Motor proteins at the mitochondria–cytoskeleton interface

1Mitochondria are the powerhouses of the cell, generating adenosine triphosphate (ATP) by oxidative phosphorylation (OXPHOS). In addition to producing energy, they also play critical roles in the biosynthesis of macromolecules, Ca2+ homeostasis, and signalling in programmed cell death and immunity. Mitochondria are double-membraned organelles of endosymbiotic origin that have an outer mitochondrial membrane (OMM) facing the cytosol and an inner mitochondrial membrane (IMM) marking the boundary of the mitochondrial matrix, which contains the mitochondrial DNA (mtDNA). The intermembrane space (IMS) is sandwiched between the OMM and IMM. Mitochondria can sense stress stimuli, such as nutrient deprivation, and the overall metabolic state of the cell. To maintain organellar homeostasis, mitochondria constantly remodel their network through fission and fusion events, reposition themselves through motor protein- and cytoskeleton-dependent transport, and finally are degraded via several quality control mechanisms. Indeed, defects in mitochondrial homeostasis manifesting as fission–fusion imbalance, impaired transport or reduced clearance of faulty mitochondria by mitophagy increase with age and can lead to a wide range of neurodegenerative disorders. The overall control of mitochondrial homeostasis involves three types of motor proteins that translocate along cytoskeletal tracks: myosins moving along actin filaments (Box 1), as well as kinesins and dynein moving along microtubules (MTs; Box 2; ). The key adaptor complex linking kinesin-1 and dynein to mitochondria consists of Miro1 or Miro2 (referred to collectively as Miro1/2; also known as RHOT1 or RHOT2, respectively) and trafficking kinesin protein 1 or 2 (TRAK1 or TRAK2, referred to collectively as TRAK1/2), and is hereafter referred to as the Miro–TRAK complex. Interestingly, Miro1/2 without TRAK1/2 can also recruit myosin of class XIX (MYO19) to mitochondria. Actin monomers polymerize to form actin filaments. A host of actin-binding proteins regulate filament dynamics and the assembly into higher-ordered structures, such as stress fibres, lamellipodia, filopodia and actin networks associated with the plasma membrane or intracellular membranes. These include: (1) actin nucleators, such as the ARP2/3 complex, formins and tandem WH2 domain-containing proteins, such as mitoSPIRE; (2) actin filament cross-linking or bundling proteins, such as fascin; (3) actin disassembly factors; and (4) monomer-sequestering proteins.Myosins are a diverse family of molecular motors that undergo directed movement along actin filaments; they form dynamic tethers between cellular membrane compartments and the actin cytoskeleton. In addition, several classes of myosin motors are actively involved in regulating organization of the actin cytoskeleton. A total of 39 myosins belonging to 12 classes are expressed in humans. All known myosins, regardless of species, have the same basic structure: an N-terminal motor domain, a lever arm and a tail domain (see figure showing the domain organization of myosin V). In vitro experiments with myosins indicate that their behaviour broadly falls into three types – transporters, tethers or contractile motors. The lever arm, which typically contains one or more calmodulin- or light-chain-binding IQ motifs, amplifies structural changes in the motor domain into large steps along the actin track. The tail domains either self-assemble to form filaments or contain the majority of binding sites for lipid membranes and cargo adaptor proteins, which recruit the motor to different organelles and vesicular compartments.

2Actin monomers polymerize to form actin filaments. A host of actin-binding proteins regulate filament dynamics and the assembly into higher-ordered structures, such as stress fibres, lamellipodia, filopodia and actin networks associated with the plasma membrane or intracellular membranes. These include: (1) actin nucleators, such as the ARP2/3 complex, formins and tandem WH2 domain-containing proteins, such as mitoSPIRE; (2) actin filament cross-linking or bundling proteins, such as fascin; (3) actin disassembly factors; and (4) monomer-sequestering proteins.

3Myosins are a diverse family of molecular motors that undergo directed movement along actin filaments; they form dynamic tethers between cellular membrane compartments and the actin cytoskeleton. In addition, several classes of myosin motors are actively involved in regulating organization of the actin cytoskeleton. A total of 39 myosins belonging to 12 classes are expressed in humans. All known myosins, regardless of species, have the same basic structure: an N-terminal motor domain, a lever arm and a tail domain (see figure showing the domain organization of myosin V). In vitro experiments with myosins indicate that their behaviour broadly falls into three types – transporters, tethers or contractile motors. The lever arm, which typically contains one or more calmodulin- or light-chain-binding IQ motifs, amplifies structural changes in the motor domain into large steps along the actin track. The tail domains either self-assemble to form filaments or contain the majority of binding sites for lipid membranes and cargo adaptor proteins, which recruit the motor to different organelles and vesicular compartments.

