Introduction

Mitochondria play pivotal roles in the production of cellular ATP and metabolites required for normal cellular activities essential for cell survival and programmed cell death (both apoptosis and programmed necrosis) [1–4]. Moreover, mitochondria produce superoxide as an inevitable byproduct during electron transfer along the mitochondrial respiratory chain [5–8]. In addition, mitochondria are the center for iron metabolism for the synthesis of both the heme and iron–sulfur cluster, which are two classes of iron-containing molecules [7, 9]. To fulfill such diverse or even opposing functions, mitochondrial quality must be tightly monitored to avoid harmful effects from mitochondria and to maintain the health of the cell [10–14]. The accumulation of dysfunctional mitochondria is the characteristic of different types of diseases, including heart failure, Alzheimer’s disease, Parkinson’s disease and cancers [15–18]. One of the reasons could be defective mitochondrial quality control. A better understanding of the molecular mechanism of mitochondrial quality control holds promise in the fight against these incurable diseases.

Mitophagy is regarded as the major mechanism of mitochondrial quality control. Early studies observed the engulfment of mitochondria in glucagon-stimulated hepatocytes by electron microscopy [19, 20]. Lemasters et al. [21∙∙] more specifically described the selective autophagy of mitochondria as mitophagy. The authors went on to suggest that mitochondrially derived reactive oxygen species (ROS) can initiate mitophagy, likely via the mechanism of mitochondrial permeability transition (mPT) [21∙∙]. The opening of mPT pores renders mitochondria permeable to all solutes with a molecular mass up to 1,500 Da [22]. This permeability will lead to mitochondrial depolarization and reverse operation of the mitochondrial ATPase [23, 24]. Various pathological conditions, such as ischemia and reperfusion, often induce the loss of mitochondrial membrane potential (Δψ) and the opening of mPT pores, mechanisms that may be involved in ROS-induced ROS release or the so-called vicious cycle [25–27]. The exact mechanism explaining how mPT triggers mitophagy has yet to be defined.

A breakthrough in our understanding of the mechanisms of mitophagy came from Youle’s seminal findings [28∙∙]. His group found that in response to a loss of mitochondrial membrane potential, Parkin can translocate onto mitochondria for the selective removal of mitochondria that have lost membrane potential. In healthy mitochondria, phosphatase and tensin homolog (PTEN)-induced putative kinase 1 (PINK1) is constitutively imported, most likely via the TIM/TOM complex, to the inner membrane, where the kinase is cleaved by several proteases and ultimately proteolytically degraded [28∙∙]. When induced by CCCP or other uncouplers, PINK1 is stabilized on the outer mitochondrial membrane (OMM) where it can selectively recruit Parkin to depolarized mitochondria [29]. Parkin, as a potent E3 ligase, can polyubiquitinate numerous mitochondrial substrates that will ultimately lead to mitophagy [30∙, 31–34]. In particular, the ubiquitination of the mitofusins (Mfn1 and Mfn2), large GTPases that mediate mitochondrial fusion, will lead to their degradation and mitochondrial fission [30∙, 32]. The segregated mitochondria will undergo autolysosome-dependent degradation [35]. Mizushima and colleagues found that structures containing upstream autophagic proteins, including the ULK1 complex, can associate with depolarized mitochondria, even in the absence of LC3 [36∙]. Atg9A is recruited to autophagosome formation sites independent of the ULK1 complex [36∙]. LC3 is then recruited for the efficient incorporation of damaged mitochondria into the autophagosome at a later stage [36∙].

Receptors in Mitophagy

Since the first description of mitophagy, it was suggested that damaged mitochondria may crosstalk with autophagy machinery [37]. In other words, a mitochondrial surface protein may act as a cargo receptor that interacts with autophagy machinery to form selective mitophagy machinery [37]. An early study by Camougrand’s group identified an OMM protein, Uth1p, required for the autophagic degradation of excess mitochondria [38, 39]. Furthermore, ancient ubiquitous protein 1 (also known as Ptc6), a phosphatase 2C (PP2C) superfamily member, can efficiently eliminate overworked mitochondria in stationary-phase yeast cells [40].

