Research Article

Mitochondrial Fusion Intermediates Revealed in Vitro

See allHide authors and affiliations

Science  17 Sep 2004:
Vol. 305, Issue 5691, pp. 1747-1752
DOI: 10.1126/science.1100612

Abstract

The events that occur during the fusion of double-membraned mitochondria are unknown. As an essential step toward determining the mechanism of mitochondrial fusion, we have captured this event in vitro. Mitochondrial outer and inner membrane fusion events were separable and mechanistically distinct, but both required guanosine 5′-triphosphate hydrolysis. Homotypic trans interactions of the ancient outer transmembrane guanosine triphosphatase, Fzo1, were required to promote the fusion of mitochondrial outer membranes, whereas electrical potential was also required for fusion of inner membranes. Our conclusions provide fundamental insights into the molecular events driving mitochondrial fusion and advance our understanding of the evolution of mitochondrial fusion in eukaryotic cells.

In eukaryotic cells, homotypic mitochondrial fusion events occur continuously throughout the cell cycle. These fusion events are balanced by mitochondrial fission events, and together they function to create a distributable and responsive mitochondrial compartment (1). As a result of their endosymbiotic origin, mitochondria possess two structurally and functionally distinct membranes. During mitochondrial fusion, two pairs of these membranes are coordinately and accurately fused. The mechanism by which these four membranes rapidly fuse into two continuous, yet distinct, bilayers is not understood. Mitochondrial membranes have not been observed to undergo fusion with other endomembranes, and conserved core secretory fusion components, such as soluble N-ethylmaleimide–sensitive factor attachment protein receptors, are not required for homotypic mitochondrial fusion. These features suggest that mitochondrial fusion evolved independently from secretory fusion and that the fundamental steps in membrane fusion are mediated in a mechanistically unique manner.

Three proteins in yeast—Ugo1, Mgm1, and Fzo1—are required for mitochondrial fusion in vivo based on cytological assays for mitochondrial morphology and mitochondrial content mixing during yeast mating. Ugo1 is an outer membrane protein, with several predicted transmembrane regions, and is a member of the mitochondrial transport family (2, 3). Fzo1 and Mgm1 are highly conserved large guanosine triphosphatases (GTPases) and are components of the mitochondrial outer and inner membranes, respectively (46). Mgm1 is a dynamin-related GTPase protein (DRP) and is tethered to the inner membrane and localized to the intermembrane space (6). Genetic analysis indicates that Mgm1 interacts with itself and that its GTPase domain is required for fusion, suggesting that it functions in mitochondrial fusion in a similar way to other DRPs, as a self-assembling GTPase (7, 8). The Fzo1 GTPase is an integral membrane protein that spans the outer membrane twice; it has a small loop of functionally important amino acids in the intermembrane space, and its N-terminal GTPase domain and C terminus are in the cytosol (9). The GTPase domain of Fzo1 is most similar to a family of eubacterial GTPases, suggesting that it is derived from the mitochondrial ancestral prokaryote (10). Among eukaryotic GTPases, however, the Fzo1 GTPase domain is most closely related to the DRP family GTPase domain (11). Members of the Fzo1 GTPase family also possess, like DRPs, regions predicted to form coiled-coil structures, raising the possibility that they are modified but bone fide members of the DRP family and function in mitochondrial fusion through self-interaction and assembly.

Conserved components that mediate mitochondrial fusion and mitochondrial fission play an important role in the transmission of the cell death response in mammalian cells (1214). Mutations in two conserved human fusion proteins, Opa1, the Mgm1 ortholog, and Mfn2, an Fzo1 ortholog, result in two types of neurodegenerative diseases: dominant optic atrophy and Charcot-Marie-Tooth neuropathy, respectively, consistent with a role of these proteins in cell death (1517).

Mgm1, Fzo1, and Ugo1 have been shown to interact physically (7, 8, 18), suggesting that outer and inner membranes share a common fusion apparatus, but the exact role of these components and their interactions in fusion are unknown. Indeed, discrete steps in mitochondrial fusion have not been described, primarily because this event has proven difficult to replicate in vitro.