4MTs are rigid but dynamic polymers, growing and shrinking through the addition or loss of α-tubulin–β-tubulin heterodimers, which provide the tracks for polarized transport.Kinesins are a diverse family of motor proteins that transport cargo unidirectionally along MTs (mostly towards the plus end) and regulate MT network organization. The kinesin superfamily consists of 15 classes, with over 45 known kinesins expressed in mammalian cells. All kinesins share the same basic domain organization: a conserved head domain that is structurally homologous to the myosin motor domain, followed by the neck linker, a divergent stalk, and a tail domain (; see figure, top). Cargo can either bind directly or through adaptor proteins to the kinesin tail or to one of the kinesin light chains that bind to the kinesin heavy chain. Kinesins can be grouped according to the position of their motor domain. Kinesins with motor domains at either the N- or C-terminus drive plus- or minus-end directed transport, respectively, whereas kinesins with the motor domain in the middle of the protein are involved in depolymerizing MTs.Cytoplasmic dynein 1 (referred to as dynein) drives transport of a wide range of cargos towards the minus-end of MTs. Human dynein is a 1.5 MDa complex comprising the dynein heavy chain (DHC, also known as DYNC1H1), the dynein intermediate chain (DIC), the dynein light intermediate chain (DLIC) and three classes of dynein light chains (DLCs; Robl, LC8 and Tctex) (see figure, middle). The DHC consists of an N-terminal tail domain that is important for their dimerization and that binds to DIC and DLIC, and a linker region connecting the tail domain to the C-terminal motor domain, which comprises a ring of six AAA+ domains, a coiled-coil stalk and a MT-binding domain (MTBD). Dynein motility along MTs is powered by ATP hydrolysis in the AAA+ ring; however, processive movement of dynein requires the co-factor dynactin (see figure, bottom). Dynactin is a 1.1 MDa protein complex with 23 subunits assembled around an actin-like filament (ARP1 filament) and p150glued (also known as DCTN1), which can also bind to MTs. Cargo attachment to dynein requires long coiled-coil activating adaptors, such as BICD, Hook, and Spindly proteins, which bind to the C-terminus of DLIC and run along the dynactin filament.

5MTs are rigid but dynamic polymers, growing and shrinking through the addition or loss of α-tubulin–β-tubulin heterodimers, which provide the tracks for polarized transport.

6Kinesins are a diverse family of motor proteins that transport cargo unidirectionally along MTs (mostly towards the plus end) and regulate MT network organization. The kinesin superfamily consists of 15 classes, with over 45 known kinesins expressed in mammalian cells. All kinesins share the same basic domain organization: a conserved head domain that is structurally homologous to the myosin motor domain, followed by the neck linker, a divergent stalk, and a tail domain (; see figure, top). Cargo can either bind directly or through adaptor proteins to the kinesin tail or to one of the kinesin light chains that bind to the kinesin heavy chain. Kinesins can be grouped according to the position of their motor domain. Kinesins with motor domains at either the N- or C-terminus drive plus- or minus-end directed transport, respectively, whereas kinesins with the motor domain in the middle of the protein are involved in depolymerizing MTs.

7Cytoplasmic dynein 1 (referred to as dynein) drives transport of a wide range of cargos towards the minus-end of MTs. Human dynein is a 1.5 MDa complex comprising the dynein heavy chain (DHC, also known as DYNC1H1), the dynein intermediate chain (DIC), the dynein light intermediate chain (DLIC) and three classes of dynein light chains (DLCs; Robl, LC8 and Tctex) (see figure, middle). The DHC consists of an N-terminal tail domain that is important for their dimerization and that binds to DIC and DLIC, and a linker region connecting the tail domain to the C-terminal motor domain, which comprises a ring of six AAA+ domains, a coiled-coil stalk and a MT-binding domain (MTBD). Dynein motility along MTs is powered by ATP hydrolysis in the AAA+ ring; however, processive movement of dynein requires the co-factor dynactin (see figure, bottom). Dynactin is a 1.1 MDa protein complex with 23 subunits assembled around an actin-like filament (ARP1 filament) and p150glued (also known as DCTN1), which can also bind to MTs. Cargo attachment to dynein requires long coiled-coil activating adaptors, such as BICD, Hook, and Spindly proteins, which bind to the C-terminus of DLIC and run along the dynactin filament.

8In this Review, we will discuss the importance of motor proteins and motor-linked track dynamics not only for the movement of mitochondria but also for the positioning of these organelles, for the production of force associated with fission and fusion, and for the three-dimensional organization of the cytoskeleton around mitochondria. We will focus on kinesin-1, kinesin-3 and related family members, cytoplasmic dynein-1, MYO19 and myosin V (MYO5), which are involved in mitochondrial transport, as well as non-muscle myosin II (NMII) and myosin VI (MYO6), which have prominent roles in mitochondrial fission and mitophagy, respectively. Finally, we highlight human diseases with mitochondrial phenotypes caused by mutations in motor proteins that open up potential new avenues for treatment (Box 3). Motor proteins and other members of the machinery important for mitochondrial dynamics that have been linked to human diseases are summarized in Table S1 (see also table 1 in ). For instance, a mutation in MYH14 (encoding the heavy chain of NMIIC) affects mitochondrial fission and leads to a peripheral neuropathy. Fibroblasts from patients harbouring the mutations K671E or I584L in DYNC1H1, which have been linked to the neuromuscular disorder spinal muscular atrophy lower extremity-predominant 1 (SMA-LED), display fragmented and fewer mitochondria with decreased MFN1 levels. KIF5A mutations causing myoclonus also lead to mitochondrial abnormalities and manifest as atrophy of type I muscle fibres and complex IV deficiency. Moreover, mutations in the motor adaptor TRAK1 are linked to epileptic encephalopathy early infantile 68; in TRAK1-deficient patient fibroblasts, the mitochondria are irregularly distributed and have reduced motility, decreased membrane potential and a lower respiratory capacity. Interestingly, Miro1/2 and MYO19 have not been linked to any Mendelian diseases. In many cases where there are disease-causing mutations in motor proteins, it remains to be established whether the motor-linked disruption in mitochondrial homeostasis is also the underlying cause of the disease. Nevertheless, the emerging importance of motor proteins in maintaining mitochondrial homeostasis might open new avenues for treatment of mitochondria-linked diseases, as motor proteins are attractive drug targets.

