The superfamily of dynamins includes classical dynamins and dynamin-related proteins. Classical dynamins are proteins that share sequence similarity with the first described dynamin, which is a large GTPase with five characteristic domains. Dynamin-related proteins differ from classical dynamins in that they contain, at a minimum, three of the five characteristic domains. Dynamins are associated with diverse cellular processes, including the release of transport vesicles [as in clathrin-mediated endocytosis (CME)], fusion and fission of mitochondria, division of chloroplasts and peroxisomes, cell division, and resistance to viral infections. Corresponding with their roles in a broad range of cellular functions, dynamins are found in all eukaryotic phyla (panel 3a in the poster). In this article and its accompanying poster, we present selected aspects of what is known about individual members of this superfamily in the context of their structure, function, interaction partners and roles in diseases.
GTPase is not enough
Three domains – GTPase domain, middle domain and GTPase-effector domain (GED) – form the core of all dynamins (panel 1a in the poster). The GTPase domain is the key signature of dynamins and resembles that of Ras-like GTPases. It contains the four conserved elements that are typically found in the smaller Ras-like GTPases: the P-loop (G1 motif), switch-I (G2), switch-II (G3), and the motif involved in base and nucleotide binding (G4) (Reubold et al., 2005; Praefke and McMahon, 2004) (supplementary material Fig. S1A). In addition, the dynamin GTPase domain contains dynamin-specific sequences between G2 and G3, and downstream of the G4 motif; these sequences might add regulatory functionality (Mears et al., 2007) (supplementary material Fig. S1B). The middle domain lacks similarity to established structural motifs but contains a predicted coiled-coil region and has been shown to be crucial for the self-assembly of human dynamin-1. The GED contains two predicted coiled-coil regions, is also involved in self-assembly and has been shown to increase GTPase activity upon dynamin self-assembly (Ramachandran et al., 2007).
Classical dynamins harbor two additional domains that dynamin-related proteins typically lack: a highly conserved pleckstrin-homology (PH) domain that confers binding to negatively charged lipids via its flexible regions (Ferguson et al., 1994) (supplementary material Fig. S2A,B) and a proline-rich domain (PRD) that is the binding site for proteins that interact with dynamins via Src-homology 3 (SH3) domains. Several of the dynamin-related proteins vary in their domain structure to accommodate for specific functional requirements; for example, some dynamin-related proteins contain membrane-spanning regions or organelle-targeting sequences (panels 3a-c in the poster).
Compared with Ras-GTPases, the GTPase domain of classical dynamins has a lower affinity for nucleotides and displays higher hydrolysis rates. In addition, the GTPase domain of dynamins exhibits cooperativity of GTP hydrolysis upon dynamin oligomerization. Thus, in vitro rates of hydrolysis increase by one to two orders of magnitude when dynamin self-assembles into helical arrays around lipid tubes (Song et al., 2004). Interactions between the GTPase domain, the middle domain and the GED drive this self-assembly process, whereas the PH domain provides affinity for lipid (panel 1b in the poster). Upon hydrolysis of GTP, dynamin undergoes a conformational change that leads to a decrease in the helix diameter, thereby constricting and twisting the enclosed lipid tube (Danino et al., 2004; Roux et al., 2006) (panel 5 in the poster). These observations suggest that dynamin translates the scalar chemical process of GTP hydrolysis into a vectorial physical force – that is, it exhibits mechano-enzymatic properties (Hinshaw, 2000). In vitro, multiple rounds of GTP hydrolysis lead to the disassembly of dynamin oligomers and their release from the lipid bilayer (Danino et al., 2004). The results of recent light-microscopy studies suggest that dynamin alone could be sufficient for membrane fission in vitro (Bashkirov et al., 2008; Pucadyil and Schmid, 2008).
Dynamins mediate more than endocytosis
The domain architecture of dynamins and dynamin-related proteins provides the structural basis for these proteins to carry out a large variety of essential cellular functions.
