The Meaning of Mitochondrial Movement to a Neuron’s Life

Abstract

Cells precisely regulate mitochondrial movement in order to balance energy needs and avoid cell death. Neurons are particularly susceptible to disturbance of mitochondrial motility and distribution due to their highly extended structures and specialized function. Regulation of mitochondrial motility plays a vital role in neuronal health and death. Here we review the current understanding of regulatory mechanisms that govern neuronal mitochondrial transport and probe their implication in health and disease.

1. Introduction

Mitochondria, best known as the powerhouses of cells, are cytoplasmic organelles of an endosymbiotic origin. Mitochondria have their own distinct haploid DNA which encodes essential enzymes for oxidative phosphorylation and mitochondrial tRNAs, but are assembled from proteins mostly encoded by nuclear DNA. Although ATP conversion by aerobic respiration is their major job, mitochondria also provide a reservoir for cytosolic calcium ions, and synthesize certain heme compounds [1] and steroids [2]. The uniqueness of these features places mitochondria in a significant position for regulating cellular proliferation and death [3].

Mitochondria move and undergo fission and fusion in almost all eukaryotic cells. This dynamic nature holds a particular urgency for neurons because of the problems that can arise during the long-range transport that is necessitated by the remarkable length and complexity of axons and the variable and specialized energetic demands of these highly polarized cells. The far-flung extremities of neurons are especially susceptible to disruption of the proper allocation of mitochondria [4]; both the presynapse and postsynapse have particularly high demands for energy and Ca⁺⁺ buffering. To sustain the periphery, newly assembled mitochondria from the cell body must be transported into neurites, and peripheral mitochondria whose proteins and DNA accumulate defects must be either repaired by fusion with fresh mitochondria or cleared from the cell [5]. Because the cell body is enriched with ribosomes, lysosomes, and other organelles, damaged mitochondria might be transported back to the cell body to be replenished or degraded. Therefore, in order to keep energy homeostasis and maintain essential activities, neurons need not only to precisely set up an adequate distribution of mitochondria, but also to efficiently sustain them in the periphery and clear them away when necessary. An exquisite regulation of mitochondrial movement is required to achieve these goals.

Misregulated mitochondrial motility in neurites is predicted to cause neuronal dysfunction and degeneration. Abnormalities in mitochondrial motility and distribution are implicated in a wide range of neurodegenerative and psychiatric disease models, such as Charcot-Marie-Tooth (CMT) [6], Amyotrophic Lateral Sclerosis (ALS) [7], Alzheimer’s [8–10], Huntington’s [11], Parkinson’s [12], and schizophrenia [13]. However in these diseases mitochondrial function of oxidative phosphorylation and generation of ATP may be disrupted by primary neurotoxic factors and cellular stresses. Malfunctioning mitochondria can affect mitochondrial motility [12], and cause neuronal cell death by generating reactive oxygen species and triggering apoptosis [14]. Therefore the part the misregulation of mitochondrial motility plays in disease progression merits further investigation. It is certain, however, that both misregulated mitochondrial motility and function can expedite the cell death process in neurons already vulnerable or stressed [15–16]. Here we review the current understanding of the cellular signals that regulate neuronal mitochondrial motility, as well as their significance and implication in neuronal health and disease.

2. Mitochondrial Transport Machinery

In neurites approximately 30–40% of total mitochondria are engaged in saltatory movement at any given time [17–20]. The densities of mitochondria at different synapses therefore change constantly. This reflects an acute need for mitochondria to buffer Ca⁺⁺ influx and distribute ATP among the microdomains of a neuron [21]. A mitochondrion can either move continually over a long distance at a relatively constant speed, or pause sporadically and start again with a different speed or direction. Fission-and-fusion events can also occur in moving mitochondria, though the relationship between movement and fission-and-fusion is unclear. Study in non-neuronal cells suggests that fusion events happen more frequently while mitochondria are moving [22]. The majority of mitochondrial movement is microtubule-based [17, 21, 23]. In an axon microtubules are uniformly arrayed, with all (+)-ends pointing to the axonal terminal and the (−)-ends to the cell body, whereas in a dendrite their polarities are typically more mixed [24]. Anterograde movement toward the (+)-ends of microtubules is mediated by kinesin motors, and retrograde toward (−)-ends by dynein motors. These motor proteins are generally shared with other cellular cargoes moving along microtubules. Mitochondria are anchored to motor proteins by idiosyncratic adaptor proteins; in this way their motility is precisely and specifically regulated by cellular signals (Table 1).

Table 1. Summary of current understanding of transport machineries for mitochondria.

See also Section 2.

Motor	Adaptor/associated proteins	Cytoskeleton	Illustration
KHC	milton, Miro	Microtubules	Figure 1A
	syntabulin/FEZ1/RANBP2/more	Microtubules	Figure 1B
KIF1Bα	KBP, more	Microtubules	Figure 1C
KLP6	KBP, more	Microtubules	Figure 1C
Dynein	Dynactin, more	Microtubules	Figure 1D
Myo19, more	Unknown	Actin	Figure 1E

Mitochondria are also transported along actin filaments, a process that happens more frequently in dendritic spines, growth cones, and synaptic boutons where the actin cytoskeleton is enriched. This actin-based movement is relatively short-ranged, likely mediated by myosin motors, and important for local and acute translocation and docking of mitochondria in response to action potentials, Ca⁺⁺ influx, or neurotrophic stimulation (Table 1). Actin-based transport can also coordinate, supplement, or even oppose microtubule-based transport [25].

