Distinct roles of the two isoforms of the dynamin-like GTPase Mgm1 in mitochondrial fusion

FEBS Letters 583 (2009) 2237–2243

journal homepage: www.FEBSLetters .org

Distinct roles of the two isoforms of the dynamin-like GTPase Mgm1in mitochondrial fusion

Michael Zick a, Stéphane Duvezin-Caubet a,b,*, Anja Schäfer a,c, Frank Vogel d, Walter Neupert a,Andreas S. Reichert a,c,*

a Adolf-Butenandt-Institut für Physiologische Chemie, Ludwig-Maximilians-Universität München, Butenandtstr. 5, 81377 München, Germanyb Institut de Biochimie et Génétique Cellulaires, CNRS-Université Bordeaux2, 33077 Bordeaux, Francec CEF Makromolekulare Komplexe, Mitochondriale Biologie, Fachbereich Medizin, Goethe-Universität Frankfurt am Main, Theodor-Stern-Kai 7, 60590 Frankfurt am Main, Germanyd Max-Delbrück-Centrum für Molekulare Medizin, 13092 Berlin, Germany

a r t i c l e i n f o

Article history:Received 15 April 2009Revised 8 May 2009Accepted 29 May 2009Available online 6 June 2009

Edited by Vladimir Skulachev

Keywords:MitochondriaMembrane fusionMitochondrial DNAMgm1Genetic analysisDynamin-like GTPase

0014-5793/$36.00 � 2009 Federation of European Biodoi:10.1016/j.febslet.2009.05.053

Abbreviations: TM, transmembrane segment of Mgregion; IBM, mitochondrial inner boundary membramembrane; CM, crista membrane; s-Mgm1, short isoisoform of Mgm1; 5-FOA, 5-fluorouracil-6-carboxylicmitochondrial DNA.

* Corresponding authors. Address: Institut de BiochUMR5095 CNRS/Université Bordeaux2, 1, Rue Camillecedex, France. Fax: +33 5 56 99 90 51 (S. Duvezin-CComplexes, Mitochondrial Biology, Goethe University26/5th Floor, Theodor-Stern-Kai 7, 60590 Frankfurt am6301 87162 (A.S. Reichert).

E-mail addresses: [email protected] (S. Duvkgu.de, [email protected] (A.S.

a b s t r a c t

The mitochondrial dynamin-like GTPase Mgm1 exists as a long (l-Mgm1) and a short isoform(s-Mgm1). They both are essential for mitochondrial fusion. Here we show that the isoforms interactin a homotypic and heterotypic manner. Their submitochondrial distribution between inner bound-ary membrane and cristae was markedly different. Overexpression of l-Mgm1 exerts a dominantnegative effect on mitochondrial fusion. A functional GTPase domain is required only in s-Mgm1but not in l-Mgm1. We propose that l-Mgm1 acts primarily as an anchor in the inner membrane thatin concert with the GTPase activity of s-Mgm1 mediates the fusion of inner membranes.� 2009 Federation of European Biochemical Societies. Published by Elsevier B.V. All rights reserved.

1. Introduction

Mitochondria in most eukaryotic cells form a tubular, intercon-nected network that is highly dynamic due to constant reshapingby fusion and fission events [1,2]. Mitochondrial dynamics isimportant for numerous cellular processes, such as cell differenti-ation, development, inheritance of mitochondrial DNA (mtDNA),and apoptosis [3]. A number of severe neuropathies in humansare associated with impairment of fusion and fission of mitochon-dria [4].

chemical Societies. Published by E

m1; RCR, rhomboid cleavagene; OM, mitochondrial outerform of Mgm1; l-Mgm1, long

acid monohydrate; mtDNA,

imie et Génétique Cellulaires,Saint Saëns, 33077 Bordeauxaubet). CEF, MacromolecularFrankfurt am Main, BuildingMain, Germany. Fax: +49 69

ezin-Caubet), [email protected]).

In Saccharomyces cerevisiae three proteins have been reported tobe involved in distinct steps of mitochondrial fusion and shown tophysically interact with each other: Fzo1, Ugo1, and Mgm1. Loss ofany of these factors causes fragmentation of mitochondria and lossof mtDNA [5–8]. Mgm1 is essential for inner membrane fusion to-gether with a high level of GTP [9,10], and was proposed to play arole in the maintenance of cristae [9,11]. It exists as a long(l-Mgm1) and a short (s-Mgm1) isoform, which are present inroughly equal amounts [11,12]. l-Mgm1 is anchored to the innermembrane via an N-terminal transmembrane segment (TM)exposing the bulk of the protein to the intermembrane space(IMS). s-Mgm1, which lacks the N-terminal transmembraneanchor, also resides in the IMS but is only peripherally mem-brane-associated [11,12]. The formation of s-Mgm1 depends onthe mitochondrial rhomboid protease Pcp1 [12,13]. We have previ-ously shown that insertion of Mgm1 into the inner membrane withits TM and processing by Pcp1 at the rhomboid cleavage region(RCR) are competing processes [14]. Deletion of the RCR inMgm1 prevents formation of s-Mgm1 and consequently resultsin formation of l*-Mgm1 only. Conversely, deletion of the TM leadsto complete processing of Mgm1 by Pcp1, selectively generating s*-Mgm1. Importantly, both isoforms of Mgm1 are required for

lsevier B.V. All rights reserved.

