The etiology of Parkinson disease (PD) has been assumed to be a complex combination of environmental factors, intrinsic cellular metabolic properties, and susceptible genetic alleles. The primary obstacles to the development of a neuroprotective therapy in PD include uncertainties with regard to the precise cause(s) of neuronal dysfunction and what to target. The discoveries of Mendelian genes associated with inherited forms of PD in the last 10 years have revolutionized the understanding of the cellular pathways leading to neuronal dysfunction. Common themes of the pathogenesis of PD are beginning to emerge with mitochondrial dysfunction at the center stage. In this review, we summarize our knowledge of the pathogenesis of PD, revisit some aspects of mitochondrial biology, and discuss the insights from the study of Pink1, a familial PD-associated gene. We propose that mitochondrial morphogenesis and distribution might be a novel and potential common paradigm for PD and other neurodegenerative disease research and that modulation of such mitochondrial processes may prove to be a valuable therapeutic avenue for PD.
Parkinson disease (PD) is the most common neurodegenerative movement disorder. It afflicts more than 1 million people in North America, affecting approximately 1% of the population older than 65 years, and the incidence increases to 4% to 5% in people 85 years and older (1, 2). The primary symptoms of the disease include muscular rigidity, resting tremor, and bradykinesia. Secondary symptoms may include gait disturbance with small steps, a tendency to experience freezing, dysarthria, hypomimia, and salivation. Nonmotor symptoms such as dementia and autonomic dysfunction also develop in many cases, particularly in the late stages. The symptoms among patients are highly variable, but the disorder is generally chronic and progressive.
The primary manifestations of PD result from decreased stimulation of the motor cortex by the basal ganglia, which is the result of deficiency of the neurotransmitter dopamine produced by dopaminergic (DA) neurons. The pathological hallmarks of the disease are the predominant degeneration of DA neurons in the substantia nigra (SN) pars compacta and the neuronal loss in the ventral tegmental area, locus caeruleus, and other brain regions. Other typical pathological features of PD are characteristic intracellular proteinaceous inclusions called Lewy bodies and Lewy neurites that are found in the brainstem and cortical areas. The clinical manifestations of PD emerge by a threshold effect, whereby at the time they first appear, approximately 60% to 70% of the dopamine fibers in the caudate putamen and at least 50% of the DA neurons in the SN are already lost (3). The progression of PD may take 20 years or more, but there is no way to predict what course the disease will take in an individual patient.
CAUSES OF PD: NURTURE VERSUS NATURE
In most cases, the cause of PD is unknown or nonspecific, and the patients are considered to have idiopathic PD. The question of “nature versus nurture” in the etiology of parkinsonism has been a matter of considerable debate. A commonly accepted opinion holds that the etiology of PD is probably a complex combination of environmental factors, genetically determined vulnerability, and an aging brain (Fig. 1).
Environmental Risk Factors
The notorious synthetic narcotic drug 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP) was brought to light as a specific cause of parkinsonism in the 1980s when young drug addicts developed irreversible and severe PD symptoms after self-administration of what they hoped to be synthetic heroin (4). The MPTP is converted into the active neurotoxic metabolite 1-methyl-4-phenylpyridinium ion (MPP+). MPP+ accumulated via the dopamine transporter into DA neurons, resulting in specific neuronal damage. This disease outbreak made the development of the first animal model of PD possible. Other environmental risk factors most strongly suspected at present are certain agricultural chemicals and transition-series metals such as manganese or iron, especially those that generate reactive oxygen species (ROS) and/or bind to neuromelanin, a metabolite enriched in the DA neurons. The agricultural chemicals include pesticides (e.g. rotenone), herbicides (e.g. paraquat), and fungicides (e.g. maneb). Individuals who had been exposed to pesticides had a 70% higher incidence of PD than individuals not exposed (5). Like MPP+, rotenone and paraquat are inhibitors of complex I (NADH ubiquinone oxidoreductase) of the electron transport chain in mitochondria.
The mitochondrial respiratory chain is exquisitely evolved to transfer electrons via a series of electron receptors until the final transfer to oxygen that leads to production of water. The transfer of electrons along the respiratory chain provides the energy to pump protons from the matrix into the intermembrane space, generating the electrochemical gradient required to drive adenosine triphosphate (ATP) synthesis. These biochemical reactions have an inherent danger, however, because electron leakage can lead to the production of ROS. Therefore, blocking complex I would not only lead to dysfunction of mitochondria but also to an elevated level of ROS. Mitochondrial dysfunction would result in depletion of ATP, increased intracellular calcium concentration, and excitotoxicity. More importantly, excessive ROS is the major source of oxidative stress because ROS would produce lesions in mitochondrial DNA (mtDNA), lipids, and proteins by oxidation. Mutated mtDNA would produce aberrant proteins that affect the efficiency of electron transport chain; thus, a so-called vicious cycle would be perpetuated that eventually leads to cell death (6).
During the last decade, the identification of single genes linked to heritable forms of parkinsonism has revolutionized the previously held view of a largely nongenetic etiology. Linkage mapping has revealed that 13 loci, designated as PARK1 to PARK13, are associated with rare inherited forms of PD (6). At 7 of the PARK loci, genetic mutations have been found in affected family members, and most interestingly, 4 of these loci have also been implicated in sporadic PD. Therefore, it is anticipated that these susceptibility genes are also candidate genes for idiopathic forms of the disease. So far, mutations in α-synuclein and leucine-rich repeat kinase 2 (LRRK2) genes have been found to cause dominant forms of familial PD, whereas mutations in Parkin, DJ-1, Pink1, and Omi/Htr2 are associated with autosomal recessive forms of PD (7).
