Regulation and function of mitophagy in development and cancer

Abstract

Beyond its role in recycling intracellular components nonselectively to sustain survival in response to metabolic stresses, autophagy can also selectively degrade specific cargoes such as damaged or dysfunctional organelles to maintain cellular homeostasis. Mitochondria, known as the power plant of cells, are the critical and dynamic organelles playing a fundamental role in cellular metabolism. Mitophagy, the selective autophagic elimination of mitochondria, has been identified both in yeast and in mammalian cells. Moreover, defects in mitophagy may contribute to a variety of human disorders such as neurodegeneration and myopathies. However, the role of mitophagy in development and cancer remains largely unclear. In this review, we summarize our current knowledge of the regulation and function of mitophagy in development and cancer.

Keywords: :

• Warburg effect
• autophagy
• cancer
• development
• energy metabolism
• mitochondria
• mitophagy
• tumor microenvironment
• tumorigenesis

Introduction

As an evolutionarily conserved and genetically programmed lysosomal degradation pathway, autophagy principally self-digests part of the cytoplasmic constituents to maintain homeostasis in response to various metabolic stresses such as nutrient deprivation, growth factor depletion, and hypoxia.^1,2 Although it was originally considered as a nonselective bulk degradative process, autophagy was recently found to selectively degrade several specific cargos, either damaged or superfluous organelles or protein aggregates, even under nutrient-rich conditions, to maintain cell viability and organelle quality and quantity control.^3,4 To date, several selective types of autophagy have been confirmed, including the cytoplasm-to-vacuole targeting (Cvt) pathway,⁵ pexophagy,⁶ ribophagy,⁷ and xenophagy.⁸

Mitophagy, another well-characterized type of cargo-specific autophagy, is recently capturing more and more attention. As critical and dynamic organelles in eukaryotic cells, mitochondria are primarily responsible for aerobic energy production through oxidative phosphorylation (OXPHOS).⁹ However, accumulation of the by-product reactive oxygen species (ROS) during metabolism is detrimental and inseparably leads to mitochondrial dysfunction. Therefore, it is essential to remove dysfunctional mitochondria in a timely manner so as to maintain mitochondrial turnover and cellular homeostasis. The presence of mitochondria within the double-membrane structure that will become an autophagosome in mammalian cells was reported as early as 1957.¹⁰ Subsequently, the selective elimination of damaged or excessive mitochondria by mitophagy was found to be an important mitochondrial quality control mechanism.^9,11 As mitochondria are the essential organelles indispensable for ATP production and cellular homeostasis, mitochondrial dysfunction has been closely linked to many pathophysiological conditions and diseases. Recent studies have also demonstrated that mitophagy defects are indeed associated with a wide array of human disorders including neurodegenerative diseases,¹² myopathies,¹³ aging,¹⁴ cardiac diseases, and autoimmune diseases.^15,16 Despite recent substantial research on autophagy, the molecular mechanisms with regard to regulating the fate of mitophagy destined for degradation remain poorly understood. In this review, we summarize recent advances on the regulation and biological significance of mitophagy and highlight its complicated role in development and cancer.

Overview on the Regulation of Mitophagy

Mitophagy serves to degrade damaged or dysfunctional mitochondria to match metabolic demand and orchestrate mitochondrial quality and quantity control in cellular homeostasis. The budding yeast Saccharomyces cerevisiae, a key model organism for understanding the molecular aspects of autophagy, has been an ideal system for exploring the regulation of mitophagy. Mitophagy in yeast can generally be induced under two different conditions: nitrogen starvation and culturing cells to the post-logarithmic phase in a medium with a nonfermentable carbon source.¹⁷ Currently, more than 30 autophagy-related (ATG) genes have been identified in yeast, of which the majority have clear mammalian counterparts.¹⁸ Being a selective type of autophagy, mitophagy shares core machinery with canonical starvation-induced autophagy. Accordingly, many Atg proteins also play indispensable roles in mitophagy. Herein we briefly review the roles of the main Atg proteins in the regulation of mitophagy.

Yeast Atg proteins in mitophagy

Recently, two independent groups using a genome-wide screen in yeast for mitophagy-related genes almost simultaneously demonstrated that Atg32 is exclusively devoted to mitophagy rather than nonselective macroautophagy or other types of specific autophagy.^17,19 Under mitophagy-inducing conditions, the N terminus of the receptor protein Atg32 (residues 51–150) interact with the C terminus of the scaffold protein Atg11 to deliver mitochondria into the vacuolar lumen. In addition to Atg11 binding, Atg32 can also interact with Atg8 through a conserved WXXL motif in its cytosolic domain.¹⁹ Therefore, Atg32 mediates mitophagy through direct or indirect interaction with Atg8 (Fig. 1). However, it has been reported that no defects in mitochondrial and vacuolar morphology are detected in ATG32 null cells,¹⁹ indicating that there may be other, yet-unidentified mechanisms for Atg32-independent mitophagy.

Figure 1. The general process of selective autophagic degradation of mitochondria. When cells suffer different kinds of stress, damaged or excessive mitochondria are removed by mitophagy to maintain cellular homeostasis. First, damaged mitochondria are recruited into phagophores. Then the phagophore membranes are elongated and sequester targeted mitochondria to form an autophagosome. Finally, an autophagosome fuses with a lysosome to eliminate the engulfed mitochondria through the action of lysosomal hydrolases.

Strikingly, posttranslational modification such as phosphorylation is implicated in mitophagy induction. Atg32 can be phosphorylated upon mitophagy induction by nitrogen starvation and the phosphorylation of Ser114 and Ser119 is critical for Atg11Atg32 interaction.²⁰ It is worth noting that Hog1 and Pbs2 have been validated to regulate Atg32 phosphorylation, and both Atg32 phosphorylation and mitophagy are severely blocked after depletion of either Hog1 or Pbs2.^20,21 Unfortunately, it remains unclear which kinase downstream of Hog1-Pbs2 can directly phosphorylate Atg32. Moreover, Hog1 is not always activated even under mitophagy-inducing circumstances. Consequently, it was proposed that the initial mitophagy-inducing triggers coming from mitochondria could activate some yet unrecognized kinase(s) to phosphorylate Atg32 at Ser114 and/or Ser119 and facilitate its interaction with Atg11.²⁰

Different from Hog1 and Pbs2 that are specifically dedicated to mitophagy, another mitogen-activated protein kinase (MAPK), Slt2, is responsible for both mitophagy and pexophagy.²¹ Intriguingly, both Hog1 and Slt2 remain in the cytoplasm throughout mitophagy, highlighting that both of them might activate certain undiscovered cytoplasmic targets to eventually phosphorylate Atg32.²¹ Additionally, the timing of their activation is of subtle difference: Slt2 is activated earlier than Hog1, and Slt2 but not Hog1 is essential for recruitment of mitochondria to the mitophagy-specific phagophore assembly site (PAS), indicating that Slt2 might participate in the initiation of mitophagy while Hog1 acts at a later stage.²²

Similar to Atg32, Atg33 was identified from a genomic screen for yeast mutants deficient in mitophagy and is another exclusive mitophagy-specific protein.²³ Notably, deletion of ATG33 obstructs mitophagy almost completely at the post-log phase but reduces starvation-induced mitophagy by only approximately 50%, suggesting that different mechanisms are dedicated to mitophagy induced under different conditions.²³

Mammalian ATG proteins in mitophagy

Mitophagy is highly conserved from yeast to mammals. Consequently, ATGs (in particular, the homologs of the core Atg proteins) are also indispensable for mitophagy in higher eukaryotes. However, the mammalian homologs of Atg32 and Atg33, if they exist, have not yet been identified. Other ATGs required for mitophagy in mammals and their yeast homologs are summarized in Table 1. Atg19, Atg22, Atg25, Atg26, Atg28, and Atg30 do not appear to play any role in mitophagy. Moreover, Atg11, Atg20, and Atg24 participate specifically in the Cvt pathway, pexophagy, and mitophagy, but are not essential for nonselective macroautophagy (for example, Atg11 facilitates, but is not absolutely needed for, nonselective macroautophagy),³⁷ highlighting the differences among the Atg proteins in the regulation of macroautophagy and mitophagy.