2238 M. Zick et al. / FEBS Letters 583 (2009) 2237–2243

maintenance of mitochondrial morphology and inheritance ofmtDNA [12].

The molecular basis of the requirement for both isoforms ofMgm1 in inner membrane fusion is not known. Here we addressedthis question by applying co-immunoprecipitation experiments,quantitative immunoelectron microscopy, and a novel in vivo com-plementation assay that allows the analysis of the two isoforms ofMgm1 in an independent manner.

2. Materials and methods

2.1. Yeast strains, plasmids and in vivo complementation assay

Standard methods were used for growth and manipulation ofyeast strains [15]. The strain FG4/10 (nuclear background of MR6[16]; mtDNA of SDC22 [17]) was transformed withpRS316_Mgm1_fl and the chromosomal copy of MGM1 was thendeleted, resulting in strain FDM316-2 (MATa; ade2-1; his3-11,15;leu2,112; trp1-1; ura3-52; Dmgm1::kanMX4; Darg8::HIS3;[q+ARG8m]; [MGM1+-URA3-CEN]). This strain was used as theparental strain for the complementation assays. The details of theprocedure for the in vivo Mgm1 complementation assay are de-scribed in Fig. 3A. A complete list of plasmids and details of cloningprocedures can be found in Supplementary data.

MTS TM RCR M1 80 94 111 156 169

MPP Pcp1

MTS RCR M1 80 94 111 156 169

MPP Pcp1

MTS TM M1 80 94 111 156 169

MPP

MTS TM RCR1 80 94 111 156 169

GTPase212 43

Mgm

++--

+--+

-++-

--++

+--+

+--+

++--

++--

IP: c αHA c αHA

Load El

Fig. 1. The two isoforms of Mgm1 interact with each other in both homotypic and hetegenerate l*-Mgm1 and s*-Mgm1. (B) Analysis of protein-protein interactions between Mindicated tagged Mgm1 variants were subjected to co-immunoprecipitation using polyEquivalent fractions of loaded material (load) and elution were subjected to SDS–PAGEmitochondrial targeting sequence; GED, GTPase effector domain; middle, dynamin middl

2.2. Immunoelectron microscopy

Quantitative immunoelectron microscopy was performed asdescribed previously [18] using rabbit polyclonal antibodies raisedagainst the peptides H2N-CLGESMKEKFNKMFSGD-COOH (l-Mgm1)and H2N-CKKSYKGVSKNL-COOH (s- and l-Mgm1; [12]). By thatthe submitochondrial distributions of both Mgm1 isoforms andof the long Mgm1 isoform could experimentally be obtained. Fromthis the distribution of the short isoform (s) was calculated usinga subtraction procedure described in detail Supplementary dataand Fig. S1.

2.3. Preparation of cell extracts, cell fractionation, and analysis ofprotein–protein interactions

Preparation of yeast total cell extracts, isolation of mitochon-dria, and co-immunoprecipitation experiments were performedessentially as described previously [19]. Mitochondria isolatedfrom a Dmgm1/Dmgm1 strain (Research Genetics Inc. (Huntsville,US-AL), now Life Technologies Corp. (Carlsbad, US-CA)), expressingthe indicated HA- and FLAG-tagged Mgm1 variants, were used forco-immunoprecipitation experiments. Samples were analyzed bySDS–PAGE and Western blotting using rabbit antibodies againstthe HA or FLAG tags (H6908; F7425; Sigma–Aldrich Corp. (St. Louis,US-MO)).

gm1 fl 1mgM-l+s

gm1 ΔTM 1mgM-*s

gm1 ΔRCR 1mgM-*l

97 kDa

84 kDa

97 kDa

84 kDa

97 kDa

84 kDa

7

middle GED801 893

HA/FLAG

1

αHA

αFLAG

s*-Mgm1-HAs*-Mgm1-FLAGl*-Mgm1-HAl*-Mgm1-FLAG

--++

--++

-++-

-++-c αHA c αHA

ution

↑

↑↑

rotypic manner. (A) Schematic representation of Mgm1 and Mgm1 variants used togm1 isoforms. Solubilized mitochondria of Dmgm1 deletion strains expressing theclonal anti-HA antibodies (aHA) or non-related antibodies as negative control (c).and immunoblotting with antibodies directed against the HA- or FLAG-Tag. MTS,

e domain; MPP, mitochondrial processing peptidase; and Mgm1fl, full-length Mgm1.

