Ubiquitin and Receptor Dependent Mitophagy Pathways and Their Implication in Neurodegeneration

Abstract

Selective autophagy of mitochondria, or mitophagy, refers to the specific removal and degradation of damaged or surplus mitochondria via targeting to the lysosome for destruction. Disruptions in this homeostatic process may contribute to disease. The identification of diverse mitophagic pathways and how selectivity for each of these pathways is conferred is just beginning to be understood. The removal of both damaged and healthy mitochondria under disease and physiological conditions is controlled by either ubiquitin-dependent or receptor-dependent mechanisms. In this review, we will discuss the known types of mitophagy observed in mammals, recent findings related to PINK1/Parkin-mediated mitophagy (which is the most well-studied form of mitophagy), discuss the implications of defective mitophagy to neurodegenerative processes, and unanswered questions inspiring future research that would enhance our understanding of mitochondrial quality control.

Introduction

Mitochondria are important for numerous cellular functions including generation of ATP, regulation of apoptosis, calcium handing, generation and detoxification of reactive oxygen species (ROS), regulation of iron-sulfur proteins, and lipid metabolism [1]. Macroautophagy, hereafter referred to as autophagy, removes cytosolic components or organelles via sequestration into double membrane vesicles called autophagosomes for delivery to the lysosome for enzymatic degradation [2]. Specific targeting of mitochondria for autophagic removal was first described after a series of observations. First, upon stimulation of autophagy, mitochondria and mitochondrial-resident enzymes were specifically digested after fusion of autophagosomes to lysosomes [3–5]. Further studies observed that upon upregulation of autophagy due to starvation, depolarized mitochondria were selectively engulfed by autophagosomes, followed by fusion with and degradation in lysosomes [6, 7]. Collectively, these early observations and others led J.J. Lemasters to coin the term “mitophagy” [8]. Mitophagy is now considered a major quality control mechanism vital for maintenance of the health of the mitochondrial network and homeostasis, balancing the number of mitochondria within the cell.

Ubiquitin-dependent mitophagy

Pink1/Parkin

Parkinson’s disease (PD) is the most common age-related motor-deteriorating neurodegenerative disease worldwide, affecting more than 4 million people [9]. PD is characterized by the preferential degeneration of dopaminergic (DA) neurons of the substantia nigra [10, 11]. Although the majority of cases are sporadic, mutations in PINK1 (PTEN–induced putative kinase protein 1), a mitochondrial targeted serine/threonine kinase, and Parkin, an E3 ubiquitin ligase, are known to cause autosomal recessive forms of the disease [12–15]. Epistasis experiments in Drosophila established that PINK1 and Parkin act within the same pathway and placing PINK1 function upstream of Parkin [16–18]. Over the last decade, the extraordinary efforts made to elucidate the functions of these two proteins has previously been reviewed extensively [19–21]. Therefore, after a brief overview of PINK1/Parkin-mediated mitophagy (Figure 1), we will focus on the most recent findings examining the PINK1/Parkin pathway.

Figure 1: PINK1/Parkin as a selective UB mediated mitophagy pathway.

Cartoon depiction of PINK1 and Parkin working in conjunction to tag the OMM proteins with ubiquitin. PINK1 phosphorylates ubiquitin and Parkin. This causes the translocation and activation of Parkin from the cytosol to the damaged mitochondria. Increased UB chains on OMM proteins allow autophagy receptor binding and recruitment of TBK1 to phosphorylate receptors for autophagosome recruitment.

Youle and colleagues first described Parkin translocation to damaged mitochondria upon loss of membrane potential, promoting their removal [22]. Under steady-state conditions, Parkin resides in an inactive conformation in the cytosol [23–25], while PINK1 is constitutively turned over by the proteasome after being imported, cleaved, and released from polarized mitochondria [26, 27]. Mitochondrial damage resulting in disruption of mitochondrial membrane potential or other types of mitochondrial insult such as proteotoxic stress or ROS-induced damage impair PINK1 import and cleavage in the inner mitochondrial membrane, allowing for its accumulation on the outer mitochondrial membrane (OMM) surface where its kinase domain is presented to the cytosol [28–33]. Stabilized PINK1 recruits and activates Parkin to the damaged mitochondria by phosphorylating Parkin in its ubiquitin-like domain (UBL) [34, 35] and by phosphorylating free ubiquitin and ubiquitin chains on resident OMM proteins [36–39]. Parkin activation results in a cascade of ubiquitination of resident proteins on the OMM [40–43]. This increase in the ubiquitin signal on the surface of the mitochondria recruits the ubiquitin-proteasome system (UPS) to degrade OMM proteins [41, 43] as well as autophagy receptor proteins ultimately resulting in development of the autophagophore, which encapsulates the damaged mitochondria prior to fusion with the lysosome [44–47].

