Mitochondrial Fusion and Division

Contrary to their illustration as ovoid structures in most textbooks, mitochondria are actually dynamic organelles that fuse and divide to form constantly changing tubular networks in most eukaryotic cells. Furthermore, regulation of mitochondrial dynamics is crucial for the health of the cell, and defects are associated with several human genetic disorders. Today researchers are uncovering the genes, gene products, and mechanisms that control fusion and division of mitochondria.

In 1931, Dorothea Smith observed variable mitochondrial networks in developing rat liver cells (Smith 1931). Over subsequent years, many scientists described complex mitochondrial networks in numerous cell types (reviewed in Bereiter-Hahn & Voth 1994). A mitochondrial network consists of a complex, branched system of connected mitochondria (Figure 1). Some modern textbooks still depict mitochondria inaccurately as small, static ovoid organelles, perhaps because early cell cross sections appeared to show many distinct mitochondria, which we now know are often connected in a plane outside the one visualized. Mitochondrial fusion is the physical merging of the outer and then the inner mitochondrial membranes of two originally distinct mitochondria. Mitochondrial division is the breaking apart of one mitochondrial body into two. The combined effects of ongoing fusion and division give rise to mitochondrial networks.

Mitochondrial connectivity allows for even distribution of special mitochondrial components such as the lipid cardiolipin. Fusion of mitochondria also enables genetic complementation: two mitochondrial genomes with different defects inside the same organelle can each encode what the other lacks, generating all components for a functional mitochondrion. As hypothesized by Vladimir Skulachev (2001), a branched mitochondrial network may allow ATP generation in oxygen-poor regions of a cell through electrical transmission of the mitochondrial membrane potential across the network. In addition, mitochondrial division facilitates the transport of mitochondria to daughter cells during mitosis and meiosis.

Mitochondrial fusion and division typically counterbalance each other. Levels of nutrients and other criteria can tip the balance toward either fusion or division. In some specialized cell types, like neurons and sperm, energy needs are high and mitochondria are trafficked long distances, dictating unique needs for the shape of the mitochondrial network. Researchers have identified many gene products that control mitochondrial fusion and division. Three central players are (1) mitofusins (outer mitochondrial membrane fusion), (2) OPA1/Mgm1 (inner mitochondrial membrane fusion), and (3) Drp1/Dnm1 (division of outer and inner mitochondrial membranes). All of these molecules are GTP-hydrolyzing proteins (GTPases) that belong to the dynamin superfamily. The different regulation of these proteins and their binding partners is a framework that scientists use to understand mitochondrial dynamics in different cells and contexts.

The first experimental manipulations of mitochondrial fusion were conducted in model organisms. Jodi Nunnari, Wallace Marshall, John Sedat, Peter Walter, and their colleagues at UC San Francisco demonstrated that mitochondrial fusion occurs in the budding yeast Saccharomyces cerevisiae. They treated two live, haploid S. cerevisiae populations with green or red mitochondria-specific fluorescent dyes and observed that when cells from the two populations mated, the green and red mitochondrial dyes redistributed and merged into a yellow signal in the diploid zygote (Figure 2; Nunnari et al. 1997). The fluorescently labeled mitochondrial networks observed over time subsequently divided into pieces and then fused again in different arrangements.

All membrane fusion events in the cell are tightly regulated by protein mediators. Otherwise, a hopeless jumble of structures would result from the random fusion of any two nearby membrane-enclosed organelles. The proteins that enable mitochondria to fuse are different from those that allow fusion between other structures, such as Golgi-derived vesicles. How do scientists identify these mitochondrial fusion mediators? They use a genetic approach and start with mutant strains of a model organism that have defects in mitochondrial fusion. For example, in the maturing sperm cells of the fruit fly Drosophila melanogaster, mitochondria undergo dramatic shaping, including aggregation, fusion, and elongation. Defects lead to sterility, making the fruit fly a good system for genetic study of mitochondrial shaping, since male sterile strains can be easily identified from thousands of strains with different random mutations. Karen Hales and Margaret Fuller at Stanford University used this model system to discover the first known protein mediator of mitochondrial fusion. All the mitochondria in a Drosophila spermatid normally fuse into two large, intertwined mitochondrial derivatives, which then disentangle and elongate as the sperm tail grows. Male flies homozygous for mutations in the fzo gene are sterile because mitochondria fail to fuse and mature sperm are nonmotile. Hales and Fuller used genetic mapping and cloning techniques with this strain to identify the fzo gene, which encodes the founding member of the conserved mitofusin GTPase family (Hales & Fuller 1997).

