Mitochondrial fission and fusion have been recognized as critical processes in the health of mitochondria and cells. Two decades of studies have generated a great deal of information about mitochondrial fission and fusion; however, still much needs to be understood for the basic molecular mechanisms of these important cellular processes. The core protein factors for mitochondrial fission and fusion are dynamin proteins that possess membrane-remodeling properties. This short review covers a recent development and understanding of the mechanisms by which these mechanochemical enzymes mediate mitochondrial fission and fusion.
Mitochondria carry out diverse functions, ranging from ATP synthesis to steroid biosynthesis. They also regulate cellular redox and calcium as well as cell death. Mitochondria execute most of these within the highly compartmentalized, elaborate, double-membrane system. Metabolic enzymes are mostly contained in the mitochondrial matrix, which is surrounded by the inner membrane embedded with channels/transporters and respiratory chain complexes. This whole metabolic machinery is enclosed by the outer membrane that presumably originates from host cell membrane during the ancient endosymbiotic engulfment. The external appearance of mitochondria is typically tubular, but varies greatly in different organisms and tissues, which may reflect different cellular and physiological environments of organisms and organ systems. Mitochondria in many cell types change their shapes and distribution frequently, collectively termed mitochondrial dynamics. Most prominent features of mitochondrial dynamics are fission and fusion. Owing to frequent fission and fusion, different shapes of mitochondria can be found within a cell as small vesicles, short rods, and reticular networks, which are a snapshot of continually changing shapes.
The core proteins mediating mitochondrial fission and fusion are membrane-remodeling mechanochemical enzymes, which belong to the dynamin family of large GTPases. The mitochondrial dynamins include dynamin-related/-like protein 1 (Drp1/DLP1), mitofusin (Mfn), and optic atrophy 1 (OPA1). Drp1/DLP1 mediates mitochondrial fission, whereas Mfn and OPA1 mediate the fusion of outer and inner membranes, respectively. Drp1 is most similar to prototypic dynamin: the ring-like assembly of Drp1 on the liposomes forms membrane tubules, suggesting its membrane-constricting activity. The inner membrane dynamin OPA1 is required for not only inner membrane fusion, but also maintenance of cristae structure [1–5]. The outer membrane fusion protein Mfn has two isoforms: Mfn1 and Mfn2. Mfn2 also localizes to the endoplasmic reticulum (ER) membrane where it interacts with Mfn1 or Mfn2 of the mitochondria to tether two organelles or to maintain the critical distance [6–10].
Studies from the past decade provided a great deal of information about the involvement of mitochondrial dynamics in pathological conditions, emphasizing the important role of mitochondrial fission and fusion in human health. Mutations in the Mfn2 and OPA1 genes cause the hereditary neuropathies Charcot–Marie–Tooth type 2A and optic atrophy type 1, respectively [11–13]. Furthermore, many pathological conditions, including metabolic and cardiovascular diseases, present with progressive mitochondrial dysfunction and alterations in mitochondrial structure [14–17]. The current understanding of mitochondrial dynamics in physiopathology is expansive and has been the subject of many recent reviews [18–25]. Although a lot of progress has been made, molecular mechanisms of mitochondrial fission and fusion are still incompletely understood. As such, in this short review, we will focus on the mechanistic aspects of mitochondrial fission and fusion, paying specific attention to the recent development in the field.
Mitochondrial fission by Drp1
Drp1: structure and expression
Drp1 belongs to the dynamin family proteins that possess the membrane-remodeling property. It contains the GTPase domain (G-domain), middle domain, variable domain (VD) and GTPase effector domain (GED; Figure 1A, top). These domains have been redefined based on the three-dimensional (3D) structure of conventional dynamin, which includes bundle signaling elements (BSEs) and stalk regions (Figure 1A, bottom) . The proposed Drp1 crystal structure shows that BSEs form a neck structure connecting the GTPase domain and the stalk (Figure 1B) . The stalk consists of seven helices connected by loops. Two loops, L2S and the one formed by the B-insert, are located at the tip of the stalk, at the opposite end to the GTPase domain (Figure 1B). These two loops are suggested to be involved in Drp1 oligomerization and binding to the Drp1 receptor and membrane [26,27]. Notably, the B-insert contains the two known phosphorylation sites, indicating the regulation of Drp1 self-interaction and recruitment, potentially by phosphorylation-mediated conformational changes.
