Please see Supplementary Material.

Adult male C57BL/6 mice (18–22 g) and Sprague-Dawley rats (250–300 g) were purchased from Jiangsu University Laboratory Animals Center, Zhenjiang, China. All animal procedures used in this study were approved by the Ethics Committee for Animal Experimentation of Jiangsu University (UJS-20160012, in Supplementary Material). The drug treatment protocols of GA, CsA, and U0126 in this experiment were based on previous literature (Maddahi and Edvinsson, 2008; Cho et al., 2013; Sun et al., 2014).

Mitochondrial CRC was assessed fluorimetrically in the presence of the fluorescent Ca²⁺ indicator Calcium Green 5N by the QuantaMaster & TimeMaster Spectrofluorometer (Photon Technology International, QuantaMaster^TM40, USA) at excitation/emission maxima of 505/535 nm. Mitochondria (0.5 mg) were diluted in CRC buffer (120 mM KCl, 10 mM NaCl, 1 mM KH₂PO₄, 10 mM MOPS, 20 mM HEPES-Tris, pH 7.2) containing 150 nM Calcium Green 5N. Mitochondria were pulse-loaded with CaCl₂ (10 μM each). Compounds were added to the mitochondrial suspension 5 min before CaCl₂ exposure. The ratio between the amounts of calcium required to trigger MPTP in the presence of the compound (CRC_i) with respect to that required to induce MPTP in the absence of the compound (CRC_o) was used for statistical analysis.

Mitochondrial swelling assay buffer (120 mmol/L KCl, 10 mmol/L Tris-HCl, 20 mmol/L MOPS, 5 mmol/L KH₂PO₄, pH 7.4) containing mitochondrial solution at 1 mg/ml was placed in each well of a 96-well plate at 25°C. The swelling was measured with a spectrophotometer (Molecular Device, Spectra Max 190, USA) as a decrease in light scattering at 540 nm for 10 min. Compounds were added to the mitochondrial suspension 5 min before CaCl₂ or menadione exposure. The difference (ΔA₅₄₀) between the maximum A₅₄₀ value (immediately after CaCl₂ or menadione addition) and the minimum A₅₄₀ value (10 min after CaCl₂ or menadione addition) were used for statistical analysis.

After pre-treatment with compounds for 5 min and 100 μM CaCl₂ exposured to 10 min, the isolated liver mitochondria were lysed in PBS containing 1% TritonX-100 and protease inhibitors cocktail. The supernatant was precleared with Protein A/G beads for 2 h at 4°C, and 500 mg of precleared protein extracts were incubated with 1 μg of anti-CypD antibody or normal rabbit IgG overnight at 4°C. The precipitated complexes were washed in lysis buffer and bound proteins analyzed by immunoblots.

Samples were prepared to isolation of mitochondria and cytosolic protein by a mitochondria isolation kit according to the manufacturer’s guidelines. Equal amounts of proteins were resolved by SDS/polyacrylamide gel electrophoresis (PAGE) using 10–15% acrylamide, transferred onto polyvinylidene difluoride (PVDF) membranes, and blotted using the following commercially available antibodies: anti-CypD (1:500), anti-ANT (1:300), anti-VDAC (1:400), anti-Cyto C (1:300), anti-Caspase-3 (1:500), anti-Caspase-8 (1:300), anti-Cleaved-Caspase-9 (1:400), anti-ERK1/2 (1:500), and anti-p-ERK1/2 (1:500), anti-β-actin (1:4000), anti-COX IV (1:500) followed by incubation with the corresponding horseradish peroxidase-labeled secondary antibody (1:2000). Protein bands were visualized and analyzed using the Chemifluorescence Imaging Systems (GE, USA).