Mitochondrial fusion in vitro. To understand the molecular events required for mitochondrial fusion, we developed an in vitro cytological assay that measures fusion of outer and inner mitochondrial membranes (19). Equivalent amounts of mitochondria containing either matrix-targeted green fluorescent protein (m-GFP) or matrix-targeted Discosoma red fluorescent protein (m-dsRed) were mixed together, concentrated by centrifugation, and resuspended in the presence of cytosolic extract, adenosine 5′-triphosphate (ATP), guanosine 5′-triphosphate (GTP), and an energy regeneration system. Fusion of both the outer and inner membranes was assessed by matrix content mixing, reflected by the colocalization of m-GFP and m-dsRed into single individually resolvable mitochondrial structures in three dimensions (Fig. 1, A and B, and movie S1). Fusion was quantified by determining the number of mitochondria containing both m-GFP and m-dsRed and dividing by the total number of mitochondria analyzed (n > 200). In vitro mitochondrial fusion occurred in a time- and temperature-dependent manner, consistently attaining 10 to 15% by 30 min at 22°C (Fig. 1, B and C, and fig. S1A). Because we could only detect m-GFP and m-dsRed content-mixing events, the efficiency of mitochondrial fusion in our assay was substantially higher than 10 to 15%. We also observed successive fusion events at time points greater than 30 min, which indicates that our content-mixing assay was an underestimate of total fusion efficiency. Mitochondrial fusion was highly dependent on exogenous nucleotide triphosphates (NTPs) and an energy regeneration system (Fig. 2) but not dependent on the addition of cytosolic extract, indicating that non–mitochondrial-associated proteins were not required for fusion (fig. S1B).

Fig. 1.

Mitochondrial fusion in vitro. (A) Schematic of in vitro fusion assay. (B) Fluorescent images of a fusion reaction with m-GFP and m-dsRed mitochondria. Arrows indicate fused mitochondria. Scale bar, 2 μm. (C) Fixed time-point analysis of in vitro mitochondrial fusion. (D) Comparison of mitochondrial diameter along long axis of unfused (gray) and fused (black) mitochondria labeled with m-GFP and m-dsRed. Total number of mitochondria analyzed in each case was 56. (E and F) Immunoelectron microscopy analysis of m-GFP and m-dsRed localization in mitochondria from 0-min (E) and 20-min (F) fusion reactions. Large gold particles (red arrowheads) indicate immunolabeled dsRed and small gold particles (black arrowheads) indicate immunolabeled GFP.

Fig. 2.

Mitochondrial fusion in vitro requires GTP hydrolysis and inner mitochondrial membrane potential.In vitro fusion reactions were conducted as described and fusion efficiency is expressed as a percentage of the standard reaction, which contains an energy regeneration system (ERS), 1 mM ATP, and 0.5 mM GTP (19).Stage 1 mitochondria were resuspended and placed under stage 2 conditions in the presence of nonhydrolyzable NTP analogs and treatments as described (19). Error bars show mean + SD. VAL, valinomycin; NIG, nigericin.

Three criteria indicate that our matrix content-mixing assay was an accurate reflection of mitochondrial fusion. First, mitochondria containing both m-GFP and m-dsRed were larger in diameter along their long axis than mitochondria containing only GFP or dsRed (Fig. 1D; gray bars represent unfused mitochondria, black bars represent fused mitochondria). Consistently, under conditions where fusion was observed, a population of larger single-labeled m-GFP and m-dsRed also appeared (fig. S2, A and B). Second, the relative fluorescence intensity of GFP and dsRed in fused mitochondria was decreased as compared with that of single-labeled mitochondria, indicating that fluorophores were diluted as a result of an increase in mitochondrial volume from fusion (Fig. 1B; average pixel intensity of fused mitochondria was 520, average pixel intensity of unfused mitochondria was 1252; a Student's t test resulted in P = 0.000056). Third, immunoelectron microscopy analysis demonstrated a colocalization of gold-conjugated antibodies directed against GFP and dsRED in fixed mitochondria obtained from a fusion reaction (Fig. 1F). In contrast, in fixed mitochondria obtained from a control reaction in which no fusion was observed, gold-conjugated antibodies directed against GFP and dsRed were observed only in separate mitochondria (Fig. 1E).

Monitoring the behavior of mitochondria labeled with m-GFP and m-dsRed by time-lapse fluorescent microscopy also confirmed that mitochondrial fusion occurred in vitro (movie S2, A and B). Specifically, adjacent mitochondria labeled with m-GFP and m-dsRed were observed to extend toward each other and mix matrix contents. Within our time resolution, membrane extension and matrix content mixing were rapid and unresolved events, similar to in vivo observations in which mitochondrial tubules fuse upon contact (1). Thus, the cytological content-mixing assay represents a reliable measure of mitochondrial fusion in vitro.