PINK1 Accumulation and Regulation

When stabilized at the OMM, PINK1 forms a high molecular weight complex shown to include components of the translocase of the outer membrane (TOM) complex [48–50]. PINK1’s association with the TOM complex and regulation at damaged mitochondria has recently been further clarified. TOM7 was identified in an siRNA screen looking for modulators of Parkin translocation, and this was shown to occur through an inability to promote PINK1 accumulation at the OMM in the absence of TOM7 [51]. Without TOM7, a subunit of the TOM40 complex that allows for efficient import of proteins into mitochondria [52]. PINK1 enters mitochondria even in depolarizing conditions and is degraded by the mitochondrial protease, OMA1 [53]. Notably, the loss of TOM7 may not generally inhibit mitochondrial import as tested substrates appear to be imported normally [51], indicating specialization in control of PINK1/Parkin-dependent mitophagy. There is some speculation that this may allow regulated reversal of mitophagy if mitochondria are able to recover and quickly repolarize [54, 55]; however, whether this occurs remains to be determined.

Human ATG8 family members

Microtubule-associated protein light chain 3 (MAP1LC3A, MAP1LC3B, MAP1LC3C) and gamma-aminobutyric acid receptor-associated proteins (GABARAP, GABARAPL1, GABARAPL2) are mammalian homologues of yeast ATG8. These small ubiquitin-like proteins are conjugated to phosphatidylethanolamine (PE) and targeted to autophagic membranes (autophagophore and autophagosome) [56]. Upon activation of Parkin-mediated mitophagy, all of the LC3 isoforms and GABARAPs are present at the damaged mitochondria [57], though not required for autophagosome formation. However, GABARAPs were shown to be essential for autophagosome elongation and fusion to lysosome [58, 59]. Additional factors recruited to damaged mitochondria include: the VSP34 complex with DFCP1 and WIPIs (to produce phosphatidylinositol-3-phosphate (PtdIns3P) on donor membranes), the ULK1 complex to initiate and expand the autophagophore, and other ATG8 proteins such as ATG16L1, ATG14. ATG12, ATG5, and ATG3 to grow the autophagosomal membrane, which has recently been reviewed [60]. These data suggested that autophagosome formation occurs at the site of damaged mitochondria destined to undergo Parkin-mediated mitophagy rather than via the recruitment of existing autophagosomal membranes [46, 61]. Although selective isoforms of LC3 and GABARAPs were found to participate in other forms mitophagy (see below), the LC3/GABARAP knockout lines used to study Parkin-mediated mitophagy have not been used for evaluation in receptor-dependent mitophagy, in which mitochondrial-membrane associated/embedded proteins provide the “eat me” signal.

Additional explorations of autophagosome biogenesis have been performed for Parkin-mediated mitophagy. Mitochondrial fission protein 1 (Fis1) was originally described as a mitochondrial fission protein in mammals; however, this claim has since been refuted [62, 63]. Upon examination of the loss of Fis1 in C. elegans, mitochondrial fission appeared unaffected, yet enlarged aggregated LC3 puncta formed upon induction of mitophagy, though not starvation-induced autophagy [64]. After confirmation of this phenotype in mammalian Fis1 KO cells, it was further shown that TBC1D15 and TBC1D17 bind Fis1 at the mitochondria and interact with ATG8 proteins via LC3-interacting region (LIR) domains to influence autophagosome formation [65]. TBC1D15/17 are Rab GTPase activating proteins [66] that were shown to constrain RAB7A’s activity at the mitochondria during mitophagy. Loss of TBC1D15/TBC1D17 mimicked Fis1 KO cells [65]; modulation of RAB7A activity can occur either by binding to GTP/GDP or via phosphorylation, promoting recruitment of ATG9A vesicles which likely facilitate autophagophore expansion [67, 68]. ATG9A is a transmembrane ATG protein is essential for autophagophore formation during mitophagy [61, 67, 69].

Autophagy Receptor Proteins

Autophagy receptor proteins have an ATG8/LC3 interacting domain termed the LC3 interacting region (LIR) motif, which promotes attachment to the growing phagophore, and a ubiquitin binding domain (UBD) or zinc finger (ZF) domain which contributes to cargo selectively via binding to ubiquitinated proteins [70]. An analysis of five autophagy receptors (NDP52, TAX1BP1, p62, NBR1, and optineurin (OPTN)) revealed partial redundancy of receptor proteins and an increased reliance on OPTN, NDP52, and to a lesser extent, TAX1BP1, for Parkin-mediated mitophagy [45–47, 71]. Although other receptor proteins, such as p62 and NBR1, are present at damaged mitochondria, they are not required for the progression of mitophagy [44, 46, 47]. However, p62 does cause mitochondrial clustering in response to mitophagy, a role suggested to promote efficiency by collecting mitochondria targeted for removal at localized sites [72, 73].

