Regulation of PRKN-independent mitophagy

ABSTRACT

Mitochondria are dynamic, multifunctional cellular organelles that play a fundamental role in maintaining cellular homeostasis. Keeping the quality of mitochondria in check is of essential importance for functioning and survival of the cells. Selective autophagic clearance of flawed mitochondria, a process termed mitophagy, is one of the most prominent mechanisms through which cells maintain a healthy mitochondrial pool. The best-studied pathway through which mitophagy is exerted is the PINK1-PRKN pathway. However, an increasing number of studies have shown an existence of alternative pathways, where different proteins and lipids are able to recruit autophagic machinery independently of PINK1 and PRKN. The significance of PRKN-independent mitophagy pathways is reflected in various physiological and pathophysiological processes, but many questions regarding the regulation and the interplay between these pathways remain open. Here we review the current knowledge and recent progress made in the field of PRKN-independent mitophagy. Particularly we focus on the regulation of various receptors that participate in targeting impaired mitochondria to autophagosomes independently of PRKN.

Abbreviations: AMPK: AMP-activated protein kinase; ATP: adenosine triphosphate; BCL2: BCL2 apoptosis regulator; BH: BCL2 homology; CCCP: Carbonyl cyanide m-chlorophenylhydrazone; CL: cardiolipin; ER: endoplasmic reticulum; FCCP: carbonyl cyanide p-trifluoromethoxyphenylhydrazone; IMM: inner mitochondrial membrane; IMS: mitochondrial intermembrane space; LIR: LC3-interacting region; MDVs: mitochondrial-derived vesicles; MTORC1: mechanistic target of rapamycin kinase complex 1; OMM: outer mitochondrial membrane; OXPHOS: oxidative phosphorylation; PD: Parkinson disease; PtdIns3K: phosphatidylinositol 3-kinase; RGC: retinal ganglion cell; RING: really interesting new gene; ROS: reactive oxygen species; SUMO: small ubiquitin like modifier; TBI: traumatic brain injury; TM: transmembrane.

Introduction

Autophagy is a cellular recycling process that is highly conserved among eukaryotic organisms and essential for cellular homeostasis. It is important to maintain equilibrium between the biosynthesis and catabolism of macromolecules and organelles, and facilitates selective turnover of dysfunctional or surplus cellular components. Autophagy is defined as lysosomal degradation of cytoplasmic material and can be grouped into three main types based on how the cellular components reach the lysosome: macroautophagy, chaperone-mediated autophagy, and microautophagy. During macroautophagy, which is best studied and often simply referred to as autophagy, the cellular components to be degraded are enclosed by a double membrane (phagophore), which closes to form an autophagosome that will fuse with the lysosome (Figure 1) [1–3]. Macroautophagy was long considered an unspecific degradation process induced upon starvation, but it became evident that specific cargo can be recognized and targeted for degradation in a selective manner even under nutrient-rich conditions [4]. Selective autophagy can specifically degrade different cellular components such as mitochondria, endoplasmic reticulum (ER), ribosomes, protein aggregates, lipid droplets, and cytosolic pathogens. Selective removal of cargo by autophagy often requires their labeling with specific degradation tags (e.g. ubiquitin) that are recognized by autophagy receptors, which further facilitate cargo degradation by binding to members of the MAP1LC3/LC3 (microtubule associated protein 1 light chain 3) [5] and GABARAP (GABA type A receptor-associated protein) subfamilies, mammalian homologs of the yeast Atg8 protein, which are ubiquitin-like proteins that are conjugated to the lipid phosphatidylethanolamine in the phagophore [6,7]. Autophagy receptors have a short linear LC3-interacting region (LIR) motif by which they interact with LC3/GABARAP proteins on the phagophore (Figure 1), thereby facilitating autophagic degradation of specific cargo. The core LIR motif consists of an aromatic amino acid (W/F/Y) and a hydrophobic amino acid (L/I/V) that are separated by any two amino acids [8–11] and was first identified in the autophagy receptor SQSTM1 (sequestosome 1) bound to LC3B [10]. Later, multiple LIR motif-containing autophagy receptor proteins have been identified, as well as LIR-containing proteins that interact with LC3/GABARAP proteins to facilitate autophagosome biogenesis or trafficking [12].

Figure 1.

Molecular mechanisms of PRKN-independent mitophagy. De novo autophagosome formation can take place near the ER from various membrane sources or at/around the cargo. Autophagy initiation requires the activation of ULK1 complex that is tightly regulated by MTORC1 and AMPK and series of phosphorylation events triggered by growth signals, nutrient availability and cellular energy and oxygen levels. Phosphorylation of BECN1 by ULK1 further activates the class III PtdIns3K complex, leading to PtdIns3P synthesis and phagophore formation/nucleation. PtdIns3P binds to WIPI proteins, which further recruit the ubiquitin-like conjugation system (E1-E3) that facilitates conjugation of Atg8-family proteins to phosphatidylethanolamine (PE) (Atg8-II), leading to phagophore elongation. Mitophagy receptors such as BNIP3, BNIP3L, FUNDC1, BCL2L13, FKBP8 and PHB2 can directly bind to lipidated LC3 and GABARAP-family members at the phagophore membrane. Phagophore closure around the tagged mitochondria will form the double-membrane autophagosome (mitophagosome) that later fuses with the lysosome (mitolysosome) where mitochondria and cargo receptors are degraded and recycled by the action of lysosomal hydrolases.

Mitochondria are highly dynamic, double-membraned organelles that take part in a broad array of cellular functions such as adenosine triphosphate (ATP) production, lipid metabolism, calcium signaling and reactive oxygen species (ROS) generation and detoxification [13]. The maintenance of a healthy and functional mitochondrial network is critical for a variety of functions such as development, cell metabolism as well as the response to physiological adaptations and stress conditions [14]. Since the main role of mitochondria is energy production, it is no surprise that they are exposed to high levels of ROS, making them particularly vulnerable to mitochondrial DNA mutations and protein misfolding [14]. Consequently, several mitochondria quality control mechanisms exist to ensure a healthy mitochondria population that is sufficient to fulfill the requirements of the cell. Selective removal of damaged mitochondria by mitophagy is one such quality control mechanism, but it is also evident that mitophagy regulates mitochondrial abundance in response to environmental cues such as hypoxia, oocyte fertilization and erythroid cell maturation.

