgm1Δpcp1(−Pcp1) strains expressing the indicated Mgm1 variantswere analyzed by immunoblotting using antibodiesraised against Mgm1. WT, wild-type Mgm1; l, l-Mgm1;s, s-Mgm1; f, Mgm1 fragments.

186 Intramembrane Proteolysis of Mgm1

Mutagenesis of amino acid residues near thecleavage site does not affect the exact cleavageposition

In order to test how mutagenesis of the RCRaffects the exact cleavage position by Pcp1, wedetermined the N-terminal amino acid sequences ofs-Mgm1 of certain variants. Next to wild-typeMgm1, we focused on Mgm1 variants A161V,T160A/A161F/T162A (AFA), T160L/A161L/T162L (LLL), TM, and Dld1, as they represent themost drastic alterations within the RCR. Theseconstructs were expressed in a Δmgm1/Δmgm1strain, Mgm1 was immunoprecipitated from isolat-ed mitochondria, eluted proteins were separated bySDS gel electrophoresis, and silver-stained proteinbands corresponding to s-Mgm1 were digested bytrypsin and analyzed by nanoscale liquid chroma-tography/electrospray ionization tandem massspectrometry (Fig. 2b). We confirmed that themajor P1′ cleavage sites in wild-type Mgm1 areA161 and L163 (Fig. 2b; Fig. S2a and b), as reportedpreviously.17 An additional, apparently minor,cleavage site between those sites was also observed.The P1′ cleavage site L163 was invariantly observedfor those Mgm1 variants containing mutationsdirectly surrounding the wild-type cleavage sites,namely, A161V, AFA, and LLL (Fig. 2b; Fig. S2a, c,d, and e). When the RCR was replaced by the TM,the cleavage site was spaced seven amino acidresidues from the C-terminal flank of the RCR(Fig. 2b; Fig. S2a and f). Thus, although the length ofthe hydrophobic segment was increased by fouramino acids, the cleavage site was not shiftedrelative to the C-terminal flank. Analysis of theDld1 variant of Mgm1 revealed three cleavage siteslocated within the hydrophobic segment, whereasspacing to the C-terminal flank appeared to be morevariable in this case. Taken together, the cleavageposition within Mgm1 is not affected by majoralterations surrounding the endogenous cleavagesites, suggesting that the RCR is indeed devoid of asequence-specific recognition motif.

Intramembrane proteolysis of Mgm1 ispromoted by a stretch of negatively chargedresidues C-terminal to the RCR

Mgm1 contains two stretches of negativelycharged residues that are located C-terminally tothe RCR and are also present in Mgm1 homologs ofother fungal species (Fig. 1b). We constructed Mgm1mutants in which either the four negatively chargedresidues directly C-terminal to the RCR (DDDE;residues 170–173) or the D/E stretch furtherC-terminal (residues 182–194) was removed(Fig. 3a). Mgm1ΔDDDE was processed similarly towild-type Mgm1. However, cells expressing anMgm1 variant lacking the D/E stretch (Mgm1ΔD/E)contained reduced levels of s-Mgm1,while the level ofl-Mgm1 was increased (Fig. 3a). We verified s-Mgm1formation to be dependent on Pcp1 (Fig. 3a), and weexcluded improper sorting and insertion into the

inner membrane as both variants behaved similarlyto wild-type Mgm1 upon submitochondrial locali-zation and carbonate extraction (Figs. S4a–c andS5a). We further excluded the possibility that alteredstabilities of the individual isoforms derived fromthis variant compared to those of the wild-typeisoforms of Mgm1 might explain this observation(Fig. S6a and b).To test whether deletion of the D/E stretch could

promote membrane insertion of the TM and therebyreduce processing of Mgm1, we studied Mgm1variants lacking the TM. As reported previously,22

in the absence of the TM, the entire pool of Mgm1 isconverted into s-Mgm1 (Fig. 3b). However, upondeletion of both the TM and the D/E stretch, aconsiderable fraction of the entire Mgm1 pool wasnot processed to s-Mgm1 in mitochondria (Fig. 3b),although both Mgm1ΔTM and Mgm1ΔTMΔD/Ewere sorted correctly (Fig. S4j and k). Thus, thedeletion of the D/E stretch impaired Pcp1-depen-dent cleavage of Mgm1 also in the absence of theTM, ruling out the possibility that lack of the D/E

187Intramembrane Proteolysis of Mgm1

stretch could promote membrane insertion of theTM and thereby reduce s-Mgm1 formation by Pcp1.In summary, we conclude that the 13 negativelycharged residues C-terminal to the RCR are impor-tant for the efficient formation of s-Mgm1.

