Main

Glycogen synthase kinase-3 (GSK3) was originally identified as a phosphorylating kinase for glycogen synthase.1, 2 It has two isoforms, α and β, that possess strong homology in their kinase domains with, however, distinct functions.3 GSK3 is constitutively active but it can be inhibited by phosphorylation on serine 21 (Ser21) for GSK3α and Ser9 for GSK3β.4 In the heart, GSK3β has several important roles in cardiac hypertrophy5 and ischemia–reperfusion (IR) injury.6 Accumulating evidence indicates that phospho-Ser9-GSK3β-mediated cytoprotection is achieved by an increased threshold for permeability transition pore (PTP) opening.6, 7, 8, 9 The mechanism by which GSK3β delays PTP opening still remains unclear. It has been reported that GSK3β could interact with ANT at the inner mitochondrial membrane in the heart9 and/or to phosphorylate voltage-dependent anion channel (VDAC) and cyclophilin D (CypD) in cancer cells.10, 11 GSK3β also has other proposed mechanisms of action, including a poorly characterized role in calcium (Ca2+) homeostasis regulation12 and protein–protein interactions,9 as well as functions in different subcellular fractions such as the nucleus, cytosol and mitochondria.13

Reperfusion is the most powerful intervention to salvage ischemic myocardium. However, it can also paradoxically lead to cardiomyocyte injury and death.14 One of the main actors of this lethal reperfusion injury is cellular Ca2+ overload,15 which results in part from excessive sarco/endoplasmic reticulum (SR/ER) Ca2+ release and Ca2+ influx through the plasma membrane (e.g. through L-type Ca2+channel and NCX (sodium-calcium exchanger)).16 Although ryanodine receptors (RyRs) are the major cardiac SR/ER Ca2+-release channels involved in excitation–contraction coupling (ECC)17 and ischemia–reperfusion (IR) injury,18 recent studies reported an increasing role for inositol 1,4,5-trisphosphate receptors (IP3Rs) Ca2+-release channels in the modulation of ECC and cell death.19, 20 Ca2+-handling proteins of ER and mitochondria are highly concentrated at mitochondria-associated ER membranes (MAMs), providing a direct and proper mitochondrial Ca2+ signaling, including VDAC, Grp75 and IP3R1.20, 21, 22

Here, we provide evidence that, following IR, a fraction of cellular GSK3β is localized at the SR/ER and MAMs. At the MAMs interface, GSK3β can specifically interact and regulate the protein composition of the IP3R Ca2+ channeling complex and modulate Ca2+ transfer between SR/ER and mitochondria. These findings support a novel mechanism of action of GSK3β in cell death process during reperfusion injury.

Results

GSK3β interacts with the IP3R Ca2+ channeling complex at the SR/ER-mitochondria interface in the heart

To characterize the role of GSK3β in the heart, we first determined its subcellular localization. The purity of SR/ER, mitochondrial and MAM fractions from mouse hearts were visualized by both electron microscopy (Supplementary Figure S1) and validated by western blot (WB) (Figure 1a). GSK3β was detected in both SR/ER and mitochondria and more particularly in the MAMs fraction together with IP3R, Grp75, VDAC and CypD. To characterize the implication of GSK3β in MAM fraction, we performed two-dimensional blue native page separation on heart homogenates. Our results showed that although GSK3β was present in varying amounts in complexes, a pool of GSK3β was also present in a higher molecular weight complex encompassing IP3R, Grp75, VDAC and CypD (Figure 1b, arrow), suggesting that GSK3β interacts with MAM-resident proteins to form a macrocomplex in native state in the heart.

Figure 1
figure 1

GSK3β modulates the IP3R-Ca2+ channeling complex at MAMs in the heart. (a) Protein components of subcellular fractions prepared from WT mouse hearts revealed the presence of GSK3β in MAMs. Total, total heart homogenate; Mito, pure mitochondria; MAM, mitochondria-associated ER membranes. IP3R and SERCA2 were used as cardiac SR/ER biomarkers. VDAC, Grp75, COXIV and CypD were used as mitochondrial biomarkers. (b) Two-dimensional blue native protein separation of WT heart homogenates. GSK3β was clearly detected in a high molecular weight complex encompassing IP3R, Grp75, VDAC and CypD (arrow). (c) IP revealed that SB21 (a potent GSK3β inhibitor, 70 μg/kg) reduced the interaction of IP3R with Grp75, GSK3β, VDAC and CypD in mice hearts (one representative out of n=3 is shown). (d) Typical images of in situ interactions between IP3R with Grp75, GSK3β, VDAC and CypD using PLA in adult cardiomyocytes with or without SB21 treatment (6 μM). Quantification of the PLA red fluorescent dots was performed using BlobFinder V.3.2 and nuclei were stained in blue with 4',6-diamidino-2-phenylindole (DAPI). Indicated results are means±S.E.M. of three distinct experiments (each experiment represents the average of 8–10 adult cardiomyocytes). *P<0.05 versus vehicle (Veh)