In search of a specific mitophagy receptor, both Ohsumi’s and Klionsky’s groups [41∙∙, 42∙∙] have found that Atg32 is a mitophagy receptor in yeast. Atg32 is a 60-kDa protein with an intra-mitochondrial domain at its C terminus and the tetrapeptide sequence WQAI in its cytoplasmic domain, which is critical for the protein’s interaction with Atg8 [41∙∙, 42∙∙]. Mitochondrial proteins containing the tetrapeptide sequence W(Y)XXL, which is present in several other Atg8- or LC3-binding partners, are defined as mitophagy receptors mediating selective autophagy [41∙∙, 42∙∙]. Atg32 also binds to Atg11, which is a scaffold protein important for cargo selection [43] and autophagosomal formation [41∙∙, 42∙∙]. Unfortunately, there are no orthologs of Atg32, Uth1p and Aup1p in the mammalian system [37].

What are the mitochondrial molecules that mediated mitophagy in the mammalian system? Previous studies have suggested that NIX/BNIP3L is involved in the autophagic degradation of mitochondria in reticulocytes, a process essential for red blood cell maturation [44∙∙]. Genetic ablation of Nix retards mitochondrial clearance in maturing murine reticulocytes [45, 46∙∙]. Research from Ivan Dikic and colleagues [46∙∙] has subsequently shown that NIX also has a putative LC3-interacting region (LIR) that interacts with LC3 and defines NIX as a mitophagy receptor. BNIP3 also contains a typical LIR motif and functions as a mitophagy receptor [47, 48∙].

We have found that FUNDC1, an OMM protein, has the typical LC3-binding motif Y(18)xxL and have identified this protein as a new mitophagy receptor [49∙∙]. In response to hypoxia, FUNDC1 interacts with LC3 through its LIR, and mutation of the LIR impairs its interaction with LC3 and the subsequent induction of mitophagy. Knockdown of endogenous FUNDC1 significantly prevents hypoxia-induced mitophagy, which can be rescued by the expression of wild-type FUNDC1 but not LC3 interaction-deficient FUNDC1 mutants [49∙∙].

Both NIX/BNIP3 and FUNDC1 are involved in hypoxia-induced mitophagy. However, their mechanisms of action are different. Nix and BNIP3 are transcriptionally upregulated through HIF or FOXO3 [50, 51], whereas the mRNA level of FUNDC1 decreases under hypoxic conditions [49∙∙]. Nix and BNIP3 are localized to the ER and are involved in regulating apoptotic or programmed necrosis by affecting mitochondrial respiration or ROS production [51, 52], whereas FUNDC1 is exclusively localized on the OMM. NIX shows much weaker binding to LC3B, whereas the conversion of LC3B is correlated with elevated levels of autophagic vesicles in response to autophagy-inducing stress [49∙∙].

In addition to these receptors, p62 has also been implicated in selective mitophagy [31]. Early studies have shown that p62 functions as an adaptor, or scaffold protein, that participates in the regulation of the NF-kappa B pathway through its interactions with atypical PKC (aPKC), RIP1 kinase or TRAF6 [53–56]. Recently, p62 was reported to act as a cargo receptor for ubiquitinated protein aggregates intended for selective autophagy [57–60]. It was found that p62 translocates onto mitochondria via interaction with ubiquitin, which is significantly increased in response to mitochondrial stresses. Through this interaction, p62 could theoretically recruit LC3 and autophagosomes to facilitate subsequent selective autophagy [61∙, 62]. Whether ubiquitin/p62/LC3 complexing is sufficient for Parkin-mediated mitophagy requires further investigation.

Kinases and Phosphatases in Mitophagy

Mitophagy is a highly regulated and complex process essential for the mitochondrial turnover and quality control. A key question is how mitochondrial signals are sensed and then transduced to activate the autophagy machinery. Recently, post-translational modification, such as protein phosphorylation, was found to regulate the mitophagic process [63, 64]. ATG32 is phosphorylated in response to oxidative stress that promotes its interaction with ATG11 and LC3 [65, 66]. In particular, Ser-114 and Ser-119 of Atg32 are phosphorylated, which promotes the Atg11-Atg32 interaction. However, phosphorylation of these sites is not required for Atg32-Atg8 interaction [65]. Likewise, phosphorylation of Ser-17 and Ser-24 of Bnip3 also promotes its binding to LC3-B and GATE-16 [48∙] to facilitate subsequent mitophagy. In contrast, dephosphorylation activates FUNDC1-mediated mitophagy in mammalian systems [49∙∙]. We found that under normal (unstressed) conditions, FUNDC1 is phosphorylated, inhibiting its interaction with LC3 and subsequent mitophagy. Under hypoxic conditions, FUNDC1 is dephosphorylated, which will enhance its interaction with LC3 and subsequent mitophagy [49∙∙].

