Pharmacological targeting of MCL-1 promotes mitophagy and improves disease pathologies in an Alzheimer’s disease mouse model

There is increasing evidence that inducing neuronal mitophagy can be used as a therapeutic intervention for Alzheimer’s disease. Here, we screen a library of 2024 FDA-approved drugs or drug candidates, revealing UMI-77 as an unexpected mitophagy activator. UMI-77 is an established BH3-mimetic for MCL-1 and was developed to induce apoptosis in cancer cells. We found that at sub-lethal doses, UMI-77 potently induces mitophagy, independent of apoptosis. Our mechanistic studies discovered that MCL-1 is a mitophagy receptor and directly binds to LC3A. Finally, we found that UMI-77 can induce mitophagy in vivo and that it effectively reverses molecular and behavioral phenotypes in the APP/PS1 mouse model of Alzheimer’s disease. Our findings shed light on the mechanisms of mitophagy, reveal that MCL-1 is a mitophagy receptor that can be targeted to induce mitophagy, and identify MCL-1 as a drug target for therapeutic intervention in Alzheimer’s disease.

A lzheimer's disease (AD) is a severe neurodegenerative disorder with memory loss and cognitive dysfunction as its main symptoms. Mitochondrial dysfunction is a fundamental pathological hallmark of AD, as damaged neuronal mitochondria have been found to accumulate both in sporadic and familial types of the disease, as well as in AD animal models 1,2 . The impaired mitochondrial function triggers energetic stress that promotes the disease-defining amyloid-β (Aβ) oligomers and hyper-phosphorylated Tau (pTau) pathologies 1,3 . Impairment of the mitochondrial biogenesis, for example, induced by decreased expression of the mitochondrial biogenesis regulator PGC-1α, contributes to synaptic dysfunction and neuronal degeneration in AD 4,5 . Furthermore, mitochondria regulate cytosolic calcium homeostasis, and mitochondrial dysfunctioninduced intracellular calcium imbalance leads to neuronal death and is implicated in AD 6 . Importantly, mitochondrial dysfunction has been reported to accelerate Aβ production and occur prior to the accumulation of Aβ deposits in the brains of AD mouse models 2,7,8 . Moreover, suppression of mitochondrial function using toxins or genetic deletion of mitochondrial proteins exacerbates Aβ pathology 9,10 . In light of this cumulative evidence, mitochondrial dysfunction has been suggested as a pivotal event in the initiation of AD, and interventions that bolster mitochondrial health may ameliorate the neurodegenerative pathologies associated with it 1,11,12 .
Mitophagy is a selective autophagy pathway for mitochondrial quality control, in which engulfment of damaged or depolarized mitochondria by a double-membrane autophagosomal structure is followed by fusion with lysosomes for degradation 13 . Emerging findings suggest that mitophagy is also compromised in AD, resulting in the accumulation of dysfunctional mitochondria that contributes to synaptic dysfunction and cognitive decline in AD 11,14 . Conversely, mitophagy enhancement reduces Aβ plaques and Tau tangles in human neuronal cells and ameliorates memory impairment in transgenic mouse models of AD 11 .
Here, our small-molecule compound screen of 2024 FDAapproved drugs or drug candidates revealed a BH3-mimetic UMI-77 as a potent mitophagy promoter. We identified the target of UMI-77, a key anti-apoptotic protein MCL-1, as a mitophagy receptor that interacts with LC3A to promote mitophagy. Finally, we found that UMI-77-induced mitophagy significantly improves AD pathologies seen in the APP/PS1 mouse model.

Results
UMI-77 selectively induces mitophagy independent of mitochondrial damage and apoptosis. mt-Keima is a fusion of a mitochondrial signal sequence from cytochrome C oxidase subunit IV with the fluorescent protein Keima. This chimeric protein has been previously established as a robust sensor of mitophagy 20,21 . mt-Keima undergoes an excitation wavelength shift at low pH, allowing quantification of the mitochondria that have been exposed to the acidic milieu of lysosomes and hence, mitophagy. To identify mitophagy activators, we performed a high-throughput screen using a HEK293T cell line stably expressing mt-Keima and a library of 2024 FDA-approved drugs or drug candidates. We identified 20 activators of mitophagy that increase relative mitophagy levels by >1.5-fold (Fig. 1a, Supplementary Fig. 1a). Interestingly, some of the Bcl-2 family inhibitors (also known as BH3-mimetics), such as UMI-77, were found to trigger mitophagy.
