Expanding perspectives on the significance of mitophagy in cancer

Abstract

Mitophagy is a selective mode of autophagy in which mitochondria are specifically targeted for degradation at the autophagolysosome. Mitophagy is activated by stresses such as hypoxia, nutrient deprivation, DNA damage, inflammation and mitochondrial membrane depolarization and plays a role in maintaining mitochondrial integrity and function. Defects in mitophagy lead to mitochondrial dysfunction that can affect metabolic reprogramming in response to stress, alter cell fate determination and differentiation, which in turn affects disease incidence and etiology, including cancer. Here, we discuss how different mitophagy adaptors and modulators, including Parkin, BNIP3, BNIP3L, p62/SQSTM1 and OPTN, are regulated in response to physiological stresses and deregulated in cancers. Additionally, we explore how these different mitophagy control pathways coordinate with each other. Finally, we review new developments in understanding how mitophagy affects stemness, cell fate determination, inflammation and DNA damage responses that are relevant to understanding the role of mitophagy in cancer.

1. Introduction

Mitophagy is a selective form of general autophagy in which mitochondria are specifically targeted for degradation at the autophagolysosome [1,2]. As such, the process of mitophagy is dependent on the general autophagy machinery, and additionally, relies on a growing cadre of “mitophagy adaptors” and regulatory molecules that are involved in selecting mitochondria for autophagic turnover (see Table 1). While mitophagy can be induced artificially with respiratory chain inhibitors and uncoupling agents, such as carbonyl cyanide 3-chlorophenylhydrazone (CCCP) that depolarize mitochondria, it is also part of the response to key physiological stresses, such as hypoxia and nutrient deprivation [3]. In addition, mitophagy is a programmed component of developmental and differentiation processes, including elimination of paternal mitochondria from the fertilized egg [4–6] and removal of mitochondria during red blood cell production [7,8] and muscle differentiation [9]. By promoting elimination of dysfunctional, supernumery and/or aged mitochondria, mitophagy plays a central role in maintaining mitochondrial and cellular integrity. Conversely, defective mitophagy can lead to loss of tissue homeostasis and development of disease, including cancer [2]. Various mitophagy modulators have been shown to be deregulated in human cancers, including PARK2, BNIP3, BNIP3L, FANCC, p62/SQSTM1, with others likely to emerge as research increases in this area. The challenge in the field is to determine what selective advantage is conferred to the tumor through deregulation of mitophagy and whether mitophagy is acting to promote or to limit tumorigenesis. In this review, we address outstanding questions pertaining to the function and regulation of mitophagy adaptors in cancer, including the extent to which these molecules interact or are co-regulated. We also probe emerging roles for mitophagy in cell fate determination, inflammatory responses and DNA damage responses that contribute to our overall understanding of the role of mitophagy in cancer.

Table 1.

Mitophagy regulators and what we know about them.

Gene/protein	Function in mitophagy	Links to cancer/disease	Reference
PARK6/PINK1	Serine/threonine kinase that undergoes voltage-dependent import and degradation at the IMM; stabilized at the OMM by altered ΔΨmt; phosphorylates ubiquitin chains, and Parkin on S65 to derepress its auto-inhibitory activity leading to Parkin recruitment to and activity at the OMM.	PARK6 loss linked to Parkinson’s disease. Reduced PARK6 expression has been detected in glioblastoma and ovarian cancer and PARK6 mutated in rare cases of neuroblastoma.	[31,35,36,37,51]
PARK2/Parkin	E3 ubiquitin ligase, activated by phosphorylation on S65 by PINK1 causing it to localize to the OMM; conjugates ubiquitin chains to numerous OMM proteins, including Mfn2; phospho-Ub chains are bound by autophagy cargo adaptors, like p62, OPTN, NDP52. Antagonized by USP30 and other mitochondrial de-ubiquitinases.	PARK2 inactivating mutations linked to Parkinson’s Disease; deleted in human ovarian, breast, lung and bladder cancers; inactivating mutations found in glioblastoma and other cancers; Parkin null mice develop spontaneous liver tumors and are sensitized to radiation-induced lymphoma.	[31,34,50,51]
PARK8/LRRK2	A novel PD predisposition locus encoding a protein that interacts with Miro to promote its degradation, prevents sequestration of mitochondria and inhibits mitophagy. LRRK2 has specific kinase activity for Rab GTPases involved in cellular trafficking (Rab1b, Rab8a, Rab10) and mutant LRRK2 (G2019S) has increased kinase activity for these substrates.	Mutated (eg. G2019S) in Parkinson’s Disease	[32,46]
Mul1	Mitochondrial E3 ubiquitin ligase that can compensate for loss of Parkin.	Not known.	[54]
PHB2/Prohibitin2	IMM protein involved in processing OPA1 and cristae remodeling; binds LC3 following Parkin-mediated rupture of OMM, acts as an IMM mitophagy adaptor; required for Parkin-mediated mitophagy. Requires PHB1 for stability. Also has a nuclear function regulating activity of E2F, p53 and the AR.	Expression deregulated in breast, prostate and lung cancer but role varies depending on tissue type. PHB1 deletion promotes HCC in mice.	[140,141]
Miro	Anchors kinesins to the OMM; is phosphorylated and cleaved by Parkin/PINK1 activity resulting in reduced mitochondrial motility, possibly isolating damaged mitochondria for mitophagy; differential phosphorylation of Miro can block Parkin recruitment to OMM.	Turnover disrupted in PD.	[32,207,208,209]
BNIP3	OMM mitophagy adaptor that binds LC3 directly to target mitochondria to the autophagosome; induced by hypoxia, nutrient deprivation, oncogenic Ras, FoxO3A, E2F, NF-κB; repressed by pRB and p53; required for mitophagy in fasted liver.	Deleted, silenced or mis-localized in breast, prostate, colon, pancreatic, liver, glioma and other cancers. BNip3 loss accelerates progression to metastasis in mouse models of breast cancer.	[3,18,106,210]
BNIP3L (NIX)	Homolog of BNIP3; induced by hypoxia, p53; required for mitophagy during red blood cell differentiation; interacts with Rheb and LC3.	BNIP3L knockdown promoted tumor growth in a mouse mammary tumor xenograft study.	[7,8,105,210]
Rheb	Small GTPase required for mTORC1 activity; interacts with LC3 and NIX to promote mitophagy induced by switch from glycolytic to oxidative metabolism.	Rheb point mutations found in genomic analyses of kidney and endometrial cancers.	[91]
FUNDC1	OMM protein induced by hypoxia, interacts directly with LC3; LC3 interaction regulated by ULK1, SRC, CK2, PGAM5; accumulates at ER-mito contact sites through interactions with calnexin; essential for hypoxia-induced mitophagy through recruitment of both Drp1 and LC3.	Not known. Knockout mouse has defects in platelet maturation.	[29,122–124]
FKBP8	A member of the FK506-binding protein family localized to the OMM possesses a LIR motif and interacts directly with LC3-related protein, LC3A recruiting it to damaged mitochondria in a Parkin-independent manner.	Not known.	[211]
PGAM5	Serine/threonine phosphatase at the OMM, required for CCCP-induced mitophagy, de-phosphorylates S13 of FUNDC1 to activate it, interacts with PINK1 at OMM in response to altered ΔΨmt; protects PINK1 from degradation at IMM; Aged mice develop symptoms of PD.	Pgam5 null mice exhibit a movement disorder as they age, reminiscent of PD.	[124–126]
Bcl2-L13/Bcl-Rambo	Mammalian homologue of Atg32 in yeast; binds LC3 through conserved LIR motif; over-expression induces mito fragmentation and mitophagy in a Parkin-independent manner.	Not known.	[129]
EndophilinB/Bif-1	Fatty acyl transferase required for mitochondrial membrane dynamics; interacts with Beclin1 via UVRAG; colocalizes at Atg9+ puncta, involved in membrane lipid trafficking around mitochondria.	Haploinsufficiency promotes Myc-driven lymphomagenesis in mice;	[130–132]
SMURF	Identified in a screen for genes that are required for xenophagy; also essential for Parkin-induced mitophagy; recruited to mitochondria by altered ΔΨmt; has E3 Ub ligase activity but this is not required for mitophagy	Not known.	[52]
FANC-C	Component of the Fanconi Anemia (FA) DNA repair pathway; identified in a screen for genes that rescue selective autophagy; Fanc-C interacts with Parkin, required for mitophagy, localizes to mitochondria, genetically distinct role for FANCC separate from its role in DNA repair. Other FA genes also implicated in mitophagy are FANCA, FANCD1 (BRCA2), FANCD2, FANCF, FANCL, FANCS (BRCA1).	Inactivated in Fanconi Anemia; other FANC genes include BRCA2 and BRCA1 that are deleted in hereditary breast and ovarian cancer.	[53]
Cardiolipin (CL)	IMM phospholipid involved in tethering proteins to cristae including ETC components; stress induces relocalization of CL to OMM; LC3 binds to CL, required for mitophagy involving LC3.	Not known.	[138]
TAZ/tafazzin	Phospholipid trans-acylase that remodels cardiolipin; loss of TAZ blocks mitophagy due to failure of LC3 to recognize CL; TAZ deficiency results in reduced acylation of CL and its depletion from mitochondria.	Mutated in Barth syndrome	[139]
p62/SQSTM1	Multi-functional protein that acts as a signaling hub integrating various stress responses; relevant function here is as a cargo adaptor that can promote mitophagy by binding to ubiquitinated proteins at the OMM and to processed LC3.	Amplified in human kidney cancer, accumulates in pancreatitis and non-alcoholic steatohepatitis, over-expressed in autophagy deficient cancers in mice, required for Ras-mediated transformation in mouse models.	[65,70]
NBR1 (Near BRCA1)	Similar to p62/SQSTM1, NBR1 acts as an autophagy-adaptor protein by binding directly to both LC3 through a conserved LIR and to ubiquitinated proteins at the cargo. Role in mitophagy less established than other cargo adaptors. Promotes cell motility by facilitating focal adhesion turnover by autophagy.	Deleted along with BRCA1 in some hereditary breast cancers.	[60,77]
OPTN/Optineurin	OPTN is an autophagy cargo adaptor that can simultaneously bind LC3 and ubiquitin at the OMM. Gets recruited to depolarized mitochondria in a Parkin-dependent manner	Mutated in ALS. Down-regulated in lung cancer cell lines. Low expression correlates with poor prognosis in human lung cancer.	[50,59]
NDP52 (Nuclear Dot Protein 52)	Along with OPTN, NDP52 was found to interact with LC3 and be required for Parkin-dependent mitophagy and to recruit LC3, DFCP1 and WIPI1 to focal regions of the OMM.	Not known.	[50,51]
TAX1BP1	Like OPTN and NDP52, TAX1BP1 was shown to be a cargo adaptor that could rescue mitophagy in cargo adaptor deficient cells exposed to CCCP. However, it appeared to do so less well than either OPTN or NDP52.	Not known.	[50]
TBK1 (TANK-binding kinase-1)	Serine/threonine kinase that signals downstream of MAVS to activate the interferon response pathway in response to altered ΔΨmt. Also phosphorylates OPTN and NDP52 promoting their recruitment to depolarized mitochondria in a Parkin/PINK1 dependent manner. Phosphorylates PLK1 and other mitotic substrates active at the centrosome and mitotic spindle/midbody.	Mutated in ALS; required for K-Ras transformation and tumor cell survival; promotes dormancy of prostate cancer cells.	[51,67,68,69,195]
Caspase-1	Proto-typical caspase involved in cytokine processing, key component of the NLRP3 inflammasome; recruited to the OMM, activated by altered ΔΨmt ROS and oxidized mtDNA in a positive feedback loop; cleaves Parkin. Mitophagy attenuates inflammasome activity.	Inflammasome and IL-1 signaling promotes fibrosarcoma, breast and gastric cancers.	[186,188,189,212]

