EGCG protects cardiomyocytes against hypoxia-reperfusion injury through inhibition of OMA1 activation

Mitochondria are the main energy-producing organelles in mammalian cells. In addition to their roles in oxidative metabolism, mitochondria play crucial roles in the pathophysiological processes of cardiomyocytes (Lemieux and Hoppel, 2009). It has been well established that mitochondria are closely associated with the progression of cardiomyopathy in response to various stresses, such as ischemia-reperfusion injury (IRI), hyperglycemia and pressure overload (Biala and Kirshenbaum, 2014; Neuzil et al., 2007).

Mitochondria are dynamic organelles that continuously fluctuate between fission and fusion events, which is referred to as mitochondrial dynamics (Nan et al., 2017). Recently, studies have suggested that mitochondrial dynamics are crucial for the regulation of mitochondrial metabolism in order to meet adequate metabolic demands of the heart and, eventually, protect cardiomyocytes from apoptotic stimuli (Nan et al., 2017; Wai and Langer, 2016). Abnormalities in mitochondrial dynamics contribute to the pathogenesis of the heart (Dorn, 2015).

Mitochondrial dynamics are mainly operated by mitochondrial fission and fusion proteins, i.e. mitochondrial dynamics-related proteins most of which are dynamin-like proteins that contain a GTPase domain (Dorn, 2015). These proteins are located in the outer mitochondrial membrane (OMM), intermembrane space (IMS) or inner mitochondrial membrane (IMM) (Cogliati et al., 2013). Dynamin-related protein1 (Drp1) is mainly located in the cytosol and translocates to the OMM to initiate mitochondrial fission in response to stress (Ong et al., 2010).The OMM-anchored proteins mitofusin 1 and 2 (Mfn1 and Mfn2, respectively) tether the OMM together to mediate mitochondrial fusion (Ranieri et al., 2013). Optic atrophy 1 (OPA1) can be divided into an IMM-anchored long form (L-OPA1) and an IMS-distributed short form (S-OPA1). OPA1 is the key protein to induce IMM fusion and to maintain cristae structure, which is correlated to mitochondrial function (Cipolat et al., 2006). The IMM consists of two sub-compartments, i.e. the boundary membrane and the cristae, and the connection between these two sub-compartments form tubular structures that are small in diameter, known as mitochondrial crista junctions (Zick et al., 2009). Cristae are studded with proteins, including respiratory chain supercomplexes (RCSs) and ATP synthase (Mannella, 2006), which enable mitochondria to synthesize sufficient ATP to meet the energy demands of the heart. In addition to OPA1 allowing fusion, the oligomers formed by L-OPA1 and S-OPA1 tighten crista junctions, which stabilizes RCSs (to increase the efficiency of ATP production) (Cipolat et al., 2006; Cogliati et al., 2013; Frezza et al., 2006) and prevents cytochrome c release (to inhibit the initiation of apoptosis). OMA1 cleaves L-OPA1 at its S1 site to produce redundant S-OPA1 in response to stress. This breaks down the optimal ratio between L-OPA1 and S-OPA1 and leads to deterioration of cardiac function and promotion of cardiac injury (Varanita et al., 2015; Yu et al., 2016). Since excessive cleavage of OPA1 by OMA1 is closely associated with the destruction of cristae structure and mitochondrial malfunction, OMA1 inhibitors need to be found to treat patients with heart disease.

In mammalian cells, OMA1 is first translated into a precursor form that is proteolytically cleaved upon its import into mitochondria, yielding mature (M-OMA1) and lacking the N-terminal amino acid sequence 1–139 (Baker et al., 2014). Previous studies have indicated that OMA1 processing, i.e. self-cleavage of M-OMA1 and accumulation of a short form (S-OMA1) accounts for its enhanced activity towards OPA1 cleavage in response to adverse stimuli (Baker et al., 2014; Zhang et al., 2014), producing redundant S-OPA1 to interfere with cristae structure and to promote cytochrome c release. Therefore, chemical compounds that inhibit OMA1 processing need to be found and their therapeutic effects tested. OMA1 has a role in both physiological and pathological situations, and fine-tunes mitochondrial bioenergetic function and respiratory supercomplex stability (Bohovych et al., 2015). Loss of this mitochondrial protease alters the cleavage pattern of OPA1, and causes obesity and defective thermogenesis in mice (Quirós et al., 2012).

In our study, 2295 plant-derived compounds were used to dock with active sites in mature OMA1 by using Molecular Operating Environment (MOE) software. According to the ranking score resulting from the docking simulation and drug-likeness analysis, epigallocatechin gallate (EGCG) and betanin were potential compounds to inhibit self-cleavage of OMA1. In vitro studies were also conducted to validate the inhibitory efficiency of EGCG and betanin on OMA1. We confirmed that EGCG but not betanin inhibited self-cleavage of OMA1 to reduce the proteolytic process of OPA1 upon hypoxia-reperfusion injury (HRI) and subjection to H₂O₂, eventually alleviating the deterioration of cristae structure and ensuing cytochrome c release to protect cardiomyocytes from apoptosis.