UMI-77 is an MCL-1-specific compound that blocks the interaction between MCL-1 and Bax/Bak, thereby allowing Bax/ Bak to induce apoptosis 22 . Unlike CCCP (carbonyl cyanide 3chlorophenylhydrazone), UMI-77 did not induce mitochondrial damage in HEK293T and HeLa cells at the sub-lethal 5 µM dose (Fig. 1b). Compared with CCCP, the UMI-77+CCCP combination treatment significantly augmented mitophagy levels (Fig. 1c). These results indicated that UMI-77 can promote mitophagy independent of mitochondrial damage and can augment mitophagy triggered by CCCP-induced mitochondrial damage.
To eliminate the possibility that apoptosis induction may nonspecifically induce the excitation shift of mt-Keima, we assessed the effect of all apoptosis inducers in our drug library on this established mitophagy reporter. As shown in Supplementary  Fig. 1b, most apoptosis-inducing drugs did not trigger the mt-Keima excitation shift. Moreover, mitophagy induction by UMI-77 could not be rescued by pan-caspase inhibitor Z-VAD-fmk, suggesting that this UMI-77 can induce mitophagy independent of its established role in activating apoptosis (Fig. 1d). In addition, UMI-77 strongly induced mitophagy, but could not induce apoptosis (Supplementary Figs. 1e and 2a-d). These results strongly indicated that the induction of mt-Keima excitation shift by UMI-77 is independent of apoptosis induction and that UMI-77 can induce mitophagy, but not apoptosis at sub-lethal doses.
Consistent with the pro-mitophagy effect of UMI-77, using live-cell imaging of HEK293T-mt-Keima cells, we found that this compound induces co-localization of mitochondria with lysosomes (Fig. 1e). Transmission electron microscopy (TEM) confirmed that mitochondria were accumulated in autophagosomes of HeLa cells (Fig. 1f) and HEK293T cells ( Supplementary  Fig. 1f) following UMI-77 treatment. We found that UMI-77 promoted degradation of mitochondrial proteins (outer-membrane protein Tom20 and inner-membrane protein Tim23), but not that of endoplasmic reticulum marker calnexin or cytosolic marker tubulin in HEK293T, HeLa, SH-SY5Y, and U2OS cells (Fig. 1g, Supplementary Fig. 1g). The UMI-77-induced degradation of mitochondrial proteins could be blocked by the lysosome inhibitors E64D or NH 4 Cl/Leupeptin, but not the proteasomal inhibitor MG-132 (Fig. 1h, Supplementary Fig. 1h). As E64D was able to increase the levels of LC3-II, UMI-77 may increase the autophagic flux ( Supplementary Fig. 1c).
Macroautophagy is known to degrade mitochondria in a nonselective manner; however, we found that p62 levels (macroautophagy marker) and other organelle markers did not decrease following UMI-77 treatment, consistent with the notion that this compound induces mitophagy, but not macroautophagy (Supplementary Fig. 1d, i). Taken together, our findings indicate that the MCL-1-targeting BH3-mimetic UMI-77 specifically induces mitochondrial degradation via mitophagy, independent of macroautophagy, mitochondrial damage, or apoptosis.
MCL-1 promotes mitophagy. As MCL-1 is the de facto target of UMI-77, we investigated whether this anti-apoptotic protein participates in the UMI-77-driven mitophagy activation. As shown in Fig. 2a, knockdown of MCL-1 rescued the UMI-77induced degradation of mitochondrial proteins Tom20 and Tim23 in HEK293T and HeLa cells. In addition, the MCL-1 knockdown prevented the induction of the mt-Keima excitation shift induced by the UMI-77 treatment ( Fig. 2b and Supplementary Fig. 3a). These results indicate that MCL-1 is required for UMI-77-induced mitophagy activation.