2. Mechanics of mitophagy and key mitophagy proteins

Mitophagy is only one type of selective autophagy that the cell can engage in, while the general autophagy machinery is also required for elimination of other organelles, pathogens and protein aggregates [10]. In addition, autophagy has novel non-canonical functions in secretion from the cell, cellular motility and transcription factor turnover, as has recently been reviewed elsewhere [11–13]. Since mitophagy is reliant on the general autophagy machinery for formation of the pre-initiation and initiation complexes, defects in mitophagy are epistatic to defects in general autophagy. However, perhaps because general autophagy plays so many critical roles in the emergence of tumor cells, core autophagy genes have rarely been found mutated in human cancers [14]. Indeed, debate remains over whether Beclin-1, a core component of the autophagy initiation complex, that is mono-allelically deleted in breast, ovarian and other cancers [15,16], is a bona fide tumor suppressor gene or a merely a passenger lesion whose deletion is driven by selection for loss of the adjacent BRCA1 gene [17]. By contrast, some of the unique mitophagy adaptors, such as PARK2, BNIP3 and BNIP3L are deleted or silenced in numerous primary human cancers [3]. For this reason, as well as the greater ease of parsing out the functions of mitophagy in cancer, rather than all of general autophagy, our laboratory has increasingly focused on trying to understand the role of selective mitophagy in cancer [3,18].

Interestingly, nascent phagophore membranes that are destined to engulf mitochondria and other autophagic cargo appear to form specifically at the sites of contact between mitochondria and the endoplasmic reticulum (ER), with recruitment of Atg14 and other autophagy components [19]. The integrity of ER-mitochondrial junctions is essential for starvation-induced autophagy in both yeast and mammalian cells [19,20]. Thus mitochondrial integrity is likely to feed back to modulate the production of phagophore membranes and general autophagy. From the perspective of mitophagy, ER-mitochondrial contacts are especially important since mitochondrial fission appears to be spatially restricted to a subset of these sites [21,22]. Moreover, fission has been shown in several systems to be induced in parallel with, and to be required for mitophagy [23–25], while mitochondrial fusion can protect mitochondria against mitophagy [26,27]. A number of mitophagy regulators including BNIP3 and FUNDC1 have been localized to both the ER and mitochondria [28,29] but further work is necessary to elucidate the timing and topology of recruitment of specific mitophagy factors to these sites.

The number of proteins now known to play a direct role in mitophagy has increased markedly over the past few years. Table 1 summarizes those proteins with defined roles in mitophagy and below we highlight key aspects of their function germane to our discussion of the role of mitophagy in cancer.

2.1. PINK1 and Parkin

Mitophagy activated through the concerted actions of PINK1 (PTEN-induced putative kinase-1) and the Parkin E3 ubiquitin ligase has been more widely studied than other mitophagy pathways [1,30]. Briefly, PINK1 accumulates at the outer mitochondrial membrane (OMM) in response to mitochondrial depolarization. At the OMM, PINK1 functions as a ubiquitin kinase phosphorylating Parkin on S65 within its ubiquitin-like domain, promoting increased localization of active Parkin to the OMM. Parkin conjugates ubiquitin chains to key mitochondrial substrates including mitofusins, Vdac1, Miro and other OMM proteins and these ubiquitin chains are further phosphorylated by PINK1 resulting in amplification of the signal to undergo mitophagy (Fig. 1A). Both PINK1 and Parkin are required for recruitment of mitophagy cargo adaptors including p62/Sqstm1, OPTN and NDP52 which bind to the phospho-ubiquitin chains generated by PINK1/Parkin. The recruited cargo adaptors then interact with processed LC3 at growing phagophores to target mitochondria for degradation (Fig. 1A) [31].

Fig. 1.