Table 1. Mitophagy-related Atg proteins in yeast and their mammalian homologs

Yeast proteins	Mammalian homologs	Location	Functions and actions	Ref.
Atg1	ULK1/2	PAS, cytosol	A serine-threonine kinase, phosphorylation of its binding partner Atg13 at S318	24
Atg2	ATG2A, ATG2B	PAS	Forms a complex with Atg18 and PtdIns3 for autophagosome formation	25
Atg3	ATG3	Cytosol	The E2-like conjugating enzyme for Atg8/LC3	26
Atg4	ATG4A, ATG4B ATG4C, ATG4D	Cytoplasm	A cysteine protease for initial processing and subsequent deconjugation of Atg8/LC3–PE	27
Atg5	ATG5	The convex side of the phagophore	Forms the Atg12–Atg5-Atg16 complex for autophagosome formation	28
Vps30/Atg6	BECN1	PAS	Formz the class III PtdIns3K complex with PIK3R4/VPS15, PIK3C3/VPS34 and ATG14	29
Atg7	ATG7	Cytoplasm	a homodimeric E1-activating enzyme and mitophagy regulator	26, 30, 31
Atg8	LC3/GABARAP	PAS	A phagophore/autophagosome marker; Atg8 interacts with Atg32 through a WXXL motif	19
Atg9	ATG9A, ATG9B	Transmembrane protein	Trafficking between PAS and peripheral sites, mediate the delivery of membrane to form autophagosomes	32
Atg10	ATG10	Cytoplasm	The E2-like conjugating enzyme for Atg12	26
Atg11	-	PAS	A scaffold that facilitates Atg8-Atg32 binding	5, 20
Atg12	ATG12	The convex side of the phagophore	Forms the Atg12–Atg5-Atg16 complex for autophagosome formation	28
Atg13	ATG13	PAS Cytosol	The interacting partner of Atg1, a regulatory subunit of the Atg1 kinase complex	24
Atg14	ATG14	Phagophore	Forms the PtdIns3K complex with Vps15, Vps34 and Vps30	29
Atg15	-	Vacuole	A lipase transported to vacuoles for intravacuolar lysis of autophagic and Cvt bodies	33
Atg16	ATG16L1, ATG16L2	The convex side of the phagophore	Forms the Atg12–Atg5-Atg16 complex for autophagosome formation	28
Atg17	-	PAS	Scaffold protein of the Atg17-Atg31-Atg29 complex and a regulatory subunit of the Atg1 kinase complex	34
Atg18	WIPI1/2	PAS	Forms a complex with Atg2 and PtdIns3P for autophagosome formation	25
Atg20	-	PAS	Required for rapamycin- induced mitophagy	35
Atg21	-	PAS	Controversial views of its implication in mitophagy	19, 23
Atg23	-	PAS and other peripheral sites	Atg9 transport factor, shuttles between the PAS and peripheral sites	36
Snx4/Atg24	-	Cytoplasm	Required for rapamycin- induced mitophagy	35
Atg27	-	Transmembrane cycling protein	Atg9 transport factor, shuttles between the PAS and peripheral sites	32
Atg29	-	PAS	Forms the Atg17-Atg31-Atg29 complex for PAS organization and autophagy	34
Atg31	-	PAS	Forms the Atg17-Atg31-Atg29 complex for PAS organization and autophagy	34
Atg32	-	OMM	Mitophagy receptor, interacts with Atg11 and Atg8	17, 19
Atg33	-	OMM	Mitophagy induction at post-log phase	23, 180

Other proteins important in the regulation of mitophagy

Uth1

Initially identified in a genetic screen as a protein implicated in life-span determination,³⁸ Uth1 also participates in the regulation of mitophagy.³⁹ The Uth1 protein is a member of the SUN gene family (SIM1, UTH1, and NCA3), mainly anchored on the outer mitochondrial membrane (OMM) and is responsible for mitophagy induced by rapamycin or nitrogen starvation.³⁹ Intriguingly, there are two distinct Uth1-related mitochondrial autophagic pathways: Uth1-dependent and Uth1-independent mitophagy. The first one is a selective process of mitophagy that cannot occur in the absence of Uth1. The second one is a nonselective mitophagy present in both wild-type and UTH1-deleted strains, in which mitochondria are engulfed nonexclusively with a large amount of surrounding cytosol.⁴⁰

Ptc6/Aup1

Besides Uth1, Ptc6/Aup1, a member of the protein phosphatase 2C superfamily, is required for mitophagy in stationary phase yeast and plays a prosurvival role. It was identified in a protein phosphatase homolog screen aiming to capture proteins that can functionally interact with the autophagy-dedicated protein kinase Atg1.^41,42 Notably, there appear to be different roles for Uth1 and Ptc6 in mitophagy, although both of them are required for this process: UTH1 knockout promotes the survival of cells cultured in nitrogen-deprivation medium with lactate but not glucose, whereas knockout of PTC6 leads to decreased cell survival in long-term stationary phase cultures even under sustained gluconeogenic conditions, suggesting that distinct regulatory networks might be involved in the process of mitophagy mediated by these proteins.⁴¹ Therefore, further investigation will be needed to shed light upon the possible relationship between these mitophagy-associated proteins and their detailed contributions in the regulation of mitophagy.

