Perhaps the most researched PD-linked gene to date is SNCA, which encodes α-synuclein. While mutations in this gene are rare, the SNCA locus is appreciated to be a significant variant factor for PD development (Lill et al., 2012). The missense mutation α-synuclein-A53T (Polymeropoulos et al., 1997) was the first gene to be attributed to disease development and its discovery represents a defining moment in PD research, not only preceding the identification of many other genetic determinants of PD but also mechanisms of sporadic disease. Since then, other point mutations in α-synuclein (A30P, E46K, G51D, and H50Q), as well as gene duplications and triplications, have been identified in familial PD (Krüger et al., 1998; Singleton et al., 2003; Zarranz et al., 2004; Appel-Cresswell et al., 2013; Lesage et al., 2013). The fact that duplications and triplications of SNCA can also cause PD (Singleton et al., 2003) is significant because it indicates that elevated wild type α-synuclein alone is sufficient to cause disease. So far, A53T has been found in seven families worldwide and only one family for each other four missense mutations. Duplications are more common than triplications, which has been found in several families (Hernandez et al., 2016). α-synuclein is also understood to be a key factor of sporadic PD and present in Lewy bodies, which are abnormal proteins commonly observed in PD (Spillantini et al., 1998). α-synuclein-associated mechanisms have therefore been at the forefront of the PD research and multiple high profile discoveries have greatly contributed to the understanding of disease pathology. For example, as discussed below, the discovery that α-synuclein pathology can spread from one cell to another in a prion-like fashion has provided a key insight into how PD may develop and provide novel therapeutic strategies.

α-synuclein is a pre-synaptic protein that plays a role in SNARE complex assembly and the exocytosis of neurotransmitters (Burré et al., 2010; Garcia-Reitböck et al., 2010; Bendor et al., 2013). The pathology of α-synuclein is largely due to its propensity to aggregate; gradually transitioning from small soluble oligomers to larger insoluble fibrils, ultimately forming Lewy bodies. Although the majority of α-synuclein is present in the cytosol, the protein also localizes to mitochondria and induces dysfunction (Devi et al., 2008; Parihar et al., 2008; Liu G. et al., 2009; Nakamura et al., 2011; Subramaniam et al., 2014; Chen et al., 2015; Di Maio et al., 2016). Multiple mitochondrial targeting domains have been identified within the N-terminal domain of the protein and confer its ability to bind to components of the mitochondrial membrane (Devi et al., 2008). Specifically, α-synuclein can bind to cardiolipin (Nakamura et al., 2008), TOM20 (Di Maio et al., 2016), TOM40 (Devi et al., 2008) and VDAC (Rostovtseva et al., 2015) either to directly promote dysfunction at the membrane level or to allow import into other mitochondrial compartments. The importance of the N-terminal domain in mitochondrial targeting is emphasized by the fact that α-synuclein-A53T, a mutant with a substitution in the N-terminal region, has a greater binding affinity for mitochondria than the wild type protein (Devi et al., 2008). However, more recently it has also been argued that, rather than localizing to the mitochondria, α-synuclein specifically targets the mitochondria-associated endoplasmic reticulum membrane (MAM) where it plays a role in regulating mitochondrial morphology (Guardia-Laguarta et al., 2014).

Recent studies have demonstrated that α-synuclein induces mitochondrial dysfunction in vivo. Mice with global overexpression of human wild type α-synuclein in the brain using the Thy1 promoter exhibited age dependent accumulation of α-synuclein in mitochondria of the nigrostriatal dopaminergic pathway, impaired electron transport chain function and enhanced oxidative stress (Subramaniam et al., 2014). Mice with inducible α-synuclein-A53T exhibited severe mitochondrial fragmentation that preceded dopaminergic neurodegeneration and other pathologies (Chen et al., 2015). These detrimental effects may be caused by the direct and specific binding between α-synuclein and outer mitochondrial membrane components such as TOM20 and VDAC (Rostovtseva et al., 2015; Di Maio et al., 2016). Di Maio et al. (2016) demonstrated that oligomeric, dopamine-modified and phosphorylated species of α-synuclein bound to TOM20 and prevented the import of multiple mitochondrial proteins leading to mitochondrial dysfunction. Similarly, binding of wild-type α-synuclein to VDAC was sufficient to block the channel preventing bi-directional metabolite transfer thus promoting dysfunction (Rostovtseva et al., 2015). In addition to blocking the channel, α-synuclein was able to translocate through VDAC to gain access to inner mitochondrial compartments, many of which being direct pathological targets, including respiratory complex components. A similar relationship has been reported for TOM40 as antibodies against this channel prevented the mitochondrial import of α-synuclein (Devi et al., 2008). In addition to these detrimental effects on mitochondrial function, α-synuclein also interferes with normal mitochondrial dynamics (fission, fusion, and transport). Overexpression of both wild type and mutated forms of α-synuclein is associated with severe fragmentation and reduced motility in vitro and in vivo through direct and indirect mechanisms (Kamp et al., 2010; Nakamura et al., 2011; Gui et al., 2012; Guardia-Laguarta et al., 2014; Chen et al., 2015; Bido et al., 2017). Another negative impact of α-synuclein on mitochondria is its ability to reduce mitochondrial biogenesis (Zheng et al., 2010; Siddiqui et al., 2012). Following the onset of oxidative stress, α-synuclein can bind to the promoter region of Peroxisome proliferator-activated receptor gamma coactivator 1-alpha (PGC1α), a key component of mitochondrial biogenesis, and inhibit the expression of downstream genes (Siddiqui et al., 2012). Lastly, multiple α-synuclein species may also promote the formation of dysfunctional mitochondria by increasing the prevalence of mitochondrial DNA mutations (Martin et al., 2006; Bendor et al., 2013).