Human mutations in autophagy receptor proteins may provide insight into how defective mitophagy contributes to disease. Optineurin mutations have been identified in amyotrophic lateral sclerosis (ALS) [74] and glaucoma [75], and NDP52 mutations in Crohn’s disease [76]. In mammalian cell lines, ALS-associated optineurin mutations cause defective mitophagy [45, 46], but these mutations have not yet been thoroughly evaluated in neurons. Receptor protein redundancy, differences in cell or tissue expression patterns of the receptors, or their additional roles in other pathways, such as innate immunity [77], may account for these observations.

During mitophagy, PINK1 and Parkin also trigger the translocation and activation of Tank Binding Kinase 1 (TBK1) to damaged mitochondria [44, 46, 47, 71]. TBK1 is best known for its role in innate immune signaling and is a key regulator of the interferon response [78]. Upon activation, TBK1 dimerizes and trans-autophosphorylates at S172 [79, 80]. At the mitochondria, TBK1 phosphorylates autophagy receptors, enhancing their ability to bind and promote association with the autophagosome developing around the damaged mitochondria [44, 46, 47, 71]. The loss of TBK1 impairs mitophagy [47, 71], while TBK1 artificially tethered to mitochondria can drive mitophagy independently of PINK1/Parkin [81]. Loss-of-function mutations contribute to the neurodegenerative diseases, ALS, and frontal temporal dementia (FTD) [82, 83]; however, it remains to be examined whether TBK1 patient mutations impact mitophagy in neurons or if loss of TBK1 influences other processes known to contribute to neurodegeneration, such as inflammatory signaling [84, 85].

Ubiquitin-specific proteases (USPs)

Ubiquitin-specific protease 30 (USP30) is a ubiquitin deubiquitinating (Dub) enzyme located on the outer mitochondrial membrane [86, 87]. USP30 modulates the PINK1/Parkin pathway in neurons by deubiquitinating OMM proteins and Parkin itself to moderate mitophagy [88]. USP30 preferentially cleaves K6 and to a lesser extent K11, K48, and K63 ubiquitin (Ub) species, either as free chains in vitro or as chains anchored on mitochondrial proteins [89–91]. USP30 depletion does not affect basal mitophagy activity in the absence of PINK1, suggesting that USP30 may act upstream or independently of PINK1 activation [92, 93]. Furthermore, USP30 activity is impaired when ubiquitin is phosphorylated (p-Ub), supporting these findings [90, 91]. The potential for modulation of mitophagy via fine-tuning of baseline ubiquitin priming is an attractive therapeutic target in attempts to boost in vivo mitophagic activity in disease states.

Additional USPs which modulate Parkin-mediated mitophagy have been identified. Overexpression of USP15 reduced mitophagy, while knockdown (KD) potentiated mitochondrial clearance [94], consistent with results observed upon knockdown of the closest USP15 orthologue in Drosophila, which rescued mitochondrial defects and enhanced mitophagy in PINK1 and Parkin mutant flies [94, 95]. However, USP15 isn’t tethered to mitochondria [87, 96], therefore, its mechanism of action remains unclear. Another Dub, USP8, modulates Parkin itself, with data suggesting that removal of K6-linked ubiquitin chains promotes the stability of Parkin [97]. However, conflicting data shows that USP8 KD or pharmacological inhibition rescues PINK1 mutant flies through reversion of mitochondrial defects by modulation of mitofusin 2 (MFN2) levels [98]. It is unclear if mitophagy was affected in these flies after USP8 KD or if the rescue in PINK1 mutant flies was mainly dependent on mitochondrial dynamics.

An alternative splice variant of another Dub, USP35, which contains a recognizable mitochondrial targeting signal (MTS) appears to localize to mitochondria under certain conditions and may negatively regulate Parkin-mediated mitophagy [99]. Additional USPs have been identified in siRNA Parkin translocation screens, but were not validated [51]. This leaves open the question of how many other Dubs might possibly translocate to mitochondria, mediate OMM Ub homoeostasis, or regulate the stability or activation of key mitophagic proteins.