PARK1, PARK4, and α-SYNUCLEIN
A major leap in our understanding of the etiology of PD came when mutations were identified in α-synuclein. The gene dosage of α-synuclein also seems to be critical because genomic duplication and triplication (which is found at the PARK4 locus) have been linked to disease severity, symptoms, and age at onset (7). α-Synuclein forms the major filamentous component of Lewy bodies and Lewy neurites in both familial and sporadic disease, suggesting that abnormalities of α-synuclein might be crucial for the pathogenesis of both rare and common forms of PD. Wild-type or mutant α-synuclein can form protofibrillar (oligomeric) or fibrillar protein assemblies. It is not presently known, however, which species is the toxic culprit and how toxicity might occur. Mutant α-synuclein variants seem to result in protein misfolding and promote protofibril or fibril formation, which is consistent with the gain-of-function inheritance pattern of α-synuclein. Formation of protofibrils is enhanced and stabilized by dopamine quinones derived from the oxidation of dopamine, and this could account for the selective toxicity of α-synuclein in the SN (8). On the other hand, mutant α-synuclein sensitizes neurons to oxidative stress and damage by dopamine and ROS (9). The physiological function of α-synuclein has not been fully determined.
Despite the lack of a direct connection to mitochondria, there is evidence that mutant α-synuclein may cause mitochondrial dysfunction. Transgenic mice that overexpress α-synuclein are more sensitive to MPTP compared with controls and exhibit decreased complex IV activity and enlarged and abnormal mitochondria (10). Conversely, α-synuclein-deficient mice are more resistant to the respiratory chain inhibitors, indicating that α-synuclein may play a role in modulating mitochondrial toxicity and oxidative stress (11).
After the discovery of α-synuclein, the discovery of the Parkin gene in 1998 corroborated the concept that mutations in a single gene can cause parkinsonism. Homozygous mutations in the Parkin gene were found at the PARK2 locus in families with an autosomal recessive early-onset form of PD. Nearly 50% of all familial cases with early onset (<45 years), as well as a significant proportion of apparently sporadic cases, are caused by homozygous or compound heterozygous Parkin mutations. Heterozygous promoter and coding polymorphisms have been associated with susceptibility to late-onset PD. Interestingly, some of parkin-associated PD patients do not have Lewy bodies (7). Parkin is expressed in neuronal cell bodies and processes, but not in glial cells in the midbrain, basal ganglia, cerebral cortex, and cerebellum. The gene encodes an E3 ubiquitin ligase responsible for the addition of ubiquitin to specific substrates, targeting them for degradation by the proteasome. The ubiquitin-proteasome system is a main cellular pathway that promotes removal of damaged or misfolded proteins. The involvement of Parkin in the ubiquitin-proteasome system strengthens the hypothesis that protein degradation and aggregation are critical in the pathogenesis of PD. Dopamine covalently modifies and inactivates the E3 ligase activity of Parkin, suggesting a potential explanation of cell type-specific vulnerability (12). Compromised ubiquitin ligase activity may not be the only mechanism of Parkin pathogenesis, however, because in vivo studies have revealed an essential role for fly and mouse Parkin in maintaining mitochondrial integrity and physiology. Parkin-null-mutant flies have severe mitochondrial pathology associated with reduced life span, apoptosis, flight muscle degeneration, and sensitivity to oxidative stress (13). Moreover, Parkin-deficient mice have decreased mitochondrial respiratory capacity and show evidence of increased oxidative damage (14). Complex I activity is consistently decreased in leukocytes from PD patients with PARKIN mutations (15), but the mechanism by which Parkin may modulate mitochondrial function remains unknown.
Pink1 was mapped at the PARK6 locus in 2004 from studies of 2 autosomal-recessive PD families (16). So far, more than 30 homozygous or compound heterozygous Pink1 mutations have been linked to autosomal-recessive PD (17). The Pink1 mutations are the second most common cause of autosomal-recessive PD after Parkin. In addition, Pink1 mutations are responsible for a small, but still considerable, percentage of sporadic cases of parkinsonism. Notably, in 1 case, peripheral sensorimotor neuropathy was present in addition to parkinsonism and another case had severe muscular fatigue (17). Pink1 mRNA was found to be present in all adult tissues with more abundant expression in the heart, skeletal muscles, and testis (18). Pink1 was further revealed to be a ubiquitous brain protein expressed in both the gray matter and white matter of all brain regions (but more abundantly in the gray matter), and it has been found in all cell types, including neurons, glia, endothelial cells, and blood vessel smooth muscle cells (19).
Pink1 encodes a protein with a mitochondrial targeting motif and a highly conserved protein kinase domain that shares homology with the Ca2+/calmodulin family of serine-threonine kinases (16). It was initially reported that overexpressed Pink1 proteins colocalized with MitoTracker, a mitochondria-specific dye, in cultured cells (16). Subcellular fractionation studies confirmed that endogenous Pink1 is present in the mitochondria-enriched fraction from human and rat brains. Further submitochondrial fractionation and immunogold electron microscopy revealed that Pink1 proteins are localized to mitochondrial cristae, predominantly the inner mitochondrial membrane and less so to the outer mitochondria membrane (19).