PINK1 and PARK2

The best-studied mitophagy-related pathway in mammals is mediated by PINK1 (PTEN induced putative kinase 1) and PARK2/Parkin. Loss-of-function mutations in the PINK1 and PARK2 genes have been substantiated to significantly cause the early onset of autosomal recessive Parkinson disease.⁴³ PINK1 and PARK2 encode a mitochondrially targeted Ser/Thr kinase and a cytosolic RING-domain containing E3 Ub ligase, respectively. Genetic analysis of PINK1 and PARK2 in Drosophila melanogaster indicates that they work together to regulate mitochondrial integrity and suppress mitochondrial damage.^44,45

PARK2 also regulates mitochondrial morphology and integrity in mammalian cells and is widely expressed in many tissues including brain, skeletal muscle, heart, and liver.^46,47 PARK2 translocates to depolarized mitochondria in cells treated with carbonyl cyanide m-chlorophenyl hydrazone (CCCP), a mitochondrial uncoupler that dissipates the mitochondrial membrane potential to induce mitochondrial damage.⁴⁶ Other mitochondrial poisons such as paraquat and azide can also induce the translocation of PARK2 to damaged mitochondria,^46,47 which is required for depolarization-mediated fragmentation.⁴⁸ In addition, a novel role of PARK2 as a tumor suppressor has been verified in breast and ovarian cancer, highlighting that PARK2-dependent mitophagy might be associated with cancer biology.⁴⁹

Translocation of PARK2 to impaired mitochondria and induction of mitophagy are indispensably required for PINK1 activity even in the absence of mitochondrial uncoupling.^3,50 PINK1 locates either in the mitochondrial intermembrane space (IMS) or OMM^51,52 and may be maintained at a low expression level in healthy mitochondria due to two potential mechanisms; either rapid proteolytic destabilization by proteases such as PMPCA (peptidase [mitochondrial processing] α)⁵³ and the inner membrane PARL (presenilin associated, rhomboid-like),⁵⁴ or import to the inner mitochondrial membrane through the translocase complexes of the outer and inner mitochondrial membrane (TOMM and TIMM, respectively).⁵⁵ Once the mitochondrial membrane potential is decreased, proteolysis or import of PINK1 is blocked so that it is stabilized and accumulated on the OMM where it recruits PARK2 to induce mitophagy.⁵² However, some conflicting results have recently reported that PINK1 deficiency may contribute to mitochondrial fragmentation,^56,57 and knockdown of PINK1 with RNA interference induces mitochondrial fragmentation and autophagy in neuronal SH-SY5Y cells, whereas overexpression of wild-type PINK1 suppresses mitophagy,⁵⁷ consistently supporting the opposite view that inhibition of PINK1 activity may promote mitophagy. These paradoxical findings are likely due to different testing times and/or different organisms used in the studies. Following translocation, PARK2-mediated ubiquitination of a subset of mitochondrial substrates accelerates mitophagy.^58,59 PARK2 polyubiquitinates itself and its mitochondrial substrates predominantly at Lys63 residues, producing the docking site for the Ub-binding adaptor SQSTM1/p62 to trigger autophagic degradation by binding to the LC3-interacting region (LIR) motif of LC3/GABARAP (GABA(A) receptor-associated protein).⁶⁰ Surprisingly, some divergent results propose that SQSTM1 deletion in mouse embryonic fibroblasts facilitates rather than hinders mitochondrial degradation in a PARK2-dependent manner,⁶¹ and recruitment of SQSTM1 to mitochondrial-anchored Ub is not sufficient to promote mitophagy,⁶² implying that SQSTM1 is not indispensable for mitophagy. The most plausible explanation for this discrepancy is the so-called “SQSTM1 functional redundancy”:⁶² Some other intracellular proteins may substitute for SQSTM1 during mitophagy when it is lost or the corresponding gene is deleted. For instance, HDAC6 was recently identified as another Ub-binding autophagic component for mitochondrial elimination and autophagosome–lysosome fusion.^63,64 Specially, some PARK2 substrates can be degraded through a Lys63-linked ubiquitination-independent manner. For instance, ubiquitination of MFN1 (mitofusin 1) and MFN2 is mediated by PARK2 upon membrane depolarization and contributes to degradation in a proteasome-dependent manner via linking with Lys48 polyubiquitin chains⁵⁹ or in a VCP/p97-dependent manner⁶⁵ and VDAC1 (voltage-dependent anion channel 1) is degraded by the autophagy–lysosomal pathway through recruiting SQSTM1 with Lys27-linked ubiquitin (Ub) chains.⁶⁶

The molecular mechanism underlying how PINK1 recruits PARK2 for the subsequent induction of mitophagy remains obscure. It is plausible that PINK1 phosphorylates cytosolic PARK2 with its kinase domain that faces the cytoplasm.⁵¹ Indeed, PINK1-triggered phosphorylation of two conserved serine/threonine residues, T175 and T217 in human PARK2, and its intact RING1 domain are crucial for its translocation to mitochondria.⁶⁷ Moreover, S65 in the PARK2 Ub-like domain was recently found to be directly phosphorylated in a PINK1-dependent manner upon the depolarization of mitochondria.⁶⁸ Intriguingly, several lines of evidence recently indicate that there may exist a crosstalk between PINK1-PARK2-dependent mitophagy and mitophagy receptor-mediated pathways.⁶⁹ For example, BNIP3L/NIX (BCL2/adenovirus E1B 19 kDa interacting protein 3-like) appears to promote CCCP-induced mitochondrial depolarization and PARK2 translocation to mediate mitophagy in mouse embryonic fibroblasts,⁷⁰ and the newly identified mitophagy receptor FUNDC1 (FUN14 domain containing 1) in mammalian cells may be associated with FCCP-mediated mitophagy,⁷¹ suggesting that identification of more cofactors and coregulators implicated in PINK1-PARK2-mediated mitophagy may be beneficial to further pursue the mechanism of this pathway. Other mitophagy inducers, receptors and their interacting partners and related regulators are listed in Table 2.

Table 2. Mitophagy inducers, receptors, and their interacting partners and related regulators

Yeast proteins	Mammalian homologs	Location	Functions and actions	Ref.
-	ALOX15	-	A key enzyme in mitophagy during reticulocyte maturation	72
-	AMBRA1	Cytoplasm	Mitophagy of depolarized mitochondria in a PARK2-dependent manner	73
Bck1	-	Cytoplasm	MAPK kinase kinase for Pck1 activation	23, 74
-	BNIP3	Mitochondria and ER	Mitophagy receptor Mitophagy regulator in renal tubular cells in acute kidney injury	75, 76
-	BNIP3L	OMM	Mitophagy receptor, interacts with LC3/ GABARAP	3, 77-81
Cdc37	CDC37	Cytosol	The kinase-specific co-chaperone of HSP90	24
-	Ceramide	-	A novel mitophagy receptor for lethal mitophagy and tumor suppression	82
-	FUNDC1	OMM	An integral OMM protein, a receptor for hypoxia-induced mitophagy	71
-	HDAC6	Nucleus	Ub-binding protein, mediates PARK2-induced mitophagy	63, 64
-	HMGB1	Nucleus	Regulator of mitochondrial function and morphology	83
Hog1	MAPK/p38	Cytosol	MAPK for Atg32 phosphorylation	20, 21, 84
Hsp60	HSPD1	Mitochondria, cytoplasm	A mitochondrial matrix chaperonin	85
Hsp90	HSP90	Cytosol	Regulation of ULK1- and ATG13-mediated mitophagy by forming HSP90-CDC37 chaperone complex	24
-	HSPB1	Nucleus	Downstream mediator of HMGB1, required for mitochondrial quality control	83
-	PARK2	Cytosol	E3 Ub ligase, ubiquitinates OMM proteins such as MFN1, MFN2 and VDAC1	46, 50, 66
Pbs2	-	IMS	Mitophagy in yeast during stationary phase, cell survival	41
-	Phen	-	Induction of mitophagy in a DNM1L-dependent manner	86
-	PINK1	IMS or OMM	Ser/Thr protein kinase, induces mitophagy by recruiting PARK2 to damaged mitochondria	51, 52, 87
Por1	VDAC1	OMM	A target for PARK2-mediated ubiquitination and mitophagy	62, 66
-	RIPK2	Cytosol	Regulation of mitophagy by phosphorylating ULK1	184
-	Salinmycin	-	Induces activation of mitophagy	178
Sln1	-	Plasma membrane	Upstream sensor and negative regulator of Hog1signaling	21
Slt2	-	Cytosol	MAPK for mitophagy and pexophagy	21
-	SQSTM1	Cytosol	Autophagic receptor, promotes PARK2-mediated mitophagy as a Ub-binding protein	49, 53
-	TNF	-	Induction of mitophagy in mouse macrophages	172
Tom20	TOMM20	OMM	A major component of the translocase of the OMM complex, a major mitochondrial protein import receptor	180
Uth1	-	OMM	Mitophagy induction by rapamycin or nitrogen starvation	26, 27
-	VCP/p97	Mitochondria	AAA-ATPase, promotes the degradation of OMM proteins and mitophagy	52
Whi2	-	Cellular component	Stress response protein, interacts with Atg8 to induce mitophagy	169
Wsc1	-	Cell surface	Upstream sensor of Slt2 signaling	21