-100

0

100

-50 0 50 100 150 200

nm

nm

l-Mgm1

OM/IBMCMbackground

-100

0

100

-50 0 50 100 150 200

nm

nm

s+l-Mgm1

OM/IBMCMbackground

0

10

20

30

40

50

60

70

80

90

100

rela

tive

amou

nt p

er

unit

mem

bran

e le

ngth

(%)

s+l-Mgm1 s-Mgm1 l-Mgm1

OM/IBMCM

Fig. 2. The two isoforms of Mgm1 show a distinct distribution across the mitochondrial inner membrane. (A, B) Localization of Mgm1 by immuno-EM and in silicoaccumulation of gold particles onto an empiric model. S. cerevisiae wild-type cells were grown to early log phase in liquid complete media containing 2% lactate, chemicallyfixed, cryosectioned, and immunogold labeled. The location of gold particles found in mitochondria showing clearly resolvable CMs connected by crista junctions to the IBMwere plotted onto a single, empirically determined, drawn to scale model. (A) Distribution of both Mgm1 isoforms determined with an antibody directed against the C-terminus of Mgm1 (data are taken from Ref. [18]; represented in adapted manner). (B) Distribution of the l-Mgm1 isoform using an l-Mgm1 specific antibody. (C)Quantification of the distribution of the Mgm1 isoforms across the sub-compartments of the mitochondrial inner membrane: OM/IBM and CM. For details see Supplementarydata and Fig. S1.

+Arg -Arg +Arg -Arg

Mgm1 fl empty plasmid

s*-Mgm1 l*-Mgm1

s*+ l*-Mgm1 dnDnm1 (S42N)

s*-Mgm1 mutX + l*-Mgm1 mutY

5-FOA +Arg 5-FOA -Arg

SD −Leu −Trp +Arg +Ura1-2 days

3-5 days

Mgm1functional

Mgm1non-functional

+ρ clones

+No ρ clones

loss of+[MGM1 -URA3-CEN]

plasmids with mutants to be tested

+ mΔmgm1 Δarg8 [ ρ _ARG8 ]+[MGM1 -URA3-CEN]

+ s*-Mgm1 mutX+ l*-Mgm1 mutY

+ mΔmgm1 Δarg8 [ ρ _ARG8 ]+[MGM1 -URA3-CEN]

+ mΔmgm1 Δarg8 [ ρ _ARG8 ]+ s*-Mgm1 mutX+ l*-Mgm1 mutY

Fig. 3. Complementation assay to assess s- and l-Mgm1 functions in an independent manner. (A) Graphical representation of complementation assay. After transformation ofthe FDM316-2 strain with a single Mgm1 variant to be tested or combinations of them, cells were plated on SD �Arg �Ura medium lacking the respective amino acids (Leu,Trp) and incubated for three days at 28 �C. A dozen of transformants were picked, pooled, and cultured in liquid SD medium +Arg +Ura for 1–2 days at 28 �C to allow for theloss of the [MGM1+-URA3-CEN] plasmid while keeping selective pressure on the other plasmids introduced (LEU2, TRP1). Cells were washed once with sterile water, plated ata density of 105 cells/plate on 5-FOA medium +Arg or �Arg, and incubated 3–5 days at 28 �C in order to select for cells that had lost the [MGM1+-URA3-CEN] plasmid and tocheck for the maintenance of mtDNA as depicted. (B) The indicated constructs were tested to validate the assay. Resulting 5-FOA plates supplemented with arginine (+Arg) ornot (�Arg) are shown. dnDnm1 (S42N), dominant negative variant of the fission factor Dnm1.

M. Zick et al. / FEBS Letters 583 (2009) 2237–2243 2239

2240 M. Zick et al. / FEBS Letters 583 (2009) 2237–2243

2.4. Quantification of mitochondrial morphology

Mitochondrial morphology was analyzed by standard fluores-cence microscopy as described previously [14].