An important step in mitophagy is the priming of mitochondria for degradation. Although the molecular mechanisms involved in such priming are still elusive, it is clear that autophagy receptors play a fundamental role in recognizing mitochondria for autophagic degradation. Such mitophagy receptors are classified into two distinct groups; ubiquitin-dependent and ubiquitin-independent mitophagy receptors [8]. The ubiquitin-dependent mitophagy receptors include the five so-called sequestosome-like receptors (SQSTM1, CALCOCO2/NDP52 [calcium binding and coiled-coil domain 2], NBR1 [NBR1 autophagy cargo receptor], OPTN [optineurin] and TAX1BP1 [Tax1 binding protein 1]) that have been implicated in recognition of ubiquitinated mitochondria following mitochondria depolarization. This type of mitophagy is referred to as PINK-PRKN-dependent mitophagy as it is driven by the enzyme 3 (E3) ubiquitin ligase PRKN (parkin RBR E3 ubiquitin protein ligase) and the protein kinase PINK1 (PTEN induced kinase 1). Mutations in these two proteins are directly connected to the development of familial Parkinson disease (PD) [13], suggesting the relevance of mitophagy in the progression of neurodegenerative diseases [14]. Upon mitochondrial membrane depolarization, PINK1 accumulates at the outer mitochondrial membrane (OMM), leading to phosphorylation of ubiquitin and recruitment of PRKN. The further ubiquitination of OMM proteins by PRKN triggers recruitment of the sequestosome-like receptors, which facilitate further recruitment of the autophagy machinery to initiate autophagosome formation and mitophagy. CALCOCO2 promotes PRKN-dependent mitophagy by recruitment of the ULK1 complex subunit RB1CC1/FIP200 (RB1-inducible coiled-coil protein 1) [15–17]. Ectopic localization of ULK1 (unc-51 like autophagy activating kinase 1) to mitochondria is sufficient to activate ULK1 and induce mitophagy in an MTOR (mechanistic target of rapamycin kinase) and AMP-activated protein kinase (AMPK)-independent manner, bypassing the inhibition of autophagy under nutrient-rich conditions [15]. Moreover, it was recently discovered that the autophagy receptor OPTN initiates PRKN-dependent mitophagy by forming a complex with ATG9A [18], a multispanning membrane protein essential for autophagosome formation [19]. Interestingly, although the interaction of these autophagy receptors with Atg8-family proteins was dispensable for mitophagy, Atg8-family proteins seem to amplify mitophagy by mediating LIR-dependent recruitment of OPTN and CALCOCO2 to the growing phagophore [20]. Thus, the ubiquitin-dependent mitophagy receptors have a key role in recruitment of the core machinery required for de novo mitophagosome formation in close proximity to the cargo during PRKN-dependent mitophagy. It is currently not known if this is also the case for the ubiquitin-independent mitophagy receptors involved in PRKN-independent mitophagy.

It is interesting to note that mice lacking either PRKN or PINK1 show a mild or no phenotype [21,22], suggesting that the PINK1-PRKN pathway is dispensable for mitochondrial turnover under normal conditions. As mitophagy is essential to obtain cellular homeostasis, this indicates that alternative mitophagy pathways are used to cope with mitochondrial stress and dysfunction. One such pathway is induced upon hypoxic condition and involves transcriptional upregulation of the OMM proteins BNIP3 (BCL2 interacting protein 3) and BNIP3L (BNIP3 like) that function as ubiquitin-independent autophagy receptors [23,24]. Hypoxia-induced mitophagy is initiated in a PINK1-PRKN independent manner by stabilization and activation of the transcription factor HIF1A (hypoxia inducible factor 1 subunit alpha), leading to expression of BNIP3 and BNIP3L that interact with Atg8-family proteins in an LIR-dependent manner [23,24]. Moreover, it has recently become clear that mitochondrial-derived vesicles (MDVs) can target mitochondrial components to the lysosome [25], independently of mitochondrial fission, but still requiring PINK1-PRKN [26]. This pathway can be rapidly initiated upon mitochondrial stress in an attempt to reestablish mitochondrial function before autophagy initiation [27]. In addition, a piecemeal (bit-by-bit) mitophagy pathway was found to remove parts of damaged mitochondria and target them for lysosomal degradation, aiming to maintain the integrity of the mitochondrial network upon local damage [28].

In yeast, Atg32 is the only mitochondrial receptor protein known to selectively target mitochondria for degradation. Mitophagy can be activated through nitrogen starvation or long-term respiratory growth, leading to upregulation of ATG32 and its binding to the OMM, which further recruits the autophagy machinery by directly interacting with Atg8 and Atg11 [29–31]. In mammalian cells, the number and type of mitophagy receptors have evolved to provide a highly specific control of mitophagy to the different stimuli and cellular needs. In this review, we will discuss the selective degradation of mitochondria, with a special focus on mitophagy receptors that participate in mitophagy pathways independent of PINK1-PRKN activity.

Autophagy receptors in PRKN-independent mitophagy

The ability of the autophagy machinery to target specific organelles such as dysfunctional mitochondria relays on specific mitochondrial membrane proteins termed “mitophagy receptors” which preferentially interact with distinct LC3 and GABARAP proteins to specifically recruit dysfunctional mitochondria into autophagosomes (Figure 1). For example, BNIP3L interacts with GABARAPL1 [32], while BCL2L13 (BCL2-like protein 13), BNIP3, FUNDC1 (FUN14 domain containing 1) and AMBRA1 (autophagy and beclin 1 regulator 1) preferentially bind LC3B [33,34]. Mitophagy receptors also include FKBP8 (FKBP prolyl isomerase 8) which is known as a preferential LC3A recruiter [35]. Some of the mitophagy receptors described in this section are mainly PRKN-independent (Table 1), while others are important for both PRKN-dependent and independent mitophagy.

Table 1.

An overview of PINK1-PRKN-independent mitophagy factors and their regulators

Protein or Lipid	Function	Mitochondrial localization	Mitophagy inducers	Autophagic interactors	Regulators	References
BNIP3	Receptor	OMM	Hypoxia	LC3B ATG8	FOXO3 ↑ HIF1A ↑ MA-5 ↑ ULK1 ↑	[34,57,58,67,70,72,73]
BNIP3L	Receptor	OMM	Hypoxia High OXPHOS activity	GABARAPL1	HIF1A ↑	[32,42,43,56]
FUNDC1	Receptor	OMM	Hypoxia FCCP	LC3B	ULK1 ↑ SRC ↓ PGAM5 ↑ CSNK2 ↓ MIR137↓	[84,87,88,95]
FKBP8	Receptor	OMM	Starvation	LC3A	RHEB↓	[35,102,107–109,111]
BCL2L13	Receptor	OMM	CCCP	LC3B		[113]
AMBRA1	Receptor	OMM	FCCP	LC3 BECN1	CHUK↑ MCL1 ↓ GSK3B↑	[116,119–123]
PHB2	Receptor	IMM	CCCP Oligomycin and antimycin	LC3B	AURKA↑	[125,126,130–132]
Cardiolipin	Membrane phospholipid	OMM and IMM	Rotenone CCCP	LC3 BECN1	PKC↑ CRLS1↑ PLSCR3↑ SNCA↑	[133–136,139,141,142]
Ceramide	Sphingolipid	OMM	C18-ceramide	LC3B	DNM1L↑	[144,145]
MUL1	E3-ubiquitin ligase	OMM	Selenite	GABARAP	ULK1↑	[137,146,147,152,153]
SIAH1	E3-ubiquitin ligase	-			PINK1↑ SNCAIP↑	[154]
ARIH1	E3-ubiquitin ligase	-				[155]