Key molecular players in mitophagy. The most extensively analyzed molecular mechanisms of mitophagy involve PINK1/Parkin-mediated mitophagy in response to mitochondrial depolarixation (A) and the BNIP3/NIX and FUNDC1 modulated pathways in response to hypoxia (BNIP3, NIX and FUNDC1), nutrient stress (BNIP3) and cellular differentiation cues (NIX) amongst other signals. (A) PINK1 translocation to the mitochondrial matrix where it is proteolytically cleaved is dependent on mitochondrial membrane potential and thus stresses that induce mitochondrial membrane depolarization result in accumulation of functional PINK1 kinase at the OMM. PINK1 functions as a ubiquitin kinase that phosphorylates ubiquitin and Parkin (within its ubiquitin-like domain) resulting in recruitment of Parkin ubiquitin ligase activity at the OMM where is can ubiquitinate other substrates (S) that then become PINK1 targets, feeding back to amplify the effect of PINK1 stabilization. Interaction of PINK1/Parkin targets at the mitochondria with processed LC3 to target mitochondria for degradation by autophagy relies on specific cargo adaptor molecules including optineurin (OPTN), NDP52 and TAXBP1 that simultaneously bind ubiquitin and LC3. (B) The other major mechanism promoting mitophagy involves the induction of proteins that bind LC3 at nascent autophagosomes directly, including BNIP3, NIX and FUNDC1. All three of these mitochondrial adaptor proteins are induced by hypoxia and interact with LC3 through conserved LIR motifs in their respective amino termini. The extent to which BNIP3, NIX and FUNDC1 are functionally inter-dependent is not clear although due to tissue-restricted expression BNIP3 plays a key role in mitophagy in fasted liver while NIX plays a critical role in mitophagy during red blood cell maturation.

Parkin is encoded by the PARK2 gene and PINK1 by the PARK6 gene with both genetic loci found mutated in human Parkinson’ s disease (PD), implicating mitophagy defects in the etiology of PD [31]. Interestingly, a novel risk locus for PD, PARK8 encoding LRRK2, has also been shown to play a role in mitophagy [32]. PARK2 (Parkin) maps to a common fragile site at human chromosome 6q25–q26 that is frequently deleted in ovarian, breast, bladder, lung, and other cancers [33] while inactivating mutations in PARK2 have been detected in glioblastoma, colon cancer and lung cancer [34]. Meanwhile, PARK6 expression is lost or down-regulated in glioblastoma and ovarian cancer respectively [35,36] and mutated in rare cases of neuroblastoma [37]. Consistent with a tumor suppressor function for Parkin, parkin null mice are susceptible to spontaneous hepatocellular carcinoma [38], possibly related to functions of Parkin in liver metabolism [39] and are also sensitized to irradiation-induced lymphomagenesis [40]. In this setting, Parkin expression increased oxidative metabolism and limited the Warburg effect downstream of the p53 tumor suppressor, possibly as a consequence of improved mitochondrial quality control [40]. Similarly, PINK1 was shown to suppress ROS generation, HIF-1 stabilization and the Warburg effect in human astrocytes [35]. However, the role of Parkin in cancer is complicated by more recent findings showing that it regulates cyclin levels and exit from mitosis [41,42]. As a component of the FBX4 Cullin-ring ligase complex, Parkin regulates levels of Cyclin D1, Cyclin E, and CDK4 in cancers [41]. Additionally, through its ability to bind Cdc20 and substitute for the Anaphase promoting complex (APC), Parkin also modulates mitotic checkpoint control and Parkin loss subsequently leads to chromosomal instability and aneuploidy [42]. Thus, it is unclear whether the tumor suppressor activity of Parkin relates to its role in mitophagy or to its role in cell cycle. However, these studies did not address a role for PINK1 in mitotic checkpoint control which is particularly important given that Parkin ubiquitin ligase activity is dependent on PINK1 activity [43–45]. Perhaps cell cycle regulatory kinases can compensate for PINK1 in activating Parkin during mitosis; this remains to be examined.

Interestingly, LRRK2 encoded by PARK8 has recently been shown to possess kinase activity toward Rab GTPases involved in vesicular trafficking in the cell [46]. Indeed, several Rab proteins have themselves been implicated in the etiology of PD, including Rab7L1 that is encoded by the PARK16 locus, loss of which phenocopies the G2019S mutation in LRRK2 and restoration of which rescues neuronal cell death linked to LRRK2-G2019S mutation [47]. The LRRK2-G2019S mutant shows increased kinase activity for Rab GTPase substrates compared to wild-type LRRK2 and this results in Rab GTPases with reduced binding for regulatory GDI proteins and reduced Rab GTPase dependent trafficking activity [46]. Intriguingly, Parkin has also been shown to modulate Rab GTPase activity in cells via control of specific mitochondrial Rab GAP proteins that control phagophore engulfment of mitochondria during mitophagy [48] while PINK1 activation has been shown to result indirectly in phosphorylation of specific Rab GTPases to modulate their interaction with regulatory GAP proteins [49]. Thus, a common theme to the role of PD loci in disease etiology is regulation of Rab GTPase activity that may contribute to the efficiency with which mitochondria are turned over at the autophagolysosome.

Most cell lines used in tissue culture to examine mitophagy do not express Parkin and many of the elegant analyses of Parkin-dependent mitophagy were carried out using Parkin over-expression [50,51], as opposed to examination of endogenous Parkin. This includes some of the more recent studies that identified novel regulators of mitophagy using the combination of Parkin over-expression, and mitochondrial depolarization (by treatment with CCCP) to induce mitophagy [52,53]. Given the lack of Parkin expression in the majority of cell types, the role of other mitochondrial E3 ubiquitin ligases in mitophagy has been explored, including Mul1 that can compensate for loss of Parkin in flies and other systems, including during the elimination of paternal mitochondria from the embryo [6,54]. SMURF is also a mitochondrial E3 ubiquitin ligase involved mitophagy although its ubiquitin ligase activity was not required for its role in mitophagy [52]. A direct role for other mitochondrial E3 Ub ligases in mitophagy has not yet been demonstrated but seems likely to emerge.

2.2. p62/Sqstm1, OPTN and NDP52

There are five autophagy cargo adaptor molecules (p62/Sqstm1, NBR1, OPTN, NDP52 and TAX1BP1) known to interact with processed LC3 at the autophagosome through a conserved LIR motif in one domain of the adaptor molecule and ubiquitylated chains on the cargo through a separate domain (Fig. 1A) [55,56]. All of these adaptor proteins have been implicated in selective turnover of mitochondria by mitophagy [50,51,57–62], although the evidence that NBR1 is required for mitophagy is limited [60]. p62/Sqstm1 is the prototypical autophagy cargo adaptor that was first shown to facilitate turnover of mitochondria in a Parkin/PINK1-dependent manner by interacting with ubiquitylated VDAC1 at the OMM following mitochondrial depolarization [57]. However, p62/SQSTM1 is also involved in turnover of other cargo in the cell, such as large protein aggregates [63] and thus is not considered a mitophagy-selective adaptor. Furthermore, p62/SQSTM1 is a multi-functional protein with roles in cell signaling and inflammation beyond its function as an autophagy adaptor, including control of NRF2 and mTORC1 activity [64,65]. It has also been questioned to what extent p62/SQSTM1 is actually involved in mitophagy [50,58] with the role of the OPTN, NDP52 and TAX1BP1 cargo adaptors emerging recently as being more critical for mitophagy [50,59,61,62]. OPTN in particular appears to be essential for Parkin/PINK1-dependent mitophagy induced by depolarization or excess ROS [59,61,62]. OPTN is inactivated in amyotrophic lateral sclerosis (ALS) and ALS-linked mutations prevent recruitment of OPTN to mitochondria with Parkin [59]. Thus, neuronal deterioration in ALS is linked to defective mitophagy similar to that seen in PD with mutation of PARK6, PARK2 or PARK8 [31]. TANK-binding kinase (TBK1), a Parkin/PINK1 target, phosphorylates OPTN, promoting the co-recruitment of OPTN and TBK1 to depolarized mitochondria. At depolarized mitochondria, OPTN binds to both phosphorylated ubiquitin chains and processed LC3 at growing phagophores [51,61,62]. Similar to OPTN, mutations in TBK1 that are associated with ALS, block its ability to promote mitophagy [61]. Interestingly, TBK1 has been linked to tumor dormancy in prostate cancer [66] and TBK1 emerged from an shRNA screen for synthetic lethality with oncogenic K-Ras in human NSCLC cell lines [67] where it was suggested that TBK1 was selectively required in K-Ras transformed tumor cells to prevent apoptosis mediated via NF-κB induced transcription of Bcl-X_L [67]. However, TBK1 also plays a role in regulating mitosis possibly through phosphorylation of PLK-1 and its localization to centrosomes and the spindle mid-body, such that inhibition of TBK1 resulted in mitotic spindle defects and polyploidy [68,69]. Thus, TBK1 appears to pleiotropic roles in cell growth and it remains to be determined how its role in mitophagy is linked to a role in promoting tumorigenesis.