Full size image

To confirm the involvement of GSK3β in MAM protein complex, we next investigated its relationship with CypD, VDAC, Grp75 and IP3R at the MAM interface.21 Using in situ proximity ligation assay (PLA), we first validated that GSK3β interacted with IP3R, Grp75, VDAC or CypD but not RyR2 and ANT in both adult cardiomyocytes and H9c2 cells (Supplementary Figure S2A). GSK3β inhibitors were used to determine if it might change these protein interactions. As shown in Supplementary Figures S2A and B, both pharmacological and genetic inhibition of GSK3β decreased the interaction of GSK3β with IP3R, Grp75, VDAC or CypD. Interestingly, the inhibition of GSK3β also modified the interaction between the other partner proteins of the complex; indeed, SB216763 (SB21) significantly reduced the co-immunoprecipitation (IP) of IP3R with GSK3β, as well as Grp75, VDAC and CypD in heart homogenates (Figure 1c). In line with this, using PLA approach, adult cardiomyocytes treated with SB21 displayed significantly decreased interaction of GSK3β, Grp75, VDAC and CypD with IP3R (Figure 1d). Altogether, these results suggest that GSK3β modulates the IP3R-Ca2+ channeling complex at the MAM interface in the heart.

Inhibition of GSK3β alters Ca2+ transfer from SR/ER to mitochondria in cardiomyocytes

We next examined the role of GSK3β on the transfer of Ca2+ between SR/ER and mitochondria. Calcium fluxes were monitored in isolated adult mouse cardiomyocytes stained with rhod-2 fluorescent dye. In our model, fluorescent dye showed a specific mitochondrial staining, overlapping with the mitochondrial marker Mitotracker Green (Figure 2a). The quantitative pixel-by-pixel analysis measured with Zen software (Zeiss, Germany) showed strong correlation between rhod-2 and Mitotracker Green with an overlap coefficient averaging 0.89±0.01 (0 corresponds to no overlap and 1 corresponds to a perfect overlap) (Figure 2a). Under basal conditions, 100 μM histamine induced a transient Ca2+ release from SR/ER, which consequently induced mitochondrial Ca2+ uptake in adult cardiomyocytes (Figure 2b), reflecting the IP3R-mediated Ca2+ transfer from SR/ER to mitochondria. This Ca2+ transfer was significantly reduced when GSK3β was inhibited by SB21 averaging 0.60±0.06- versus 1.00±0.07-fold versus control (CTRL); P<0.05 (Figure 2b). Interestingly, the inhibition of GSK3β by SB21 did not modify the caffeine-induced Ca2+ transfer from SR/ER to mitochondria (P=NS, Figure 2c), suggesting that GSK3β has a specific action on IP3R channels with no effect on RyRs in adult cardiomyocytes. As adult mice cardiomyocytes are refractory to transfection using conventional methods and that they can only be maintained in culture for short periods before they dedifferentiate,23 the analysis on adult cardiomyocytes was complemented by a number of assays on H9c2 cells to define the role of GSK3β in the regulation of Ca2+ transfer between SR/ER and mitochondria. Similar to adult cardiomyocytes, rhod-2 showed a specific mitochondrial staining in H9c2 cells with an overlap coefficient averaging 0.87±0.02 when compared with Mitotracker Green (Supplementary Figure S3A). As shown previously, SB21 did not modulate the mitochondrial Ca2+ amplitude induced by caffeine (P=NS, Supplementary Figure S3B), but significantly reduced the Ca2+ transfer under histamine stimulation in H9c2 cells (Supplementary Figures S3C and S5). Although the knockdown of GSK3β significantly reduced the amplitude of Ca2+ into mitochondria averaging 0.36±0.01- versus 1.00±0.10-fold with SiC; P<0.05 (Supplementary Figure S3D), the overexpression of GSK3β in H9c2 cells significantly enhanced the Ca2+ transfer into mitochondria when compared with its respective control averaging 1.52±0.07 versus 1.00±0.04 in the pcDNA group; P<0.05 (Supplementary Figure S3E). These results were further confirmed using a mitochondrially targeted Ca2+ probe (Mitycam, Glasgow, Scotland, UK), suggesting that GSK3β specifically controls and regulates the IP3R-mediated SR/ER Ca2+ transfer to mitochondria in cardiac cells.