Transient cerebral ischemia was induced by MCAO according to the method of Longa et al. (1989). The rats were anesthetized with chloral hydrate (350 mg/kg intraperitoneal injection). To alleviate pain, animals received 0.05 mg/kg subcutaneous buprenorphine immediately after the reperfusion anesthesia. The level of rCBF was achieved using a laser-Doppler flowmeter (Moor Instruments, UK). Rats had a rCBF in 20% of the pre-ischemic baseline value 10 min after ischemia and to the 80% of baseline at 30 min after reperfusion were used in this study (Mao et al., 2015). Physiological parameters such as the mean arterial blood pressure and heart rate were all monitored by Power Lab System (AD Instruments, Australia). The rats in the sham operation group underwent vessel exposure without MCAO. Body temperature was regulated at 37.0 ± 0.5°C with a heating pad and lamp when necessary. The details of the exclusion criteria, the level of rCBF, and physiological parameters were shown in the Supplementary Material.

2,3,5-Triphenyl-tetrazolium chloride staining was carried out using methods previously published by Leker et al. (2003). Brain Infarct volumes were calculated according to the following formula: (contralateral volume-undamaged ipsilateral)/contralateral volume (%).

Rat brains were removed, fixed in 4% paraformaldehyde in phosphate-buffered saline for 72 h at room temperature, dehydrated through a graded ethanol series, and embedded in paraffin. Sections of 4–5 μm thick were cut in the coronal plane and stained with H & E staining, which was performed as previously described (Monaghan et al., 2016).

Immunostaining methods were used as described previously (Fernandez-Godino et al., 2016). The following primary antibodies were used: anti-Cyto C (1:150), anti-Cleaved-Caspase-9 (1:100), and anti-NeuN (1:200). Following incubation with the primary antibody at 4°C overnight, sections were incubated for 90 min with Alexa Fluor 594-and Alexa Fluor 488-conjugated secondary antibodies (1:200) in blocking solution at room temperature. Finally, sections were incubated with DAPI for 10 min to detect nuclei. All the slices were visualized by fluorescent inverted microscope (Nikon, Ti-E Live Cell Imaging System, Japan).

Rats brains were removed, dissected, and the ipsilateral penumbra areas post-fixed in the same fixative for 2 h before transferring to 0.12 M PBS. Specimens were immersed in 1% OsO₄ in cacodylate buffer, dehydrated in ethanol and embedded in epoxy resin. Ultrathin sections (60 nm) were obtained with an ultra microtome (Leica EM UC7) and then were stained with uranyl acetate and lead citrate for examination with a transmission electron microscope (TEM) at 80 kV (HITACHI, H-7650, Japan).

Human neuroblastoma SH-SY5Y cells were cultured in 90% modified Eagle’s medium-F12 medium and supplemented with 10% fetal bovine serum. All cultures were maintained in 10 cm tissue culture dishes to approximately 90% confluence in a humidified atmosphere of 5% CO₂-95% air at 37°C. The SH-SY5Y cells were treated with H₂O₂ at concentration of 500 μM for 4 h to establish the MPTP opening. Various concentrations of GA (0.1, 1, 10 μM), EGF (10, 100 ng), and CsA (1 μM) were added into the cultures 24 h before the onset of H₂O₂-inducing MPTP opening. Cells were incubated with U0126 (1, 10 μM) in culture medium for 2 h before GA or EGF treatment, which was used to inhibit ERK phosphorylation as recommended by the manufacturer.

Clustered regularly interspaced short palindromic repeats-CypD plasmid synthesized by Santa Cruz Biotechnology was transfected into SH-SY5Y cells using the UltraCruz Transfection Reagent. Samples were collected at 24 and 48 h, and the transfection efficiency was examined by Western bolt. According to the manufacturer’s instructions and our experiment data, the cells were used following 48 h transfection. Control cells were treated control CRISPR activation plasmid in an identical way.

Cell viability was measured by MTT assay (Ganot et al., 2013). The optical density was spectrophotometrically measured at 570 nm using a microplate reader (Molecular Device, Spectra Max 190, USA) with DMSO as the blank. Cell apoptosis was determined by the Annexin V-FITC/PI apoptosis detection kit. A total of 20,000 cells of each sample were analyzed by bivariate flow cytometry system (Becton-Dickinson, FACSCanto II, USA).