Drp1 structure and mitochondrial fission.
Drp1 is encoded by the gene DNM1L that undergoes alternative splicing (AS) during its expression to produce multiple splice variants in mammals [28–30]. The AS occurs in two regions, the A-insert in the GTPase domain and the B-insert in the VD (Figure 1), each producing multiple different sequences [28,29]. Combinations of these sequences in two regions predict 18 different variants in rat and 14 in mouse by the gene prediction method Gnomon. While these splice variants are ubiquitously expressed, they show a degree of tissue specificity. Most notably, brain tissues express the long variant predominantly, whereas the short variants are the major forms in tissues other than brain [28,29]. A recent in vitro study indicated that the short Drp1 variant has higher GTPase activity, as a long B-insert greatly suppresses GTPase activity and self-assembly . On the other hand, the long variant shows a greatest GTPase stimulation with liposomes bearing Mff, a Drp1 receptor, which is further enhanced by the presence of cardiolipin in the liposomes . However, it is unclear how these differences in biochemical properties of Drp1 variants contribute to the fission process, as mixtures of the variants in cells form heterooligomers. It is possible that the relative abundance of specific variants within a given Drp1 oligomeric structure may provide an additional regulation for mitochondrial fission. Interestingly, indeed, one of the Drp1 variants has been shown to localize to microtubules, which becomes more prominent by higher level expression of this specific variant . Drp1 binding to microtubules was found to be cell cycle dependent, regulated through Drp1 phosphorylation by CDK1 and CDK5, suggesting the role of microtubules in regulating Drp1 sequestration and mobilization . It is also possible that microtubules provide the platform where these cell cycle kinases phosphorylate Drp1.
Mitochondrial receptors for Drp1
Drp1 is a cytosolic protein that translocates to the mitochondrial surface for fission. While Drp1 has an intrinsic property for membrane binding, as purified protein binds and tubulates membranes [31–35], Drp1 recruitment to mitochondria in cells probably occurs through binding to mitochondrial outer membrane-anchored receptor proteins. Early on, the small tail-anchored protein Fis1 was suggested to be the Drp1 receptor [36–38], primarily based on the role of its yeast homolog Fis1p. Subsequently, the tail-anchored mitochondrial fission factor (Mff) and N-terminally anchored mitochondrial dynamics proteins of 49 and 51 kDa (MiD49 and MiD51) were additionally identified to recruit Drp1 to mitochondria [39–42]. Further studies showed that Mff and MiDs are the bona fide receptors, whereas Fis1 has little or no role in Drp1 recruitment [33,43–46]. Loss of Mff causes the greatest decrease in fission, compared with MiDs, suggesting that Mff is the primary receptor for Drp1 . Although the role of mammalian Fis1 in mitochondrial fission is unclear, a recent study suggested that Fis1 may function in autophagic removal of mitochondria (mitophagy) during cellular stress by participating in the fission complex containing Drp1 and Mff at the ER–mitochondrial interface .
Studies indicate that Mff and MiDs recruit Drp1 independently of each other [33,44] and that they may function in distinct processes based on different morphological outcomes upon their overexpression: Mff overexpression induces mitochondrial fragmentation, whereas overexpression of MiD51 or MiD49 elongates mitochondria [40,41]. Although mitochondrial elongation by MiD49/51 overexpression can be caused by Drp1 sequestration at the nonfission sites , it has been shown that low-level overexpression of MiD49/51 increases fission by colocalizing with Drp1 and Mff at the mitochondrial constriction at the ER–mitochondria contacts , suggesting a potential co-operation of Mff and MiDs in mitochondrial fission. More recently, a functional distinction between Mff and MiDs was observed during apoptosis-associated mitochondrial fission. Using null cell lines for Mff and MiDs, partial inhibition of mitochondrial fission was observed with Mff-null and MiD49/51-null cells, but complete fission inhibition in Mff/MiD49/51 triple-null cells during apoptotic induction, supporting co-operative activities of Mff and MiDs in mitochondrial fission . In contrast, MiD-null cells drastically decreased cytochrome c release during apoptosis by preserving cristae structure, whereas Mff-null cells showed only a minor effect, indicating that MiD-mediated, not Mff-mediated, mitochondrial fission is necessary for apoptotic cytochrome c release . However, a similar study indicated an additive resistance to apoptosis upon removal of both Mff and MiDs . The discrepancy could be due to the use of different cell types and apoptotic inducers (HeLa versus MEFs; actinomycin D versus etoposide).