Lack of OMA1 protects the kidney against IRI, suggesting that OMA1 contributes to IRI (Xiao et al., 2014). Induction of mitochondrial depolarization, by subjecting MEFs to H₂O₂ (Baker et al., 2014) or HeLa cells to the inhibitor of oxidative phosphorylation CCCP (Zhang et al., 2014), promotes self-cleavage of OMA1 (OMA1 activation) and subsequently, induces OPA1 cleavage and morphological abnormality of mitochondria (Baker et al., 2014; Zhang et al., 2014). In cardiomyocytes, IRI is associated with H₂O₂ production (Slezak et al., 1995) and mitochondrial depolarization (Sack, 2006). We speculate that IRI induces OMA1 activation and ensues OPA1 cleavage in cardiomyocytes. To simulate IRI in vivo, we subjected MEFs in vitro to HRI in our study. To verify that self-cleavage of OMA1 induces OPA1 cleavage to initiate apoptosis in MEFs upon HRI, we knocked down endogenous OMA1 in MEFs and ectopically expressed wild-type OMA1, as well as OMA1 (E324Q) – in which Glu324 had been replaced with Gln in order to maintain its non-activated status. We observed that HRI can induce the accumulation S-OMA1 (Fig. 1A) and S-OPA1 (Fig. 1B, bands c and e), whereas M-OMA1 (Fig. 1A) and L-OPA1 (Fig. 1B, bands a,b) disappeared. By contrast, in OMA1 (E324Q)-expressing MEFs, HRI failed to induce the self-cleavage of OMA1 (Fig. 1A); moreover, OPA1 was also preserved (Fig. 1B, bands a,b). These data indicate that HRI accounted for self-cleavage dependent activation of OMA1, which subsequently induced the cleavage of OPA1 protein bands a and b into protein bands c and e. A previous study has proven that OPA1 is crucial for the formation of crista junctions and the inhibition of cytochrome c release under stress conditions (Cipolat et al., 2006). In our current study, cytochrome c release was significantly decreased in OMA1 (E324Q)-expressing MEFs compared with OMA1-expressing MEFs upon HRI (Fig. 1C), suggesting that OMA1-activation-dependent cleavage of OPA1 promotes cytochrome c release. Accordingly, active-caspase3 were significantly decreased in OMA1 (E324Q)-expressing MEFs when compared with OMA1-expressing MEFs upon HRI (Fig. 1D). Taken together, the results of these experiments strongly supported our hypothesis that self-cleavage of OMA1 correlates with OPA1 cleavage to initiate apoptosis in MEFs upon HRI.

In mitochondria, OMA1 exists as a mature form that lacks 139 N-terminal amino acids (Fig. 2A) (Baker et al., 2014). In response to oxidative stress, activity of OMA1 is significantly enhanced by OMA1 processing (self-cleavage of M-OMA1 and accumulation of S-OMA1), which subsequently induces OPA1 cleavage and, eventually, leads to mitochondrial fragmentation, cytochrome c release and apoptosis (Zhang et al., 2014). Therefore, to protect cells from oxidative stress, the candidate inhibitors should be found to block the self-cleavage of M-OMA1. To screen for potential plant-based inhibitors of mature OMA1, we used our own small-molecule database of plant-derived inhibitors. MOE software was employed to screen for docking between mature OMA1 and its inhibitors, which we then validated by using cell models, such as MEFs and NMCMs. Assisted by the research group of Yang Zhang (University of Michigan, Medical School, Ann Arbour, MI), we used amino acids 140–521 of OMA1 to construct the 3-dimensional (3D) structure of mature OMA1. Simultaneously, after uploading the compound information to MOE, a compound library was established to ensuing docking in the MOE system. As discussed in the Materials and Methods, we confirmed active sites within the structure of mature OMA1 that should be targeted by the compounds. After the docking simulations between the 2295 plant compounds and active sites in mature OMA1, eight candidate compounds, including EGCG, betanin, malvidin, cineroside, ginkgetin, chicoric acid, carnosic acid and baicalin, were selected according the results (i.e. high or low scores) of the second rescoring stage (Table S1). Lower scores indicate more forms of interaction that are favored in binding. As an OMA1 inhibitor, the compound selected should be able to inhibit protease and metabolic enzymes. Therefore, we predicted the drug-likeness for these eight compounds. The Molinspiration software tool (http://www.molinspiration.com/) was used to find out more about the molecular properties and bioactivities of the compounds. The Molinspiration tool calculated the drug-likeness score towards GPCR ligands, ion channel modulators, kinase inhibitors, nuclear receptor ligands, protease inhibitors and enzyme inhibitors. Higher scores indicate higher bioactivity probability of a particular compound (Table S2). Scores of being both a protease inhibitor and an enzyme inhibitor, were higher for EGCG and betanin than for other compounds, indicating that EGCG and betanin are candidate inhibitors of OMA1. When evaluating cineroside, chicoric acid, carnosic acid and baicalin, the scores for them being enzyme inhibitors were high; however, scores for them being protease inhibitors were low. These results appear to rule out that these compounds inhibit OMA1 activation.