In addition to these direct effects, α-synuclein can also affect mitochondrial function through indirect mechanisms. For example, it is increasingly evident that mitochondrial function and autophagy are tightly linked. A reduced autophagic capacity has been identified in many neurodegenerative diseases, including PD. Many of the genes implicated in familial PD, including SNCA, PINK1, Parkin, LRRK2, DJ-1 and glucocerebrosidase (GBA), have functions related to autophagy and mitophagy, emphasizing the importance of this pathway in the disease. Mitochondrial dysfunction may be able to inhibit the autophagy pathway, perhaps by a reduction in ATP availability and increased reactive oxygen species (ROS). Equally, impaired autophagy can promote mitochondrial dysfunction by the reduced clearance of damaged, dysfunctional organelles. The complex relationship between these two cellular pathways is not fully understood, particularly regarding the question of whether mitochondrial dysfunction precedes impaired autophagy or vice versa. Regardless, the interaction between these two fundamental processes represents a critical area of future research in the neurodegenerative field.

As mentioned above, α-synuclein has a direct interaction with the autophagy pathway. Overexpression of wild type α-synuclein was shown to directly modulate autophagy via inhibition of Rab1a protein, which is important for autophagosome formation (Winslow et al., 2010). Impaired autophagy has been demonstrated in vivo using mice overexpressing wild-type α-synuclein (Ebrahimi-Fakhari et al., 2011) or A53T α-synuclein (Yu et al., 2009; Chen et al., 2015). Conversely, an impairment in autophagy pathways promotes the formation and accumulation of higher order α-synuclein species, thus creating a cycle of neurotoxicity (Xilouri et al., 2016). In summary, α-synuclein can impair mitochondrial function at multiple levels: inhibition of the electron transport chain, imbalanced mitochondrial dynamics, reduced mitochondrial biogenesis and autophagy blockade.

Emerging evidence indicates that α-synuclein can spread inter-cellularly through multiple mechanisms, including non-classical exocytosis, exosomal release and nanotubes that directly connect the two cells (Guo and Lee, 2014). According to the Braak staging hypothesis (Hawkes et al., 2007), the olfactory bulb and the gut are the initial spreading sites of misfolded α-synuclein. Consistent with this hypothesis, accumulation of α-synuclein has been identified in enteric neurons of patients with early-stage PD (Hilton et al., 2014). Mitochondria have been implicated in the pathological spread in neurons of both the enteric and central nervous systems. For example, structural mitochondrial abnormalities have been identified in enteric neurons of post-mortem tissue from PD patients (Baumuratov et al., 2016). In primary human fetal enteric neurons, exposure to recombinant α-synuclein induces mitochondrial dysfunction involving a reduction in complex I activity and impairment of mitochondrial respiration (Braidy et al., 2014). Exposure of enteric neurons to the environmental neurotoxin, rotenone, promotes the release of α-synuclein species which are subsequently taken up by and form aggregates in second-order neurons (Pan-Montojo et al., 2012). Exposure of rat ventral midbrain neurons to exogenous, pre-formed fibrillar structures of α-synuclein, often used to study the transmission of the protein, also produce mitochondrial dysfunction (Tapias et al., 2017). These studies, and others, suggest that mitochondria may play a central role in the spread of α-synuclein from one neuron to another. That said, however, the topic of whether α-synuclein is transmitted trans-synaptically in PD is still a topic of debate (Brundin and Melki, 2017; Surmeier et al., 2017), which is beyond the scope of this article.

First identified in a family of Japanese origin (Funayama et al., 2002) and subsequently independently confirmed in several other families from different countries, leucine-rich repeat kinase 2 (LRRK2) mutations are the most common group of mutations in PD cases (Cookson, 2015). Although highly dependent on ethnicity, LRRK2 mutations are estimated to account for up to 40% of familial cases, depending on the ethnic background, and up to 10% of sporadic PD cases worldwide (Hernandez et al., 2016), producing a disease phenotype that is similar to that of classical late-onset idiopathic PD. The most common and well characterized LRRK2 mutation, G2019S, alone is suggested to account for 4% of familial and 1% of sporadic cases. All described LRRK2 mutations, to date, affect a central enzymatic region of the protein consisting of a GTPase and kinase domain and confer a gain of function (West et al., 2007). Although more than 100 LRRK2 mutations have been reported, only a few such as (N1437H, R1441C, R1441G, R1441H, Y1699C, G2019S, and I2020T0) have been proven to cause PD and two have been nominated as risk factors (R1628P and G2385R) (Dächsel and Farrer, 2010; Paisan-Ruiz et al., 2013). The incomplete penetrance of multiple mutations has led to LRRK2 being identified as a risk factor for sporadic PD (Nalls et al., 2014; Hernandez et al., 2016). Sporadic PD patient derived cells also display a slightly increased basal activity of LRRK2, however, it is not known whether this is causative or a consequence of pre-existing pathology (Esteves et al., 2015). Interestingly, a recent study demonstrated that variability within the DNM3 gene influences the age of onset in G2019S carriers (Trinh et al., 2016). The presence of other genetic modifiers may therefore influence disease manifestation in the case of LRRK2 mutations.