Abbreviations: AMBRA1: autophagy and beclin 1 regulator 1; ARIH1: ariadne RBR E3 ubiquitin protein ligase 1; AURKA: aurora kinase A; BCL2L13: BCL2 like 13; BNIP3: BCL2 interacting protein 3; BNIP3L: BCL2 interacting protein 3 like; CCCP: carbonyl cyanide m-chlorophenyl hydrazone; CHUK: component of inhibitor of nuclear factor kappa B kinase complex; CRLS1: cardiolipin synthase 1; CSNK2: casein kinase 2; DNM1L: dynamin 1 like; FCCP: carbonyl cyanide p trifluoromethoxyphenylhydrazone; FKBP8: FKBP prolyl isomerase 8; FOXO3: forkhead box O3; FUNDC1: FUN14 domain containing 1; GABARAPL1: GABA type A receptor associated protein like 1; GSK3B: glycogen synthase kinase 3 beta; HIF1A: hypoxia inducible factor 1 subunit alpha; MA-5: mitochonic acid-5; MAP1LC3A/LC3A: microtubule associated protein 1 light chain 3 alpha; MAP1LC3B/LC3B: microtubule associated protein 1 light chain 3 beta; MCL1: MCL1 apoptosis regulator, BCL2 family member; MUL1: mitochondrial E3 ubiquitin protein ligase 1; OXPHOS: oxidative phosphorylation; PGAM5: PGAM family member 5, mitochondrial serine/threonine protein phosphatase; PHB2: prohibitin 2; PLSCR3: phospholipid scramblase 3; PRKC: protein kinase C; RHEB: Ras homolog, mTORC1 binding; SIAH1: siah E3 ubiquitin protein ligase 1; SNCA: synuclein alpha; SNCAIP: synuclein alpha interacting protein; SRC: SRC proto-oncogene, non-receptor tyrosine kinase; ULK1: unc-51 like autophagy activating kinase 1.

BNIP3L

Oxygen concentration must be precisely regulated to sustain energy and redox homeostasis [23]. In conditions of decreased oxygen levels (hypoxia), the first transcriptional regulators to respond are the hypoxia-inducible factors (HIFs), having important physiological and pathophysiological functions [36]. Moreover, the process of oxidative phosphorylation (OXPHOS) involves the risk of developing ROS [37]. This happens as a result of electrons prematurely reacting with oxygen at respiratory complex I or complex III. Prolonged exposure to elevated concentrations of ROS results in the oxidation of nucleic acids, lipids, and proteins, which consequently leads to cell dysfunction or death. Under normoxia, HIF1A is hydroxylated by prolyl hydroxylase domain proteins and directed to the proteasome for degradation. The catalytic activity of prolyl hydroxylase domain proteins is inhibited when oxygen is limiting [38], leading to stabilized of HIF1A and its translocation to the nucleus, where it binds to hypoxia response elements (conserved RCGTG sequences) in the promoter regions of HIF-regulated genes [39–41]. HIF1A regulates the transcription of hundreds of genes in response to hypoxia, including the mitophagy receptors BNIP3 and BNIP3L [42].

BNIP3L was first identified as a dimeric pro-apoptotic mitochondrial protein that interacted with anti-apoptotic proteins BCL2/adenovirus E1B 19 kDa [43]. As it contains a BCL2 homology (BH) domain 3, it is classified as BH3-only protein (Figure 3). BNIP3L is inserted into the OMM by its C-terminal transmembrane domain, with the N terminus being exposed to the cytosol [44]. Although initial studies of this protein focused on its role in cell death, BNIP3L was found to have a role in mitophagy under hypoxic conditions [24]. BNIP3L transcription is upregulated under hypoxic condition in a HIF1A-dependent manner and for the protein to be active in mitophagy it goes through a series of modifications [45–48]. The combined mechanism of LIR phosphorylation and receptor dimerization is needed for proper BNIP3L-dependent mitophagy initiation and progression [48,49].

Figure 3.

Schematic representation of the domains of proteins involved in the PRKN-independent mitophagy. Domain structures are shown schematically with sizes of proteins roughly in scale. Abbreviations: LIR: LC3-interacting region; TM: transmembrane domain; BH: BCL2 homology domain; RING: RING domain; IMS: mitochondrial intermembrane space; SIAH: SIAH zinc finger domain; IBR: IBR zinc finger domain; UBA: ubiquitin-associated like domain; WD: WD repeats.

Its role in mitophagy was first demonstrated in the context of erythropoiesis. BNIP3L is transcriptionally upregulated during erythroid differentiation and is essential for mitochondrial clearance via mitophagy during the terminal stages of this process [50–52]. Strikingly, in erythrocytes from BNIP3L-deficient mice, mitochondria fail to be incorporated into autophagosomes, implying that BNIP3L is critical for their recruitment [52]. Ablation of BNIP3L interrupts erythroid maturation, resulting in anemia, reticulocytosis and erythroidmyeloid hyperplasia [51,52]. Besides erythroid differentiation, BNIP3L is also required for mitophagy during somatic cell reprogramming to induced pluripotent stem cells and for retinal ganglion cell (RGC) differentiation [53–55]. During RGC differentiation, local hypoxia transcriptionally induces BNIP3L, accumulation of which triggers both mitochondrial removal and activation of glycolysis. Retinas from mice lacking BNIP3L retain their mitochondria and exhibit fewer RGCs [54,55].

BNIP3L is also involved in mitophagy induced when the demand for mitochondrial OXPHOS is high [56]. Accelerated rejuvenation of the mitochondrial pool is believed to contribute to increased OXPHOS efficiency. In these circumstances, the small GTPase RHEB is recruited to the OMM where it is proposed to stimulate mitophagy through binding BNIP3L and recruiting LC3 [56]. Novak et al., 2010, identified a highly conserved LIR motif (“WxxL”) in the N-terminus of BNIP3L that is exposed toward the cytosol. The BNIP3L LIR motif interacts with GABARAP family proteins and recruits GABARAPL1 triggering mitochondrial depolarization, mediating partial clearance of mitochondria [32].

To conclude, BNIP3L has been found to have a critical role in various cellular processes such as: programmed mitophagy, metabolic reprogramming and controls the differentiation of red blood cells and erythrocytes. It is still necessary to elucidate the molecular mechanisms underlying the function of BNIP3L in mitophagy to further understand how, when and where it is regulated. It is also important to understand its contextual regulation with different signaling mechanisms, depending on the development phase, stress condition, metabolic state or tissue type.

BNIP3

BNIP3 is a mitochondrial protein that, similar to BNIP3L, contains a BH3 domain, acts as a pro-apoptotic factor and was identified as an interactor of adenovirus E1B-19 kDa and BCL2 (BCL2 apoptosis regulator) proteins [57,58]. Many different cell types express the BNIP3 protein and it has been implicated in multiple cellular processes, including mitochondrial dysfunction, mitophagy, apoptosis, membrane permeability, transition pore opening, mitochondrial membrane potential, oxidative stress, calcium overload, mitochondrial respiratory collapse, and ATP shortage of mitochondria [59–61]. The structure of BNIP3 is comprised of a large complex N-terminal region and a characteristic C-terminal trans-membrane (TM) domain that targets BNIP3 to mitochondria (Figure 3) [62]. The N-terminal region contains three major domains: (1) a proline, glutamic acid, serine, threonine, and aspartic acid domain; (2) a BH3 domain adjacent to the proline, glutamic acid, serine, threonine, and aspartic acid domain; and (3) a conserved domain [63–65], as well as a LIR motif (’18-WxxL-21ʹ), which is enclosed by two serine residues (Ser17 and Ser24) that are both phosphorylated. Phosphorylation of the Ser17 residue is essential for BNIP3 binding to LC3B, while a strong binding to LC3B and GABARAPL2 was observed upon phosphorylation of both the Ser17 and the Ser24 residues [34]. Interestingly, if compartmentalized at the ER, BNIP3 promotes reticulophagy dependent on the LIR motif, indicating that its autophagy function is independent of mitochondrial targeting [33].