To obtain direct evidence of GA effect on MPTP, the Calcein AM-CoCl₂ assay was performed as previously described (Bonora et al., 2016). The results were visualized using fluorescent inverted microscope (Nikon, Ti-E Live Cell Imaging System, Japan). Alternatively, fluorescence intensity was monitored using flow cytometry (Becton-Dickinson, FACSCanto II, USA) at excitation/emission maxima of 490/515 nm.

A cationic dye, JC-1, was used to monitor the MMP. Briefly, cells were treated JC-1 10 μM for 30 min incubation at 37°C. The results were visualized using fluorescent inverted microscope (Nikon, Ti-E Live Cell Imaging System, Japan). Alternatively, fluorescence intensity was monitored using flow cytometry (Becton-Dickinson, FACSCanto II, USA) at excitation/emission maxima of 585/590 nm (J-aggregates) and 514/529 nm (monomer).

Except neurological deficit scores, all values are expressed as the mean ± SD, and the statistical significance of difference between groups was determined by a two-way analysis of variance (ANOVA) test. Neurological deficit scores were expressed as the median, and the statistical significance of the difference between groups was determined by a non-parametric Mann Whitney test. A P-value < 0.05 was considered to indicate statistical significance.

Matrix Ca²⁺ is the single most important factor for inducing MPTP (Baumgartner et al., 2009; Fulda et al., 2010). Here, mitochondrial CRC was assessed fluorimetrically in the presence of the fluorescent Ca²⁺ indicator Calcium Green 5N. As shown in Figure 1, control mitochondria buffered 30 μM CaCl₂, and their buffering capacity was significantly increased after pre-incubation with 1 μM CsA, consistent with the importance of MPTP in this process (Figures 1A,C and Supplementary Table 1; n = 10, P < 0.01). As expected, mitochondria that pre-incubation with 10 μM GA buffered up to 60 μM CaCl₂, strongly increased the level of CRC as observed in control mitochondria (Figures 1A,B and Supplementary Table 1; n = 10, P < 0.01).

We further assessed MPTP sensitivity to the oxidant menadione, which acts as a potent MPTP inducer through promoting oxidative stress (Criddle et al., 2006). The ΔA₅₄₀ value of the mitochondria exposed to menadione at concentrations of 50–200 μM for 10 min increased to 0.1566 ± 0.01, 0.1975 ± 0.02, and 0.2203 ± 0.01 of the control value, respectively (Figure 1D; n = 10). Menadione caused marked mitochondria swelling, whereas CsA pre-incubated mitochondria were significantly less sensitive (Figure 1E; n = 10, P < 0.01). Importantly, GA prevent mitochondrial swelling induced by stimuli independent of calcium overload (Figure 1E; n = 10, P < 0.05), suggesting that it act as a genuine inhibitor of MPTP and not by affecting the calcium homeostasis.

Next, we are interested in the MPTP activity of GA in cerebral ischemia insult, which is well known to favor activation of MPTP in vivo (Sun et al., 2012; Sanderson et al., 2013). The preliminary experiment is conducted to determine the time limited term under MCAO condition (Supplementary Figure 1; n = 4, P < 0.05). No neurological deficits were observed in the sham treated rats, whereas rats subjected to MCAO were scored 2, 3, or 4 in the neurological assessment (Supplementary Figure 5A; n = 20, P < 0.01). The neurological deficits score in GA 50 mg/kg or CsA 10 mg/kg group was significantly reduced (Supplementary Figure 5A; n = 20, P < 0.05). By immunostained with NeuN, the numbers of pyramidal neurons in the hippocampus region were diminished dramatically after reperfusion, whereas the loss of hippocampal neurons was attenuated by GA (Figures 2F-a,b). The expression of cleaved-caspase-3, 8 and number of cleaved-caspase-9-positive cells (% of DAPI+ cells) were dramatically increased in the MCAO group compared with sham group, suggesting that both intrinsic and extrinsic apoptosis were trigged in cerebral ischemia (Figures 2A,C–E-a5,b5,H; n = 4, P < 0.01). Compared with the MCAO group, pre-treatment with GA had unequally sensitive in expression of cleaved-caspase-8 (Figures 2A,C; n = 4, P > 0.05) and the number of cleaved-caspase-9-positive cells (% of DAPI+ cells) (Figures 2E-b5,c5,H; n = 4, P < 0.01), suggesting that GA alleviates neuronal apoptosis dependent in the mitochondrial apoptotic pathways.