The complex physiological roles of LRRK2 are not fully understood; however, it is associated with many cellular functions, particularly those involving membrane dynamics, including autophagy, cytoskeletal dynamics, vesicle dynamics and mitochondrial function. Attempts to better understand how LRRK2 mutations contribute toward PD pathology have been restricted partly due to the difficulty in creating a LRRK2 transgenic model that recapitulates important features of PD. The majority of mutant LRRK2 transgenic models do not display appreciable dopaminergic neurodegeneration, but do exhibit multiple synaptic deficits including reduced vesicle endocytosis (Arranz et al., 2015) and altered dopamine release, uptake and signaling (Li et al., 2009; Tong et al., 2009; Melrose et al., 2010; Walker et al., 2014; Beccano-Kelly et al., 2015; Yue et al., 2015). LRRK2 has been shown to impair mitochondrial dynamics by physically interacting with Drp1 and promoting mitochondrial fragmentation associated with enhanced mitochondrial fission, increased ROS generation and impaired autophagy (Niu et al., 2012; Wang et al., 2012; Stafa et al., 2014; Esteves et al., 2015). In addition to increased Drp1 expression, a concomitant increase in Drp1-S616 phosphorylation has been reported. This specific modification is a marker for increased Drp1 activity (Wang et al., 2012; Stafa et al., 2014; Esteves et al., 2015). Overexpression of the gain of function mutations G2019S and R1441C resulted in a reduction of mitochondrial membrane potential, reduced complex IV activity, increased uncoupling, altered mitochondrial motility, and an imbalance in calcium signaling (Cooper et al., 2012; Papkovskaia et al., 2012; Cherra et al., 2013; Godena et al., 2014). Ultrastructural studies show an accumulation of condensed, disorganized and damaged mitochondria consistent with deficits in mitophagy and mitochondrial dynamics (Ramonet et al., 2011; Yue et al., 2015). Finally, increased mtDNA mutations have been identified in cerebrospinal fluid (CSF) and iPSCs derived from patients with the G2019S mutation (Sanders et al., 2014; Podlesniy et al., 2016). Taken together, although the mechanism is not entirely clear, a toxic gain of kinase function in LRRK2 mutations negatively affects mitochondrial function.

Mitochondrial dysfunction induced by LRRK2 also provides a strong link to the impaired autophagy pathway in multiple models. For example, an accumulation in autophagic and lysosomal structures occur both in vitro and in vivo caused by overexpression of PD-associated mutations (MacLeod et al., 2006; Plowey et al., 2008; Gomez-Suaga et al., 2012). A block in autophagy flux, increase in p62 levels and an accumulation of autophagosomes have been demonstrated in dopaminergic neurons generated from PD patient-derived induced pluripotent stem cells (iPSCs) expressing LRRK2-G2019S at endogenous levels (Sánchez-Danés et al., 2012). Importantly, the inhibition of mitochondrial fission, using the Drp1 peptide inhibitor P110, was shown to improve autophagy and reduce detrimental changes in cell morphology caused by the G2019S mutation in PD patient fibroblasts and dopaminergic neurons derived from iPSCs (Su and Qi, 2013). Conversely, another study showed that disrupted mitochondrial motility and an increased vulnerability to multiple stressors induced by G2019S and R1441C mutations in PD patient iPSCs was rescued by the autophagy promoter rapamycin (Cooper et al., 2012) Together, these studies suggest that autophagy and mitochondrial function are bi-directionally regulating one another.

An interest in GBA as a causative factor of PD followed the clinical observations that some Gaucher’s disease patients developed Parkinsonian symptoms (Neudorfer et al., 1996) and that relatives of these patients had an increased risk of developing PD (Halperin et al., 2006). Gaucher’s disease (GD, Tsuji et al., 1987) is the most common lysosomal storage disorder and is caused by homozygous and compound heterozygous mutations in the GBA gene. GBA is a lysosomal hydrolase, specifically a GBA which acts in the lysosomal membrane to convert the sphingolipid glucosylceramide to glucose and ceramide. Approximately 300 different mutations in GBA have been identified with many shown to have reduced GBA activity. Such reduction has indeed been identified in cases of PD with GBA mutations but, importantly, it has also been identified in the substantia nigra of sporadic PD brains (Gegg et al., 2012). More severe mutations, such as L444P, increase the risk of developing PD and promote an earlier onset as well as greater cognitive decline when compared to milder mutations, such as N370S. However, the overall mortality rate between these two mutations is not significantly different (Gan-Or et al., 2015; Arkadir et al., 2016; Cilia et al., 2016); therefore suggesting that even a partial loss of function is sufficient to cause the disease. The prevalence of GBA mutations varies in different PD populations but is accountable for up to 31% of PD patients with Ashkenazi Jewish ancestry (Sidransky and Lopez, 2012).

A reduction in GBA activity has been associated with mitochondrial dysfunction. GBA inhibition, either pharmacologically or genetically result in a reduction of membrane potential and ATP production, as well as an increase in oxidative stress and mitochondrial fragmentation in vitro (Cleeter et al., 2013). These findings have been confirmed using both fibroblasts isolated from GD patients and mouse models of GD. Fibroblasts from GD patients, with an L444P mutation, exhibited a reduction in complex I-III activity, CoQ10 expression, membrane potential and ATP production as well as an increase in oxidative stress (de la Mata et al., 2015). Mouse models demonstrate a reduction in membrane potential, complex I-III activity, oxygen consumption and an increase in mitochondrial fragmentation in neurons and glia (Osellame et al., 2013; Xu et al., 2014). Further, ultrastructural analysis of mitochondrial morphology showed impaired cristae organization, with mitochondria appearing more rounded and electron dense (Xu et al., 2014).