9Motor proteins and other members of the machinery important for mitochondrial dynamics that have been linked to human diseases are summarized in Table S1 (see also table 1 in ). For instance, a mutation in MYH14 (encoding the heavy chain of NMIIC) affects mitochondrial fission and leads to a peripheral neuropathy. Fibroblasts from patients harbouring the mutations K671E or I584L in DYNC1H1, which have been linked to the neuromuscular disorder spinal muscular atrophy lower extremity-predominant 1 (SMA-LED), display fragmented and fewer mitochondria with decreased MFN1 levels. KIF5A mutations causing myoclonus also lead to mitochondrial abnormalities and manifest as atrophy of type I muscle fibres and complex IV deficiency. Moreover, mutations in the motor adaptor TRAK1 are linked to epileptic encephalopathy early infantile 68; in TRAK1-deficient patient fibroblasts, the mitochondria are irregularly distributed and have reduced motility, decreased membrane potential and a lower respiratory capacity. Interestingly, Miro1/2 and MYO19 have not been linked to any Mendelian diseases. In many cases where there are disease-causing mutations in motor proteins, it remains to be established whether the motor-linked disruption in mitochondrial homeostasis is also the underlying cause of the disease. Nevertheless, the emerging importance of motor proteins in maintaining mitochondrial homeostasis might open new avenues for treatment of mitochondria-linked diseases, as motor proteins are attractive drug targets.

10The dynamic positioning of mitochondria within different cell types and tissues is spatially regulated to fulfil subcellular requirements for local energy production and Ca2+ signalling, which is particularly important in neurons due to their extended morphology. Mitochondrial movement over long distances in axons or dendrites is dependent on kinesin and dynein motors moving in opposite directions along MTs due to the intrinsic track polarity. Fast axonal transport with velocities between 0.25 and 1 µm/s is driven in the anterograde direction from the cell body to the distal part of the axon and the synapse by plus-end-directed kinesins, whereas retrograde movement depends on the minus-end-directed dynein motor moving back towards the soma (cell body). This bidirectional transport of mitochondria along axonal MTs, which characteristically displays frequent pauses and directional changes, and the docking mechanisms that ensure the correct distribution of mitochondria throughout the neuron, is fundamentally the same in all eukaryotic cells as it uses equivalent motors and machinery. In contrast to the uniform MT polarity in the axon, neuronal dendrites contain MTs of mixed polarity with approximately equal numbers of plus- and minus-ends of MTs directed towards the cell surface; hence, the track orientation determines the direction of transport and the motor involved. Mitochondrial transport has been extensively studied in neurons (for examples, see recent reviews by; ), and we will focus here only on the key motors and mitochondrial adaptors (Fig. 1).

11Kinesin-1 (which in mammals has three heavy chain isoforms – the ubiquitously expressed KIF5B, and neuron-specific KIF5A and KIF5C) is the primary motor for anterograde mitochondrial transport (Fig. 1) in axons in Drosophila, mouse and zebrafish. Kinesin-1 is a highly processive motor that takes over 100 steps before detaching from the MT track in vitro. Within the cell, kinesin-1 is activated and the run length increases upon binding to TRAK1, thereby promoting kinesin-1-dependent long-range transport of mitochondria in the crowded cytosol.

12In contrast to the well-documented role of kinesin-1, the importance of other classes of kinesins for mitochondrial transport remains controversial. The highly processive kinesin-3 protein KIF1B has been suggested to transport mitochondria in vitro, although no defects in mitochondrial transport are observed in KIF1B-mutant mice or zebrafish. Furthermore, one publication has described changes in mitochondrial morphology that are linked to the depletion or overexpression of the dominant-negative tail domain of kinesin-like protein 6 (KLP6, also known as KIF28P), a distantly related member of the kinesin-3 family. Although expression of mutant KLP6 also affects mitochondrial distribution in neurites, further work is required to establish the role of KLP6 in mitochondrial transport. Finally, loss of KIF1-binding protein (KIFBP, also known as KBP and KIF1BP), the adaptor protein for both kinesin-3 and KLP6, has been reported to change mitochondrial distribution in some specific cell types and developmental stages. Since KIFBP alters MT dynamics and mitochondrial biogenesis, it is more likely to play an indirect role in mitochondrial transport by modulating MT tracks.

13Dynein is the main motor driving the retrograde transport of axonal mitochondria (Fig. 1). Interestingly, this activity also requires kinesin-1, which delivers newly synthesized dynein from the soma via anterograde slow axonal transport out to the synapse. The majority of disease-associated dynein mutations affect overall retrograde dynein-dependent motility, whereas mutations causing loss of function in the dynactin subunit ACTR10 (Box 2) selectively impair binding of dynein to mitochondria in zebrafish.