The intermolecular interaction of EGCG and betanin upon binding to M-OMA1 was analyzed by using MOE. EGCG was predicted to interact with M-OMA1 through hydrogen bonds via the amino acid residues Leu172, Lys177, Glu178, Thr305, Asn309, Phe480, Gln508 and Thr512 (Fig. 2B), whereas betanin was predicted to interact with Lys226, Ser233 and Val295 (Fig. 2D). The spatial interaction between the compounds and residues in M-OMA1 were also analyzed in a 3D model (Fig. 2C, E). Atoms in EGCG and betanin, such as oxygen and hydrogen (indicated in the 2D models in Fig. 2B,D, and 3D models in Fig. 2C, E), act as either hydrogen-bond acceptors or hydrogen-bond donors of corresponding residues in M-OMA1. Values (in Å) for the in-between distances are given in red in Fig. 2B,D. In addition, values indicating the strength of each hydrogen bond is provided in the 3D models in purple (Fig. 2C,E). OMA1 contains a conserved M48 metallopeptidase domain (according to the PROSITE database) and depends on this domain to exert its proteolytic activity. The M48 metallopeptidase domain is, according to the primary structure of OMA1, located in the middle of OMA1. As shown in Fig. 1B,C, residues within the M48 metallopeptidase domain of OMA1 – including Asp287, Ser288, Pro289, Ala290, Thr305, Gly306 and Asn309 – form hydrophobic interaction with atoms in EGCG, while Ile276, Glu278, Phe294, Val295, Leu296, Pro297, Asn298, Glu324, Ala333 and Ala334 form hydrophobic interactions with atoms in betanin. This suggests that EGCG and betanin interfere with the proteolytic activity of M-OMA1 by interacting with the M48 metallopeptidase domain.

We knocked down endogenous OMA1 and ectopically overexpressed FLAG-tagged OMA1. Silencing efficiency of OMA1 was validated earlier in experiments that involved MEFs transfected with negative-control shRNA (NC shRNA) and OMA1 shRNA (Fig. 3A). In our current study, MEFs injury caused by stimulation with H₂O₂ was used to mimic reperfusion injury. Being consistent with a previous study (Mao et al., 2014), we found that stimulation with H₂O₂ promoted OMA1 processing (self-cleavage of M-OMA1 and accumulation of S-OMA1). When stimulation with H₂O₂ was prolonged, protein levels of M-OMA1 decreased in a time-dependent manner, whereas S-OMA1 protein levels increased accordingly (Fig. 3B), suggesting that M-OMA1 was cleaved to yield S-OMA over time. Four hours after stimulation with H₂O₂, M-OMA1 was barely detectable and S-OMA1 had reached stable protein levels (Fig. 3B).

To verify that ectopically expressed OMA1 reflects the true physiology of OMA1, and also that accumulation of S-OMA1 is not an artifact of the ectopically expressed system, we detected the autocatalytic process of endogenous OMA1 in MEFs. We found that S-OMA1 accumulated in MEFs upon stimulation with H₂O₂ for 4 h (Fig. 3D), suggesting that the autocatalytic process of endogenous OMA1 is consistent with ectopically expressed OMA1. Levels of S-OMA1 were easily detected, whereas those of M-OMA1 could not be detected at all. To check whether the S-OMA1 form had been stabilized, we measured levels of S-OMA1 in response to stimulation with H₂O₂ after 4 h. Considering that the cell viability was not further decreased at 6 h in response to stimulation with H₂O₂ when compared with cell viability at 4 h (Fig. 3C), we employed western blotting to detect S-OMA1 at 6 h and found that S-OMA1 was still detectable, whereas M-OMA1 could hardly been detected (Fig. 3D). Taken together, these data demonstrate that S-OMA1 accumulated and stabilized in MEFs under conditions of stress.