While the mechanisms of mitochondrial dysfunction associated with GBA mutations are unclear, an impairment of mitophagy is suggested to play an important role. This is perhaps not surprising given the role of GBA in the cell and this observation further strengthens the link between autophagy and mitochondria. In mouse models and primary fibroblasts, decreased mitophagy flux accompanied the observed mitochondrial dysfunction (de la Mata et al., 2015). However, whether an impaired mitophagy response causes the observed mitochondrial dysfunction in these models or whether it is a consequence of an upstream effect is not understood. Another contributing factor to mitochondrial dysfunction in these models is neuroinflammation. In GBA knockout mouse models, activation of both astrocytes and microglia precedes disease manifestation (Enquist et al., 2007; Farfel-Becker et al., 2011). Astrogliosis has also been identified in some GD patients that exhibit parkinsonian symptoms (Wong et al., 2004). It is important to note that the majority of the studies related to GBA discussed in this review have been performed in the setting of GD rather than PD. Therefore, the conclusions made from this work, although relevant, may not be directly transferrable to cases of PD with GBA mutations.

GBA activity appears to be linked to the expression and aggregation of α-synuclein. An accumulation of aggregated α-synuclein, β-amyloid and amyloid precursor protein (APP) in the cortex, hippocampus, striatum and substantia nigra of GBA mutant mice was observed (Xu et al., 2014). Further, inhibiting GBA, using zinc finger techniques, promotes the intercellular transmission of α-synuclein species and therefore contributes toward the spread of PD pathology (Bae et al., 2014). These observations are consistent with the role of GBA in autophagy function. An increase in α-synuclein expression has also been shown to inhibit GBA activity (Mazzulli et al., 2011; Gegg et al., 2012; Murphy et al., 2014), thus further impairing the clearance of α-synuclein mediated by autophagy.

A single mutation, D620N, in the gene encoding vacuolar protein sorting 35 (VPS35) causes the development of an autosomal dominant, late-onset PD (Vilarino-Guell et al., 2011; Zimprich et al., 2011; Sharma et al., 2012). A number of other mutations in the gene have been identified; however, their pathological relevance is yet to be demonstrated. VPS35 mutations are rare and only account for approximately 1% of familial and 0.2% of sporadic PD cases.

VPS35 is a core subunit of a heteropentameric complex known as the retromer, which plays an important role in the transport of endosomes to the golgi apparatus and the plasma membrane, as well as in recycling of transmembrane protein cargo (Williams et al., 2017). In recent years VPS35 has also been reported to mediate the shuttling of cargo from mitochondria to peroxisomes and lysosomes through mitochondria-derived vesicles (Braschi et al., 2010; Sugiura et al., 2014). Models of VPS35 mutations successfully demonstrate dopaminergic neurodegeneration (Tsika et al., 2014; Wang et al., 2014, 2016; Tang et al., 2015). The pathogenic mechanism by which VPS35 induces PD is not clear, however, recent studies show that its mutation caused mitochondrial fragmentation, impaired mitochondrial function, increased ROS, α-synuclein accumulation and cell death (Tang et al., 2015; Wang et al., 2016). The D260N mutant has an increased binding interaction with Drp1, an effect that is further promoted by oxidative stress, resulting in an increased turnover of mitochondrial Drp1 complexes via their mitochondria derived vesicular trafficking to the lysosome for degradation (Wang et al., 2016). In addition to increasing mitochondrial fission, VPS35 deficiency or mutation induces mitochondrial fragmentation by impairing mitochondrial fusion (Tang et al., 2015) via mitochondrial E3 ubiquitin ligase 1 (MUL1) upregulation, leading to increased ubiquitination and proteasomal degradation of the fusion protein MFN2. Together, these studies strongly support the role of VPS35 in imbalanced mitochondrial dynamics.

Although little is known about the physiological roles of Coiled-Coil-Helix-Coiled-Coil-Helix Domain Containing 2 (CHCHD2) or its role in PD pathology, several mutations in the CHCHD2 gene cause an autosomal dominant form of late-onset PD and dementia with Lewy bodies (Funayama et al., 2015; Ogaki et al., 2015; Koschmidder et al., 2016). CHCHD2 contains a mitochondrial targeting sequence in the N-terminal domain and localizes to intermembrane space of mitochondria, where it binds to the complex IV (cytochrome c oxidase, COX). This binding is required for COX activity (Aras et al., 2015). CHCHD2 is therefore necessary for oxidative phosphorylation (Baughman et al., 2009). Indeed, reduced CHCHD2 expression results in reduced COX activity, collapsed membrane potential and an increase in ROS and mitochondrial fragmentation (Aras et al., 2015), resulting in cell death in Drosophila (Meng et al., 2017). PD-related mutant flies exhibited a chronic state of oxidative stress, reduced oxygen consumption, ATP production, increased cytochrome c release and displayed mitochondria with abnormal ultrastructure leading to degeneration of dopaminergic neurons. This study also suggests that CHCHD2 may be an important mitochondrial stress response protein and can modulate cell death signaling by interacting with cytochrome c and MICS1, a member of the Bax inhibitor-1 superfamily (Meng et al., 2017). Although CHCHD2 predominantly localizes in mitochondria and directly binds to complex IV, CHCHD2 is also a transcription factor of the complex IV subunit 4 isoform COX4I2 (Aras et al., 2015). Because of this unique feature, loss of CHCHD2 function negatively affects complex IV both functionally and structurally.