Figure 2
figure 2

GSK3β regulates the SR/ER-mitochondria Ca2+ transfer through IP3R in adult cardiomyocytes. (a) Adult cardiomyocytes were loaded with the Ca2+ indicator rhod-2 and MitoTracker Green. Merge signal revealed a mitochondrial pattern, especially apparent in enlarged insets. (Right) The quantitative pixel-by-pixel analysis measured with Zen software showed strong correlation between rhod-2 and Mitotracker Green with an overlap coefficient averaging 0.89±0.01. (b) Representative traces of mitochondrial Ca2+ transients. Adult cardiomyocytes were challenged with 100 μM histamine to induce Ca2+ release from the ER. Mitochondrial Ca2+ transfer was significantly reduced under GSK3β inhibition (SB21). (c) Time scan of mitochondrial Ca2+ challenged with caffeine stimulation (5 mM) showed that inhibition of GSK3β did not modify the caffeine-induced Ca2+ transfer from SR/ER to mitochondria. Maximal mitochondrial Ca2+ peak fluorescence are expressed as means±S.E.M. *P<0.05 versus respective control (n=5–6 distinct experiments; each experiment represents the average of 8–12 adult cardiomyocytes)

Full size image

To rule out a potential depolarization of mitochondria under GSK3β inhibition, we stained cardiac cells with TMRM to analyze mitochondrial membrane potential. Neither the pharmacological nor the genetic inhibition of GSK3β modified the mitochondrial membrane potential of cells (Supplementary Figures S4A and B), indicating that the observed reduction of Ca2+ transfer from SR/ER to mitochondria was not the consequence of a diminution of the driving force between the two organelles.

Inhibition of GSK3β at reoxygenation limits both cytosolic and mitochondrial Ca2+ overload by reducing Ca2+ release from SR/ER

It has been previously demonstrated that inhibition of GSK3β at reperfusion is cardioprotective by preventing PTP opening.6, 7 However, the relative cardioprotective mechanism still remains unknown. We here hypothesized that this beneficial effect might be linked to the limitation of both cytosolic and mitochondrial Ca2+ overload (and subsequent Ca2+-dependent PTP opening) by reducing Ca2+ release from IP3R at reperfusion.

To test our hypothesis, we started by investigating reticular (ER), cytosolic and mitochondrial Ca2+ homeostasis after hypoxia reoxygenation (HR) in H9c2 cells. In the control group (HR), 4 h hypoxia followed by 2 h reoxygenation induced a significant cell death averaging 50.4±2.9% versus 7.4±0.5% in basal group; P<0.05 (Figure 3a). As expected, SB21 (6 μM) at reoxygenation significantly reduced the cell death by half-averaging 25.9±1.8% versus 50.4±2.9% in the HR group (Figure 3a).

Figure 3
figure 3

GSK3β regulates the SR/ER Ca2+ homeostasis at reoxygenation in H9c2 cells. (a) Cell viability measured with propidium iodide revealed that SB21 clearly protected H9c2 cells against reoxygenation injury. Bar graphs were presented as the mean±S.E.M. (n=4 distinct experiments per group). *P<0.05. (b) Representative curves of SR/ER Ca2+ FRET ratio in hypoxic H9c2 cells. After 2 h reoxygenation, ER Ca2+ level was significantly higher in cells treated with SB21 (6 μM) at reoxygenation. Indicated results are means±S.E.M. of 4–6 distinct experiments (each experiment represents the average of 6–8 H9c2 cells). The addition of thapsigargin (2 μM) induced an ER Ca2+ efflux, which was 2.5-fold higher in the HR+SB21 group compared with the HR group (arrow). (c) Measurement of cytosolic Ca2+ at the end of reoxygenation with fura-2 (5 μM) with or without SB21 (6 μM). Results are means±S.E.M. of 4–5 distinct experiments. (Each experiment represents the average of 20–30 H9c2 cells.) *P<0.05 versus HR. (d) Representative curves of the time scan of mitochondrial Ca2+ (rhod-2, 2.5 μM) upon histamine stimulation (100 μM) after HR with or without SB21 at reoxygenation. Peak mitochondrial Ca2+ amplitude is expressed as means±S.E.M. *P<0.05 versus HR (n=5–6 distinct experiments; each experiment represents the average of 8–12 H9c2 cells)

Full size image

The ER Ca2+ level (measured with D1ER sensor24) was significantly higher in cells treated with SB21 at reoxygenation averaging 2.19±0.10 versus 1.44±0.11 in the HR group (Figure 3b). After addition of thapsigargin, the efflux of ER Ca2+ was 2.5-fold higher in the HR+SB21 group compared with the HR group (Figure 3b), reinforcing the notion that inhibition of GSK3β at reoxygenation increases the steady-state [Ca2+] in ER lumen during HR.