Moreover, the mitochondrial protective effect of GA was further measured by the TEM and TTC assay. Notably, while the outer and inner membranes of mitochondria in sham group were clearly distinguishable, the mitochondria showed robustly swelling, vacuolization, disruption and loss of cristae following 48 h reperfusion (Figures 2E-a3,b3). This brought considerable relief in mitochondrial ultrastructure by pretreated with GA or CsA, which the cristae were slightly disrupted (Figures 2E-c3,d3), and the infarct volume was significantly reduced in GA pre-treatment (Figures 2E-b1,c1,G; n = 8, P < 0.01) further confirmed that GA effectively protects brain mitochondria against ischemia/reperfusion injury.

Since our in vivo data indicated MPTP is a key target of GA in pathologic neuron apoptosis, we next focus on validate this idea in vitro. Compared to the control group, H₂O₂ group was significantly decreased the cell viability (% of control) at 500 μM (Supplementary Figure 2; n = 5, P < 0.01), and the number of apoptotic and necrotic cells were significantly increased, in accordance with the morphological observations (Figures 3A-a1,a2,b1,b2,C,D; n = 5, P < 0.01). Notably, H₂O₂ elicited a rapid and almost complete decrease in the fluorescent density of mitochondrial calcein, a highly selective indicator of sustained MPTP opening in situ (Vaseva et al., 2012; Chen et al., 2015), while it was significantly reversed in the CsA pre-treated cells confirming loss of signal resulted from H₂O₂-induced opening of MPTP in our system (Figures 4A-a2–a4,b2–b4,e2–e4,B; n = 4, P < 0.01). This observation was in striking contrast to GA pre-treated cells, which considerably less SH-SY5Y cells mortality or apoptosis and retained more levels of calcein fluorescence than H₂O₂ treated cells in a concentration-dependent manner (Figures 3A-b1,b2,e1,e2,B,C, 4A-b2,b4,c2,c4,d2,d4,B; n = 4–5, P < 0.01).

As MMP could immediate dissipated by MPTP activation (Fulda et al., 2010), the level of the MMP by JC 1 assay were also examined in this work. Compared to the control group, the red/green fluorescence intensity ratio in H₂O₂ treatment SH-SY5Y cells was significantly decreased (Figures 4A-a5–a7,b5–b7,C; n = 4, P < 0.01), while GA or CsA effectively blocked this event, consistent with the results of Calcein-CoCl₂ assay (Figures 4A-b5–b7,d5–d7,e5–e7,C; n = 4, P < 0.01). These data further confirmed that GA could suppress oxidative stress-induced MPTP and cell death.

As a second line of evidence for the inhibition of GA on MPTP, the expression of mitochondrial apoptotic signaling molecules, including the initiator Cyto C, mediator caspase-9, and effector caspase-3 were measured by Western Blot. As shown in Figure 3E, the release of Cyto C, the expression of cleaved-caspase-9 and -3 were observed with a 4-, 9-, and 7-fold increase, respectively, when compared to the control cells (Figures 3E,F,G,4A-a1,b1; n = 4, P < 0.01). Notably, compared with H₂O₂ treated cells, GA significantly reduced the expression of Cyto C, cleaved-caspase-9 and -3 confirming that GA inhibits H₂O₂-induced cells death by MPTP inhibition (Figures 3E,F,G, 4A-b1,d1; n = 4, P < 0.05).