DJ-1 was mapped to the seventh PARK locus (PARK7). DJ-1 mutations are recessively inherited and result in early-onset PD; however, the mutations only account for a small portion of PD cases. DJ-1 expression is detected in both neurons and glia. The DJ-1 gene is ubiquitously expressed, endogenous protein is distributed primarily in the cytosol, and a small fraction has been found to associate with mitochondria. Oxidation causes the modified protein to relocalize to mitochondria (20). Both Drosophila and mouse models suggest that DJ-1 acts as a sensor of cellular ROS levels and interacts with the phosphatidylinositol 3-kinase pathway (21, 22). DJ-1-deficient mice show increased sensitivity to MPTP, and DJ-1 null embryonic cortical neurons show increased sensitivity to oxidative stress. DJ-1 can function as a redox-sensitive molecular chaperone that is capable of preventing the aggregation of α-synuclein. DJ-1 might also act as a transcriptional cofactor that regulates the response to oxidative stress by stabilizing the transcriptional regulator Nrf2, a master regulator of antioxidant genes (23).
The gene associated with the PARK8 locus was identified as leucine-rich repeat kinase 2 (LRRK2). Mutations in LRRK2 constitute the most common known cause of familial and sporadic parkinsonism (7). LRRK2 gene expression is surprisingly low in the affected dopamine neurons of the human SN, whereas much higher expression was seen in striatal neurons that receive DA input. The LRRK2 protein has been shown to associate with membranous and vesicular structures such as lysosomes, endosomes, transport vesicles, and mitochondria (24). Little is known about LRRK2 function. It is a large multidomain protein that has kinase and GTPase domains. Parkinson disease-associated LRRK2 mutations that enhance kinase activity induce a progressive reduction in neurite length and branching in cultured neurons (25), and it seems to be involved in the regulation of translation through phosphorylation of 4E-BP (26). The detailed cell biologic function of LRRK2 and the range of physiological and pathological substrates for LRRK2 remains to be determined.
The ATP13A2 gene corresponds to the PARK9 locus. Mutations in ATP13A2 were detected in families with recessively inherited Kufor-Rakeb syndrome. These patients experience early-onset parkinsonism with additional neurological features including pyramid degeneration and dementia (27). The ATP13A2 gene encodes a previously uncharacterized predominantly neuronal ATPase. Transfected wild-type ATP13A2 protein is localized to the lysosomes, whereas truncated mutants were retained in the endoplasmic reticulum and degraded by the proteasome (27). The essential contribution of the ATP13A2 protein to lysosomal integrity in neurons remains unknown, whereas its link to macroautophagy is certainly intriguing.
The gene encoding HtrA serine peptidase 2 (HTRA2) is the candidate gene at the PARK13 locus. HTRA2 is localized to the intermembrane space of the mitochondria and released into the cytosol during apoptosis. Pathogenic mutations lead to defective activation of protease activity, increased susceptibility to stress, and mitochondrial dysfunction in vitro (28).
PINK1 MODULATES MITOCHONDRIAL MORPHOGENESIS AND DISTRIBUTION
Recent characterization of Drosophila and murine models of Pink1-associated PD promise to enhance understanding of how Pink1 mutations lead to PD in human patients (29-33). Both the Drosophila and murine models recapitulate some, although not all, of the features of PD. No significant loss of DA neurons was detected in the murine model, whereas a modest loss of DA neurons was reported in fly Pink1 models (29, 31, 32). There is a remarkable reduction of dopamine content in the fly model but not in the murine model. Nevertheless, impairment of DA physiology in the murine model was documented. Evoked dopamine release is decreased in striatal slices, and the quantal size and release frequency of catecholamine in dissociated chromaffin cells from the murine model are also reduced. Intracellular recordings of striatal medium spiny neurons, the major DA target, revealed specific impairments of corticostriatal long-term potentiation and long-term depression (29, 30). Therefore, pathogenic DA physiology seems to precede nigrostriatal neuronal degeneration in the murine model.
Dramatic morphological abnormalities of mitochondria have been observed in the muscle of the fly model. Overall mitochondrial network morphology, which has yet to be examined in the murine model, was aberrant in DA neurons of the fly model. Electron microscopy revealed swollen mitochondria short of cristae in DA neurons of the fly model, although these data need to be independently confirmed (32). By contrast, there were no gross changes in the ultrastructure or the total number of mitochondria in the murine model. Notably, the number of larger mitochondria is selectively increased in the striatum (29, 30).
The relationship between larger intact mitochondria and swollen mitochondria observed in the Pink1 models is unclear, but the increased number of larger mitochondria could be an indication of imbalanced mitochondrial fission and fusion (see later). How elevated mitochondrial fusion compromises presynaptic function is a critical issue regarding the normal function of Pink1 and is essential for understanding how dysfunction of Pink1 leads to degeneration of DA terminals. Careful histological studies combined with the recording of miniature excitatory and inhibitory postsynaptic currents could provide insights. Adenosine triphosphate levels are remarkably reduced in the fly but not in the murine model. Impaired mitochondrial respiration in the striatum but not in the cerebral cortex was detected in young Pink1-mutant mice, however, suggesting specificity of this defect for the DA circuitry. Aging also seems to be able to exacerbate mitochondrial dysfunction in the mice model. Furthermore, both Drosophila and murine models exhibit sensitivity to multiple stresses, including oxidative stress.