As a consequence, the basal cytosolic Ca2+ measured with the ratiometric dye fura-2 was significantly reduced in the HR+SB21 group averaging 0.30±0.02 versus 0.38±0.01 in the HR group; P<0.05 (Figure 3c), and the IP3R-mediated Ca2+ transfer from SR/ER to mitochondria was significantly abolished when GSK3β was inhibited at reoxygenation averaging 1.05±0.02 versus 1.10±0.03 in the HR group; P<0.05 (Figure 3d). Altogether, these results indicate that SB21-dependent cell protection is associated with a limitation of both cytosolic and mitochondrial Ca2+ overload likely through a reduction of Ca2+ release from SR/ER at reoxygenation.

IP3R is a downstream substrate of GSK3β during IR injury

We then sought to determine the mechanism by which GSK3β regulates IP3R in the heart. As IP3Rs are known to contain the GSK3β consensus sequence (Ser/Thr-XXX-pSer/Thr) for phosphorylation (at Ser1756) that activates IP3R, we next questioned whether IP3R might be phosphorylated by GSK3β at the SR/ER-mitochondria interface in the heart. To test this hypothesis, we investigated the posttranslational modification of IP3R in adult cardiomyocytes during HR.

As expected, HR induced a significant cell death averaging 65.3±3.2% versus 13.6±0.7% in the basal group (Figure 4a). Concomitantly, we observed a significant increase in phosphorylation of GSK3β at tyrosine 216 (Tyr216) residue (known to facilitate GSK3β activity) averaging 1.82±0.08- versus 1.00±0.05-fold in the basal group; P<0.05 (Figure 4b). At reoxygenation, SB21 significantly reduced both cell death (around 50%; Figure 4a) and phosphorylation at Tyr216 of GSK3β averaging 1.18±0.10 versus 1.82±0.08 in the HR group (Figure 4b). These results suggest that GSK3β activity is linked with cell death of adult cardiomyocytes during HR.

Figure 4
figure 4

GSK3β inhibition at reoxygenation reverses the activation of IP3R induced by HR in adult cardiomyocytes. (a) Cell viability measured with propidium iodide revealed that SB21 clearly protected adult cardiomyocytes against reoxygenation injury (n=4 distinct experiments per group); mean±S.E.M. of cell death is shown as percentage. *P<0.05. (b) GSK3β activity quantified by western immunoblotting of phospho-GSK3β (Tyr216, active form) under hypoxic conditions. Results were normalized with total GSK3β (n=4–5 distinct experiments per group). *P<0.05. (c) Immunoblotting of relative expression of phosphor-Ser1756 IP3R normalized with total IP3R during HR injury. Results expressed as fold versus basal are means±S.E.M. *P<0.05 (n=4 distinct experiments). (d) In situ protein interactions using PLA in adult cardiomyocytes revealed that SB21 (6 μM) significantly reduced the interaction of GSK3β with IP3R induced by HR. Results are expressed as means±S.E.M. of 8–10 distinct experiments (each experiments represent the average of 4–8 adult cardiomyocytes). *P<0.05

Full size image

HR induced a significant increase of IP3R phosphorylation of Ser1756, which was markedly reduced by SB21 at reoxygenation averaging 6.24±0.42- and 1.32±0.12-fold versus basal, respectively (Figure 4c). The decreased IP3R phosphorylation and the consecutive limitation of its activity seem to be an important mechanism in the protection of cardiomyocytes elicited by GSK3β inhibition at reoxygenation. This idea was supported by the protein–protein interaction of GSK3β with IP3R in cardiomyocytes during HR. Indeed, using the PLA approach, our data showed that HR doubled the interaction of GSK3β with IP3R averaging 2.23±0.23- versus 1.0±0.19-fold in basal (Figure 4d). SB21 reduced this interaction at reoxygenation averaging 0.74±0.10 versus 2.23±0.23 in the HR group (P<0.05, Figure 4d), suggesting that IP3R might be a downstream substrate of GSK3β during HR injury.

Eventually, we tested whether GSK3β regulates the posttranslational modification of IP3R during in vivo IR injury in a mouse model. As previously demonstrated by our group,6, 7 mice treated with SB21 at reperfusion (IR+SB21, 70 μg/kg) exhibit a significant reduction of infarct size after prolonged IR injury when compared with control (IR), averaging 31.0±0.6% versus 44.6±1.7%, respectively (Figure 5a). SB21 administered at reperfusion significantly decreased the phosphorylation of both GSK3β (Tyr216) and IP3R (Ser1756) induced by reperfusion injury (Figures 5b and c), suggesting that GSK3β controls the extent of infarct size through the regulation of the phosphorylation state of IP3R at reperfusion. The in vivo link between GSK3β and IP3R was further reinforced by the reduction of IP3R co-IP with the partner of the Ca2+ channeling complex at MAMs including GSK3β, as well as Grp75, VDAC and CypD in the area at risk of ischemic hearts treated with SB21 at reperfusion (Figure 5d).

