Mitophagy is primarily due to alternative autophagy and requires the MAPK1 and MAPK14 signaling pathways

Cyto-Keima- and mito-Keima-expressing HeLa cells for observing macroautophagy and mitophagy

Because few studies have observed mitophagy under physiological conditions, we initially attempted to observe mitophagy induced by starvation. We cultured HeLa cells in starvation medium for up to 18 h and observed the amount of mitochondria. The volume of mitochondria and amounts of mitochondrial proteins TOMM20, TFAM, VDAC1, and TIMM23 were mostly unchanged (Fig. S1A and S1B). These findings suggest that only a small amount of the total mitochondrial pool is degraded by mitophagy induced by starvation, and that a sensitive method for detecting mitophagy is required to study starvation-induced mitophagy. To observe mitophagy, we used the pH-sensitive fluorescent protein Keima.^31-33 When Keima is present in a neutral environment, such as in the cytoplasm or mitochondrial matrix, this fluorescent protein is excited by 440-nm light, but not by 586-nm light. Interestingly, when Keima is present in acidic conditions, such as in autolysosomes, this fluorescent protein is excited by 586-nm light, but not by 440-nm light. We developed tetracycline (Tet)-suppressible cytosol-localizing Keima (cyto-Keima)-expressing and mitochondrial matrix-localizing Keima (mito-Keima)-expressing HeLa cells. When cyto-Keima cells were cultured in normal growth medium, cyto-Keima was diffuse in the cytosol and was excited by 440-nm light, but not by 590-nm light (Fig. S2A). After the cells were shifted to Krebs-Henseleit buffer to starve the cells for 4 h, autophagy delivered cyto-Keima to the lysosomal lumen, and lysosomal Keima became excited by 590-nm light, which was observed as punctate structures (hereafter referred to as “acidic puncta,” Fig. S2A, starvation). These acidic puncta increased depending on the starvation period up to 24 h (Fig. S2B). The acidic puncta were not observed when cells were treated with the phosphatidylinositol 3-kinase inhibitor 3-methyladenine (3-MA), and drastically decreased when autophagy-related PIK3C3 was knocked down by siRNA treatment, suggesting that the acidic puncta represent macroautophagy (Fig. S2A, S3A, and S3B). Similarly, when mito-Keima cells were cultured in normal growth medium, mito-Keima was localized within the mitochondrial matrix and excited by 440-nm light. Mito-Keima fluorescence showed tubular mitochondria-like shapes and completely overlapped with the mitochondria dye MitoTracker Red CMXRos (Fig. 1A; Fig. S4). To induce mitophagy, we first attempted nutrient starvation as reported previously.³⁴ After cells were shifted to Krebs-Henseleit buffer to starve the cells for 18 h, part of the mito-Keima was delivered to lysosomes and became excited by 590-nm light, which was observed as punctate structures (Fig. 1A, starvation). These acidic puncta colocalized with the lysosomal dye LysoTracker Blue or the lysosomal marker protein LAMP1-EGFP (Figs. S5A and B), and were barely observed when cells were treated with 3-MA, or when PIK3C3 was knocked down by siRNA (Fig. 1A and B). These findings suggest that the acidic puncta represent mitophagy.

Starvation and hypoxia, but not CCCP, induce mitophagy in HeLa cells

To investigate what types of cellular stress can induce mitophagy, we exposed mito-Keima cells to CCCP treatment, rapamycin treatment, starvation, and hypoxia (Figs. 2 and 3). Starvation or hypoxia for 24 h resulted in a certain number of acidic puncta, whereas CCCP treatment did not result in the formation of acidic puncta (Fig. 2A, B, and D). Some acidic puncta were observed with rapamycin treatment, but the number of these puncta was less than that with starvation or hypoxia (Fig. 2C). During starvation or hypoxia, the number of acidic puncta increased in a time-dependent manner (Fig. 2A and B; Fig. 3A and B). In contrast, rapamycin treatment did not lead to a time-dependent increase in acidic puncta (Figs. 2C and 3C). These findings suggest that starvation and hypoxia can induce, rapamycin can weakly induce, and CCCP treatment can marginally induce mitophagy in HeLa cells. We additionally tested the mitochondrial toxic reagents paraquat, antimycin A, and oligomycin, but none of them could induce mitophagy efficiently (Fig. S6–S8). On another front, the iron-chelating drug deferiprone, which was recently reported to efficiently induce PARK2-independent mitophagy,³⁵ could induce mitophagy in mito-Keima cells (Fig. S9).