The energetics of mitochondrial fusion. Two mitochondrial membrane proteins, Fzo1 and Mgm1, required for mitochondrial fusion in vivo, are predicted to be GTPases, and mutations in the GTPase domains of both proteins result in mitochondrial fusion defects in vivo (4, 7, 8). However, a direct role for GTP hydrolysis in mitochondrial fusion has not been demonstrated.

We examined the nucleotide requirements and the effects of nonhydrolyzable nucleotide analogs on content mixing in vitro (Fig. 2). Mitochondrial fusion was completely dependent on NTPs. The energy regeneration system alone supported a small amount of fusion, indicating that isolated mitochondria contain an endogenous pool of NTPs. Treatment of mitochondria with apyrase, added before the assay to deplete NTPs, completely abolished fusion in a reaction containing an energy regeneration system (20). The addition of exogenous ATP to a reaction containing the energy regeneration system did not further stimulate fusion, indicating either that mitochondria possess sufficient endogenous ATP to support fusion or that ATP was not required for mitochondrial fusion. In contrast, the addition of exogenous GTP in the presence of the energy regeneration system substantially stimulated the efficiency of mitochondrial fusion, suggesting that the amount of GTP in mitochondria was limiting and required for mitochondrial fusion.

To test whether mitochondrial fusion in vitro required GTP hydrolysis, we added excess nonhydrolyzable GTP analogs to a reaction containing an energy regeneration system and ATP. In the presence of guanosine 5′-O-(3′-thiotriphosphate) (GTP-γ-S) or guanosine 5′-monophosphate guanosine 5′ [β,γ-imido]triphosphate (GMPPNP), fusion was substantially reduced, indicating that GTP hydrolysis was essential. The addition of excess guanosine 5′-O-(2′-thiodiphosphate) (GDP-β-S) also inhibited fusion, providing further evidence that GTP binding and hydrolysis were essential. In contrast, the addition of an excess amount of nonhydrolyzable ATP [adenosine 5′-O-(3-thiotriphosphate)] did not reduce fusion, indicating that in vitro ATP hydrolysis was not required. Thus, GTP but not ATP hydrolysis was required for fusion, consistent with essential roles of the two large GTPases Fzo1 and Mgm1 in mitochondrial fusion in vivo.

During the development of the in vitro assay, we observed that mitochondrial fusion was strongly influenced by yeast cell culture media, particularly the carbon source (fig. S1D). Specifically, mitochondria isolated from cells grown on nonfermentable carbon sources efficiently fused in vitro, whereas those isolated from cells grown on a fermentable carbon source were unable to fuse in vitro. This led us to test directly whether mitochondrial membrane potential played a role in mitochondrial fusion by determining the effects of the electron transport chain uncoupler, carbonylcyanide mchlorophenylhydrazone (CCCP) (Fig. 2). When CCCP was present during the assay, fusion was abolished. CCCP also inhibits mitochondrial fusion in mammalian cells, indicating that membrane potential is a conserved requirement (21).

To dissect the requirement for mitochondrial inner membrane potential, we examined the effects of two additional ionophores on fusion: valinomycin, an electrogenic K+ ionophore, which specifically dissipates the electrical gradient, and nigericin, an electroneutral K+/H+ ionophore, which specifically dissipates the proton gradient (Fig. 2). Mitochondrial content mixing was highly sensitive to valinomycin, but significantly less sensitive to nigericin, indicating that the electrical and not the proton gradient was required for efficient mitochondrial fusion.

Mitochondrial fusion proceeds through discrete and sequential mitochondrial outer and inner membrane fusion events. The in vitro fusion reaction was divided into two experimental stages: Stage 1 included mitochondrial centrifugation, incubation, and resuspension, and stage 2 included the addition of exogenous NTPs and an energy regeneration system (19). To assess quantitatively the requirement of stage 1 for fusion efficiency, either mitochondria were centrifuged, incubated, and resuspended in the presence of the energy regeneration system and NTPs, or they were immediately resuspended with the energy regeneration system and NTPs. Content-mixing efficiency in each reaction was determined after 60 min of incubation at 22°C. In the absence of stage 1 treatment, mitochondrial fusion was markedly reduced (Fig. 3A). Both the centrifugation and incubation steps of stage 1 were essential for efficient fusion during stage 2 (fig. S1C). Thus, bringing mitochondrial membranes into close proximity for a critical duration was a required and rate-limiting step. In vivo, this step is likely to be mediated by cytoskeletal elements, such as actin, which in yeast function both to tether and to move mitochondria (22).