Collectively, an increasing number of regulators have been recognized to play crucial roles in mitophagy, raising some pending questions as to whether different regulators actually regulate mitophagy functionally depending on context-specific conditions and distinct initiating signals, or all regulators ultimately converge at a canonical model of mitochondrial degradation, or even if these differences are only attributed to different experimental procedures.

Mitophagy in Development

Apart from degradation of damaged mitochondria, removal of nondamaged or superfluous mitochondria occurs as a pivotal process for organ and tissue development. Thus, mitophagy during some specialized developmental stages is deemed as a critical quantity control mechanism for maintaining the proper amount of mitochondria.

Mitophagy in early embryogenesis

One intriguing role of mitophagy in development is selective degradation of paternal mitochondria in fertilized oocytes. Mitochondrial DNA (mtDNA) is maternally inherited in almost all eukaryotes despite the fact that, at least in some organisms, sperm-derived paternal mitochondria enter into the oocyte cytoplasm upon fertilization.⁸⁸ Indeed, several studies in Caenorhabditis elegans embryos have recently demonstrated that fertilization-triggered autophagy is essential for removal of paternal mitochondria and maintenance of maternally inherited mtDNA.^89-91 After fertilization, this induction of mitophagy is triggered by penetrating sperm components.^89,90 Importantly, such degradation seems to be associated with protein ubiquitination in mammals.^92,93 PHB (prohibitin), a highly conserved inner mitochondrial membrane protein, can be ubiquitinated by the Ub-proteasome system to recognize paternal mitochondria and facilitate their degradation.^94,95 However, its counterpart in C. elegans cannot be ubiquitinated. Instead, the nematode-specific sperm membranous organelles are ubiquitinated and removed by autophagic machinery together with paternal mitochondria.⁹⁰ Nevertheless, it remains unknown whether paternal mitochondria are incorporated into autophagosomes for degradation by virtue of their close association with ubiquitinated membranous organelles, or mitophagy is able to degrade nonubiquitinated paternal mitochondria.

Mitophagy in erythrocyte differentiation

In mammals, terminal differentiation of reticulocytes during erythropoiesis requires enucleation (the expulsion of the nucleus) and organelle removal (such as removal of mitochondria by mitophagy).⁹⁶ BNIP3L is required for programmed mitochondrial clearance since its expression is highly upregulated during reticulocyte maturation and bnip3l^−/− mice retain their mitochondria in circulating reticulocytes.^77-80 Like wild-type reticulocytes undergoing maturation, autophagy can also be induced in bnip3l^−/− reticulocytes. However, the mitochondria are not eliminated. This failure to eliminate mitochondria reveals that they are not properly targeted to autophagosomes in the absence of BNIP3L.

Surprisingly, although BNIP3L was originally thought to be a BH3-only member of the pro-apoptotic BCL2 (B-cell CLL/lymphoma 2) family essential in the regulation of programmed cell death, some proapoptotic proteins including BAX, BAK1, BCL2L1, BCL2L11/BIM, or BBC3/PUMA are not involved in BNIP3L-mediated mitophagy, implying that it may be independent of the proapoptotic pathways during reticulocyte differentiation.^30,77,78

Indeed, BNIP3L has been verified as a mitophagy receptor in mammalian cells.⁸⁰ It is anchored at the OMM and contains a WXXL motif near its N terminus facing the cytosol, which binds to the mammalian homolog of Atg8, LC3, and the LC3 homolog GABARAP.^3,80,81 The WXXL (35–38) motif is indispensable for BNIP3L-LC3/GABARAP interaction and BNIP3L-mediated mitophagy, although this interaction can be enhanced under mitochondrial stress conditions such as treatment with rotenone or CCCP.⁸⁰ Interestingly, depolarizing mitochondria with the uncoupling agent FCCP, or the BH3 mimetic ABT-737, restores mitophagy even in bnip3l^−/− reticulocytes,⁷⁹ demonstrating that mitochondrial depolarization-dependent mitophagy functions through BNIP3L-independent pathways. However, the initiating signals for mitophagy and the relationship between depolarization- and BNIP3L-mediated mitophagy during reticulocyte maturation remain poorly understood.

A mammalian homolog of the yeast mitophagy receptor, Atg32, has not been identified; however, several common characteristics are shared between Atg32 and BNIP3L, suggesting that BNIP3L might be the equivalent of Atg32 in mammals. First, both of them are OMM proteins with their N-terminal domains in the cytosol and C-terminal regions in the IMS. Second, both of them contain a WXXL motif for binding to Atg8 or LC3/GABARAP. Nevertheless, since the mammalian counterpart of Atg11, which interacts with Atg32 as a scaffold protein, remains undiscovered, it is still an open question as to how BNIP3L functions to regulate mitophagy in mammalian cells.

In addition to BNIP3L, ULK1 (Unc51 like autophagy activating kinase 1)⁹⁷ and Atg7^30,96 are also implicated in mitophagy during reticulocyte maturation since Ulk1-deficient mice show delayed mitochondrial clearance, although this defect can be rescued by treatment with CCCP in vitro, emphasizing that there are other Ulk1-independent mechanisms to compensate for mitophagy in Ulk1-deficient reticulocytes. Similarly, there is an accumulation of damaged mitochondria in atg7^−/− mice erythrocytes, eventually leading to differentiation failure and cell death, indicating that ATG7 is a critical mitophagy regulator in mammalian hematopoietic cells.⁹⁶ However, a recent study showed that mitophagy is partially impaired but not completely blocked in atg7^−/− reticulocytes, and that ATG7 deficiency has no effect on the expression of BNIP3L and vice versa.³⁰

More regulators of mitophagy in reticulocytes are being discovered. For example, inhibition of ALOX15 (arachidonate 15-lipoxygenase) delays mitophagy in reticulocytes.⁷² In addition, an evolutionarily conserved KRAB-TRIM28/KAP1-miRNA regulatory cascade controls mitophagy in reticulocytes.⁹⁸ Once this cascade is disrupted, the expression of mitophagy-associated genes is repressed so that mitochondria are retained in erythroblasts and the host animal displays severe hypoproliferative anemia.