In animals, classical dynamins facilitate vesicle budding in CME and are required for other events such as budding of vesicles from the recycling endosome and Golgi, and for the internalization of caveolae (Praefcke and McMahon, 2004). In CME, the interplay between clathrin assembly, the recruitment of proteins with membrane-remodeling activities and the action of dynamin leads to invagination of clathrin-coated buds and vesiculation. GTP-hydrolysis-induced conformational changes in dynamin might then drive constriction of the bud neck to the point that spontaneous fission at the neck, and release of the vesicle, occurs (panel 5 in the poster). Alternatively, the release of dynamin from the membrane following constriction of the bud neck might facilitate fission of the lipid tube (Bashkirov et al., 2008). Furthermore, it has been proposed that actin dynamics could be involved in the fission and release of budding vesicles (Itoh et al., 2005; Shin et al., 2008).
In plants, the dynamin-1 homolog dynamin-related protein 2A (DRP2A) interacts with clathrin-binding partners, but there is no direct evidence that DRP2A is involved in endocytosis. Instead, DRP2A mediates Golgi-vacuole trafficking and has been shown to localize to the forming cell plate (Fujimoto et al., 2008). By contrast, Arabidopsis DRP1C, a protein of the DRP1 family that lacks both the PH domain and the PRD, localizes to the division plane and the plasma membrane, and is thought to take part in clathrin-mediated membrane dynamics (Konopka et al., 2008). Although budding yeast utilizes CME and expresses proteins that are homologous to mammalian dynamin-binding proteins [such as yeast homologs of the dynamin partner amphiphysin (Rvs161p, Rvs167p)], no yeast dynamin proteins have been shown to be involved in CME.
The important differences between CME events that are observed in animals, plants and yeast suggest that the role and requirement of dynamins in this process is not universal.
In animals and budding yeast, the dynamin-related proteins DRP1 and Dnm1, respectively, mediate fission of mitochondria (panel 4 in the poster). These proteins lack a PRD and PH domain but contain a B-domain that is located between the middle domain and the GED, and that is devoid of any specific structural features (panel 3b in the poster). Despite the absence of a PH domain, Dnm1 binds preferentially to negatively charged lipids and, similar to human dynamin-1, assembles into helical structures around lipid tubes. Notably, the dimensions of the diameter of Dnm1 helical assemblies on lipid tubes (∼110 nm) match those of mitochondrial constriction sites (Ingerman et al., 2005), which suggests that Dnm1 spirals could encompass mitochondrial sites destined for fission. Dnm1 also contributes to the division of peroxisomes in yeast, a process that is mainly orchestrated by the dynamin Vps1. The finding that dynamins are involved in the division of both of these organelles corroborates the notion that the activities of peroxisomes and mitochondria are fundamentally connected (Motley et al., 2008). Vps1 is also involved in the fission of yeast vacuoles (Peters et al., 2004). In plants, a combination of prokaryotic and eukaryotic proteins assists in the division of chloroplasts. DRP5B is the dynamin-related eukaryotic protein that facilitates division from the outside of chloroplasts (Miyagishima et al., 2008). Finally, members of the DRP3 protein family are thought to regulate the fission of both mitochondria and peroxisomes in plants (Fujimoto et al., 2009).
The observation that dynamins can be involved in more than one cellular process suggests that there are mechanisms in place that might regulate their targeting to specific organelles and modulate their function.
In animals, the fusion of mitochondria involves at least two steps: fusion of the outer mitochondrial membrane via the membrane-anchored dynamins mitofusin 1 (MFN1) and MFN2, and fusion of the inner membrane via optic atrophy 1 homolog (OPA1) (panel 4 in the poster). In budding yeast, the OPA1 ortholog Mgm1 is essential for fusion of the inner mitochondrial membrane but also seems to be involved in the fusion of the outer membrane, because cells that lack Mgm1 do not undergo mitochondrial fusion (Sesaki et al., 2003). Mitochondrial fusion in yeast also requires the mitofusin ortholog fuzzy onions 1 (Fzo1). Mitofusins are thought to form homo- or hetero-oligomers through C-terminal heptad repeat regions, and tether adjacent mitochondria in what might be the first step of mitochondrial fusion (Benard and Karbowski, 2009). Mammalian OPA1 and yeast Mgm1 each exist in two forms, a membrane-bound form and a diffusible form, both of which are required for mitochondrial fusion; these are produced by proteolytic processing of the proteins once they are targeted to the mitochondrial membrane. Purified yeast Mgm1 protein has been shown to assemble into low-order oligomers and displays GTPase activity. Although Mgm1 lacks a PH domain, it preferentially binds to negatively charged phospholipids that are typically found in mitochondrial membranes, which suggests that Mgm1 contains a lipid-binding motif (Meglei and McQuibban, 2009). Furthermore, the dynamins of the inner mitochondrial membrane, together with scaffolding proteins such as prohibitins, are thought to be involved in the maintenance of cristae morphology, possibly by assembling into higher-order structures (Merkwirth and Langer, 2008). Because mitochondria are not only the cellular powerhouse but are also central for apoptosis, tight regulation of cytochrome-c pools via cristae junction morphology and mitochondrial dynamics is indispensable; therefore, regulation of mitochondrial fusion and fission events probably involves many other components.