Fig. 3.

Analysis of mitochondrial fusion in vitro reveals discrete intermediates. (A) The role of mitochondrial concentration in fusion. Comparison of fusion efficiency in fusion reactions that either underwent or bypassed stage 1 concentration and incubation (19). (B) The effects of dilution on mitochondrial fusion during stage 2. Comparison of the fusion efficiency in reactions that were either fixed at 10 min, or diluted 20-fold into reaction mix at 10 min and incubated for an additional 50 min, or allowed to proceed under standard conditions to 60 min (19).Error bars in (A) and (B) show mean + SD. (C) Representative fluorescent images of mitochondrial structures labeled with m-GFP and m-dsRed that formed in reactions that either bypassed (I and insets), or underwent stage 1 conditions (II and insets).Scale bar, 8 μm.Quantification of deformed or clustered structures in the total mitochondrial population is listed beneath respective conditions. (D) Quantification of deformed mitochondria (gray) and fused mitochondria (black) plotted as a percentage of total mitochondria after stage 1 and after stage 2 (n > 250 per condition).Error bars show mean ± SD (gray) and mean + SD (black).

We then assessed whether mitochondrial concentration was a critical parameter for stage 2 of the reaction (Fig. 3B). After 10 min at stage 2, mitochondria were diluted 20-fold, but the concentrations of the energy regeneration system and NTPs were maintained. The diluted reaction was incubated for an additional 50 min for a total of 60 min. The fusion efficiency of the diluted reaction was reduced, compared with the undiluted 60-min control reaction, indicating that mitochondrial concentration was important, but not essential, during stage 2. In addition, the fusion efficiency of the diluted reaction at 60 min was greater than that of the undiluted control reaction assessed at 10 min, the time of dilution, indicating the existence of dilution-resistant intermediates capable of completing fusion. Thus, during stage 1, intermediate fusion structures formed and went on to fuse in an energy-dependent manner during stage 2.

We examined stage 1 mitochondria with the use of fluorescence microscopy to look for intermediate structures. Two classes of morphological structures were observed: clustered structures with a region lacking a fluorescence signal between adjacent mitochondria (Fig. 3CII) and tightly juxtaposed deformed structures with no region lacking a fluorescence signal between them (Fig. 3CII, arrows). In contrast, mitochondria that bypassed stage 1 and were placed directly under stage 2 conditions for 60 min at 22°C formed a significantly reduced number of both types of structures (Fig. 3CI). Thus, the clustered and deformed structures appear to represent the dilution-resistant fusion intermediates detected during stage 2 of the reaction (Fig. 3B). A comparison of stage 1 mitochondria with stage 2 mitochondria showed significantly fewer deformed stage 2 mitochondria and a proportionally greater number of fused mitochondria, indicating that these structures were productive intermediates in the fusion pathway (Fig. 3D; gray bars represent deformed mitochondria, black bars represent fused mitochondria). There was less of a difference in the number of clustered stage 1 versus stage 2 mitochondria, suggesting that deformed structures represent a more advanced fusion intermediate.

To determine the behavior of outer and inner mitochondrial membranes in these fusion intermediates, we examined stage 1 mitochondria by electron microscopy (EM). EM analysis also revealed at least two classes of mitochondrial structures (Fig. 4A): tightly associated mitochondria without obvious deformation (Fig. 4A, black arrows) and tightly associated and reciprocally deformed mitochondrial pairs (Fig. 4A, red arrows, and Table 1). Higher magnification of deformed mitochondria revealed that they possessed a continuous outer membrane encapsulating a boundary of two separate but tightly aligned and associated inner membranes, suggesting that they were a product of outer membrane fusion in the absence of inner membrane fusion (Fig. 4B, arrows). These structures probably represent the deformed structures observed by fluorescence microscopy (Fig. 4AII). EM analysis revealed that a greater number of deformed structures was observed in stage 1 reactions compared with that of either freshly isolated or stage 2 mitochondria (Table 1 shows stage 1; 27% of stage 1 found in freshly isolated mitochondria, n > 500, and 28% of stage 1 found in stage 2 mitochondria, n > 200). We also observed by EM a population of tightly associated nondeformed mitochondrial pairs (Fig. 4A, black arrows). In most cases, we were unable to visualize all four mitochondrial membranes within these pairs and thus were unable to determine whether outer membrane fusion had occurred. However, it is likely that these tightly associated pairs of mitochondria represented the clustered structures visualized by fluorescence microscopy (Fig. 4AI).