Greg Hermann and colleagues in Janet Shaw's laboratory at the University of Utah studied S. cerevisiae strains lacking the yeast mitofusin homolog and saw that mitochondria (detected by fluorescent labeling) were abnormally fragmented in those cells (Figure 3; Hermann et al. 1998). When researchers labeled mitochondria separately with green and red dyes, they saw that mutant mitochondria from two populations could not merge their contents as one would expect. In other words, the mutant mitochondria could not fuse (Figure 2). Similar findings have been observed in the mouse. Hsuichen Chen and colleagues in David Chan's laboratory at the California Institute of Technology made knockout mouse strains lacking one or both of the two mammalian mitofusins, Mfn1 and Mfn2. Although homozygous mutant mice died because of placental defects, cultured cells from these animals did indeed show fragmented mitochondria (Chen et al. 2003). Chan and collaborators also used genetic tools to generate live mice that lack functional mitofusins in certain tissues only. They found tissue-specific roles for mitochondrial fusion in the cerebellum and skeletal muscle. In the cerebellum, mitochondrial fusion is necessary for cell survival and for the movement of mitochondria within neurons (Chen et al. 2007). In skeletal muscle, mitochondrial fusion is most crucial for protecting against the deleterious effects of mitochondrial DNA mutations (Chen et al. 2010).

As it turns out, mitofusin defects can also occur in humans. The human neurodegenerative disorder Charcot-Marie-Tooth disease type 2A, characterized by progressive sensory and motor losses in the extremities, results from mutations in human mitofusin Mfn2 (Zuchner et al. 2004). Scientists have generated mouse strains carrying mutations that directly mimic mutations in human patients, and the mouse and human symptoms are similar (Detmer et al. 2008). Even with a mouse model, however, scientists face significant challenges in devising mitochondrial treatments (and delivering them to the relevant neurons) in Charcot-Marie-Tooth type 2A patients.

Mitofusins are large GTPases that traverse the outer mitochondrial membrane twice, with the amino and carboxy termini both facing into the cytoplasm (Figure 4). In a like-attracts-like fashion, mitofusins bind to other copies of themselves on adjacent mitochondria. Takumi Koshiba and colleagues in David Chan's laboratory used immunoprecipitation as a binding assay with different pieces of mitofusins to determine that a region in the carboxy terminus of the protein mediates homotypic (self) binding (Koshiba et al. 2004). Once mitofusins are tethered, GTP hydrolysis is crucial for enabling mitochondrial fusion, since defects in the GTPase domain lead to tethered but unfused mitochondria (Chen et al. 2003; Hales & Fuller 1997; Hermann et al. 1998). We still do not know the exact mechanism by which GTP hydrolysis triggers a mitofusin shape change to initiate the merging of membranes.

The human mitofusin protein Mfn2 appears to have additionals beyond mediating mitochondrial fusion. In 2008, Olga de Brito and Luca Scorrano from the Venetian Institute of Molecular Medicine showed that Mfn2 exists on endoplasmic reticulum (ER) membranes and tethers ER to mitochondria. In cultured cells lacking functional Mfn2, mitochondria were not held as closely against ER membranes as is normally seen in wild-type cells (Figure 5). Close contact between mitochondria and ER is important for the transfer to mitochondria of calcium released from the ER during cell signaling events. We do not yet know how mitofusins enable fusion between outer mitochondrial membranes but only enable tethering (with no fusion) between mitochondrial and ER membranes. Robert Baloh and colleagues at Washington University showed that defective Mfn2 in neurons slows mitochondrial transport and results in mitochondrial clustering around the nucleus instead of movement to axons (Baloh et al. 2007; Misko et al. 2010). Further research will determine the relative importance of mitochondrial fusion, mitochondria-ER tethering, and mitochondrial transport in the symptoms of Mfn2-associated Charcot-Marie-Tooth disease type 2A.