Dynamins that are associated with mitochondrial fusion have not been found in plants, which suggests that this process might occur by a different mechanism (Sheahan et al., 2005). Although a Fzo1-like protein (FZL) has been identified, it was shown to be involved in the determination of thylakoid and chloroplast morphology, but not that of mitochondria (Gao et al., 2006).
In Arabidopsis, members of the DRP1 protein family (known as the phragmoplastin family in soybean) are involved in cytokinesis and cell expansion. They are thought to generate dumbbell-shaped tubular structures located at the cell plate, which is the zone that generates a new cell wall between dividing cells. Such tubulation would prevent the fusion of Golgi-derived vesicles into a large vacuole and would instead give rise to the outward-growing, dividing cell plate (Verma and Hong, 2005).
Members of the Mx GTPase proteins form a separate group within the dynamin superfamily. They have the three core dynamin domains and the ability to form oligomers, but they also contain unique sequence motifs. Mx proteins are found in most vertebrate species and confer resistance to a variety of viruses following the induction of their expression in response to interferon. The antiviral mechanism is thought to occur through the binding of Mx proteins to essential viral components. Mx proteins are not only found in the cytoplasm but also at nuclear sites, such as the nuclear pores and inside the nucleus, which suggests that they might have a role in regulating the transport of ribonucleo-protein particles between nucleus and cytoplasm (Haller et al., 2007).
In addition to the dynamins described above, proteins that are distantly related to dynamins have been identified in eukaryotes [guanylate-binding proteins (Ghosh et al., 2006)] and in prokaryotes [bacterial dynamin-like proteins (Low and Lowe, 2006)]. Although the X-ray structures of these proteins have been solved, their function remains unknown.
Not a lone ranger – dynamin partners
Although there is a great deal of information available regarding the function of dynamins, understanding exactly how they operate in the different cellular settings remains challenging owing to the complexity of their interactions with the cellular machinery. Dynamins are intrinsic GTPases that require neither exchange factors to replace GDP nor other proteins to hydrolyze GTP. Nonetheless, their activity depends on protein-protein interactions that are based on oligomerization and self-assembly. In addition to dynamin-dynamin and dynamin-lipid interactions, most dynamins have been shown to interact with an ever-expanding number of accessory proteins (Kruchten and McNiven, 2006; Kim and Chang, 2006) (panels 6 and 7 in the poster). Dynamin partners can be grouped into two distinct classes: modifiers and binding partners.
Dynamin modifiers include kinases, phosphatases, ubiquitin ligases, deubiquitylases, small ubiquitin-like modifier (SUMO) ligases and proteases, and constitutively active or induced proteases. These enzymes regulate dynamin activity within the complex network of dynamin-protein interactions.