Other ubiquitin-dependent forms of mitophagy

Mul1

Mul1 (Mitochondrial Ubiquitin Ligase 1)/MAPL (Mitochondrial-anchored Protein Ligase)/MULAN (Mitochondrial Ubiquitin Ligase Activator of NF-κB) was discovered to reside on the outer mitochondrial membrane via two transmembrane domains while harboring a cytosolic-facing RING finger domain required for its E3 ligase activity [100–102]. Mul1 may also participate in the SUMOylation of outer mitochondrial membrane proteins such as the fission-associated GTPase and dynamin-related protein 1 (Drp1), thus influencing mitochondrial dynamics and cell death [103, 104]. More recently, work on Mul1 has shown that its E3 ubiquitin ligase activity also activates ubiquitin-dependent mitophagy [105, 106].

Genetic interaction experiments in Drosophila confirmed that Mul1 acts in parallel to the PINK1/Parkin pathway; Mul1 deficiency exacerbated thorax indentation phenotypes and mitochondrial morphology defects PINK1 or Parkin mutant flies [106]. By forcing cells to utilize oxidative phosphorylation (OXPHOS) in culture, Rojanksy et al. showed that Mul1 works in parallel with Parkin to trigger mitophagy and can also be responsible for ubiquitination of outer mitochondrial membrane (OMM) proteins [105] (Figure 1). Exact identification of outer mitochondrial membrane protein targets of Mul1 identified to date include SLC25A46 [107] and MFN2 [106]. Determination of the types of ubiquitin chains formed and identification of additional substrates are of great interest to better define the selectively and function of Mul1 in mitophagy. Additional research on the types of Ub chains Mul1 produces on the OMM and its substrates may also reveal receptor protein binding preferences and provide insight into autophagy receptor regulation.

Thus far, it is unclear how Mul1 itself is regulated and how that regulation affects mitophagy. Mul1 steady state levels were shown to be increased in Omi/Htra2 KO cells [108]. Omi/Htra2 is a mitochondrial inner membrane space (IMS) serine protease [109] which may degrade portions of Mul1 that extend into the IMS [108]. Another study reported elevations in Mul1 levels in vacuolar protein sorting-35 (VPS35) heterozygous mice with concomitant data suggesting that VPS35 may promote Mul1 turnover via proteasome-mediated degradation [110]. Increased Parkin and Mul1 expression has previously been linked to apoptosis [102, 107, 111, 112]; however, Mul1 overexpression has also been shown to be neuroprotective after exposure to rotenone, a complex I inhibitor [113]. It is unclear how Mul1 turnover and stability is regulated under different conditions, however, it is likely the physiological functions of mitochondrial-targeted E3 ligases may vary depending on the state of the cell.

Mitochondrial derived vesicles (MDVs)

Due to the dynamic nature of mitochondria, it is difficult to estimate their size and shape at any given time [114]. Determination of mitochondrial morphology by electron microscopy is also highly dependent on the given cell type [115, 116], but the question of whether whole or fragments/portions of mitochondria are selectively targeted for degradation is an area of intense interest. Mitochondrial derived vesicles (MDVs) are discrete vesicles which bud off from the “whole” mitochondria in a distinct type of mitochondrial quality control. MDVs are visible by electron microscopy, approximately 70–100 nm in diameter, are initiated by oxidative stress, exhibit a unique repertoire of OMM protein and inner mitochondrial membrane (IMM)/matrix identifiers and composition (single or double membranes), and are Drp-1 independent [100, 117–119]. The final destination of MDVs within the cell is also distinct from “whole” mitophagy with data suggesting that they are targeted to peroxisomes [100, 117], lysosomes [118], endosomes [120], and phagosomes [121]. A fraction of these MDVs requires PINK1 and Parkin; however, MDV formation is not completely abolished by loss of either [122]. Other proteins that control starvation-induced autophagy, such as ATG5, Beclin-1, and RAB9 that promote autophagophore formation, do not appear to be required [118, 122]. Syntaxin-17 and its associated SNARE proteins SNAP29 and VAMP7 target a select PINK1/Parkin-dependent MDV pool to the lysosome for degradation [119].

It is unclear if every cell type has the ability to form MDVs or if the physiological outcome upon upregulation of this pathway results in distinct phenotypes in addition to its proposed role in organelle quality control. One function of MDVs is to drive adaptive immune recognition of mitochondrial antigens [120]. Another recent paper showed potential MDV involvement in the delivery of superoxide dismutase 2 (SOD2) to phagosomes to aid in the degradation of methicillin-resistant Staphylococcus aureus; physical delivery of this ROS-detoxifying enzyme to the phagophore allows local production of bacteria-lethal hydrogen peroxide [121]. These reports indicate that MDV formation may be important for macrophage polarization in response to diverse pathogenic challenges. MDVs have also been artificially derived in cardiomyocytes and liver [119, 123], so their role in these tissues could very well differ and are yet to be elucidated.