3. Results

3.1. The isoforms of Mgm1 interact with each other in homo- andheterotypic manner

Mgm1 was reported to undergo self-self interactions in trans[9]. However, the individual role of each Mgm1 isoform in thisprocess has not been addressed so far. To elucidate this weexpressed various combinations of two differently tagged vari-ants of s*-Mgm1 and/or l*-Mgm1 in a strain lacking endogenousMgm1, and performed co-immunoprecipitation experiments(Fig. 1). s*-Mgm1-HA physically interacted with s*-Mgm1-FLAG,and l*-Mgm1-HA with l*-Mgm1-FLAG. Thus, both isoforms canundergo homotypic interactions even in the absence of therespective other isoform. Furthermore, s*-Mgm1-HA interactedwith l*-Mgm1-FLAG, and l*-Mgm1-HA with s*-Mgm1-FLAGshowing that both isoforms also interact in a heterotypicmanner (Fig. 1B). Still, only about 1–5% of each isoform wasobserved to form homotypic or heterotypic interactions(Fig. 1B).

WT --

↑↑

↑↑-

- l*-Mgm1s*-Mgm1

97 kDa

84 kDa

25 kDa

l* (α FLAG)

s* (α HA)

αTim23

αMgm1

l* / s*

l*↑ / s*

l*↑ / s*

↑ l* / s*↑

0102030405060708090

100

% o

f cel

ls

0.1

1

0 5tim

-A

OD

600n

m l*↑ ↑ / s*l* ↑ / s*

l*↑ / s*

l*↑ / s*

Fig. 4. The long isoform of Mgm1 exerts a dominant negative effect on mitochondrialconstructs were analyzed by SDS–PAGE and immunoblotting using indicated antibodpromoter (–) or overexpressed (") as indicated. The mitochondrial protein Tim23 was usewere tested in the complementation assay (see Fig. 3A). Sectors and enlarged insets of tfrom plates in panel B and grown in liquid SD media containing arginine (+Arg) or notstrains shown in panel B.

3.2. The two isoforms of Mgm1 show a differential distribution acrossthe inner membrane

Next we asked how the two isoforms of Mgm1 are distributedacross the distinct sub-compartments of the mitochondrial innermembrane, namely the inner boundary membrane (IBM) and thecrista membrane (CM). We performed quantitative immunoelectronmicroscopy using antibodies directed against either the C-terminusof Mgm1 (recognizing both s- and l-Mgm1) or against a region lo-cated between TM and RCR (recognizing selectively l-Mgm1). Previ-ously, we have shown that detection of both s- and l-Mgm1 revealeda modest enrichment in the IBM compared to the CM (adapted rep-resentation in Fig. 2A; [18]). Here, we determined the distribution ofl-Mgm1 and found it strongly enriched in the CM (Fig. 2B and C). Wecalculated the expected amount of s-Mgm1 in IBM and CM by sub-traction (for details see Supplementary data and Fig. S1). This re-vealed a strong enrichment of s-Mgm1 in the region correspondingto the IBM/OM, the two membranes being indistinguishable by elec-tron microscopy due to their close apposition (Fig. 2C). Taken to-gether, both isoforms show a strikingly asymmetric, only partlyoverlapping distribution over the mitochondrial inner membrane.This observation supports the idea of distinct roles of the two iso-forms in mitochondrial fusion. Furthermore, it may partly explainwhy only substoichiometric amounts of the two Mgm1 isoformswere found in stable heteromeric complexes in our co-immunopre-cipitation experiments (Fig. 1B).

tubular

fragmented/aggregated

Δmgm1 -

- -- ↑↑

↑↑l*-Mgm1s*-Mgm1WT

+Arg

10e (h)

0 5 10time (h)

rg

0.1

1O

D60

0nm l*↑ ↑ / s*

l* ↑ / s*l*↑ / s*

l*↑ / s*

fusion. (A) Total cell extracts of Dmgm1 cells harboring different combinations ofies. Mgm1 isoforms were either expressed under the control of the endogenousd as loading control. (B) The different combinations of Mgm1 variants from panel A

he 5-FOA-Arg plates are shown. Scale bar 5 mm. (C) Growth curve of clones picked(�Arg). (D) Quantification of mitochondrial morphology of wild-type, Dmgm1, and

M. Zick et al. / FEBS Letters 583 (2009) 2237–2243 2241

3.3. A novel assay allows for assessing the function of individual Mgm1isoforms

We developed an assay to study the function of the Mgm1 iso-forms individually. It is based on the ability of Mgm1 to rescuemaintenance of mtDNA in a Dmgm1 strain. To obtain an unambig-uous readout independent of the respiratory capacity of the cells,we deleted MGM1 in a strain in which the auxotrophy markerARG8 was chromosomally deleted and integrated into the mito-chondrial genome yielding ARG8m [17]. In this strain growth onmedia devoid of arginine depends on the integrity and mainte-nance of mtDNA independent of the carbon source as the mtDNAencodes for Arg8 which is required for arginine biosynthesis. Cellslacking mtDNA (q�/0) can thus only grow on medium supple-mented with arginine. Using a plasmid shuffling strategy(Fig. 3A), a wild-type copy of Mgm1 was replaced by Mgm1 vari-ants to be tested. The resulting strains were analyzed for their abil-ity to grow on media devoid of arginine in order to test whethermtDNA can be maintained or not (Fig. 3A).

