Mitochondrial dysfunction and neurodegenerative proteinopathies: mechanisms and prospects for therapeutic intervention Open Access

The energetic requirements of neuronal excitability, synaptic activity and plasticity are extensive and are almost exclusively fulfilled by mitochondrial oxidative phosphorylation (OXPHOS) [1]. Mitochondria are often trafficked long distances to meet spatiotemporal adenosine triphosphate (ATP) requirements and dynamic mechanisms determine mitochondrial localisation to best satisfy local demands of the neurone. It is also now clear that, during the lifetime of a neurone, mitochondria can become dysfunctional to the extent they cannot maintain a proton motive force sufficient for ATP generation. Cellular mechanisms engage to remove damaged mitochondria and replenish the mitochondrial pool. These cycles of mitochondrial fusion and fission can become defective in neurodegenerative disorders and consequent accumulation of damaged components within mitochondria can adversely affect mitochondrial function and cellular homeostasis.

Alzheimer's disease (AD), Parkinson's disease (PD), amyotrophic lateral sclerosis (ALS) and frontotemporal dementia (FTD) are pathologically similar, progressive neurodegenerative proteinopathies [2] (Figure 1). Proteinopathy refers to a disease-causing conformational change in a protein that normally has other roles in cell biology. Such proteins undergo pathogenic misfolding and oligomerisation into higher-order structures, revealing self-templating conformations, and have an ability to undergo prion-like spreading between cells, together resulting in toxin gain- or loss-of-function [2,3]. Misfolded proteins can adversely affect mitochondria, either through a direct association, damaging mitochondrial DNA, altering trafficking and dynamics, deregulating bioenergetics and quality control pathways or promoting mitochondria-dependent cell death pathways. Mitochondrial dysfunction as a cause or a consequence of neurodegenerative disease pathogenesis is still debated and a self-perpetuating feed-forward toxic cycle may exist (Figure 2).

AD is the most prevalent proteinopathy, characterised by accumulation of extracellular plaques containing amyloid β (Aβ) and intracellular tangles of tau. Aβ and tau both have detrimental effects on mitochondria [4]. Disruptions in mitochondrial DNA maintenance, protein import, electron transport chain (ETC) activity and redox balance are all consequences of Aβ-induced toxicity [5–12] (Figure 3 and Table 1). Mitochondrial localisation of Aβ peptide has been observed, albeit the exact sub-mitochondrial topology remains less well defined [13,14] and, indeed, Aβ may be produced at the mitochondrial-associated membrane (MAM) [15], linking mitochondrial–ER contact sites, Ca²⁺ handling and bioenergetics to Aβ toxicity.

Tau is a highly soluble, natively unfolded protein [16]. Tau hyperphosphorylation impairs its ability to bind and stabilise microtubules, and tau aggregation and intraneuronal filaments are common in the pathology of tauopathies (Figure 1). The influence of tau on mitochondrial dynamics and quality control is well documented [17–21] (Figure 3). N-terminal tau fragments are associated with mitochondrial functional changes and defects in mitochondrial quality control, and the accumulation of tau fragments in human mitochondria isolated from synaptosomes correlates with synaptic changes observed in AD [17,22,23]. Changes in mitochondrial dynamics are also observed following tau overexpression in cultured cells or in in vivo models of tauopathy [18,24,25] (Table 1).