The D/E stretch is not required for Mgm1function as long as both isoforms are presentin roughly equal amounts

Deletion of the D/E stretch could eventually causea major misfolding of Mgm1, which might indirectlyaffect substrate recognition. To test this, we ana-lyzed whether or not removal of the D/E stretchcauses a loss of function of Mgm1. Deletion ofMgm1 or even the loss of only one of the twoisoforms is known to cause a loss of mtDNA andthus a deficiency in respiratory growth.17,22,24 Wechecked whether variants of individual isoforms ofMgm1 lacking the D/E stretch restored competencefor respiratory growth when expressed in a Δmgm1background. We used a plasmid-shuffling approachin which a wild-type copy of Mgm1was replaced bythe Mgm1 variants to be tested. Consistent withearlier studies,17,21 the expression of either a largeMgm1 isoform (Mgm1ΔRCR) or a short Mgm1isoform (Mgm1ΔTM) alone did not complement therespiratory growth defect of the Δmgm1 strain,whereas coexpression of both variants (Mgm1ΔRCRand Mgm1ΔTM) did (Fig. 4a). This was expected as,only in the latter case, both isoforms were present inroughly equal amounts (Fig. 4b). When the

Fig. 4. The D/E stretch of Mgm1 is not required for funcomplementation assay showing the tested Mgm1 variants eIndicated strains were tested for their capacity for fermentativfor their capacity for respiratory growth on a complete mediumextracts fromΔmgm1 strains expressing the indicatedMgm1 varaised against Mgm1. WT, wild-type Mgm1; l, l-Mgm1; s, s-M

Mgm1ΔD/E and Mgm1ΔTMΔD/E variants, bothlacking the D/E stretch, were expressed separately,complementation of Mgm1 function was not ob-served (Fig. 4a), consistent with the markedlyaltered ratio of the two resulting isoforms (Fig. 4b).However, when these two variants (Mgm1ΔD/Eand Mgm1ΔTMΔD/E) were coexpressed, the rela-tive amounts of the resulting long and shortisoforms were roughly equal (Fig. 4b), and Mgm1function was efficiently rescued (Fig. 4a). Therefore,deletion of the D/E stretch does not directly disturbMgm1 function and thus is very unlikely to cause amajor misfolding of Mgm1.We conclude that the D/E stretch is not directly required for Mgm1 function.

Reducing the amount of negative charges inthe D/E stretch progressively impairsPcp1-dependent cleavage

We further corroborated the role of the D/Estretch in Pcp1-dependent processing by systemat-ically replacing these negatively charged amino acidresidues with either uncharged residues (5u, 9u, and13u) or positively charged residues (5+, 9+, and 13+)(Fig. 5). The Mgm1 variants 5u and 5+ wereprocessed as efficiently as wild-typeMgm1, whereasthe other variants (9u, 13u, 9+, and 13+) were lessefficiently processed by Pcp1 (Fig. 5). The morenegative charges were removed, the more pro-nounced was the processing defect. These variantsbehaved similarly to wild-type Mgm1 in submito-chondrial localization and membrane extraction

ction. (a) Schematic representation of sectors used in thexpressed alone or as a combination of two variants (left).e growth on a complete medium with glucose (middle) orwith glycerol as carbon source (right). (b) Total yeast cellriants were analyzed by immunoblotting using antibodiesgm1.

Fig. 5. Reducing the amount of negative chargeswithin the D/E stretch of Mgm1 progressively impairsPcp1-dependent cleavage. Top: Schematic representationof the N-terminal region of Mgm1 variants with indicatedamino acid changes within the D/E stretch. Bottom: Totalyeast cell extracts from Δmgm1/Δmgm1 (+Pcp1) orΔmgm1Δpcp1 (−Pcp1) strains expressing the indicatedMgm1 variants were analyzed by immunoblotting usingantibodies raised against Mgm1. WT, wild-type Mgm1; l,l-Mgm1; s, s-Mgm1; f, Mgm1 fragments.

188 Intramembrane Proteolysis of Mgm1

experiments excluding improper sorting or mem-brane insertion (Figs. S4d–i and S5b). The otherknown substrate of Pcp1, Ccp1, contains a relatively

with Fig. 3); (d) amino acid exchanges within the D/E strMgm1; s, s-Mgm1.

short stretch of negative charges (EDDE), which islocated C-terminally to its RCR (Fig. S3a). However,deleting these negative charges did not impair Pcp1-dependent processing of Ccp1, suggesting that sucha stretch is not strictly required for substraterecognition by Pcp1 (Fig. S3b). Still, our datademonstrate that the extended stretch of negativecharges C-terminal to the RCR is an importantdeterminant for the intramembrane proteolysis ofMgm1 by Pcp1.

Substrate recognition by the humanmitochondrialrhomboid protease PARL is similar to that of itsyeast counterpart

We asked whether the mechanism by whichmitochondrial rhomboid proteases recognize theirsubstrates is conserved in evolution. To this end, wetested the ability of the human Pcp1 homolog PARLto process the different Mgm1 variants. It wasshown previously that PARL overexpressed in aΔpcp1 strain was able to efficiently process both ofthe Pcp1 substrates Mgm1 and Ccp1.18,25

The Δmgm1Δpcp1 strains expressing the Mgm1variants were transformed with a plasmid contain-ing the PARL gene under the control of a galactose-inducible promoter and grown on media with orwithout galactose. Expression of PARL was con-firmed by detection of processed and unprocessedCcp1 in induced or noninduced samples, respec-tively (data not shown). Like Pcp1, PARL processedall Mgm1 variants with mutations in the RCR, eventhose where the RCR was replaced by transmem-brane segments of Mgm1 or Dld1 (Fig. 6 a and b; Fig.S5). Mgm1ΔDDDE was cleaved efficiently, whileMgm1ΔD/E was not (Fig. 6c). Moreover, replacingthese negatively charged residues progressivelywith uncharged or positively charged residues hada similar negative effect on PARL-dependent