Drp1 receptors in mitochondrial fission
While the role of Mff and MiDs in Drp1 recruitment is clear, how the receptor molecules participate in Drp1-mediated fission is less understood. Recent studies provide some insights into the mechanisms by which Drp1 receptors regulate mitochondrial fission. In studies using yeast two-hybrid and size-exclusion chromatography, it was observed that deletion of the Drp1 B-insert greatly increases Mff binding to Drp1, suggesting a regulatory (inhibitory) role of the B-insert in Mff–Drp1 interaction . Importantly, Drp1 mutants defective in oligomerization were unable to bind to Mff, which suggests that Mff recruits the oligomeric form of Drp1 to the mitochondria . On the other hand, oligomerization-defective Drp1 mutants still bind to MiDs. Therefore, Mff recruits functionally active oligomeric Drp1, whereas the inactive dimeric form of Drp1 binds to MiDs, suggesting differential roles of Mff and MiDs in mitochondrial fission . However, the use of Mff anchored on the liposomes has resulted in different information on the mechanism by which Mff recruits Drp1. In this study, consistent with the other report, Mff binds to Drp1 robustly in solution when the B-insert of Drp1 is removed . However, it was found that oligomeric Drp1 formed in solution failed to bind to Mff on the liposomes. Instead, dimeric Drp1 selectively binds to Mff on the liposomes . The authors proposed a model in which binding of the Drp1 dimer to Mff occurs with the B-insert binding to lipid (cardiolipin). Membrane binding of the B-insert would expose the Mff-binding site on Drp1, stabilizing the Mff–Drp1 interaction to promote functional assembly of Drp1 on the membrane for fission . As the B-insert contains sites for phosphorylation and SUMOylation, it is likely that these Drp1 modifications affect B-insert conformation to regulate interactions of Drp1 with Mff and membrane. It was reported that Drp1 SUMOylation by MAPL stabilizes ER–mitochondria contacts and increases mitochondrial constrictions during cell death . It will be of interest to find out whether and how Drp1 SUMOylation regulates Drp1 structure and interactions with Mff and MiDs.
Mff and MiD49 have been shown to exert minimal effects on Drp1 GTPase activity in solution when cytosolic domains of these receptors were used [27,33]. However, when the Drp1 B-insert was deleted, Mff greatly increased its binding to Drp1 as well as Drp1 GTPase activity [27,49]. Interestingly, using liposome-anchored Mff and MiD51, it was found that Mff stimulates Drp1 GTPase activity, whereas MiD51 inhibits it . Because Drp1, Mff, and MiDs colocalize in the same foci at mitochondria , these opposite roles of Mff and MiDs in regulating Drp1 GTPase activity might provide a fine-tuning of fission progression: Drp1–MiD interaction would maintain the GTP-bound state of Drp1, which facilitates productive assembly of Drp1. Subsequent Drp1 binding to Mff would activate GTP hydrolysis for membrane constriction and scission . Possibly, there are additional regulations of Mff and MiDs for Drp1 binding at the fission site to control the progression of mitochondrial fission.
ER, actin and mitochondrial fission
Ring-like assembly of Drp1 forms the collar around the mitochondrial tubule and GTP hydrolysis-associated narrowing of the collar results in fission. However, Drp1 assembles into a spiral with defined diameters that are much smaller than those of mitochondria, requiring initial constriction. The findings that the ER tubule wraps around mitochondria where fission occurs provide new information regarding the event preceding Drp1-mediated membrane scission. The initial high-resolution imaging of the ER and mitochondria using electron microscopy and tomography demonstrated that the ER tubule physically wraps around and constricts the mitochondrial tubule . This ER-mediated mitochondrial constriction is independent of Drp1 and Mff, and rather it is an initial event providing sites for Drp1 recruitment and assembly for fission .
Not surprisingly, the mechanism by which the ER constricts mitochondrial tubules has turned out to involve the actin-mediated contractile force. Studies found that an ER-associated actin modulator, inverted formin 2 (INF2), polymerized actin at the ER–mitochondria interface where Drp1 was recruited (Figure 1C) . Myosin II was also found to be the component in the mitochondrial constriction machinery, presumably binding to actin filaments generated by INF2 . Additionally, a mitochondria-anchored isoform of the actin-nucleating Spire protein, Spire1C, has been shown to promote actin assembly on the surface of mitochondria through interacting with the ER-associated INF2, suggesting a co-operative role of Spire1C and INF2 in mitochondrial constriction by regulating actin assembly at the ER–mitochondria contacts (Figure 1C) . Furthermore, actin-binding proteins, cofilin, cortactin, and the Arp2/3 complex, were also shown to play a role in mitochondrial fission .