Some literature suggests that the accumulation of (phospho-)tau and Aβ is a direct consequence of mitochondrial dysfunction [26–30]. Perturbations to mitochondrial proteases and chaperones have demonstrated relationships with AD disease markers. Missense mutations in the mitochondrial matrix peptidase, pitrilysin metalloprotease 1 (PITRM1), are associated with Aβ-positive deposits and a slowly progressing neurodegenerative phenotype [31]. In addition, decreased activity in pre-sequence protease (PreP) has been observed in the temporal lobe region of AD patients [32] and has been linked to Aβ processing [33]. Overexpression of heat shock chaperone, mortalin, alleviates Aβ-induced toxicity, while pharmacological inhibition or siRNA down-regulation of mortalin induces DRP-1 (dynamin-related protein 1)-mediated mitochondrial fission and potentiates Aβ-induced mitochondrial and cellular toxicity [34,35]. With respect to pathogenic tau, antioxidant treatment of sod2 nullizygous neonatal mice reverses phospho–tau accumulation, placing mitochondrial oxidative stress upstream of tau pathology [26]. Finally, mitochondrial dysfunction induced by loss of either prohibitin 2 (PHB2), a mitochondrial membrane scaffold protein, or the m-AAA protease subunit, AFG3L2, both results in tau hyperphosphorylation [27,36], potentially linking mitochondrial dysfunction and the cytoskeleton. Taken together, these observations link mitochondrial dynamics, quality control and function to accumulation and toxicity of Aβ and tau.

PD is characterised by loss of dopaminergic neurones within the substantia nigra pars compacta and reduced dopamine innovation to the striatum. Mitochondrial dysfunction is a prominent pathological feature of both sporadic and familial diseases [39–45], and many PD-causing genes have overt mitochondrial phenotypes [44–53]. Accumulation of insoluble α-synuclein (α-syn) is a common feature of many clinical phenotypes, known collectively as synucleinopathies [54] (Figure 1).

Mitochondrial localisation of α-syn negatively affects mitochondrial function, morphology and dynamics [46,55–60] (Figure 3 and Table 2). Constitutive import of α-syn to mitochondria is transmembrane potential-dependent and is facilitated through a cryptic mitochondrial targeting sequence within the N-terminal region [57]. α-Syn associates with the mitochondrial inner membrane where a direct interaction and toxicity towards mitochondrial complex I has been observed [46,57]. Oligomeric and dopamine-modified α-syn-dependent reduction in protein import occurs via disruption of the association between translocase of the outer mitochondrial membrane (TOMM)-20 and its co-receptor, TOMM22 [61]. As a result, diminished protein import in nigrostriatal neurones impairs mitochondrial function, decreasing respiration and transmembrane potential, and increasing mitochondrial reactive oxygen species (ROS) [61]. Purification of crude mitochondrial preparations has led to the hypothesis that α-syn is, in fact, localised to the MAM [58] and α-syn point mutations reduce mitochondria–ER contacts, causing mitochondrial fragmentation [58]. Finally, α-syn has significant effects on mitochondrial quality control and its accumulation is becoming recognised as a consequence of deficient mitophagy [50,62,63]. α-Syn competes with LC3 for cardiolipin on the outer mitochondrial membrane. Cardiolipin facilitates refolding of α-syn oligomers; however, following prolonged exposure, LC3 is recruited. Mutated α-Syn is less able to compete with LC3, contributing to increased mitophagic flux observed in disease cells [64].

ALS and FTD share pathological and genetic similarities and potentially common neurodegenerative pathways [68]. Aggregated transactive response DNA-binding protein 43 kDa (TDP-43) and fused in sarcoma (FUS) are pathological hallmarks of both ALS and FTD (Figure 1). Both are ribonuclear proteins and contain prion-like domains, rich in glycine molecules, increasing their propensity for aggregation and cell-to-cell transmission. Dysfunction in OXPHOS, Ca²⁺ handling and ROS have all been proposed as key mitochondrially-associated determinants of ALS pathogenesis [69] (Table 3). Furthermore, mitochondrial trafficking defects are responsible for accumulation of defective mitochondria around cell bodies in motor neurones [69].