Mutations in PTEN-induced kinase 1 (PINK1) and Parkin both cause an almost identical form of autosomal recessive, early onset familial PD which can present with or without Lewy Bodies (Kitada et al., 1998; Sasaki et al., 2004; Valente et al., 2004a; Samaranch et al., 2010). A mutation in Parkin was first identified in a Japanese family (Matsumine et al., 1997; Kitada et al., 1998) and are now collectively shown to encompass a wide range of variations spanning the entire length of the gene conferring a loss of function (Klein and Westenberger, 2012). Parkin mutations are considered as the most prevalent autosomal recessive mutations in PD accounting for approximately up to 77% of familial early-onset PD and 10–20% of early-onset PD in general (Klein and Lohmann-Hedrich, 2007; Klein and Westenberger, 2012; Kilarski et al., 2012). Mutations in PINK1, first identified in an Italian family (Valente et al., 2001), are the second most common cause of early-onset autosomal recessive PD accounting for up to 9% of cases (Klein and Westenberger, 2012). Mutations span the length of PINK1, with the majority being missense or non-sense mutations and, similarly to Parkin, confer a loss of function. Heterozygous PINK1 and Parkin mutations may also be considered as a risk factor for developing PD (Khan et al., 2005; Criscuolo et al., 2006; Hedrich et al., 2006; Hiller et al., 2007).

PINK1 is a Ser/Thr kinase that is localized to the mitochondria (Hatano et al., 2004; Valente et al., 2004a,b; Zhou et al., 2008). In a normal, healthy mitochondrion PINK1 binds to the outer mitochondrial membrane (OMM), and is then imported into the inner mitochondrial membrane (IMM) via the translocase of the outer membrane (TOM) complex and translocase of the inner membrane 23 (TIM23) in a membrane potential-dependent manner. Full length PINK1 is then rapidly cleaved by mitochondrial processing peptidase (MPP) and presenilin-associated rhomboid-like protease (PARL). Cleaved PINK1 is released into the cytosol for other non-mitochondrial functions (Dagda et al., 2014) and can be degraded by the proteasome (Meissner et al., 2011; Yamano and Youle, 2013). Hence, PINK1 is maintained at a low level in healthy mitochondria. However, upon mitochondrial damage and the loss of membrane potential, PINK1 import is blocked causing it to accumulate at the OMM (Jin et al., 2010; Lazarou et al., 2012; Okatsu et al., 2013). At the OMM, the kinase domain of PINK1 faces toward the cytosol, enabling it to phosphorylate both mitochondrial and cytosolic proteins (Zhou et al., 2008). Although not consistently observed, PINK1 has been reported to phosphorylate ubiquitin and ubiquitin-like (Ubl) domain of Parkin at a conserved Ser65 residue. This phosphorylation initiates a cascade that consequently activates and recruits Parkin, a ubiquitin E3 ligase, to mitochondria (Durcan and Fon, 2015; Roberts et al., 2016). Activated Parkin promotes ubiquitination of multiple OMM proteins targeting them for degradation via the proteasomal pathway (Gegg et al., 2010; Chan et al., 2011; Yoshii et al., 2011; Kazlauskaite et al., 2014) as well as the recruitment of autophagy adaptor proteins (Roberts et al., 2016). The PINK1/Parkin axis is therefore an important pathway for mitochondrial quality control and the selective removal of the damaged organelle by mitophagy (Geisler et al., 2010; Narendra et al., 2010, 2012; Vives-Bauza et al., 2010).

Mitochondrial abnormalities have been reported in multiple in vivo and in vitro models of PINK1 and Parkin. PINK1 deficient and PD-associated mutant models display reduced mitochondrial membrane potential, ATP levels, reduced respiratory capacity via complex I and IV activity, increased mitochondrial calcium levels, sensitized mitochondrial permeability transition pore opening and increased ROS production (Gautier et al., 2008, 2012; Piccoli et al., 2008; Dagda et al., 2009; Gandhi et al., 2009; Gegg et al., 2009; Liu W. et al., 2009; Morais et al., 2009, 2014; Cui et al., 2010; Heeman et al., 2011). Isolated mitochondria from brains of aged PINK1 null mice display complex I deficits (Gautier et al., 2008; Morais et al., 2014) and reduced calcium buffering capacity (Akundi et al., 2011). Aged Parkin null brains show alterations in the expression of a wide range of mitochondrial proteins, with many relating to a reduced respiratory capacity and increased oxidative damage (Palacino et al., 2004; Periquet et al., 2005; Stichel et al., 2007). PINK1 and Parkin also play a regulatory role in mitochondrial dynamics by mechanisms involving the turnover of fusion and fission associated proteins such as Mfn1/2 and Drp1 (Clark et al., 2006; Deng et al., 2008; Poole et al., 2008; Yang et al., 2008; Lutz et al., 2009; Cui et al., 2010; Gegg et al., 2010; Tanaka et al., 2010; Ziviani et al., 2010; Exner et al., 2012; Buhlman et al., 2014; Pryde et al., 2016). However, whether PINK1 and Parkin are pro-fission or pro-fusion has been a topic of debate (Exner et al., 2012; Pilsl and Winklhofer, 2012).

In addition to the above effects on mitochondrial function and dynamics, PINK1 and Parkin mediate biogenesis through an indirect interaction with PGC-1α (Shin et al., 2011; Stevens et al., 2015; Lee et al., 2017). The loss of PINK1/Parkin activity leads to a reduced clearance, and consequent build-up, of the Parkin Interacting Substrate (PARIS/ZNF746) which is a transcriptional repressor of PGC-1α. An increase in PARIS/ZNF746 is identified in the striatum and substantia nigra of patients with autosomal recessive and sporadic PD, as well as in Parkin knockout mice (Shin et al., 2011).

Homozygosity mapping in two consanguineous families from Italy and the Netherlands with early-onset PD identified DJ-1 mutations (Bonifati et al., 2003). Collectively, mutations in DJ-1 account for approximately 1–2% of early onset recessive cases of PD and encompass deletions, homozygous and heterozygous point mutations and truncations (Bonifati et al., 2003). DJ-1 is a small protein of 189 residues localized to the cytoplasm, the nucleus and associated with mitochondria (Canet-Aviles et al., 2004; Li et al., 2005; Junn et al., 2009). Among its functions, DJ-1 is involved in regulating mitochondrial activity (Hayashi et al., 2009; Junn et al., 2009) and protecting against oxidative stress (Taira et al., 2004).