To confirm that starvation and hypoxia can induce mitophagy in HeLa cells, we investigated the colocalization of the lysosomal probe LysoTracker Red and mitochondrial protein TOMM20. Before cellular stress, there was no colocalization of LysoTracker Red and TOMM20. After starvation or hypoxia, we observed some colocalization of these markers (Fig. S10). These findings are consistent with the results in mito-Keima cells that starvation and hypoxia induced mitophagy, although the majority of mitochondria were not degraded and appeared intact. LC3 is widely used as an autophagy marker. During autophagy, LC3 accumulates on phagophores, which enwrap autophagic cargos. As a result, LC3 colocalizes with the cargos. Therefore, we investigated the colocalization of LC3 and mitochondrial TOMM20 or mito-Keima during starvation-induced mitophagy in HeLa cells. Unexpectedly, we could not find any obvious colocalization of LC3 and TOMM20 or mito-Keima (Figs. S11 and S12), suggesting that mitophagy induced by starvation was not efficient enough to be detected by this method, or that LC3 was not related to mitophagy in this case. We next observed whether hypoxia could induce macroautophagy using cyto-Keima cells. Starvation and rapamycin induced macroautophagy efficiently, but hypoxia did not result in substantial induction of macroautophagy (Fig. S13). Taken together, these results suggest that starvation is a stressor for the induction of macroautophagy and mitophagy, whereas rapamycin mainly contributes to the induction of macroautophagy, and hypoxia preferentially contributes to the induction of mitophagy.

Mitophagy induced by starvation or hypoxia is related to ATG5-ATG7-independent alternative autophagy

Because mitophagy is a type of autophagy, most autophagy-related genes are generally assumed to be required for mitophagy. To test this, we separately knocked down ATG7, ATG12, and LC3 by siRNA and observed macroautophagy and mitophagy using cyto-Keima and mito-Keima cells, respectively. In HeLa cells, knockdown of either ATG7 or ATG12 did not affect macroautophagy or mitophagy (Fig. 4A and C; Fig. S14A and B). To test whether ATG7 and ATG12 were efficiently knocked down, we observed LC3 accumulation in phagophores and autophagosomes during starvation by immunostaining ATG7 and ATG12 knockdown cells. Knockdown of either ATG7 or ATG12 did not efficiently suppress LC3 accumulation on phagophores or autophagosomes, which were observed as puncta (Fig. S15A). When both ATG7 and ATG12 were knocked down by siRNA, they had a synergistic effect, and LC3 accumulation was suppressed to a large degree (Fig. S15A, ATG7 and ATG12 KD). Interestingly, in ATG7 and ATG12 knockdown cells, although macroautophagy observed by cyto-Keima was efficiently inhibited, mitophagy observed by mito-Keima was only partially suppressed (Fig. 4A and C; Fig. S14A). Finally, we knocked down all isoforms of LC3 (LC3A/B/C) and observed macroautophagy and mitophagy. Interestingly, again, although macroautophagy observed by cyto-Keima was severely inhibited, mitophagy observed by mito-Keima was only partially suppressed in LC3A/B/C knockdown cells (Fig. 4A and C; Figs. S14A and S15B). However, knockdown of BECN1 or ULK1, as well as PIK3C3, efficiently inhibited macroautophagy and mitophagy (Figs. 1B, 4B and C; Fig. S3A and B; Fig. S14A and B).