EGCG and betanin were analyzed by MTT assay to determine their toxic effects on MEFs. EGCG, at concentrations between 0 and 20 μM, did not show any toxic effects on MEFs (Fig. 4A). The safety of betanin was also determined after administrating MEFs with increasing concentrations of betanin (from 0–40 μM) followed by MTT detection (Fig. 4B). These results supported that EGCG and betanin can be employed to treat cells without cytotoxicity. Previous studies have indicated that OMA1 processing (self-cleavage of M-OMA1 and accumulation of S-OMA1) accounts for its enhanced activity following adverse stimuli (Baker et al., 2014; Zhang et al., 2014). Therefore, we examined whether EGCG and betanin interfere with OMA1 processing. We evaluated OMA1 processing in H₂O₂-stimulated MEFs by western blotting with an antibody against OMA1. The OMA1 (E324Q)-expressing group was used as negative control. In response to stimulation with H₂O₂, OMA1 processing was diminished in MEFs pre-treated with EGCG but not betanin (Fig. 4C), suggesting that EGCG has the potential to be an OMA1 inhibitor. As a substrate of OMA1, OPA1 should be regulated in MEFs pre-treated with EGCG. Indeed, EGCG attenuates OPA1 cleavage upon stimulation with H₂O₂ (Fig. 4D). Of note, excessive cleavage of OPA1 leads to mitochondrial fission and ultimately mitochondrial fragmentation. Therefore, it was justifiable to hypothesize that EGCG-attenuated OPA1 cleavage can maintain the mitochondrial network upon stimulation with H₂O₂. After incubating with placebo or EGCG for 3 h in culture medium containing Mito Tracker Red (a chemical dye to stain mitochondria), MEFs were stimulated with H₂O₂ or not for 4 h, followed by mitochondrial network examination using immunofluorescence and confocal microscopy. Three major traits of mitochondrial morphology (tubular, intermediate and fragmented) were evaluated to describe mitochondrial dynamics in differently treated groups (Fig. 4E). Almost 90% of MEFs exhibited an intermediate mitochondrial morphology in the placebo- and EGCG-incubated groups even if not stimulated with H₂O₂, which made it hard to detect the existence of tubular and fragmented mitochondria (Fig. 4F), and suggesting that EGCG is unable to affect mitochondrial dynamics in non-stressed MEFs. Notably, MEFs that underwent stimulation with H₂O₂ showed mitochondrial fission, featuring 88% fragmented mitochondria in MEFs (Fig. 4F). With 40% intermediate and 60% fragmented mitochondria (Fig. 4F), incubation with EGCG can partially counteract the pro-fission effect of H₂O₂ on mitochondrial dynamics, which may be due to its inhibitory role in OMA1 activation and OPA1 cleavage. To assess the protective efficacy of EGCG, MTT assay was employed. It showed that cell viability of MEFs was significantly decreased following stimulation with H₂O₂ for 4 h, as compared with that of the untreated group (P<0.01). EGCG significantly attenuated this H₂O₂-induced decrease in cell viability (P<0.01) (Fig. 4G).

To fully evaluate the inhibitory capacity of EGCG on OMA1 processing, different pre-treatments were tested (Fig. 5A). We observed that, when MEFs were pre-treated with 30 μm EGCG for 2 h, EGCG reached its maximum inhibitory capacity – with OMA1 processing being completely inhibited (Fig. 5A) – indicating that EGCG is a potent OMA1 inhibitor. Accordingly, cleavage of OPA1 (i.e. cleavage of bands a and b into bands c and e, Fig. 5B), was completely inhibited when MFEs were pre-treated with 30 μm EGCG for 2 h (Fig. 5B), indicating that EGCG can affect OPA1 cleavage through inhibition of OMA1 processing. To investigate whether EGCG directly inhibits OMA1, we knocked down endogenous OMA1 and ectopically expressed wild-type OMA1 and mutant OMA1 (OMA1-L172A/K177A), i.e. L172 and K177 in OMA1 were replaced by A to abolish the interaction between EGCG and OMA1. EGCG inhibited the conversion of M-OMA1 into S-OMA1 within wild-type OMA1-expressing MEFs, which is not observed in OMA1 (L172A/K177A)-expressing MEFs (Fig. 5C). Accordingly, EGCG was unable to attenuate OPA1 cleavage in MEFs expressing OMA1 (L172A/K177A) (Fig. 5D). These data demonstrate that EGCG directly inhibits the catalytic activity of OMA1.

The data from our experiments using MEFs indicate that EGCG inhibits activation of OMA1, which subsequently attenuates OPA1 cleavage. To explore the potential therapeutic use of EGCG in cardiac disease, neonatal mouse cardiomyocytes (NMCMs) challenged with HRI were used to mimic IRI. As shown in Fig. 6A, EGCG did not interfere with the physiological ratio between L-OPA1 (bands a and b) and S-OPA1 (bands c–e) when cells were kept under normoxic conditions. L-OPA1 (bands a and b) was cleaved to its short form (bands c and e) (Fig. 6A) during HRI. NMCMs pre-treated with EGCG show attenuated OPA1 cleavage during HRI (Fig. 6A). L-OPA1 cleavage by OMA1 does not produce band d. OPA1 has been implicated in mitochondrial dynamics, which correlate with improved mitochondrial function (Cogliati et al., 2013). As expected, when compared with the placebo-treated group, the EGCG-treated group exhibited much-improved ATP production (P<0.05) and mitochondrial potential (P<0.01) during HRI (Fig. 6B, C).