2.1. Anterograde Microtubule Motors

In mammals, at least 14 families and 45 genes of kinesins have been identified to date, of which the Kinesin-1 family has been shown to be critical for mitochondrial transport [26–29]. Kinesin-1 is also known as the conventional kinesin heavy chain (KHC), or KIF5. Each KIF5 gene contains an N-terminal motor domain that binds to microtubules and hydrolyzes ATP, and a C-terminal domain that can interact with kinesin light chain (KLC) or cargo adaptors. Three KIF5 genes exist in mammals: two are neuronal-specific (KIF5A and KIF5C), while one is ubiquitously expressed (KIF5B). Mutations in mice KIF5B lead to perinuclear clustering of mitochondria, which indicates a disruption of kinesin-mediated anterograde transport of mitochondria [29]. In Drosophila, when the only orthologue of KIF5 is mutated axonal mitochondrial transport is impaired [30]. In both neuronal and non-neuronal cells KIF5 is recruited exclusively to mitochondria from cytosol by overexpression of mitochondrial adaptor proteins [19, 31], and overexpression of the motor domain of KIF5 in hippocampal neurons alters the regulation of mitochondrial motility by interfering with the integrity of the mitochondrial motor/adaptor complex [19].

In addition, two members of the Kinesin-3 family are associated with mitochondria; KIF1Bα and Kinesin-Like Protein 6 (KLP6). KIF1Bα is expressed ubiquitously and is particularly abundant in differentiated neuronal cells. The motor has been shown to colocalize with mitochondria and facilitate their (+)-end directed transport in vitro [32]. Knockdown of KLP6 disrupts anterograde axonal transport of mitochondria in neuronal cells and expression of a dominant negative KLP6 affects mitochondrial morphology in HeLa cells [33].

2.2. Anterograde Microtubule Adaptors

2.2.1

The KHC/milton/Miro complex is the best understood motor/adaptor complex for the regulation of mitochondrial transport. The current model suggests that Miro functions as a receptor with a transmembrane (TM) domain integrated into the outer mitochondrial membrane, and Miro binds to milton, which in turn binds to KHC. This complex allows mitochondria to associate with microtubules and plays key roles in regulating mitochondrial motility (Figure 1A). Milton came from a genetic screen in Drosophila for identification of mutants that disrupt synaptic transmission in photoreceptors, and was named after the 17th-century blind English poet John Milton [34]. Mitochondria are absent from axons deficient in milton but are present and functional in cell bodies. Milton is localized to mitochondria, and has a predicted coiled-coil domain interacting directly with KHC, Overexpression of milton in cultured mammalian cells recruits KHC to mitochondria [19, 31, 34]. In addition, the interaction between milton and KHC is KLC independent: KLC is not recruited to mitochondria by milton nor is it present in the KHC-milton complex [31]. Knockout of KLC in flies does not impair mitochondrial transport, suggesting that KLC is dispensable for their movement [31]. Milton has two homologues in mammals, TRAK1 (also known as milton-1, OIP106) and TRAK2 (milton-2, GRIF1), which are about 30% identical to Drosophila milton in their amino acid sequence. Both homologues also interact with KHC [35–36]. Knockdown of TRAK1 but not TRAK2 in cultured neurons impairs axonal mitochondrial movement, which can be rescued by expression of either TRAK1 or TRAK2 [37]. These reveal an important and conservative role of TRAK as a KHC adaptor in regulating mitochondrial motility. Differences do exist between Drosophila and mammalian milton homologues: whereas Drosophila mutants appear to be selectively defective in mitochondrial transport, there is evidence that the mammalian homologues may be associated with additional organelles [35, 38–40].

Figure 1. Schematic representations of mitochondrial transport machineries.

Protein complexes that move mitochondria along microtubules including A) KHC/milton/Miro, B) KHC/syntabulin (FEZ1, RANBP2), C) Kinesin-3/KBP, D) dynein/dynactin, and along actin E) Myo19, and docking machinery along microtubules F) syntaphilin. “Protein X” indicates additional unidentified proteins on the outer mitochondrial membrane.

KHC and milton need a third protein, Miro, to attach them to mitochondria. There is one Miro gene in Drosophila, and two (Miro1 and Miro2) in mammals. Each contains two GTPase motifs, a pair of EF-hands involved in Ca⁺⁺ binding, and a C-terminal TM domain that incorporates into the outer mitochondrial membrane [41–42]. Similarly to milton mutants, neurons deficient in Drosophila Miro lack axonal mitochondria [43]. Miro binds to milton directly, and Miro, milton and KHC together form a motor/adaptor complex on the mitochondrial surface [19, 31]. This complex plays a key role in conveying cellular signals to control mitochondrial movement, as will be discussed later.

2.2.2. Other KHC complexes

Besides Miro and milton, several other proteins have been found to connect KHC to mitochondria (Figure 1B). Syntabulin can directly attach KHC to the outer mitochondrial membrane, and anterograde axonal transport of mitochondria is disrupted when syntabulin is knocked down by RNAi in cultured hippocampal neurons [44]. Fasciculation and elongation protein-zeta 1 (FEZ1) can also bind to KHC and anterograde axonal transport of mitochondria is impaired when FEZ1 is disrupted in neurons [45–46]. RAN-binding protein 2 (RANBP2) has been shown to interact with KIF5B and KIF5C (but not KIF5A), and interrupting its function or its interaction with KHC affects mitochondrial distribution in both neuronal and non-neuronal cells [47]. It is likely that additional unidentified proteins are needed to attach these adaptor proteins and KHC to the outer mitochondrial membrane, and their specific roles in regulating neuronal mitochondrial motility need further investigation.

2.2.3. The Kinesin-3/KBP complex

Both KIF1Bα and KLP6 from the Kinesin-3 family interact with KIF1-Binding Protein (KBP). Together with other scaffolding proteins they may form a motor/adaptor complex to regulate mitochondrial motility [33] (Figure 1C). KBP is localized to mitochondria and downregulation of KBP protein levels leads to perinuclear aggregation of mitochondria [48]. KBP is also essential for normal axonal outgrowth through maintenance of axonal microtubule integrity during development [49].