DJ-1 can interact with mitochondrial complex I subunits and is translocated to the mitochondria under stress conditions (Hayashi et al., 2009). DJ-1 and its mutant forms could associate with the chaperone protein Hsp70, a link that was further increased under oxidative stress conditions. These results suggest that translocation of DJ-1 to the mitochondria may occur with the aid of mitochondrial chaperone proteins (Li et al., 2005). It is believed that the dimeric form of DJ-1 is required for antioxidative stress reactions, while monomeric mutant forms of DJ-1 are toxic for cells (Zhang et al., 2005; Maita et al., 2013). Further studies also demonstrated mitochondrial dysfunction and reduced membrane potential in DJ-1 knockout mice and flies (Hao et al., 2010; Giaime et al., 2012). Consistent with its protective role in mitochondria, DJ-1 knockout mice exhibit enhanced vulnerability to the mitochondrial toxicant 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP) (Kim et al., 2005; Manning-Boğ et al., 2007).

DJ-1 is required for the proper balance of mitochondrial dynamics; primary cortical neurons and embryonic fibroblasts cultured from DJ-1 knockout mice and human dopaminergic neuronal cells with DJ-1 knockdown exhibit mitochondrial fragmentation and reduced Mfn2 levels (Irrcher et al., 2010; Krebiehl et al., 2010; Thomas et al., 2011). Blocking mitochondrial fission or promoting mitochondrial fusion rescues this abnormal mitochondrial morphology (Irrcher et al., 2010; Thomas et al., 2011). DJ-1 has also been suggested to participate in the mitophagy pathway. In fact, loss of DJ-1 leads to mitochondrial impairment and dysfunctional autophagy (Irrcher et al., 2010; Krebiehl et al., 2010; Thomas et al., 2011) and overexpression of PINK1 and Parkin is protective (Irrcher et al., 2010; Thomas et al., 2011). These results are consistent with the proposal that DJ-1 may be aiding PINK1/Parkin in the clearance of misfolded proteins and damaged mitochondria (Xiong et al., 2009; Thomas et al., 2011). Taken together, DJ-1 has a critical role in the oxidative stress response, mitochondrial function and basal autophagy.

Mutations in the PLA2G6 gene have been described in patients with autosomal recessive early-onset dystonia-parkinsonism with widespread accumulation of Lewy bodies and hyperphosphorylated tau (Paisan-Ruiz et al., 2009; Miki et al., 2017). Brain iron accumulation is found in most affected individuals (Paisan-Ruiz et al., 2009). The PLA2G6 gene encodes a 85-kDa group VI calcium-independent phospholipase A₂β (PLA2G6) and is involved in the hydrolysis of glycerophospholipids to release free fatty acids and lysophospholipids (Ma and Turk, 2001). PLA2G6 is confined to the mitochondrial compartment where it remodels membrane phospholipids, protects mitochondria from oxidative damage during apoptotic induction and takes part in calcium signaling (Gadd et al., 2006; Seleznev et al., 2006; Strokin et al., 2012). Although the pathological mechanism of PLA2G6 remains elusive, it has been suggested to affect lipid homeostasis and metabolism as well as impact mitochondrial membranes (Beck et al., 2011). Further, PLA2G6 knockout mice displayed degenerated mitochondrial membranes and elevated levels of synuclein (Sumi-Akamaru et al., 2016). PLA2G6 has been reported to be immunopositive in the Lewy bodies of idiopathic PD, suggesting its pathogenic role extends beyond familial PD (Miki et al., 2017).

ATP13A2 encodes ATPase type 13A, which is usually localized to lysosomes. Mutations in ATP13A2 were initially identified in Kufor–Rakeb syndrome (Ramirez et al., 2006), which is a recessively inherited disease with multiple brain regions are affected. Fibroblasts derived from these patients display reduced ATP synthesis, increased mtDNA mutations, impaired oxygen consumption and mitochondrial fragmentation (Grunewald et al., 2012). Mutations in ATP13A2 are also found in early onset parkinsonism (Di Fonzo et al., 2007), suggesting that ATP13A2 may also be relevant to PD. ATP13A2 is a multifunctional protein that has been linked to endosome-lysosome dynamics (Dehay et al., 2012; Usenovic et al., 2012), mitochondrial function (Grunewald et al., 2012; Gusdon et al., 2012) and toxicity induced by divalent cation metals (Mn²⁺ and Zn²⁺) (Gitler et al., 2009; Kong et al., 2014). Consistent with these roles of ATP13A2, loss of function in this protein has been reported to reduce the ability of various vesicular structures such as lysosomes to regulate divalent metal cations such as Mn²⁺ and therefore sensitize cells to metal toxicity, accumulation of α-synuclein and mitochondrial dysfunction (Gitler et al., 2009; Kong et al., 2014). These effects are most likely secondary to autophagy deficits, causing an accumulation of misfolded proteins and damaged mitochondria. Collectively, ATP13A2 may exert an indirect, protective role on mitochondria and the cell as a whole. Indeed, overexpression of ATP13A2 is protective against α-synuclein-induced toxicity (Gitler et al., 2009; Kong et al., 2014).