In addition to its regulation by post-translational modifications, the activity of BNIP3 in mitophagy is also regulated transcriptionally [66–69]. BNIP3 is reported to be upregulated during starvation by binding of the FOXO3 (forkhead box O3) transcription factor to the promoter of BNIP3 and BNIP3L in skeletal muscles [70]. Moreover, the expression level and transcriptional level of Foxo3 is upregulated during erythroid maturation where BNIP3L is shown to be upregulated [71].

In neonatal rat cardiomyocytes, BNIP3 expression is significantly upregulated after HIF1A overexpression. The BNIP3 gene promoter possesses a hypoxia-responsive element, which confirms the existence of BNIP3 as a hypoxia-responsive gene [72]. In line with this, suppression of HIF1A resulted in downregulation of the expression levels of BNIP3 and BECN1 (beclin 1), suggesting an involvement of the HIF1A-BNIP3-BECN1 signaling cascade pathway in hypoxia-induced autophagy [73]. It is also known that BNIP3 is broadly expressed in cancer cells and is involved in cancer growth in tissues under hypoxia or hypoxia-like conditions [74,75]. In a mouse model of mammary tumorigenesis, loss of BNIP3 resulted in reduced mitophagy, the accumulation of functionally impaired mitochondria, elevated ROS production and consequent stabilization of HIF1A [76]. Increased HIF1A target gene expression favored angiogenesis and bioenergetic changes associated with Warburg metabolism that ultimately promoted invasiveness and metastasis. Furthermore, several studies have reported that HIF1A limits ROS generation via mitophagy by promoting BNIP3 expression during hypoxia [42]. In one of those studies, it was found that the drug mitochonic acid 5 acts as an upstream transcriptional trigger that upregulates BNIP3 expression via the MAPK/ERK-YAP1 signaling pathway, resulting in activation of mitophagy and decreased ROS production [77]. In line with this, it was reported that BNIP3-induced mitophagy has a protective role against mitochondrial dysfunction under lactate through controlling the ROS generation [78]. In conclusion, these studies indicate that BNIP3-dependent mitophagy is induced to prevent mitochondrial ROS levels and is activated by the stabilization of HIF1A.

Despite many studies attempting to understand the underlying mechanisms of HIF1A and BNIP3 in mitophagy, many contradictions and dilemmas remain to be clarified. BNIP3-induced mitophagy under hypoxic conditions has been proposed to be through generation of mitochondrial depolarization [23,79,80]. BNIP3 is also known to compete with BECN1 for binding to BCL2, thereby increases the levels of free BECN1, which triggers autophagy [23,81]. In cardiac myocytes overexpression of BNIP3 leads to mitochondrial fragmentation, translocation of the DNM1L (dynamin 1 like) protein to mitochondria, which ultimately leads to the induction of mitophagy [82]. Another study using a BNIP3 expression vector lacking the transmembrane (TM) domain found that autophagic degradation of BNIP3 is not dependent on its association with mitochondria, but that it is regulated by ULK1 via the MTOR complex 1 (MTORC1) and AMPK [67]. ULK1 activity is closely regulated by AMPK and MTORC1 (Figue 1). AMPK phosphorylates ULK1 or ULK2 at Ser555, leading to induction of autophagy [83]. It was demonstrated that inhibition of the AMPK-mediated phosphorylation of ULK1 or ULK2 at Ser555 leads to the accumulation of BNIP3. It is however not known if and how the activity of ULK1 activity regulates BNIP3 phosphorylation and subsequent mitophagy.

FUNDC1

FUNDC1 is a ubiquitously expressed protein with mitochondrial localization. It contains three transmembrane regions integrated into the OMM, a C-terminus stretched into the intermembrane space of mitochondria and a cytosolic N-terminus (Figure 3). The protein acts as a mitophagy receptor in hypoxia-induced and carbonyl cyanide p-trifluoromethoxyphenylhydrazone (FCCP) -induced mitophagy [84] by interacting with LC3B via a LIR (‘18-YxxL-21ʹ) motif located on the N-terminal region of FUNDC1. It was reported that FUNDC1 interacts with other Atg8-family protein homologs as well, but with less affinity. Point mutations in or a deletion of the LIR domain are reported to result in an impairment of FUNDC1-mediated mitophagy in HeLa cells [84]. Mutational and structural analyses have shown that Lys49 of LC3B interacts with phosphorylated Ser17 of FUNDC1 and that the side chain of LC3B undergoes a large structural rearrangement to accommodate the phosphorylated FUNDC1, thus acting as a sensor for the phosphorylation state of FUNDC1 [85,86]. ULK1 gets recruited to fragmented mitochondria under hypoxia or FCCP treatment where it phosphorylates FUNDC1 at Ser17, enabling its interaction with LC3 and linking mitochondria to autophagosomes [87]. On the contrary, phosphorylation of FUNDC1 at Tyr18 leads to a side chain extension, which seems to disrupt its interaction with LC3B by conflicting with the hydrophobic pocket of LC3B [85,86]. Tyr18 is phosphorylated by SRC kinase under normoxic conditions [84], leading to inhibition of FUNDC1 [87], and its dephosphorylation is thus required for the activation of FUNDC1-mediated mitophagy [88]. Furthermore, the Arg10 side chain of LC3B forms hydrogen bonds with the side chain and backbone carbonyl group of Ser13 in FUNDC1 and phosphorylation of this residue blocks LC3B-FUNDC1 interaction via steric hindrance [85,86]. PGAM5 (PGAM family member 5, mitochondrial serine/threonine protein phosphatase) is an enzyme with cytosolic and mitochondrial localization [89]. The mitochondrial PGAM5 acts as a serine/threonine phosphatase to regulate mitochondrial dynamics and mitophagy, while the proteolytically cleaved cytosolic form of PGAM5 acts as a phosphohistidine phosphatase to inhibit NME2 (NME/NME23 nucleoside diphosphate kinase 2) [89]. PGAM5 dephosphorylates FUNDC1 at Ser13 upon hypoxia treatment or FCCP-induced depolarization of mitochondrial membrane potential, thus acting as a positive regulator of mitophagy [88]. The casein kinase 2 (CSNK2), a constitutive serine/threonine kinase, was found to counteract the effect of PGAM5 through a reversible phosphorylation of FUNDC1 at Ser13 under basal condition, thus preventing autophagic clearance of mitochondria [88]. Overall, the phosphorylation status of FUNDC1, mediated by SRC, CSNK2 and ULK1 kinases as well as phosphatases such as PGAM5, dictates its interaction with Atg8-family proteins, thus regulating FUNDC1-induced mitochondrial turnover.

In addition to regulation of the FUNDC1-Atg8-family protein interaction, the PGAM5 phosphatase plays an important role in regulation of the crosstalk between mitophagy and apoptosis. Under basal conditions BCL2L1/BCL-xL is found in its non-phosphorylated form in complex with pro-apoptotic proteins such as BAX and BAK, as well as PGAM5 [90]. This interaction prevents PGAM5-mediated FUNDC1 dephosphorylation and subsequent mitophagy. Severe mitotic stress leads to phosphorylation and subsequent release of BCL2L1 from the complex and the induction of apoptosis. PGAM5, in its dimeric state, then dephosphorylates BCL2L1 to reactivate its anti-apoptotic function. Mild oxidative stress, on the other hand, leads to oligomerization of PGAM5, which abolishes its ability to bind to BCL2L1 while retaining its ability to dephosphorylate FUNDC1, leading to activation of mitochondrial fission and mitophagy, thus promoting cell survival [90].