14Microtubule and myosin motors are bound to the mitochondrial surface by the conserved Miro–TRAK adaptor complex (Fig. 1), which is the key regulator of axonal transport of mitochondria in many different organisms. TRAK1 and TRAK2, the mammalian homologues of the Drosophila Milton protein, bind directly to Miro proteins, which are anchored to the OMM via a C-terminal transmembrane domain. In mammals, two Miro proteins are expressed, Miro1 and Miro2, and both have two GTPase domains flanking two Ca2+-sensing EF-hand domains. The TRAK family has an extensive N-terminal coiled-coil region similar to huntingtin-associated protein (HAP1) and a C-terminal domain that binds to Miro proteins. Whereas TRAK1 predominantly mediates axonal transport by simultaneously interacting with both the cargo-binding tails of kinesin-1 and dynein, TRAK2 promotes dendritic transport by forming a complex with dynein in a conformation that interferes with kinesin-1 binding.

15Interestingly, Miro proteins not only regulate transport, but also play a role in maintaining mitochondrial morphology and the architecture of the IMM, since Miro1/2 double-knockout (KO) mouse embryonic fibroblasts (MEFs) have a non-uniform arrangement of mitochondrial cristae.

16Finally, in addition to mitochondrial motility, kinesin-1, together with TRAK1/2, also drives the tubulation of mitochondria in the cell periphery. This process requires kinesin-1 motor activity and gives rise to highly dynamic thin tubules that fuse to form lattices and eventually generate an interconnected mitochondrial network. Indeed, overexpression of TRAK1/2 leads to striking changes in mitochondrial network morphology, inducing the formation of extended, highly dynamic mitochondrial tubules in a MT-dependent manner.

17For mitochondria to become stationary, their movement is either simply paused or they are actively anchored to either the actin or MT track. Mitochondrial positioning and docking at locations of nutrient abundance increases the efficiency of ATP generation and is caused by stalling due to TRAK1 glycosylation (O-GlcNAcylation) at high glucose levels. Miro proteins can also act as Ca2+ sensors and induce mitochondrial immobilization by a Ca2+-dependent mechanism, for example, at synaptic sites. Ca2+ binding to the EF-hand motifs of Miro1/2 triggers conformational changes that arrest mitochondrial movement by disrupting interactions between kinesin-1 and MTs, or by leading to dissociation of the Miro–TRAK complex from kinesin-1. Elevated Ca2+ levels also activate anchoring proteins that mediate docking of mitochondria, such as syntaphilin (SNPH). Syntaphilin, previously thought to be axonal-specific, is inserted into the OMM with its C-terminal transmembrane domain and directly binds to MTs via its N-terminal MT-binding domain. It functions as a brake to axonal mitochondrial transport by competing with TRAK1/2 for kinesin-1 binding and, once bound, inhibiting kinesin-1 motor activity. The dynein light chain (DLC) LC8 (also known as DYNLL1) is thought to enhance SNPH-based anchoring of mitochondria.

18Mitochondrial positioning and immobilization may require the inactivation of kinesin and dynein motors, but also involves actin-based myosin motor proteins, which may act as short-range transporters or may regulate mitochondrial tethering. Disrupting the actin cytoskeleton with drug treatments has been shown to reduce mitochondrial anchoring in axons, and disruption of the actin- or MT-based cytoskeleton destabilizes mitochondrial tethering in dendrites. MYO6 (also known as Jaguar in Drosophila) is thought to oppose MT-based mitochondrial movement to facilitate organelle docking in Drosophila neurons, in line with recent findings that the cargo-binding domain of MYO6 directly binds to SNPH (Fig. 1). MYO6 is a unique myosin that moves towards the minus ends of actin filaments, in the opposite direction to all other known myosins. MYO6 has essential roles in membrane trafficking pathways and mediates the delivery of endosomal membranes to autophagosomes during autophagosome maturation. The direct interaction between MYO6 and SNPH suggests a new role for MYO6 in capturing mitochondria on F-actin and mediating the cytoskeletal switch from MTs to actin, which serves as a platform for anchoring organelles in presynaptic terminals. MYO6-dependent mitochondrial capture is regulated by AMP-activated protein kinase (AMPK)–p21-activated kinase (PAK) signalling pathways and phosphorylation of MYO6 in the motor domain.

19A role for MYO5 in mitochondrial docking was observed in Drosophila, where depletion of MYO5 (encoded by didum in Drosophila) causes an increase in anterograde and retrograde axonal mitochondrial motility. Myosins of class V (MYO5A, -B and -C) function as a dimer and move processively along actin filaments. Each MYO5 heavy chain binds six calmodulins, giving rise to a long lever arm, which allows this myosin to take large steps of 36 nm and move straight along the helical actin track. MYO5A can bind to a mitochondrially-anchored isoform of the actin nucleator and tandem WH2 domain-containing protein SPIRE1 (mitoSPIRE), which cooperates with formins to assemble actin filaments (Fig. 1) ( preprint). These actin–myosin networks are thought to oppose mitochondrial motility by anchoring mitochondria, since loss of mitoSPIRE in MEFs increases mitochondrial motility ( preprint).