2.3. Retrograde Microtubule Motors and Adaptors

The mechanisms and adaptor proteins for retrograde movement of mitochondria are less clear, although cytoplasmic dynein has been shown to be the motor [30]. In contrast to numerous kinesins, only one dynein exists. However, it is composed of multiple components and forms complicated structures that may give dynein functional diversity (Figure 1D). Dynein includes two heavy chains that function as motors and interact with microtubules, several intermediate chains, light intermediate chains and light chains that regulate its functions and attachments to cargoes. An auxiliary complex composed of 11 subunits, dynactin, binds to dynein and microtubules directly via its largest subunit, p150. Dynactin may facilitate the processivity of the dynein motor or its cargo binding [50]. In Drosophila, subunits of both dynein and dynactin have been found to be associated with mitochondria, and mutations in dynein heavy chain and p150 impair retrograde axonal transport of mitochondria [30]. It is unknown how dynein accomplishes specific cargo recognition, or how dynein manages to attach to the outer mitochondrial membrane. The relationship between dynein and kinesin seems to be complicated; mutations of KHC in flies disrupt retrograde mitochondrial movement in addition to its disruption of anterograde movement [30], and dynactin has been reported to coordinate both anterograde and retrograde movement [51]. Are kinesin and dynein present on the same mitochondrion? Do they coordinate with or oppose each other? How does each mitochondrion decide which direction to go? Do kinesin and dynein share the same mitochondrial adaptors? One indication comes from studies of Miro, the kinesin-associated mitochondrial receptor. Mutating the functional domains or key residues of Miro to make it insensitive to cellular signals that regulate mitochondrial motility (details in Section 3.1) inhibits the regulation in both anterograde and retrograde directions [12, 19]. And Drosophila Miro seems to promote mitochondrial motility in both directions [52]. These data suggest that Miro is also involved in the regulation of dynein-mediated mitochondrial transport, either directly by acting on dynein or indirectly through kinesin.

2.4. Actin Motors and Adaptors

Although mitochondria are able to travel along actin [53], little is known of actin motors or their mitochondrial adaptors. A recent study identified Myo19 as a new myosin motor for mitochondria [54]. Myo19 is exclusively localized to mitochondria and regulates mitochondrial morphology and actin-based mitochondrial transport. It is still unknown if Myo19 binds to the mitochondrial membrane directly or via its binding partners (Figure 1E). Since Myo19 is ubiquitously expressed, it is possible that Myo19 mediates actin-based mitochondrial movement in neurites, spines and boutons. Other candidates in neurons include Myosin II, III, V and VI [25, 55–56], although evidence of a direct association with mitochondria is still lacking. Knowledge of motor associated proteins that regulate actin-based mitochondrial movement is even scarcer. The WASP family verprolin homologues protein 1 (WAVE1) that regulates actin polymerization has been recently shown to control depolarization-induced mitochondrial movement into spines and filipodia and regulate spine morphogenesis [57].

2.5. Molecular Brakes

In neurons, the majority of mitochondria are immobile. At many subcellular regions that need a constant energy supply, such as places of protein synthesis or sites occupied with ion pumps, static mitochondria are always located nearby like gas stations. In addition, on many occasions mobile mitochondria can temporarily stop and start moving again. This suggests that mitochondria not only need motors to move themselves, but also additional regulation to dock and stabilize them at various subcellular locations. It is possible that once mitochondria are dissociated from microtubules or motors, they are able to remain static in the cytosol. It has also been suggested that actin-based regulation docks mitochondria away from microtubules to presynaptic boutons, dendritic spines, or growth cones where actin is enriched and mitochondria are more needed [58]. Recently, a microtubule-interacting protein, syntaphilin, has been identified as a mitochondrial anchor (Figure 1F): mobile mitochondria stop once they bind to syntaphilin, while still being attached to motors and microtubules [59]. A dynein light chain, LC8, has been shown to be involved in this immobilization by stabilizing syntaphilin-microtubule interactions [60]. Additional mechanisms are yet to be identified to explain the docking of mitochondria under various circumstances.

3. Regulation of Mitochondrial Movement

3.1. Regulations by the KHC/milton/Miro Complex

This primary motor/adaptor complex mediating anterograde axonal transport of mitochondria is employed by diverse cellular signals to allow distinct cellular demands to influence mitochondrial transport.

3.1.1. Splicing variants of milton

At least four different proteins (named milton A-D) are translated from Drosophila milton due to alternative splicing, and they only differ at their N-termini. All milton variants contain the coiled-coil KHC-interacting region and are expected to bind to and recruit KHC to mitochondria [31]. Yet one of the variants, milton-C, fails to do so [31]. It has not been documented how the variants of milton are expressed in neurons. These raise the interesting possibility that different variants of milton may have disparate roles, can respond to variable physiological stimuli, or may be expressed distinctively (Figure 2A). In Drosophila early developing oocytes acquire and multiply mitochondria and other organelles via directional microtubule-based transport in an accumulation known as the Balbiani body. A study showed that milton is essential for mitochondrial transport to form the Balbiani body, and suggests that milton-A facilitates (+)-end movement via kinesin, while milton-B facilitates (−)end movement via dynein [61]. In addition, since KHC always forms a dimer, milton is potentially a dimer as well. Four forms of milton monomer can make ten different combinations of the milton dimer if they are expressed in the same cell, and a combinatorial mechanism can also serve as a means to deliver different commands to mitochondrial transport machineries. Mammalian milton-2 (TRAK2) has three splicing variants and the differences are unclear [62]. Additional studies are needed to determine the roles and expression patterns of the splicing variants of milton. Much less is known about the variants of Miro, although they do exist (Ensembl). Like milton, alternative splicing of Miro may play a role to regulate the motor/adaptor complex.

