Introduction
The relative rates of mitochondrial fusion and division control the striking array of mitochondrial shapes and numbers observed in different eukaryotic cell types1. Mitochondrial morphology can range from the hundreds of small, ovoid organelles seen in hepatocytes to the long, spiral tubule found at the base of a sperm flagellum. In addition to shape and number, mitochondrial dynamics are critical for normal organelle function. For example, mice defective in mitochondrial fusion die early in development2, whereas yeast mitochondrial-fusion mutants rapidly lose their mitochondrial DNA and are defective in oxidative phosphorylation1. Moreover, organelle fusion and division also seem critical for the dramatic membrane remodelling events facilitating cytochrome c release during apoptosis3, 4.
Before the study by Choi et al.5 on page 1255 of this issue, three proteins were known to mediate mammalian mitochondrial fusion. Two related proteins, Mitofusin 1 (Mfn1) and Mitofusin 2 (Mfn2), are anchored in the outer membrane with a GTPase domain facing the cytosol6. Mfn1 and Mfn2 are at least partially redundant, as cells lacking Mfn1 show a reduction in mitochondrial fusion, whereas fusion is absent in cells lacking both proteins. Mitofusins are homologous to fusion proteins first identified in Drosophila (fuzzy onions)7 and yeast (Fzo1p)8, 9. In yeast fzo1 mutants, cells contain many small mitochondrial fragments instead of the 5–10 tubular mitochondria observed in wild-type cells, and a defect in mitochondrial fusion in fzo1 mutants can be directly demonstrated using a cell-mating assay1. The fragmentation of mitochondria in fzo1 mutants depends on mitochondrial division10. In particular, yeast mitochondrial fission requires the dynamin-related GTPase, Dnm1p, and cells lacking this protein contain a single mitochondrion consisting of a network of interconnected tubules. Strikingly, dnm1 fzo1 double mutants contain relatively normal, tubular-shaped mitochondria, suggesting that mitochondrial shape and number in cells is controlled, at least in part, by a balance between division and fusion10. The third mammalian protein required for mitochondrial fusion is OPA1, a dynamin-like GTPase11. Alternative splicing and proteolytic processing produces multiple forms of OPA1 located in the inner membrane and intermembrane space. Highlighting the pivotal role of fusion in mitochondrial function, defects in both Mfn2 and OPA1 have been implicated in neurodegenerative disorders6, 11.
Mfn1, Mfn2 and OPA1 do not resemble other fusion proteins (such as the SNARES of the exocytic–endocytic pathway), raising the possibility that mitochondria fusion occurs by a very distinct mechanism.
A potential new and exciting mitochondrial fusion component was found in a search of the human genome for proteins distantly related to the well-characterized PLD1 and PLD2 enzymes. These classical PLDs cleave phosphatidylcholine to produce phosphatidic acid12. Choi et al. identified a PLD-like protein called MitoPLD, which they showed resides in the mitochondrial outer membrane and is expressed in a wide array of tissues. Providing a clue to its function, alignments showed that mitoPLD actually resembles bacterial cardiolipin synthase more than PLD1 or PLD2. In eukaryotes, cardiolipin — a 'double' phospholipid with four fatty acid tails — is found in mitochondrial membranes and is synthesized by an inner membrane-localized cardiolipin synthase that is very different from PLD1, PLD2 or MitoPLD. Using a bacterially produced MitoPLD, Choi et al. found that it does not make cardiolipin, but instead seems to catalyse the reverse reaction, hydrolysing cardiolipin to generate phosphatidic acid. Although MitoPLD resembles PLD1 and PLD2 in its production of phosphatidic acid, MitoPLD contains only half of the site necessary for catalysis. The authors show that the mitochondrial protein dimerizes, which presumably generates the active form of enzyme.
A big surprise came from the observation that MitoPLD is required for mitochondrial fusion, which was monitored through fusion of HeLa cells that individually expressed mitochondrial-targeted GFP or dsRed. When MitoPLD activity was inhibited using RNA interference or by expressing a dominant-negative version of MitoPLD, most of the mitochondrial fragments within the fused cells contained only GFP or dsRed. In contrast, cells with functional MitoPLD efficiently fused their mitochondria, rapidly mixing the two fluorescent signals. Moreover, overproduction of MitoPLD produced a network of aggregated and interconnected mitochondrial tubules, consistent with a MitoPLD-dependent increase in fusion activity. Cells containing altered MitoPLD activity grew normally and their mitochondria retained membrane potential and cytochrome c. Thus, the fusion defect does seem to result from generalized mitochondrial dysfunction or apoptosis.