Figure 5
figure 5

In vivo GSK3β inhibition-mediated cardioprotection is correlated to the inactivation of IP3R at reperfusion. (a) Cardioprotective effect of SB21 at reperfusion. Percentage of both area at risk (AR/LV) and area of necrosis (AN/AR) are presented for each groups (n=6 per group). As illustrated in the right panel, injection of SB21 (70 μg/kg) at reperfusion reduced the size of myocardial infarction compared with IR group. (b) Relative GSK3β activity at reperfusion quantified by western immunoblotting of phospho-GSK3β (Tyr216, active form). Results were normalized with total GSK3β (n=4–5 distinct experiments per group). *P<0.05. (c) Immunoblotting of relative expression of phospho-Ser1756 IP3R revealed that SB21 significantly reduced the activity of IP3R during IR injury. Results are means±S.E.M. *P<0.05 (n=4–6 distinct experiments). (d) Hearts protected with SB21 (70 μg/kg) at reperfusion exhibited a clear reduction of IP3R interaction with Grp75, GSK3β, VDAC and CypD. Results shown are one result representative of three trials

Full size image

Discussion

The present results suggest that (1) a fraction of GSK3β is localized at the SR/ER-mitochondria interface in the heart; (2) GSK3β regulates the IP3R-dependent Ca2+ transfer from SR/ER to mitochondria in cardiac cells and (3) cardioprotection mediated by GSK3β inhibition limits both cytosolic and mitochondrial Ca2+ overload by limiting IP3R channel opening at reperfusion.

We examined whether inhibition of GSK3β might protect the heart from reperfusion injury.7 Our results point to a regulatory effect of GSK3β on the Ca2+ transfer between SR/ER and mitochondria at the MAMs interface during IR injury.

In the heart, although a mitochondrial targeting sequence (MTS) has not been identified in GSK3β, it has been recently demonstrated that an N-terminal domain of GSK3β could function as an MTS, permitting the translocation of GSK3β to mitochondria in a kinase-activity-dependent manner to regulate mitochondrial functions.13 Here, we show that a fraction of GSK3β is located in SR/ER and MAMs. We demonstrated that GSK3β physically interacts with the IP3R-Ca2+ channeling complex including VDAC, Grp75, CypD and IP3R at MAMs. Moreover, we found that pharmacological and siRNA-mediated inhibition of GSK3β altered the interaction among proteins of the IP3R-Ca2+ channeling complex, suggesting a role for GSK3β in the structure of this complex.

We observed that GSK3β does not interact with RyRs, and the RyR-dependent Ca2+ release does not seem to be affected by GSK3β inhibition in cardiac cells. The inhibition of GSK3β was sufficient to reduce the Ca2+ transfer from SR/ER to mitochondria via a specific action on IP3R Ca2+ leak channels at MAMs. One can therefore question the respective importance of IP3Rs- versus RyRs-mediated Ca2+ release following reperfusion injury in cardiomyocytes. IP3Rs are involved in cardiac Ca2+ signaling including the ECC in the normal heart, and also in some pathologies such as hypertrophy.19, 25, 26, 27 Our data reinforce this idea as both pharmacologic and genetic inhibition of GSK3β leads to a decreased Ca2+ transfer from SR/ER to mitochondria, whereas overexpression of GSK3β enhances the entrance of Ca2+ into mitochondria under IP3R leak channel stimulation.

Whether GSK3β inhibition specifically impacts the SR/ER Ca2+release or rather the Ca2+ entrance into mitochondria was not addressed in this study. Nevertheless, we demonstrated that pharmacological or genetic inhibition of GSK3β did not modify the mitochondrial membrane potential, indicating that the observed reduction of Ca2+ transfer from SR/ER to mitochondria was likely not the result of a limitation of the driving force between the two organelles.

Upon IR, the accumulation of Ca2+ into mitochondria is known to trigger PTP opening and cell death.28 Ca2+ crosstalk between SR/ER and mitochondria has been involved in myocardial reperfusion injury.29 As cardioprotective mechanisms of GSK3β at reperfusion are still not clear, we hypothesized that GSK3β inhibition limits both cytosolic and mitochondrial Ca2+ overload by limiting IP3R channel opening and subsequent PTP-dependent cell death at reperfusion.

We demonstrated here that the interaction of GSK3β with all the partners of IP3R Ca2+ channeling complex, including CypD, VDAC, Grp75 and IP3R, was significantly increased after HR and associated with mitochondrial Ca2+ overload and cell death. Interestingly, the pharmacological inhibition of GSK3β not only decreased the interaction of GSK3β with the Ca2+ channeling complex but also decreased the protein composition of the IP3R channeling complex and attenuated both cytosolic and mitochondrial Ca2+ overload at reperfusion. This suggests that binding of GSK3β to IP3R may be responsible of the Ca2+ crosstalk between SR/ER and mitochondria during reperfusion injury.

