Purinergic control of AMPK activation by ATP released through connexin 43 hemichannels – pivotal roles in hemichannel-mediated cell injury

ATP is the energy source for cell function and metabolic activities, and is mainly produced in mitochondria. ATP is also constantly secreted into the extracellular milieu, through several different routes, where it functions as a paracrine and/or autocrine mediator to regulate a wide range of cell functions (Baroja-Mazo et al., 2013; Gordon, 1986; Lazarowski, 2012; Schwiebert and Zsembery, 2003; Wang et al., 2013). Most of the actions of the extracellular ATP are mediated through its binding to purinergic receptors.

AMP-activated protein kinase (AMPK) is one of the major kinases in the control of ATP homeostasis. It is ubiquitously expressed in all cells and acts as an ultrasensitive sensor of cellular energy. It is activated by an increase in the cellular AMP to ATP ratio, which commonly occurs during metabolic stresses that either interfere with ATP production or accelerate ATP consumption. Once activated, AMPK switches off ATP-consuming pathways and switches on ATP-producing pathways; thus, maintaining levels of cellular ATP (Hardie, 2011; Mihaylova and Shaw, 2011). Although alterations in the level of intracellular ATP are usually the result of changes in ATP generation or consumption, it can also be caused by certain pathophysiological situations that lead to loss of ATP into the extracellular space (Baroja-Mazo et al., 2013; Wang et al., 2013). Because of the barrier created by the cell membrane, it is unclear whether or not the extracellular ATP also influences AMPK activity.

Gap junctions are intercellular channels that permit the direct exchange of ions, secondary messengers and small signaling molecules between neighboring cells. Each gap junction channel is composed of two hemichannels that reside in the plasma membrane of two closely apposed cells. Connexin (Cx) proteins form gap junctions and, to date, more than 20 different connexin molecules have been identified. Among them, Cx43 has been extensively investigated because of its ubiquitous expression in a variety of cell types. Intercellular communication via gap junctions is thought to play an important role in the regulation of cell functions – including cell proliferation, migration, differentiation and survival (Saez et al., 2003; Yao et al., 2009; Yao et al., 2007). The majority of the biological effects of gap junctions are mediated by the direct transmission of signaling molecules among neighboring cells. Nonjunctional connexin hemichannels also participate in the regulation of cell signaling and cell behaviors through the release of the mediators ATP, NAD⁺ or glutamate (Baroja-Mazo et al., 2013; Evans et al., 2006; Saez et al., 2003; Wang et al., 2013).

The activation of hemichannels has been reported in a variety of pathological situations and has been shown to influence cell responses to various stimuli (Fang et al., 2011; Ramachandran et al., 2007; Thompson et al., 2006). The molecular mechanisms underlying the effects of hemichannels on various cell functions are still poorly understood. Several studies have pointed to a role for ATP that has been released into the extracellular space. Indeed, hemichannel-derived ATP has been implicated in the transmission of the intercellular Ca²⁺ signal and in the control of cell growth, migration and survival (Anselmi et al., 2008; Fang et al., 2011; Pearson et al., 2005; Stout et al., 2002; Zhao et al., 2005). Given that the efflux of ATP through hemichannels leads to an altered intracellular level of ATP, we speculated that hemichannel opening might affect the activation of AMPK and subsequently lead to altered cell behaviors. Therefore, we sought to address this hypothesis and, here, we present our results showing the existence of the previously unrecognized channel-mediated regulation of AMPK. Furthermore, we characterize AMPK as a pivotal molecule involved in the biological actions of hemichannels on cell survival.

We have used a well-documented model of hemichannel opening that is initiated by the removal of extracellular Ca²⁺ (Quist et al., 2000; Stout et al., 2002). As shown in Fig. 1, the removal of extracellular Ca²⁺ initiated an exchange of small molecules across the plasma membrane, as indicated by the influx of the small-molecular-mass fluorescent dyes Lucifer Yellow and ethidium bromide, and the efflux of ATP (Fig. 1A–C). This effect was rapid as it was observable immediately after Ca²⁺ removal and was not associated with an increased release of lactate dehydrogenase (LDH) (data not shown). Inhibition of hemichannels with the hemichannel blockers heptanol or lindane, or downregulation of Cx43 with specific siRNA largely prevented the diffusion of the small molecules (Fig. 1D–F). These results indicate that Ca²⁺ deprivation activates Cx43 hemichannels and induces extracellular release of ATP (Li et al., 2013; Müller et al., 2002; Stout et al., 2002).