Fig. 4.

Fusion of the outer mitochondrial membrane and inner mitochondrial membrane are separable events.(A) EM analysis of stage 1 mitochondria. Clustered and deformed mitochondria are indicated by black arrows and red arrowheads, respectively. Magnified fluorescent images of stage 1 mitochondria from Fig. 3C are shown representing clustered (I) and deformed (II) mitochondrial structures. (B) EM images of representative deformed mitochondrial pairs. Arrowheads indicate regions of continuous outer membrane encapsulating two separate, but tightly aligned and opposing inner membranes.

Table 1.

Mitochondrial outer and inner membrane fusion of fzo1-1 mitochondria relative to that of the wild-type (WT) control. When fluorescence was used, outer membrane fusion was quantified after stage 1 by determining the number of mitochondria with an om-GFP—labeled outer membrane surrounding doubly labeled but nonoverlapping dsRED- and BFP-labeled matrix compartments and dividing by the total number of mitochondria analyzed. When EM was used, outer membrane fusion was quantified after stage 1 by determining the number of mitochondrial pairs that were reciprocally deformed and surrounded by a continuous outer membrane. A significant number of these structures was observed in input mitochondria used in the fusion assays (27% of wild type, n > 500). Inner membrane fusion was quantified after stage 2 by determining the number of mitochondria labeled with overlapping dsRED- and GFP-labeled matrix compartments and dividing by the total number of mitochondria analyzed. Absolute percentages are indicated in parentheses for wild-type reactions for each assay and condition. Temp., temperature.

Type of mitochondria Temp. (°C) Outer membrane fusion Inner membrane fusion
Fluorescence (%) EM (%) Fluorescence (%)
WT × WT 22° 100 100 100
(9.2%, n > 600) (6.3%, n > 300) (12%, n > 300)
WT × WT 37° 100 100 100
(8.8%, n > 400) (6.4%, n > 300) (11.8%, n > 400)
fzo1-1 × fzo1-1 22° 13 64 88
(n > 600) (n > 800) (n > 300)
fzo1-1 × fzo1-1 37° 14.5 41 5.5
(n > 400) (n > 600) (n > 200)
fzo1-1 × WT 22° 16 60 92
(n > 600) (n > 300) (n > 200)
fzo1-1 × WT 37° 16 52 24
(n > 400) (n > 300) (n > 300)

To substantiate our EM observations, we developed a fluorescence microscopy–based assay for outer membrane fusion (Fig. 5A) (19). Mitochondria labeled with m-dsRED and an outer membrane–targeted GFP (om-GFP) were mixed under various reaction conditions with an equivalent amount of mitochondria labeled with matrix-targeted blue fluorescent protein (m-BFP). If outer membrane fusion occurred in the absence of inner membrane fusion, an outer membrane labeled with om-GFP would be observed surrounding doubly labeled but nonoverlapping matrix compartments labeled with dsRED and BFP (Fig. 5A). Consistent with our EM observations, analysis of stage 1 mitochondria with the use of this assay revealed a significant population of such structures (Fig. 5B, movie S3, and Table 1). In stage 2 reactions, the number of structures in which only the outer membrane had fused was significantly less than the number in stage 1 reactions, confirming that these are intermediates in the fusion pathway (Table 1 shows stage 1; 62% of stage 1 found in stage 2 reactions, n > 300). Consistent with our previous observations, outer membranes labeled with om-GFP surrounding colocalized m-dsRed and m-BFP specifically appeared during stage 2, indicating that inner membrane fusion had occurred under these conditions (Fig. 5C).

Fig. 5.