Human dynamin-1 undergoes reversible phosphorylation at two serine residues in its PRD. Dephosphorylation of dynamin-1 occurs during synaptic-vesicle endocytosis in nerve terminals and is stimulus-dependent. It has been shown to be associated with the recruitment of syndapin 1 to the endocytic protein complex, in which syndapin 1, an F-BAR protein, possibly has membrane-tubulating function (Anggono et al., 2006). Similarly, DRP1 is reversibly phosphorylated by at least three kinases (CaMKI, CDKI and PKA), and these modifications are thought to regulate mitochondrial dynamics and cell-death pathways (Han et al., 2008). Furthermore, DRP1 has been shown to undergo SUMOylation by conjugation to SUMO1, a modification that is associated with an increase in DRP1-mediated mitochondrial fission. This modification is removed by the protease SENP5 (Zunino et al., 2007). Other dynamins are regulated by proteolytic events. In yeast, the integral membrane protein Fzo1 controls fusion of the mitochondrial outer membrane. Cellular levels of Fzo1 are regulated through ubiquitin-dependent targeting of Fzo1 for proteasome-mediated degradation (Cohen et al., 2008). As in the case of Mgm1 in yeast (Sesaki et al., 2006), proteolysis has also been implied in post-translational processing and regulation of the mammalian mitochondrial dynamin OPA1. At least three proteases (PARL, paraplegin and Yme1) might be involved in the constitutive processing of OPA1 once it is inserted into the mitochondrial membrane. Increased cleavage of OPA1 coincides with mitochondrial fragmentation, which indicates that OPA1 can be inactivated by inducible proteases (Griparic et al., 2007).
In the case of classical dynamins, binding partners interact with the PRD of dynamin through SH3 domains. Dynamin-related proteins lack a PRD and bind by other as-yet-unknown mechanisms. Most binding partners mediate interactions either with other proteins or with membranes.
At the outer mitochondrial membrane, the proteins Mdv1 and Caf4 link the mitochondrial dynamin Dnm1 to the integral membrane protein Fis1 for mitochondrial fission in yeast. Mdv1 or Caf4 binds to Fis1 through a helix-loop-helix motif and to oligomeric Dnm1 through a WD40 repeat, thereby connecting soluble Dnm1 to the outer mitochondrial membrane (Naylor et al., 2006; Zhang and Chan, 2007). The mitochondrial outer membrane protein Ugo1, for which a mammalian ortholog has not been identified, physically links Mgm1 and Fzo1 (Okamoto and Shaw, 2002; Hoppins et al., 2009). Another example of a dynamin-binding partner is actin-binding protein 1 (ABP1), which binds to human dynamin-2 through its SH3 domains and to F-actin through its actin-binding domains. By linking the actin cytoskeleton and endocytic proteins, ABP1 is thought to initiate the internalization of B-cell receptors, thereby regulating antigen processing and presentation (Onabajo et al., 2008).
The majority of binding partners of classical dynamins contain BAR or F-BAR domains and, therefore, have membrane-bending or -sensing properties (Heath and Insall, 2008). Amphiphysin and endophilin are two examples of BAR proteins that tubulate lipids and bind to dynamin. Amphiphysin is thought to target dynamin to clathrin-coated pits and to modulate dynamin self-assembly (Ren et al., 2006). Endophilin is thought to recruit dynamin to sites of sensed or induced curvature such as those found at the nascent necks of clathrin-coated pits (Gallop et al., 2006). Notably, members of both the BAR and F-BAR families bind not only to dynamin but also to Wiskott-Aldrich proteins (WASP, N-WASP), which are scaffolding proteins that recruit actin and the actin-related protein 2/3 (ARP2/3) complex. Collectively, these findings indicate that dynamin-mediated vesiculation and actin polymerization are intricately linked (Takenawa and Suetsugu, 2007).
When things fall apart – dynamins and diseases
The fact that dynamins are involved in key cellular processes suggests that mutations in dynamins, or alterations in their expression, could interfere with their specific functions and might be associated with diseases. The earliest evidence for the involvement of dynamins in human disease came from studies that investigated the molecular origins of myopathies. Genetic analysis revealed that mutations in human dynamin-2 were connected with defects in endocytosis. On the basis of clinical presentation, it is possible to discern two basic types of disease phenotypes: centronuclear myopathy (CNM), a slow-progressing disease of the muscle, and Charcot-Marie-Tooth type B (CMTB), comprising a diverse group of peripheral neuropathies. The CMTB phenotype is exclusively caused by mutations located in the dynamin PH domain, which suggests that the mutated protein expressed by patients with this disease is defective in dynamin-lipid binding. By contrast, mutations in the dynamin middle domain and in the far C-terminal portion of the PH domain lead to the CNM phenotype, which suggests that there are defects in dynamin assembly in patients that carry this mutation (Zuchner et al., 2005; Bitoun et al., 2005; Fabrizi et al., 2007) (panel 2 in the poster). Interestingly, a recently discovered mutation that is associated with a CNM phenotype maps to the GED, which suggests a more complex relationship between the location of the dynamin-2 mutation and disease phenotype (Bitoun et al., 2009).