20In vertebrates, MYO19 is tightly associated with mitochondria and regulates the cellular distribution of this organelle (Fig. 1). Overexpression of full-length MYO19 significantly increases the overall mitochondrial motility over short distances in epithelial cells. In neuronal cells, however, MYO19 causes a decrease in mitochondrial movement by inducing track switching from MTs to actin filaments, which results in reduced mitochondrial run length. MYO19 displays a slow rate of ADP release linked to a prolonged actin-binding state and high duty ratio in vitro. The very short unique C-terminal tail of MYO19 contains a lipid-binding region of 30–45 amino acids essential for mitochondrial targeting. In addition, its far C-terminal tail directly interacts with Miro1/2 and competes with TRAK1/2 for binding. Thus, Miro proteins not only form a complex with MT motors but also with MYO19, thereby coordinating MT- and actin-based mitochondrial movement. This is confirmed by studies on fibroblasts isolated from Miro1/2 double-KO mice, which not only display reduced retrograde and anterograde mitochondrial movement along MTs, but have also lost MYO19 from mitochondria. The N-terminal GTPase domain of Miro1/2 is critical for recruiting and stabilizing MYO19 on the OMM by facilitating its membrane insertion and protecting it from proteasome-mediated degradation. Finally, MYO19 is also able to regulate the precise localization of mitochondria in response to metabolic stimuli, such as glucose starvation or reactive oxygen species (ROS), which activate MYO19 and lead to re-localization of mitochondria into the tips of filopodia.

21As discussed above, the correct distribution of mitochondria is particularly important in highly polarized neuronal cells. In the next section, we highlight other specialized adaptations regulating mitochondrial positioning.

22Mitochondrial transport in cardiac myoblasts is driven by the same players (kinesin-1, dynein, TRAK1/2) as in neuronal cells and is regulated by similar Ca2+-dependent mechanisms. Indeed, Ca2+ is a key regulator of contractile activity, and elevated cytoplasmic Ca2+ resulting from ionomycin or thapsigargin treatment inhibits mitochondrial transport.

23Kinesin-1-mediated mitochondrial transport is also important during cardiac maintenance. Cardiac hypertrophy (cardiomyocyte growth) is an adaptive response aimed at reducing wall stress and maintaining cardiac function. Kinesin-1 expression is upregulated in a mouse model of pathological cardiac hypertrophy and in neonatal rat ventricular cardiomyocytes. In these cardiomyocytes, mitochondria are re-localized to the cell periphery accompanied by increased mitochondrial respiration associated with the hypertrophic response, which can be reverted by depletion of kinesin-1. Interestingly, altered mitochondrial morphology and distribution are hallmarks of heart failure.

24The intercellular communication between cardiomyocytes and cardiac myofibroblasts is important for maintaining normal myocardial function and is facilitated by membrane nanotubes (MNTs), which are long, thin membrane-based connections. These connections enable kinesin-1-dependent transport of mitochondria from cell to cell, thus acting as direct highways for the exchange of mitochondria. Kinesin-1-mediated transport of mitochondria along MTs in MNTs is physiologically important in ischaemic cardiomyopathy for rescuing cardiomyocytes from hypoxia–reoxygenation-induced apoptosis.

25Migrating cells have a radial MT organization with plus ends facing the cell periphery. In migrating lymphocytes, the centrosome, the main MT-organizing centre (MTOC), is located in the uropod at the rear or trailing end of the cell, whereas in motile epithelial cells, the centrosome is oriented towards the migration axis, establishing a high-density MT network towards the leading edge.

26In lymphocytes or T cells, mitochondria accumulate in the energy-demanding uropod, where they provide ATP for actomyosin II-dependent force generation required for cell movement. The Miro1–dynein complex is important in controlling lymphocyte and T cell migration by regulating mitochondrial accumulation around the MTOC at the uropod. Another example of polarized distribution of mitochondria can be observed in immune cells during the formation of the immunological synapse between T cells and antigen-presenting cells. Mitochondria regulate sustained Ca2+ influx, and kinesin-1 drives anterograde mitochondrial transport to the immunological synapse at the plasma membrane. Miro proteins also regulate mitochondrial recruitment in leukocytes to sites of contact with inflamed endothelial cells.

27In cancer cells, mitochondrial positioning is crucial for facilitating the energy-expensive processes of dissemination and invasion. Mitochondria provide localized energy production by infiltrating the leading edge lamellipodia of migrating epithelial cancer cells in a MT-dependent manner. Anterior localization of mitochondria correlates with faster cell movement and increased directional persistence, which may account for the ability to invade further and metastasize earlier. Kinesin-1 and Miro1/2 are responsible for transporting mitochondria to the leading edge, since their depletion suppresses mitochondrial re-localization and results in slower cell migration rates as well as reduced tumour cell invasion. These findings are supported by observations in Miro1 KO MEFs, which are slower in collective and single-cell migration assays as a result of impaired membrane ruffling, leading-edge protrusion and focal adhesion dynamics. Interestingly, downregulation or loss of the SNPH anchor enhances tumour progression in humans, whereas SNPH expression in tumour cells inhibits mitochondrial transport and blocks metastasis in vivo.