Mitochondrial outer membrane fusion in vitro. (A) Schematic of in vitro outer membrane fusion assay. (B and C) Fluorescent images of products from fusion reactions with om-GFP/m-dsRed and m-BFP mitochondria after (B) stage 1 (two examples are shown) and (C) stage 2. White arrows in (B) indicate regions of continuous outer membrane connecting two distinct matrix compartments. Scale bar, 1 μm. (D) Energetics of outer membrane fusion. In vitro outer membrane fusion reactions were conducted as described (19) and fusion efficiency is expressed as a percentage of the standard reaction.

Using this fluorescence-based assay, we asked whether outer membrane fusion also required GTP and/or inner membrane potential (Fig. 5D). Pretreatment of mitochondria with apyrase severely inhibited outer membrane fusion of mitochondria during stage 1 (20). Similarly, the addition of GTP-γ-S, GMPPNP, or GDP-β-S during stage 1 significantly reduced the efficiency of outer membrane fusion, indicating that endogenous GTP was required to promote the fusion of the outer membrane. When CCCP was present during stage 1, outer membrane fusion was abolished. Mitochondrial outer membrane fusion was highly sensitive to nigericin but significantly less sensitive to valinomycin, indicating that proton gradient was required for efficient mitochondrial outer membrane fusion.

Our data indicate that outer and inner membrane fusion events were separable and mechanistically distinct. Outer membrane fusion required mitochondrial concentration, was driven energetically by a relatively low (endogenous) concentration of GTP and was dependent on the inner membrane proton gradient (Fig. 5D). Inner membrane fusion, in contrast, required the hydrolysis of a relatively high concentration of GTP and the inner membrane electrical gradient (Fig. 2).

Fzo1 is required for outer membrane fusion and influences the efficiency of inner membrane fusion. Dissection of mitochondrial fusion into separate outer and inner membrane fusion events provides an experimental framework for determining the exact functions of the fusion proteins. Given our fundamental knowledge of membrane fusion events, it seems likely that outer and inner membrane fusion are products of active mechanisms resulting from trans interactions between fusion components on opposing mitochondria (23). The ancient and highly conserved transmembrane GT-Pase, Fzo1, is an excellent candidate for mediating outer membrane fusion.

To assess whether Fzo1 was required during fusion in trans, both heterozygous and homozygous fzo1 mitochondrial mutant reactions were analyzed. In both heterozygous and homozygous Δfzo1 reactions, both outer and inner membrane fusion was completely abolished. However, these defects may have been the result of indirect phenotypes associated with Δfzo1 cells. Specifically, the loss of Fzo1 function in cells secondarily causes a loss of mitochondrial DNA and consequently greatly reduces mitochondrial membrane potential, which is essential for mitochondrial fusion (4) (Fig. 2). To assess more directly the role of Fzo1 in fusion, we used the hypomorphic, recessive, temperature-sensitive fzo1-1 allele. Both mitochondrial tubules and mitochondrial DNA in yeast cells harboring the hypomorphic fzo1-1 allele can be maintained at permissive temperature, but after shifting to nonpermissive temperature, mitochondria rapidly fragment as a result of a block in mitochondrial fusion (4).

To analyze fzo1-1 reactions, both EM analysis and the fluorescence-based assays for mitochondrial outer and inner membrane fusion were used (Table 1). Wild-type mitochondria exposed to nonpermissive temperatures before or during some of the experimental steps in stage 1 were unable to fuse. Thus, to analyze the role of Fzo1, the concentration and incubation steps were performed at permissive temperature and mitochondria were placed at a nonpermissive temperature after resuspension in reactions lacking the energy regeneration system and NTPs (for stage 1) or containing the energy regeneration system and NTPs (for stage 2) and incubated for 60 min.

In homozygous fzo1-1 stage 1 reactions, mitochondrial outer membrane fusion was significantly compromised under both permissive and nonpermissive conditions, compared with that of the wild type, as assessed by both fluorescence and EM analysis (Table 1). It is likely that Fzo1-1 protein retained partial function in vitro under nonpermissive conditions, explaining why outer membrane fusion was not completely abolished. Significantly, mitochondrial outer membrane fusion was equally compromised in heterozygous fzo1-1 stage 1 reactions under nonpermissive conditions as compared with that of the wild type. These findings indicate that Fzo1-Fzo1 interactions on opposing mitochondrial membranes were required to promote outer membrane fusion.