Fig. 6. The sequence determi-nants within Mgm1 required forintramembrane proteolysis areevolutionarily conserved betweenyeast and humans. Total cellextracts from Δmgm1Δpcp1 strainscoexpressing indicated Mgm1 var-iants and the human mitochondri-al rhomboid protease PARL(+PARL or −PARL) were ana-lyzed by immunoblotting usingantibodies against Mgm1. Variantswith (a) amino acid exchanges inhelix-breaking residues (comparewith Fig. 1c); (b) replacement ofthe RCR by the TM or Dld1(compare with Fig. 2); (c) deletionof negatively charged amino acidsC-terminal to the RCR (compare

etch (compare with Fig. 5). WT, wild-type Mgm1; l, l-

189Intramembrane Proteolysis of Mgm1

processing as in Pcp1-dependent processing (Fig.6d). Thus, the same stretch of negative charges isimportant for the processing of Mgm1 by PARL, aswell as by Pcp1. Taken together, these resultsdemonstrate that the mechanism of substrate recog-nition is conserved among yeast and humanmitochondrial rhomboid proteases.

Discussion

Regulated intramembrane proteolysis is a processinvolved in a variety of cellular pathways. Themolecular determinants that govern the substratespecificity of rhomboid proteases have been studiedin a number of cases and, in general, involvesequence determinants within the cleaved hydro-phobic segment. For mitochondrial rhomboid pro-teases, which form an evolutionarily distinctsubfamily, this was largely unclear. We haveinvestigated here the determinants required for theprocessing of Mgm1 by the mitochondrial rhomboidproteases of yeast and humans, Pcp1 and PARL. Wereport the rather surprising finding that theseintramembrane proteases are highly promiscuousregarding the recognition of the cleaved hydropho-bic segment, since massive changes can be intro-duced in the RCR without loss of cleavability. Onthe other hand, cleavage is highly specific in thewild-type situation as it occurs at specific sites in theonly two substrate proteins known to date, Mgm1and Ccp1.17,18,26 Furthermore, this specificity ishardly affected by an exchange of amino acidresidues near the peptide bond to be cleaved. Thisraises the intriguing question as to how Pcp1 worksand how the vast majority of nonsubstrate mem-brane proteins in the inner membrane are protectedfrom unspecific cleavage by Pcp1.Our study reveals several features that are

important to intramembrane proteolysis of Mgm1.Cleavage of Mgm1 requires the presence of ahydrophobic stretch, the RCR, whose sequence,however, can be highly variable. A multitude ofalterations in this region did not affect processing.The hydrophobic region comprising the RCR couldeven be replaced by a transmembrane anchorsequence of the inner membrane protein Dld1 orby the transmembrane anchor of l-Mgm1. Remark-ably, Pcp1 does not cleave the transmembraneanchor segment of Dld1 when present in its nativecontext.23 It also does not cleave the transmembraneanchor of l-Mgm1 when it is located in its authenticposition (i.e., about 60 residues N-terminal to theRCR).17–19,22

A flanking sequence located C-terminally to theRCR (the strongly negatively charged D/E stretch)modulates the formation of s-Mgm1. Deletion of thisD/E stretch led to inefficient formation of s-Mgm1.Moreover, reducing the amount of negative chargesprogressively reduced processing efficiency. Also,certain deletions of flanking sequences located N-terminally to the RCR impaired formation of s-Mgm1. However, as we could not entirely rule out

defects in the folding or membrane insertion/correctsorting of these constructs, we decided to excludethem from our analysis. Taken together, our datastrongly suggest that flanking regions are morecritical for intramembrane proteolysis of Mgm1 thanthe sequence of the RCR itself as long as an overallhydrophobicity of this segment is maintained.How can these—in many aspects surprising—

results be explained? In particular, what is thereason for the apparent lack of sequence specificityof the mitochondrial enzyme, which is in contrast tothe situation with other rhomboid proteases? De-spite this apparent lack of specificity, why does theenzyme cleave at a defined site in the authenticsubstrate but does not process other membraneproteins?The mechanism of interaction of the rhomboid

protease with its substrates in mitochondria ispeculiar in that cleavage is linked to the importprocess of the substrate. The enzyme must obtainaccess to the translocating chain either when it ispassing through the TIM23 complex and when thiscomplex opens to allow lateral release, or after it hasbeen fully released and integrated into the innermembrane. The competence for cleavage by Pcp1could be determined by a number of characteristicsof the substrate protein.First, recognition of elements of the substrate

could depend on an interaction of the rhomboidprotease with the hydrophobic RCR, with flankingregions, or with both elements at the same time.Second, the flanking elements could serve todetermine the position of the unfolded substratepolypeptide in the import machinery and could beresponsible for placing the RCR in the transloca-tion channel for a sufficient time and withappropriate affinity to guarantee lateral exit.Interestingly, such a role of sequence elements inthe topogenesis of mitochondrial proteins has beendemonstrated with certain proteins of dual local-ization in the mitochondria and the cytosol. Inthese cases, sequence elements in the precursordetermine which fraction of the precursor, aftercleavage in the matrix by the MPP, moves forwardto reach the matrix and which fraction undergoesretrograde movement into the cytosol.27 Themolecular basis for such mechanism may residein specifically placed folding elements or in thepresence of sequence elements that interfere withthe binding of mitochondrial Hsp70, which is partof the import motor. Furthermore, the rhomboidprotease could conceivably interact with the TIM23translocase in a transient manner. This kind ofscenario is particularly attractive, since the trans-location rate depends on the binding of Hsp70 inthe matrix to the unfolded substrate, and specificsequences can assist in placing the substrate inrelation to the translocase.28 A third scenario,which contains elements of the first two proposals,could explain both apparent promiscuity andspecificity. Flanking sequences could have amajor role in trapping the RCR in the channel ofthe TIM23 complex, as well as in presenting the