To validate the assay we performed the following control exper-iments. A plasmid carrying a wild-type version of Mgm1 was ableto prevent loss of mtDNA as indicated by the approximately equalnumber of colonies on medium containing arginine or not (Fig. 3B).On the contrary, an empty plasmid was not able to rescue the cellsas colonies were only obtained on plates containing arginine(Fig. 3B). To confirm that the assay monitors the role of Mgm1 inmitochondrial fusion, and not other unknown functions ofMgm1, we expressed a dominant negative variant of the fissionfactor Dnm1 (S42N) [20]. The simultaneous disruption of Mgm1and Dnm1 function is known to result in the restoration of mito-chondrial tubules and proper mtDNA maintenance [21]. Indeed,equivalent numbers of colonies were obtained on plates containingor not containing arginine (Fig. 3B). In a previous study [12], partialcomplementation was obtained by co-expressing engineeredl*-Mgm1 (DRCR) and a variant of s-Mgm1 generated by fusingthe cleavable targeting signal of cytochrome b2 (aa 1-167) tos-Mgm1 (aa 161-902). Here we tested the functionality of the engi-neered s*-Mgm1 (DTM) and l*-Mgm1 (DRCR) variants containingC-terminal tags (HA or FLAG; see Fig. 1A and Supplementary Table1). Only the co-expression of both isoforms rendered Mgm1 fullyfunctional, but neither of the two isoforms alone was able to main-tain mtDNA (Fig. 3B). We conclude that the ARG8m based comple-mentation assay allows a reliable and fast assessment of thefunctionality of the two Mgm1 isoforms independently.

3.4. The ratio of the short to the long isoform of Mgm1 is crucial forfusion competence of mitochondria

As the ratio of the two Mgm1 isoforms appears to be tightly reg-ulated, we asked whether altering the relative levels of the two iso-forms affects Mgm1 function. We expressed differentcombinations of s*-Mgm1 and l*-Mgm1 either under control ofthe endogenous Mgm1 promoter or of a strong constitutive pro-moter for overexpression (Fig. 4A). Applying the complementationassay described above (Fig. 3A) we noticed that simultaneous over-expression of both l*- and s*-Mgm1, or overexpression of only s*-Mgm1 in the presence of wild-type levels of l*-Mgm1 did not affectthe ability of the constructs to take over Mgm1 function (Fig. 4B).However, overexpression of l*-Mgm1 while maintaining wild-typelevels of s*-Mgm1 had a strong impact on complementation. Thenumber of colonies obtained on the 5-FOA plates lacking argininewas similar to what was obtained with the other combinations,however, their size was drastically smaller (Fig. 4B). This defectin growth was also observed when the respective strains weregrown in liquid media lacking arginine (Fig. 4C). Apparently, anincrease of the level of l*-Mgm1 relative to s*-Mgm1 impairs

maintenance of mtDNA to a considerable extent. In accordancewith these results, microscopic inspection of the mitochondrialnetwork revealed that the relative increase of l*-Mgm1 levels re-sulted in a phenotype similar to Dmgm1. In contrast, cells withall other combinations tested were very similar to wild-type(Fig. 4D). In conclusion, a shift of the ratio of Mgm1 isoforms to-wards l-Mgm1 had a strong dominant negative effect on mitochon-drial fusion, whereas a shift towards s-Mgm1 had no detectableeffect.