Different observations were in favor of the causal connection of GSK3β with IP3R during IR. First, we provide in vivo and in vitro evidence that pharmacological inhibition of GSK3β reduced the physical interaction of GSK3β with IP3R induced by IR. Second, the Ser1756 phosphorylation of IP3R was significantly reduced when GSK3β was inhibited at reperfusion. As a result, the open probability of the Ca2+ channels was reduced at reperfusion inducing a higher level of Ca2+ into SR/ER lumen and a reduction of Ca2+ overload in both cytosol and mitochondria. Third, GSK3β inhibition at reperfusion reduced the interaction of IP3R with the other partners of the Ca2+ channeling complex including CypD, VDAC, Grp75 and GSK3β.

Although IP3R2 is recognized as the predominant isoform in the cardiac myocyte,27 IP3R1 is highly expressed in MAMs heart.21, 30,31 Our results show that GSK3β preferentially interacts with IP3R1, and can also interacts with IP3R2 and IP3R3 in cardiac cells (data not shown). In summary, we propose that inhibition of GSK3β might decrease the phosphorylation and the activity of IP3R, thereby limiting cytosolic and mitochondrial Ca2+ overload, possibly contributing to the reduction of both cell death and infarct size at reperfusion. Our proposal is in line with previous reports demonstrating that SR/ER loses almost 90% of its Ca2+ at the onset of reperfusion,32 and that specific IP3R channels inhibition can protect the heart at reperfusion.33 One may question whether IP3R may be a direct or an indirect substrate of GSK3β. Such mechanism remains to be investigated in-depth in future studies. However, our results showed (1) a positive correlation between the activity of GSK3β and the phosphorylation rate of IP3R, (2) a linear correlation between the Ser1756 phosphorylation of IP3R and the protein–protein interaction of GSK3β with IP3R and (3) in silico analysis revealed at least two putative GSK3β phosphorylation sites on IP3R, Ser1756 and Ser1765, reinforcing the notion that IP3R could be a substrate of GSK3β.

Whether other factors may contribute to the increase of Ca2+ into SR/ER lumen upon inhibition of GSK3β at reperfusion remains to be determined (i.e. ANT, VDAC, CypD). Moreover, we cannot fully exclude a contribution of SERCA2 since Michael et al.12 previously demonstrated that overexpression of GSK3β could significantly reduce the expression of SERCA2 in the heart.12 We noticed in the present work that SERCA2 expression seems to be increased when GSK3β was inhibited at the onset of reperfusion (data not shown). Such mechanism remains to be investigated in-depth in future studies, especially since a problem of timing has to be considered between the speed of Ca2+ response and the time needed for SERCA2 protein transcription at the onset of reperfusion.

To conclude, our work identifies a new role for GSK3β in cardioprotection as an important regulator of Ca2+ transfer between SR/ER and mitochondria during IR injury. We propose that GSK3β inhibition at reperfusion may protect heart through the limitation of SR/ER Ca2+ release by inhibition of IP3R Ca2+ channeling complex and consecutive reduction of both cytosolic and mitochondrial Ca2+ overload and subsequent PTP opening.

Materials and Methods

Animals

The investigation conformed to the Guide for the Care and Use of Laboratory Animals published by the US National Institute of Health (NIH Publication No. 85-23, revised 1996) and was approved by the local institutional animal research committee (N BH2012-64 and 901I).

Subcellular fractionation

Cell fractionation was performed by differential ultracentrifugation as described previously.22, 34 Briefly, hearts were homogenized with a glass potter in an isolation buffer (225 mM mannitol, 75 mM sucrose, 0.5% BSA, 0.5 mM EGTA and 30 mM Tris-HCl, pH 7.4). Both nuclei and cellular debris were then pelleted by two different centrifugations at 740 × g for 5 min. The supernatant was centrifuged at 9000 × g for 10 min to pellet crude mitochondria that were resuspended in mitochondria resuspending buffer (MRB) containing 250 mM mannitol, 5 mM HEPES and 0.5 mM EGTA, pH 7.4. Resulting supernatant was further centrifuged at 20 000 × g for 30 min followed by 100 000 × g for 1 h to pellet the SR/ER.

Purified mitochondria were obtained by centrifugation of crude mitochondria fraction through a Percoll medium at 95 000 × g for 30 min in an SW40 rotor (Beckman Coulter, Villepinte, France). The resulting pellet was then washed two times in MRB by centrifugation at 6300 × g for 10 min to obtain purified mitochondria.

To purify the MAM fraction localized in the diffused white band (under pure mitochondria band), MAM were diluted in MRB and centrifuged at 6300 × g for 10 min to remove any contaminating mitochondria. MAM supernatant was then centrifuged at 100 000 × g for 1 h in a 70 Ti rotor (Beckman Coulter) and the resulting pellet was resuspended in MRB. Protein content was assayed by the Lowry method for the following analysis.