190 Intramembrane Proteolysis of Mgm1

RCR in a proper manner to the rhomboid protease.The sequence of the RCR, as long as it ishydrophobic enough, would have a rather modestimpact, explaining why mutations in the flankingregions have strong effects whereas mutationswithin the RCR show no effects. Targeting ofPcp1 to the TIM23 translocase, potentially in itslaterally opened state exposing the substrate,would then allow specific cleavage even in thepresence of low enzyme substrate affinities. Such amechanism is also consistent with the model ofalternative topogenesis,22 which implies that arrestand insertion of the transmembrane segment of l-Mgm1 in the inner membrane would prevent theRCR from reaching the inner membrane andthereby prevent access of the RCR to the activesite in Pcp1.Taken together, this latter working hypothesis

can explain why Pcp1 cleaves certain transmem-brane domains that are normally not processed inthe sequence context of the Mgm1 precursor. Itfurther provides an explanation for the lack of anapparent sequence similarity between the twoknown substrates of Pcp1, Mgm1 and Ccp1,with the exception that the cleavage sites arepresent in a moderately hydrophobic sequence.Processing of the precursor of cytochrome cperoxidase (Ccp1) by Pcp1 depends on the matrixAAA (m-AAA) protease (Yta10/Yta12), whichremoves an N-terminal peptide.26 However, thisdependency is not due to the proteolytic activityof the m-AAA protease but rather due to its ATP-dependent activity in dislocating Ccp1 from themembrane.22 Thus, correct positioning of the Ccp1precursor in the inner membrane appears to beexerted by the m-AAA protease, whereas itappears linked to the import process in the caseof Mgm1. This would also explain why, withCcp1, a stretch of negative charges is not requiredfor intramembrane proteolysis by Pcp1, whereasthis is the case for Mgm1.Our model can also explain the differences

between the mitochondrial rhomboid protease andother rhomboids from different species that recog-nize a specific motif within the cleaved segment ofthe substrate.9,10 Such a motif was first identified inSpitz, which is cleaved by rhomboid-1, therebyreleasing this epidermal growth factor receptorligand. The substrate motif of Spitz comprises twoparts. The first part is more hydrophilic, and thesecond part contains a helix-breaking GA or GGmotif as the primary determinant for cleavage bythe rhomboid protease.9 Interestingly, in Mgm1,two helix-breaking residues are present in its RCR.Nevertheless, these helix-breaking residues werenot required for processing by Pcp1. A recent studyreported on the identification of a recognition motifin rhomboid substrates.15 Positions P4, P1, and P2′were considered to be of particular importance. P1position was proposed to preferentially containsmall residues (Ala and Gly), consistent with anearlier report,11 but also cysteine or serine residues.As stated by the authors of that study, mitochon-

drial substrates, however, do not contain thisrecognition motif.15 They rather contain unfavor-able residues at positions P4 and P1. Evenintroducing leucine, a rather large and helix-stabilizing amino acid, in positions P3 to P1′(Mgm1 LLL variant) neither impaired Pcp1-depen-dent cleavage nor shifted the cleavage site. Takentogether, substrate recognition of Mgm1 does notoccur via its RCR, but neighboring regions areimportant. Such an external determinant was alsoproposed for the human rhomboid proteaseRHBDL-2; recognition of the human anticoagulantprotein thrombomodulin depends on the cytoplas-mic C-terminal domain of the substrate rather thanon its transmembrane domain.29 Thus, substraterecognition of mitochondrial rhomboid proteasesdepends on sites distinct from those where intra-membrane proteolysis takes place. Our data arecompatible with the possible presence of an exositein Pcp1, which would mediate substrate recognitionseparate from the active site. Still, in mitochondrialrhomboids, an intramembrane exosite does notappear to be present—a possibility proposed forother rhomboids.15