Receptor-dependent mitophagy

Mitophagy is observed in yeast during cellular remodeling processes that occur upon shifting to different carbon sources [124]. Autophagy-related gene 32 (ATG32) is a mitophagy receptor located in the outer mitochondrial membrane which interacts with ATG11 and ATG8 via a WXXL motif in its cytoplasmic domain to facilitate mitophagy [125, 126]. Phosphorylation events at Ser114 and Ser119 on ATG32 enhance its interaction with the scaffold protein ATG11, which recruits fission machinery [127]. It is important to note that in yeast, the role of mitochondrial fission appears to be more important in the progression of mitophagy than in mammalian cells [33, 128]. While mammals do not have ATG32, a number of potential ATG32 orthologues, each containing a WXXL-like LIR motif at the N-terminus, have been identified. This interaction between the LIR and LC3 or GABARAP proteins in mammals mediates the formation of an autophagosome surrounding the selected mitochondrion.

BNIP3/NIX (BNIP3L)

Receptor-mediated mitophagy is observed during other types of stimuli in addition to mitochondrial damage, such as hypoxia and cellular differentiation. B-cell leukemia/lymphoma 2 (BCL-2)/adenovirus E1B 19 kDa interacting protein 3 (BNIP3) was first identified in a yeast two-hybrid screen for interactors of adenovirus E1B-19 KD [129]. Nip-like protein X (Nix, also known as BNIP3L) was identified through the cloning of a cDNA library and found to have approximately 50% homology to BNIP3 [130]. Both of these multifunctional autophagy receptors were found to be located on the OMM, contain an atypical BCL2-homology 3 (BH3) domain, and have a primarily cytoplasmic N-terminus [129, 131]. Early work on these proteins focused on their roles in apoptosis and showed that both BNIP3 and Nix interact with the anti-apoptotic proteins Bcl-2 and Bcl-X_L [129, 131–134].

Nix is a receptor for mitophagy triggered upon reticulocyte differentiation and mitochondrial damage [135] and is mediated by the direct binding of Nix’s LIR domain to either LC3A/B or GABARAP1L [135–137] (Figure 2A). The NIX-LC3 interaction is relatively weak during normal growth conditions, but when mitochondria were stressed with rotenone or carbonyl cyanide m-chlorophenylhydrazone (CCCP), the NIX and LC3 interaction increased, indicating a possible mechanism for mitophagy regulation [135]. LC3B is normally a weak interacting partner with Nix; however, inclusion of phosphomimetic residues upstream of the LIR do increase binding [137]. An unidentified kinase may be involved during mitochondrial damage to account for this increased binding. Nonetheless, it is still untested whether the sites upstream of the NIX LIR motif are truly phosphorylated.

Figure 2: BNIP3/Nix and FUNDC1 are receptor dependent mitophagy pathways.

Cartoon depiction of BNIP3/Nix-mediated (A) and FUNDC1-mediated (B) mitophagy. (A) BNIP3 and Nix form a dimer; however, it remains unclear if they dimerize upon activation of mitophagy. BNIP3 and Nix are anchored on the OMM and interact with hATM8 family members. Dashed circle phosphorylation sites denote potential important post-translational amino acids by site-directed mutagenesis. Dashed monomers: Nix forms dimers with transmembrane domain; however, in vitro the transmembrane (TM) domain is not necessary for hATG8 binding. While BNIP3 dimers have been implicated in cell death, they have not been shown to explicitly regulate mitophagy. (B) In normoxia, the cytoplasmic portion of FUNDC1 is phosphorylated. In response to hypoxia, FUNDC1 is dephosphorylated, resulting in enhanced hATG8 family member binding and recruitment to initiate autophagosome formation.

There is some indication that Nix plays a role in neuronal differentiation and function. Nix expression increases during retinal ganglion cell (RGC) differentiation, and Nix KO mice show abnormalities in RGCs in vivo [138]. Mice harboring mutations in the lysosomal enzyme glucosylceramidase beta (GBA) show altered protein levels of Nix in mitochondrial fractions isolated from brain [139]. However, the role of Nix in neurons is understudied in comparison to its function and importance in reticulocyte differentiation.