Mitochondrial dynamics have been implicated in mitochondrial function, including producing ATP and maintaining mitochondrial potential (Wai and Langer, 2016). OPA1 is one of the core proteins regulating mitochondrial dynamics (Nan et al., 2017). To determine whether EGCG regulates mitochondrial dynamics specifically via OMA1/OPA1 signaling during HRI, protein levels and sub-cellular distribution of other mitochondrial dynamics related proteins (including Mfn1, Mfn2 and Drp1) were evaluated by western blotting. In NMCMs pre-treated with placebo and EGCG, protein expressions of Mfn1 were decreased to the same extent during HRI (Fig. 6D). Total protein levels of Mfn2 and Drp1 did not change significantly in each group (Fig. 6D). When evaluating the sub-cellular distribution of Drp1, we found that, during HRI, translocation of Drp1 from cytosol to mitochondria was increased, and this process cannot be affected when adding EGCG (Fig. 6E). This detection of mitochondrial dynamics-related proteins demonstrated that EGCG solely depends OMA1/OPA1 signaling to regulate mitochondrial dynamics and mitochondrial function.

In addition to OPA1, the mitochondrial contact site and cristae organizing system (MICOS) is also central in the maintenance of the cristae structure and it is located in the mitochondrial inner membrane (Kozjak-Pavlovic, 2017). MICOS is also crucial for the regulation of mitochondrial function (Genin et al., 2018). Four core components of MICOS, including the MICOS complex subunits MIC60, MIC25 and MIC19 (officially known as IMMT, CHCHD6 and CHCHD3, respectively), and the sorting and assembly machinery component 50 homolog (Samm50), were evaluated by western blotting (Fig. 6F). The total protein levels regarding the core components of MICOS did not change significantly in each group (Fig. 6F). From the perspective of protein expression, EGCG does not affect assembly and function of MICOS and solely depends on OMA1/OPA1 signaling to regulate cristae structure.

The correct ratio between L-OPA1 and S-OPA1 is a prerequisite to tighten crista junctions. As shown in Fig. 1B, OPA1 was cleaved from long form to short form during HRI. And EGCG was able to inhibit OMA1 and attenuate OPA1 cleavage (Fig. 1A,B). Therefore, we inferred that EGCG is able to keep a relatively normal cristae structure to resist HRI challenge in NMCMs. To verify this hypothesis, NMCMs were pre-treated with either placebo or EGCG and cultured under normoxic conditions or HRI. Afterwards, NMCMs were harvested to detect cellular apoptosis and the structure of cristae. As shown in Fig. 7A, NMCMs cultured under normoxic conditions showed tightly arrayed cristae structure. Following reperfusion injury, destruction of cristae and vacuole formation were observed in mitochondria. EGCG incubation noticeably improved cristae structure, with fewer and smaller vacuoles in mitochondria (Fig. 7A). Intact and tightened cristae can hold cytochrome c in IMS and prevent cytochrome c release in response to stress. EGCG may alleviate cytochrome c release through inhibition of OMA1/OPA1 signaling. As shown in Fig. 7B, following reperfusion injury, cytochrome c levels were increased in the cytosol and decreased in mitochondria, suggesting that cytochrome c is released from mitochondria to cytosol in both placeboes and EGCG pre-treated NMCMs (Fig. 7B). However, when compared with the placebo pre-treated group, cytochrome c release was significantly alleviated in EGCG pre-treated group during HRI (P<0.01). The release of cytochrome c can, ultimately, cause the activation of caspase-3. We observed that protein levels of active caspase-3 (Fig. 7C) and TUNEL-positive nuclei (Fig. 7D) (P<0.05) were correlated with the release of cytochrome c, suggesting that EGCG can significantly attenuate activation of caspase-3 and apoptosis. Furthermore, the morphometric analysis of mitochondria and cristae were evaluated in the in vivo study. A total of 12, 8-week-old male mice, including six mice fed with EGCG and six mice fed with placebo were used. Each group of mice fed with EGCG or placebo was randomly assigned to either the sham (n=3) or HRI group (n=3), and mitochondrial morphology and cristae structure were established by using transmission electron microscopy (TEM). TEM showed that sham-operated mice in both placebo- and EGCG-fed groups displayed round and rectangular mitochondria, i.e. mitochondria near the cell membrane are mainly round, wheras those between muscle filaments are mainly rectangular – displaying the mitochondrial morphology of normal mouse cardiomyocytes. (Fig. 7E). Compared with sham-operated mice, HRI mice in the placebo-fed group displayed significantly increased swelling of mitochondria (established by measuring the circularity index and the mitochondrial area) and significantly increased vacuole formation (established by measuring the loss of cristae) (Fig. 7E) (P<0.05). These changes, including swelling of mitochondria and vacuole formation, were significantly attenuated in the EGCG-fed group (P<0.05).