Figure 2. Schematic representations of cellular signals that may regulate mitochondrial transport.

A) Splicing variants of milton; B) O-GlcNAcylation and OGT; C) Ca⁺⁺; D) mitochondrial damage and PINK1/Parkin; E) GTP/GDP; F) hypoxia and HUMMR; G) various other signals.

3.1.2. OGT and O-GlcNAcylation

In addition to KHC and Miro, milton binds to another protein, OGT (O-linked N-acetylglucosamine transferase) in an interaction conserved between flies and mammals [31, 63] (Figure 2B). OGT and the KHC/milton/Miro complex can form a quaternary complex [64]. OGT is an enzyme that catalyzes a posttranslational modification called O-GlcNAcylation: it is able to attach small sugar UDP-GlcNAc (Uridine diphosphate N-acetylglucosamine) to naked proteins at Serine and Threonine residues by covalent ligation, and another enzyme, O-GlcNAcase, reversibly removes it [65]. In contrast to phosphorylation which is fulfilled by the 400 Serine/Threonine kinases encoded by the human genome [66], O-GlcNAcylation is carried out by a single and highly conserved OGT gene [67]. O-GlcNAcylation regulates various cellular events including cytoskeletal dynamics, signal transduction, organelle trafficking, and gene transcription [65], by modification of distinct protein substrates localized to cytosol, nucleus, and cytoplasmic organelles including mitochondria. Both mammlian and Drosophila milton have been shown to be substrates of OGT and can be modified by UDP-GlcNAc [31, 63]. Therefore O-GlcNAcylation likely controls certain cellular events such as mitochondrial movement through milton by modifying it.

What upstream cellular signal does O-GlcNAcylation represent to regulate milton and mitochondrial movement? The catalytic activity of OGT may be activated or inactivated by upstream proteins or signals. The availability of small sugar UDP-GlcNAc may drive the equilibrium of the reaction as well, to control the amount of O-GlcNAcylation on milton or the amount of milton O-GlcNAcylated. Glucose is the source of UDP-GlcNAc, and 2–5% of total glucose is converted to UDP-GlcNAc via the hexamine biosynthetic pathway (HBP). If glucose influences O-GlcNAcylation of milton by UDP-GlcNAc and OGT, mitochondrial movement may be altered by subcellular metabolic conditions.

Neurons depend on glucose and oxidative phosphorylation almost exclusively for energy supply, which makes them unique in maintaining their metabolic and energetic homeostasis [68]. Regulation of mitochondrial motility by OGT and the KHC/milton/Miro complex along the extensions of axons may be needed to support a healthy metabolic state of a neuron. Further investigation of how OGT and O-GlcNAcylation regulate mitochondrial motility and their relationships to glucose and the KHC/milton/Miro complex, in both neuronal and non-neuronal cells, may unravel the remaining puzzles.

3.1.3. Cytosolic Ca⁺⁺

Another important cellular signal that regulates mitochondrial motility is cytosolic Ca⁺⁺. Resting levels of cytosolic Ca⁺⁺ permit mitochondrial movement, whereas elevation of the concentration by calcium ionophores completely inhibits mitochondrial motility in both neuronal and non-neuronal cells [19, 69–70]. Why is Ca⁺⁺ used to stop mitochondria? It has been shown that concentrations of cytosolic Ca⁺⁺ are not uniform across subcellular regions and areas of lower ATP concentrations may have higher Ca⁺⁺ concentrations [71]. In neurons, this heterogeneity is more evident: in synapses and dendritic spines, the high energy demands necessary to maintain action potential firing and ion gradients across membranes are accompanied with high Ca⁺⁺ influx. Therefore, mitochondria are arrested at places where Ca⁺⁺ concentration is higher and ATP concentration is lower, probably by buffering extra Ca⁺⁺ and providing more ATP to regulate both Ca⁺⁺ and energy homeostasis. Ca⁺⁺ halts mitochondrial movement by binding to the EF-hands of Miro in the mitochondrial motor/adaptor complex, therefore inducing a direct interaction between Miro and the motor domain of KHC, and interfering with KHC motor/microtubule binding. Consequently, the whole motor/adaptor complex and mitochondria are dissociated from microtubules and mitochondria stop moving (Figure 2C). This regulation is transient and reversible, likely representing a temporary and local need for mitochondria [19]. An alternative model suggests that Ca⁺⁺ disrupts the integrity of the KHC/milton/Miro complex and dissociates mitochondria from KHC motors and microtubules [70], though this model has not been confirmed by other studies [19, 64, 72]. When the key glutamate residues within the EF-hands of Miro are mutated to lysines to prevent Ca⁺⁺-binding, Ca⁺⁺-dependent mitochondrial arrest is abolished in both anterograde and retrograde directions [19, 69, 70]. Since KHC regulates the transport of other cargoes, this mechanism via the mitochondrial adaptor Miro explains the mitochondrial specificity for Ca⁺⁺ regulation; Ca⁺⁺ stops mitochondria, not other cargoes transported along microtubules. Physiologically, the EF-hands of Miro mediate the glutamate-dependent mitochondrial arrest, which protects neurons during excitotoxic stress [19].