Both NDP52 and TAX1BP1 are also recruited to depolarized mitochondria but it not yet clear if these cargo adaptors play as extensive a role as OPTN in mitophagy [61,62] although, it has been suggested that NDP52 could synergize with OPTN to promote mitophagy [51]. Several of these studies suggested that p62/SQSTM1 was not required for PINK1/Parkin-dependent mitophagy [50,51,61] but further work is needed to determine the specific contexts in which one autophagy adaptor molecule is utilized preferentially in PINK1/Parkin-dependent mitophagy over another.

Of these various autophagy adaptors, only p62/SQSTM1 has been analyzed in the context of cancer. P62/SQSTM1 has been shown to be over-expressed in renal cell carcinoma as a result of chromosome 5q amplification [70] and to accumulate in pancreatitis and liver degenerative diseases which are precursor conditions to pancreatic ductal carcinoma (PDAC) and hepatocellular carcinoma (HCC) respectively [65]. In mouse models, p62/Sqstm1 is required for Ras-driven lung tumorigenesis [71] and its over-expression is sufficient to induce HCC in mice [72]. While p62/Sqstm1 accumulates when autophagy is inhibited and is required for HCC that develops in autophagy deficient mice [73], it also appears that its ability to promote tumorigenesis is independent of its role as an autophagy cargo adaptor since over-expression of p62/Sqstm1 lacking the domain required to interact with ubiquitin retains the ability to promote tumorigenesis [72]. These studies highlighted the role of p62/Sqstm1 in regulating NRF2 and mTORC1 activity to explain tumor-promoting activity of p62/SQSTM1 [72]. Intriguingly, p62/Sqstm1 down-regulation in cancer associated fibroblasts (CAFs) was required for IL-6 production and the tumor promoting activity of CAFs in prostate cancer, identifying a contrasting tumor suppressor role for p62 in the tumor microenvironment [74]. Similarly, in its role as a mitophagy adaptor, p62/Sqstm1 may play a tumor suppressor role in tumor associated macrophages (TAMs) by promoting elimination of damaged mitochondria and thereby suppressing inflammasome activation [75], as discussed below in more detail. The other known autophagy cargo adaptors are much less extensively characterized for their role in cancer. NBR1 maps next to BRCA1 and is frequently deleted along with BRCA1 in breast cancer [76]. NBR1 also serves as an adaptor for focal adhesion turnover by autophagy to promote tumor cell migration [77], suggesting that like p62/SQSTM1 that NBR1 is a multi-functional adaptor and is not restricted to a role in selective mitophagy. OPTN1 is down-regulated in lung cancer cell lines and higher OPTN1 expression was correlated with a better relapse-free survival in lung cancer patients [78]. However, much more investigation into the role of these various cargo adaptors in cancer is needed and in particular, whether their activity in cancer is specifically linked to their role in selective mitophagy.

2.3. BNIP3/BNIP3L

Mitophagy is a key adaptive response to hypoxia in order to reduce mitochondrial mass to both limit ROS production and maximize the efficient use of available oxygen [79]. BNIP3 and BNIP3L (also known as NIX) are hypoxia-inducible tail-anchored proteins that integrate into the OMM via a carboxy terminal transmembrane (TM) domain [80–83]. The TM domain contains a critical glycine zipper that is required for homo-dimerization [80–83]. The amino terminus of BNIP3 and BNIP3L extends into the cytosol and interacts with LC3-related molecules at nascent phagophores, in a manner analogous to ATG32 in yeast [84,85] (Fig. 1B). This occurs via conserved LC3-interacting regions (LIR) located at amino acids 15 to 21 in BNIP3 and 43 to 49 in BNIP3L to promote hypoxia-induced mitophagy [28,86]. Similar to other forms of mitophagy, BNIP3-dependent mitophagy requires Drp1-induced fission; inhibition of fission or knockdown of Drp-1 blocks hypoxia-induced, BNIP3-dependent mitophagy [87]. However, BNIP3/BNIP3L dependent/hypoxia-induced mitophagy is not dependent on PINK1 or Parkin [29,88] and BNIP3 is not required for PINK1 accumulation or Parkin recruitment to the OMM although BNIP3 can compensate for PINK1 inactivation possibly by acting in parallel to induce mitophagy [88]. BNIP3L has been reported to stimulate mitophagy induced by Parkin in response to mitochondrial depolarization with CCCP [89] and to be a direct target of Parkin [90]. However, further work is required to properly dissect the inter-dependence of these mitophagy mechanisms.

BNIP3 and BNIP3L each interact with Rheb, a small GTPase that acts positively upstream of mTOR to promote cell growth [91,92]. Mitophagy induced by switching tumor cells from glycolytic growth conditions to oxidative metabolism required recruitment of Rheb to the OMM where it interacted with both BNIP3L and LC3 to promote LC3 processing and mitophagy in an mTOR-independent manner [91]. While BNIP3 also interacts with Rheb, disruption of this interaction results in growth inhibition that is mTOR-dependent [92]. Thus, BNIP3 and BNIP3L may function distinctly with respect to their interactions with Rheb.

Both BNIP3 and BNIP3L are transcriptionally induced by hypoxia-inducible factor-1 (HIF-1) [93–95]. In addition to its transcriptional activation by HIF-1 [96,97], BNIP3 is also transcriptionally regulated by FoxO3A [98], PPARα [99], RB/E2Fs [100], NF-κB [101], oncogenic Ras [102,103] and p53 [104] while BNIP3L is transcriptionally regulated by HIF-1 and p53 [105]. BNIP3 and BNIP3L are expressed in a tissue-specific manner, such that BNIP3 is most strongly expressed in adult heart, liver, and muscle, reflecting FoxO3A, PPARα and HIF activity respectively, while BNIP3L is most strongly expressed in the hematopoietic system [106,107]. In line with this expression pattern, BNIP3 has been shown to promote mitophagy in heart, skeletal muscle and liver [87,98,106]. BNIP3L is required for mitophagy during normal red blood cell differentiation [7,8] although this process can be disrupted by acquisition of mtDNA mutations in mutator mice with inactivation of mitochondrial Polγ [108,109]. This was associated with increased ROS arising from abnormal iron loading in differentiating erythroblasts and led to anemia in Polγ mice [108]. How BNIP3L/NIX-dependent mitophagy is mechanistically inhibited in red cells by mtDNA mutations and/or associated ROS accumulation, remains to be determined.

BNIP3 and BNIP3L are both upregulated in pre-malignant breast cancer at the ductal carcinoma in situ (DCIS) stage of disease as tumors become ischemic [95,110], but reduced BNIP3 expression is detected as breast cancers progress to invasive ductal carcinoma and metastases [111]. Loss of BNIP3 expression in malignant cancers has been linked to copy number loss due to deletion of the BNIP3 locus at chromosome 10q26.3, particularly in breast cancer [18,112] and to epigenetic silencing in other cancers, including gastric, pancreatic, liver, lung and hematological malignancies [113–116]. BNIP3 inactivation in pancreatic cancer was linked to chemoresistance and poor prognosis [117,118]. Altered subcellular localization of BNIP3 reported in gliomas, prostate and breast cancer may suggest an alternative mechanism of BNIP3 deregulation in cancer [111,119,120]. Mouse models also support a tumor suppressor role for BNIP3 with loss of BNip3 in the MMTV-PyMT breast cancer model resulting in increased primary tumor growth rate and accelerated progression to lung metastases [18]. Similarly, BNIP3 knockdown promoted tumor growth and metastasis in an orthotopic model of mammary tumorigenesis [121] while BNIP3L knockdown promoted tumor growth in a xenograft model [105].