FUNDC1 has been suggested to act at a crossroad between mitochondrial fusion, mitochondrial fission and, possibly, mitophagy. Before a dysfunctional mitochondrion can be engulfed by an autophagosome, it must be fragmented through mitochondrial fission or some other mechanism and segregated from the mitochondrial network since elongated mitochondria are too long to be sequestered by autophagosomes [91]. It was found that FUNDC1 forms interactions dependent on its phosphorylation status with the mitochondrial dynamin-like GTPases DNM1L and OPA1, required for mitochondrial fission and fusion, respectively. Phosphorylation of FUNDC1 Ser13 promotes its interaction with OPA1 and decreases its interaction with DNM1L, thus inhibiting mitochondrial fission. In addition, Lys70 of FUNDC1 was found to be important for the recruitment of OPA1 under basal conditions and knockdown of OPA1 was reported to enhance FUNDC1-induced mitophagy. Mitochondrial stress, on the other hand, leads to a dissociation of the FUNDC1-OPA1 complex, causing the released FUNDC1 to assemble with DNM1L to propagate mitochondria fission [92]. However, DNM1L was found to be dispensable for hypoxia-induced and deferiprone-induced mitophagy [93]. Under these conditions, elegant live imaging showed that the portion of the mitochondrion labeled for degradation is enwrapped by a phagophore and segregated from the mitochondrial network dependent on early core autophagy factors, but independently of DNM1L-mediated mitochondrial fission via an unidentified mechanism [93].

Of note, FUNDC1 accumulates at ER–mitochondrial contact sites in HeLa cells at the early stage of hypoxia by interacting with the ER protein calnexin. As mitophagy proceeds, FUNDC1-calnexin association is broken and the cytosolic part of FUNDC1 is then free to act as a receptor for DNM1L, which in turn is recruited to the ER–mitochondrial contact sites to drive mitochondrial fission [91]. The mitochondrial E3 ubiquitin-protein ligase MARCHF5 also interacts with FUNDC1 and seems to fine‐tune the graded response of cells to hypoxic stress [94]. It was demonstrated that FUNDC1 is degraded through MARCHF5 ‐mediated ubiquitination and proteasome‐dependent degradation as early as 3 h following hypoxia treatment. This results in the reduction of mitophagy, which might protect the cells from excessive mitophagy under hypoxic conditions [94]. FUNDC1 expression level is regulated at basal state through MIR137, which is downregulated in response to hypoxia. MIR137 also targets BNIP3L, thereby preventing mitophagy under basal conditions [95].

FUNDC1 was also implicated in the regulation of cellular proteostasis. By interacting with the cytosolic chaperone protein HSPA8/HSC70 (heat shock protein family A [Hsp70] member 8) under proteostatic stress and during aging [96], it facilitates translocation of unfolded cytosolic proteins into mitochondria where they are degraded via mitochondrial Lon protease or form mitochondrion associated protein aggregates [97].

FUNDC1 has been linked to pathophysiological conditions. For example, in cardiac ischemia, phosphorylation of FUNDC1 activates FUNDC1-mediated mitophagy to promote the survival of cardiomyocytes by preventing mitochondria-mediated apoptosis. Conversely, upon reperfusion injury, upregulation of both RIPK3 (receptor interacting serine/threonine kinase 3) [98] and CSNK2A/CSNK2α inhibit FUNDC1 by phosphorylation of Ser13, thereby suppressing FUNDC1-dependent mitophagy leading to apoptosis [99]. FUNDC1-dependent mitophagy is important for maintaining homeostasis and differentiation of adult cardiac progenitor cells since these cells exist in a hypoxic environment in vivo [100]. In addition, deficiency of FUNDC1 results in malformation of body axis in zebrafish embryos [101], implicating this protein in differentiation and development.

FKBP8/FKBP38

FKBP8/FKBP38 is an OMM receptor protein and is a unique member of the FK506 family of binding proteins. Receptor proteins have a variety of protein-binding modules, which connect protein-binding partners and facilitate the formation of larger signaling complexes. By linking specific proteins together, cellular signals can be propagated to generate an effective cell-to-environment response [102]. FKBP8 possesses Ca²⁺/calmodulin-activated peptidylprolyl cis-trans isomerase activity, which has a potent anti-apoptotic function via recruitment of BCL2 and BCL2L1 to mitochondria. The domain architecture of FKBP8 consists of an N-terminal Glu-rich domain following the peptidylprolyl cis-trans isomerase domain, three tetratricopeptide repeat domains, calmodulin-binding domain and a TM domain (Figure 3) [103,104]. FKBP8 is anchored in the OMM via its TM domain, with the N-terminus pointing toward the cytosol [105,106]. FKBP8 has been associated with cellular processes such as apoptosis and MTORC1 regulation. FKBP8 binds to the MTORC1 complex and inhibits its activity. RHEB is activated during nutrient-rich conditions, binds to FKBP8 and BCL2 and releases them from MTORC1, while during nutrient deprivation RHEB remains inactive and enables FKBP8 inhibition of MTORC1 [107–109]. FKBP8 has been reported to escape to the ER to avoid being degraded, which is important to prevent apoptosis from occurring during mitophagy [110]. The exact mechanism by which FKBP8 escapes degradation is unclear, as are the reasons for this intriguing behavior, although it depends on a number of basic residues within its C-terminal sequence. [15]. It is proposed that FKBP8 can boost autophagy in one of the following ways, first by inhibiting upstream MTORC1 activity, second by recruiting core autophagy components to ER-mitochondria membranes, and third by recruiting Atg8 family proteins.

FKBP8 showed interactions with LC3A, LC3B, GABARAP, and GABARAPL1, but very weak binding to LC3C and Atg8-family proteins. In contrast to other mitophagy receptors, the LIR-containing region of FKBP8 does not contain any phosphorylatable residue (Figure 1). FKBP8 targeted to the OMM interacts preferentially with lipidated LC3A (LC3A-II), but is not absolutely required for LC3A recruitment to the mitochondria, as LC3A is present in the mitochondrial fraction even in FKBP8-deficient cells treated with carbonyl cyanide m-chlorophenylhydrazone (CCCP). However, the presence of FKBP8 augments LC3A recruitment to mitochondria especially in times of mitochondrial stress [35,111].

BCL2L13

BCL2L13/BCL-RAMBO (BCL2 like 13) is a ubiquitously expressed protein first described as a pro-apoptotic member of the BCL2 protein family. According to a sequence-based prediction, it contains all of the conserved BH domains: BH1, BH2, BH3 and BH4 and two LIR motifs, WXXL/I located at positions 147–150 and 273–276 (Figure 3) [112]. The protein is integrated into the OMM via its C-terminal transmembrane domain with a few amino acids at the C-terminal end being exposed to the inter-membrane mitochondrial space, while the N terminus is exposed to the cytoplasm [113]. BCL2L13 induces apoptosis in HEK293T cells, HeLa cells, PC-3 prostate cancer cells and S2 cells, possibly by interacting with adenine nucleotide translocase to facilitate the formation of membrane permeability transition pore and the release of cytochrome c [114,115]. Like Atg32, BCL2L13 has mitochondrial localization, and a LIR motif [112]. Its contribution to mitophagy remains unclear as the mouse BCL2L13, but not the human form of the protein, takes part in CCCP-induced mitochondrial fragmentation and mitophagy via LIR-mediated binding of LC3B. Mutations in the LIR motif abolish mitophagy, but have no effect on mitochondrial fragmentation. Rather, all four BH domains play a role in BCL2L13-induced mitochondrial fission, which occurs in a DNM1L-independent manner [113]. Clearly, clarification of the exact role of BCL2L13 in mitophagy will be the subject of future studies.