Several recent studies have concluded that Pink1 plays an important role in modulating mitochondrial morphology (30, 34-36). Loss of Pink1 function leads to elongated mitochondria and an increased number of larger mitochondria (Fig. 2), whereas enforced expression of Pink1 results in shorter mitochondrial units in DA neurons (34). Pink1 genetically interacts with mitochondrial morphogenesis machinery, such as Drp1, Opa1, mitofusin (Mfn), lending further support for a role of Pink1 in regulating mitochondrial morphogenesis.
Taken together, the events by which mutations of Pink1 lead to PD patients could be envisioned in the following simplified scheme based on the information from fly and murine models (Fig. 3). Pink1 promotes mitochondrial fission by interacting with mitochondrial morphogenesis machinery. Mitochondrial fission would facilitate the localization of mitochondria to the engaged synapses and regulate synaptic strength and integrity. In parallel, fission could be thought of as “guerilla warfare” that mitochondria engage to fight against stress and preserve mitochondrial genome and entities, thus to promote neuron survival. Because loss of function of Pink1 results in elevated mitochondrial fusion, the number of larger mitochondria with reduced motility would consequently increase, and mitochondrial trafficking might be impeded. Excessive fusion might lead to swollen, fragmented, and presumably dysfunctional mitochondria as well. Local supply of ATP and buffering of Ca2+ would therefore be negatively impacted and potentiation of dopamine transmission impaired. A chronic deficiency in dopamine transmission might gradually lead to synapse and axonal degeneration and the demise of DA neurons, likely because of a combination of loss of mitochondrial ATP for maintaining ionic gradients, loss of mitochondrial Ca2+ buffering, and dearth of the paracrine feeding of trophic factors. In parallel, aberrant mitochondria generated locally in the soma and/or by retrograde transport might also promote neuronal death. Considering that the synapses might be more fragile than other neuronal structures, synaptic dysfunction might occur much earlier than the death of a neuron. Therefore, synaptic dysfunction before cell death seems to be a common theme in neurodegeneration. The involvement of Pink1 and Parkin in synaptic integrity and function implicated by recent studies might have strengthened this concept. The role of Pink1 in synaptic physiology should be further investigated. Mitochondrial dysfunction, particularly defects in complex I, has been at the center stage of PD research because of the finding of the causal relationship between administrations of complex I inhibitor MPTP and the sporadic form of PD. The characterization of Pink1 has, thus, rekindled interest in the role of mitochondrial dysfunction in PD pathogenesis.
MITOCHONDRIAL MORPHOGENESIS AND MOTILITY IN NEURONS
The CNS has an intense demand for mitochondria: the human brain consumes 20% of resting metabolic energy while comprising only 2% of total body mass (37). Mitochondria produce more than 95% of ATP used by the brain. Fifty percent to 60% of total brain ATP is used to maintain Na+, K+, and Ca2+ ion gradients, especially through Na+/K+ pumps. Within neurons, mitochondria are distributed to regions of high metabolic demands, including synapses and nodes of Ranvier. Adenosine triphosphate is required for multiple steps in synaptic neurotransmission, including neurotransmitter synthesis, synaptic vesicle mobilization, release, and recycling, and local protein translation and degradation. Mitochondria also play important roles in regulating calcium homeostasis. Mitochondria are able to buffer intracellular Ca++ ([Ca++]i) by virtue of their negatively charged mitochondrial membrane potential (ΔΨm) and low [Ca++]mito relative to the [Ca++]i. Ca++ release from mitochondria can prolong elevated [Ca++]i, which mediates important physiological functions such as exocytosis and posttetanic potentiation at presynaptic terminals. Mitochondria exhibit dynamic structural changes in vivo. In addition to the classical kidney bean shape of individual mitochondrion, mitochondria are commonly found as extended reticular and tubular networks and do not resolve into discrete units. In many cell types, these elaborate mitochondrial networks are extremely dynamic, undergoing frequent fission and fusion of tubules, tubular branching and resealing, and redistribution of tubules in response to environmental cues (38). These processes require distinct conserved GTPase proteins and their binding partners acting at mitochondrial membranes. In mammals, the key molecules for mitochondrial fission are hFis1 and Drp1. The opposing process, mitochondrial fusion, is controlled in mammalian cells by mitofusion (Mfn) and OPA1. Mfn1 and Mfn2 localize on the outer membrane of mitochondria and may directly mediate mitochondrial outer-membrane fusion. OPA1 resides in the intermembrane space and is essential for inner-membrane fusion (39). Nevertheless, the physiological advantages a dynamic mitochondrial population brings are still unclear. It is speculated that fission facilitates mitochondrial biogenesis and promotes compartmentalization of mitochondrial function such as local Ca++ buffering and ATP production, whereas fusion promotes intermitochondrial cooperation, such as the transmission of membrane potential and the exchange of mitochondria contents. In particular, mtDNA exchange between mitochondria could lead to complementation of mtDNA lesions (40). Moreover, 2 key functions of mitochondria, electron transport and regulation of apoptosis, are affected by disruption of molecules involved in mitochondrial fusion and fission (39). Mouse embryonic fibroblasts lacking OPA1 or both Mfn1 and Mfn2 have defects in mitochondrial membrane potential and respiration (41). Surprisingly, inhibition of mitochondrial fission, by downregulating expression of Drp1 in mammalian cells, also leads to a loss of mitochondrial DNA and a decrease of mitochondrial respiration coupled to an increase in the levels of cellular ROS and inhibition of cell proliferation (42, 43). Discrete mitochondrial entities generated by fission could have the advantage of being protected from mitochondrial damage or depolarization, which would otherwise spread through a connected mitochondrial network (44). Therefore, coupled with fusion, dynamic mitochondrial fission might represent an ancient stress response aimed at minimizing cellular injuries.