Mitophagy in adipose tissue differentiation

Several studies have underscored that autophagy is closely associated with adipose tissue development and differentiation.^99-101 Based on their different colors, morphology, and functions, adipose tissues can be classified into white adipose tissue (WAT) and brown adipose tissue (as well as a more recently characterized beige adipose tissue). Recently it has been reported that alterations in structures and numbers of mitochondria occur in the differentiation of WAT. A dramatic increase in mitochondrial biogenesis is observed in the early stage of adipogenesis, whereas the majority of mitochondria are scattered around a few lipid droplets and disappear in the late stage. Importantly, atg5^−/− and atg7^−/− mouse embryonic fibroblasts or pre-adipocytic 3T3-L1 cells with either Atg5 or Atg7 knockdown fail to eliminate mitochondria and complete this differentiation process,^99-101 implying that mitophagy is indispensable for adipocyte differentiation. Consistently, the white adipocytes in adipose-specific Atg7 conditional knockout mice show abnormal morphological characteristics such as an increased number of mitochondria with small lipid droplets. However, about half of the white adipocytes in atg7^−/− mice display normal morphology, indicating that other Atg7-independent autophagic pathways are also involved in mitophagy during WAT differentiation.¹⁰¹ Of note, Atg7 conditional knockout mice display reduced adiposity with highly diet-induced obesity resistance and insulin sensitivity, decreased concentration of blood triglycerides and cholesterol, and higher basal metabolic rate.¹⁰¹ Therefore, novel therapeutics targeting mitophagy may be appropriate for patients with metabolic diseases such as obesity and type II diabetes.

Mitophagy in T lymphocyte differentiation

Mitophagy is also necessary for the clearance of superfluous mitochondria in T lymphocyte differentiation and survival. Both Atg5 and Atg7 are crucial for this process in mice.^102,103 Remarkably, mitochondrial content is developmentally regulated only in T rather than B lymphocytes and the decrease in mitochondrial content marks T lymphocyte maturation from precursor thymocytes. Accordingly, autophagy-deficient T lymphocytes are unable to reduce their mitochondrial content and are eventually susceptible to apoptosis largely due to increased production of ROS.^103,104

Mitophagy in Human Cancer

Mitochondria are the critical manipulators for regulating cell energy metabolism and cell life/death decisions; thus, alterations in mitochondrial homeostasis have been implicated in the pathogenesis of many human diseases including cancers, myopathies, neurodegenerative diseases¹² such as PD,^105,106 Alzheimer disease,¹⁰⁷ and Huntington disease.^106,108 The relevance of mitophagy to the development of human diseases such as neurodegeneration has been elegantly reviewed. Herein we will focus on the implications of mitophagy in cancer.

Overview of mitophagy in cancer development

Autophagy has always been perceived to play a double-faceted role in tumorigenesis; either supporting survival or promoting death, depending on the different cellular contexts. Generally, basal autophagy remains at a low level to maintain cellular homeostasis and sustain prolonged survival by degradation of polyubiquitinated or aggregated proteins and damaged organelles. However, unrestrained autophagy will eventually trigger cell death when the cellular metabolic balance is disturbed.¹⁰⁹ Similarly, mitophagy follows this Janus-faced nature of autophagy. As the main powerhouse in eukaryotic cells, mitochondria are crucial for cell survival. The mtDNA exhibits a high susceptibility to damage attributed to heavy exposure to ROS, restricted repair systems, and lack of histone protection.¹¹⁰ Thus, mitophagy serves to remove dysfunctional mitochondria to alleviate oxidative stress and prevent carcinogenesis. Conversely, mitophagy can protect cells from apoptosis or necrosis and promote tumor cell survival under some adverse conditions such as poor nutrient supply and hypoxic stresses.^111,112 Therefore, mitophagy emerges as a key quality control factor and decision maker in cancer cells. In the following sections, we will summarize the current understanding of the biological functions of mitophagy during cancer development, provide an overview of how mitophagy is regulated by cancer-associated signaling pathways and proteins, and discuss the possibilities of targeting mitophagy for cancer intervention.

Regulation and function of mitophagy in the tumor microenvironment

By producing approximately 36 ATPs per molecule of glucose, OXPHOS is primarily used by normal cells for supplying energy. Under anaerobic conditions, however, glycolysis is used as the main source of energy production, but generates only 2 ATPs per molecule of glucose.¹¹³ However, cancer cells predominantly utilize glycolysis but not OXPHOS to generate energy even in the presence of oxygen, which is known as the “Warburg effect.”^114,115 In the past several decades, the regulation and relevance of the Warburg effect with regard to cancer has been extensively studied. Otto Warburg in his seminal report first proposed the role of aerobic glycolysis in cancer and attributed this phenomenon to dysfunctional mitochondria.^114,115 However, later studies found little evidence to support mitochondrial dysfunction in cancer cells. This Warburg paradox has hindered advances in this field for many years. Recently, a new model of tumor metabolism termed “two-compartment tumor metabolism” or “parasitic cancer metabolism” proposed that autophagy/mitophagy and aerobic glycolysis indeed physically converge in the tumor microenvironment.^116,117 According to this model, ROS produced by cancer cells are transferred to adjacent cancer-associated fibroblasts or other stromal cells and initiate the onset of a stromal oxidative stress responses including mitophagy. The consequent mitochondrial dysfunction in the stromal cells leads to the production of sufficient high-energy metabolites such as L-lactate, ketones, glutamine, and free fatty acids to support cancer cell survival. This kind of “host-parasite” relationship formed between tumor stromal cells and epithelial cancer cells was coined as the “reverse Warburg effect.”^118,119

It has been previously shown that loss of stromal CAV1 is a powerful predictive biomarker of tumor progression and metastasis in human cancers.^120,121 ROS released from cancer cells can induce the loss of stromal CAV1, which in turn induces metabolic reprogramming of tumor stromal cells with increased mitophagy and mitochondrial dysfunction.^122,123 Notably, a recent miRNA profiling of cav1^−/− stromal cells revealed the upregulation of two key cancer-related miRNAs, MIR31 and MIR34C, both of which are sufficient to drive mitophagy due to their close association with the acknowledged mitophagy inducers oxidative stress and activation of the hypoxia response/HIF1A, respectively.¹²⁴ This finding suggests that miRNA regulation provides new mechanistic insights into the role of mitophagy during tumorigenesis and is attracting increasing attention. Furthermore, since CAV1 can negatively regulate TGFB (transforming growth factor, β 1) signaling, activation of TGFB signaling seems to be important for the induction of mitophagy in tumor stromal cells.¹²⁵ Indeed, ligand-dependent paracrine or cell–autocrine activation of TGFB signaling in stromal cells rather than in cancer cells induces metabolic reprogramming of the tumor microenvironment.¹²⁶