What are the kinases and phosphatases that mediate this reversible phosphorylation? A recent report using kinase screening showed that casein kinase 2 (CK2) phosphorylates the OMM protein Atg32 at Ser-114 and Ser-119 in vitro to promote its interaction with the cytosolic adaptor protein Atg11 in yeast, thus resulting in recruitment into vacuoles and mitophagy [67∙∙] Furthermore, CK2 inhibition selectively prevents Atg32-Atg11 interaction-dependent mitophagy without affecting the Cvt pathway, pexophagy and macroautophagy [67∙∙]. The adaptor protein Atg11 preferentially interacts via Atg32 with phosphorylated Ser-114 and relocates the mitochondrion to pre-autophagosomal complexes and phagophore assembly sites. At last, isolation membranes are produced, and mitochondria are specifically enveloped by autophagosomes. Therefore, Atg32 phosphorylation may represent mitophagy recognition and initiation signaling (see Fig. 1).

Fig. 1
figure 1

Phosphorylation regulates selective mitochondrial autophagy. There are PINK1/Parkin- dependent and receptor-mediated mitophagy pathways. In the PINK1/Parkin pathway, upon the loss of membrane potential in damaged mitochondria, PINK1 is stabilized and accumulated at the outer membrane to recruit Parkin, an E3 ligase to mitochondria. Parkin can be phosphorylated by PINK1 at Ser-65, and the activated Parkin mediates hyperubiquitination of the mitochondrial outer membrane proteins. P62 and other autophagy molecules are also recruited toward damaged mitochondria. Although its role in the PINK1/Parkin pathway remains controversial, P62 can also be phosphorylated by mTORC1 at Ser-351 and CK2 at Ser-403. ATG32, NIX/Bnip3 and FUNDC1 were identified as mitophagy receptors. In yeast, upon the oxidative stress, mitochondrial outer membrane protein Atg32 is phosphorylated at Ser-114 and Ser-119 by CK2, which is presumably activated by the MAPK signaling pathway, Hog1. This phosphorylation mediates the Atg32-Atg11 interaction and leads to the recruitment of mitochondria to PAS, and subsequently mitochondria surrounded by the phagophore membrane. In mammalian cells, mitochondrial outer membrane protein FUNDC1 is normally phosphorylated at Tyr-18 by Src kinase. Under hypoxic conditions, Src kinase is inactivated, and FUNDC1 becomes dephosphorylated. The dephosphorylated FUNDC1 will promote its interaction with LC3 for subsequent mitophagy. Also, under metabolic stress conditions Bnip3 protein can be phosphorylated by an unidentified kinase at Ser-17 and Ser-24. During red blood maturation, NIX protein mediates mitochondria removal via the mitophagy mechanism, and it is yet to be determined whether any kinase is involved

Full size image

We found that FUNDC1-mediated mitophagy is inhibited by its phosphorylation at the Tyr-18 position in the LIR motif by Src kinase under normal physiological conditions. Upon hypoxia stimulation, Src is inactivated and degraded, and FUNDC1 becomes dephosphorylated, resulting in an increased interaction between FUNDC1 and LC3-II, leading to the selective incorporation and autophagic removal of the mitochondrion [49∙∙]. We propose that dephosphorylation of the Tyr-18 position in the LIR motif may cause interface exposure of the FUNDC1 protein, thus promoting its interaction with LC3 (our unpublished observations), although the exact molecular mechanisms need further investigation (see Fig. 1).

Interestingly, PINK1 kinase can phosphorylate Parkin for its mitophagic activity. Specifically, PINK1 was found to phosphorylate the Ubl domain of Parkin at Ser-65 and to enhance the activity of ubiquitin ligase in in vitro assays [68∙, 69]. Mutation of Ser-65 to Ala dramatically inhibits the recruitment of Parkin onto the mitochondria with depolarized potential, indicating that phosphorylation of Ser-65 primes the translocation of Parkin [68∙]. The exact mechanism of how PINK1 phosphorylation activates Parkin’s activities and translocation remains obscure. Other tyrosine protein kinases, such as c-Abl, diminish activity of Parkin ubiquitination both in vitro and in the cultured cells [70, 71]. Inactivation of Parkin is also implicated in oxidative stress conditions [72]. In addition, Parkin activation in the depolarized mitochondria induces proteasome-mediated degradation of Parkin, indicating that Parkin auto-inhibition might prevent ubiquitin-dependent Parkin degradation [73, 74]. Recently, Chen et al. [75] also found that Pink1 phosphorylated Mfn2 and promoted its Parkin-mediated ubiquitination, which functions as a mitochondrial receptor for Parkin and is required for quality control of cardiac mitochondria (see Fig. 1).