3.5. A functional GTPase domain is only required in the short but not inthe long isoform of Mgm1

Furthermore, we investigated whether mutations in the GTPasedomain (Fig. 5A) in only one of the Mgm1 isoforms have distincteffects on mitochondrial fusion. To this end, we introduced theG430D mutation (mgm1-5; temperature sensitive [11]) in s*-Mgm1, l*-Mgm1, or full-length Mgm1. The single variants ordifferent combinations thereof were analyzed with the comple-mentation assay. At permissive temperature (24 �C) with allcombinations similar numbers of colonies were obtained on 5-FOA plates lacking arginine (data not shown). Individual cloneswere assessed for respiratory growth on glycerol medium atpermissive (24 �C) and non-permissive (37 �C) temperature.Mgm1-G430D full length, the combination of s*-Mgm1-G430Dand l*-Mgm1-G430D, and the combination s*-Mgm1-G430D andl*-Mgm1 were not able to grow under respiratory conditions atthe non-permissive temperature (Fig. 5B). However, the combina-tion of s*-Mgm1 and l*-Mgm1-G430D showed wild-type like respi-ratory growth at both permissive and non-permissive temperature(Fig. 5B). We conclude that a functional GTPase domain is strictlyrequired in s*-Mgm1 but not in l*-Mgm1 for maintenance ofmtDNA. To further corroborate this conclusion we tested muta-tions known to affect either binding (RasG12V, K244A, D409W)or hydrolysis (T265A) of GTP [22,23] separately in s*-Mgm1 or l*-Mgm1 (Fig. 5A). We excluded impaired expression or reduced sta-bility of the constructs by determining the steady state levels of thedifferent constructs (Fig. 5C). All mutations within the GTPase do-main proved to inactivate Mgm1 function when introduced into s*-Mgm1, while they had no impact when they were introduced intol*-Mgm1. This was true using the ARG8m based assay (Fig. 5D) aswell as when mitochondrial morphology was assessed (Fig. 5E). Ta-ken together, both isoforms of Mgm1 are required for fusion ofmitochondria, but an active GTPase domain is only required inthe short isoform. This finding further supports our conclusion thatboth isoforms have distinct functions in inner membrane fusion.

4. Discussion

In light of its molecular role in inner membrane fusion of mito-chondria several aspects are remarkable for Mgm1. First of all, it isa dynamin-like GTPase and thus belongs to a class of conservedlarge GTPases that are well known for their role in membrane fis-sion, e.g. during endocytosis, rather than in membrane fusion pro-cesses [22]. Second, Mgm1 exists in two different isoforms both ofwhich are essential for Mgm1 function [12], an unknown propertyof dynamins or non-orthologous dynamin-like proteins. Third, thelong isoform is an integral inner membrane protein harboring anN-terminal transmembrane segment. Thus, it belongs to the sub-group of dynamin-like proteins that mediate membrane bindingvia a transmembrane segment instead of the Pleckstrin homologydomain found in classical dynamins [22].

The present study yielded several novel and surprising insightsinto the molecular mechanisms of Mgm1-mediated membrane fu-sion. A major and unexpected result is that the two isoforms of

2242 M. Zick et al. / FEBS Letters 583 (2009) 2237–2243

Mgm1 are distinct in several so far unknown aspects. First, bothisoforms are differentially distributed between the CM and theIBM, the two sub-compartments of the inner membrane. Second,in l-Mgm1 a functional GTPase domain is not required whereasin s-Mgm1 it is. Thus, Mgm1 is to our knowledge the first exampleof a dynamin-like GTPase in which at least one isoform has aGTPase-independent function. We further show that relative over-expression of l-Mgm1, but not of s-Mgm1, results in a severe dom-inant negative phenotype. Taken together, our data stronglysuggest that the two isoforms have distinct roles in mitochondrialfusion.

Our observations that l-Mgm1 and s-Mgm1 can in principleform a stable heteromeric protein complex but in addition showdistinct distributions across the inner membrane might appearcontradictory at first sight. However, this heteromeric protein

97 kDa

84 kDa

l* (αF

s* (α

αMgm1

WT K244A

T265A

D409W

RasG12

V

K244A

T265A

D409W

RasG12

V

emptyvector

WT

s*-G430D+

l*-G430D

+l*-G430D

s*

s*-G430D +l*

G430Dfl

YPG 24°C YPG 37°C

0

10

20

30

40

5060

70

80

90

100

% o

f cel

ls

Δmgm1

WT

tubularfragmenaggrega

K244A

T265A

D409W

RasG12

V

K244A

T265A

D409W

mutation in s* mutation in l*

mutation in s* mutation in l*

Fig. 5. The short but not the long isoform of Mgm1 requires a functional GTPase domain.motifs are shown in bold. Introduced mutations are indicated by arrows. (B) Growth of asensitive mutation of Mgm1 (mgm1-5/G430D) on respiratory media (YPG). Left panel, sc(24 �C); right panel, growth at non-permissive temperature (37 �C). (C) Total yeast cellcarrying different GTPase-mutations were subjected to immunoblotting using indicated aindicated. (D) Strains harboring the different GTPase-mutations in one of the two isoformof mitochondrial morphology of wild-type, Dmgm1, and strains harboring the different

complex might represent only a minor subpopulation which is lo-cated in one or both sub-compartments of the inner membrane.This is consistent with our results showing that this complex wasisolated only in substoichiometric amounts and that the locationsof both isoforms overlap to some, albeit minor, extent within theinner membrane. In addition, the heteromeric complexes mightbe formed only when s-Mgm1 and l-Mgm1 transiently meet dur-ing membrane fusion, while under steady state conditions the iso-forms are mostly located in different sub-compartments. Thiswould imply a dynamic redistribution of l-Mgm1 from the CM tothe IBM and could well coincide with remodeling of cristae.Thereby l-Mgm1 may reach the site of inner membrane dockingand fusion. Docking of inner membranes itself might then bemediated via homotypic interactions between l-Mgm1 in trans(l–l) which is consistent with our identification of a homomeric