Given that AMPK activity is inversely correlated with the levels of intracellular ATP (Hahn-Windgassen et al., 2005; Mihaylova and Shaw, 2011), we speculated that hemichannel-mediated loss of ATP to the extracellular space would affect AMPK activation. Therefore, we examined the level of AMPK phosphorylation at threonine residue 172. The level of phosphorylation at this site is correlated with AMPK activity (Towler and Hardie, 2007). Removing extracellular Ca²⁺ induced a weak activation of AMPK, as reflected by the appearance of a phosphorylated band of AMPK (Fig. 2A). Contrary to our initial expectation, blockade of the release ATP to the extracellular space with the hemichannel blockers heptanol or lindane markedly potentiated AMPK activation (Fig. 2A,B). Treatment of cells with lanthanum (La³⁺) or gadolinium (Gd³⁺), two nonselective cation channel blockers that also inhibit hemichannels (Sàez et al., 2005), also produced a similar potentiating effect (Fig. 2C).

To confirm the above finding, we modified the expression of Cx43 by treating NRK cells with siRNA against Cx43, which is the only connexin molecule expressed in NRK cells (Yao et al., 2010), or by transfection of gap-junction-deficient LLC-PK1 cells with a vector encoding wild-type Cx43 (Fang et al., 2011). As shown in Fig. 2D,E, downregulation of Cx43 in NRK cells by using specific siRNA markedly augmented AMPK activation upon removal of extracellular Ca²⁺. On the contrary, forced expression of Cx43 in gap-junction-deficient LLC-PK1 cells greatly suppressed AMPK activation, as compared with control cells that had been transfected with a vector encoding a non-functional Cx43 mutant (Fig. 2F) (Fang et al., 2011). Similar to the effects in NRK cells, the gap junction blockers heptanol and lindane also potentiated AMPK activation in LLC-PK1 cells expressing wild-type EGFP–Cx43 (data not shown). Thus, these results indicate the existence of a mechanism for the suppression of AMPK that is mediated by Cx43 hemichannels.

Because ATP is one of the major determinants of AMPK activity (Hahn-Windgassen et al., 2005; Mihaylova and Shaw, 2011), we examined the possible role of ATP. Therefore, we inhibited ATP production by treating cells with antimycin (an inhibitor of mitochondrial electron transport) and iodoacetic acid (an inhibitor of glycolysis). As expected, these chemicals caused a complete depletion of intracellular ATP (Fig. 3A). Consequently, the extracellular efflux of ATP that was triggered by removal of extracellular Ca²⁺ was also abolished (Fig. 3B). The depletion of ATP was associated with a markedly increased level of AMPK phosphorylation (Fig. 3C). The potentiating effect of heptanol on AMPK activation diminished in the absence of ATP. These results suggest that extracellular ATP is important for the regulatory effects of hemichannels on AMPK activation.

To establish the role of the extracellular ATP, we examined the influence of ATP-degrading enzymes or exogenously added ATP on AMPK activation. As shown in Fig. 3D, the promotion of ATP degradation with Apyrase augmented AMPK activation, whereas supplementing with ATP counteracted the effect of hemichannel blockers on AMPK activation (Fig. 3E,F). These results indicate that hemichannel-derived ATP suppresses AMPK activation.

Most of the cellular actions of extracellular ATP are mediated through its binding to P2 purinoceptors (Baroja-Mazo et al., 2013; Wang et al., 2013). At present, two classes of purinoceptors have been identified (P2X and P2Y) and both are expressed in renal tubular cells (Praetorius and Leipziger, 2010). We therefore examined the possible involvement of purinergic signaling in the suppression of AMPK activation. Suramin and PPADS, two effective antagonists of P2 purinoceptors (Charlton et al., 1996), markedly potentiated AMPK activation, indicating an involvement of purinergic signaling in suppression of AMPK activation (Fig. 3G,H).

To investigate the signaling molecules downstream of P2 purinoceptors, we focused on Akt (AKT1), which is a reported negative regulator of AMPK (Hahn-Windgassen et al., 2005; Liu et al., 2012; Soltys et al., 2006). Removal of extracellular Ca²⁺ triggered a rapid activation of Akt that was largely blocked by the P2 receptor antagonists suramin and PPADS (Fig. 4A). Consistent with this, exogenous ATP induced a concentration-dependent activation of Akt (Fig. 4B). Intriguingly, this effect of ATP was more robust under Ca²⁺-free conditions.

To establish the role of Akt in hemichannel- and/or extracellular-ATP-mediated suppression of AMPK, we examined AMPK activation after inhibition of Akt and its upstream regulator phosphoinositide 3-kinase (PI3K). As shown in Fig. 4C–D, the PI3K inhibitor LY294002 and the Akt inhibitor Akti1/2 effectively abrogated the activation of Akt that was triggered by Ca²⁺ deprivation and both significantly potentiated AMPK activation.