Mitochondrial Motility in Neurons
Because neurons have long processes with complex blanches and compartment-specific metabolic needs, mitochondria-dependent functions cannot solely depend on passive diffusion of mitochondria. Indeed, mitochondria have to be delivered appropriately to serve the spatial and temporal needs of neurons. Mitochondrial distribution and trafficking depend on 2 seemingly opposite processes: transport and docking. Mitochondria seem to move along microtubules and can move in both directions along the axon via differential cytoskeletal motors (e.g. kinesin, dynein, and myosin motors), adaptor, and scaffolding proteins. For example, Milton and Drosophila Miro are proteins that recruit mitochondria to what is likely to be a large microtubule-based complex of proteins involved in movement (45, 46). After traveling along the microtubule, mitochondria arrive at the synapse, where they transfer from microtubules to an actin-based complex that docks the mitochondrion. Mitochondrial docking can occur in response to intracellular signaling pathways stimulated by the binding of growth factors (e.g. nerve growth factor) extracellularly or by local Ca++ stimuli (47).
Compared with mitochondria in axons, dendritic mitochondrial distribution and transport have been less studied. In a growing neurite, mitochondrial recruitment may occur during the process of synapse formation and stabilization because mitochondria would be in a position to provide energy for future and ongoing synaptic activity. The most significant consequence of impaired mitochondrial trafficking in dendrites might be a failure of remodeling of postsynaptic structures (48). Impaired axonal mitochondria transport has been implicated in multiple neurodegenerative conditions such as Alzheimer disease, Huntington disease, and amyotrophic lateral sclerosis (ALS) (49). Possible causes include mitochondrial injury, disrupted Ca2+ homeostasis, oxidative stress, glutamate excitotoxicity, defects in motor proteins or their interacting/modifying proteins, and protein aggregation. An increase in the number of mitochondria, especially with a rounded morphology, was observed in proximal axonal segments of anterior horn cells in human and mouse models for ALS (50). These mitochondria were thought to accumulate because of impeded axonal transport and contribute to axonal degeneration. A very similar mitochondrial distribution pattern was observed in DA neurons with overexpression of Pink1 or Parkin (34), raising the possibility that Pink1/Parkin might interfere with mitochondrial trafficking as well as morphogenesis. Indeed, a recent study demonstrated that Pink1 interacts with the mitochondrial transport complex consisting of Miro and Milton (51). Further research will clarify the role of docking as well as transport defects in multiple neurodegenerative conditions, including PD, and how common mitochondrial trafficking impairment is involved in CNS disorders.
Mitochondrial Morphogenesis in Neurons
Mitochondrial movement and morphogenesis are not completely independent from one another. It could be speculated that if 2 mitochondria cannot travel across the cell to dock with one another, they cannot fuse. Therefore, a decrease in mitochondrial movement could result in a secondary decrease in mitochondrial fusion. In contrast, mitochondrial fission might be independent of mitochondrial movement because mitochondria can stay stationary and still undergo fission. Therefore, it could be predicted that a decrease in mitochondrial movement would tip the balance of mitochondrial dynamics toward fission. On the other hand, shorter mitochondria seem to move more often and are more easily transported in axons and dendrites. In mouse peripheral axons, moving mitochondria are half the size of stationary mitochondria (∼1.5 μm vs 3 μm) (52). It is possible that larger cargoes might require more motor proteins and more ATP to power motility and, thus, have reduced motility. Furthermore, the complicated morphology of neuronal processes with narrow passageway imparts unique challenges to larger mitochondria. Accordingly, in younger cultured cortical neurons, mitochondria are shorter and more mobile, which might allow faster and more flexible energy dispersal and Ca++ buffering for cellular demands, such as synaptogenesis. As neurons mature, mitochondria in the processes may then elongate and reduce their motility for long-term support of synaptic structures because fused mitochondrial filaments would have the advantage of compensating for uneven oxygen supply and distribute energy more evenly throughout the cell (49). In addition, mitochondrial morphogenesis might impact mitochondrial trafficking by modulating transmembrane potential Δψm. Although the exact mechanisms remain largely unknown, suppression of either fusion or fission results in reduction of Δψm (41-43). ΔΨm seems to be important for mitochondrial movement perhaps because it is a direct requirement for mitochondrial ATP synthesis. Ninety percent of mitochondria with relatively high ΔΨm move anterogradely toward growth cones in dorsal root ganglia neurons, implying that healthy mitochondria are trafficked away from their origins in the cell body. In contrast, 80% of low-potential mitochondria move retrogradely, suggesting that less healthy mitochondria may return to the cell body for repair or removal (53). Furthermore, depolarization of mitochondrial ΔΨm or inhibition of ATP synthase activity causes mitochondria to stop moving (54).