An early observation in yeast showed that puncta formed by Dnm1p (yeast Drp1 homolog) were dynamic, appearing and disappearing at certain spots as well as changing their shapes and sizes. This dynamic behavior had no correlation with fission and constriction events, suggesting distinct processes of mitochondrial constriction and fission protein recruitment . A similar transient association of Drp1 puncta with mitochondria was observed recently . In addition, Drp1 puncta on the mitochondria were found to move, merge, and split . In this study, it was suggested that Drp1 merging is a maturation process. However, the large mature Drp1 puncta may or may not engage in fission, as fission was observed with only 6% of them in a 10-min time span. These observations are consistent with the previous findings in yeast, suggesting that the formation of Drp1 puncta on the mitochondria is independent of fission and that productive fission may require additional factors. Through imaging analyses, Drp1 accumulation at the fission site was shown to be dependent on actin polymerization, myosin II, and INF2 . While these observations clearly demonstrated that the actin at the ER–mitochondria contact is necessary for Drp1 accumulation on the mitochondria, it is not sufficient to lead to fission. As discussed, Mff and MiDs may have opposite effects on Drp1 GTPase activity at the fission site . It was found that the actin–Drp1 interaction in vitro stimulated GTPase activity, which was synergistic with Mff . Actin, Mff, and MiDs may compete or co-operate within the fission complex, adjusting and modifying Drp1 assembly while awaiting a final command for whether or not to execute membrane scission.
Outer membrane fusion by Mfn
Fusion of mitochondrial outer membrane requires Mfn proteins. Early studies showed homotypic and heterotypic interactions of Mfn1 and Mfn2 . It was demonstrated that mitochondrial fusion requires the presence of Mfn molecules in apposing mitochondria, as no mitochondrial fusion occurs in cell hybrids between Mfn-null and wild-type cells . Furthermore, reintroduction of Mfn1 or Mfn2 to the Mfn-null cells restored fusion with wild-type mitochondria, indicating that either isoform can mediate mitochondrial fusion . The crystal structure of the Mfn1 C-terminal region (heptad repeat 2: HR2, Figure 2A) showed a dimeric antiparallel coiled coil. This structural information suggests that Mfn–Mfn interaction through the HR2 brings about tethering of apposing mitochondria for subsequent GTP hydrolysis-induced fusion . However, much of the mechanism by which Mfn mediates fusion is unknown partly due to the lack of structural information for the whole Mfn molecule. Bacterial dynamin-like protein (BDLP) from filamentous cyanobacterium is shown to be most similar to the Mfn protein among metazoan dynamin family proteins . The crystal structure of BDLP has provided useful insights into the Mfn-mediated mitochondrial fusion. The structures of GDP-associated and nucleotide-free BDLP are highly compact unlike other dynamin proteins  (Figure 2B, left). The middle domain (HR1-containing domain in Mfn; Figure 2A) and the GED domain (HR2-containing domain of Mfn) along with the N-terminal extension (NTE) form helix bundles that constitute the neck and trunk regions (Figure 2B). Two helices forming the paddle region are predicted to be transmembrane, possibly inserting into the outer leaflet of the membrane [60,61]. Importantly, the neck and trunk regions are suggested to be connected by a flexible hinge that opens and stretches upon GTP binding . Indeed, structural fitting of the GTP-bound form of BDLP assembled on lipid tubules demonstrated the radical opening (∼135°) of the hinge  (Figure 2B, right). GDP-bound BDLP forms a dimer through an interaction at the GTPase domain, whereas nucleotide-free BDLP shows a decreased dimer interface and swinging up of the paddle toward the GTPase domain, suggesting that nucleotide release may induce an unstable dimer and further compact the molecule . Furthermore, self-interaction of the GTP-bound, fully stretched form of BDLP is mediated by GED (HR2 of Mfn) . Based on this and previous information from BDLP and Mfn, one can envision mitochondrial tethering through head-to-head interactions between the fully extended Mfn molecules in opposing mitochondria upon GTP binding. GTP hydrolysis (GDP bound) would close the hinge, bringing the opposing membrane together for fusion. Possibly, GDP release may augment membrane fusion by inducing additional conformational change that induces membrane contact. Although the BDLP crystal structure suggests that Mfn may form a dimer , the multimeric state of Mfn engaging in mitochondrial tethering and fusion is unclear. Previous studies indicated that Mfn1 forms three distinct size complexes: ∼180 kDa steady-state complex, which forms ∼250 kDa cis-complex in the same mitochondria and ∼450 kDa docking complex between apposing mitochondria in the presence of GTP . While these are crude size estimations from sucrose gradient centrifugation, it is possible that dimeric steady-state Mfn (∼180 kDa) forms higher order cis- and trans-complexes. Because GTP hydrolysis and release may have different effects (compact dimer versus unstable dimer, respectively) , ∼250 and ∼450 kDa complexes observed in the presence of GTP may represent heterogeneous Mfn complexes having different nucleotide states.