3.1.4. The PINK1/Parkin cascade

Another signaling pathway that regulates mitochondrial motility via this common KHC/milton/Miro complex is the PINK1/Parkin cascade. PINK1 (The PTEN Induced Kinase 1) and Parkin are two Parkinson’s disease proteins. Mutations in their encoding genes cause autosomal recessive early onset Parkinson’s disease in humans [73–74]. Parkinson’s disease is the second most common neurodegenerative disease with a selective loss of midbrain dopaminergic neurons. PINK1 encodes a Serine/Threonine kinase with an N-terminal mitochondrial targeting sequence (MTS), and can be localized to either the outer or inner mitochondrial membrane depending on mitochondrial membrane potential [75]. Parkin encodes an E3 ubiquitin ligase and is likely involved in the ubiquitin-proteasomal-degradation pathway. With PINK1 functioning upstream of Parkin in a common pathway, both PINK1 and Parkin regulate mitochondrial morphology and function [76–80], and are involved in a mitochondrial quality control process known as mitophagy [81–85], which clears damaged mitochondria via autophagic engulfment. The PINK1/Parkin pathway can be activated by mitochondrial depolarization and damage [10, 81–85].

Both PINK1 and Parkin can physically interact with the KHC/milton/Miro complex and their interaction with Miro is enhanced by depolarization of mitochondrial membrane potential [12, 86]. This is consistent with findings that PINK1 is stabilized on the outer mitochondrial membrane and Parkin is recruited to mitochondria from cytosol when membrane potential is compromised [75]. The direct binding of the mitochondrial transport complex to PINK1 led Wang et al to investigate the roles of the PINK1/Parkin pathway in the regulation of mitochondrial transport. The results [12] showed that excessive amount of PINK1 or Parkin, or depolarization of mitochondrial membrane potential, arrests mitochondrial transport in both anterograde and retrograde directions through proteasomal degradation of Miro in rodent and Drosophila axons. The removal of Miro from mitochondria releases KHC from the mitochondrial surface. PINK1 has been demonstrated to phosphorylate Miro and this phosphorylation triggers its Parkin-dependent degradation (Figure 2D). A recent study confirmed that Miro is the target of the PINK1/Parkin cascade and can be ubiquitinated by Parkin, and suggests that the degradation of Miro is a key step in generating PINK1 mutant phenotypes in flies and facilitating mitophagy in HeLa cells [87]. It is likely that arresting mitochondrial motility in this manner is an early step in the quarantining of damaged mitochondria for subsequent mitophagy. In individuals lacking either functional PINK1 or Parkin, a failure to isolate, stop and remove the damaged mitochondria will likely contribute to neuronal cell death. PINK1 and Parkin represent an evolutionarily conserved pathway to regulate mitochondrial motility, by acting on a common target Miro. Unlike Ca⁺⁺, this regulation initiates an irreversible and permanent process to clear damaged mitochondria.

More questions remain to understand this cascade in mitophagy. The absence or reduction of PINK1 in flies does not affect the ability of Parkin overexpression to impair mitochondrial motility [12] and morphology [76], indicating that PINK1 phosphorylation on its substrates such as Miro is likely to stimulate Parkin function, but probably not absolutely necessary. It is less clear if this cascade has crosstalk with other pathways that are necessitated by mitophagy. For example, massive mitochondrial depolarization may activate more than one kinase besides PINK1 to enhance Parkin activities. Wang et al have noticed that overexpression of exogenous PINK1 and Parkin together had a bigger effect on Miro degradation than overexpression of either individually [12]. A recent study has shown that Parkin does not form a stable complex with PINK1, and recruitment of Parkin alone to mitochondria without activation of PINK1 is not able to ubiquitinate mitochondria and induce mitophagy [88]. These findings exclude the possibilities that the primary role of PINK1 is to recruit Parkin to mitochondria and that the mitochondrial presence of Parkin alone is sufficient to trigger mitophagy. The relationship between PINK1 and Parkin may be perplexing, and more proteins may be involved in between. This may also explain the discrepancies in the extent of degradation of exogenous Miro by overexpressed PINK1, Parkin, or treatment with CCCP, a mitochondrial depolarizer. The ratios of amounts of overexpressed Miro, PINK1, or Parkin are crucial to determine the results in different cell lines. When Miro synthesis outpaces its degradation caused by the overexpression of either enzyme, Miro protein remains [12]. Blocking the major phosphorylation site Serine156 of Miro resisted degradation best when an appropriate amount of PINK1 or Parkin was overexpressed individually [12], but inadequately when extra amounts of PINK1 and Parkin were overexpressed together, with or without CCCP treatment [87]; additional sites on Miro for PINK1 (such as Threonine298, 299) [12] or for other kinases may be phosphorylated to stimulate Parkin function under the latter condition.

It is unclear what the fate of these arrested and damaged mitochondria is in the axons. Since lysososomes are found more abundant in cell bodies [89], damaged mitochondria may fuse with autophagosomes in situ and be subsequently transported back to cell bodies for lysosomal degradation, via retrograde axonal transport of autophagosomes independent of the KHC/milton/Miro complex which is specific to mitochondria. However, since lysosomes do exist in the axons [90], it is possible that local lysosomal fusion could occur. It is also possible that there is a preference of the cell bodies over the axons as the site of activation of this cascade in vivo. Maybe under physiological conditions when there are no global mitochondrial damages, it is more economical to activate the PINK1/Parkin cascade in the cell bodies to prevent damaged mitochondria from moving out into the axons.

Studies of this cascade suggest that mitochondrial importation of PINK1 could happen at the distal part of axons, and support the view of a constant and local mitochondrial biogenesis. PINK1 is continually imported into healthy mitochondria and subsequently degraded, and it accumulates on the outer mitochondrial membrane of damaged mitochondria because its importation is blocked by compromised membrane potential [75]. How could PINK1 reach mitochondria in axonal processes to either be imported or be stabilized on the outer mitochondrial membrane? Is PINK1 protein or mRNA transported there from the cell bodies along axons? If so, does this apply to other mitochondrial proteins? Are anterograde axonal transport of the necessary components and local protein translation the way by which axonal mitochondria replenish their proteins and maintain biogenesis within the terminals? Obviously the remote part of axons is more susceptible to problems that arise during this long-range transport, and a shortage of PINK1 to the distal mitochondria due to the disrupted transport may contribute to the mitochondrial malfunction and neurodegeneration in neurons with particularly long axons.