LAG)

HA)

+Arg -Arg +Arg -Arg

K244A

T265A

D409W

RasG12V

Mutation in s*-Mgm1

ted/ted

RasG12

V

Mutation in l*-Mgm1

(A) Schematic representation of the GTPase domain of Mgm1. Key residues of G1–G4Dmgm1 strain complemented with indicated Mgm1 variants carrying a temperaturehematic representation of sectors; middle panel, growth at permissive temperatureextracts of a Dmgm1 strain expressing s*-Mgm1-HA and l*-Mgm1-FLAG isoforms

ntibodies. The mutation is always only present in one of the two isoforms, s* or l*, ass of Mgm1 were subjected to the complementation assay (Fig. 3A). (E) QuantificationGTPase-mutations in one of the two isoforms of Mgm1 only, s* or l*, as indicated.

M. Zick et al. / FEBS Letters 583 (2009) 2237–2243 2243

l-Mgm1 complex. Also the presence of an N-terminal transmem-brane segment and the dispensability of a functional GTPase do-main in l-Mgm1 suggest a primary role of l-Mgm1 in anchoringthe fusion machinery to the inner membrane. l-Mgm1 may act asa docking receptor for s-Mgm1, eventually involving also other fu-sion factors such as Fzo1 or Ugo1. Overexpression of l-Mgm1 mighttitrate out one of these factors, a reaction that would explain itsdominant negative activity. Consistent with its prevailing locationin the CM, l-Mgm1 could further help to stabilize cristae structures,a function of Mgm1 proposed previously [9,11]. s-Mgm1, eventu-ally as part of a transiently formed heteromeric complex, wouldthen regulate membrane fusion and disassembly of the fusionmachinery. s-Mgm1 requires a functional GTPase domain, and thusappears to regulate inner membrane fusion in a GTP-dependentmanner. Still, it is not known yet which step exactly is mediatedby s-Mgm1 and how GTP binding and hydrolysis are involved. Inthis working model l-Mgm1 and s-Mgm1 would act in a consecu-tive manner, explaining the observed requirement for both iso-forms of Mgm1. Taken together, our study shows that bothisoforms have distinct roles in mitochondrial membrane fusionand provides a framework for future studies in order to dissect fur-ther the molecular details of this complex interplay.

Acknowledgments

We are grateful to Ilona Dietze and Christiane Kotthoff forexcellent technical assistance and Dr. Soledad Funes for discussionand critically reading the manuscript. We thank Dr. Jean-Paul diRago for the generous contribution of strain FG4/10 and Dr. JanetShaw for providing the pYEP213-DNM1(S42N) plasmid. This workwas supported by the University of Munich FöFoLe program (M.Z.),the Deutsche Forschungsgemeinschaft SFB 594 project B8 (A.R.),the Center for Integrated Protein Science Munich (W.N.), and theCluster of Excellence ‘‘Macromolecular Complexes” at the GoetheUniversity Frankfurt DFG project EXC 115 (A.R.).

Appendix A. Supplementary data

Supplementary data associated with this article can be found, inthe online version, at doi:10.1016/j.febslet.2009.05.053.

References

[1] Bereiter-Hahn, J. and Voth, M. (1994) Dynamics of mitochondria in living cells:shape changes, dislocations, fusion, and fission of mitochondria. Microsc. Res.Technol. 27, 198–219.

[2] Nunnari, J., Marshall, W.F., Straight, A., Murray, A., Sedat, J.W. and Walter, P.(1997) Mitochondrial transmission during mating in Saccharomyces cerevisiaeis determined by mitochondrial fusion and fission and the intramitochondrialsegregation of mitochondrial DNA. Mol. Biol. Cell 8, 1233–1242.

[3] Okamoto, K. and Shaw, J.M. (2005) Mitochondrial morphology and dynamicsin yeast and multicellular eukaryotes. Annu. Rev. Genet. 39, 503–536.

[4] DiMauro, S. and Schon, E.A. (2008) Mitochondrial disorders in the nervoussystem. Annu. Rev. Neurosci. 31, 91–123.

[5] Guan, K., Farh, L., Marshall, T.K. and Deschenes, R.J. (1993) Normalmitochondrial structure and genome maintenance in yeast requires thedynamin-like product of the MGM1 gene. Curr. Genet. 24, 141–148.