2.4. FUNDC1

Hypoxia also induces FUNDC1, another mitophagy adaptor that functions at the OMM and interacts with processed LC3 through a conserved LIR motif [122] (Fig. 1B). In response to hypoxia, FUNDC1 accumulates at mitochondrial-ER junctions where it interacts directly with processed LC3, DRP-1 and indirectly with Calnexin [29]. FUNDC1 is essential for hypoxia-induced mitophagy possibly due to its role in recruiting DRP-1 [29]. This is evidenced by inhibition of FUNDC1-induced mitophagy via expression of mutant forms that block its interaction with DRP-1. The activity of FUNDC1 is tightly regulated by phosphorylation of residues adjacent to the LIR motif which controls interaction with LC3. In particular, the tyrosine residue at the +1 position in the LIR motif of FUNDC1 is phosphorylated by oncogenic SRC to block interaction with LC3. Phosphorylation of the adjacent serine 17 by ULK-1 in response to hypoxia promotes the interaction of FUNDC1 with LC3 and facilitates mitochondrial turnover [123]. A second signaling loop involving the PGAM5 phosphatase and Casein Kinase-2 (CK2) also regulates the interaction of FUNDC1 with LC3. PGAM5 dephosphorylates serine 13 and promotes the interaction with LC3, while phosphorylation of serine 13 by CK2 inhibits this interaction [124]. Interestingly, Bcl-X_L but not Bcl-2, binds to PGAM5 to block mitophagy, possibly through the effect of PGAM5 on FUNDC1 [125]. However, PGAM5 also promotes PINK1 accumulation in response to mitochondrial membrane depolarization [126] and thus Bcl-X_L may also be limiting PINK1/Parkin dependent mitophagy via PGAM5. However, it remains to be explained how PGAM5, that is reported to localize to the IMM [126,127], could simultaneously de-phosphorylate the LIR motif of FUNDC1 that is located in the cytosolic side of the OMM [124] and modulate PINK turnover in the mitochondrial matrix [126,127]. Clearly, PGAM5 could function at both locations but the mechanisms controlling how it translocates from matrix to OMM need to be worked out, as do the signals controlling these functional differences at OMM and matrix.

Clearly, one of the major unknowns in the mitophagy field is the extent to which there is cross-regulation between the different mitophagy pathways. BNIP3, BNIP3L and FUNDC1 are all induced by hypoxia and required for hypoxia-induced mitophagy. Conversely, the hypoxia-induced microRNA, miR-137 represses both FUNDC1 and BNIP3L expression, limiting the extent of mitophagy under hypoxia [128]. In summary, it seems likely that there are physical and/or functional interactions between key mitophagy regulators that remain to be elucidated.

2.5. Bcl2-L-13

Bcl2-L13 is a Bcl-2 related protein that was identified based on its homology to ATG32 in yeast [129]. Like BNIP3 and BNIP3L, it interacts directly with processed LC3 through two different conserved LIR motifs. Over-expression of Bcl2-L13 is sufficient to induce both mitochondrial fragmentation and mitophagy, independently of Drp-1 and Parkin [129]. Intriguingly, Bcl2-L13 expression is induced by mitochondrial depolarization with CCCP, as a result of phosphorylation of serine 272, which is adjacent to the second of its two LIR motifs. Furthermore, knockdown of Bcl2-L13 attenuates CCCP-induced mitophagy suggesting that Bcl2-L13 acts in parallel to Parkin to induce mitophagy [129], although much more remains to be understood about how and when Bcl2-L-13 promotes mitophagy.

2.6. Bif1/endophilin B

Bif-1 (Bax-Interacting Factor-1), also known as Endophilin-B, is a fatty acyl transferase that regulates general autophagy through its interaction with UVRAG at the initiation complex and also plays a role in membrane trafficking [130–132]. However, its localization to mitochondria in response to stress and the marked accumulation of defective mitochondria in cells when Bif-1 is knocked down suggests that mitophagy is more dependent on Bif-1 than other forms of selective autophagy [132]. Bif-1 deficiency leads to chromosomal instability and increased susceptibility to Myc-driven lymphoma in a genetically engineered mouse model [132]. This tumor phenotype was associated with defective mitophagy and mitochondrial dysfunction [132] consistent with a role for Bif-1 in mitophagy although similar phenotypes are observed when general autophagy is inhibited [133–136]. Thus, it remains to be resolved whether Bif-1 has a selective involvement in mitophagy or a broader role in general autophagy.

2.7. Cardiolipin and PHB-2 – mitophagy adaptors at the inner mitochondrial membrane

Cardiolipin, an inner mitochondrial phospholipid, plays a critical role in anchoring proteins, including components of the electron transport chain, to the IMM and in stabilizing mitochondrial cristae [137]. In response to stresses that induce mitophagy, such as respiratory chain inhibitors and mitochondrial membrane depolarization, cardiolipin relocalizes to the OMM where it binds directly to processed LC3 to promote mitophagy [138]. Inhibition of the LC3-cardiolipin interaction, knocking down of cardiolipin synthase, or limiting transport of cardiolipin to the OMM, all block LC3-dependent mitophagy indicating that cardiolipin is a mitophagy adaptor [138]. LC3 interacts specifically with mature tetralinoleoyl-cardiolipin that is dependent on processing from immature monolyso-cardiolipin by the Tafazzin (TAZ) phospholipid transacylase. Consistently, TAZ deficiency prevents cardiolipin remodeling and blocks mitophagy leading to mitochondrial dysfunction (reduced respiration and ROS production) that likely contributes to the cardiomyopathy and other pathologies found in Barth syndrome patients that carry loss of function mutations in the TAZ gene [139].

Another IMM molecule that was recently shown to function as a mitophagy adaptor is Prohibitin-2 (PHB-2) [140] which normally promotes cristae morphogenesis and acts as a molecular chaperone for ETC complexes, amongst other roles, at the IMM [141]. Like cardiolipin, PHB-2 also interacts directly with LC3 but unlike cardiolipin, PHB-2 relies on Parkin-mediated rupture of the OMM [140]. Indeed, previous work has indicated that degradation of OMM proteins during mitophagy is mediated by the proteasome while IMM proteins are degraded at the autophagolysosome [142]. The interaction with LC3B is mediated by a LIR motif in PHB-2 that is required for mitophagy induced by mitochondrial membrane depolarization and for elimination of paternal mitochondria in worms [140]. As the authors point out, many of the stresses known to induce mitophagy, including mitochondrial depolarization happen at the IMM and thus it is significant that IMM molecules like cardiolipin and PHB-2 can function as molecular triggers for mitophagy.

The section above has detailed important features of key proteins known to play a direct role in mitophagy. From this discussion, it is clear that new modulators of mitophagy are likely to emerge. Novel aspects of how these proteins are regulated and how their functions are coordinated with each other needs to be investigated, particularly in a physiological setting. Below, we focus on a discussion of how mitophagy is integrated into cellular and organismal responses to stress and how this may play into understanding how mitophagy modulates tumorigenesis.

3. Mitophagy and metabolic re-programming

Defects in mitophagy arising from deletion, mutation, or silencing of some of the mitophagy adaptors described above leads to mitochondrial dysfunction [3]. Consequences of mitochondrial dysfunction that influence tumor cell growth include altered Ca²⁺ signaling [143,144], ROS production [145], mtDNA mutation [146], the unfolded mitochondrial protein response (UPR^mt), reduced Fe-S cluster biosynthesis [147] and apoptosis [79,148,149]. Additionally, mitochondrial dysfunction leads to production of oncometabolites that alter epigenetic control of gene expression and stabilize key transcription factors such as HIF-1 [150,151]. Importantly, mitochondrial integrity determines the efficiency of respiration and metabolism at the mitochondria, which in turn modulates tumor growth and metastasis [148,152,153]. Indeed, mitochondrial defects were proposed by Otto Warburg to explain why tumor cells preferentially undergo aerobic glycolysis [154,155]. The switch by tumor cells to aerobic glycolysis and reduced oxidative phosphorylation has since been linked to altered expression and activity of key metabolic enzymes, including pyruvate kinase M2, phosphoglycerate dehydrogenase, succinate dehydrogenase and isocitrate dehydrogenase [156,157]. These enzymes alter rates of carbon flux through glycolysis, the TCA cycle and to various biosynthetic pathways in tumor cells [156,157]. However, defects in mitophagy can also contribute to the Warburg Effect and altered mitochondrial metabolism [3,18,148,155].