Since mitochondria have double membranes, full fusion between mitochondria requires two membrane fusion events. A large GTPase called OPA1 (in mammals) and Mgm1 (in yeast) localizes to the inner mitochondrial membrane and is required for fusion of that phospholipid bilayer. Shelly Meeusen and colleagues in Jodi Nunnari's laboratory at UC Davis used a variation of the previously mentioned double-dye labeling technique to show that OPA1/Mgm1 is required only for inner mitochondrial membrane fusion in yeast cells (Figure 6). They generated mgm1 mutant yeast cells whose outer mitochondrial membranes were marked with a green fluorescent tag. To two populations of these cells, they added blue or red mitochondrial dyes that are taken up into the matrix (the region enclosed within the inner membrane). When cells of the two mgm1 mutant yeast populations mated, the resulting mitochondria showed distinct, separate blue and red mitochondrial matrix compartments enclosed within one outer mitochondrial membrane. This observation suggested that the outer, but not the inner, mitochondrial membrane could fuse (Meeusen et al. 2006). In a control experiment using normal yeast cells that mated, the two matrix compartments indeed mixed successfully.

If OPA1/Mgm1 promotes fusion of inner mitochondrial membranes, then why don't the cristae (infoldings of inner membrane) fuse together? In fact, OPA1/Mgm1 does indeed tether (but not fuse) cristae and helps maintain their folded structure (Figure 7). Several research groups showed by immuno-electron microscopy that the OPA1/Mgm1 protein is present on cristae and that in mutant cells lacking the protein, the cristae membranes are not held together properly (Griparic et al. 2004; Olichon et al. 2003; Wong et al. 2000). As with mitofusins, we do not yet know what controls OPA1/Mgm1-mediated tethering versus fusion. Nunnari and colleagues speculate that OPA1/Mgm1-mediated inner membrane fusion may occur only after a signal from active mitofusins on the outer membrane (Hoppins et al. 2007). OPA1/Mgm1 is thus a dual-function protein important both for inner mitochondrial membrane fusion and for maintaining the shape of cristae. Defects in either or both processes underlie the progressive blindness disorder dominant optic atrophy, which is caused by OPA1 mutations in humans (Alexander et al. 2000; Delettre et al. 2000).

Mitochondria are not always fusing. They have an equal, balanced activity of division (fission) within most cells. Why do mitochondria divide? Mitochondrial division is important for the remodeling and rearrangement of mitochondrial networks, as well as for enabling mitochondrial segregation during cell division. William Bleazard and colleagues in Janet Shaw's lab, and Hiromi Sesaki in Rob Jensen's lab at Johns Hopkins University, defined a role in yeast mitochondrial division for a conserved, large dynamin-related GTPase called Dnm1 (Drp1 in mammals). They discovered that, in cells with mutations in the dnm1 gene, mitochondria form large nets that result from failed mitochondrial division and therefore unopposed mitochondrial fusion; the system of fission and fusion is thrown out of balance (Figure 3; Bleazard et al. 1999; Sesaki & Jensen 1999). Other researchers confirmed a mitochondrial division role for homologs of this protein in different organisms (reviewed in Hoppins et al. 2007; Liesa et al. 2009).

Dnm1/Drp1 is very closely related to dynamin, which functions in a different cell context: endocytosis. Dynamin self-assembles into spiral structures around the forming endosome, pinching off the endosome as a result of GTP hydrolysis and a conformation change (Mears & Hinshaw 2008). Dnm1/Drp1 appears to act analogously on mitochondria. Elena Ingerman and colleagues in Jodi Nunnari's lab purified large quantities of Dnm1 and let it assemble under conditions that allowed imaging by electron microscopy. They showed that the protein physically associates with other copies of itself in curved structures that fit exactly around the outer surface of mitochondrial division sites (Figure 8). In the current model, these curved Drp1/Dnm1 structures constrict and pinch off mitochondria using the energy from GTP hydrolysis (Hoppins et al. 2007). We do not yet know exactly how phospholipid bilayers are deformed during this process to allow membrane topology rearrangement.