AMBRA1

AMBRA1 was initially identified as a pro-autophagic protein, which takes part in the initial steps of phagophore formation, thus acting as a general regulator of autophagy. It is an interactor of BECN1 and a promoter of BECN1 association with the class III phosphatidylinositol 3-kinase (PtdIns3K) complex [116]. Functional deficiency of AMBRA1 in mice leads to dramatic alterations in the nervous system development with defective autophagy and excessive apoptosis in the brain, accumulation of ubiquitinated proteins in neurons and embryonic lethality [117]. It was demonstrated that AMBRA1 plays a role in mitophagy when it is recruited to damaged mitochondria to locally activate the PtdIns3K complex, leading to phagophore formation in the proximity of perinuclear mitochondrial clusters [118]. AMBRA1 possesses a LIR motif (Figure 3), which binds LC3 following FCCP induced mitochondrial depolarization [119] and phosphorylation of Ser1014 of AMBRA1 by a component of the nuclear factor kappa B subunit inhibitor (NFKBI) complex promotes the AMBRA1-LC3B interaction and mitophagy [120]. Interestingly, although AMBRA1 binds PRKN and is localized to depolarized mitochondria, its overexpression leads to ubiquitination of mitochondria independent of PRKN. Moreover, AMBRA1 can induce mitophagy in Pink1- or Prkn-deficient MEFs and in human fibroblasts carrying inactive PRKN, indicating a role for AMBRA1 in an alternative mitophagy pathway [119]. In addition, expression of AMBRA1 in HeLa cells depleted for PINK1-PRKN-related mitophagy receptors such as CALCOCO2, OPTN, TAX1BP1, SQSTM1 and NBR1 rescued mitochondrial clearance [120].

AMBRA1 has also been described as a BH3-like protein, containing a BH3 domain that facilitates its interaction with BCL2 family proteins [121]. Under basal conditions a pool of AMBRA1 is bound to mitochondria via BCL2 proteins and this interaction inhibits its autophagic function. Following the induction of autophagy AMBRA1 is released from the mitochondrial BCL2 to interact with BECN1 and participate in the autophagic process [122]. In addition, AMBRA1 was reported to be a ubiquitin-independent mitophagy receptor, which cooperates with E3 ubiquitin ligase HUWE1 during membrane depolarization-induced mitophagy [120]. It was reported that a fusion protein composed of MYC-AMBRA1 and a sequence from the Listeria monocytogenes actin assembly-inducing protein favors MFN2 (mitofusin-2) is a substrate for HUWE1-mediated ubiquitination [120]. In line with these results, MCL1 (myeloid cell leukemia 1), a member of the BCL2 family, was shown to be a potent inhibitor of AMBRA1-mediated mitophagy through its inhibition of HUWE1, while GSK3B (glycogen synthase kinase 3 beta) rescued the activity of HUWE1 by inhibiting MCL1 [123]. Expression of AMBRA1^ActA in an in vitro model of PD based on the inhibition of mitochondrial respiratory chain complexes I and IV by 6-hydroxydopamine or upon blockade of complex I with rotenone in SH-SY5Y cells, increased cell viability and reduced the generation of ROS [120] indicating that AMBRA1-induced mitophagy is able to suppress oxidative stress and apoptosis in these cells.

In conclusion, AMBRA1 binds to depolarized mitochondria and it is able to induce mitophagy either by acting as a receptor for E3-ubiquitin ligases such as PRKN and HUWE1 or by directly recruiting LC3B, thus acting as a mitophagy receptor.

PHB2

PHB2 (prohibitin 2), a ubiquitously expressed protein from the STOM (stomatin)-PHB (prohibitin)-FLOT (flotillin)-HflK/C superfamily [124], localizes in multiple cell compartments such as the nucleus, mitochondria and cytosol, where it exerts different functions [125]. Mitochondrial PHB2 interacts with PHB/PHB1 (prohibitin 1) to form a really interesting alternating tetrameric new gene (RING)-finger domain‐like complex located in the inner mitochondrial membrane (IMM) [125,126]. The PHB-PHB2 complex regulates the activity of mitochondrial proteases, such as m-AAA that mediates degradation of various mitochondrial membrane proteins [127], including OPA1, important for maintenance of cristae morphology and mitochondrial fusion [128]. In addition, the complex maintains the integrity of the mitochondrial genome [129], and has a chaperon-like function that stabilizes OXPHOS machinery.

PHB2 acts as a mitophagy receptor. Specifically, the PHB-PHB2 complex interacts with LC3-II upon CCCP or oligomycin and antimycin treatment. Such LC3-II-PHB2 interaction is dependent on expression of PRKN, suggesting that PRKN and PHB2 work in concert to facilitated mitophagy. PHB2 binds LC3-II via a LIR motif, and mutation of the critical LIR residues prevented this interaction and led to a decrease in mitophagy upon oligomycin and antimycin treatment of cells. According to this model, PRKN-mediated proteasomal degradation causes rupture of the OMM, exposing PHB2 in the inner membrane, which then interacts with LC3 [130]. It was further shown that PHB2 promotes PINK1-PRKN mediated mitophagy through stabilization of PINK1. This happens either by preventing the proteolytic cleavage of PINK1 via direct binding to the mitochondrial inner membrane protease PARL (presenilin associated rhomboid like), or by association with PGAM5 to prevent its degradation by PARL, as PGAM5 is known to anchor PINK1 and enable the transfer of PINK1 to the mitochondrial outer membrane to initiate mitophagy [131]. It was recently reported that PHB2 can induce mitophagy in a PRKN-independent manner whereby AURKA (aurora kinase A) is imported into mitochondria where it forms a tripartite complex with LC3B and PHB2, resulting in activation of mitophagy. Phosphorylation of PHB at Ser39 is a prerequisite for the formation of this complex [132].

PHB2 is a unique mitophagy receptor as it operates on the inner mitochondrial membrane but relies on PRKN-mediated ubiquitination of the outer membrane. It still remains to be clarified whether PHB2 works in concert with other mitophagy pathways in a similar way.

Regulation of mitophagy by lipids

Our knowledge of phospholipids needed for the buildup of the autophagic membrane at the vicinity of the mitochondrial membrane is rather limited; however, specific lipids play an important role in the regulation of mitophagy. In the following section, we focus on cardiolipin (CL) and ceramide that regulate the selective removal of dysfunctional mitochondria.

Cardiolipin

CL is a phospholipid characteristically found in mitochondrial membranes, where it represents up to 20% of total mitochondrial phospholipid content [133,134]. CL is a flexible molecule composed of a glycerol head group bound to four fatty acyl chains adopting a cone-shaped geometry that localizes at high membrane curvature regions [133]. CL interacts and stabilizes proteins from the electron transport chain (complexes I, III, IV and ATP synthase [135,136]), participates in mitophagy and apoptosis [136,137].