Finally, the human mitochondrial rhomboidprotease PARL showed properties very similar tothose of Pcp1 regarding recognition of Mgm1. Thisis in agreement with the observed ability of PARLto functionally complement Pcp1 in yeast.25 Thelatter study further demonstrated that the humanortholog of Mgm1, OPA1, is processed neither byPcp1 nor by PARL, but can be processed in yeastby the m-AAA protease. Interestingly, cytochromec peroxidase (Ccp1) requires, in addition to Pcp1,the m-AAA protease for maturation. Thus, duringcoevolution of substrates and proteases, the m-AAA protease and the mitochondrial rhomboidprotease may both have been required for proces-sing the ancestor of the Mgm1 protein. Later, thehuman protein OPA1 may have lost its dependen-cy on the rhomboid protease, and Mgm1 mayhave lost its dependency on the m-AAA protease.Ccp1 could represent an “ancestral” substrate stillrequiring both proteases. Indeed, neither OPA1nor Ccp1 contains an extended stretch of acidicresidues comparable to the D/E stretch in Mgm1,which could function as a positive determinant forintramembrane proteolysis possibly by mediatinga proper presentation of Mgm1 to the rhomboidprotease upon release from the TIM23 complex.One could further speculate that other innermembrane proteins containing transmembranesegments that are not substrates of Pcp1 mayhave evolved in a way to exclude access to theactive site or an exosite of Pcp1. Taken together,the mitochondrial rhomboid proteases Pcp1 andPARL share conserved properties for recognitionof substrates that are apparently distinct tononmitochondrial rhomboid proteases. In contrast,Mgm1 and OPA1 have apparently diverged suchthat only Mgm1 is still recognized as a substratefor intramembrane proteolysis by mitochondrialrhomboid proteases.

191Intramembrane Proteolysis of Mgm1

Materials and Methods

Generation of sequence alignments of Mgm1 andCcp1 homologs

Sequence alignments of Mgm1 or Ccp1 homologs weregenerated using the CLUSTAL algorithm†. A list of theaccession numbers of sequences used for alignments canbe found in Table S1.

Plasmids, yeast strains, and growth conditions

Two silent mutations creating restriction sites SpeI andNcoI were introduced in pRS315proMgm1fl coding forfull-length Mgm1 under the control of the endogenouspromoter.17 Using this plasmid as template, we amplifiedPCR fragments using primers 1 and 3 and primers 2 and4 (Table S1), respectively, and cloned them into pRS315,30

generating pRS315proMgm1flmut carrying the two newrestriction sites. For cloning of all other Mgm1 variants,PCR fragments were amplified using the primersindicated in Tables S2 and S3. The PCR products werecloned into pRS315proMgm1flmut using indicated re-striction sites (Table S3). Mgm1Dld and Mgm1TM werecloned using a four-primer mutagenesis strategy.Mgm1ΔTMΔD/E was generated by subcloning a SacI/NheI fragment from pRS315proMgm1Δtm112 intoMgm1ΔD/E, introducing the deletion of the TM. PARLsplice variant 1 was amplified from human cDNA usingprimers 36 and 37 (Table S2). The PCR product was cutby MfeI/XhoI and cloned into EcoRI/XhoI cut pYES2(Invitrogen). Mgm11–228ΔD/E-DHFR was cloned intopGEM4 (Promega). To this end, Mgm1ΔD/E cloned inpRS315 was used as template, and the critical region wasamplified using primers 38 and 39 (Table S2). The PCRproduct was cloned into pGEM4-Mgm11–228DHFR17

using the EcoRI/BamHI sites, thereby replacing thewild-type sequence. For cloning of pYX142Ccp1, PCRfragments were amplified using primers 42 and 43 (TableS2) and genomic DNA from the haploid BY4742 wild-typeyeast strain (MATα; Biocat) as template. The resultingfragments were cloned into the pYX142 plasmid using theEcoRI and SacI restriction sites. The plasmid pYX142Ccp1-ΔEDDE was cloned by generating a PCR fragment usingprimers 42 and 44 (Table S2) and genomic DNA from theBY4742 strain. The PCR product was cloned into theEcoRI/ApaI-cut pYX142Ccp1 plasmid. All clonedsequences were confirmed by DNA sequencing.Standard methods were used for the growth and

manipulation of yeast strains.31 A Δmgm1/Δmgm1 strain,obtained from the homozygous diploid deletion library(Invitrogen), and the haploid Δmgm1Δpcp1 strain,obtained by deletion of the PCP1 gene in the Δmgm1/Δmgm1 background using a His6 cassette,32 sporulation,and tetrad dissection, were transformed with the plasmidsfor expression of the different Mgm1 variants. Δccp1 andΔpcp1 deletion strains were obtained from the Matαhaploid deletion library (Invitrogen). The Δyta10/Δyta12strain and antisera raised against Ccp1 were kindlyprovided by Thomas Langer.33 The Δccp1 strain wastransformed with Ccp1-expressing or Ccp1ΔEDDE-expressing plasmids. Cells were cultured on agar platesor liquid-selective media containing 2% (wt/vol) glucoseat 30 °C. For overexpression of PARL, the Δmgm1Δpcp1

†http://www.ch.embnet.org/software/ClustalW.html

strain was transformed with the plasmid pYES2-PARL,and cells were grown on media with 2% (wt/vol) glucoseand, for expression of PARL, on media with 2% (wt/vol)galactose in addition.

Respiratory growth complementation assay

The parental strain was obtained by chromosomaldeletion of MGM1 in a haploid wild-type yeast (w303a)harboring an Mgm1 full-length construct on a pRS316(URA3) plasmid.30 The deletion cassette was obtained byPCR amplification using primers 40 and 41 and genomicDNA of the Δmgm1/Δmgm1 strain as template. Severalvariants of Mgm1 were transformed into the parentalstrain. By growing cells on a medium containing 5-fluoroorotic acid,34 we selected clones that had lost thepRS316 Mgm1 full-length construct. These clones wereanalyzed for their ability to grow on a complete mediumcontaining glycerol as carbon source.