28Kinesins and their mitochondrial adaptor proteins are also involved in the spatial redistribution of mitochondria during oocyte maturation and spermatogenesis. Mitochondria in many species (human, Xenopus and Drosophila) enter oocytes from interconnected germ cells to generate the Balbiani body, a non-membrane-bound compartment packed with mitochondria. In Drosophila, oocyte acquisition of mitochondria depends on kinesin-1- and Milton-dependent movement, since mutations in these proteins cause premature and excessive mitochondrial transport into the oocyte. During Drosophila oogenesis, the long isoform of the Oskar protein tethers mitochondria to the posterior cortex at the site of primordial germ cell formation through an actin-dependent mechanism requiring the non-muscle tropomyosin, Tropomyosin II; however, at present, it is not known whether any myosin motors are involved in this process. During meiotic maturation of mouse oocytes, kinesin-3 is responsible for mitochondrial redistribution near the metaphase plate.

29In Drosophila spermatogenesis, mitochondria undergo dramatic rearrangements, during which they aggregate and fuse to form the spherical Nebenkern. Milton has been implicated in anchoring mitochondria of the Nebenkern at the minus end of MTs to the nucleus. Furthermore, in Milton-mutant flies, the unfurling and elongation of mitochondria along MTs is impaired during axonemal growth. Finally, in the testis of some fish species (Larimichthys polyactis and Boleophthalmus pectinirostris), kinesin-2 (specifically KIF3A) is highly expressed and colocalizes with mitochondria in the nuclear periphery during spermiogenesis and in the midpiece of mature sperm, suggesting an active role in mitochondrial positioning during sperm maturation.

30During cell division, mitochondria are divided between the two daughter cells by either active or passive mechanisms. In metaphase, phosphorylation triggers the release of the dynein and kinesin motors from the Miro–TRAK adaptor complex on the mitochondrial surface, causing mitochondria to dissociate from MTs, ensuring symmetrical mitochondrial distribution and inheritance. Thus, passive inheritance results from the active release of motor proteins from mitochondria. In contrast, at later stages of mitosis, both MT- and actin-based motors play an active role in mitochondrial segregation between the two daughter cells. During anaphase, MYO19 has been shown to tether mitochondria to actin filaments, thereby ensuring equal distribution between the two daughter cells. In cells undergoing cytokinesis, mitochondria are linked via kinesin-1 to the growing tips of astral MTs of the mitotic spindle and thereby are transported to the cleavage furrow. In addition, Miro1/2-dependent recruitment of the centromeric protein F (CENP-F), a non-motor MT-binding protein that associates with growing MTs, to mitochondria is required for their transport towards the periphery of the two daughter cells in late cytokinesis. All these processes are likely to involve Miro1/2, which coordinates mitochondrial distribution by MT- and actin-based motors, since Miro1/2 double-KO MEFs display aberrant mitochondrial positioning resulting from asymmetric mitochondrial segregation during mitosis.

31Mitochondria undergo regulated fission and fusion events to maintain a healthy, dynamic network. In mammals, the major mediators of fission and fusion are members of the dynamin-related GTPase family and include mitofusins 1 and 2 (MFN1 and MFN2, respectively), optic atrophy 1 (OPA1) and dynamin-related protein 1 (DRP1, also known as DNM1L).

32Mitochondrial fission first involves separation of the IMM followed by OMM division. Endoplasmic reticulum (ER) tubules wrap around mitochondria, inducing a first constriction at these ER–mitochondria contacts, which mark the sites of mitochondrial fission as well as of active mtDNA replication.

33The machinery driving IMM fission in the matrix is poorly understood; however, the process is triggered by a transient rise in mitochondrial matrix Ca2+ that occurs at ER–mitochondria contact sites, which are generated and stabilized by the actin cytoskeleton. The formin INF2, an actin nucleator embedded in the ER membrane through a prenylated C-terminal CAAX motif, polymerizes actin away from the ER (Fig. 2). These actin filaments together with NMIIA (which has a heavy chain encoded by MYH9), which forms bipolar mini-filaments, are thought to provide the contractile tension to draw the ER and mitochondria closer together, thereby stimulating ER-to-mitochondrial Ca2+ transfer.

34OMM scission requires the dynamic recruitment of DRP1 from the cytosol to tail-anchored proteins that act as DRP1 receptors (Fig. 2), such as the mitochondrial fission factor (MFF) and the mitochondria dynamics proteins 49 and 51 (MiD49 and MiD51, also known as MIEF2 and MIEF1, respectively). DRP1 oligomerizes into ring-like structures around mitochondria, causing initial constriction of the organelle, before GTP hydrolysis leads to further mitochondrial membrane constriction. The final scission event downstream of DRP1 has recently been proposed to involve Golgi-derived vesicles and occur at mitochondria–lysosome contact sites. In addition, several MT-based motors regulate DRP1 recruitment from the cytosol to mitochondria. Kinesin-3 (specifically the KIF1Bβ isoform) drives mitochondrial fission by activating calcineurin, which dephosphorylates DRP1 at Ser637, leading to DRP1 translocation to mitochondria. In contrast, under oxidative stress, kinesin-1 has been shown to enhance DRP1 phosphorylation at Ser616 through JNK activation, which also promotes DRP1 recruitment to mitochondria. Finally, the dynein–dynactin complex also facilitates DRP1 translocation to mitochondria, since disruption of dynein function by overexpression of the dynactin subunit p50 (also known as DCTN2) in HeLa cells leads to accumulation of DRP1 in the cytosol, resulting in the formation of long, highly branched mitochondria. Interestingly, in ageing Dync1h1-mutant mice, mitochondria progressively increase in size and become profoundly dysfunctional in muscle. In contrast to MT-based motors, Miro1/2 negatively regulates DRP1 recruitment to mitochondria, thereby suppressing fission, although the recruitment of fusion-promoting factors may also be affected.