OMA1 processing (self-cleavage of M-OMA1 and accumulation of S-OMA1) accounts for its enhanced activity towards OPA1 cleavage following adverse stimuli (Baker et al., 2014; Zhang et al., 2014). Therefore, we simulated docking between mature OMA1 and 2295 plant-derived compounds by using the MOE system, searching for potential inhibitors that interfere with the autocatalytic process of M-OMA1. Based on the s-score values, eight of these compounds were found to have a higher possibility to interact with active sites on M-OMA1. On the basis of the drug-likeness score of these eight compounds, only EGCG and betanin held the promise to be inhibitors for protease and enzymes. Thereafter, in vitro studies were conducted to validate the inhibitory efficacy of EGCG and betanin against the autocatalytic process of M-OMA1. MEFs and NMCMs were subjected to H₂O₂ and/or HRI, and these two cell models were used to mimic oxidative stress during cardiac IRI in humans. When MEFs were subjected to H₂O₂ and HRI, M-OMA1 underwent self-cleavage and S-OMA1 accumulated, and this process was correlated OPA1 cleavage (L-OPA1 was cleaved into S-OPA1). OMA1 processing was potently blocked by administration of EGCG and accompanied with attenuated OPA1 cleavage. Moreover, EGCG significantly alleviated mitochondrial dysfunction, cytochrome c release and initiation of apoptosis when NMCMs were subjected to HRI.

Reperfusion injury, including HRI and IRI, is accompanied with over-production of oxygen species (oxidative stress), leading to deterioration of homeostasis and cellular functions (Tompkins et al., 2005). Molecular events underlying oxidative stress following reperfusion injury are complex and involve loss of mitochondrial potential and ATP production, which ultimately leads to mitochondrial dysfunction and cellular apoptosis (Di Lisa and Bernardi, 2006). In our study, we observed OMA1 processing as well as OPA1 cleavage, both of which play a key role in inducing mitochondrial dysfunction and cellular apoptosis during reperfusion injury. This study demonstrated that oxidative stress depends on OMA1/OPA1 signaling to deteriorate homeostasis and function of cells following reperfusion injury.

OMA1 depends on the M48 metallopeptidase domain to exert its proteolytic activity (Baker et al., 2014). The M48 metallopeptidase domain of OMA1 cleaves L-OPA1 at the S1 site to produce S-OPA1 (Baker et al., 2014). According to the data provided by MOE, we found that EGCG was able to interact with the active site within the M48 metallopeptidase domain. Since OMA1 processing accounts for its enhanced activity towards OPA1 cleavage (Zhang et al., 2014), we infer that EGCG interferes with OPA1 cleavage by using two mechanisms: (1) EGCG interacts with the active site within the M48 metallopeptidase domain to block OMA1 processing; (2) the M48 metallopeptidase domain of OMA1 interacts with and cleaves OPA1 at the S1 site, which is competitively blocked by EGCG.

EGCG, the major catechin accounting for 59% of the total catechins in green tea (Singh et al., 2011; Steinmann et al., 2013), was demonstrated to be a compound protecting mitochondrial function in response to pathological stimuli (Adikesavan et al., 2013; Devika and Stanely Mainzen Prince, 2008a, b). TEM has been used to show that mitochondria in cardiomyocytes of rats that had been treated with isoproterenol, exhibit disarrangement and smaller size of cristae; and this was attenuated when rats were pre-treated with EGCG (Devika and Stanely Mainzen Prince, 2008a). Rats exposed to cigarette smoke exhibited highly abnormal architecture of mitochondria, swelling of mitochondria, increased areas of vacuolation indicative of edema and disruption of mitochondrial cristae, whereas EGCG administrated rats that had been subjected to EGCG and cigarette smoke showed mild distortion in mitochondrial structure without swelling and normal cristae (Adikesavan et al., 2013). However, the specific underlying mechanism that relate to abnormal mitochondrial structure and cristae damage was not elucidated. In addition, EGCG has been proved to be an effective inhibitor of matrix metalloproteinase-2 (MMP2) (Deb et al., 2016; El Bedoui et al., 2005; Zhen et al., 2006) and MMP9 (Lee et al., 2014; Sarkar et al., 2016), implying that EGCG serves as a metalloproteinase inhibitor. OMA1 is also a metalloproteinase and belongs to the peptidase M48 subset (Baker et al., 2014). In this study, we proved that EGCG inhibits the metalloproteinase OMA1. The data obtained in our study support the previously reported notion that EGCG plays a crucial role in the maintenance of cardiac mitochondrial structure and function; moreover, the underlying mechanism of this notion was elucidated.

In conclusion, an in silico drug design strategy revealed that EGCG can be an inhibitor of OMA1. OMA1 processing accounts for its enhanced activity towards OPA1 cleavage following reperfusion injury, which was inhibited by administration of EGCG. Inhibition of OMA1/OPA1 signaling is a potential approach to attenuate reperfusion injury. When cardiomyocytes were challenged with HRI, EGCG maintained mitochondrial cristae structure and function through inhibition of OMA1/OPA1 signaling, resulting in reduced cytochrome c release and decreased cellular apoptosis. Taken together, our data support the notion that mitochondrial integrity is the key for EGCG-induced cardiac protection, thereby providing support for EGCG as a possible therapeutic agent intended to prevent and/or limit IRI in the heart.