3.1.5. Other possible signals through the complex

Miro was identified as an atypical Rho GTPase and has two functional GTPase domains [41–42]. While many relatives of Miro in the Rho GTPase subfamily such as RhoA, Rac1, and cdc42 regulate cell morphology and motility [41], and other large GTPase proteins localized to mitochondrial membranes including MFN (mitofusin), drp1 (dynamin like protein 1), and OPA1 (optic atrophy 1) control membrane fission and fusion, the roles of Miro’s GTPase domains are still unclear (Figure 2E). Evidence suggests that they might be needed to maintain mitochondrial morphology [42]. It is possible that they control other functions of Miro or the KHC/milton/Miro complex, such as Ca⁺⁺-sensing or OGT-binding, in order to affect mitochondrial motility. Further exploration of how Miro’s GTPase domains function is warranted.

Hypoxia up-regulated mitochondrial movement regulator (HUMMR) also interacts with the KHC/milton/Miro complex and regulates mitochondrial transport and distribution [91]. HUMMR can be expressed in neurons and induced by hypoxia-inducible factor 1α (HIF-1α) (Figure 2F). Knockdown of HUMMR in hypoxic neurons inhibits anterograde axonal transport of mitochondria and promotes retrograde movement. Hypoxia induces a mild increase in the density of axonal mitochondria, which is abolished by reduced HUMMR expression [91]. This may represent a physiological response of mitochondrial redistribution to acute stress such as hypoxia, which is an important factor during ischemic stroke. Induction of HUMMR can enable more mitochondria to enter the distal part of neurites and synapses, and high Ca²⁺ influx resulting from glutamate excitotoxicity can trap mitochondria within the terminals. This response may have neuronal protective effect during ischemia by gathering extra mitochondria to buffer Ca⁺⁺, but could also aggravate cell death by releasing more reactive oxygen species.

3.2. Regulation by Other Signals

3.2.1

ADP/ATP has been reported to regulate mitochondrial motility in cultured neurons [92]. This study suggests that in places with higher energy consumption, such as synapses, intracellular ADP concentration increases and mitochondria are attracted to sustain ATP supply. However, how mitochondrial transport machinery senses changes in concentrations remains elusive. Together with the potential involvement of OGT discussed in Section 3.1.2, these regulatory factors may permit subcellular metabolic states to allocate mitochondria by influencing their motility.

3.2.2

Neurotransmitters including serotonin (5-HT) and dopamine (DA) have been shown to alter mitochondrial movement in hippocampal neurons. Mitochondrial motility is boosted by activation of the 5-HT receptor via the AKT-GSK3β (Akt-Glycogen Synthase Kinase 3β) pathway [17]; while mitochondrial motility is inhibited by DA or activation of the D2 DA receptor via the same AKT-GSK3β pathway [93]. This indicates the involvement of neuromodulators in the control of neuronal growth and energy allocation by regulating mitochondrial transport. The AKT-GSK3β pathway likely achieves this through the regulation of kinesin-cargo interactions [94–95].

3.2.3

Nerve Growth Factor (NGF) treatment of cultured neurons causes accumulation of mitochondria close to where NGF is applied [96], suggesting that NGF also regulates mitochondrial motility. This effect appears to be dependent on actin and specific to mitochondria [58]. The exact mechanism of this phenomenon is still unknown, although involvement of the PI3 (phosphoinositide-3) kinase pathway has been suggested [96]. In addition to the signals summarized here, other unidentified small molecules and cellular messengers may be involved in the regulation of mitochondrial movement (Figure 2G).

4. Misregulation of Mitochondrial Motility in Diseases

Given the particular importance to a neuron of a precise control of mitochondrial motility and distribution to maintain energy and Ca⁺⁺ homeostasis, a subtle perturbation could be detrimental to neuronal health and survival. Misregulated mitochondrial motility and distribution have been found in various neurological disease models and patients, and emerging evidence suggests their participation in disease development (Table 2). Because mitochondria play a vital role in cellular homeostasis, questions remain as to whether abnormality in mitochondrial motility is a consequence of mitochondrial malfunction or a key step in neuronal pathogenesis, and whether correction of mitochondrial motility is worth pursuing as a new strategy for better treatments.

Table 2. Summary of neurological disorders with impaired mitochondrial motility in their disease models.

See also Section 4.

Disease	Affected Region	Proteins Bind to Mitochondrial Transport Machinery	Potential Mechanisms	Reference
Parkinson’s	Dopaminergic neurons in substantia nigra	PINK1, Parkin	Target Miro	12, 19, 86
Alzheimer’s	Cerebral cortex	Unknown	Unknown; possibly microtubule abnormality, protein aggregation, Ca++	8–10, 19, 105–108
ALS	Motor neurons in spinal cord and cortex	Unknown	Unknown; possibly mitochondrial dysfunction, Ca++	7, 19, 72, 110, 111
Huntington’s	Striatum and cortex	HTT?	Unknown; possibly protein aggregation, milton, Ca++	11, 19, 34, 109
CMT Type2A	Peripheral sensory and motor neurons	MFN2?	May coordinate with KHC/milton/Miro	6, 112
Schizophrenia	Not confirmed	Unknown	Unknown	13