In healthy mitochondria, CL is unevenly distributed between the IMM (about 97% CL) where it is synthesized, and OMM (about 3% CL) [138]. An increase in CL content at the OMM occurs in response to mitochondrial damaging agents such as Rotenone (inhibits complex I) and CCCP (uncoupler that disrupts ATP synthesis) [135,136], possibly indicating that CL exposure at the OMM can act as a signal for damaged or dysfunctional mitochondria (Figure 2). The silencing of CRLS1 (cardiolipin synthase 1; catalyzes CL synthesis) and PLSCR3 (phospholipid scramblase 3; mediates the transport of CL from IMM to OMM upon phosphorylation of Thr21 by PRRT2/PKC [proline rich transmembrane protein 2]) genes reduced the distribution of CL on the OMM and mitophagy levels [139], supporting the plausible function of CL transfer to OMM in mitophagy signaling [136]. This hypothesis was reinforced by the finding that CL residing in the OMM was able to directly recruit and bind to the N-terminal helix of LC3 independently of mitochondria depolarization [136]. This CL-LC3 interaction might explain the contribution of CL to the initial stages of the autophagic pathway and to signal damaged mitochondria (Figure 2). The participation of CL in mitophagy is also suggested by the direct interaction of CL residing at the OMM with the autophagy protein BECN1 [133]. The evolutionarily conserved domain of BECN1 preferentially binds lipid membranes enriched in CL [135]. Mitochondrial fission plays an important role in mitophagy since large dysfunctional mitochondria (5 μm) must be engulfed by autophagosomes (1 μm) for degradation [135]. DNM1L transfers from the cytosol to the OMM where it oligomerizes and fragments mitochondria into smaller structures. In vitro studies using CL-agarose beads showed that DNM1L has a strong binding affinity to CL [135] and cooperate to promote mitochondrial division [140].

Figure 2.

Ceramide and CL mediate PRKN independent mitophagy upon mitochondrial damage. (A) Ceramide at the OMM can bind LC3B-II: targeting autophagosomes to the mitochondria to initiate mitophagy. Ceramide signaling induces BECN1 expression guiding LC3B lipidation and autophagy. DNM1L expression influences ceramide distribution at the OMM, contributing to its interaction with LC3B-II. (B) CL is mainly found in mitochondrial membranes and its translocation from IMM to OMM triggers mitophagy. PLSCR3 is responsible for CL transport from IMM to OMM and its depletion leads to reduced distribution of CL at the OMM and lower levels of mitophagy. CL in the OMM can directly bind LC3 and BECN1 and has a strong binding affinity for DNM1L.

PD is often associated with dysfunctional mitochondria and accumulation of misfolded protein species that ultimately lead to neural degeneration. Mutations in SNCA (synuclein, alpha) protein play an important role in early-onset familial PD and can favor beta-sheet protein conformations that are prone to fibrilization [141]. SNCA was described to interact with CL at the OMM [142] and acquires alpha-helical conformation in the presence of negatively charged phospholipids such as CL [143]. The interaction between SNCA and CL lead Ryan et al. to propose a novel mechanism where CL supports the folding of mutant SNCA monomers to alpha-helical conformation reversing their beta-sheet fibril conformation. In this study, they also described that the binding of mutant SNCA to CL increased LC3 recruitment to OMM and induced mitophagy [141]. The translocation of CL to OMM was suggested to function as a molecular marker for the development of PD and that mitochondria may be critical for SNCA folding and reversing protein fibrilization, slowing down SNCA pathogenicity [141].

CL was also found to be involved in an early signaling pathway that promptly triggers mitophagy after traumatic brain injury (TBI) in human and rat brain samples in an attempt to avoid irreversible neural apoptosis [139]. It was observed that CL synthase and PLSCR3 impairment in TBI rat models reduced the levels of mitophagy. In injured brain samples, an increase in PLSCR3 phosphorylation correlated with the higher levels of CL found in OMM fractions [139]. The early mitophagy activation by CL seems to be an advantageous response to avoid further neuronal damage resulting from TBI.

Ceramide

The sphingolipid ceramide is composed of a sphingosine backbone and a fatty acyl chain. The number of carbon atoms in the fatty acyl chain determines different types of ceramides, ranging from C14 to C26-ceramide [144], and their de novo synthesis requires the catalytic action of CERS1 (ceramide synthase 1) to CERS6. CERS1 is responsible for the synthesis of C18-ceramide and both components participate in the regulation of cell proliferation, apoptosis and more recently to the selective removal of damaged mitochondria [145].

Sentelle et al. demonstrated that CERS1 and its metabolic product C18-ceramide can selectively induce lethal mitophagy independently of apoptosis in multiple head and neck squamous cell carcinoma (HNSCC) cell lines. Overexpression of CERS1 or treatment of HNSCC cancer cell lines using the ceramide analog D-erythro-C14C18-pyridinium ceramide bromide, that preferentially accumulates at the mitochondria, lead to an increase in LC3B lipidation and binding of ceramide to LC3B-II at the mitochondria membrane [145]. Structure similarities between the LC3B globular domain and the ceramide-binding domain of ceramide transporter protein guided the discovery of LC3B Ile35 and Phe52 residues to be essential for C18-ceramide binding [144,145]. This lipid-protein interaction at the OMM facilitates the recruitment and expansion of autophagosomes to initiate mitophagy (Figure 2).

While the length of the fatty acyl chain of ceramide seems to not be essential for ceramide-induced mitophagy, the localization of ceramide species within the cell is. CERS1 overexpression induced C18-ceramide synthesis and its localization on the OMM, whereas CERS6 overexpression neither caused C16-ceramide accumulation in mitochondria nor induction of mitophagy. In contrast, treatment with the C16-ceramide analog C16-Pyr-Cer (accumulates preferentially in mitochondria) inhibited mitochondrial function at the level of oxygen consumption rate and induced mitophagy [145].

Ceramide-induced mitophagy was shown to be mediated by the activation of mitochondrial fission [145]. DNM1L depletion inhibited mitophagy when using D-erythro-C14C18-pyridinium ceramide bromide and reduced the distribution of ceramide at the OMM, preventing the recruitment of LC3B to the mitochondria. The exact mechanism that relates ceramide-LC3B-II-mediated mitophagy to mitochondrial fission is still not fully described.

Lipid-induced mitophagy is a relatively unexplored topic. CL and ceramide are distinct molecules capable of signaling and recruiting the autophagic machinery to dysfunctional mitochondrial. The molecular mechanisms of these lipids in mitophagy seem close enough, but any direct interaction is not described and might be of interest to pursue.

E3 Ubiquitin ligases in mitophagy

Selective removal of dysfunctional mitochondria by mitophagy is facilitated by the ubiquitination of mitochondrial proteins and recruitment of autophagy receptors to the mitochondria. Ubiquitin-dependent mitophagy is commonly associated with the PINK1-PRKN mitophagy pathway; however, distinct set of ubiquitin ligases that includes MUL1 (mitochondrial E3 ubiquitin protein ligase 1), SIAH1 (siah E3 ubiquitin protein ligase 1) and ARIH1 (ariadne RBR E3 ubiquitin protein ligase 1) were found to be important for mitochondrial clearance in a PRKN-independent manner (Table 1).