Overexpression of mitophagy receptors such as FUNDC1 and FKBP8 has been shown to promote mitophagy 23,24 . To investigate the role of MCL-1 in mitophagy, we generated a HEK293T stable cell line expressing MCL-1 in a doxycyclineinducible manner (HEK293T-MF2). Notably, doxycyclineinduced MCL-1 overexpression resulted in the degradation of mitochondrial markers (Fig. 2d). Moreover, MCL-1 overexpression resulted in the co-localization of mitochondria with lysosomes (Fig. 2c). Interestingly, as shown in Fig. 2c, mitochondria became smaller and fragmented following MCL-1 overexpression, consistent with the notion of mitophagy induction. Finally, electron microscopy confirmed that MCL-1 overexpression drives mitophagy (Fig. 2e). These results are similar to the phenotype of Bcl-2-L-13 overexpression 25 . Taken   MCL-1 is an LC3-interacting mitophagy receptor. We hypothesized that the UMI-77-induced release of MCL-1 from Bax/ Bak allows MCL-1 to interact with LC3, thereby promoting mitophagy. We found that MCL-1 contains three canonical "LC3interacting region" (LIR) motifs [W/Y/F]XX[I/L/V] 18,19 at its Cterminus and that the first two of them (henceforth, LIR 261-264 and LIR 318-321 ) are in the cytosolic region of the protein (Fig. 3a, b). We also found that LIR 261-264 is strongly conserved, suggesting it is functional ( Supplementary Fig. 4). To test whether UMI-77 induces interaction between MCL-1 and LC3, we performed co-immunoprecipitation assays. As shown in Fig. 3c, the interaction between LC3A and MCL-1 was enhanced, whereas the interaction between Bax and MCL-1 was decreased following UMI-77 treatment. Consistent with this, we also found that overexpression of MCL-1-M (L213A/D218A) which was not able to interact with Bax 26 , enhances mitophagy levels (Supplementary Fig. 5a and b). MCL-1 also interacted with other Atg8 family proteins in the presence of UMI-77 (Fig. 3d). To further demonstrate that MCL-1 binds to LC3A in a specific manner, we generated a series of mutations of the MCL-1 LIR 261-264 and LIR 318-321 motifs (W261A, I264A, W261A/I264A, ΔLIR 261-264 , F318A, V321A, and F318A/V321A). The mutations of the LIR 261-264 motif of MCL-1, but not that of LIR 318-321 motif, attenuated MCL-1 interaction with LC3A (Fig. 3e). Importantly, a pull-down assay, using proteins purified from a bacterial expression system, revealed that MCL-1 binds to LC3A directly ( Supplementary Fig. 6). Next, we investigated whether UMI-77 enhances the interaction of endogenous MCL-1 and LC3A. Consistent with the notion that the mechanism of mitophagy induction by UMI-77 is via stimulation of the MCL-1-LC3A interaction, the enhancement of this interaction between endogenous MCL-1 and LC3A was observed in situ on mitochondria by using Duolink® PLA technology, following UMI-77 treatment (Fig. 3f, Supplementary  Fig. 7). This interaction was also observed at basal levels, in the absence of UMI-77 treatment, reflecting the basal endogenous levels of mitophagy and further, suggesting that MCL-1 is a mitophagy receptor (Fig. 3f).
Taken together, these results demonstrate that MCL-1 is a mitophagy receptor, which directly interacts with LC3A through its LIR 261-264 motif and that this interaction is enhanced by UMI-77, leading to enhanced levels of mitophagy. Moreover, the interaction between MCL-1 and LC3A is critical for UMI-77mediated mitophagy activation.