4.1. Linking Miro to Diseases

4.1.1

Parkinson’s disease is a well-characterized mitochondrial disease, although why dopaminergic neurons in substantia nigra die first remains elusive considering the fact that every disease-causing gene is ubiquitously expressed. A “mitochondrial damage” theory has recently come into shape given the emerging evidence that PINK1 and Parkin, mutations of which cause juvenile hereditary Parkinson’s disease, are involved in a mitochondrial-specific autophagy process, namely mitophagy, which clears damaged mitochondria [81–85]. The consequence of an impaired mitophagy process in mutant cells is an accumulation of oxidative stress that can lead to neurodegeneration. Dopaminergic neurons may be more susceptible to inefficient mitophagy probably due to their higher level of cellular stress relative to other cells [97]. However, the exact roles of these two enzymes in mitophagy are still unclear, as discussed in section 3.1.4. How is the PINK1/Parkin pathway activated by mitochondrial damage? How do PINK1 and Parkin recognize defective mitochondria in order to initiate mitophagy? How and where is subsequent autophagosome formation signaled, initiated and completed? Nevertheless, the hypothesis that mutant PINK1 or Parkin cells may be impaired in their ability to clear damaged mitochondria with resultant injury to neurons is so intriguing that it makes researchers ponder whether this also accounts for other types of Parkinson’s disease and other neurodegenerative diseases.

Mitochondrial damage and subsequent activation of the PINK1/Parkin pathway inhibits mitochondrial motility by directly targeting Miro in the motor/adaptor complex at a much earlier stage than the initiation of mitophagy (also see section 3.1.4) [12]. This suggests that tagging Miro to stop damaged mitochondria might be how cells first spot and segregate damaged from healthy mitochondria, to prevent further cell damage from mitochondrial movement and fusion, and to quarantine them before mitophagy. It could also serve as the means by which damaged mitochondria are labeled and recognized by autophagosomes. Failure to do so may interfere with mitochondrial clearance and eventually cause neurodegeneration. It is apparent that the process starting when mitochondria are damaged to final clearance is very complicated, and any error during this process could cause a delay or halt of mitophagy and an accumulation of oxidative stress. An interesting question remains: if we help correct mitochondrial motility, will it restore mitophagy and postpone neuronal cell death to some extent? Further experiments aimed at correcting motility abnormality, for example, using mutant forms of Miro, will be the first step in helping answer this question and other questions designed to improve treatment and change the course of the disease.

The PINK1/Parkin pathway may have additional substrates through which it regulates mitochondrial physiology and affects mitophagy. MFN (mitofusin), an outer mitochondrial membrane protein which controls mitochondrial fusion, has been shown to be substrate of Parkin, and ubiquitinated and degraded upon mitochondrial depolarization [98–103]. VDAC (voltage-dependent anion channel 1), an outer membrane component of the mitochondrial permeability transition pore (mPTP) which is involved in apoptosis, is potential target of Parkin as well [81]. MFN and VDAC have been demonstrated to be important to facilitate Parkin-dependent mitophagy [81, 98–103]. Misregulation of MFN and VDAC in mutant PINK1 or Parkin cells could interfere with mitophagy, activate apoptosis, and contribute to neurodegeneration. The possibility of correcting their malfunctions to hinder neuronal cell death is worth investigating.

4.1.2. Ca⁺⁺ mediated neurodegeneration

Ca⁺⁺ is an important signal that can cause neuronal death when misregulated. Voltage and ligand-gated Ca⁺⁺ channels populate both pre- and post-synaptic membranes and change intracellular Ca⁺⁺ concentrations instantaneously. Ca⁺⁺ is taken up by mitochondria via the Ca⁺⁺ uniporter and released by the Na⁺/H⁺- Ca⁺⁺ exchanger, and is able to travel between mitochondria and ER through their contacts. High Ca⁺⁺ influx is a major side effect of glutamate excitotoxicity implicated in stroke, acute brain and spinal cord injuries, and many neurodegenerative diseases such as Parkinson’s, Alzheimer’s, ALS and multiple sclerosis. It damages cells by activating numerous enzymes that destroy cellular structures, such as the cytoskeleton, membrane and DNA. High Ca⁺⁺ can also facilitate the opening of the mPTP, which makes the inner mitochondrial membrane permeable to small molecules less than 1.5kDa and consequently breaks down the outer mitochondrial membrane by increased osmotic pressure. Release of Cytochrome C into the cytosol actives apoptosis, and spread of reactive oxygen species and mitochondrial Ca⁺⁺ further damages the cell. As a result, the mitochondrial membrane potential is compromised, and the electron transport chain and ATP production are disabled.

Ca⁺⁺ arrests mitochondria via the EF-hands of Miro, which is critical for neurons to resist glutamate excitotoxicity [19], probably by accumulating more mitochondria to buffer extra Ca⁺⁺ and supply more ATP as a means of cell survival. Although in many diseases glutamate excitotoxicity or elevation of Ca⁺⁺ is not the primary cause of pathology, it may well be the primary factor that leads to cell death. Will manipulation of mitochondrial motility in order to best allocate mitochondria to subcellular regions with highest Ca⁺⁺ influx help preserve the longevity of neurons under these stressful conditions? Combined with other strategies to prevent the detrimental damages caused by high Ca⁺⁺, such as usage of mPTP inhibitors [104], this is an exciting and innovative area to explore to better understand disease progression and improve treatment strategies.

4.2. Other Evidence

4.2.1

Alzheimer’s disease is the most common neurodegenerative disease characterized by loss of neurons in cerebral cortex. Both the “amyloid” hypothesis of amyloid β deposits and the “tau” hypothesis of neurofibrillary tangles exist to explain the causes of neuronal death. Reduced mitochondrial motility has been observed in various disease models of Alzheimer’s [8–10, 105–108], and the precise mechanisms remain unclear. Amyloid β peptides have been shown to inhibit mitochondrial transport in cultured neurons through PKA (protein kinase A) and GSK3β pathways [9]. Hyperphosphorylation or overexpression of Tau also inhibits mitochondrial movement, probably by disrupting axonal cytoskeleton including both microtubules and actin [10, 108].