Blue native and SDS-PAGE 2D separation

Hearts from wild-type (WT) mice were homogenized three times at 6000 Hz for 10 s (Precellys 24; Ozyme, Montigny-le-Bretonneux, France) in a buffer (100 mM KCl, 50 mM MOPS, 5 mM MgSO4, 1 mM EGTA, 1 mM ATP). Heart homogenates were then solubilized with n-dodecyl-β-d-maltopyranoside (detergent/protein ratio of 1.6:1, g/g), combined with Coomassie Blue G-250 dye (detergent/dye ratio of 4:1, g/g) and separated on 4–16% Bis-Tris native PAGE gels (Invitrogen, Cergy Pontoise, France) for the first dimension. The first dimension gel was then excised and reduced with 1 × NuPage reducing agent in 1 × NuPage LDS sample buffer (Invitrogen) for 15 min at room temperature (RT). Cysteine alkylation was carried out with N, N-dimethylacrylamide for 15 min at RT. The reaction was quenched for 15 min at RT with 20% ethanol in 1 × LDS and 0.1 × reducing agent. Finally, the equilibrated gel strip was applied to the second dimension on a 4–16% Criterion Bis-Tris gel, separated and transferred to the PVDF membrane.

IP and WB analysis

Total lysates of cardiac cells were prepared in lysis buffer containing 1% NP40, 20 mM Tris-HCl, 138 mM NaCl, 2.7 mM KCl, 1 mM MgCl2, 5% glycerol, 5 mM EDTA, 1 mM DTT supplemented with a cocktail of protease inhibitors (Sigma-Aldrich, Saint-Quentin Fallavier, France) and phosphatases inhibitors (PhosphoStop; Roche Diagnostics, Meylan, France). For IP, 2 μg of antibodies directed against the protein of interest were added to the IP buffer (except for the control, which was supplemented with 2 μg of anti-mouse IgG). A range of 300–500 μg of total protein (depending on the origin of the samples) was added in the previous mix and incubated overnight at 4 °C. Fifty microliters of magnetic beads (Millipore, Fontenay sous Bois, France; no. LSKMAGG10) were then incubated for 2 h at 4 °C and finally precipitated with a strong magnet. After three washes in IP buffer, proteins were resuspended in Laemmli buffer and denatured at 95 °C for 10 min. Immunoprecipitated proteins were then analyzed by WB. Migration was performed in 4–20% acrylamide gel and semidry blotter was used to transfer the proteins. All primary antibodies were used at a dilution of 1:1000, except CypD (1 : 4000). Mouse and rabbit secondary antibodies were incubated for 1 h at RT at 1 : 10 000 (Bio-Rad, Marnes-la-Coquette, France). ECL+ reagent (GE Healthcare, Velizy-Villacoublay, France) was used for the membrane revelation, acquisition was performed with Bio-Rad Molecular Imager Gel Doc XR+, and ImageLab software was used for quantification (Bio-Rad). Antibodies IP3R (sc-28614), Grp75 (sc133137), P-GSK3β (sc-135653) and IgG (sc-66931) were from Santa Cruz Biotechnology (Dallas, TX, USA); P-IP3R (no. 8548), GSK3β (no. 9315), Serca2 (no. 4388), COXIV (cs no. 4844) and GAPDH (no. 2118) were from Cell Signaling Technology (Saint Quentin Yvelines, France); and VDAC (Ab14734), CypD (Ab110324) and RyR2 (Ab 117840) were from Abcam (Paris, France). Relative expressions of phosphorylated proteins were normalized to respective total form and results were expressed as fold versus basal condition.

Isolation of adult murine cardiomyocytes

Ventricular cardiomyocytes were isolated using enzymatic digestion according to previously described procedure.35, 36 Briefly, WT mice were anesthetized with pentobarbital (70 mg/kg) (Sanofi Santé Animale, Libourne, France) and the heart was quickly harvested and retrogradely perfused at 37 °C for 6–8 min with a perfusion buffer (113 mM NaCl, 4.7 mM KCl, 0.6 mM KH2PO4, 0.6 mM Na2HPO4, 1.2 mM MgSO4, 12 mM NaHCO3, 10 mM KHCO3, 10 mM HEPES, 30 mM Taurine, pH 7.4) containing 10 mM 2,3-butanedione, 5.5 mM glucose, 12.5 μM CaCl2, 0.1 mg/ml of liberase (Roche) and 0.14 mg/ml trypsin (Sigma). Isolated myocytes were then maintained in culture in laminin- (10 μg/ml) precoated dishes with M199 medium (Invitrogen) and allowed to attach for 2 h before analysis.