UMI-77 induces mitophagy via the ATG5 pathway, independent of the canonical mitophagy receptor proteins, Bax or Parkin. We investigated whether other mitophagy receptor proteins play a role in the UMI-77-induced mitophagy or induction of the MCL-1-LC3A interaction. UMI-77 significantly enhanced the interaction between MCL-1 and LC3A, both in wild-type HeLa and HeLa cells with a quadruple knockout of mitophagy receptors NDP52, p62, NBR1, and TAX1BP1 ( Fig. 4a, b). This indicated that the induction of mitophagy by UMI-77 was independent of these mitophagy receptor proteins. Consistent with this, we found that the levels of the mitochondrial marker proteins Cox II and Tim23 were decreased by UMI-77 treatment in a time-dependent manner in both wild-type and the quadruple knockout HeLa cells (Fig. 4c). We also found that the previously reported mitophagy receptors (FUNDC1, BNIP3, and NIX) did not participate in UMI-77-induced mitophagy (Supplementary Fig. 8a-d).
MCL-1 is required for mitophagy induced by oxygen-glucose deprivation. We then aimed to understand the physiological function of MCL-1 as a mitophagy receptor. Previous studies showed that oxygen-glucose deprivation (OGD) or OGD/reperfusion damage mitochondria and induce mitochondrial clearance through mitophagy 28 . Our mt-Keima assay shows that MCL-1 is required for OGD-induced mitophagy, as knockdown of MCL-1 blocked the excitation shift ( Fig. 5a and Supplementary Fig. 3d). The specific degradation of the mitochondrial marker proteins (Cox II and Tim23) induced by OGD was also blocked by knockdown of MCL-1 (Fig. 5b).
MCL-1 has been shown to regulate mitochondrial fragmentation, which is required for mitophagy 29,30 . Therefore, we attempted to  Mitochondrial fragmentation was observed in HEK293T cells following OGD, as judged by immunofluorescence microscopy.
Notably, knockdown of MCL-1 rescued these morphological changes, indicating that MCL-1 has a critical role in OGDinduced mitochondrial fragmentation (Fig. 5c). Next, we examined the role of MCL-1 LIR 261-264 motif in the OGD-induced mitochondrial fragmentation and mitophagy. Although overexpression of the LIR 261-264 motif mutants W261A, I264A, and ΔLIR 261-264 had no effect on the OGDinduced mitochondrial fragmentation in cells with a stable MCL-1 knockdown (Fig. 5d), the OGD-induced mitophagy was significantly blocked by these LIR mutations (Fig. 5e). We also found OGD enhanced the interaction between MCL-1 and LC3A and decreased the interaction with Bax (Fig. 5f). This indicated that MCL-1 acts as a receptor for mitophagy activation during OGD and that the LIR 261-264 motif is only involved in mitophagy, but not in the mitochondrial fragmentation role of MCL-1.  UMI-77 ameliorates cognitive decline and amyloid pathologies in the APP/PS1 mouse model of AD. Using mt-Keima transgenic mice, we found that intraperitoneal injection of a 10 mg/kg dose of UMI-77 potently induces mitophagy in vivo, in mouse brain tissues, within 6 hours (Fig. 6a). Next, we examined the effect of UMI-77-induced mitophagy on the disease pathologies and mouse behavioral phenotypes of the APP/PS1 mouse model of AD. Each mouse was injected with either vehicle or UMI-77 (10 mg/kg) every other day, from the age of 4 months, for a total period of 4 months. Using the Morris water maze test, we found that the UMI-77 treatment improved the learning and memory of the APP/PS1 mice (Fig. 6b, c). UMI-77 could effectively reduce the levels of the insoluble Aβ 1-42 in mouse brains (Fig. 6d).
Similarly, immunofluorescence results also showed that the size of extracellular Aβ plaque in the hippocampus was significantly reduced, and the activation of astrocytes (as judged by glial fibrillary acidic protein (GFAP) staining) was also inhibited by the UMI-77 treatment (Fig. 6e). UMI-77 reduced the neuroinflammation levels in the APP/PS1 mice. Inflammatory cytokine levels (TNFα and IL-6) were significantly reduced by the UMI-77 treatment, whereas antiinflammatory cytokine levels (IL-10) were unaffected (Fig. 6f). Finally, As shown in Fig. 6g, UMI-77 significantly restored the mitochondrial morphology in the neurons, consistent with the notion that induction of mitophagy by UMI-77 would result in the clearance of the damaged mitochondria seen in the APP/ PS1 mice.