P62 is considered to act as a cargo receptor in selective autophagy (including mitophagy), and it was also found to be phosphorylated at multiple sites [76]. Specific phosphorylation of p62 at Ser-403 (S403) by CK2 in its ubiquitin-associated (UBA) domain increases the affinity between UBA and polyubiquitin chains [76]. This event will enhance p62′s targeting to polyubiquitinated proteins and promote autophagosomal formation. Additionally, mTOR was found to phosphorylates p62 at Ser-351 [77], which increase its binding to Keap1, an adaptor of the Cul3-ubiquitin E3 ligases, and sequesters Keap1 on its cargos such as mitochondria.

Sensing the Signals in Mitophagy

As a vital organelle with diverse cellular functions, mitochondria are highly sensitive to the cellular bioenergetic status, oxygen tension, ROS (including NO) levels, and other cellular and environmental cues. It is thus not surprising that a loss of mitochondrial membrane potential, an increase in cellular ROS [78] (either derived from the cytoplasmic or inside mitochondria), hypoxia [49∙∙], a disturbance of Ca2+ signaling, impaired bioenergetic and biogenesis of mitochondria [79], defects in mitochondrial protein import or export [79], mtDNA damages [80], a perturbation of mitochondrial protein quality control system and an accumulation of protein aggregates in mitochondria [81] are all able to trigger mitophagy. Given the critical role of reversible phosphorylation in mitophagy, it is reasonable to suggest that kinases or phosphatases are likely to play a role in sensing the mitochondrial stress signals responsible for the initiation of mitophagy.

Early studies have clearly shown that ROS can activate mitophagy [21∙∙]. It was reported that N-acetylcysteine (NAC), a compound that increases the cellular GSH pool, prevents mitophagy induction [82, 83], suggesting that the mitochondrial redox status or the ROS production level is one of the factors contributing to mitophagy. How do ROS signals lead to the activation of mitophagy? Although ROSs are short-lived molecules by nature, they can induce oxidative modification of ATGs and related pathways. Under oxidative conditions, Atg4 is oxidized and inactive, which allows Atg8 to bind to lipid and initiates autophagy, revealing that the modulation of the action of Atg4 on Atg8 can be controlled by H2O2 [84]. This phenomenon does not explain the selective effects of mitophagy. One possibility is that ROS production or the activation of the vicious cycle could lead to the opening of permeability transition pores and a loss of mitochondrial membrane potential. The loss of mitochondrial membrane potential may activate the PINK1/Parkin pathway of mitophagy. Indeed, it was shown that an acute burst of ROS within mitochondria was induced by a mitochondrial-targeted photosensitizer and that mitochondrial KillerRed (mtKR) resulted in a loss of membrane potential and subsequent activation of PARK2-dependent mitophagy [85]. ROS is implicated in NIX- and FCCP/Parkin-induced mitophagy [86]. The antioxidant NAC can effectively decrease the mitophagy induced by CCCP. ROS and the mitophagy receptor NIX are important in the induction and initiation of mitophagy by enhancing the translocation of Parkin onto the damaged mitochondria [86]. In the yeast system, CK2/ATG32 pathways can be regulated by MAPK Hog1 pathways, which can be modulated by oxidative stress and thus the downstream mitophagic activities [87]. In mammalian systems, we found that hypoxia can inactivate Src kinase to activate FUNDC1-mediated mitophagy [49∙∙]. NAC can also prevent hypoxia-induced mitophagy, suggesting that ROSs are involved (unpublished observations).

Additionally, studies have found that selective mitophagy induced by 6-OHDA, rotenone and staurosporine in neuron cells involve the externalization of cardiolipin, which binds LC3 to mediate mitophagy [88]. The Parkinson’s toxins result in a delayed phase of mitochondrial superoxide production that causes localized activation of ERK/MAPK at the mitochondria. Furthermore, mitochondrial localization of ERK is sufficient to induce mitophagy, even in the absence of mitochondrial injury [89, 90].

Mitophagy and Diseases

An accumulation of dysfunctional mitochondria and increased oxidative stress are characteristic of several diseases, including neurodegenerative diseases, cardiomyopathy and cancers [91–94]. Defective mitophagy is thus implicated in these diseases, although the molecular details remain to be defined. Due to space limitations, we only highlight a few diseases pertinent to mitophagy signaling and discuss how mitophagy contributes to these diseases.