The connection between mitochondrial movement and morphogenesis is also implicated in Charcot-Marie-Tooth disease Subtype 2A (CMT2A) neuropathy caused by mutations of Mfn2, which encodes an essential mitochondria fusion molecule. Charcot-Marie-Tooth is a group of diseases in which there are abnormalities in the longest motor and sensory nerves, that is, those that innervate the hands and feet. Indeed, aggregation of axonal mitochondria is a prominent feature in the peripheral nerves of CMT2A patients (55). Most pathogenic Mfn2 mutations cause loss of mitochondrial fusion activity, which could be complemented by wild-type Mfn1 but not Mfn2 (56). In mice lacking Mfn2, mitochondria tend to cluster together in the cell body and dendritic junctions and fail to enter the distal tracts of Purkinje neurons in the cerebellum, leading to loss of dendritic growth and cell degeneration. Mutant Purkinje neurons form far fewer synaptic connections (dendritic spines), thus compromising cerebellar function before Purkinje cell death (57). In a transgenic mouse model of CMT2A, disease-associated Mfn2 dominantly promotes mitochondrial aggregation, resulting in sequestration of mitochondria to small clusters in the axons in many motor neurons eventually leading to axonal degeneration (58). Exactly how Mfn2 mutants interfere with mitochondrial transport remains to be determined, but an interesting possibility is that Mfn2 itself may function in the linker-adapter complex between mitochondria and microtubule motors in addition to its membrane remodeling capacity during mitochondria fusion.
Mutations of another mitochondrial fusion molecule called OPA1 lead to autosomal-dominant optic atrophy, which is characterized loss of visual acuity caused by degeneration of retinal ganglion neurons (59). Excessive mitochondrial fission and fragmented network were observed in samples from some patients and animal models, consistent with decreased mitochondrial fusion (60), but whether mutations of Opa1 disrupt mitochondrial transport remains to be determined.
Recently, several lines of evidence also have established the role of mitochondrial fission molecules in axonal and synaptic functions. Deficiency of the Drosophila homologue of fission molecule Drp1 caused a severe loss of presynaptic mitochondria in neuromuscular junctions. Mutant axons contained elongated mitochondria, few of which localized to synaptic boutons (61). Although basal synaptic properties were largely unaffected, neurotransmission failure was detected after intense stimulation, which was related to impaired mobility of reserved pool vesicles because of the deficit of ATP supply. Waterham et al (62) recently identified a lethal mutation in the DRP1 gene in a newborn human with microcephaly, lactic acidemia, and optic atrophy.
A direct correlation between mitochondrial fission or fusion and the number of spines and synapses has been demonstrated in primary hippocampal neuronal culture. Overexpression of mitochondrial fission protein Drp1 in these neurons resulted in increased number of mitochondria, which correlated with increased number of spines and synapses. In contrast, the expression of mitochondrial fusion protein Opa1 or dominant-negative Drp1 led to fewer spines and synapses. Reciprocally, local synaptic activity induces mitochondrial recruitment to dendritic protrusions through promoting mitochondrial fission (63). A presumed function of mitochondrial fission is to generate more mitochondria. Inhibition of mitochondrial fission might therefore result in a reduction of total mitochondrial number and an increase of large fused mitochondria. As previously discussed, large mitochondria might have reduced motility and place a limitation on the local energy supply and Ca2+ buffering in axons and dendrites. Hence, too many large mitochondria and/or too few mitochondria might impair synaptic neurotransmission, formation of dendritic arbors, and the development and plasticity of dendritic spines and synapses. Given that depletion of either mitochondrial fusion molecule, such as Mfn2, or mitochondrial fission molecule, such as Drp1, leads to synaptic dysfunction, it seems that the mitochondrial dynamics (fusion vs fission) must be delicately balanced to serve the needs of neurons.
AN EMERGING PICTURE OF THE CELLULAR FUNCTIONS OF PINK1: CENTERING ON MITOCHONDRIA
It is possible that Pink1 might function through diverse but not mutually exclusive mechanisms. Here, we summarize our current knowledge and discuss 3 scenarios regarding the possible cellular functions of Pink1, all of which are related to mitochondrial function and have implications for the pathogenesis of PD.
Pink1 Modulates Mitochondrial Morphogenesis and Distribution
Several recent studies concluded that Pink1 and Parkin act in a common pathway to modulate mitochondrial morphology (30, 34-36). A key question arises as to whether Pink1 directly participates in mitochondrial morphogenesis or merely exerts secondary effects on mitochondrial morphology. Pink1 might act cooperatively with Drp1 and Fis1, and especially serve as the execution player for scission of the inner mitochondrial membrane. Alternatively, because mitochondrial fission/fusion machinery is subjected to complex posttranslational modifications such as phosphorylation, ubiquitinylation, and sumoylation (64), it is possible that Drp1 and/or Fis1 could be modified by Pink1 with its kinase activity and/or by Parkin through its ubiquitin-ligase activity. Recently, endogenous complexes containing Drp1 and antiapoptotic Bcl-xL were identified (65). The Drosophila homologue of the antiapoptotic Bcl-2 protein (Buffy) is able to suppress most abnormalities that resulted from depletion of Pink1 (32). Hence, it is possible that Pink1 interacts with Drp1 with the facilitation of the Bcl-2 family member(s). Further studies are warranted to answer key questions such as whether mitochondrial fission apparatus could the substrates of Pink1, and if so, what the functional consequences are of such modification. A positive answer would provide stronger evidence that Pink1 is directly involved in the process of mitochondrial dynamics.