Mature mammalian red blood cells lack mitochondria, though they are present in immature reticulocytes; thus, a key aspect of red blood cell development is the removal of undamaged mitochondria via lysosomal degradation [140]. Early studies demonstrated that Nix expression is increased during reticulocyte maturation [141], and Nix has subsequently been identified as a critical mediator of this process. Nix KO mice retain higher numbers of erythroid precursor cells with significantly increased mitochondrial mass compared to wildtype animals [142, 143]. Incubation of isolated mitochondria with recombinant Nix results in mitochondrial permeabilization and loss of membrane potential [144]. The addition of mitochondrial depolarizing agents rescues mitophagy in Nix KO reticulocytes [143], suggesting activation of a compensatory pathway upon depolarization. Strikingly, damaged mitochondria are retained in mature reticulocytes in mouse models that accumulate mtDNA mutations [145, 146]. It is unclear whether Nix expression, stabilization, or function are perturbed in these erythroid precursors where mitochondria are laden with high levels of mutant mtDNA and remain highly polarized. Mitochondrial disease patients have been described with different types of anemia [147], so it is possible that impairments in Nix-mediated mitophagy are a contributing factor to these phenotypes.

BNIP3 expression activates cell death [148–150]; however, its role in mitophagy has only recently been revealed [151–153]. Ceramide activation of BNIP3 led to autophagic cell death in malignant glioma cells which was the first indication of its involvement in autophagy [154]. Overexpression of BNIP3 induced extensive autophagy in which mitochondria were detected as cargo [152]. Upon activation by hypoxia [153] or CCCP [151], BNIP3 expression increases [148, 149], and dimerization leads to the recruitment of LC3B to promote selective mitochondrial removal. Furthermore, BNIP3 mediates the removal of mutated mtDNA in Drosophila oocytes suggesting a physiological role for BNIP3 in mitochondrial quality control during development [155].

FUNDC1

FUN14 Domain-Containing Protein 1 (FUNDC1) is an outer mitochondrial membrane protein containing a LIR motif which has been proposed to act as a mitophagy receptor [156]. During hypoxia, FUNCD1 is dephosphorylated, increasing its affinity to bind and retain LC3B at the mitochondria, resulting in engulfment of the targeted mitochondria by an autophagosome (Figure 2B) [156]. There have been other proposed key resides around or within the LIR domain that undergo phosphorylation or dephosphorylation in response to mitochondrial stimuli [156–158]. The generation of FUNDC1 KO mice has provided some level of insight for the in vivo requirements for this receptor - FUNDC1 KO mice are viable and grossly normal but did display decreased mitophagy in platelets when exposed to hypoxia, a process which may contribute to protection during ischemia reperfusion injury [159].

Cardiolipin/BCL2-L-13/Prohibitin 2

Additional ubiquitin independent mechanisms of mitophagy with identifiable receptors or receptor-like lipid recognition have been described. BCL2-L-13 was recently identified and put forth as the mammalian orthologue of yeast ATG32, though it has poor sequence homology [160]. Conflicting results found that KD of BCL2-L-13 increased basal mitophagy but inhibited mitophagy during iron chelation [161] (see below). More work is needed to truly evaluate BCL2-L-13 as a mitophagy receptor as opposed to its known roles in apoptosis [162] and to understand how depolarization activates BCL2-L-13-dependent mitophagy on a mechanistic level.

The lipid composition of the IMM has also been proposed to perform a receptor-like function for mitophagy [163]. Cardiolipin is a phospholipid important for normal mitochondrial function that is involved in cristate organization, respiration, and apoptosis [164]. Upon mitochondrial damage, the externalization of cardiolipin from the IMM to the OMM allows for the recognition and binding of the N-terminus of LC3 to cardiolipin for mitophagic progression in neurons [165]. In support of these findings, epistasis experiments in yeast demonstrated synthetic lethality when crossing cardiolipin-deficient mutants with autophagy or mitophagy mutant strains [166]. Decreased levels of tafazzin, a phospholipid transacylase, inhibits the maturation of cardiolipin and causes decreased encapsulation of mitochondria inside autophagosomes and mitolysosomes [167]. However, other data in yeast suggest that dysfunctional mitophagy due to cardiolipin deficiency may actually be due to alterations in upstream signaling pathways [166] or vacuolar defects [168].

Prohibitin 2 (PHB2) is another IMM protein [169] that forms a multimeric complex with Prohibitin1 (PHB1) [170–172]. Many different functions have been described for prohibitins [173], and it is clear that PHB1 and PHB2 stability and function are interdependent [169]. Similar to cardiolipin, PHB2 was recently proposed as a mitophagy receptor. After Parkin-mediated OMM rupture and UPS degradation of OMM proteins [43, 174], the LIR domain of PHB2 tethers LC3 to damaged mitochondria and also acts a mitophagy receptor in Parkin-independent mitophagy of paternal sperm in C. elegans [175]. It is unclear how these receptor-mediated types of mitochondrial clearance compete with one another for available LC3 and GABARAP pools. More information is needed to establish cell-type specificity and mechanisms that better define how different stimuli trigger each of the receptor-dependent mitophagy pathways.