In recent studies, electron cryo-tomography (cryo-ET) captured striking images of yeast mitochondria undergoing outer membrane fusion . Cryo-ET of cold-arrested mitochondria from the in vitro fusion assay revealed two types of attached mitochondria: tethered and docked. Tethered mitochondria had a distance of ∼7 nm between two outer membranes with the protein densities in-between, which appeared to have ordered repeats of globular proteins (Figure 2C). In contrast, docked mitochondria have a large flattened interface (∼3 times larger contact area than tethered mitochondria) with a distance of 1–3 nm between the outer membranes, devoid of proteinaceous materials . Instead, the protein density was found to form a ring (docking ring) along the edge of the membrane contact (Figure 2C). It was found that tethered mitochondria proceed to docked mitochondria through GTP hydrolysis . Remarkably, a fusion pore was formed locally at the periphery of the docked membranes in the path of the docking ring where the protein density was sparse, suggesting a coincidence of protein dissociation and outer membrane fusion  (Figure 2C). Expansion of the fusion pore leads to full outer membrane fusion. As Fzo1p/Mfn has been shown to localize at the mitochondrial interface [64–66], these cryo-ET observations demonstrate Mfn-mediated mitochondrial tethering and fusion by GTP hydrolysis. These results agree well with the model suggested above: mitochondrial tethering by Mfn–Mfn interaction with open hinges and GTP hydrolysis increasing membrane proximity by transitioning to the closed/compact Mfn structure (docked mitochondria), leading to membrane fusion. Sparse Mfn where fusion occurs may indicate the nucleotide release-mediated dissociation of Mfn complexes coinciding with augmentation of membrane fusion.
OPA1, an inner membrane-remodeling protein
OPA1 was identified as the human disease gene optic atrophy 1, which is a homolog of yeast Mgm1 that mediates mitochondrial inner membrane fusion [11,12]. A single-span transmembrane domain downstream of the N-terminal mitochondrial matrix targeting sequence serves as a stop-transfer signal (Figure 2A), anchoring OPA1 at the inner membrane, which leaves the bulk of the molecule in the intermembrane space . OPA1 mRNA is alternatively spliced at the N-terminal region to generate eight different splice variants [68,69] (Figure 2A, ‘AS’). In addition to the differential splicing, OPA1 also undergoes post-translational cleavage downstream from the transmembrane domain, which produces short forms of OPA1. Therefore, cells contain both membrane-anchored long OPA1 (L-OPA1) and transmembrane region-free short OPA1 (S-OPA1). All OPA1 variants have a common cleavage site, S1, and some splice variants have the additional cleavage site S2  (Figure 2A). S1 and S2 are cleaved by inner membrane-associated metalloproteases OMA1 and Yme1L, respectively [71–77]. OPA1 molecules smaller than S-OPA1 were identified, indicating the presence of additional cleavage sites .
It has been shown that L- or S-OPA1 alone is insufficient for mitochondrial fusion . OMA1-mediated OPA1 cleavage was observed with mitochondrial depolarization, permeability transition, or apoptosis [71,72,79,80], suggesting that OPA1 cleavage is a mechanism for segregation of dysfunctional mitochondria by preventing fusion. However, an increase in mitochondrial oxidative phosphorylation was shown to enhance OPA1 cleavage to induce mitochondrial fusion , contradicting the notion that OPA1 cleavage is inhibitory of fusion. In another study, knocking out both OMA1 and Yme1L maintains fusion, suggesting that L-OPA1 alone is sufficient for mitochondrial fusion . Unexpectedly, expressing S-OPA1 in OMA1/Yme1L double knockout cells causes mitochondrial fragmentation without affecting fusion, suggesting that S-OPA1 is implicated in mitochondrial fission . Thus, the role of OPA1 in mitochondrial dynamics appears debatable. Furthermore, OPA1 has been shown to play a role in regulating the cristae junction as well as overall cristae [1–5], further complicating OPA1-mediated cellular processes.