As our data show that MCL-1 is a mitophagy receptor, next, we attempted to evaluate the effect of MCL-1-induced mitophagy on the behavioral phenotypes of the APP/PS1 mice. Following  AAV-mediated delivery of an MCL-1-expressing vector into the hippocampus of these mice, we found that MCL-1 overexpression ameliorates the cognitive decline seen in the APP/PS1 mice and reduces extracellular Aβ plaque in the hippocampus (Supplementary Fig. 10a-c). Surprisingly, overexpression of MCL-1 also improved the learning and memory of wild-type mice, indicating that MCL-1 has an important role in neurons ( Supplementary  Fig. 10a).
In conclusion, UMI-77 potently induced mitophagy in vivo, significantly restored cognitive deficits of the APP/PS1 mouse model of AD, reduced the inflammatory response, and the pathological effects caused by the Aβ plaques, and promoted clearance of the damaged mitochondria. Moreover, confirming our UMI-77 findings, overexpression of MCL-1 in the hippocampus of the APP/PS1 mice phenocopied these results. Overall, these experiments suggest that UMI-77 is a potent drug lead for the treatment of AD. . e Mice were treated as in b and IHC of whole brains was performed to stain for amyloid-beta (Aβ) plaques (6E10 antibody, green), astrocytes (GFAP antibody, red) and nuclei (DAPI, blue). Scale bar, 1000 μm; insets: Scale bar, 100 μm. f Mice were treated as in b and the levels of the indicated cytokine levels were measured by ELISA using whole brain lysates (mean ± S.E.M.; *p < 0.05, ns, not significant, two-tailed t test). Box plots indicate median (middle line), 25th, 75th percentile (box) and minima and maxima (whiskers). g Electron microscopy images of mice brain hippocampal tissues. Insets (blue boxes) show mitochondria. Scale bars, 5 µm; insets: Scale bars, 2 µm. Source data are provided as a Source Data file.

Discussion
Our study shows that MCL-1, an important anti-apoptotic protein, is a LC3-interacting mitophagy receptor protein that induces mitochondrial fragmentation and mitophagy in response to mitochondrial damage caused by OGD. Our results suggest that MCL-1 mediates mitochondrial fragmentation and mitophagy through distinct molecular mechanisms. Although the mitophagy role of MCL-1 requires its interaction with LC3, the mitochondrial fragmentation role of MCL-1 is independent of this interaction. We postulate that MCL-1 recruits LC3A via its LIR motif to the surface of mitochondria, leading to the formation of nascent mitophagosomes and the elongation of the mitophagosome membrane. The interaction between MCL-1 and GABARAP proteins subsequently mediates the closure of the mitophagosome membrane, thereby engulfing the mitochondria. Consistent with this, our data showed that MCL-1 can interact with both LC3A and GABARAP proteins (Fig. 3c, d). However, we are not able to rule out the possibility that MCL-1 can cooperate with NIX and/ or FUNDC1 for mediating mitophagy. As both NIX and MCL-1 specifically interact with LC3A, there may be a synergistic mechanism between NIX and MCL-1 to promote mitophagy induction 31 .
Previous studies suggest that loss of MCL-1 in adult myocytes results in mitochondrial dysfunction, defective PINK1-PARK2 signaling and impaired initiation of autophagy 32 . MCL-1 is also suggested to function in embryonic development and synapse formation 33,34 , lifespan regulation 35,36 , and diverse cellular processes (including mitochondrial dynamics) 32,36,37 . Our findings thus provide insights into how MCL-1 can act as a mitophagy receptor to couple with the molecular mechanisms of mitophagy and the mechanisms by which MCL-1 regulates normal physiology, aging, and disease.