Of note, CypD is well known bind to mitochondrial ANT, thus promoting MPTP opening (Eliseev et al., 2009; Fulda et al., 2010). Here, the physical interaction between CypD and ANT-1 was markedly enhanced by Ca²⁺ incubation, and prevented by CsA, consistent with these reports mentioned above (Figure 1G). GA abrogated Ca²⁺-induced CypD binding to ANT-1, indicating that the beneficial effects of GA are mediated, at least in part, by decreasing in CypD associated with ANT-1, subsequently blocking MPTP (Figure 1G). Moreover, the results of molecular docking between GA and CypD showed that GA was able to form electrostatic forces with ASN 102, PHE 113, MET 61, ARG 55 of CypD (Figure 1F-c). Notably, there is a hydrogen bond between GA and the residue ASN 102 of CypD (Distance: 2.039, Estimated free energy of binding: -0.568) (Figure 1F-d), in addition the binding site was correlated with previously reported binding site of CypD where CsA was bound (Kajitani et al., 2008).

Interestingly, mitochondria isolated from mouse liver in the pre-treatment with GA once a day for 6 days (Figure 1H) were desensitized to the permeability transition, and the level of mitochondrial swelling triggered by Ca²⁺ or menadione was significantly decreased compared with the control mitochondria (Figures 1I,J; n = 10, P < 0.01). Importantly, it seemed more valence compared with mitochondria pre-treated GA 10 μM mentioned above, indicating that there may be other mechanisms at work. Western blotting data showed that not VDAC but CypD expression was significantly suppressed in liver mitochondria (Figure 1L; n = 6, P < 0.05), suggesting that CypD is a potential target of GA in regulating the sensitivity of MPTP in the isolated mitochondria. Moreover, no significant difference in the release of Cyto C was observed in all samples, confirmed that the mitochondrial membrane remains well (Figure 1K; n = 6, P > 0.05).

We reproduce this phenomenon in the cultured SH-SY5Y cells. As shown in Figure 4D, oxidative stress mediated by H₂O₂ caused greatly increase the expression level of CypD (Figure 4D; n = 4, P < 0.01). Compared with H₂O₂ treated cells, SH-SY5Y cells exposed to 0.1 or 10 μM of GA for 24 h displayed significantly decreased levels of CypD expression (Figure 4D; n = 4, P < 0.01). Importantly, it was worth noting that the decrease in brain mitochondrial CypD expression is indicative of the ability of GA to cross the blood-brain barrier, and suggesting that CypD is responsible for therapeutic benefit by GA via effects on MPTP inhibiting (Figures 2A,B; n = 4, P < 0.01).

To further evaluate whether CypD is essential for the sensitivity of GA in MPTP, cells were transfected with CPISPR-CypD to overexpress CypD. Incubation of SH-SY5Y cells with a CPISPR-CypD resulted in 120% increasing in CypD protein levels after 48 h transfection (Figure 4E; n = 4, P < 0.01). As shown in Figure 4F, SH-SY5Y cells overexpress CypD tended no significant loss of calcein fluorescence than non-transduced cells treated by 500 μM of H₂O₂ (Figures 4F-b1,b2,G; n = 4, P > 0.05). Importantly, the calcein fluorescence signal in the GA pretreated cells was significantly higher than H₂O₂ treated cells (Figures 4F-b1,c1,G; n = 4, P < 0.01), however, there was no significant charged in CypD overexpression SH-SY5Y cells following GA pretreated (Figures 4F-b2,c2,G; n = 4, P > 0.05). These results confirmed that the inhibition of H₂O₂-induced MPTP by GA is dependent on CypD.