Ubiquitin-positive aggregates are observed in aged, mutant FUS-expressing transgenic animals and correlate with neuronal loss. Aggregates are also positive for mitochondrial cytochrome C oxidase (COX-IV), suggesting that defective mitochondria may be tagged for removal through the mitophagic machinery [70]. Similar pathology has been observed in a single post-mortem analysis of an FUS mutation carrier [71]. C- and N-terminal fragments of TDP-43 have been identified within mitochondria in Amyloid precursor protein (APP)/PS1 mice and mitochondrial dynamic changes, including trafficking and quality control defects, organelle redistribution and clustering within cytoplasmic inclusions, as well as morphological and ultrastructural alterations, are observed in animal models of TDP-43 pathology [72–74]. Taken together, these observations suggest a phenotype of dysfunctional, mislocalised and fragmented mitochondria in ALS and FTD (Figure 3).

Despite evidence of mitochondrial dysfunction in the pathology of proteinopathies and exacerbation of neurodegenerative disorders, the exact biochemical, neurotoxic mechanisms of misfolded, aggregated proteins remain poorly understood. Protecting mitochondrial function therefore may be one plausible drug discovery strategy for neuroprotective or disease-modifying end-points.

Mitochondrial permeability transition pore (mPTP) opening has been implicated as a major cell death pathway in multiple neurodegenerative diseases [76–78]. A shift in the mitochondrial redox balance towards oxidative stress, coupled with Ca²⁺ overload, triggers opening of the mPTP leading to osmotic swelling, uncoupling of electron transport and metabolic collapse [79–86]. The mitochondrial matrix enzyme, cyclophilin D (CypD), is a known positive regulator of mPTP opening [86]. Genetic ablation or pharmacological inhibition of CypD desensitises the pore to Ca²⁺, restricting pore opening [85,86]. Direct binding between Aβ and CypD links amyloid toxicity to mPTP opening in AD [38,87] and CypD deficiency corrects mitochondrial trafficking defects observed in AD models [88]. Previous literature supports a role for the F₁F₀-ATP synthase in pore formation [89,90], suggesting that the oligomycin-sensitivity conferring protein (OSCP) serves as a docking site for CypD [90]. Interestingly, OSCP is decreased during AD progression and may directly interact with Aβ [37]. Given the relationship between CypD, OSCP, Aβ and the propensity for mPTP opening, it is plausible that targeting these processes may have clinical benefits. Indeed, recently, phenotypic screening approaches have identified mPTP inhibitors and CypD-binding compounds in a model of Aβ-induced mPTP opening [91–94]. Overexpression of an N-terminal region of α-syn has also been observed to regulate mitochondrial membrane permeability [67], linking mPTP to synucleinopathies. Moreover, following overexpression, α-syn associates with the adenine nucleotide translocase (ANT), another putative pore component [95,96]. Interestingly, pharmacological inhibition of ANT partially reverses the associated α-syn-induced mitochondrial toxicity [97].

Homology within the cyclophilin isoenzyme family makes selective targeting of CypD therapeutically challenging [98]. Moreover, since CypD does not constitute a principal pore component, and effects are indirect, mitochondria remain capable of permeability transition given enough stimuli [86,99]. CypD confers sensitivity to the mPTP inhibitor, cyclosporin A (CsA) [85,100], and number of CsA analogues have been developed [101,102]. CsA and its derivatives are large molecular mass natural products and penetrate the blood–brain barrier poorly, limiting efficacy in neurodegenerative disease. Many groups have developed CypD-independent inhibitors [103–107], but to date, none, as far as we are aware, have been tested in models of neurodegenerative disease.

Dysfunctional lysosomal and proteasomal degradation pathways have been implicated in neurodegenerative diseases. A selective form of macroautophagy, termed mitophagy, is responsible for the clearance of defective mitochondria from cells [44,45,108–110]. PTEN-induced putative kinase 1 (PINK1) and parkin are regulators of mitophagy and are integral to a mechanism that identifies and tags defective mitochondria for removal [110,111] (Figure 4). The association between mutations in these proteins and dysfunction in the mitophagy pathway has direct implications in both familial and sporadic PD [43,51,112].