4.2.2

Huntington’s disease is another neurodegenerative disease affecting striatal and cortical neurons. It is caused by mutant expansions of a polyglutamine region in the huntingtin (HTT) gene. Mutant, but not wild type HTT decreases bi-directional mitochondrial motility in both rodent neuronal cultures and transgenic animals [11, 109]. Long mutant HTT protein stretches could certainly jam microtubule tracks by interacting with cytoskeletal components and transport motors [109], but a specific binding between HTT and milton is also possible: milton is a distant homologue of human HAP1 (huntingtin-associated protein 1), which binds to HTT and dynactin [34].

4.2.3

ALS (Lou Gehrig’s disease) is caused by degeneration of upper and lower motor neurons in the spinal cord and cortex. Bi-directional axonal movement of mitochondria is decreased in various ALS disease models [7, 72, 110, 111]. Defects in kinesin or dynein function have been suggested to cause mitochondrial arrest in certain Cu/Zn superoxide dismutase-1 (SOD1) -related ALS models [7, 110, 111]. A recent study proposed another interesting hypothesis: in ALS-associated mutant VAPBP56S cells, intracellular Ca⁺⁺ is elevated, which results in halting mitochondrial transport via Miro [72].

4.2.4

CMT Type 2A is a dominant inherited peripheral neurodegenerative disease caused primarily by mutations in MFN2, which controls mitochondrial outer membrane fusion. As discussed above, both MFN1 and MFN2 are targets of Parkin and important for mitophagy [98–103]. Interestingly, they also interact with mammalian Miro and milton [112]. Disease-causing mutant MFN2 or knockdown of MFN2 in neuronal cultures significantly abolishes bi-directional mitochondrial motility [6, 112], similarly to knockdown of Miro2 [112]. This poses an intriguing model in which MFN and the KHC/milton/Miro complex may coordinate with each other to regulate mitochondrial motility, as well as mitochondrial fusion and mitophagy.

4.2.5

Schizophrenia is a mental illness characterized by a person’s breakdown in perception and emotion. Little is known about the molecular, anatomical, and physiological mechanisms underlying its pathogenesis. Recently, mitochondrial dysfunction and misdistribution have been found related to the disorder [113]. Interestingly, when a schizophrenia-susceptible gene product DISC1 (disrupted in schizophrenia 1) is either knocked down or overexpressed in rodent cultured neurons, mitochondrial motility is misregulated [13]. DISC1 is distributed to various subcellular locations including mitochondria [114], and is involved in the regulation of numerous cellular events such as neuronal development, proliferation and migration [115]. Could abnormal mitochondrial motility caused by mutant DISC1 or other proteins contribute to or exacerbate disturbances of neuronal growth, function, and structures that might lead to psychiatric disorders such as schizophrenia? A detailed investigation of mitochondrial movement in various psychiatric disease models is an excellent place to start testing this hypothesis.

5. Closing Remarks

Regulation of neuronal mitochondrial motility is an emerging frontier, and the more we understand the underlying mechanisms the more we appreciate how significant they are to neuronal function and survival. The delicacy and complexity of these mechanisms reflects heterogeneous and specialized needs for mitochondria on an accurate temporal and spatial scale among subcellular regions of a neuron. Mitochondrial dynamics allow neurons to respond precisely to changes in their activity and environment. Local signals are crucial to maintaining energy homeostasis and thereby to supporting neuronal function. A better understanding of these mechanisms will help us decipher the mystery of mitochondria, and shed light on the right course of action to sustain the healthy life of a neuron, or reverse or delay the progression of a dying neuron.

Highlights.

• Transport machineries that move mitochondria along microtubules and actin cytoskeleton

• Cellular signaling pathways that regulate mitochondrial movement in neurons

• Implication of the regulation of mitochondrial motility in health and disease and future directions

Abbreviations

Akt-GSK3β

Akt-Glycogen Synthase Kinase 3β

ALS

Amyotrophic Lateral Sclerosis

CCCP

Carbonyl cyanide m-chlorophenyl hydrazine

CMT

Charcot-Marie-Tooth

DA

Dopamine

DISC1

Disrupted In Schizophrenia 1

FEZ1

Fasciculation and Elongation Protein-Zeta 1

GRIF1

Gamma-Amino Butyric Acid (GABA) Receptor Interacting Factor 1

HAP1

Huntingtin-Associated Protein 1

HBP

Hexamine Biosynthetic Pathway

HIF-1α

Hypoxia-Inducible Factor 1α

HTT

Huntingtin

HUMMR

Hypoxia Up-regulated Mitochondrial Movement Regulator

KBP

KIF1-Binding Protein

KHC

Kinesin Heavy Chain

KIF

Kinesin Family Member

KLC

Kinesin Light Chain

KLP6

Kinesin-like Protein 6

LC8

Dynein Light Chain 8

MFN

Mitofusin

Miro

Mitochondrial Rho GTPase

MPTP

1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine

MTS

Mitochondrial Targeting Sequence

OGT

O-linked N-acetylglucosamine Transferase

OIP106

O-linked N-acetylglucosamine Transferase (OGT) Interacting Protein 106

PINK1

Phosphate Tensin Homologue (PTEN) Induced Kinase 1

RANBP2

Ras-Related Nuclear Protein (RAN) Binding Protein 2

TM

Transmembrane

TRAK

Trafficking Kinesin Binding Protein 1, 2

UDP-GlcNAc

Uridine Diphosphate N-acetylglucosamine

WAVE1

Wiskott-Aldrich syndrome protein (WASP) Verprolin Homologe 1

5-HT

5-hydroxytryptamine