The following antibodies were used in this study: anti-OPA1 (dilution 1:2000), anti-FLAG-tag (dilution 1:1000), anti-Mfn1 (dilution 1:1000), anti-Tom20 (dilution 1:1000) and anti-active caspase3 (dilution 1:500) were from Abcam (Cambridge, MA); anti-HSP60 (dilution 1:1000), anti-VDAC (dilution 1:1000), anti-Cytochrome c (dilution 1:1000), anti-Drp1 (dilution 1:1000) and anti-Mfn2 (dilution 1:1000) were from Cell Signaling Technology (Danvers, MA); anti-tubulin and anti-OMA1 (dilution 1:2500 and 1:1000, respectively), were from Santa Cruz Biotechnology (Shanghai, China); anti-β-actin (dilution 1:2500) was from KANGCHEN (Shanghai, China). Tetramethylrhodamine methyl ester (TMRM; Sigma-Aldrich, St Louis, MO) and MitoTracker Deep Red (Invitrogen, Carlsbad, CA) were used to stain the mitochondria. EGCG (Sigma-Aldrich) and betanin (Sigma-Aldrich) were used to inhibit the enzymatic activity of OMA1. Terminal deoxynucleotidyl transferase dUTP nick-end labeling (TUNEL) kit (Roche, Mannheim, Germany) and 4,6-diamidino-2-phenylindole (DAPI) (Sigma-Aldrich) were used to determine the apoptosis. 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) was purchased from Sigma-Aldrich Corporation to measure the cell viability.

Neonatal mouse cardiomyocytes (NMCMs) were isolated from the hearts of neonatal C57BL6 mice using collagenase II (0.05% [w/v] (Invitrogen) and trypsin (0.05% [w/v], Genom, China) digestion. The enzymatic digestion was stopped by adding newborn calf serum. NMCMs were purified by differential adhesion to plastic on 24- or 6-well plates (Corning, Tewksbury, MA).

NMCMs, as well as rat myoblast (H9C2) cells and MEFs (because flow cytometry was more accurate in detecting H9C2 cells than NMCMs) were maintained in Dulbecco's modified Eagle’s medium (DMEM; Corning, Manassas, VA) supplemented with 10% (v/v) fetal bovine serum (FBS; Life Technologies, Paisley, UK) and penicillin/streptomycin. All cells were cultured in a 5% CO₂ and 95% atmosphere at 37°C. For the experiments shown in Figs 1, 3, 4 and 5, MEFs were used. For the experiments shown in Fig. 6C, H9C2 were used. For other experiments shown in Figs 6 and 7, NMCMs were used.”

Eight-week-old C57BL/6 mice (male) were anesthetized by injection of sodium pentobarbital (60 mg/kg, i.p.) followed by tracheal intubation aided by a rodent ventilator. By using a standard surgical approach, ischemia-reperfusion injury (IRI) was induced by 45 min of ischemia followed by 3 h of reperfusion injury. Sham surgical procedures were performed on the control group.

The molecular docking procedures were implemented using Molecular Operating Environment (MOE, Chemical Computing Group, Quebec, Canada) software. The protein structure of mature mouse-OMA1 was obtained from Yang Zhang's Research Group (University of Michigan). 2295 plant-derived compounds were used to dock with OMA1. Prior to starting the docking simulation, mature OMA1 and plant-derived compounds were prepared in which (i) hydrogen atoms were added to the structures with a standard geometry; (ii) structures were minimized using a MMFF94s force-field; (iii) structures were protonated utilizing Protonate3D; (iv) the MOE Alpha Site Finder was used for active site searches within the enzyme structure and dummy atoms were created from the obtained alpha spheres. Molecular docking is used to predict the binding mode of the ligand within the binding sites of mature OMA1. Docking score and ligand interaction between mature OMA1 and corresponding compounds were gained.

Lentiviruses carrying shRNA against OMA1 and recombinant adenoviruses expressing shRNA-resistant OMA1 cDNA were all provided by JiKai Biotechnology (Shanghai, China). To generate OMA1-knockdown and knock-in MEFs, endogenous OMA1 was knocked down by OMA1-shRNA followed by transfection with lentiviruses expressing exogenous shRNA-resistant OMA1. For doing this, two silent nucleotide mutations (the siRNA-resistant nucleotide sequence 5′-GACGTTACTCCCTAGAAACTT-3′ was used to replace the original nucleotide sequence 5′-GACCTTACTGCCTAGAAACTT-3′) were introduced into FLAG-tagged mouse wild-type (WT) OMA1 and FLAG-tagged mutant form of OMA1 for which we generated a glutamic acid (E)-to-glutamine (Q) mutation at residue 324 (E324Q). The viruses were amplified in HEK 293T cells and titrated according to the manufacturer's instruction. Adenoviruses containing empty plasmids (vectors) and lentiviruses containing non-specific shRNAs (NC shRNA, NC-shRNA) served as controls. Cardiomyocytes were infected with purified viruses at multiplicities of infection (MOI) of 50 (for adenoviruses) and 20 (for lentiviruses treated with polybrene at a final concentration of 8 μg/ml) overnight. Each viral suspension was replaced with fresh medium the day after infection and the expression of the indicated proteins was determined by western blotting.