We also noticed that MCL-1 is reported to act as a suppressor of AMBRA1 to suppress mitophagy, under the conditions of FCCP or CCCP treatment 38 . It is reported that AMBRA1mediated mitophagy is regulated by HUWE1, an E3 ubiquitin ligase, which could control mitochondrial protein ubiquitylation in cooperation with AMBRA1 during the mitophagy process in a PARKIN-free cellular system, in the condition of MMP deprivation 39 . And MCL-1 overexpression is sufficient to inhibit recruitment to mitochondria of HUWE1, upon AMBRA1mediated mitophagy induction 38 . However, PINK1/PARKIN, but not HUWE1/AMBRA1, is the main pathway regulating mitophagy in cells with PARKIN, following MMP deprivation, during which the inhibition of MCL-1 on HUWE1 translocation to mitochondria may not inhibit mitophagy. What's more, these previous findings suggest that MCL-1 may inhibit ubiquitindependent mitophagy through inhibition of E3 ligase translocation to depolarized mitochondria, while our findings suggest that MCL-1-mediated mitophagy may be in a ubiquitinindependent way.
We found that several apoptosis inducers can induce mitophagy to an extent much weaker than UMI-77. Among them, Gambogic acid inhibits most of the Bcl-2 family proteins, including MCL-1, whereas AT101 inhibits Bcl-2, Bcl-xL, and MCL-1, suggesting that the mechanism of mitophagy induction by them could be similar to that of UMI-77. ABT-737 and ABT-263 are BH3-mimetics, similar to UMI-77, but they have not been reported to inhibit MCL-1, suggesting that other Bcl-2 family proteins may participate in the regulation of mitophagy. In addition, ABT-737 has been reported to promote mitophagy via inducing Parkin translocation to mitochondria by inhibiting the interaction between Bcl-2 and Parkin. RITA promotes p53 phosphorylation to induce apoptosis, whereas GDC-0152 inhibits the interaction between IAP and other proteins. Notably, both IAP and p53 have been suggested to be involved in the regulation of mitophagy 27,[40][41][42][43][44] .
Mitophagy induction is a promising strategy for AD therapy 1,11 . CCCP and oligomycin A are two wide-used mitophagy-inducing agents that drive mitophagy by causing mitochondrial damage 45,46 . However, their severe cytotoxicity limits their clinical application. Nicotinamide riboside (NR) is a safe and effective activator of neuronal mitophagy 47 . NR is a precursor of NAD + and can be metabolized to produce NAD + in cells which can rescue the function of mitochondria in xeroderma pigmentosum group A 48 . NR reduces Aβ levels in APP/PS1 mice and has been tested in clinical trials 47,49 . Urolithin A (UA) is a natural, dietary, microflora-derived metabolite, which can induce mitophagy and ameliorate cognitive decline in the APP/PS1 mouse model 11,50,51 .These studies highlight the necessity to identify more safe, effective, and clear mechanisms for mitophagy inducers that can become effective therapeutic approaches for the treatment of AD. Our discovery of the FDA-approved drug candidate MCL-1 BH3-mimetic UMI-77 is an unexpected mitophagy-inducing agent that does not cause mitochondrial damage. Previous studies showed that UMI-77 can induce apoptotic cell death in certain cancer cell lines that express high levels of MCL-1 22,37 . However, using unbiased systematic approaches, we found that UMI-77 can induce mitophagy fully independent of apoptosis and that it can ameliorate the pathologies seen in the APP/PS1 mice.
Risk factors in AD patients, such as genetic mutation, aging, and environmental factors, promote the production of reactive oxygen species (ROS) and lead to mitochondrial dysfunction, as well as the production of Aβ plaques. The abnormal mitochondrial function accelerates the production of ROS and fueling this process. Therefore, degradation of damaged mitochondria by mitophagy can prevent excessive production of ROS, reducing the production of Aβ production 52 . On the other hand, Aβ can be transported into cells and accumulated in mitochondria, and this interaction inhibits mitochondrial function, elevates ROS levels, and alters mitochondrial dynamics [53][54][55] . We report that UMI-77induced mitophagy effectively reduces Aβ plaque levels, which may be achieved through decreased production of ROS and the Aβ deposited in the damaged mitochondria. Our work strongly supports UMI-77 as a putative drug for the treatment of AD.