Mutations within the FBXO7 (F-box only protein 7) gene have been shown to promote either a typical PD phenotype (Lohmann et al., 2015) or an early onset form of parkinsonian-pyramidal syndrome (Shojaee et al., 2008; Di Fonzo et al., 2009). FBX07 has been linked to mitochondrial function by acting on the same pathway as PINK1 and Parkin. It directly interacts with both PINK1 and Parkin and therefore plays a role in mitochondrial maintenance and mitophagy (Burchell et al., 2013; Zhou et al., 2015). These studies demonstrate that FBXO7 assists in the translocation of Parkin to the mitochondria in response to cell stress and PD-associated mutations leading to a mislocalization of FBXO7 to the cytosol where it can form toxic aggregates that are deleterious to the cell (Zhou et al., 2015).

A potent non-competitive mitochondrial complex I inhibitor derived from plants of the Leguminosa family (Soloway, 1976), rotenone is a pesticide/piscicide associated with increased risk of developing PD (Tanner et al., 2011; Kamel, 2013). Because the half-life of rotenone is relatively short (∼3 days) (Hisata, 2002), consumption of contaminated food products is an unlikely exposure route. The role of rotenone in PD is therefore most likely through low chronic exposure rather than acute toxicity. Epidemiological studies have demonstrated increased PD risk of ∼1.5–3-fold in individuals who utilized rotenone agriculturally or who lived in close proximity to its use (Tanner et al., 2011). As a highly lipophilic chemical, rotenone passes freely through the blood-brain-barrier (BBB) and cellular membranes and once inside cells it can directly alter mitochondrial function. Rotenone impairs oxidative phosphorylation by inhibiting NADH-ubiquinone reductive activity via binding to the PSST subunit of complex I of the electron transport chain (Schuler and Casida, 2001). This mechanism of rotenone is relevant to the observations that mitochondrial complex I activity is reduced in PD patients (Parker et al., 1989; Schapira et al., 1989). Rotenone has also been shown to enhance mitochondrial fission. In primary cortical neurons, rotenone induces rapid mitochondrial fragmentation within 2 h of treatment, prior to any cellular changes indicative of cytotoxicity (Barsoum et al., 2006). Altering mitochondrial dynamics, by promoting mitochondrial fusion with Mfn2 and blocking Drp1 with Drp1-K38A averts this neurotoxicity (Barsoum et al., 2006). Although the neuropathology is variable, rotenone treatment has been shown to induce motor impairment, neuroinflammation, nigrostriatal degeneration, and accumulation and phosphorylation of α-synuclein (Betarbet et al., 2000; Cannon et al., 2009) in rodent models of PD.

In addition to impairing mitochondrial function, rotenone has been demonstrated to alter mitochondrial transport and protein interactions within neurons. For example, rotenone induces microtubule depolymerisation, thereby interfering with the intracellular transport of vesicles and organelles, including mitochondria (Choi et al., 2011). A recent study revealed increased α-synuclein association with mitochondria, specifically outer membrane protein TOM20, in nigrostriatal dopaminergic neurons of rotenone-treated rats relative to controls (Di Maio et al., 2016). This observation suggests that rotenone exposure may induce α-synuclein-mediated changes in mitochondrial function and morphology to further promote PD pathology.

The herbicidal mechanism of paraquat (PQ) involves interference with photosynthesis and oxygen free radical production, resulting in damage of plant membranes (Tanner et al., 2011; Breckenridge et al., 2013). Initially based on its similar structure to MPP⁺, this widely used herbicide was hypothesized to be a potential environmental parkinsonian toxicant (Snyder and D’Amato, 1985). Epidemiological studies demonstrate that PQ exposure, both alone and in conjunction with other pesticides, increases the risk of developing PD (Costello et al., 2009; Ritz et al., 2009; Tanner et al., 2011). To further understand the mechanism by which this molecule induces dopaminergic neurotoxicity, PQ has been used experimentally to model PD (Tieu, 2011). PQ is unable to diffuse across the BBB and relies on transport by the L-neutral amino acid transporter to enter the brain (Shimizu et al., 2001; McCormack and Di Monte, 2003). Once inside the brain, PQ can enter cells through the dopamine transporter (DAT) and organic cation transporter-3 (Rappold et al., 2011). In the cytosol, PQ induces toxicity primarily through redox cycling with cellular diaphorases such as NADPH oxidase and nitric oxide synthase (Day et al., 1999), leading to the generation of superoxide. Despite its similar structure to MPP⁺, PQ is not a complex I inhibitor (Richardson et al., 2005), although this is the site where superoxide is generated from PQ redox-cycling in the mitochondria (Cochemé and Murphy, 2008). Other work also implicates complex III in the formation of paraquat-induced ROS (Castello et al., 2007). Consistent with the role of generating mitochondrial ROS, a recent study reported mitochondrial DNA damage in PD patients with prior history of exposure to PQ (Sanders et al., 2017). The mechanism by which PQ enters mitochondria, however, is currently unknown.

PQ has also been reported to increase mitochondrial fission by reducing the levels of key mitochondrial fusion factors, Mfn1/Mfn2 (Tanaka et al., 2010). In rodent models PQ has been shown to cause neuroinflammation, as well as α-synuclein upregulation and aggregation (Manning-Bog et al., 2002; Wu et al., 2005; Fernagut et al., 2007; Purisai et al., 2007). The use of PQ to model PD has been contested by some, partly because studies utilizing wild-type mice demonstrated variable results, with loss of dopaminergic nigral cell bodies but a lack of consistency in the loss of the corresponding striatal terminals (McCormack et al., 2002; Thiruchelvam et al., 2003; Tieu, 2011); however, this lack of striatal damage has been suggested to be the result of compensatory sprouting from surviving dopaminergic neurons and uptake of PQ into non-dopaminergic striatal cells. In contrast to wild-type mice, mutant mice deficient in OCT3 (Slc22a3), a cation transporter which can uptake PQ into cells, display enhanced striatal damage upon exposure to PQ. This increased sensitivity likely results from reduced buffering capacity by non-dopaminergic cells, leading to increased availability of PQ to damage dopaminergic terminals (Rappold et al., 2011). Despite the controversy which still surrounds the use of PQ to model PD, current evidence from epidemiological studies and mechanistic investigations appear to sustain the conclusion that PQ may increase the risk of developing PD.