PINK1 and parkin function may be necessary for α-syn clearance. Preceding neurodegeneration, α-syn A53T transgenic mice accumulate neuronal inclusions containing mitochondrial remnants and autophagic markers which increase in size and number with PINK1 or parkin knockout [62]. PINK1 loss-of-function potentiates the A53T phenotype, decreasing lifespan and enhancing movement deficits and protein aggregation [63]. Similarly, iPSCs from mutant PINK1/parkin carriers accumulate cytoplasmic inclusions and insoluble α-syn, a phenotype which can be partially corrected following PINK1 re-expression [50]. Interestingly, following mitochondrial uncoupling, autophagic α-syn removal is reduced and the likelihood of aggregate formation in oligodendrocytes is enhanced, suggesting that mitochondrial damage over time may play a role in α-syn accumulation [48]. Finally, a novel mechanism of PINK1 protection in α-syn models has been proposed as being mediated through protein phosphatase 2A activity [113].

Mitophagy has been linked to both Aβ and tau in AD. PINK1 is down-regulated in AD patients and in transgenic AD models [114]. Furthermore, the absence of PINK1 augments the mutant APP phenotype and stereotaxic injection of rAAV2–PINK1 into the hippocampus of mutant APP mice significantly reduced Aβ compared with control, improving synaptic function and memory [114]. An association between deficient mitophagy and abnormal tau accumulation has been found in AD patient brain homogenate and transgenic mice. This deficit can be rescued by up-regulating parkin expression [20]. Exogenous parkin has also been found to decrease Aβ levels in vitro and Aβ-induced plaque formation in transgenic mice [115–117].

Parkin activation may be a promising strategy to enhance mitophagy in disease models. Nilotinib, originally discovered as a tyrosine kinase inhibitor, increases parkin abundance and ubiquitination, potentially increasing parkin recycling via the proteasome [118]. Nilotinib-mediated c-ABL inhibition also prevents parkin tyrosine phosphorylation, resulting in release of parkin auto-inhibition and demonstrating protection in PD models [119]. Additionally, nilotinib has been demonstrated to increase the parkin–beclin 1 interaction and increase clearance of Aβ in transgenic APP mice following chronic treatment [120]. Finally, with respect to ALS and FTD, motor and cognitive deficits measured in TDP-43 transgenic mice have also been reversed using nilotinib [121].

Down-regulation of the ubiquitin–proteasome system (UPS) is common across neurodegenerative diseases and promoting UPS activity is an emerging strategy for the treatment of proteinopathies. Deubiquitinases (DUBs) hydrolyse isopeptide bonds covalently binding ubiquitin to proteins, regulating degradation, localisation or activity. Multiple DUBs regulate mitochondrial function [122–126]. Ubiquitin-specific protease 15 (USP15) and USP30 both antagonise parkin-mediated mitophagy [122,125]. USP30 is the only DUB exclusively localised to mitochondria [127], tethered to the outer mitochondrial membrane. USP30 deubiquitinates parkin substrates, including TOMM20 and MIRO1 [122], and inhibition has been proposed to enhance parkin-mediated mitophagy [128,129] (Figure 4). In one study, USP35 was also found to oppose parkin-mediated mitophagy, with both a distinct mechanism to USP30 and lack of influence on the translocation of parkin [124]. USP8 enhances mitophagy by removing lysine-6-linked ubiquitin from parkin, promoting its turnover [126]. However, confusingly, USP8 knockout also limits toxicity in an α-syn model in Drosophila melanogaster [130]. Other DUBs localise to mitochondria, albeit not exclusively; Ataxin-3 [131,132] and the X-chromosome-linked deubiquitinase, USP9x [132] have demonstrated mitochondrial localisation under specific conditions. USP9x deubiquitinates α-syn and silencing increases the abundance of mono-ubiquitinated α-syn, enhancing its propensity for aggregation [133].