Consistent with the role of mitophagy in maintaining mitochondrial integrity and efficient oxidative metabolism, signals induced by changes in nutrient availability in the microenvironment modulate rates of mitophagy with consequences for cellular metabolism [3,18]. As discussed above, hypoxia induces BNIP3, BNIP3L and FUNDC1 to promote mitophagy [93–95,122]. In addition, reduced ATP levels activates AMPK which promotes mitophagy via ULK1 activation and mTOR inhibition [158,159] while Sirt1 and other sirtuins sensitive to both cytosolic and mitochondrial pools of NAD+ promote mitophagy through de-acetylation of Atg5, Atg7 and Atg12 [160], as well as induction of dct-1, the worm homolog of BNIP3/BNIP3L [161].

Interestingly, nutrient deprivation signals that induce mitophagy can also induce mitochondrial biogenesis, including AMPK and SIRT1 that activate PGC-1α [162–164]. Indeed, mitophagy and mitochondrial biogenesis are coordinated in C. elegans to prevent aging [165]. This has led us to suggest that coupled mitophagy and biogenesis plays a role in metabolic reprogramming of mitochondria in response to nutrient stress (Fig. 2). We propose that mitophagy eliminates mitochondria that are not metabolically suited to the imposed nutrient stress and that biogenesis promotes replacement with mitochondria that are better suited to nutrient stress, favoring catabolic processes, such as fatty acid oxidation. We also suggest that the coordinated induction of mitophagy and biogenesis is likely disrupted in cancer where decreased mitophagy and elevated biogenesis has been reported [166]. These concepts remain to be tested experimentally but offer a novel perspective on how cancers are able to maintain anabolic metabolism and cell growth despite nutrient deprivation.

Fig. 2.

Mitophagy in metabolic re-programming. Mitophagy is activated by nutrient stress resulting in turnover of mitochondria. Mitochondrial biogenesis is also activated by nutrient deprivation resulting in generation of new mitochondrial mass. The elimination of mitochondria only to produce new ones would appear to be energetically unfavorable to a starving cell unless, as we propose, it allows the cell to better adapt to the imposed stress. Thus, we propose that mitophagy coupled to mitochondrial biogenesis results in metabolic re-programming of the mitochondria to produce mitochondria that allow the cell to better survive the imposed stress. For example, we suggest that the cell eliminates mitochondria that were engaged in excess citrate production and replaces them with mitochondria in which fatty acid oxidation predominates. Furthermore, we propose that this coupling of mitophagy and biogenesis detected in normal cells is disrupted in cancer cells, possibly through loss of key mitophagy adaptors, such as BNIP3 silencing detected in PDAC, and over-expression of mitochondrial biogenesis regulators, such as PGC-1α, as seen in a subset of human melanomas.

4. Mitophagy in cell fate determination

Mitochondria are important signaling platforms in the cell, partly through production of ROS and specific metabolites that communicate with the nucleus and other cellular compartments [145,167]. By eliminating dysfunctional or aged mitochondria, mitophagy undoubtedly influences the signaling capacity of the mitochondria and indeed the ultimate consequence of defective mitochondria is a terminal cell fate, namely cell death [149]. However, recent evidence indicates that reduced mitophagy also contributes to loss of stemness and modulates differentiation rates within tissues [9,168,169]. Mitochondria are asymmetrically divided in immortalized human mammary epithelial stem cells with younger mitochondria segregating preferentially to stem-like daughter cells with mammosphere-forming capacity [168]. By contrast, older mitochondria localized to the perinuclear region are preferentially segregated to non-stem daughter cells. Inhibition of either fission (with mdivi) or mitophagy (through knockdown of Parkin) resulted in reduced inheritance of young mitochondria, increased age of mitochondria and reduced stemness. Inhibition of mitophagy was associated with the failure of older mitochondria to be spatially restricted to perinuclear regions of the cell [168] suggesting a role for mitophagy in modulating stemness through both segregation and elimination of older mitochondria (Fig. 3). Related findings have been reported in the lymphoid system where accumulation of older mitochondria promoted differentiation of B and T cells at the expense of self-renewing lymphoid progenitors [169]. However, the role of mitochondrial fission in segregation of young and old mitochondria to stem versus non-stem daughter cells clearly requires more investigation, particularly in light of recent data showing that mdivi targets complex I of the respiratory chain [170] and the known role for ROS in cell fate determination [145,171].

Fig. 3.

Mitophagy and metabolism in cell fate decisions. Mitophagy is required for self-renewal and to maintain stemness either through elimination of old mitochondria or alternatively by reducing overall mitochondrial mass to maintain low levels of OXPHOS. Analysis of pluripotent stem cells suggests that stemness is also linked to glycolytic metabolism, mitochondrial fragmentation while priming of stem cells to differentiate is associated with a switch to oxidative metabolism, mitochondrial fusion and increased mitochondrial mass. Conflicting data has emerged from analysis of cancer stem cells however, where oxidative phosphorylation marks out therapy resistant tumor cells.

There is evidence demonstrating that tissue stem cells, embryonic stem cells and induced pluripotent stem cells limit oxidative metabolism in favor of glycolytic metabolism in order to maintain their quiescent stem-like state (Fig. 3) [169,172–174]. This block to respiration and oxidative metabolism is driven by HIF activity in the hypoxic environment in which stem cells frequently reside [173]. Conversely, differentiation requires a switch to oxidative metabolism to fuel specialized functions and is associated with mitochondrial remodeling, including increased overall mitochondrial mass, dispersed cytoplasmic localization, and increased expression of enzymes involved in the TCA cycle and the electron transport chain [174]. Indeed, stimulation of glycolysis, mitochondrial depolarization or inhibition of respiration promotes pluripotency but blocks differentiation of pluripotent stem cells. Conversely, inhibition of glycolysis and enhancement of mitochondrial respiration limits pluripotency [172,174–177]. Pluripotent stem cells select against mtDNA mutations and increased ROS production linked to mtDNA mutagenesis, reduced both self-renewal capacity and stem cell reprogramming potential [178].

Similar findings have been reported for long-term repopulating hematopoietic stem cells (LT-HSCs) where maintenance of the stem cell phenotype has been attributed to reduced mitochondrial respiration and low membrane potential [179] such that induction of mitophagy with uncoupling agents promoted HSC self-renewal [179]. Consistent with these findings, work from Ito and colleagues showed that Tie2-positive LT-HSCs express higher levels of autophagy regulators including the mitophagy modulators, Parkin and Pink1 than committed progenitors, and underwent proportionately higher rates of mitophagy in response to signals that induce fatty acid oxidation [180]. Inhibition of mitophagy, through knockdown of Parkin or Pink1, attenuated HSC maintenance and expansion both in vitro and in vivo [180]. More recent work has confirmed higher rates of autophagy and low metabolic rates in HSCs compared to more committed hematopoietic progeny with loss of HSC function during the aging process in autophagy-deficient mice [181]. However, and in contrast to previous findings, parkin null HSCs behaved normally in long term repopulation experiments in vivo leading Passegue and colleagues to suggest that autophagy is required for turnover of metabolically more active, healthy mitochondria thereby explaining reduced metabolic activity and primarily glycolytic phenotype in HSCs [181]. It will be important to determine what Parkin-independent pathways of mitochondrial turnover operate in HSCs and also whether this suggests that previous findings showing a role for mitophagy in segregation of older versus younger mitochondria during stem cell division need to be revisited. Since mitophagy adaptors like BNIP3, NIX and FUNDC1 are induced by low oxygen conditions such as those found in the hematopoietic stem cell niche and do not require membrane depolarization to function, it seems important to determine whether these proteins, as opposed to Parkin/Pink1 are used in HSCs to eliminate mitochondria.