Duolink in situ PLA

We used PLA experiments to evaluate the proximity of two proteins and their potential interaction (<40 nm) as described previously.22 Adult cardiomyocytes (or H9c2 cells) were plated into Lab-Tech slides (Thermo Scientific; no.155411). Cells were first fixed in 4% PFA for 10 min at RT. Cell permeabilization was carried out in 0.1% Triton X-100 for 15 min at RT followed by blocking for 30 min in a 37 °C preheated humidity chamber. Primary antibodies directed against the two proteins of interest were then added at 1 : 200 and incubated overnight at 4 °C. From this point, each step was separated by two washes of 5 min in TBST and incubations were performed at 37 °C in a humidity chamber. PLA probes, which are composed of antibodies linked to small complementary oligonucleotides, were added in accordance with the polarity for 1 h. Ligation reaction was then performed for 30 min to ligate the complementary oligonucleotides. Polymerization of the ligated fragment was performed with a Taq polymerase (Duolink II in situ PLA; Olink Bioscience). Cells were then washed in 1 × and 0.01 × SSC buffer (150 mM NaCl, 15 mM sodium citrate for 1 ×) and dried at RT in the dark to preserve fluorescence. Slides were mounted in mounting medium containing DAPI and observed with a Leica SP5X confocal microscope (Wetzar, Germany). A maximum intensity projection created an output image from Z stack performed at 0.5 μm depth. A total of 10–15 cells were counted in each experiment (n=7–16) for quantification. We used BlobFinder software (Image Analysis, Uppsala University, Uppsala, Sweden) to quantify the red dots corresponding to the positive interaction of the different proteins of interest.

Cytosolic, mitochondrial and SR/ER Ca2+ measurements

To monitor mitochondrial Ca2+, isolated adult cardiomyocytes (or H9c2 cells) were loaded with 2.5 μM rhod-2/AM (Molecular Probes, Eugene, OR, USA) for 30 min at 37 °C in the presence of 0.003% (w/v) pluronic acid. After 5 h de-esterification, cardiac cells were washed and placed in 1 ml phenol-free DMEM to the heated stage (37 °C) of an inverted epifluorescence microscope (x40 oil-immersion objective) connected to a cooled CCD camera (PXL; Photometrics). Cytosolic Ca2+ dynamics were measured on H9c2 loaded with 5μM fura-2/AM. Imaging was performed with a digital cooled CCD camera (Photometrics, Tucson, AZ, USA) on an inverted microscope (Nikon Diaphot, Melville, NY, USA) using a x40 oil-immersion objective. SR/ER Ca2+ level was measured with a FRET-based genetic sensor (D1ER), which allows quantification of Ca2+ because of its ratiometric properties.37

Cellular models of HR

Adult cardiomyocytes or H9c2 were subjected to simulated hypoxia in a controlled hypoxic chamber (Adelbio, Clermont-Ferrand, France), induced by nitrogen flushing up to 1% partial O2 pressure for 45 min and 4 h, respectively, in a Tyrode solution (130 mM NaCl, 5 mM KCl, 10 mM HEPES, 1 mM MgCl2, 1.8 mM CaCl2 at pH 7.4/37 °C) followed by 2 h of reoxygenation at 37 °C in normal culture medium. Cells received vehicle with no additional intervention (HR) or 6 μM SB21 (HR+SB21) at the onset of reoxygenation. Time control group consisted of cells without the hypoxic stimulus. Cell death was measured by counting PI-positive cells by imaging or flow cytometry (FACSCalibur; BD Biosciences, Le Pont de Claix, France) at the end of the reperfusion period.

In vivo model of acute myocardial IR injury

C57BL/6 J (male, 8–10 weeks old) were anesthetized by intraperitoneal mixture of 70 mg/kg pentobarbital sodium and 0.1 mg/kg buprenorphine as described previously.7, 38, 39, 40 The animals were orally intubated, ventilated and body temperature was maintained at 37 °C. A left thoracotomy was performed in the fourth intercostal space. A small curved needle was passed around the left anterior descending coronary artery to induce IR. Mice were subjected to 45 min regional myocardial ischemia followed by 24 h reperfusion. Myocardial infarct size was then determined by triphenyltetrazolium staining as described previously.7, 38, 39, 40

Statistical analysis

Data are presented as mean±S.E.M., unless otherwise specified. Statistical analysis was performed with GraphPad Prism 5 software (La Jolla, CA, USA) with the use of nonparametric tests. Comparisons among >2 groups were analyzed by the Kruskal–Wallis test. When the Kruskal–Wallis test was significant, the Dunn’s appropriate multiple testing procedure was performed. For all other analysis, data were compared by the Mann–Whitney rank-sum test. Statistical significance was defined as a value of P<0.05.