Cell lysis was conducted following corresponding treatments, then, the proteins were extracted in RIPA solution (Beyotime, Shanghai, China) supplemented with a protease inhibitor cocktail (Roche, Base, Switzerland). Mitochondria and cytoplasmic protein fractions were obtained using a Mitochondria and Cytoplasmic Protein Extraction Kit (Beyotime Biotechnology, Shanghai, China), according to the manufacturer's instructions. Protein concentrations were determined by BCA assay (Pierce, Rockford, IL). Equal amounts of protein were resolved by SDS–PAGE and then transferred onto a polyvinylidene fluoride (PVDF) membrane (Millipore, Billerica, MA). The blots were subsequently incubated with the corresponding primary antibodies overnight followed by horseradish peroxidase-conjugated secondary antibodies incubation. Immunoreactivity was visualized using a chemiluminescence ECL Western Blot System (Millipore, Boston, MA).

Cells were first cultured under anoxic conditions (6 h) and kept under normoxic conditions (18 h) to establish an in vitro model of simulated hypoxia and reperfusion injury (HRI). In brief, to simulate cells cultured under ischemic conditions, they were switched from maintenance medium (DMEM) to a buffer containing (in mmol/l) 137 NaCl, 12 KCl, 0.5 MgCl2, 0.9 CaCl2, 4 HEPES, 10 2-deoxyglucose and 20 sodium-lactate (pH 7.2), and incubated at 37°C in a cell incubator flushed with 0.5% O₂ 5% CO₂ and 94.5% N₂. After challenge with simulated ischemia, cells were moved back to DMEM and kept in 5% CO₂ at 37°C to simulate reperfusion injury. Normoxic cells were cultured in DMEM and flushed with 5% CO₂ at 37°C. In Figs 4C–G and Figs 5C,D, MEFs were treated with H₂O₂ for 4h. For other figures, the duration of H₂O₂ treatment is provided in Figures and Figure legends.

Mitochondria were labeled with MitoTracker Red (100 nmol/l). The mitochondrial phenotype of each cell was categorized as tubular, intermediate or fragmented. Mitochondria displaying morphology of puncta and rods were regarded as fragmented mitochondria. Mitochondrial integration into a network was considered as tubular mitochondria. Mitochondria exhibiting both fragmented and tubular morphology were reckoned as intermediate mitochondria. Cells were visually scored and ≥80 cells were quantified per treatment group. Confocal images were obtained with an Olympus IX81 inverted microscope. The percentage of cells with each pattern were evaluated.

Mitochondrial membrane potential (ΔΨ) was measured after loading cardiomyocytes with TMRM (200 nmol/l) for 30 min. Afterwards, cells underwent trypsinization and fluorescence was assessed by flow cytometry (excitation/emission 543/560) with a BD FACSCanto II Flow Cytometer (BD Biosciences, San Jose, CA).

Cellular ATP content was measured using a luciferin/luciferase-based kit (Beyotime Biotechnology, Shanghai, China), according to the manufacturer's instructions.

The cell viability was measured using MTT assays according to manufacturer's protocol. The absorbance was measured at 570 nm by Micro Plate Reader (Molecular Device, SpectraMax 250). For TUNEL analysis, NMCMs were fixed with 4% paraformaldehyde, permeabilized with 0.2% Triton X-100 and then washed twice with PBS. Nucleotide and TdT enzyme mixture was added to the cells or samples for 1 h of incubation at 37°C. The nuclei were stained with DAPI. Then, the TUNEL-positive and total nuclei were observed under a fluorescence microscope (Leica, Wetzlar, Germany).

For Transmission electron microscopy (TEM), NMCMs were collected and fixed in 2% formaldehyde and 2.5% (V/V) glutaraldehyde in 0.1 M Na-cacodylate buffer (pH 7.4) for 2 h at room temperature and then incubated overnight at 4°C. The cells were then washed thrice in 0.1 M PBS for 15 min each and fixed in 1% osmium tetroxide for 1 h. Then, cells were stained, blocked with uranyl acetate, dehydrated with a graded series of ethanol solutions, cleared in propylene oxide and then embedded in epoxy resin. Thereafter, the blocks were trimmed and ultra-thin sections (120 nm) were cut, which were subsequently observed under a transmission electron microscope (H7500 TEM, Hitachi, Tokyo, Japan, http://www.hitachi.com).

All values are presented as the mean±s.e. We evaluated the significance of the differences between groups by Student's t-test or two-way analysis of variance. P<0.05 was considered statistically significant. All statistical analyses were performed using GraphPad Prism® 5.0 (GraphPad Software, Inc., La Jolla, CA).

We thank Yang Zhang (University of Michigan, Medical School, Ann Arbour, MI) for helping us with bioinformatics analysis.