Curiously, researchers who made yeast strains with fluorescently tagged Dnm1 found that not all the protein in the cell associates with mitochondria (Otsuga et al. 1998; Sesaki & Jensen 1999). Some of it was observed in the cytoplasm, while some was in clumps on the mitochondrial surface at mitochondrial division sites. Perhaps this is not surprising, since the protein does not include a transmembrane domain and is therefore not expected to be permanently anchored in a membrane on its own. Several research groups discovered that a different gene product, called Fis1, which does have a transmembrane domain, sits in the outer mitochondrial membrane and is crucial for recruiting Drp1/Dnm1 to mitochondrial fission sites. In yeast cells lacking Fis1, the Drp1/Dnm1 protein was present but could not associate with mitochondria, and mitochondria formed a net structure representing failed mitochondrial division (James et al. 2003; Mozdy et al. 2000; Yoon et al. 2003).

Up-and-down regulation of mitofusins, OPA1/Mgm1, and Drp1/Dnm1 cumulatively dictate the balance between mitochondrial fusion and division in different cell contexts. For example, mitochondria break into smaller pieces early in programmed cell death (apoptosis). This event is triggered by up-regulation of Drp1/Dnm1 and down-regulation of mitofusins. Later in apoptosis, altered OPA1/Mgm1 activity in the inner mitochondrial membrane causes a change in cristae structure, release of cytochrome c, and subsequent activation of the destructive caspase enzymes (reviewed in Suen et al. 2008), indicating that inner mitochondrial membrane structure is intertwined with regulatory pathways influencing cell life and death.

Regulation of these three mitochondrial shaping proteins occurs in different ways in various cell contexts and at many levels, including protein stability, protein cleavage, protein conformation via covalent modification, and protein localization via association with binding partners. The additional protein participants are too numerous to describe here. In brief, degradation of mitofusins by the ubiquitin-proteasome system determines their steady-state levels and thus influences mitochondrial fusion (Cohen et al. 2008). OPA1/Mgm1 exists as several different-sized isoforms with overlapping functions; proteases that generate these isoforms regulate OPA1/Mgm1 activity (Griparic et al. 2007; Ishihara et al. 2006). The mitochondrial fission mediator Drp1/Dnm1 is regulated by posttranslational modifications such as phosphorylation and the addition of the ubiquitin-like protein SUMO (reviewed in Santel & Frank 2008), as well as by association with binding partners on the mitochondrial surface.

Finally, the common motor neurodegenerative disorder Parkinson's disease appears to be related to the control of mitochondrial dynamics, at least in some patients. Normally, mitochondrial fusion enhances mitochondrial integrity by allowing component sharing across the tubular network. However, fusion of highly damaged mitochondria to the network could be detrimental, since hobbled mitochondria generate reactive oxygen species that cause a chain reaction of damage. Cells have mechanisms to segregate and degrade badly damaged mitochondria. Two research groups (Angela Poole and colleagues in Leo Pallanck's laboratory at the University of Washington, and Elena Ziviani and Ran Tao in Alexander Whitworth's laboratory at the University of Sheffield) showed that mitochondrial quality control involves mitofusin regulation by the Parkinson's-associated gene products PINK1 and Parkin (Poole et al. 2010; Ziviani et al. 2010). In normal cells, PINK1 senses a membrane potential change on damaged mitochondria and recruits Parkin, which in turn ubiquitinates mitofusins. These damaged, ubiquitin-marked mitochondria are then targeted for degradation. A current hypothesis, not yet proven, is that alteration of mitofusins may also prevent damaged mitochondria from fusing to the healthy network. In Parkinson's patients with faulty PINK1 or Parkin, damaged mitochondria are not degraded and remain integrated in the network, leading to cell damage and neurodegeneration.

Mitochondrial fusion and division are mediated by several large GTPases whose combined effects lead to the dynamic mitochondrial networks seen in many cell types. Mitofusins control outer mitochondrial membrane fusion, and OPA1/Mgm1 mediates inner mitochondrial membrane fusion. Drp1/Dnm1 is a central player in mitochondrial division. Many additional protein participants modulate the activities of these proteins in various contexts. Several human genetic neurodegenerative diseases are associated with improper regulation of mitochondrial fusion or division. Scientists are currently determining how the molecular shape changes of these GTPases control the precise three-dimensional arrangement of phospholipids during fusion or division. Scientists are also searching for additional ways in which these fusion and division mediators are regulated in different tissues, since each mechanism of regulation represents a possible therapeutic target for patients with related disorders.