Import of precursor proteins into mitochondria

Radiolabeled precursor proteins were synthesizedusing a coupled reticulocyte lysate transcription transla-tion system (Promega) in the presence of [35S]methionine.Mitochondria were isolated as described previously.35

Import reactions were performed in import buffer[600 mM sorbitol, 50 mM Hepes, 50 mM KCl, 10 mMMgCl2, 2 mM MnCl2, 2.5 mM ethylenediaminetetraaceticacid (EDTA), 2 mM KH2PO4, 5 mM NADH, 2.5 mM ATP,2.5 mM malate, 2.5 mM succinate, 10 mM phosphocrea-tine, 0.1 mg/ml creatine kinase, and 0.1% bovine serumalbumin (pH 7.2)] for 24 min at 25 °C. Fifty micrograms ofmitochondria and 1% (vol/vol) reticulocyte lysate withthe radiolabeled precursor were used per import reaction.When indicated, MPP was inhibited by adding 10 mMEDTA and 2 mM o-phenanthroline. After import, sampleswere diluted, exposed to hypoosmotic buffer [20 mMHepes/KOH (pH 7.4)] to selectively rupture the outermembrane, and treated with proteinase K as indicated. Toassess the stabilities of imported Mgm1 isoforms, weperformed in vitro import and degradation experiments asdescribed previously.36 In brief, after standard import for24 min, nonimported material was digested with 50 μg/ml trypsin for 20 min on ice; trypsin activity was inhibitedby addition of 1 mg/ml soybean trypsin inhibitor andincubation on ice for 5 min; and mitochondria werereisolated by centrifugation (13,000g, 10 min), washedwith HS buffer [20 mM Hepes and 600 mM sorbitol(pH 7.4)], resuspended in import buffer without bovineserum albumin, and incubated at 30 °C for up to 45 min(chase). Finally, samples were analyzed by SDS-PAGE,Western blot analysis, and autoradiography. Westernblots and autoradiographs were scanned and quantifieddensitometrically using an Image Scanner accompaniedby the Image Master 1D Elite software (GE Healthcare).

Identification of cleavage sites by mass spectrometry

Mgm1 variants [WT, A161V, T160A/A161F/T162A(AFA), T160L/A161L/T162L (LLL), TM, and Dld1]expressed in a Δmgm1/Δmgm1 strain were grown inliquid-selective media containing 2% glucose and 2%glycerol (WT, TM, Dld1, and LLL) or 2% glucose and 2%lactate (A161V and AFA). Mitochondria were isolated,and immunoprecipitation of Mgm1 was performed asdescribed earlier.17 In brief, 5–6 mg of mitochondria was

192 Intramembrane Proteolysis of Mgm1

lysed in 10 mMTris/HCl (pH 7.6), 0.5% (wt/vol) Triton X-100, 150 mM NaCl, 5 mM EDTA, and 1 mM phenyl-methylsulfonyl fluoride for 15 min. After a clarifying spin,the supernatant was subjected to immunoprecipitation for3 h at 4 °C using Protein A-Sepharose beads (AmershamBiosciences) preloaded with affinity-purified antibodiesagainst the C-terminus of Mgm1. Samples were elutedfrom the beads with SDS-containing buffer and separatedby SDS-PAGE, and protein bands corresponding to s-Mgm1 were excised from a 10% SDS-PAGE gel. For massspectrometry, the gel pieces were destained with a freshlyprepared solution of 15 mM potassium hexacyanoferrate(III) and 50 mM sodium thiosulfate for 10 min, washedtwice with water, and dehydrated with 75% acetonitrile inwater. Proteins were reduced with 5 mM DTT, alkylatedwith 15 mM iodacetamide (both in 50 mM ammoniumbicarbonate for an incubation time of 1 h at roomtemperature), and in-gel digested with sequencing-grademodified trypsin (Sigma) in 50 mM ammonium bicarbon-ate at 37 °C overnight. After digestion, peptides wereextracted from gel pieces by vortexing with 100 μl of 1%formic acid in 50% acetonitrile (vol/vol) for 1 h. Thesupernatants were collected, dried in Speed Vac, andresolubilized in 20 μl of 0.5% formic acid in 5% acetonitrile(vol/vol). The resulting peptide mixture was subjected toreversed-phase HPLC coupled to a LTQ Orbitrap XL massspectrometer (Thermo Scientific). Briefly, peptides wereseparated in a 90-min HPLC run on a C18 nanoscalecolumn in an acetonitrile gradient from 5% to 45%containing 0.1% formic acid for 30 min. Peptides elutingfrom the column were ionized by electrospray ionizationand analyzed in the mass spectrometer in positive modeby repeating cycles of a survey scan, followed by collision-induced dissociation fragmentation of up to the top fivepeptides with a charge state of 2 or greater using dynamicexclusion. Peak lists were generated using extract_msn(Thermo Scientific) and searched using the Mascot server(version 2.2) against an S. cerevisiae protein databasecontaining 6874 sequences extracted from Swiss-Protdatabase (release 57.7). Alternative sequences with muta-tion sites of the protein Mgm1 were included in thedatabase. The Mascot search was performed using theerror-tolerant search option; the search settings were asfollows: precursor mass tolerance of 10 ppm, fragment iontolerance of 0.6 Da, and one missed cleavage. Themodifications allowed were carbamidomethylation ofcysteine as fixed and oxidation of methionine as option.Mascot results were filtered with a cutoff of individualions scores at 23 (individual ions scores greater than 23indicate identity or extensive homology; pb0.01). After thecleavage sites of the protease Pcp1 had been identified, thesequences of the Mgm1 protein variants in the databasewere shortened according to the identified cleavage site,and Mascot searches were repeated without the error-tolerant search option. Extracted ion current (XIC) signalsof the peptide ions were manually extracted from the MSsurvey scans of corresponding raw files using QualBrowser. XIC areas were integrated using the ICISalgorithm (see Fig. S2).