Pink1 in the Apoptosis Pathway
Mitochondria are central players in the initiation and execution of apoptosis. Mitochondrial dysfunction in cell death is characterized by a decline in mitochondrial membrane potential, respiratory defects, an increase in ROS production, changes in ATP levels, and release of apoptogenic factors, including cytochrome c. Manipulations of the machinery of mitochondrial morphogenesis seem to control the progression of apoptosis. Whereas the results revealing antiapoptotic protective function of fusion molecules (Mfn and Opa1) have been consistent, results regarding the effects of the fission molecules Drp1 and Fis1 have been conflicting. On the one hand, emerging evidence indicates that mitochondria undergo rapid and excessive fission at an early stage during apoptosis, and Drp1-mediated mitochondrial fission was shown to initiate or promote apoptosis; on the other hand, Drp1-dependent division of the mitochondrial network blocks intraorganellar Ca++ waves and protects against Ca++-mediated apoptosis (44, 66). Therefore, it would be interesting to interrogate further how Pink1 might be involved in Drp1-mediated antiapoptotic as well as proapoptotic processes.
Several in vitro studies indicate that Pink1 might be involved in the apoptosis pathway. Pink1 loss of function in cultured cells implicated a role in the maintenance of mitochondrial membrane potential and prevention of stress-induced apoptosis (16). Overexpression of wild-type, but not mutant, Pink1 in cultured cells was subsequently shown to prevent apoptosis by blocking the release of cytochrome c from mitochondria during stress (67). In contrast, downregulation of Pink1 in cultured cells was shown to increase the rate of basal and stress-induced apoptosis (68). Recent studies have revealed 2 candidate Pink1-interacting partners both of which are involved in the apoptosis pathway. First, a mitochondrial molecular chaperone TRAP1 (also called heat shock protein 75) was shown to be phosphorylated by Pink1 in cell culture (69). TRAP1 and Pink1 are colocalized in the inner mitochondrial membrane. Overexpression of Pink1 enhanced the phosphorylation of TRAP1 in cultured nonneuronal cells. The protective capacity of Pink1 against apoptosis in response to oxidative stress could be attenuated by downregulation of TRAP1, suggesting that TRAP1 might act downstream of Pink1. The interaction between TRAP1 and Pink1 needs to be further confirmed in DA neuronal culture and in model organisms. Second, a mitochondrial serine protease HtrA2/Omi was found to be a Pink1-interacting protein (70).
Mutations in HtrA2 also lead to PD, and loss-of-function models in mice exhibit phenotypes resembling parkinsonism. The phosphorylated form of HtrA2 demonstrates increased protease activity and elevated protective effect in cells upon stress. Phosphorylation of HtrA2 requires the activation of p38 stress kinase pathway and is dependent on Pink1, although Pink1 does not seem to directly phosphorylate HtrA2 (70). On the other hand, recent in vivo studies in Drosophila reached divergent conclusions on the relationship between Pink1 and HrtA2 (71, 72).
How might these findings be reconciled with our discovery that Pink1 modulates mitochondria morphogenesis? Mitochondrial membrane remodeling occurs early during apoptosis, whereas Drp1, Fis1, and Mfns all play a role in apoptosis. Because the exact mechanism by which TRAP1 or HtrA2 suppresses apoptosis is unclear, it remains possible that both proteins are involved in mitochondrial membrane remodeling. Interestingly, 2 proapoptotic Bcl-2 family members, Bax and Bak, have been demonstrated to modulate the remodeling of mitochondrial membrane during mitochondrial morphogenesis in healthy cells, and Bax colocalizes with Drp1 and Mfn2 at distinct foci on mitochondria (73). Recently, endogenous complexes containing Drp1 and the antiapoptotic Bcl-xL were identified, and Bcl-xL increases the GTPase activity of Drp1 in vitro (65). Therefore, it is likely that Pink1 might act through the apoptotic machinery to modulate mitochondrial morphogenesis or vice versa. It is, therefore, not surprising that neurons use apoptotic machinery to modulate synapse development and plasticity, given the importance of mitochondria at synapses. Indeed, overexpression of Drp1-associated Bcl-2 family member Bcl-xL in either axon (presynaptically) or dendrites (postsynaptically) increases synapse formation in primary neuronal culture. Drp1, together with Bcl-xL, seems to positively modulate the number and size of synaptic vesicle clusters and induce mitochondrial localization to vesicle clusters and synapses (65). Recent in vivo data suggested that Bcl-2 family member Bcl-w is specifically required for synapse formation in Purkinje neurons (74). Bcl-w does not control the cell number or neuronal survival in mouse brain but promotes what is likely to be mitochondrial fission in Purkinje cell dendrites. Bcl-w-null-mutant mice have extremely long mitochondria in Purkinje cell dendrites and exhibit aberrant dendritic arbors and synapses and severely ataxic behavior. Interestingly, Bcl-w acts synergistically with the glutamate receptor δ2 (Grid2) to stimulate mitochondrial fission and recruitment to the dendritic spines, which further supports the notion that synaptic activity modulates mitochondrial morphogenesis and distribution (63). It should, therefore, be noted that no robust apoptosis was detected in the DA neurons of Pink1-mutant flies or mice. Instead, defects in neurotransmission and synaptic plasticity were prominent; these correlate with increased number of larger but apparently intact and normal mitochondria in the striatum of the mutant mice, presumably caused by imbalanced mitochondrial fission and fusion (29, 30). The observation again suggests that mitochondrial morphogenesis might be an independent or discrete process, which is consistent with the notion that altered mitochondrial dynamics per se does not necessarily result in apoptosis. In neurons in particular, dysfunction and degeneration of presynaptic terminals and synapses before cell death are very likely the primary consequence of dysregulation of mitochondrial dynamics; this is a recurrent theme in Mfn-mutant mice, Drp1-mutant flies, Bcl-w-mutant mice, and Pink1-mutant flies and mice.