35OMM constriction is mediated by myosin motor proteins and the actin cytoskeleton (Fig. 2). INF2, which directly interacts and cooperates with mitoSPIRE and the actin-bundling protein fascin, drives actin polymerization on the mitochondrial surface. These mitochondrial actin filaments not only directly bind DRP1 and target it to mitochondrial fission sites, but also stimulate DRP1 oligomerization. In melanoma cell lines, mitoSPIRE binds directly to MYO5A, which recruits this motor to mitochondrial fission sites adjacent to DRP1 puncta; MYO5A recruitment facilitates mitochondrial fission, whereas MYO5A depletion leads to mitochondrial elongation ( preprint). NMIIA, NMIIB and NMIIC (which have heavy chains encoded by MYH9, MYH10 and MYH14, respectively) also accumulate at mitochondrial constriction sites and enable DRP1 oligomerization, as inhibition of NMII activity leads to increased mitochondrial length and loss of mitochondrial DRP1. Furthermore, bipolar NMII filaments can pull on actin filaments to provide the tension force for mitochondrial constriction. Interestingly, NMIIC with the R94L heavy-chain mutation, which has been linked to peripheral neuropathies, acts as a dominant-negative protein that inhibits fission. In cardiomyocytes after ischaemia–reperfusion injury, inhibition of myosin II with blebbistatin or loss of NMIIA reduces mitochondrial fission and mitophagy.

36Finally, dynamic assembly and disassembly of actin filaments onto mitochondrial subpopulations has been observed at ER–mitochondria contact sites; this is blocked by inhibition of actin nucleators, such as the ARP2/3 complex and formins. This cycling is suggested to act as a steady-state surveillance method to modulate the mitochondrial fission–fusion balance to promote network remodelling and content mixing.

37OMM fusion requires tethering of MFN1 and MFN2 on the OMM between adjacent mitochondria followed by GTP hydrolysis. IMM fusion is mediated by OPA1, which is inserted into the IMM and is cleaved at S1 and S2 sites by two membrane-bound metalloproteases, OMA1 and YME1L1, resulting in at least five OPA1 fragments (two larger L-OPA1 fragments and three shorter S-OPA1 fragments) thought to regulate fusion.

38The importance of the cytoskeleton for mitochondrial fusion is less well understood. Inhibiting actin filament polymerization or inducing changes in MT dynamics has little effect on fusion kinetics. Recent results, however, highlight a potential role of kinesin-3 in IMM fusion, as it binds directly to YME1L1 and its overexpression promotes cleavage of L-OPA1, which leads to mitochondrial fission and fragmentation triggering nerve growth factor (NGF)-induced apoptosis. However, where binding between kinesin-3 and YME1L1 occurs remains to be established, since kinesin-3 is a cytosolic protein, whereas YME1L1 is present at the IMM.

39With their many links to cytoskeletal motors, Miro proteins appear to be important for maintaining the balance between fission and fusion, which controls mitochondrial morphology; however, whether this is a direct effect or due to an indirect effect on transport requires further investigation. Miro proteins may facilitate fusion, since the overexpression of the wild-type protein, or expression of a constitutively active form of the protein, results in mitochondrial enlargement in Drosophila and induces the formation of long thread-like mitochondria, whereas in yeast, loss of the Miro homologue Gem1 or its GTPase activity leads to fragmented mitochondria and a collapsed network. Miro1/2 and Milton interact with both MFN1 and MFN2, which are also required for mitochondrial transport. Thus, there is crosstalk and coordination between mitochondrial transport and fusion; however, the dependence of mitochondrial fusion on transport has made it difficult to determine the precise role of Miro proteins and MFNs in mitochondrial fusion. Depletion or mutation of MFN2 reduces the rate of anterograde and retrograde mitochondrial transport in neurons, while reducing mitochondrial motility by suppression of Miro1/2 inhibits the rate of mitochondrial fusion.

40Several pathways are known to regulate the turnover of whole mitochondria by mitophagy; one of the best studied pathways involves the serine/threonine kinase PINK1 and the E3 ubiquitin ligase Parkin (PRKN), which when mutated cause autosomal recessive Parkinson's disease. A hallmark of mitochondrial damage is membrane depolarization, leading to the accumulation of PINK1 on the OMM and phosphorylation of ubiquitin attached to OMM proteins. This, in turn, triggers the recruitment of auto-inhibited PRKN to the OMM, where it binds to phosphorylated ubiquitin. PINK1-dependent phosphorylation leads to full PRKN activation, which then allows PRKN to ubiquitylate OMM proteins including Miro1/2, thereby amplifying the mitophagy signal. Cargo-selective autophagy receptors, such as optineurin (OPTN), Ca2+-binding and coiled-coil domain-containing protein 2 (CALCOCO2, also known as NDP52), and Tax1-binding protein 1 (TAX1BP1), recognize and capture ubiquitylated mitochondria through their ubiquitin-binding domains and simultaneously bind to LC3 proteins (also known as MAP1LC3) to recruit autophagosomal membranes for the formation of mitophagosomes, which ultimately fuse with lysosomes for degradation.