Recently, 2 types of autophagy in mammalian cells were reported, namely ATG5-ATG7-dependent conventional autophagy and ATG5-ATG7-independent alternative autophagy.¹⁴ ULK1, BECN1, PIK3C3, and RB1CC1/FIP200 are required for conventional and alternative autophagy, whereas ATG5, ATG7, ATG9, ATG12, ATG16L1, and LC3 are not required for alternative autophagy.^14,36 Therefore, we speculated that the mitophagy observed in ATG7 and ATG12 or LC3A/B/C knockdown cells might be mostly due to ATG5-ATG7-independent alternative autophagy. To confirm this possibility, we knocked down RAB9A and RAB9B, which are required for alternative autophagy, but not conventional autophagy,¹⁴ and observed macroautophagy and mitophagy. Although knockdown of RAB9 only partially inhibited macroautophagy, knockdown of ATG7, ATG12 and RAB9 or LC3 and RAB9 almost completely inhibited macroautophagy (Figs. S14B and S16), suggesting that macroautophagy observed in cyto-Keima cells is a combination of conventional and alternative autophagy. On the other hand, mitophagy was severely inhibited in RAB9 knockdown cells (Fig. 4B and C), suggesting that mitophagy is primarily due to ATG5-ATG7-independent alternative autophagy. It was recently reported that the E1-activating enzyme Uba1 is required for ATG3- and ATG7-independent autophagy in flies.³⁷ We speculated that UBA1 may be related to ATG5-ATG7-independent autophagy in mammals and that it may be required for mitophagy induced by starvation or hypoxia. To test this possibility, we knocked down UBA1 and observed mitophagy. As expected, knockdown of UBA1 efficiently inhibited mitophagy induced by both starvation and hypoxia (Fig. S17).

To further study the relationship between mitophagy and ATG5-ATG7-independent alternative autophagy, we expressed mito-Keima in Atg5 (Atg5 KO)³⁸ and Ulk1 (Ulk1 KO) knockout MEFs³⁹ and observed mitophagy. As expected, mitophagy was only partially suppressed in Atg5 KO MEFs, while it was severely suppressed in Ulk1 KO MEFs (Fig. 5). Thus, we concluded that mitophagy is primarily due to ATG5-ATG7-independent alternative autophagy.

MAPK1 and MAPK14 signaling pathways are required for mitophagy

Previously, Mao et al. and our research group reported that the MAPKs Hog1 (homolog of human MAPK14) and Slt2 (homolog of human MAP kinases) are required for mitophagy in Saccharomyces cerevisiae.^40-42 Therefore, we speculated that MAPKs may be related to starvation- or hypoxia-induced mitophagy in mammalian cells. To investigate this possibility, we knocked down several MAPKs, such as MAPK3/ERK1, MAPK1, MAPK14, MAPK8/JNK, and MAPK7/ERK5, and observed starvation-induced mitophagy using mito-Keima cells. Although knockdown of MAPK3, MAPK8, or MAPK7 did not affect mitophagy, knockdown of MAPK1 or MAPK14 efficiently inhibited mitophagy during starvation (Fig. 6A and B). Similarly, knockdown of MAPK1 and MAPK14 efficiently inhibited mitophagy during hypoxia (Fig. 6C and D). We then investigated whether MAPKs are required for macroautophagy using cyto-Keima cells. All of the MAPKs, including MAPK1 and MAPK14, did not inhibit or only marginally inhibited macroautophagy during starvation (Fig. S18A and B). To confirm the importance of MAPK1 and MAPK14 for mitophagy, we suppressed MAPK1 or MAPK14 activity using inhibitors, and observed mitophagy and macroautophagy. When cells were treated with U0126, which suppresses MAPK1 activity by inhibiting the upstream MAP2K1/MEK1 and MAP2K2/MEK2 activities, mitophagy, but not macroautophagy, was efficiently suppressed (Fig. 6E and F; Fig. S19A and B). Similarly, when cells were treated with the MAPK14-specific inhibitor SB203580, mitophagy, but not macroautophagy, was efficiently suppressed (Fig. 6G and H; Fig. S19C and D). Based on these findings, we conclude that MAPK1 and MAPK14 are required for mitophagy, but not for macroautophagy.