OPA1 has a more defined middle domain and GED compared with Mfn (Figure 2A), predicting a 3D structure of S-OPA1 similar to that of Drp1 and dynamin, although L-OPA1 may have a different structure due to the presence of the N-terminal region containing the transmembrane domain. Also predicted is that S-OPA1 forms dimers that assemble into a spiral through interdimer interactions, tubulating the membrane [26,82]. Indeed, S-OPA1 has been shown to tubulate liposomes in vitro . In contrast, studies of the yeast Mgm1 showed that S-Mgm1 binds to the liposome surface as lattices, which rarely results in tubulation [84,85]. More recently, S-Mgm1 binding to liposome has been shown to induce local bending of the membrane, suggesting that Mgm1 has an intrinsic activity for membrane remodeling to initiate inner membrane fusion . Analyses of liposome-bound Mgm1 suggest that S-Mgm1 forms trimers that interact face to face to form a hexamer, postulating that trimer–trimer interaction in opposing membranes tethers and fuses inner membranes . However, other studies indicate that S-Mgm1 forms a dimeric unit upon membrane binding . Additionally, L-Mgm1 interacts with L-Mgm1 in opposing membrane while interacting with S-Mgm1 within the same membrane, suggesting that L-Mgm1 molecules in opposing membranes interact to provide tethering . In contrast, in mammalian cells, OPA1 is only needed in one of the fusing partners for fusion, suggesting that OPA1–OPA1 trans-interaction is not necessary for intermitochondrial tethering in inner membrane fusion . While direct and indirect information is available for the mechanisms of OPA1-mediated inner membrane fusion, studies such as defining OPA1 3D structure along with detailed EM-level imaging will further the knowledge to address the conflicting and diverse experimental results.
Mitochondrial dynamics has been on the center stage of cell biology for 20 years, providing fascinating images and its pathophysiological facet. Dysregulations of mitochondrial fission and fusion are associated with an energetic defect, emphasizing a close relationship between mitochondrial structure and function. Evidently, mitochondrial dynamics is a critical component in mitochondrial health by contributing to maintaining and prolonging the life span of healthy mitochondria, a function integrally shared by mitochondrial biogenesis and mitophagy. While pathological roles of mitochondrial fission and fusion are important for human health, understanding their mechanisms is central to therapeutic intervention targeting these cellular processes. Self-assembly and conformational changes of dynamin proteins are probably the basic mechanisms for membrane remodeling leading to mitochondrial fission and fusion. Much progress has been made in understanding how Drp1 and its receptors work together in mitochondrial fission in recent years. The involvement of additional factors including actin and the ER in fission is becoming clear. Further studies will be necessary to integrate how all these factors participate in the execution of mitochondrial fission through temporal and spatial regulations. The mechanism of mitochondrial fusion, on the other hand, is far less understood. Requirements of docking and coordinated fusion of outer and inner membranes predict a complex nature of mitochondrial fusion. Structural information coupled with defined biochemical properties of Mfn and OPA1 will advance our understanding of fusion mechanisms. Additional factors for fission and fusion working alongside dynamins continue to expand, including binding partners, cytoskeleton, post-translational modifications, and local lipid environment, to list a few. With rapidly advancing experimental techniques, the next 20 years of mitochondrial dynamics is predicted to bring about more exciting results as well as unexpected new developments in this ever-growing field.
bacterial dynamin-like protein
bundle signaling elements
dynamin-related/-like protein 1
fuzzy onion 1
GTPase effector domain
heptad repeat 2
inverted formin 2
mitochondria-anchored protein ligase
mouse embryonic fibroblasts
mitochondrial fission factor
mitochondrial dynamics proteins of 49 and 51 kDa
optic atrophy 1
This work was supported by an American Heart Association grant [16GRNT31170032] to Y.Y.
The Authors declare that there are no competing interests associated with the manuscript.