41Before damaged mitochondria are eliminated by mitophagy, they are transported by dynein towards the cell body/soma (Fig. 3), and defects in retrograde transport of senescent mitochondria lead to increased autophagy in axonal swellings. Enhancing retrograde transport by inducing binding of mitochondria to dynein increases neuronal mitophagy, whereas Dync1h1 mutations lead to impaired perinuclear clustering of damaged mitochondria in fibroblasts. Furthermore, the release of the MT anchor SNPH from stressed mitochondria enhances retrograde mitochondrial transport before activation of PRKN-mediated mitophagy, whereas overexpression of SNPH blocks mitophagy. Compartmental restriction of mitophagy to the soma of stressed neurons is supported by in vivo data from Drosophila neurons and Purkinje cells from mito-QC mice expressing a pH-sensitive fluorescent mitochondrial marker; however, clearance of damaged mitochondria has also been shown to occur locally in distal axons in response to acute stress.

42In addition to phosphorylating PRKN and ubiquitin, PINK1 also phosphorylates Miro1/2, thereby facilitating PRKN recruitment to Miro proteins. Ubiquitylation of Miro1/2 by PRKN then leads to its subsequent proteasomal degradation (Fig. 3). This results in the release of MYO19 and kinesin from the mitochondrial surface, leading to a reduced motility of damaged mitochondria. Hence, arresting mitochondrial transport is an early step in the clearance of dysfunctional mitochondria, which potentially isolates them before their elimination by mitophagy (Fig. 3).

43Mitochondrial damage not only halts the retrograde dynein-dependent movement of mitochondria, but also leads to distinct F-actin assembly events on mitochondria that are linked to recovery, segregation and the initiation of downstream mitophagy. A rapid, dynamic and transient ‘burst’ in mitochondria-localized F-actin occurs within minutes (∼5 min) of damage and requires ARP2/3 (Fig. 3), but is independent of INF2 and MYO6. This leads to circularization of mitochondria, which is their last chance to recover before being committed to clearance by mitophagy. Prolonged depolarization leads to mitochondrial fragmentation and initiates a second wave of MYO6-dependent actin polymerization after several hours, which initiates the formation of actin cages around damaged mitochondria during PRKN-mediated mitophagy. MYO6 forms a complex with PRKN and is selectively recruited to damaged, ubiquitylated mitochondria. Signalling downstream of the Rho family GTPase CDC42, as well as actin nucleators, such as the ARP2/3 complex, formins and N-WASP (also known as WASL), are all required for actin cage assembly. These MYO6-induced actin cages serve as a quality control mechanism at the onset of mitophagy by isolating dysfunctional mitochondria, thereby preventing their re-fusion with neighbouring populations and reintegration into the network (Fig. 3).

44In recent years, there has been growing evidence that the cytoskeletal machinery plays crucial roles during mitophagy; therefore, the actin filaments assembled by MYO6 on the mitochondrial surface may not only form actin cages, but also have important roles during mitophagy by facilitating phagophore formation and mitophagosome maturation. Indeed, since MYO6-dependent actin cages restrict the size of mitochondrial fragments, they may create small, ‘bite-size’ mitochondria that are more easily engulfed by the autophagy machinery. The synchronous formation of mitochondrial actin structures near autophagy initiation sites suggests a coordination between a reduction in mitochondrial dimensions and mitophagosome formation. This is driven by NMII and F-actin (Fig. 3), which have been implicated in disassembling mitochondrial aggregates into smaller pieces (1.5 h after damage) in preparation for efficient mitophagy into autophagosomes (with a typical diameter of 0.5–1.5 µm).

45Furthermore, the actin meshwork around mitochondria may persist after the initial damage and therefore could serve as a structural scaffold for the growing phagophore (Fig. 3) as well as supporting fusion of individual phagophores into a single mitophagosome. Mitochondrial actin may also drive autophagosome generation by disrupting mitochondrial aggregates to increase the available surface accessible to phagophores. Indeed, blocking the disassembly of mitochondrial aggregates and actin generation with an ARP2/3 inhibitor leads to prolonged clustering and inhibited mitophagosome formation.

46In this Review, we have highlighted a distinct subset of MT- and actin-based motor proteins that are linked to mitochondrial transport, fission, fusion and turnover. Even though we have discussed these key mitochondrial processes in separate sections, they are of course intimately linked and interdependent, and thus studying the precise roles of motors and their adaptors in isolation has been difficult due to their overlapping functions. One important avenue for future research is the regulation of motors and their adaptor proteins on mitochondria in response to changes in cell physiology and metabolism. The balance between fission and fusion, as well as transport and anchoring, are fine-tuned and require the coordination and handover of MT- and actin-based motors. Furthermore, a detailed understanding of motor protein regulation in maintaining mitochondrial homeostasis is crucial to elucidate how defects in these pathways can lead to human diseases.