We next decided to study whether upstream factors of the MAPK1 and MAPK14 signaling pathways are related to mitophagy. MAP2K1 and MAP2K2 are the MAPK kinases (MAPKKs) and RAF1 is the MAPKK kinase (MAPKKK) of MAPK1. Similarly, MAP2K3/MKK3 and MAP2K6/MKK6 are the MAPKKs and MAP3K11/MLK3 is the MAPKKK of MAPK14. We knocked down these MAPK1 or MAPK14 upstream kinases, and observed mitophagy and macroautophagy using mito-Keima and cyto-Keima cells. As shown in Figure 7, mitophagy was severely inhibited in most of the MAPK1 or MAPK14 upstream factor knockdown cells, whereas macroautophagy was barely inhibited in these knockdown cells (Fig. S20 and S21). Consequently, we concluded that the MAPK1 and MAPK14 signaling pathways and their activity are essential for mitophagy.

Because our results suggested that MAPK1 and MAPK14 activity are required for mitophagy, we speculated that MAPK1 and MAPK14 are activated under mitophagy-inducing conditions, such as starvation and hypoxia. To investigate this possibility, we observed MAPK1 and MAPK14 phosphorylation by immunoblotting. As expected, phosphorylated MAPK1 was dramatically increased during starvation or hypoxia, whereas rapamycin slightly increased phosphorylated MAPK1 (Fig. 8A). U0126, a MAP2K1/MAP2K2 inhibitor, inhibited MAPK1 phosphorylation, suggesting that MAPK1 activation by phosphorylation was a result of MAPK1 pathway activation (Fig. 8A). MAPK1 was phosphorylated at 2 h after starvation and this phosphorylation continued up to 24 h (Fig. 8B). However, MAPK1 was not phosphorylated soon after hypoxia and the MAPK1 phosphorylation gradually increased up to 48 h (Fig. 8B). This finding is consistent with the observation that starvation induced mitophagy earlier than hypoxia (Figs. 2A, B, 3A and B). MAPK1 was phosphorylated at 2 h after rapamycin treatment, but this phosphorylation quickly disappeared (Fig. 8B). This is consistent with the finding that rapamycin induced mitophagy less efficiently than starvation (Figs. 2C and 3C). In contrast to MAPK1 phosphorylation, MAPK14 was phosphorylated even before cellular stress and the phosphorylation level did not change during starvation or hypoxia (Fig. 8A and B). We then speculated that MAPK1 and MAPK14 are localized on mitochondria to regulate mitophagy in response to mitophagy-inducing stimuli. To test this possibility, we expressed GFP (control), GFP-MAPK1, or GFP-MAPK14 in HeLa cells and observed their localization before and after starvation. Unexpectedly, we could not find any colocalization of GFP-MAPK1 or GFP-MAPK14 and mitochondrial TOMM20 (Fig. S22), although this may reflect a low level or transient localization.