The dependence of cancer stem cells on glycolysis to maintain self-renewal capacity is less clear-cut than in pluripotent stem cells. Some evidence suggests that reduced mitochondrial mass, increased glycolysis and mitochondrial membrane depolarization promotes maintenance of a cancer stem cell phenotype [179,180], similar to pluripotent stem cells. By contrast, more recent work has suggested that dormant or quiescent PDAC cells with stem cell properties, rely on oxidative respiration more and are less dependent on glycolysis than proliferating cells in the rest of the tumor [181]. Similarly, quenching ROS or inhibiting glycolysis in lymphoid progenitors promoted self-renewal at the expense of differentiation [169]. These discrepancies may relate to differences in tumor/tissue type, the stage of progression, and/or the role of the microenvironment, including nutrient availability. It will be important to resolve the differences between classes of stem cells (cancerous or non-cancerous) that determine whether their self-renewal is promoted or inhibited by ROS generation, glycolytic metabolism or mitochondrial membrane potential.

Given that rates of mitophagy affect mitochondrial integrity and ROS production [18], and that inhibition of mitophagy limited stem cell self-renewal through effects on the age of mitochondria segregated to daughter cells [168], it will be important to obtain a better understanding of how mitophagy modulates cell fate decisions.

5. Mitophagy and inflammatory signaling

Macrophages play a central role in detecting and eliminating both pathogens and damaged cells from the organism with recent work highlighting a role for mitophagy in this response [12]. Macrophages are activated by both pathogen associated molecular patterns (PAMPs), such as bacterial lipopolysaccharides and viral DNA, and also by so-called damage-associated molecular patterns (DAMPs), such as extracellular ATP, HMG-B1 and mtDNA that emanate from host stress responses and dying cells [182]. These “danger” signals bind to receptors such as Toll-like receptors (TLRs) at the plasma membrane or to intra-cellular platforms including the Nod-like receptors (NLRs) that make up the inflammasome, resulting in NF-κB-dependent induction of cytokine expression, as well as caspase-1 activation and maturation and secretion of pro-inflammatory cytokines (IL-1β and IL-18) [182]. Initial activation of the nucleotide binding domain and leucine rich repeat pyrin domain containing 3 (NLRP3) inflammasome via caspase-1 at the mitochondria induces mitochondrial damage, including mitochondrial membrane depolarization, ROS generation and mtDNA release which feeds back to amplify NLRP3 activation [183–185]. Oxidized mtDNA released from damaged mitochondria then binds directly to the NLRP3 inflammasome localized at the mitochondria in perinuclear regions, to amplify pro-caspase-1 activation and further promote IL-1β release (Fig. 4) [189–191]. Mitochondrial amplification of inflammasome activation requires VDAC1 or VDAC2 and may be suppressed by over-expression of Bcl-2 [187] although the role of Bcl-2 in control of inflammasome activation is controversial, with other studies showing Bcl-2 and related molecules to have no impact on NLRP3-mediated production of IL-1β [189]. Together, observations to date indicate that the inflammasome can induce and sense mitochondrial damage [188]. It has also been suggested that mitochondria may promote inflammasome complex assembly [188].

Fig. 4.

Mitophagy limits inflammasome activation. The inflammasome, made up of NLRP3, ASC and pro-caspase-1 is activated by viral DNA, bacterial lipopolysaccharides and also by mtDNA. Release of oxidized mtDNA from damaged mitochondria promotes assembly and activation of the NLRP3 inflammasome leading to suggestions that the inflammasome plays a role in sensing mitochondrial damage. Once activated, the inflammasome through processed caspase-1 induces cytokine maturation and other aspects of the inflammatory response. Mitophagy plays a role in limiting inflammasome activation by getting rid of damaged mitochondria. This type of mitophagy is induced by mitochondrial depolarization and PINK1/Parkin-dependent recruitment of p62/SQSTM1 to the mitochondria to promote turnover at the autophagolysosome. Caspase-1 activation can limit mitophagy by cleaving Parkin resulting in amplification of inflammasome activation. NF-κB plays a role in the response to inflammasome activation through transcriptional induction of key cytokines and upregulation of p62/SQSTM1. MAVS also contributes to mitophagy activation by inducing IRF3 to induce TBK1. Thus, mitophagy plays a role in attenuating the inflammasome and limiting cell death and tissue degradation and conversely, the inflammasome plays a role in sensing and responding to mitochondrial damage.

Damage to mitochondria arising from activation of the NLRP3 inflammasome induces Parkin-dependent mitophagy that feeds back to limit inflammasome activation (Fig. 4) [75]. This explains in part how defective autophagy hyper-activates the inflammasome [188,193] [186], although autophagy also plays a role in directly eliminating the inflammasome itself [194]. Intriguingly, Parkin is cleaved by caspase-1 to limit mitophagy with the resultant excess inflammation leading to pyropoptosis, suggesting another mechanism by which caspase-1 acts to amplify its own activity [183]. Mitophagy of damaged mitochondria in activated macrophages was also dependent on mitochondrial recruitment of p62/Sqstm1 by signals that induced NLRP3 inflammasomes [75]. NF-κB also induces the expression of p62/SQSTM1 in response to activation of the NLRP3 inflammasome, although the kinetics of p62/SQSTM1 induction was delayed relative to that of cytokines, such as IL-1β [75]. Expression of the other autophagy adaptors (NBR1, OPTN1, NDP52, TAX1BP1) was not induced by inflammasome activation [75], suggesting that p62/SQSTM1 plays a unique role here.

MAVS (mitochondrial anti-viral signaling molecule) is an OMM protein involved in anti-viral signaling that is also required for NLRP3 recruitment to mitochondria (Fig. 4) [192]. Mice deficient for MAVS were protected from inflammasome-dependent renal dysfunction during acute tubular necrosis [192]. Interestingly, MAVS signals downstream to activate interferon response factor 3 (IRF3) via TBK1 [196] which also induces mitophagy via the recruitment of OPTN1 to depolarized mitochondria [51]. This suggests that TBK1 may be involved in integrating inflammasome activation and induction of mitophagy during the interferon response to pathogens.

In summary, by eliminating damaged mitochondria from activated macrophages, mitophagy suppresses inflammasome activation and conversely, mitophagy inhibition leads to excessive inflammation, cell death and loss of tissue integrity [12,191]. While it has not been directly explored, this important function of mitophagy seems likely to affect tumorigenesis given the known role of inflammation in promoting cancer progression [197].