Acknowledgements

We thank Dr. Ulrich Brandt for scientific adviceand for supporting mass spectrometry analysis, andGabriele Ludwig and Christiane Kotthoff for excel-

lent technical assistance. This work was supportedby the Deutsche Forschungsgemeinschaft Cluster ofExcellence “Macromolecular Complexes,” GoetheUniversity Frankfurt DFG project EXC 115 (A.S.R.),SFB 594 project B8 (A.S.R.), the Center for IntegratedProtein Science Munich (W.N.), and the Universityof Munich FöFoLe program (M.Z.).

Supplementary Data

Supplementary data associated with this articlecan be found, in the online version, at doi:10.1016/j.jmb.2010.06.014

References

1. Koonin, E. V., Makarova, K. S., Rogozin, I. B.,Davidovic, L., Letellier, M. C. & Pellegrini, L. (2003).The rhomboids: a nearly ubiquitous family of intra-membrane serine proteases that probably evolved bymultiple ancient horizontal gene transfers. GenomeBiol. 4, R19.

2. Wasserman, J. D. & Freeman, M. (1998). An auto-regulatory cascade of EGF receptor signaling patternsthe Drosophila egg. Cell, 95, 355–364.

3. Lee, J. R., Urban, S., Garvey, C. F. & Freeman, M.(2001). Regulated intracellular ligand transport andproteolysis control EGF signal activation inDrosophila.Cell, 107, 161–171.

4. Rather, P. N., Ding, X., Baca-DeLancey, R. R. &Siddiqui, S. (1999). Providencia stuartii genes activatedby cell-to-cell signaling and identification of a generequired for production or activity of an extracellularfactor. J. Bacteriol. 181, 7185–7191.

5. Gallio, M., Sturgill, G., Rather, P. & Kylsten, P. (2002).A conserved mechanism for extracellular signaling ineukaryotes and prokaryotes. Proc. Natl Acad. Sci. USA,99, 12208–12213.

6. Carruthers, V. B., Sherman, G. D. & Sibley, L. D.(2000). The Toxoplasma adhesive protein MIC2 isproteolytically processed at multiple sites by twoparasite-derived proteases. J. Biol. Chem. 275,14346–14353.

7. Brossier, F., Jewett, T. J., Sibley, L. D. & Urban, S.(2005). A spatially localized rhomboid proteasecleaves cell surface adhesins essential for invasion byToxoplasma. Proc. Natl Acad. Sci. USA, 102, 4146–4151.

8. Freeman, M. (2008). Rhomboid proteases and theirbiological functions. Annu. Rev. Genet. 42, 191–210.

9. Urban, S. & Freeman, M. (2003). Substrate specificityof rhomboid intramembrane proteases is governed byhelix-breaking residues in the substrate transmem-brane domain. Mol. Cell, 11, 1425–1434.

10. Urban, S., Schlieper, D. & Freeman, M. (2002).Conservation of intramembrane proteolytic activityand substrate specificity in prokaryotic and eukary-otic rhomboids. Curr. Biol. 12, 1507.

11. Akiyama, Y. & Maegawa, S. (2007). Sequence featuresof substrates required for cleavage by GlpG, anEscherichia coli rhomboid protease. Mol. Microbiol. 64,1028–1037.

12. Ye, J., Rawson, R. B., Komuro, R., Chen, X., Dave, U. P.,Prywes, R. et al. (2000). ER stress induces cleavage ofmembrane-bound ATF6 by the same proteases thatprocess SREBPs. Mol. Cell, 6, 1355–1364.

193Intramembrane Proteolysis of Mgm1

13. Lemberg, M. K. & Martoglio, B. (2002). Requirementsfor signal peptide peptidase-catalyzed intramem-brane proteolysis. Mol. Cell, 10, 735–744.

14. Struhl, G. & Adachi, A. (2000). Requirements forpresenilin-dependent cleavage of notch and othertransmembrane proteins. Mol. Cell, 6, 625–636.

15. Strisovsky, K., Sharpe, H. J. & Freeman, M. (2009).Sequence-specific intramembrane proteolysis: identi-fication of a recognition motif in rhomboid substrates.Mol. Cell, 36, 1048–1059.

16. Dimmer, K. S., Fritz, S., Fuchs, F., Messerschmitt, M.,Weinbach, N., Neupert, W. & Westermann, B. (2002).Genetic basis of mitochondrial function and morphol-ogy in Saccharomyces cerevisiae. Mol. Biol. Cell, 13,847–853.