Additionally, hydrogen peroxide produced by BRCA1-null ovarian cancer cells can also trigger mitophagy by activating NFKB (nuclear factor of kappa light polypeptide gene enhancer in B-cells) signaling in stromal cells.¹²⁷ As a well-recognized tumor suppressor gene, BRCA1 normally suppresses tumor growth by maintaining genome integrity. However, BRCA1-null cancer cells exert an influence on the metabolic reprogramming of their adjacent tumor-associated stromal fibroblasts by producing a large amount of hydrogen peroxide that activates NFKB signaling to induce mitophagy and glycolysis in tumor stromal cells.¹²⁷ In addition to ROS, some cytokines or other bioactive factors enriched in the tumor-permissive microenvironment such as migration stimulating factor, a genetically truncated N-terminal isoform of fibronectin, can also induce mitophagy in tumor stromal cells by activating TGFB and CDC42-NFKB signaling.¹²⁸

Collectively, mitophagy can be activated by oncogenic signaling pathways, mainly including the TGFB and NFKB pathways, to boost tumor cell growth through reprogramming cancer cell metabolism. Therefore, inhibition of mitophagy in the tumor stroma by neutralizing ROS or inactivating TGFB and NFKB signaling in the tumor microenvironment via metabolically uncoupling tumor cells from their surrounding and supportive stroma may be feasible strategies to inhibit tumor growth, which provides useful hints into targeting mitophagy as a novel and potential anticancer strategy (Fig. 2).

Figure 2. Two-compartment tumor metabolism: Mitophagy in tumor stromal cells supports the metabolism of cancer cells. Cancer cells secrete ROS or cytokines to stimulate mitophagy in tumor stromal cells by the activation of TGFB and NFKB signaling. The consequent mitochondrial dysfunction in the stromal cells leads to the production of sufficient high-energy metabolites such as L-lactate, ketones, glutamine, and free fatty acids to support cancer cell survival. CAFs, cancer-associated fibroblasts.

Regulation and function of mitophagy in cancer cells

RAS-MAPK signaling

Mounting reports have confirmed the dysregulation of mitophagy in cancer cells, in which many oncogenic signalings and proteins are implicated in the regulation of mitophagy. The first oncogene isolated from human tumors, RAS, is frequently and aberrantly activated in many human cancers.^129-131 In addition to its well-known transforming potential, oncogenic RAS has been recently found to activate mitophagy as a critical survival strategy through expediting glycolysis to conquer a cellular energy deficit resulting from glucose deficiency.¹³² Blockage of mitophagy effectively suppresses RAS-induced transformation, further demonstrating the protumorigenic role of mitophagy. Activated RAS can potently induce the expression of glucose transporters such as SLC2A1/GLUT1 and promote the switch of glucose metabolism from OXPHOS into glycolysis by epigenetically reprogramming gene expression.¹³³ Once glycolysis is activated by oncogenic signaling to meet the cellular energy requirement, maintenance of a high amount of mitochondrial mass is no longer essential for ATP production. Therefore, degradation of relatively superfluous mitochondria can be implemented by mitophagy so as to expedite glycolysis and resupply nutrients. Indeed, autophagy facilitates glycolysis during RAS-mediated oncogenic transformation.¹³⁴ In addition, transformed cells may maintain small numbers of mitochondria during rapid proliferation, minimizing the energy requirement to maintain subcellular organelles.¹³²

Remarkably, activated RAS promotes mitophagy through MAPK/JNK signaling.¹³² Inhibition of JNK rather than other MAPKs effectively prevents KRAS-induced mitophagy, restores mitochondrial functions, and overcomes energy deficit in RAS-transformed and other cancer cells. Since RAS-JNK signaling can activate mitophagy in the absence of any well-recognized inducers of mitophagy, further studies are required to elucidate the detailed mechanisms of RAS-induced mitophagy. Interestingly, the SQSTM1 protein level is dramatically induced by activated RAS and this is not a result of impaired autophagic turnover because lysosomal inhibitors further enhance the accumulation of SQSTM1.¹³⁵ It remains largely unexplored whether activated RAS induces SQSTM1 through the activation of JNK signaling. Alternatively, SQSTM1 could be upregulated in response to the initiation of mitophagy induced by RAS-JNK signaling.

AKT2 signaling

The aberrant activation of phosphoinositide 3-kinase-AKT signaling has been well documented as contributing to tumor promotion and progression.¹³⁶ As a serine/threonine protein kinase, AKT functions to promote cellular proliferation and survival of cancer cells by regulation of downstream signaling molecules such as MTOR and CCND1/cyclin D1. Interestingly, three AKT isoforms, AKT1, 2 and 3, have distinct localizations in human cells, raising the possibility that each isoform may have unique functions and employ different regulatory mechanisms. Among them, AKT2 is the only one localized in mitochondria in a variety of cancer cell lines,¹³⁷ indicating a potential link between AKT2 and mitochondrial functions during cancer development. Indeed, specific ablation of AKT2 rather than AKT1 or AKT3 has profound effects on mitochondria in MDA-MB231 breast cancer cells.¹³⁸ Initially, AKT2 ablation increases the volume of mitochondria by activation of PPARGC1A to upregulate mitochondrial biogenesis. However, AKT2 ablation also markedly inhibits the MTOR–RPS6KB pathway to limit protein synthesis. Therefore, the high activity of mitochondrial biogenesis cannot be sustainable for an extended duration, and the persistent inhibition of AKT2 signaling eventually activates mitophagy owing to the collapse of the entire mitochondrial homeostasis system. Alternatively, mitophagy is activated to salvage the mitochondria for enabling cellular survival.¹³⁸ In addition to mitophagy activation, significant growth inhibition is also induced by AKT2 ablation, indicating that AKT2 might be a promising target for cancer therapy. Actually, selective inhibitors for distinct AKT isoforms have been developed and will soon be studied in clinical trials.

BECN1 complex

The first direct and functional link between autophagy and tumor suppression derived from genetic studies of the mammalian homolog of yeast Vps30/Atg6, BECN1/Beclin 1, which was identified as a haploinsufficient tumor suppressor that is monoallelically deleted in a majority of human cancers including breast, ovarian, and prostate cancers.¹³⁹ Similar to Vps30 which is an essential autophagy regulator in yeast, BECN1 has the ability to promote autophagy in mammalian cells. Intriguingly, the autophagy-promoting role of BECN1 leads to the inhibition of cell proliferation, clonigenicity, and tumorigenesis both in vitro and in vivo. Furthermore, expression of BECN1 is markedly reduced in human breast carcinoma cell lines.¹³⁹ Consistently, it has been supported that the heterozygous disruption of Becn1 and/or lacking one copy of Becn1 makes mice susceptible to tumorigenesis and/or they suffer from spontaneous tumors such as B cell lymphoma, hepatocellular carcinoma, and lung adenocarcinoma.^140,141 These observations highlight the role of BECN1 as a tumor suppressor. Although the mechanism of tumor suppression by autophagy has not yet been fully determined, it seems to be contradictory to the fact that autophagy sustains cell survival in response to stress conditions. The most plausible explanation is that autophagy-defective cells due to monoallelic loss of BECN1, for instance, may produce a great deal of ROS that accelerates DNA damage or chromosomal instability, leading to the accumulation of oncogenic genetic changes.^142-144 Thus, autophagy-mediated tumor suppression is associated with the elimination of damaged mitochondria and peroxisomes, namely through mitophagy and pexophagy.