Non-canonical mitophagy

Additional observations of mitophagy induction upon exposure to defined stimuli have been described, but it is yet unclear whether they rely on known regulators of mitophagy. These non-canonical forms of mitophagy can be broadly grouped into two categories, whether they are activated by: 1) mitochondrial stress or 2) a change in energy source.

Iron is necessary for heme and iron-sulfur cluster containing proteins necessary for proper maintenance of the electron transport chain [176]. The iron chelator deferiprone (DFP) triggers mitophagy independently of the generation of reactive oxygen species or PINK1 accumulation/Parkin activation [128, 177]. Iron-depleted conditions in yeast also trigger mitophagy dependent on ATG32 [178]. This indicates that iron chelation mitophagy in mammalian cells could also potentially be receptor dependent. Increased iron deposition and accumulation is observed in brain regions due to aging and neurodegeneration, and DFP has been used as a therapeutic agent in early stage clinical trials targeting neurodegenerative disease [179]. The extent of the effects of DFP on neurons and the physiological significance of this stimuli is unknown.

Increased use and reliance on mitochondria and oxidative metabolism also appear to trigger mitophagy [180]. As previously discussed, Parkin/Mul1-mediated mitophagy is activated in mammalian cells reliant on OXPHOS [105]. BNIP3L/Nix has may be important for stem cell differentiation as cells switch their reliance from glycolysis to oxidative metabolism [138]. There is also a greater occurrence in the generation of MDVs in cells utilizing OXPHOS as opposed to glycolysis after being subjected to oxidative stress [118]. All of these examples demonstrate that shifts in metabolism drive mitophagy; however, as most of these mitophagy regulators have been studied in the context of mitochondrial damage, more work is needed to fully explain these observations.

Future Directions

There are numerous major themes pressing to those studying mitophagy. The physiological relevance of mitophagy has been under scrutiny as the majority of mechanistic work has been carried out in cultured mammalian cells. To attempt this valid criticism, transgenic mouse models expressing reporters able to detect mitophagy have recently been developed [181]. Coupling of this recent advance with available mouse models has led to the evaluation of mitophagy in multiple mouse models of neurodegenerative disease [139, 182]. These results indicate that mitophagy is active in the central nervous system, neurons, and other tissues/cell types. However, without further evaluation, researchers cannot distinguish between driving factors in these mitophagy pathways or whether general activation of macroautophagy clears mitochondria in these reporter mice.

The basal ubiquitination state of OMM proteins may be sufficient to trigger mitophagy. For example, mitophagy occurs even in the absence of Parkin after USP30 depletion [88], validating Mul1 as a parallel regulator of a mitophagic pathway. Depletion or knockout of March5, a mitochondrial localized E3 ubiquitin ligase, delays Parkin translocation to the mitochondria upon depolarization [183]. It is unclear how baseline levels of ubiquitination on the OMM of different cell types may influence rates of mitophagy, though recent work to address this question using induced pluripotent stem cell (iPSC) derived neurons and DA neurons examined Ub/p-Ub kinetics provides some insight [184]. In vivo, Ub/p-Ub kinetics are likely to be neuronal subtype-specific or age-specific, thus differing from the cultured environment or from embryonically derived sources. The number of mitochondria contained within lysosomes under steady state conditions differs between neuronal subtype [185, 186], but it is unresolved whether the basal amount or the rate of mitochondria undergoing mitophagy differs.

It is clear that ubiquitin-dependent PINK1/Parkin mediated mitophagy has a clear connection to Parkinson’s disease and other neurodegenerative diseases [187]; however, receptor dependent mitophagy’s role in neurodegeneration is less clear. Cardiolipin mediated mitophagy hasn’t been exclusively implicated in PD, yet pharmacological drugs that model PD utilizing rotenone or 6-hydroxydopamine upregulated cardiolipin mitophagy [165]. These mitophagy receptors, Nix, BNIP3, FUNDC1, are expressed in neurons and the central nervous system [188, 189]. Due to the numerous selective mitophagy pathways and probable overlap as seen with Mul1 and PINK1/Parkin in Drosophila [106], knockout of multiple receptors may be necessary to elucidate their role in neurons and their contribution to neuronal health or neurodegenerative diseases. In an ischemic reperfusion injury/stroke paradigm, this synergy between the Nix and Parkin pathway revealed that double Nix/Parkin knockout mice had greater infarct volume than single knockout mice [190]. Increased efforts in this area of research will elucidate which and when certain mitophagy pathways are important for neurons during development, aging, and disease.