17. Herlan, M., Vogel, F., Bornhövd, C., Neupert, W. &Reichert, A. S. (2003). Processing of Mgm1 by therhomboid-type protease Pcp1 is required for mainte-nance of mitochondrial morphology and of mitochon-drial DNA. J. Biol. Chem. 278, 27781–27788.

18. McQuibban, G. A., Saurya, S. & Freeman, M. (2003).Mitochondrial membrane remodelling regulated by aconserved rhomboid protease. Nature, 423, 537–541.

19. Sesaki, H., Southard, S. M., Hobbs, A. E. & Jensen,R. E. (2003). Cells lacking Pcp1p/Ugo2p, a rhom-boid-like protease required for Mgm1p processing,lose mtDNA and mitochondrial structure in aDnm1p-dependent manner, but remain competentfor mitochondrial fusion. Biochem. Biophys. Res.Commun. 308, 276–283.

20. Nunnari, J., Marshall, W. F., Straight, A., Murray, A.,Sedat, J. W. & Walter, P. (1997). Mitochondrialtransmission during mating in Saccharomyces cerevisiaeis determined by mitochondrial fusion and fission andthe intramitochondrial segregation of mitochondrialDNA. Mol. Biol. Cell, 8, 1233–1242.

21. Zick, M., Duvezin-Caubet, S., Schafer, A., Vogel, F.,Neupert, W. & Reichert, A. S. (2009). Distinct roles ofthe two isoforms of the dynamin-like GTPase Mgm1in mitochondrial fusion. FEBS Lett. 583, 2237–2243.

22. Herlan, M., Bornhövd, C., Hell, K., Neupert, W. &Reichert, A. S. (2004). Alternative topogenesis ofMgm1 and mitochondrial morphology depend onATP and a functional import motor. J. Cell Biol. 165,167–173.

23. Rojo, E. E., Guiard, B., Neupert, W. & Stuart, R. A.(1998). Sorting of D-lactate dehydrogenase to the innermembrane of mitochondria. Analysis of topogenicsignal and energetic requirements. J. Biol. Chem. 273,8040–8047.

24. Guan, K., Farh, L., Marshall, T. K. & Deschenes, R. J.(1993). Normal mitochondrial structure and genome

maintenance in yeast requires the dynamin-likeproduct of the MGM1 gene. Curr. Genet. 24, 141–148.

25. Duvezin-Caubet, S., Koppen, M., Wagener, J., Zick,M., Israel, L., Bernacchia, A. et al. (2007). OPA1processing reconstituted in yeast depends on thesubunit composition of the m-AAA protease inmitochondria. Mol. Biol. Cell, 18, 3582–3590.

26. Esser, K., Tursun, B., Ingenhoven, M., Michaelis, G. &Pratje, E. (2002). A novel two-step mechanism forremoval of a mitochondrial signal sequence involvesthe mAAA complex and the putative rhomboidprotease pcp1. J. Mol. Biol. 323, 835–843.

27. Regev-Rudzki, N., Yogev, O. & Pines, O. (2008). Themitochondrial targeting sequence tilts the balancebetween mitochondrial and cytosolic dual localiza-tion. J. Cell Sci. 121, 2423–2431.

28. Neupert, W. & Herrmann, J. M. (2007). Translocationof proteins into mitochondria. Annu. Rev. Biochem. 76,723–749.

29. Lohi, O., Urban, S. & Freeman, M. (2004). Diversesubstrate recognition mechanisms for rhomboids;thrombomodulin is cleaved by mammalian rhom-boids. Curr. Biol. 14, 236–241.

30. Sikorski, R. S. & Hieter, P. (1989). A system of shuttlevectors and yeast host strains designed for efficientmanipulation of DNA in Saccharomyces cerevisiae.Genetics, 122, 19–27.

31. Sherman, F., Fink, G. R. & Hicks, J. (1986). Methods inYeast Genetics: A Laboratory Course. Cold SpringHarbor Laboratory Press, New York, NY.

32. Wach, A., Brachat, A., Alberti-Segui, C., Rebischung,C. & Philippsen, P. (1997). Heterologous HIS3 markerand GFP reporter modules for PCR-targeting inSaccharomyces cerevisiae. Yeast, 13, 1065–1075.

33. Atorino, L., Silvestri, L., Koppen, M., Cassina, L.,Ballabio, A., Marconi, R. et al. (2003). Loss of m-AAAprotease in mitochondria causes complex I deficiencyand increased sensitivity to oxidative stress inhereditary spastic paraplegia. J. Cell Biol. 163, 777–787.

34. Boeke, J. D., LaCroute, F. & Fink, G. R. (1984). Apositive selection for mutants lacking orotidine-5′-phosphate decarboxylase activity in yeast: 5-fluoro-orotic acid resistance. Mol. Gen. Genet. 197, 345–346.

35. Sirrenberg, C., Bauer, M. F., Guiard, B., Neupert, W. &Brunner, M. (1996). Import of carrier proteins into themitochondrial inner membrane mediated by TIM22.Nature, 384, 582–585.

36. Leonhard, K., Guiard, B., Pellecchia, G., Tzagoloff, A.,Neupert, W. & Langer, T. (2000). Membrane proteindegradation by AAA proteases in mitochondria:extraction of substrates from either membrane sur-face. Mol. Cell, 5, 629–638.