PINK1/Parkin-Mediated Mitophagy and Parkinson’s Disease

Mutations in Parkin and PINK1 have been shown to be responsible for a small subset of autosomal-recessive Parkinson’s disease cases [95]. Mice deficient in Parkin had an abnormal mitochondrial respiratory chain [96]. Flies with Parkin knockout had pathology in muscle, with dysfunctional mitochondria and pronounced apoptosis [97, 98]. These results suggest that Parkin mutation that causes the defective of mitophagy indeed contributes to the disease. It is well documented that Parkin mediates the mitophagy in a PINK1-dependent manner [68∙, 99]. PINK1 mutations result in decreased Δψ and dysfunction of mitochondria. Although flies with PINK1 knockout are viable, they have a motor deficit and a decreased lifespan and are hypofertile or sterile [100]. In particular, these flies have abnormal mitochondrial morphology, decreased mitochondrial mass, lower ATP levels, flight muscle dysfunction and reduced dopaminergic neurons. Recent data suggest that PINK1 interacts with a mitochondrial phosphatase PGAM5, although they may not be reciprocal substrates [101]. A deficiency in PGAM5 inhibits the mitochondrial degeneration induced by PINK1 inactivation in Drosophila, suggesting that PINK1/Parkin- mediated mitophagy could be modulated by PGAM5 [102].

Parkin is a multifunctional protein that plays diverse roles in mitochondrial biogenesis, mitophagy, mitochondrial dynamics and transport through its regulation of PARIS (controlling the PGC-1alpha-NRF-1 pathway) [103], MFN1/2 [30∙, 75] and Drp-1 [[104]. Much of the work on PINK1/Parkin-mediated mitophagy has been performed in immortalized cell lines overexpressing high levels of Parkin [105]. The roles of endogenous Parkin and PINK1 that contribute to mitophagy in neurons in animal systems require further clarification. Identification of the upstream signaling mechanisms that regulate Parkin/PINK1-dependent mitophagy will help to explain the nature of the insults affecting mitophagy and mitochondrial functions.

Src Kinase/FUNDC1 Axis in Cancer

As early as the 1920s, Otto Warburg proposed a hypothesis to explain why tumors exhibited a metabolic energy change from aerobic respiration to glycolysis [106∙]. Defective mitochondria could be responsible for the Warburg effect, a widely accepted concept among cancer biologists [107, 108]. It is possible that damaged mitochondria progressively accumulate during cancer development [109, 110]. Impaired mitochondrial metabolism may provide a certain advantage in the increased glycolysis and lactate production contributing to the Warburg effect [111, 112]. Additionally, mitochondria are a major source of ROS production in cells, which may affect genomic instability via retrograde signaling [113].

Similar to the proposed dual roles of autophagy in cancers, mitophagy could be a double-edged sword in the development of cancer [16, 109]. During the early stage of cancer development, mitophagy functions to effectively remove damaged mitochondria to maintain the well-being of the cells and to prevent the occurrence of cancers [16]. During the late stage of cancer development, the hypoxic conditions inside the tumor will activate mitophagy, which can digest unwanted mitochondria to sustain the survival of the cancer [114, 115].

Src kinase is well known for its function in cancer development [116]. It is interesting to note that one of this kinase’s functions is to prevent mitophagy, as we demonstrated previously [49∙∙]. This function will lead to the accumulation of damaged mitochondria, which promotes the occurrence of cancer [117]. Additionally, hypoxia within cancer during the stage of cancer development can activate mitophagy, which will benefit cancer growth [110]. Further studies are underway to test how Src kinase modulates mitophagy in cancers.

NIX/BNIP3-Dependent Mitophagy and Heart Disease

Mitochondrial dysfunction is also implicated in the pathogenesis of several heart diseases during aging [118]. Given the postmitotic nature of cardiomyocytes, the efficient removal of dysfunctional mitochondria by mitophagy is critical for the maintenance of normal cellular functions and homeostasis [119]. Indeed, cardiac-specific ablation of both Nix and BNIP3 revealed normal cardiac structure and function at the age of 8 weeks in these mice [120]. However, by 30 weeks, these mice had developed massive cardiac abnormality. EM analysis revealed an abnormality in mitophagy and autophagic vesicle formation in senescent hearts deficient in NIX and/or BNIP3 [52]. As both NIX and BNIP3 also function in programmed cell death by affecting mPT and mitochondrial respiration, definitive proof connecting defective mitophagy and the resulting mitochondrial abnormality requires further experimental analysis. Additionally, the signaling events in mitophagy and myocardial injury warrant further investigations. For example, ischemia and reperfusion, which often incur side effects in the clinic, can trigger the opening of mPT pores, decreasing the mitochondrial membrane potential and massively increasing ROS levels [120]. These events may elicit a complex signaling pathway that activates mitophagy as a protective response or programmed cell death when the stresses are severe and persistent. It would be interesting to understand how these signals impinge upon NIX/BNIP3-dependent mitophagy. This information is important because the protective cardioprotective pathways relying on mitophagy, along with pharmacological drugs, might prevent heart damages in pathogenic heart conditions.