Based on the studies with conventional PLDs, there are several possibilities for how MitoPLD-generated phosphatidic acid facilitates mitochondrial fusion. For example, phosphatidic acid may activate or anchor other proteins needed for mitochondrial fusion, analogous to the way phosphatidic acid recruits Spo20, a yeast SNARE, to vesicles13. A more intriguing possibility is that phosphatidic acid, or a derivative such as lysophosphatidic acid or diacylglycerol, acts as a fusogenic lipid to locally alter the curvature of the opposing bilayers. Regardless of its mode of action, the topology of MitoPLD offers a tantalizing mechanism for its regulation. MitoPLD is inserted in the outer membrane through an amino-terminal transmembrane segment, positioning its catalytic domain in the cytosol and away from the mitochondrial surface. The authors speculate that MitoPLD may function only after its catalytic domain is 'plunged' into the surface of the opposing membrane (Fig. 1). Supporting this hypothesis, the coiled–coil domains within Mfn1 and Mfn2 seem to function early in the fusion pathway to dock or tether mitochondria in close proximity14. Choi et al. find that the mitofusins act before MitoPLD and are needed for MitoPLD-induced changes in mitochondrial structure.
Figure 1: A schematic representation of a model for MitoPLD in mitochondrial fusion.
Tethering of two mitochondria by the carboxy-terminal coiled–coil domains (CC) of mitofusin (green), which also carry a GTPase (G) domain, brings the active MitoPLD dimer (red) into contact with the opposite membrane (perhaps at contact sites between the outer and inner membranes). MitoPLD then cleaves cardiolipin to generate phosphatidic acid, which then stimulates fusion of the mitochondrial outer (and possibly inner) membranes. Based on its function in other types of membrane fusion, phosphatidic acid may activate or recruit fusion proteins, or act directly as a fusogenic lipid.
Full size image (79 KB)As current dogma dictates that cardiolipin is exclusively located in the mitochondrial inner membrane, the cytosolic location of the catalytic domain of MitoPLD seems paradoxical. Although somewhat controversial, several different studies suggest that a significant fraction of cardiolipin is, in fact, in the outer leaflet of the outer membrane15, 16 and is therefore accessible to MitoPLD. As cardiolipin is clearly synthesized in the inner membrane, any transport of this lipid to the outer membrane most likely occurs through contact sites between the two membranes. The observation that yeast Fzo1p mediates an outer membrane–inner membrane connection suggests that mitochondrial fusion occurs at contact sites, with outer and inner membranes fused in an ordered and consecutive pathway17. MitoPLD would therefore present a second mechanism to restrict mitochondrial fusion to contact sites. MitoPLD also provides an explanation for the mitochondrial fragmentation that commonly occurs during apoptosis. Choi et al. speculate that proapoptotic factors that are known to bind cardiolipin, such as tBid, may titrate cardiolipin away from MitoPLD and thereby inhibit mitochondrial fusion.
Although the discovery of MitoPLD is provocative, a number of hurdles remain before its role in mitochondrial dynamics is fully understood. In particular, a more detailed examination of MitoPLD enzyme activity is needed, including its potential regulation by the mitofusins and OPA1. Moreover, in vitro studies similar to those examining SNARE-driven vesicle fusion18 are needed to determine whether phosphatidic acid produces a fusogenic lipid, or instead functions to recruit Mfn1, Mfn2 or OPA1 to sites of fusion. Experiments with isolated mitochondria will also be critical to resolve whether MitoPLD dimers function in cis or trans. Interestingly, despite a number of genetic, genomic and proteomic studies, yeast mitochondria do not seem to contain a MitoPLD-like protein. If generation of phosphatidic acid is truly a fundamental step in the all mitochondrial fusion pathways, then either the yeast MitoPLD awaits discovery, or yeast cells generate phosphatidic acid from a lipid other than cardiolipin. Regardless, the discovery of mammalian MitoPLD suggests that despite carrying a unique repertoire of fusion proteins, the basic mechanism of mitochondrial fusion is surprisingly similar to fusion mechanism(s) used by other organelles.