Other factors that may be related to mitophagy

Several genes have been reported to be required for mitophagy in mammalian cells. FUNDC1 is a mitochondrial outer membrane protein and binds with LC3 to induce mitophagy during hypoxia.⁴³ TECPR1 (tectonin beta-propeller repeat containing 1) interacts with WIPI2, and is required for autophagic degradation of infected bacteria and mitochondria.^22,44 SQSTM1/p62 is required for PINK1-PARK2-related mitophagy.²² BNIP3L/Nix is a mitochondrial outer membrane protein required for mitophagy during differentiation of reticulocytes and interacts with LC3.^16,17 MUL1 is a mitochondrial E3 ubiquitin ligase and induces mitophagy by ubiquitination and degradation of the mitochondrial fusion protein MFN2 in muscle cells.⁴⁵ To test whether these genes are related to starvation- or hypoxia-induced mitophagy, we knocked down the genes by siRNA and observed mitophagy. Knockdown of SQSTM1 or FUNDC1 had an insignificant effect on starvation- or hypoxia-induced mitophagy (Fig. 9A and B). Overexpression of FUNDC1 was reported to induce mitophagy.⁴³ Thus, we tested whether overexpression of FUNDC1 can induce mitophagy in mito-Keima cells, but found that this was not the case (Fig. S23A and B). Conversely, TECPR1 and BNIP3L were required for mitophagy (Fig. 9; Fig. S17). Knockdown of TECPR1 severely suppressed starvation- or hypoxia-induced mitophagy (Fig. 9), while knockdown of BNIP3L severely suppressed starvation-induced mitophagy and partially affected hypoxia-induced mitophagy (Fig. S17). Knockdown of MUL1 only partially affected starvation-induced mitophagy, but did not affect hypoxia-induced mitophagy (Fig. S17).

Materials

Krebs-Henseleit buffer (K3753), 3-MA (M9281), and rapamycin (R8781) were from Sigma-Aldrich. The MAPK inhibitors U0126 (Cell Signaling Technology, 9903) and SB203580 (Synkinase, SYN-1074) were purchased from the cited companies. Deferiprone was from WAKO (324-65151). Atg5 KO MEFs were provided by RIKEN BRC through the National Bio-Resource Project of the MEXT, Japan. Ulk1 KO MEFs were a kind gift from Dr. Sharon Tooze (London Research Institute).

Keima-expressing cells and cell-culture conditions

To obtain a cytosolic Keima expression vector (pCyto-Keima) or mitochondria-targeting Keima expression vector (pMito-Keima), monomeric Keima-red or mitochondria-targeted mKeima (MBL, AM-V0251) were subcloned into the pTRE2hyg vector. The HeLa Tet-off cell line was obtained from Clontech (631156). We transfected HeLa Tet-off cells with the above-mentioned plasmids and grew the cells in the presence of G418 (400 μg/ml; Affymetrix, 11379), hygromycin B (200 μg/ml; WAKO, 085-06153), and doxycycline (1 μg/ml; ICN, 195044). Single-cell colonies were collected and selected for cyto- or mito-Keima-expressing cells under the control of the Tet-off promoter. Cells were cultured in Dulbecco's modified Eagle's medium (Sigma-Aldrich, D6429) with 10% fetal bovine serum (Clontech, 631106) (complete medium) at 37°C in a 5% CO₂/95% air incubator in the absence of doxycycline for several days to express cyto- or mito-Keima, and then doxycycline (1 μg/ml) was added at 24 h before macroautophagy or mitophagy induction to exclude the possibility that de novo synthesized mito-Keima (i.e., before mitochondrial localization) was degraded by macroautophagy, but not mitophagy. To induce macroautophagy or mitophagy, cyto- or mito-Keima-expressing cells were shifted to Krebs-Henseleit buffer supplemented with doxycycline (1 μg/ml) for 4 or 18 h, respectively. To induce hypoxia, cells were cultured at 37°C in a hypoxic (1% O₂, 5% CO₂) incubator (APM-30D; Astec Corporation). The cells were examined using a 60× or 100× oil immersion objective on a fluorescence microscope (BZ-9000; Keyence). Keima in neutral conditions was observed using excitation of 430 ± 20 nm and emission of 624 ± 20 nm. Keima in acidic conditions was observed using excitation of 562 ± 20 nm and emission of 624 ± 20 nm. Quantification of the number of acidic puncta was performed using ImageJ software (National Institutes of Health).

siRNA knockdown

HeLa Tet-off Keima-expressing cells were transfected with Oligofectamine (Invitrogen, 12252-011), in the presence of 10 nM or 20 nM (only for BNIP3L, MUL1, or UBA1 knockdown) of siRNA specific for each of the genes. The oligonucleotide sequences used for siRNA interference with each gene are shown in Table S1. A control siRNA was constructed with a similar insert containing a 21-nucleotide sequence that had no homology to any known human gene sequence. Knockdown efficiency was confirmed by real-time PCR and western blotting.