Dieldrin is one of the organochlorine pesticides that was first commercially available in 1950, with wide use in the United States until the late 1980’s. Despite decades of research into potential risk of dieldrin, it is still extensively used in many developing countries. Dieldrin has a long half-life (∼25 years) which contributes to a persistent high risk of exposure, even decades after its use has ceased. Human exposure risk may be mediated through consumption of dairy products, meat, and fish (Kanthasamy et al., 2005), although its high lipophilicity also confers risk of dermal exposure, as it can be rapidly absorbed through the skin. In addition to its environmental persistence, the ability of dieldrin to bioaccumulate, with a half-life of ∼300 days in humans, contributes to the risk it confers, as it can readily pass the BBB and amass in the brain. Indeed, dieldrin has been detected in 30% of post-mortem brain tissue of PD patients, with none in age-matched control brains (Fleming et al., 1994). Another study reported dieldrin in both control and patient brain tissue, with significantly higher levels in the caudate nucleus of PD patients (Corrigan et al., 1996). These results support a potential role of this pesticide in promoting nigral cell death and PD progression (Kanthasamy et al., 2005).

Dieldrin has been shown to exert multiple effects on mitochondria, including increased ROS production (Chun et al., 2001) and stimulation of mitochondria-mediated apoptosis (Kitazawa et al., 2003; Kanthasamy et al., 2008). Dopaminergic cells have increased sensitivity to dieldrin when compared to other cell types as demonstrated in vitro (Sanchez-Ramos et al., 1998; Kitazawa et al., 2001) and in vivo (Sharma et al., 1976; Wagner and Greene, 1978; Heinz et al., 1980). Dieldrin (and other pesticides) have also been shown to block the function of mitochondrial aldehyde dehydrogenase (ALDH class 2) (Fitzmaurice et al., 2013, 2014). ALDH2 metabolizes toxic aldehydes including 3,4-dihydroxyphenylacetaldehyde (DOPAL), which is a reactive neurotoxic metabolite of dopamine. Exposure of individuals with genetic variation in ALDH2 to these pesticides was associated with a two–sixfold increase in their risk of developing PD (Fitzmaurice et al., 2014). ALDH2 inhibition therefore has been suggested as a pathogenic mechanism in PD. Furthermore, based on a recent work that investigated the proteomic effects of dieldrin, 18 proteins involved in the respiratory chain were affected by dieldrin treatment (Cowie et al., 2017), further supporting the influence of this pesticide on mitochondrial function.

Dieldrin has also been demonstrated to directly stimulate α-synuclein fibril formation in vitro, by inducing a conformational change which may serve as a precursor or seed for the aggregation of further protein (Uversky et al., 2001b). Prolonged dieldrin exposure has been reported to inhibit the ubiquitin-proteasome protein degradation system (Sun et al., 2004), further promoting protein aggregation. In summary, although the mechanistic understanding remains incomplete, dieldrin has been shown to induce dopaminergic neurotoxicity through mitochondrial dysfunction, decreased protein degradation and apoptosis initiation.

Pyrethroids are synthetic derivatives of the naturally occurring pyrethrum from chrysanthemum flowers. These pesticides are widely used both agriculturally and in the household (Elwan et al., 2006), however, despite their classification as safe substances (Casida et al., 1983), pyrethroid intoxication in humans has been observed (Chen et al., 1991). These chemicals, which include permethrin and deltamethrin, readily cross the BBB and may accumulate at considerable concentrations within the brain (Anadón et al., 1996). Permethrin and deltamethrine have been shown to increase DAT-mediated dopamine uptake by up to 30% in the mouse striatum and induce apoptosis in dopaminergic neuroblastoma cells (Elwan et al., 2006). Because the DAT mediates neurotoxicity of some neurotoxicants (Gainetdinov et al., 1997; Donovan et al., 1999; Rappold et al., 2011), pyrethroids may enhance the vulnerability of dopaminergic neurons to toxic compounds. Furthermore, permethrin has been shown to potently inhibit mitochondrial complex I function in rat liver (Gassner et al., 1997), however, a subsequent study demonstrated that apoptosis induced by pyrethroids (Deltamethrin) is mitochondrial independent (Hossain and Weiner, 1993). Rather, deltamethrin induces calcium overload through Na⁺ channels, leading to endoplasmic reticulum stress.

A recent study investigated the potential use of permethrin to model PD in a progressive model of early life exposure to the pesticide (Nasuti et al., 2017). Treatment of Wistar rat pups with permethrin from post-natal days (PND) 6–21 induced cognitive impairments, reduced striatal dopamine levels and reduced dopaminergic neurons in the substantia nigra; all common neuropathological changes detected in PD patients. These changes were first obvious at PND 60, although no motor changes were recorded until testing at PND 150. This study adds a new potential pesticide model of PD to the repertoire currently utilized for research, and highlights an important consideration as early life exposure to some toxicants may pre-dispose individuals to pathogenic changes later in life, or initiate slow progressive changes which culminate in development of the clinical phenotype. Despite the demonstration of the neurotoxicity of pyrethroids in experimental models, there appears to be insufficient human data to link pyrethroids exposure to PD (Costa, 2015).