6. A link between DNA damage responses and mitophagy?

Recent work has suggested a link between mitophagy and DNA damage sensing and repair mechanisms in the cell [53,161,198–200]. Mitochondrial dysfunction has been noted in cells lacking the ATM kinase, a protein that plays a critical role in DNA damage responses and is inactivated in Ataxia-telangiectasia (AT), a cancer predisposition syndrome in which patients display radio-sensitivity amongst other symptoms [198]. Specifically, it was shown that ATM loss results in increased mitochondrial ROS generation, increased respiration, mtDNA depletion and reduced mitophagy [198–200]. Intriguingly, ATM was shown to localize to mitochondria in thymocytes and to be activated by mitochondrial dysfunction [200]. Consistent with some of these findings, neurons lacking ATM in both mice and worms were shown to exhibit reduced mitophagy [161]. This was attributed to NAD⁺ depletion due to elevated PARP-1 activity induced by DNA damage resulting in reduced SIRT1 activity that competes with PARP1 for NAD⁺ [161]. By restoring NAD⁺ and activating SIRT1 in ATM deficient neurons, Fang and colleagues were able to rescue cell death and improve neuronal differentiation [161]. Moreover, NAD⁺ precursors, such as nicotinamide riboside or nicotinamide mononucleoside, induced mitophagy in mitophagy-defective ATM deficient neurons although they had negligible effects on wild-type neurons. Similar to mammalian neurons, atm1⁻ worms showed reduced levels of dct-1 (the worm homolog of BNIP3/BNIP3L) and were also defective for mitophagy in a manner that could be rescued by supplying nicotinamide riboside [161]. Dct-1 was previously shown to limit aging in worms through effects on mitophagy [165]. Levels of dct-1 in atm1-worms were increased by nicotinamide riboside in a Sir2-dependent manner that also required daf16, the upstream regulator of dct-1 [161]. This has led to the model in which unrepaired DNA damage in ATM deficient cells induces abnormally high PARP-1 activity, NAD⁺ depletion and reduced activity of SIRT1 and other Sirtuins (Fig. 5A). This leads to reduced activity of transcription factors (Daf16) that induce expression of the mitophagy adaptor, Dct-1 in worms, or BNIP3/BNIP3L in mammals, ultimately blocking mitophagy (Fig. 5A) [161]. There remain numerous open-ended questions here, including whether BNIP3 or BNIP3L play a similar role downstream of SIRT1 and ATM deficiency in mammalian systems–since most of the in vivo work by Fang et al. was performed in worms. Both BNIP3 and BNIP3L are p53 targets [104,105] although BNIP3 is suppressed by p53 [104] while BNIP3L is induced by p53 [105], suggesting that perhaps in mammalian system, BNIP3L is the key player promoting mitophagy in response to DNA damage and ATM activity. However, p53 is inhibited by SIRT1 [199] suggesting that de-repression of BNIP3 may also contribute to mitophagy downstream of DNA damage signals. In addition, FoxO3A is activated by SIRT1/2 [203] and acts to promote BNIP3-dependent mitophagy (Fig. 5A) [98]. Absence of DNA damage-induced mitophagy caused by failure to turn on BNIP3L or to de-repress BNIP3 would in turn be predicted to contribute to tumor progression in AT and other cancer predisposition syndromes as dysfunctional mitochondria accumulate. Clearly the signaling network controlling mitophagy in response to DNA damage is more complex in mammals than in worms but the role of BNIP3 and BNIP3L in DNA damage induced mitophagy appears to be an area requiring further investigation.

Fig. 5.

Activation of Mitophagy during DNA damage responses.(A) Mitophagy is activated downstream of DNA damage as a result of NAD+ depletion arising from elevated PARP1 activity and loss of ATM function that in turn diminished SIRT1 activity and de-acetylation of transcription factors, such as Daf-16 that induces Dct-1, the worm homolog of BNIP3/BNIP3L. In mammals, the relative importance of sirtuins in activating mitophagy remains to be tested. However, both BNIP3 and BNIP3L are likely to be influenced by NAD⁺ levels and sirtuins that regulate their respective upstream regulators, including FoxO3A and p53 in mammals. (B) A screen for genes that influenced mitophagy and virophagy identified FANC-C, a component of the Fanconi anemia DNA damage sensing and repair complex. FANC-C was shown to bind directly to Parkin and to be required for Parkin-dependent mitophagy. Additional FA proteins were also identified as being involved in mitophagy. The relative contribution of the DNA repair functions of FA proteins versus their mitophagy functions to different disease manifestations of FA, such as hematological malignancies and susceptibility to infection, remains to be determined.

Another compelling link between mitophagy and DNA damage sensing and repair mechanisms was made recently with the observation that various Fanconi anemia proteins play an indispensible role in Parkin-mediated mitophagy (Fig. 5B) [53]. Fanconi anemia (FA) is an autosomal recessive disease marked by bone marrow failure, reduced fertility, skeletal abnormalities, hyper-sensitivity to DNA cross-linking agents and cancer susceptibility [204]. The causative genetic lesions fall into different FANC complementation groups with the encoded genes contributing to a DNA repair complex required to resolve interstrand DNA crosslinks [202]. Inactivation of any one component in the complex leads to the syndrome of symptoms associated with FA [204]. FANC-D1 is encoded by the BRCA2 tumor suppressor gene and other components of the FA DNA repair complex interact with the BRCA1 tumor suppressor [205]. Intriguingly, a siRNA screen identified FANC-C and other FA complex components as being required for mitophagy and virophagy, which eliminates viruses from cells [53]. FANC-C interacted directly with Parkin and showed increased localization with Parkin to the mitochondria in response to mitochondrial membrane depolarization. Furthermore, inactivation of FANC-C resulted in accumulation of dysfunctional mitochondria, elevated ROS production and increased inflammasome activity consistent with a mitophagy defect (Fig. 5B) [53]. Other FA genes, including FANC-A, FANC-D2, FANC-F, FANC-L and BRCA1 were also required for mitophagy and viral clearance in this study. This novel role for FA proteins in mitophagy was genetically distinct from their role in repair of DNA cross-links [53]. Interestingly, FA patients are more susceptible to infection as one would expect if they are defective for virophagy. With a defect in mitophagy, it is plausible that altered cytokine production and heightened inflammasome activity may also contribute to the hematological pathologies in these patients [53,201].

In summary, activation of mitophagy downstream of or in parallel to DNA damage signaling pathways may serve to limit ROS production at a time when the cell is already challenged to maintain genome integrity. Failure to activate mitophagy in cancer predisposition syndromes linked to DNA repair defects, such as AT and FA, may thus contribute to the overall phenotype of the disease in people and in mouse models but much remains to be investigated.

7. Concluding remarks

Mitophagy plays an important homeostatic function in cells and tissues by preventing accumulation of defective mitochondria that produce damaging ROS and altered metabolites that can reprogram nuclear gene transcription and disrupt normal metabolic responses to stress. Here we have highlighted some of the more recent and exciting findings linking defective mitophagy to failures in proper reprograming of cellular metabolism, control of cell fate determination, attenuation of inflammation and response to DNA damage, all of which likely contribute to the role of mitophagy in cancer. We have also highlighted areas that need further investigation, including how the activities of the various different mitophagy modulators are coordinated or indeed how they interact functionally with each other. It is also clear that more investigation of certain mitophagy proteins in a physiological setting, including in cancer and mouse models of cancer, is warranted. With the development of transgenic mice expressing mt-mKeima [206], it is now possible to specifically examine mitophagy in vivo where previously only mitochondrial mass, autophagy and indirect measures of mitophagy were possible. This and other novel techniques should facilitate greater insight to the physiological role of mitophagy in future studies.

Abbreviations

ALS

amyotrophic lateral sclerosis

AR

androgen receptor

ASC

apoptosis-associated speck-like protein containing a CARD

ATM

ataxia telangiectasia mutated

BCL-2

breakpoint cluster locus-2

BCL-XL

BCL2-like 1 long

BNIP3

BCL2 interacting protein 3

BNIP3L

BCL2 interacting protein 3 like

BRCA1

breast cancer 1

CCCP

carbonyl cyanide 3-chlorophenylhydrazone

CK2

casein kinase-2

CL

cardiolipin

DRP1

dynamin related protein 1

ETC

electron transport chain

FA

Fanconi anemia

FANC-C

FA protein C

FUNDC1

FUN 14 domain containing 1

GAP

GTPase activating protein

GDI

GDP dissociation inhibitor

HCC

hepatocellular carcinoma

IMM

inner mitochondrial membrane

IRF3

interferon response factor 3

LC3

microtubule associated protein 1 light chain 3

LIR

LC3 interacting region

mtDNA

mitochondrial DNA

MAVS

mitochondrial anti-viral signaling protein

MFN

mitofusin

NBR1

near BRCA1

NDP52

nuclear dot protein 52

NLRP3

nucleotide binding domain and leucine rich repeat pyrin domain containing 3

NSCLC

non-small cell lung carcinoma

OMM

outer mitochondrial membrane

OPTN

optineurin: p62/SQSTM1

PARK2

Parkin encoding locus

PARK6

PINK1 encoding locus

PARK8

LRRK2 encoding locus

PD

Parkinson’s disease

PDAC

pancreatic ductal adenocarcinoma

PGAM5

phosphoglycerate mutase family member 5

PHB-2

prohibitin-2

PINK1

PTEN induced putative kinase 1

PLK1

Polo-like kinase-1

ROS

reactive oxygen species

SRC

Rous sarcoma oncogene

TAX1BP1

Tax1 binding protein 1

TAZ

tafazzin

TBK

TANK binding kinase 1

ULK1

Unc-51-like autophagy activating kinase 1