Antibodies

An antiserum against GST-LC3 was raised in rabbits, and the anti-GST-LC3 antibody was affinity-purified on an immobilized LC3-CNBr-activated Sepharose column, as previously reported.⁵⁴ Mouse monoclonal antibodies against phospho-MAPK3/MAPK1 (phospho-ERK1/2; Cell Signaling Technology, 4370), phospho-MAP2K1/MAP2K2 (phospho-MEK1/2; Cell Signaling Technology, 9154), MAPK3/MAPK1 (ERK1/2; Cell Signaling Technology, 4695), TIMM23 (BD Biosciences, 611222), SQSTM1 (BD Biosciences, 610832), TOMM20 (Santa Cruz Biotechnology, sc-17764), and ACTB/β-actin (Sigma-Aldrich, A5441) were used.

Plasmids

cDNAs encoding FUNDC1, MAPK1, and MAPK14 were amplified by PCR. For overexpression of FUNDC1 with GFP-NLS, an NLS sequence (GPKKKRKV) was inserted into pIRES2-AcGFP1 (Clontech, 632435) at the C terminus of AcGFP1 by PCR, and the PCR fragment of FUNDC1 was then cloned into this vector via the BglII-EcoRI site. The PCR fragments of MAPK1 and MAPK14 were cloned into pEGFP-C1 (Clontech, 6084-1) via the XhoI-HindIII and XhoI-BamHI sites, respectively.

Inhibitor treatment

U0126 and SB203580 were used at a concentration of 10 μM and 5 μM for 18 h, respectively, in complete Dulbecco's modified Eagle's medium or Krebs-Henseleit buffer at 37 °C. Rapamycin and 3-MA were used at a concentration of 0.2 μM and 10 mM in complete medium for 24 h, respectively.

Indirect immunofluorescence

HeLa cells cultured with complete medium or Krebs-Henseleit buffer (starvation) for 6 or 18 h were fixed with 4% paraformaldehyde for 30 min, and permeabilized with 50 μg/ml digitonin (WAKO, 048-02124) for 5 min. After quenching with 50 mM NH₄Cl in PBS (137 mM NaCl, 2.7 mM KCl, 8.1 mM Na₂HPO₄/12H₂O, and 1.47 mM KH₂PO₄), the cells were blocked with 1% BSA (Sigma-Aldrich, A2153)/PBS, incubated with primary antibodies against TOMM20 (1:300 dilution; Santa Cruz Biotechnology, FL-145) and LC3, and then incubated with appropriate secondary antibodies (goat anti-mouse Alexa 488; Molecular Probes, A-11001; donkey anti-rabbit Alexa 594; Molecular Probes, A-11058). Finally, the cells were mounted in Mowiol (Merck Millipore, 475904). The cells were examined using a 60× oil immersion objective on a fluorescence microscope (IX73; Olympus) or a confocal laser microscope (A1Rsi; Nikon) and assessed by MetaMorph software (Molecular Devices) for deconvolution.

mRNA quantification

Total RNA was extracted from each of the siRNA-transfected cell lines using RNeasy (Qiagen, 74104) according to the manufacturer's instructions. Trace contamination of DNA was removed by DNase digestion (Promega, M6101). The cDNA was synthesized from 1 μg of total RNA using a SuperScript III First-Strand Synthesis System (Invitrogen, 18080051). Real-time PCR was performed in a thermal cycler (StepOne plus; Applied Biosystems) using SYBR premix Ex Taq (Takara, RR420S). The primer sequences used are listed in Table S2.