Mitochondria are a principal source of cellular ROS. ROS are generated as a by-product of OXPHOS and their abundance presents a fine balance between signalling and toxicity. Much interest has focussed around limiting oxidative stress in neurodegenerative disease. Exogenous expression of a mitochondrially targeted catalase decreases monomeric and oligomeric Aβ and Aβ plaques in mice carrying the APP KM670/671NL (Swedish) mutation [134]. Synthetic analogues of mitochondrial coenzyme Q₁₀ prevent Aβ oligomer-induced changes in mitochondrial mRNA transcript expression, protecting cells against oligomeric Aβ damage [135]. Although perturbation of ROS in preclinical models has so far proved beneficial, to date translation to human disease has been challenging and yielded multiple clinical failures across multiple neurodegenerative diseases [136]. Interestingly, the antioxidant MitoQ has been assessed in many models of ageing and neurodegenerative disease. MitoQ is a redox active ubiquinone, targeted to mitochondria [137]. MitoQ has demonstrated positive effects in an SOD1^G93A ALS mouse model [138], a triple transgenic AD mouse [139] and in models of AD in Caenorhabditis elegans [140], together linking mitochondrial ROS and proteinopathy-related neuropathologies. MitoQ is currently in clinical trials testing the efficacy for improving vascular, motor and cognitive function in middle-aged and older adults (NCT02597023).

Compelling evidence suggests that mitochondrial dysfunction plays a significant role in neurodegenerative proteinopathies. Neuronal ATP is provided almost exclusively through mitochondrial OXPHOS, and complicated processes controlling mitochondrial dynamics, redox equilibrium, protein import and mitochondrial quality control work in concert to meet spatiotemporal bioenergetic demands. Aberrant misfolded proteins disrupt these processes, triggering mitochondrial dysfunction and having wider effects on cellular homeostasis.

Numerous disease-modifying strategies targeting mitochondria are currently under investigation. Further development of mPTP inhibitors is warranted due to the emerging evidence of the involvement of Ca²⁺ homeostasis, ROS and mPTP opening in multiple neurodegenerative diseases. Accelerating removal of damaged mitochondria has been proposed as a novel disease-modifying strategy not only in PD, but in many proteinopathies. Identification of a mechanism to enhance mitophagy may demand increased understanding of DUB biology and the substrate diversity and selectivity of these enzymes. Clinical trials of mitochondrially targeted antioxidants will provide proof-of-concept concerning ROS manipulation. Taken together, thoroughly understanding the mitochondrial relationship with neurodegenerative proteinopathies is likely to pave the way for the development of targeted therapies, potentially modifying the disease course of these progressive degenerative disorders.

AD

Alzheimer's disease

ALS

amyotrophic lateral sclerosis

ANT

adenine nucleotide translocase

APP

amyloid precursor protein

ATP

adenosine triphosphate

Aβ

amyloid β

COX

cytochrome C oxidase

CsA

cyclosporin A

CypD

cyclophilin D

DRP-1

dynamin-related protein 1

DUB

deubiquitinase

ETC

electron transport chain

fis1

mitochondrial fission 1

FTD

frontotemporal dementia

FUS

fused in sarcoma

MAM

mitochondrial-associated membrane

mPTP

mitochondrial permeability transition pore

OPA-1

optic atrophy 1

OSCP

oligomycin-sensitivity conferring protein

OXPHOS

oxidative phosphorylation

PD

Parkinson's disease

PINK1

PTEN-induced putative kinase 1

ROS

reactive oxygen species

TDP-43

transactive response DNA-binding protein 43 kDa

TOMM

translocase of the outer mitochondrial membrane

UPS

ubiquitin–proteasome system

USP

ubiquitin-specific protease

α-syn

α-synuclein

T.B. and A.R.H. wrote the manuscript.

The authors thank Jim Staddon for helpful comments and advice in the course of manuscript preparation.

T.B. and A.R.H. are employees of Eisai Ltd.