In summary, we identify MCL-1 as a mitophagy receptor that can be targeted by the FDA-approved drug candidate BH3mimetic UMI-77 to induce mitophagy and promote reversal of the AD pathology. Our findings suggest MCL-1 as a drug target for AD and further confirm that induction of mitophagy is a viable strategy for treating this neurodegenerative disorder. MCL-1 is an important cancer drug target, and several clinical trials are currently underway 56 . Moreover, UMI-77 is an FDA-approved drug candidate for the treatment of pancreatic cancer 22 . In addition, our discovery of the role of MCL-1 as a mitophagy mediator and the ability of MCL-1 inhibitors to induce mitophagy may help to consider this function of the anti-apoptotic protein when designing and evaluating cancer therapies that target MCL-1. High-throughput screening. FDA-approved drug or drug candidates library was purchased from Topscience, Inc. (Shanghai, China). Phosphate-buffered saline (PBS) was used as a positive control to induce mitophagy 57,58 , whereas dimethyl sulfoxide (DMSO) was used as a negative control, since compounds from the screen were dissolved in DMSO. Mitophagy inducers were screened using HEK293T-mt-Keima stable cell line. High-throughput imaging was done using Biotek Cytation® 3 system excitation at 469 nm/586 nm and emission at 620 nm. Mitophagy levels were estimated by the fraction of cells that were fluorescent upon excitation at 586 nm using Gen5™ software. The screen Z factor was >0.5. A compound with excitation shift fluorescence intensity value higher than negative control by 1.5-fold was scored as a mitophagy inducer.
The primers pCDH-mt-Keima-F 5′-CCGGAATTCGAAATGCTGAGCCTG CGCCAGAG-3′ and pCDH-mt-Keima-R 5′-CGCGGATCCTCAACCGAGCAAA GAGTGGC-3′ were used for cloning mt-Keima into the pCDH vector. Lentiviral packaging was done according to previously established methods 59 . Packaging vectors pSPAX2 and pMD2.0 G were co-transfected with pCDH-mt-Keima into HEK293T-mt-Keima cells for 48 h. The cell supernatant was collected and filtered with 0.22-micron filter membrane. Then the virus was concentrated with 4% PEG8000 and 3% NaCl solution overnight. After centrifugation at 5000 rpm, discard the supernatant, add appropriate amount of PBS to resuspend the virus, and store it at -80°C. HEK293T-mt-Keima stable cell lines were generated using lentiviral infection for 48 h and selection with 2 µg/mL puromycin. Clones with high mt-Keima expression levels were isolated by flow cytometry.
shRNA-mediated knockdown. The shRNA sequences were cloned into the pLV3 vector. Packaging of lentiviruses was performed according to previous methods 59 .
Immunoprecipitation. Cells were cultured in 10-cm culture dishes, transfected as described above, washed twice with 5 ml PBS and scraped into 1 ml of pre-chilled RIPA buffer (20 mM Tris-HCl, 150 mM NaCl, 0.5% NP-40, 1 mM NaF, 1 mM Na 3 VO 4 , 1 mM EDTA) plus protease inhibitor cocktail. Lysates were incubated for 30 min at 4°C. After 15,000 × g centrifugation for 10 min at 4°C, the lysates were subjected to immunoprecipitation with anti-Flag or anti-HA agarose beads, overnight at 4°C. Beads were collected and washed three times with 1 ml RIPA buffer. The complexes were eluted with 2× sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) loading buffer for 10 min at 100°C.
Immunoblotting. Samples were heated at 100°C for 10 min, subjected to 10-12% SDS-PAGE, and transferred onto polyvinylidene difluoride membranes for 1 h at 0.2 A in a wet transfer tank submerged into an ice bath. Membranes were blocked in PBS with Tween (PBST) buffer containing 5% (w/v) skimmed milk for 1 h and probed with the indicated antibodies in PBST containing 5% (w/v) BSA at 4°C overnight. Detection was performed using HRP-conjugated secondary antibodies and chemiluminescence reagents (#4 AW001-500, Beijing 4A Biotech Co., Ltd.).