Apart from our understanding of Nix, receptor-mediated mitophagy has not been extensively explored in vivo. Hypoxia is a relevant stimulus in the nervous system and has been utilized preclinically in neurodevelopmental injury models for congenital heart disease [191] and pre-term birth [192]. However, exposure to lack of oxygen in these conditions occurs in conjunction with secondary factors, such as glucose deprivation. Recent work has demonstrated that induction of hypoxia may be a viable treatment for mouse models of mitochondrial diseases reverting neurodegenerative phenotypes as seen in Leigh’s disease [193, 194]. It would be of interest to examine these paradigms to determine whether mitophagy is actively conferring protection under therapeutically induced conditions of hypoxia.

In summary, mitophagy has proven to be a major regulator directing mitochondrial quality control and turnover, which, when perturbed, may encourage disease and disrupt physiological processes. Neuron-specific ATG5 (autophagy-related protein 5) and ATG7 (autophagy-related protein 7) KO mice are deficient in autophagy and mitophagy pathways, leading to the accumulation of abnormal mitochondria [195, 196]. This example makes apparent the ambiguity in determining the distinct contributions of macroautophagy versus mitophagy in mitochondrial degradation. Increased understanding of other types of mitophagy in addition to those reliant on ubiquitin will allow us to tease out the determining factors in different cellular contexts. The effort to maintain a healthier pool of mitochondria has driven the desire to identify pharmacological approaches to increase activation of mitophagy [197–199]. To present, most efforts have focused on modulating the PINK1/Parkin pathway, a valuable approach, though in this review we sought to highlight the less explored mechanisms of mitophagy. Mitophagy is now accepted as a vital aspect of mitochondrial quality control, influencing disease and physiological function and a prime target for therapeutic focus.

Highlights for Ubiquitin and Receptor Dependent Mitophagy Pathways and Their Implication in Neurodegeneration.

• Mitophagy pathways are either receptor or ubiquitin dependent.

• Mitophagy degrades damaged organelles for quality control.

• Mitophagy degrades healthy organelles for cellular differentiation.

• Defective mitophagy pathways are implicated in neurodegenerative diseases.

Abbreviations

ALS

amyotrophic lateral sclerosis

ATG

autophagy-related gene

BCL

B-cell leukemia/lymphoma

BH3

BCL2-homology 3

BNIP3

BCL-2/adenovirus E1B 19 kDa interacting protein 3

CCCP

carbonyl cyanide m-chlorophenylhydrazone

DA

dopaminergic

DFP

deferiprone

DRP-1

dynamin-related protein 1

Fis1

mitochondrial fission protein 1

Dub

deubiquitinases

FTD

frontal temporal dementia

FUNDC1

FUN14 Domain-Containing Protein 1

GABARAP

gamma-aminobutyric acid receptor-associated protein

GBA

glucosylceramidase beta

HD

Huntington’s disease

IMM

inner mitochondrial membrane

IMS

inner membrane space

iPSCs

induced pluripotent stem cells

K

lysine

KD

knockdown

KO

knockout

LC3

microtubule-associated protein light chain 3

LIR

LC3 interacting region

MAPL

mitochondrial-anchored protein ligase

MDVs

mitochondrial derived vesicles

mtDNA

mitochondrial DNA

MTS

mitochondrial targeting signal

Mul1

mitochondrial ubiquitin ligase 1

MULAN

mitochondrial ubiquitin ligase activator of NF-κB

NIX/BNIP3L

Nip-like protein X

OMM

outer mitochondrial membrane

OXPHOS

oxidative phosphorylation

PtdIns3P

phosphatidylinositol-3-phosphate

p-UB

phosphorylated ubiquitin

PD

Parkinson’s disease

PHB

prohibitin

PINK1

PTEN-induced putative kinase protein 1

RGC

retinal ganglion cell

ROS

reactive oxygen species

S

serine

SOD

superoxide dismutase

SUMOs

small ubiquitin-like modifiers

TBK1

Tank Binding Kinase 1

TM

transmembrane

TOM

translocase of the outer membrane

Ub

ubiquitin

UBD

ubiquitin binding domain

UBL

ubiquitin-like domain

UPS

ubiquitin proteasome system

USP

ubiquitin-specific protease

VPS35

vacuolar protein sorting-35

ZF

zinc finger

Δψm

mitochondria membrane potential