Pink1 Regulates Mitochondrial Energy Generation
Adenosine triphosphate decline was observed in the fly Pink1 model, suggesting that Pink1 might directly regulate mitochondrial bioenergy generation as well as maintain the morphological integrity of mitochondria. Two possible scenarios might be envisioned. First, Pink1 might regulate the biogenesis of mitochondria. This scenario is especially intriguing because Parkin has been shown to enhance mitochondrial biogenesis in proliferating cells (75). Parkin associated with key mitochondrial transcription factor A (TFAM) and positively enhanced TFAM-mediated transcription of mtDNA-encoded proteins involved in respiratory chain function and replication of mitochondrial genome, hallmarks of mitochondria biogenesis. Consistently, specific inactivation of TFAM in DA neurons has been reported to lead to defects in the oxidative phosphorylation pathway that result in a PD-like disease in mice (76). Interestingly, much lower mitochondria mass has been observed in the SN DA than in non-DA neurons in the same region (77). At the electron microscopic level, mitochondria in the SN DA neurons occupy 40% less of the soma and dendritic area than in the non-DA neurons. The molecular mechanism underlying this difference is unclear, but low numbers of mitochondria might contribute to the selective vulnerability of SN-DA neurons in PD. Second, Pink1 might modulate the mitochondrial respiratory activity by directly modifying subunits of the respiratory chain complexes. Remarkably, impaired activity of complex I was observed in the striatum of Pink1-knockout mice and in fibroblasts derived from patients bearing PINK1 pathogenic mutations (30, 78). These new findings strengthen the notion of mitochondrial respiratory defects being critical to PD pathogenesis. It remains unclear as to how impairment of complex I occurs when Pink1 is inactivated. Subunits of respiratory chain complexes might be phosphorylated by Pink1 to enhance or maintain the activity of oxidative phosphorylation.
CONCLUSIONS AND PERSPECTIVES
The evidence linking mitochondrial dysfunction to neurodegenerative disorders, in particular PD, is impressive. Recent studies of PD-associated Pink1 and Parkin have established mitochondria dynamics, broadly speaking, including morphogenesis and distribution, as a novel paradigm for PD research. Given the intricate link between mitochondrial dynamics and mitochondrial oxidative phosphorylation chain activity and ROS production, it is possible that DA neurons, being under constant oxidative stress caused by endogenous oxidative dopamine metabolites, may be particularly vulnerable to disturbance of mitochondrial dynamics. Other factors, such as the low mitochondria mass in SN DA neurons and the particular need for mitochondrial trafficking to DA nerve terminals, may also contribute to the neuropathology of PD. The particular requirement for mitochondrial function in neuronal synapses also makes synapses the primary site for cellular consequences of mitochondrial dysfunction caused by alteration of mitochondrial fission/fusion or trafficking. The phenotypes of Pink1-knockout mice support this thesis and place PD into the category of diseases characterized by synaptic failure. It remains speculative whether different forms of PD in humans, sporadic and familial, finally converge on such a common mitochondrial pathway. Studies of mitochondrial morphology in the brains of PD patients have been scarce to date; tissue samples are most often obtained at late stages of the disease in which the pathological changes such as altered mitochondrial morphology that may be linked to early stages of the disease may no longer be present. On the other hand, mitochondria derived from sporadic PD patients in the cytoplasmic hybrid cell lines display a remarkable increase of enlarged branched or reticulated, as well as swollen, subpopulation, which is an indication of dysregulated morphogenesis (79). Significantly, dysregulation of mitochondrial dynamics seems to be a recurrent theme in various forms of neurodegenerative disease. Beyond the well-established CMT2A, dominant optic atrophy, and the newly added PD, the list has been recently expanded to include Alzheimer disease.
Mitochondria of vulnerable neurons in postmortem Alzheimer disease brains exhibited significantly altered morphology (80, 81). Moreover, amyloid-β overexpression was shown to cause abnormal mitochondrial dynamics via differential modulation of mitochondrial fission/fusion proteins (81). Remaining challenging questions include determining whether mitochondria have tissue-specific features and whether mitochondrial processes such as morphogenesis and transport can be targeted to treat or ameliorate a wide spectrum of neurodegenerative diseases.