Of note, ERK is involved in mitochondrial dysfunction during both apoptotic and autophagy cell death-inducing pathways (Monick et al., 2008; Martinez-Lopez et al., 2013). As showed in Figure 5, EGF activated ERK in a concentration dependent manner, and it exerted the best effect at a concentration of 100 ng for 1 h (Figures 5A,B; n = 4, P < 0.05). As expected, the level of phosphorylation ERK (p-ERK) were significantly inhibited by pre-treatment with U0126, a specificity inhibitor of ERK phosphorylation, at a concentration of 10 μM (Figures 5A,B; n = 4, P < 0.01). In comparison with control cells, U0126 strongly increased the expression of CypD following 24 h of exposure, while EGF has the exactly opposite effect to that of CypD expression (Figures 5A,C; n = 4, P < 0.05). Importantly, U0126 reversed the EGF-induced CypD overexpression, suggesting that the CypD expression in SH-SY5Y cells, at least in part, depends on ERK phosphorylation (Figures 5A–C; n = 4, P < 0.05).

Additionally, to explore whether CypD also regulated the phosphorylation of ERK, SH-SY5Y cells were transfected CPISPR-CypD to overexpress CypD. Incubation of SH-SY5Y cells with a CPISPR-CypD resulted in 105% increasing in CypD protein levels after 48 h transfection (Figures 5G,H; n = 4, P < 0.01). Compared with the control cells, CypD overexpression had no effect on the level of p-ERK, suggesting that CypD was a downstream target protein of ERK (Figures 5G,I; n = 4, P > 0.05).

As a second line of evidence for the signaling of ERK-CypD, the effect of ERK in CypD expression during cerebral ischemia/reperfusion injury were studied in vivo. Of note, robust induction of p-ERK was expressed in the ischemic penumbra at the early stage of reperfusion, which involved in neuronal loss through caspase-3-mediated apoptosis (Li et al., 2002; Liang et al., 2014). In comparison with sham rats, the level of p-ERK were significantly increased at 24 h, strongly declined at 48 h, and recovered at 72 h in the ischemic penumbral region following cerebral ischemia/reperfusion injury (Figures 5D,E; n = 4, P < 0.01). Similarly, corresponds to overexpression of p-ERK, the expression of CypD were significantly decreased at 24 h, while robustly increased at 48 h after cerebral ischemia/reperfusion injury (Figures 5D,F; n = 4, P < 0.05). Importantly, compared with sham group, the cleaved-caspase-3 were dramatically increased at 48 h following cerebral ischemia/reperfusion injury (Supplementary Figure 1; n = 4, P < 0.05), further indicating that CypD-ERK axis involved in neuronal apoptosis at the early stage of reperfusion.

Since our data identified that phosphorylation of ERK is an early signaling event before the expression of CypD, whether ERK signaling participate in GA-induced CypD expression was determine. In SH-SY5Y cells, the phosphorylation of ERK is elevated in cells exposed to GA for 1 h when compared with control group (Figures 5K,L, 6A,D; n = 4, P < 0.05). Notably, the GA-induced downregulation of CypD was abrogated by U0126, suggesting that the effect of GA on CypD, at least in part, results in an enhancement of ERK phosphorylation (Figures 5K–M, 6A,E; n = 4, P < 0.01). The neurological deficits score in GA 50 mg/kg or U0126 30 mg/kg group was significantly reduced (Supplementary Figure 5B; n = 12, P > 0.05). Moreover, while U0126 significantly reduced the percent of the infarct volume and the expression of cleaved-caspase-3 following MCAO insult in vivo (Figures 6H–J,L; n = 4–8, P < 0.05), no significant difference in the level of calcein fluorescence between U0126 and H₂O₂ group in vitro (Figures 6B-c,d,e,C; n = 4, P > 0.05).

While GA was able to desensitize the MPTP in a CypD-dependent manner, whether this protection effect was related to ERK phosphorylation was investigated. As shown in Figure 6, combination treatment with U0126 abrogated GA-induced CypD downregulation, MPTP inhibition, and promoted caspase activation in vitro (Figures 6A,B-c,d,f,C,F; n = 4, P > 0.05) and in vivo (Figures 6H–L; n = 4, P > 0.05), further confirmed the idea that ERK mediates GA-induced cytoprotective effect via CypD downregulation.