Notably, BECN1 regulates autophagy by forming multimolecular complexes containing various autophagy-related proteins such as PIK3C3, PIK3R4, UVRAG (UV radiation resistance associated), SH3GLB1 (SH3-domain GRB2-like endophilin B1), ATG14,¹⁴⁵ and AMBRA1 (autophagy/Beclin-1 regulator 1).¹⁴⁶ BECN1 exerts its function of autophagy induction by interacting with PIK3R4 and PIK3C3, the latter of which corresponds to the lipid kinase component of the class III phosphatidylinositol (PtdIns) 3-kinase (PtdIns3K) complex that is regulated by PIK3R4.¹⁴⁷ Similar to BECN1, UVRAG is another candidate tumor suppressor, which is monoallelically deleted in human cancers, and its tumor suppressor activity is dependent on its interaction with BECN1.¹⁴⁸ Structurally, UVRAG acts as a bridge to mediate BECN1-SH3GLB1 interaction. SH3GLB1 is a member of the membrane curvature-driving endophilin family, which regulates the post-Golgi trafficking of membrane-integrated ATG9A for autophagy.^149,150 In humans, the SH3GLB1 gene is located on chromosome 1p22, a region frequently deleted in neoplasms¹⁵¹ and its expression is downregulated in a wide variety of cancers.^152-156 Furthermore, inhibition of endogenous expression of SH3GLB1 by RNA interference enhances the tumorigenecity of HeLa cells and SH3GLB1-knockout mice are susceptible to the development of spontaneous tumors, indicating that SH3GLB1 is indeed a novel bona fide tumor suppressor.^150,157 Notably, the tumor suppressor function of SH3GLB1 seems to be associated with the induction of autophagosome formation.¹⁵⁰ SH3GLB1 was recently verified to mediate the clearance of damaged mitochondria by mitophagy to alleviate DNA damage and chromosomal instability in Myc-induced lymphoma cells, highlighting the tumor suppressive role of mitophagy.¹⁵⁸ Accordingly, hemizygous deletion of SH3GLB1 results in an increase in mitochondrial mass, accumulation of DNA damage, reduced apoptosis activation, and consequent tumor development in vivo.¹⁵⁹

AMBRA1 is a WD40 domain-containing protein crucial for positive regulation of autophagy and nervous system development in mammals.¹⁶⁰ As a component of the BECN1 complex, AMBRA1 promotes autophagosome formation by strengthening the interaction of BECN1 with PIK3C3. Both BECN1 and PIK3C3 are two core components of the class III PtdIns3K complex crucial for phagophore nucleation. Moreover, both AMBRA1 and BCL2 bind to BECN1 at the same site. Consequently, it is inevitable that there is a competitive relationship between the binding of BECN1 with AMBRA1 and BCL2. Consistent with this assumption, AMBRA1 can indeed compete with BCL2 both in mitochondria (mito-BCL2) and the endoplasmic reticulum to bind BECN1. Under normal conditions, AMBRA1 can be preferentially docked by the pool of mito-BCL2 to inhibit its autophagic function. Upon autophagy induction, AMBRA1 is released form mito-BCL2 to enhance its binding to BECN1 and reinforce the autophagy-promoting role of BECN1.¹⁶¹ Although the exact role of AMBRA1 in tumorigenesis is still largely unclear, it is conjectured on the one hand that AMBRA1 dysregulation may promote carcinogenesis by virtue of its functional interaction with the BECN1 complex. Indeed, functional deficiency of AMBRA1 in mouse embryos results in excessive cell proliferation and apoptotic cell death.¹⁶⁰ On the other hand, it is hypothesized that AMBRA1 might be indirectly implicated in tumorigenesis by balancing cell death/survival owing to its interaction with the anti-apoptotic factor BCL2 to modulate the activity of proapoptotic and proautophagic regulators.^161,162 Therefore, AMBRA1 acts as a switch between autophagy and apoptosis during cancer development.

Strikingly, it has been recently reported that AMBRA1 might be a novel PARK2-binding protein implicated in mitophagy.¹⁴⁶ Mechanistically, prolonged mitochondrial depolarization markedly enhances the interaction of endogenous PARK2 and AMBRA1, which is recruited to the perinuclear clusters of depolarized mitochondria by PARK2 and activates the class III PtdIns3K complex in the vicinity of those damaged mitochondria and eventually leads to mitophagy. Actually, PARK2 stimulates the perimitochondrial nucleation of new phagophores to boost mitophagy through recruitment of AMBRA1. Consequently, AMBRA1 is indispensable for the final step of PARK2-triggered mitophagy rather than the translocation of PARK2 to depolarized mitochondria.¹⁶³ These findings unveil a novel mechanism by which AMBRA1 enhances PARK2-mediated mitophagy.

BNIP3

BNIP1, 2, and 3 were initially identified in a yeast two-hybrid screen as BCL2 and adenovirus E1B 19-kDa interacting proteins. They are proapoptotic proteins that can be suppressed by the E1B 19-kDa protein or BCL2. Spatially, BNIP1 and BNIP2 localize to the nuclear envelope and endoplasmic reticulum, whereas BNIP3 resides in the mitochondria. BNIP3 has a putative BH3 domain in addition to its C-terminal transmembrane domain, and BH3-containing proteins can induce both apoptosis and autophagy that converge at mitochondria, indicating that BNIP3 may regulate mitophagy.¹⁶⁴ Indeed, BNIP3 interacts with LC3 via its LIR for autophagic degradation of mitochondria.¹⁶⁵ Generally, there are three models for the mechanism of BNIP3- or BNIP3L-dependent mitophagy. First, BNIP3 or BNIP3L triggers mitochondrial depolarization and initiates mitophagy. However, whether mitochondrial depolarization is the cause or consequence of mitophagy remains unclear. Second, BNIP3 or BNIP3L functions as a receptor protein to recruit autophagic machinery to mitochondria. Both BNIP3 and BNIP3L can directly interact with LC3 via their LIR. Third, BNIP3 or BNIP3L can compete with BECN1 for the binding to BCL2 or BCL2L1. The increased expression of BNIP3 or BNIP3L will release BECN1 from BCL2 or BCL2L1 to activate mitophagy.¹⁶⁶

In response to hypoxia or oxidative stress, cells will undergo rapid mitophagy to enable cell survival, or undergo apoptotic or necrotic cell death. BNIP3 is important to this fate determination. When damage is too severe or persistent, BNIP3 changes its role from enabling survival to promoting cell death.¹⁶⁷ However, it remains unconfirmed how BNIP3 can be switched from a mitophagy inducer to an apoptotic mediator. The impact of BNIP3 on mitophagy is regulated by extracellular or intracellular signals. For example, phosphorylation of serine residues 17 and 24 flanking the BNIP3 LIR promote BNIP3-induced mitochondrial sequestration, lysosomal delivery, and, consequently, mitophagy.¹⁶⁵ Overexpression of the anti-apoptotic protein BCL2L1 significantly enhances LIR-dependent BNIP3 binding to LC3B and BNIP3-induced mitochondrial sequestration, supporting a prosurvival role of BNIP3-induced mitophagy.¹⁶⁵ Additionally, the transcription of BNIP3 and BNIP3L can also be regulated by hypoxia. Actually, BNIP3 was originally identified as an upregulated target in cells exposed to hypoxia. There are two HIF1A binding sites in the BNIP3 promoter. Its expression is suppressed by the VHL (von Hippel-Lindau tumor suppressor, E3 ubiquitin protein ligase) protein in human cancer cells, and deregulated BNIP3 expression in cancer is associated with several aggressive diseases.¹⁶⁴