We also asked whether Fzo1 function was required for inner membrane fusion by assaying matrix content mixing in homozygous and heterozygous fzo1-1 reactions under stage 2 conditions (Table 1). At permissive temperature, mitochondrial outer membrane fusion was affected to a greater extent than inner membrane fusion, indicating that the relatively high NTP concentration present during stage 2 partially suppressed the outer membrane fusion defect observed in stage 1 fzo1-1 reactions. At nonpermissive temperature, both mitochondrial inner and outer membrane fusion were severely compromised by the loss of Fzo1 function in heterozygous and homozygous reactions, as compared with the inner and outer membrane fusion of the wild type. Defects in outer and inner membrane fusion were suppressed in fzo1-1 mitochondria isolated from fzo1-1 cells expressing wild-type Fzo1, indicating that they were the result of compromised Fzo1 function. The severe defect in inner membrane fusion observed in fzo1-1 mitochondria at nonpermissive temperature suggests that Fzo1 trans interactions in the outer membrane were also required for inner membrane fusion.

A model for mitochondrial fusion. Mitochondrial fusion is accomplished through sequential mitochondrial outer and inner membrane fusion events that are mediated by separate and mechanistically distinct machineries. Outer membrane fusion requires GTP and trans Fzo1 interactions on opposing mitochondria, suggesting that GTP promotes outer membrane fusion by means of Fzo1. The conserved Fzo1 cytosolic C-terminal domain forms an intermolecular dimeric antiparallel coiled-coil structure (24), suggesting a mechanism for bringing opposing outer membranes into close proximity. The only known fusion protein associated with the inner membrane is the DRP Mgm1. Thus, it is probable that additional components will be required in the inner membrane to mediate the critical steps of membrane association and fusion pore formation.

Although outer and inner membrane fusion can be uncoupled, interactions between Fzo1 and inner membrane fusion proteins, such as Mgm1, play important roles in inner membrane fusion. Biochemical interactions between Fzo1 and Mgm1 have been demonstrated and involve the outer membrane fusion protein Ugo1 (7, 8). The exact nature of the interactions between Fzo1, Ugo1, and Mgm1 and their specific roles in mitochondrial fusion remain largely unknown. However, Ugo1 functions as an adaptor between Fzo1 and Mgm1 (18). Fzo1 interactions with inner membrane components may be required in a mechanical manner for the formation of regions of close inner and outer membrane contact within mitochondria. Such regions of contact would function to bring inner membranes into closer proximity after outer membrane fusion and also perhaps to eliminate cristae in the vicinity of fusion. Indeed, by EM analysis, no cristae are observed at sites of inner membrane contact (Fig. 4B). Alternatively, but not exclusively, Fzo1 may function in a regulatory manner by stimulating, by means of GTP cycle–dependent conformations, events in the inner membrane required for fusion.

Fzo1 is a key player in the evolution of mitochondrial fusion. Based on a phylogenetic analysis, Fzo1 is derived from the eubacterial endosymbiotic precursor of mitochondria (10, 11). Our data showing that Fzo1 plays essential and fundamental roles in the fusion of both outer and inner membranes are consistent with this idea. Phylogenetic analysis of Fzo1 also identifies it as a member of the dynamin-related GT-Pase family (11). The similarity of Fzo1 to DRPs suggests the intriguing possibility that DRPs evolved from a eubacterial progenitor, and that Fzo1, like DRPs, functions to remodel membranes through self-interaction and assembly. An additional evolutionary connection between DRPs and endosymbiotic organelles is that their division also has evolved to require the action of a DRP (25).

DRPs most commonly have been shown to function in membrane fission events, such as mitochondrial and chloroplast division and endocytosis (26). However, the actions of two DRPs, Fzo1 on the outer membrane and Mgm1 on the inner membrane, are required for mitochondrial membrane fusion. In a fusion event, Fzo1 and Mgm1 may possess modified activities and function through self-assembly only to tubulate, and not divide, regions of outer and inner membrane, thereby creating a bending stress, which can be harnessed for membrane fusion. The utilization of DRPs to drive membrane fusion events mechanistically distinguishes mitochondrial fusion from other fusion events in eukaryotic cells. Understanding their exact mode of action will enhance our understanding of the fundamental principles that underlie membrane fusion events.

Supporting Online Material

www.sciencemag.org/cgi/content/full/1100612/DC1

Materials and Methods

Figs. S1 and S2

References

Movies S1 to S3

References and Notes

View Abstract

Navigate This Article