Protein purification and GST-pull down. The truncated MCL-1 (1-325aa) was cloned into the pET28a vector and transformed into E. coli BL21 (DE3). The primers for pET28a-MCL-1 and GST-LC3A were shown in Supplementary  Table 1. Protein expression was induced by 1 mM IPTG addition. Proteins were purified using Ni-NTA columns. GST-LC3A expression was induced using the same method and the protein was purified using glutathione-Sepharose. GST-LC3A was incubated with purified MCL-1 for 2 h at 4°C. GST beads were washed four times with lysis buffer. Proteins were eluted at 100°C for 10 min with 20 µl SDS-PAGE loading buffer and analyzed by western blotting.
Cell viability assay. HEK293T and SH-SY5Y cells were cultured in 96-well plate for 24 h and treated with UMI-77 for 12 h. Cells were stained with LIVE/DEAD™ cell imaging kit (#R37601, Thermo Fisher Scientific, Ltd.) according to operation manual. The images were obtained from Biotek Cytation® 3 system by using 488 nm (indicate live cell) and 586 nm (indicate dead cell) excitation. The cell viability was estimated by ratio of live cells. HEK293T and SH-SY5Y cells were cultured in 96-well plate for 24 h and treated with UMI-77 for 24 h. Caspase-3 activity was determined by using Caspase-Glo® 3/7 Assay System kit (#G8090, Promega). The data were obtained from Biotek Cytation® 3 system.
Proximity ligation assays (PLA). Cells were cultured on glass slides, treated with or without UMI-77, washed twice with PBS, and fixed with 4% paraformaldehyde in PBS for 20 min at 25°C. Following blocking with 5% FBS supplemented with 0.1% Triton X-100 to increase the permeabilization for 1 h, primary antibodies (MCL-1 and LC3A) were incubated with the slides overnight at 4°C. After incubating with the secondary antibodies conjugated with the PLA probes, the signals were amplified through the ligation and amplification steps. The fluorescence analysis was done using Biotek Cytation® 3.
Fluorescence microscopy. Cells were cultured in 12-well plates on circular glass coverslips and washed twice with PBS and fixed with 4% paraformaldehyde in PBS for 30 min at 25°C. Cells were blocked with blocking buffer (5% FBS, 0.1%Triton X-100, PBS) at 25°C for 1 h, and incubated with primary antibodies overnight at 4°C. After washing with PBS, secondary antibodies were incubated at room temperature for 1 h. For mitochondria imaging, cells were stained with Mito-Tracker™ Deep Red FM (#M22426, Thermo Fisher Scientific, Ltd.). For live-cell imaging, cells were cultured on a 3.5 cm glass dish and stained with LysoTracker™ Green DND-26 (#L7526, Thermo Fisher Scientific, Ltd.) and imaged with Zeiss LSM 880 AiryScan or Nikon A1 confocal imaging system. Images were processed with ImageJ software.
Electron microscopy. Cells were imaged at the Center of Cryo-Electronic Microscopy Zhejiang University using Tecnai™ G2 Spirit. Cells were fixed with glutaraldehyde.
Morris water maze. The circular pool containing titanium dioxide water is divided into four quadrants (northwest, northeast, southwest, and southeast). The platform (12 cm) is placed about 1 cm below the Southeast quadrant water. The mice were put into the water with their heads facing the wall of the pool. The mice were put into the pool four times from different quadrants. The time that each mouse spends to find the underwater platform was recorded. If the time is more than 60 s, the mice are guided to the platform and allowed to stay for 20 s. Moreover, the time to find the platform was recorded as 60 s. Each mouse was trained four times per day for a total of four days, and the time to find the platform (escape latency) was recorded. After training for 24 h, the platform was removed, and the space search experiment was started for 60 s. The mice were put into the water from the opposite side of the original platform quadrant, and the number of times the mice passed through the original platform position and the locus of the movement were recorded.
Statistics and reproducibility. P values were computed using two-tailed t test or one-way analysis of variance. Data were analyzed using GraphPad Prism 8. All the western blot, micrographs assay, mitophagy levels and apoptosis assay were carried out at least three independent times with the same results.