MUL1

In addition to PRKN, several other E3-ubiquitin ligases have been described to participate in mitophagy. MUL1 is integrated with the OMM and contains an amino-terminus RING-finger domain (Figure 3) facing the cytoplasm, where it functions as a ubiquitin ligase or a SUMO (small ubiquitin like modifier) ligase [137,146,147]. MUL1 is also known as mitochondrial-anchored protein ligase [148], mitochondrial ubiquitin ligase activator of NFKB1, growth inhibition and death E3 ligase [149] and Hades [150] and the protein has been implicated in several biological processes, such as mitochondria dynamics, cellular growth, apoptosis and mitophagy [146,148,151,152], mainly by ubiquitination of MFN2, AKT, NFKB, TP53/p53 and ULK1 [146,152]. MUL1 regulates mitophagy by interacting with GABARAP mediated by a LIR (‘327-xxL-330ʹ) motif located in the RING finger domain (Figure 3). Using a yeast two-hybrid screen and the cytoplasmic domain of MUL1 (containing the RING finger domain), four E2 ubiquitin conjugation enzymes (UBE2E2, UBE2E3, UBE2G2, UBE2L3) were found to interact with MUL1. MUL1 conjugation to UBE2E3 was shown to be essential for its GABARAP binding and required the LIR motif localized in the RING finger domain of MUL1. This interaction was specific since no interaction between MUL1 and LC3B was detected [152]. MUL1 participates in selenite-induced mitophagy by recruiting ULK1 and mediating K48-linked ubiquitination of ULK1 at the mitochondria [146,153]. The ULK1 protein levels were directly regulated by MUL1 E3-ligase activity and ULK1 degradation by the ubiquitin-proteasome pathway prevented mitophagy initiation [153]. The activity of MUL1 can also directly regulate mitochondrial morphology by MFN2 ubiquitination and DNM1L SUMOylation, regulating fusion and fission, respectively. As previously mentioned, mitochondrial fission is considered to be important for mitophagy, and it has been shown that MUL1 can SUMOylate DNM1L stimulating mitochondrial fission [147]. MUL1 is by now the only known mitochondrial protein to be specifically targeted to peroxisomes by MDVs [27,148]. MDVs are small mitochondrial-derived vesicles (70–150 nm) that can transport mitochondrial proteins and lipids for degradation by the peroxisome, lysosome or late endosome [27]. MDV formation is independent of DNM1L and can function as an early response mechanism upon mitochondrial damage before mitochondrial removal by mitophagy [27]. This pathway is not fully understood but seems to play an important role in mitochondrial quality control.

SIAH1

SIAH1 is a RING-type E3-ubiquitin ligase (Figure 3) described to participate in a novel PRKN-independent mitophagy pathway by forming a complex with PINK1 and SNCAIP/synphilin-1 (synuclein alpha interacting protein) that does not require PINK1 kinase activity [154]. The recruitment of SNCAIP to the mitochondria is mediated by PINK1, leading to mobilization of SIAH1 to ubiquitinate mitochondrial proteins followed by the recruitment of LC3 and LAMP1-positive structures to the mitochondria to initiate mitophagy. PINK1 mutations associated with PD impaired the mitochondrial recruitment of SNCAIP, indicating a possible relevance of this pathway in the development of PD [154].

ARIH1

ARIH1 mediates PINK1-dependent mitophagy in cancer cells [155]. ARIH1 and PRKN are part of the RING-Between RING-RING family and are structurally very similar (Figure 3) [156]. The main differences rely on the expression patterns of each protein: PRKN is often downregulated in cancer cells and mainly expressed in neuronal cells, whereas ARIH1 is expressed in pluripotent stem cells and cancer cell lines [156]. In lung cancer cells, ARIH1 is activated by PINK1 and regulates mitophagy by poly-ubiquitination of OMM proteins in damaged mitochondria [155]. It has been suggested that ARIH1-dependent mitophagy has a protective effect in lung cancer cells against cisplatin since ARIH1-depleted cells were more susceptible to cisplatin treatment [155].

Concluding remarks

Removal of dysfunctional mitochondria by selective autophagy is essential to maintain a healthy mitochondrial network and cellular homeostasis. There is an increasing number of reports demonstrating that numerous pathways can activate mitophagy in the absence of PRKN both in vitro and in vivo. This highlights the importance of additional mechanistic insight into such regulatory processes for the clearance of damaged mitochondria. It is important to understand how they are coordinated, connected and the extent to which these pathways can complement or even crosstalk with components involved in PINK-PRKN-dependent mitophagy pathways. Here, we have provided a comprehensive overview of different mitophagy pathways not dependent on the PRKN ligase, with a specific focus on the mitophagy receptors involved. A possible interplay between FKBP8, BNIP3, BNIP3L, FUNDC1, and BCL2L13 in mitophagy pathways and mitochondria quality control needs to be unraveled in future studies. Moreover, further insight into regulation of their activity is needed. Several reports indicate that phosphorylation of residues N- or C-terminal to their core LIR motif may regulate the specificity and degree of interaction with LC3 or GABARAP family proteins.

In the past years, notable progress has been made in understanding the regulation of mitophagy on a cellular basis, yet, there are a lot of advances to be made in order to understand how mitophagy is regulated in vivo. Mitophagy signaling mechanisms are complex and vary depending on tissue, development phase, stress or metabolic state. Thanks to the recent development of sophisticated in vivo models using different fluorescent reporters, it has become possible to study mitophagy in living tissues. The integration of different models to study autophagy can provide a broader approach to study different mitophagy mechanisms and help to understand their possible interactions. Ongoing and future research will give us more details about the roles of these pathways in physiological and pathological conditions. One way to approach these questions could be to study different organisms deficient in one or more mitophagy regulators and manipulate them with different stresses and cross them with genetic disease models to figure out the role of a specific pathway in pathological conditions. A better mechanistic understanding of PRKN-independent mitophagy is also required to provide innovative therapeutic targets. Theoretically, it seems logical to enhance mitophagy in diseases where mitochondrial dysfunction is one of the causal elements or to regulate mitophagy when abnormally increased or decreased mitochondrial mass is a pathological factor, but further studies are needed to understand how different regulators control PRKN-PINK-independent mitophagy in different cellular and disease contexts.

Funding Statement

Z.E. is the incumbent of the Harold Korda Chair of Biology. We are grateful for funding from the Israel Science Foundation (Grant #215/19), the Sagol Longevity Foundation, Joint NRF - ISF Research Fund (Grant #3221/19), and the Yeda-Sela Center for Basic Research, as well as the Research Council of Norway through its Centres of Excellence funding scheme (project number 262652 to AS) and FRIPRO (project number 221831 to AS) and the Norwegian Cancer Society (project number 171318 to AS). Research in PB lab is supported by (AEI) and Fondo Europeo de Desarrollo Regional (FEDER) PGC2018-098557-B-I00 Neuroscience Projects from Fundación Tatiana Pérez de Guzmán el Bueno and 2017/BMD-3813 from Comunidad de Madrid. The first authors are supported by a Marie Skłodowska-Curie ETN grant under the European Union’s Horizon 2020 Research and Innovation Programme (Grant Agreement No 765912 DRIVE).

Disclosure statement

No potential conflict of interest was reported by the authors.