Detection of Mitophagy

Because the accumulation of damaged mitochondria correlates with the aging-related diseases, overall mitochondrial quality or mitophagic activities can serve as a prognostic indicator of disease risk. Additionally, advancing mitophagy research requires reliable and quantitative analysis of mitophagy. Like in autophagy, transmission electron microscopy (TEM) remains the best approach to detect mitophagy [121]. The engulfment of mitochondria by double-membrane autophagosomes can be visualized as a unique mitochondrial structure or a high electron density of mitochondria. Immuno-EM for specific mitochondrial markers, such as Tom20 or VDAC, and other mitochondrial proteins is important for confirming mitophagy. EM can be problematic in quantitative studies because of the limited cell numbers/sections that can be examined. Additionally, mitophagy is highly dynamic, so it is difficult to the capture typical engulfment of mitochondria within the autophagosomal membrane. Data from EM studies are highly subjective, and unbiased selection of samples is extremely important.

Other analyses include Western blotting of mitochondrial proteins from mitochondrial outer and inner membrane proteins, such as Tom 20 and Tim23, together with mitochondrial matrix proteins such as MnSOD or HSP60; measurements of the amount of mitochondrial DNA; and the colocalization of LC-3 and other autophagic markers with mitochondria [49∙∙]. It should be acknowledged that mitochondrial proteins can be degraded by other mechanisms of protein quality control, such as the AAA-ATPase inside mitochondria and proteasomal degradation of outer membrane proteins [122]. A decrease in mitochondrial DNA number is not specific to mitophagy, so caution needs to be exercised when interpreting these data. These analyses and a combination of these different measurements may provide a good indication of mitophagy.

Given the recent progress discussed above, we suggest using the reversible phosphorylation of mitophagy receptors as a biochemical marker for mitophagy. Phosphorylation of ATG32 at Ser-114 sites is clearly closely associated with mitophagy in yeast [65]. We have shown that de-phosphorylation of FUNDC1 resembles the activation of mitophagy in mammalian systems in response to hypoxia [49∙∙]. Further work is underway to examine whether dephosphorylation of FUNDC1 correlates with the activation of mitophagy in vivo and under pathophysiological conditions. One advantage is that both ATG32 and FUNDC1 are exclusively localized on mitochondria, while Parkin, p62 and BNIP3 are also phosphorylated in correlation with mitophagic activity [37]. Measurement of their phosphorylation statuses could be very useful. Profiling the distinct phosphorylation of mitophagy mediators, together with other morphological and biochemical analysis, may yield better and more reliable approaches to study mitophagy.

Conclusions

Mitophagy is an exquisitely regulated process comprising the segregation of to-be-removed mitochondria from the mitochondrial network and the crosstalk between mitochondria and the autophagy machinery. It is likely that mitophagy is highly regulated by reversible phosphorylation and is well coordinated with other mitochondrial events, such as mitochondrial ATP generation, ROS metabolism, mitochondrial dynamics and programmed cell death. Several kinases are found to play an essential role in selective mitophagy, namely PINK1 in the Parkin-dependent pathway, Hog1/CK2 in the ATG32 and Src kinase in the FUNDC1 and ERK2 though unknown targets [49∙∙, 68∙, 75, 122]. Other kinases and phosphatases are likely to be involved and are yet undiscovered. Future studies should aim to understand how these cues activate mitophagy via distinct pathways in distinct cellular contexts and how mitophagy is coordinated with other mitochondrial and cellular functions.

Defective of mitophagy may result in the accumulation of dysfunctional mitochondria, which is widely observed in aging-related diseases. Deciphering the causal roles of mitophagy in these diseases remains a challenge that requires useful model systems and reliable and quantitative measurements of mitophagy in vivo. A better understanding of these issues will not only improve our knowledge regarding general mitochondrial homeostasis but also reveal new avenues for treatments of diseases involving dysfunctional mitochondria.