[6] Hermann, G.J., Thatcher, J.W., Mills, J.P., Hales, K.G., Fuller, M.T., Nunnari, J. andShaw, J.M. (1998) Mitochondrial fusion in yeast requires the transmembraneGTPase Fzo1p. J. Cell Biol. 143, 359–373.

[7] Rapaport, D., Brunner, M., Neupert, W. and Westermann, B. (1998) Fzo1p is amitochondrial outer membrane protein essential for the biogenesis offunctional mitochondria in Saccharomyces cerevisiae. J. Biol. Chem. 273,20150–20155.

[8] Sesaki, H. and Jensen, R.E. (2001) UGO1 encodes an outer membrane proteinrequired for mitochondrial fusion. J. Cell Biol. 152, 1123–1134.

[9] Meeusen, S., DeVay, R., Block, J., Cassidy-Stone, A., Wayson, S., McCaffery, J.M.and Nunnari, J. (2006) Mitochondrial inner-membrane fusion and cristamaintenance requires the dynamin-related GTPase Mgm1. Cell 127, 383–395.

[10] Meeusen, S., McCaffery, J.M. and Nunnari, J. (2004) Mitochondrial fusionintermediates revealed in vitro. Science 305, 1747–1752.

[11] Wong, E.D., Wagner, J.A., Gorsich, S.W., McCaffery, J.M., Shaw, J.M. andNunnari, J. (2000) The dynamin-related GTPase, Mgm1p, is an intermembranespace protein required for maintenance of fusion competent mitochondria. J.Cell Biol. 151, 341–352.

[12] Herlan, M., Vogel, F., Bornhovd, C., Neupert, W. and Reichert, A.S. (2003)Processing of Mgm1 by the rhomboid-type protease Pcp1 is required formaintenance of mitochondrial morphology and of mitochondrial DNA. J. Biol.Chem. 278, 27781–27788.

[13] McQuibban, G.A., Saurya, S. and Freeman, M. (2003) Mitochondrial membraneremodelling regulated by a conserved rhomboid protease. Nature 423, 537–541.

[14] Herlan, M., Bornhovd, C., Hell, K., Neupert, W. and Reichert, A.S. (2004)Alternative topogenesis of Mgm1 and mitochondrial morphology depend onATP and a functional import motor. J. Cell Biol. 165, 167–173.

[15] Sherman, F., Fink, G.R. and Hicks, J.B. (1986) Laboratory Course Manual forMethods in Yeast Genetics, Cold Spring Harbor Laboratory, New York, NY.

[16] Rak, M. et al. (2007) Yeast cells lacking the mitochondrial gene encoding theATP synthase subunit 6 exhibit a selective loss of complex IV and unusualmitochondrial morphology. J. Biol. Chem. 282, 10853–10864.

[17] Duvezin-Caubet, S., Rak, M., Lefebvre-Legendre, L., Tetaud, E., Bonnefoy, N. anddi Rago, J.P. (2006) A ‘‘petite obligate” mutant of Saccharomyces cerevisiae:functional mtDNA is lethal in cells lacking the delta subunit of mitochondrialF1-ATPase. J. Biol. Chem. 281, 16305–16313.

[18] Vogel, F., Bornhovd, C., Neupert, W. and Reichert, A.S. (2006) Dynamicsubcompartmentalization of the mitochondrial inner membrane. J. Cell Biol.175, 237–247.

[19] Duvezin-Caubet, S. et al. (2006) Proteolytic processing of OPA1 linksmitochondrial dysfunction to alterations in mitochondrial morphology. J.Biol. Chem. 281, 37972–37979.

[20] Otsuga, D., Keegan, B.R., Brisch, E., Thatcher, J.W., Hermann, G.J., Bleazard, W.and Shaw, J.M. (1998) The dynamin-related GTPase, Dnm1p, controlsmitochondrial morphology in yeast. J. Cell Biol. 143, 333–349.

[21] Sesaki, H., Southard, S.M., Yaffe, M.P. and Jensen, R.E. (2003) Mgm1p, adynamin-related GTPase, is essential for fusion of the mitochondrial outermembrane. Mol. Biol. Cell 14, 2342–2356.

[22] Praefcke, G.J. and McMahon, H.T. (2004). The dynamin superfamily: universalmembrane tubulation and fission molecules? Nat. Rev. Mol. Cell Biol. 5, 133–47. <http://www2.mrc-lmb.cam.ac.uk/Dynamin/GTPase.html>.

[23] Wong, E.D., Wagner, J.A., Scott, S.V., Okreglak, V., Holewinske, T.J., Cassidy-Stone, A. and Nunnari, J. (2003) The intramitochondrial dynamin-relatedGTPase, Mgm1p, is a component of a protein complex that mediatesmitochondrial fusion. J. Cell Biol. 160, 303–311.