Targeting mitophagy for cancer therapy

Although the transient activation of autophagy including mitophagy may play dual functions to promote tumor regression or enable tumor cell survival, prolonged or robust autophagy activation will eventually lead to the degradation of components essential for cell survival and therefore cause cell death. Accordingly, some agents have been developed to induce mitophagy in cancer cells for anticancer treatment.¹⁶⁸ For instance, the lis/lin/GO (glucose oxidase) killer-suicide system has been recognized as a robust mitophagy trigger to induce cell death and growth inhibition of human cancer cells both in vitro and in vivo.¹⁶⁹ This oxidase system is composed of linamarase, linamarin, and glucose oxidase. The β-glucosidase enzyme linamarase can hydrolyze its cyanogenic glucoside substrate linamarin into glucose, acetone, and cyanide. Cyanide inhibits cytochrome c oxidase of the mitochondrial respiratory chain, thereby triggering a rapid mitochondrial fission to facilitate mitophagy and eventually autophagic cell death.^170-172 The original lis/lin system was sufficient to inhibit the in vitro growth of cancer cells but failed to suppress tumor growth in vivo. Therefore, glucose oxidase was added to increase the therapeutic potential. Glucose oxidase converts glucose into gluconic acid and hydrogen peroxide, thus inducing oxidase stress to activate mitophagy.¹⁶⁹ Similar to cyanide, hydrogen peroxide is highly diffusible so as to penetrate freely across membranes and generate a critical bystander effect on tumor regression.¹⁷³ Either the antioxidant N-acetyl cysteine or inhibitors of the core autophagic machinery such as 3-methyladenine rescue cell death induced by the lis/lin/GO system. Similarly, genetic knockdown of core genes involved in mitophagy such as Atg5, Becn1, and Bnip3 also block cell death.

Another striking example demonstrating the efficacy of mitophagy induction for cancer therapy is ceramide, which acts as a bioactive sphingolipid to induce cell death, growth inhibition, and senescence in various human cancer cells^174,175 Intriguingly, recent studies have proposed that ceramides can promote cell death and tumor regression through inducing mitophagy.¹⁷⁶ CERS (ceramide synthase) enzymes CERS1 to CERS6 generate ceramides with distinct biological roles that largely depend on their different localizations. For instance, CERS1 and CERS6 preferentially generate C₁₈- and C₁₆-ceramide, respectively.¹⁷⁷ C₁₈- ceramide promotes cell death as observed in preclinical and clinical studies, whereas C₁₆-ceramide is mainly associated with cancer cell proliferation.^176,178 Furthermore, exogenous C₁₈-pyridinium ceramide can also trigger mitophagy to promote cell death, and stable knockdown of LC3B expression with shRNA abrogates CERS1- and C₁₈-ceramide-dependent mitophagy, and blocks tumor suppression in vivo.⁸² After the lipidation of LC3B with phosphotidylethanolamine to form LC3B-II induced by C₁₈-ceramide or C₁₈-pyridinium ceramide, ceramide, and LC3B-II directly interact on mitochondrial membranes via DNM1L/Drp1-dependent mitochondrial fission, thus enabling mithchondria sequestration.⁸²

More recently, it was reported that low-intensity ultrasound therapy in the presence of curcumin enhances the cell death of nasopharyngeal carcinoma CNE2 cells through initiation of mitophagy.¹⁷⁹ Therefore, the induction of cell death by combining treatment-triggered mitophagy is another potential and feasible strategy for treating malignancies.

Taken together, an increasing number of approaches are currently being developed to induce tumor regression via mitophagy. As cancer cells have frequently dysregulated oncogenic signaling that stimulates the generation of oxidative stress and vacuole formation, they may have a lower threshold for mitophagy execution and are susceptible to autophagic cell death upon treatment with mitophagy modifiers.

Conclusions and Perspectives

Our understanding of the molecular mechanisms implicated in the regulation of mitophagy has greatly advanced in the past several years. However, there are still many questions to be answered. In different organisms, distinct signals activate their own regulators to induce mitophagy. Different from the inherent conservation of the core autophagic machinery, mitophagy-specific genes differ considerably between yeast and mammals, although the key players in mitochondrial fusion/fission dynamic events are evolutionarily conserved. The relevance of dysregulated mitophagy to neurodegeneration is much clearer, whereas the implication of mitophagy in human carcinogenesis is still very much a mystery, mainly because research in this field is still in its infancy. Meanwhile, more precise and reliable experimental methods to detect and monitor the whole process of mitophagy are urgently needed, which may not only provide further insights into the complex networks involved in mitophagy in detail, but may also help researchers and clinicians explore new therapeutic approaches to disorders that involve dysfunctional mitochondria, such as cancers. The double-faceted role of mitophagy during oncogenesis, either survival-supporting or death-promoting, leads to a novel challenge when targeting it for cancer therapies. Therefore, a better understanding of mitophagy regulation in cancer, and revealing the intrinsic molecular mechanisms of this process, will be crucial for the development of novel anticancer therapeutics.

Abbreviations:
AMBRA1	=	autophagy/Beclin 1 regulator 1
Atg	=	autophagy-related
BCL2	=	B-cell CLL/lymphoma 2
BNIP3L	=	BCL2/adenovirus E1B 19 kDa interacting protein 3-like
CCCP	=	carbonyl cyanide m-chlorophenyl hydrazone
CERS	=	ceramide synthases
Cvt	=	cytoplasm-to-vacuole targeting
GABARAP	=	GABA(A) receptor-associated protein
IMS	=	intermembrane mitochondrial space
LIR	=	LC3-interacting region
MAPK	=	mitogen-activated protein kinase
mtDNA	=	mitochondrial DNA
NFKB	=	nuclear factor of kappa light polypeptide gene enhancer in B-cells
OMM	=	outer mitochondrial membrane
OXPHOS	=	oxidative phosphorylation
PAS	=	phagophore assembly site
PINK1	=	PTEN-induced putative kinase 1
PtdIns3K	=	class III phosphatidylinositol 3-kinase
ROS	=	reactive oxygen species
SH3GLB1	=	SH3-domain GRB2-like endophilin B1
TGFB	=	transforming growth factor, beta 1
Ub	=	ubiquitin
Ulk1	=	Unc51 like autophagy activating kinase 1
UVRAG	=	UV radiation-resistance associated
WAT	=	white adipose tissue

Acknowledgments

This work was supported by the National Natural Science Foundation of China (No.81301706), Program for Innovative Research Team in Zhejiang Province (No.2012R10046-15), Ministry of Education of China (No.20120101120044), and Zhejiang Provincial Natural Science Foundation of China (No.Y13H160010 and LR12H16001).

10.4161/auto.26550

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.