Mutations in the dynamin-related protein MFN2 are associated with Charcot-Marie-Tooth disease CMT type 2A, and with hereditary motor and sensory neuropathy type VI. Both of these diseases lead to axonal degeneration, and the latter is coupled with optical neuropathy. In these diseases, MFN2 mutations are located in the GTPase domain and in the C-terminal coiled-coil domain, which is implicated in binding to the coiled-coil domain of MFN1. This suggests that the mutated protein is defective either in its GTPase activity or in its capacity to tether to fusion partners such as MFN1 during mitochondrial fusion (Verhoeven et al., 2006). It is thought that mutations in MFN2 perturb mitochondrial dynamics or possibly alter the axonal transport of mitochondria.
The most common origin of optic atrophy, known as autosomal dominant optic atrophy (ADOA), is caused by loss-of-function mutations in the mitochondrial dynamin OPA1 and is associated with a progressive loss of vision. A subset of mutations that occur in the GTPase domain of OPA1 are also linked to deafness (Amati-Bonneau et al., 2009). It is speculated that affected nerve cells in the optical or auditory system are much more susceptible to disturbances in the arrangement of mitochondria than the cells in other tissues, and that the resulting changes in bioenergetics and the induction of apoptosis cause the observed detrimental effects on vision and hearing (Olichon et al., 2006).
The importance of organelle dynamics for normal physiology is also highlighted by the finding that a single mutation in the middle domain of the human dynamin DRP1 was associated with a severe defect in the fission of mitochondria and peroxisomes, and resulted in neonatal lethality (Waterham et al., 2007). In contrast to the late-onset progressive disorders described above, this DRP1-dependent disorder developed rapidly and led to defects in multiple tissues.
Human dynamin-1 plays an important role in synaptic-vesicle endocytosis and in neurotransmission. A mutation in the GTPase domain of canine dynamin-1 has been shown to cause the condition of reversible collapse, which is caused by a deficiency in sustained synaptic transmission during heightened neural activity. Dysfunction of dynamin-1 is now considered a candidate theory to explain the syndrome of exercise-induced collapse (EIC) in Labrador retrievers (Patterson et al., 2008).
Finally, it has been shown that the expression levels of dynamin-1 are decreased in the brains of patients with Alzheimer's disease (Yao et al., 2003). It has been proposed that amyloid-β peptides, the expression of which is reportedly increased in the brains of these patients, might induce the decrease in dynamin levels by stimulating calpain-mediated cleavage of dynamin-1 (Kelly et al., 2005). In addition, there is evidence that late-onset Alzheimer's disease is associated with a single nucleotide polymorphism in intron 1 of dynamin-2, and that the level of dynamin-2 mRNA is reduced in patients with dementia by an as-yet-unknown mechanism. The decrease in dynamin expression is thought to be associated with an increase in the levels of amyloid-β, similar to the effect observed with a dysfunctional dynamin-2 mutant, thereby possibly generating a vicious cycle that could contribute to the observed amyloid pathology (Kamagata et al., 2009).
A molecular understanding of how dynamins operate has been hampered by the fact that is has been very difficult to obtain structural information at the atomic level even for individual domains. Dissecting the complex interaction networks in which dynamins are involved on a cellular level is also a daunting task, because their partners are numerous, transient and probably vary with cell type, tissues, species or metabolic state. Finally, much has to be learned about how expression levels, distribution of splice variants and the activities of dynamins are regulated.
Supplementary material available online at http://jcs.biologists.org/cgi/content/full/122/19/3427/DC1
The laboratory is supported by the intramural research program of the NIDDK, National Institutes of Health. Deposited in PMC for release after 12 months.
- © The Company of Biologists Limited 2009