To establish the role of Akt further, we looked at the effects of transfecting cells with a myristoylated (constitutively active) form of Akt (Johno et al., 2012). Cells expressing constitutively active Akt had a higher basal level of Akt phosphorylation and a lower level of AMPK activation (Fig. 4E–G). These observations, thus, support an involvement of Akt in hemichannel-mediated suppression of AMPK.

As inactivation of the LKB1 kinase (STK11), which normally phosphorylates AMPK, is one of the reported mechanisms underlying the inhibitory effect of Akt on AMPK activity (Liu et al., 2012; Soltys et al., 2006), we examined LKB1 activation under Ca²⁺-free conditions and its regulation by the P2 purinoceptor pathway. Removal of extracellular Ca²⁺, indeed, caused an elevation in the level of LKB1 phosphorylation, which was suppressed by the addition of exogenous ATP (Fig. 5A). By contrast, the promotion of ATP degradation with Apyrase or blockade of P2 purinoceptors with PPADS greatly potentiated LKB1 activation (Fig. 5B). A similar potentiating effect was also achieved by blockade of the PI3K–Akt pathway with specific inhibitors (Fig. 5C,D). These results, thus, indicate a P2-purinoceptor-mediated suppression of LKB1 under Ca²⁺-free conditions.

Activation of mammalian target of rapamycin (mTOR) has been reported to be an alternative mechanism by which Akt exerts its suppressive effect on AMPK activity (Hahn-Windgassen et al., 2005). Therefore, we assessed the possible involvement of this mechanism. Ca²⁺ depletion, indeed, induced an Akt-dependent activation of mTOR that was reflected by the increased phosphorylation of p70S6K (RPS6KB1), a downstream target of mTOR, and the prevention of phosphorylation upon treatment with the Akt inhibitor Akti1/2 (Fig. 5E,F). Treatment of cells with rapamycin, an inhibitor of mTOR, increased AMPK activation (Fig. 5E,G). These observations, thus, indicate an involvement of mTOR in Akt-mediated suppression of AMPK.

We have recently reported that Cd²⁺ activates Cx43 hemichannels and causes an extracellular accumulation of ATP in LLC-PK1 cells that express EGFP–Cx43 (Fang et al., 2011; Li et al., 2013). Furthermore, we suggested that there is a role for hemichannels in Cd²⁺-induced cell injury (Fang et al., 2011). Using this model, we confirmed our findings and explored the possible participation of the signaling cascades in Cd²⁺-induced cell injury.

Consistent with our previous reports (Fang et al., 2011; Li et al., 2013), Cd²⁺ caused release of ATP into the extracellular space in LLC-PK1 cells expressing EGFP–Cx43, which was significantly prevented by the hemichannel blockers heptanol and lindane (Fig. 6A), indicating an involvement of hemichannels. Similar to the model of Ca²⁺ depletion, hemichannel opening that was triggered by Cd²⁺ was also associated with a modest activation of AMPK, which could also be potentiated by hemichannel blockers and a P2 purinoceptor antagonist, suramin (Fig. 6B–D). These observations, thus, provide additional evidence to support an involvement of hemichannel-derived ATP in the suppression of AMPK activation.

Several recent studies have demonstrated that Cd²⁺-induced cell death is through the activation of mTOR (Chen et al., 2011; Xu et al., 2011). Given that AMPK negatively regulates mTOR (Viollet et al., 2010), we speculated that hemichannel-mediated regulation of AMPK might be implicated in Cd²⁺-induced cell death. To test this hypothesis, we first confirmed that Cd²⁺ induced activation of mTOR in our experimental settings. Exposure of LLC-PK1 cells that expressed EGFP–Cx43 to Cd²⁺ caused a time-dependent phosphorylation of p70S6K, indicative of mTOR activation (Fig. 7A). We then examined whether mTOR was affected by AMPK and hemichannels. Activation of AMPK, by using aminoimidazole carboxamide ribonucleotide (AICAR), and inhibition of hemichannels with heptanol both suppressed mTOR activation, as indicated by the decreased phosphorylation of p70S6K (Fig. 7B,C). Consistent with the reported mutual inhibition of mTOR and AMPK, there was an inverse correlation between the phosphorylated levels of p70S6K and AMPK. In contrast to the decreased mTOR activity upon activation of AMPK (Fig. 7B), suppression of mTOR with rapamycin enhanced AMPK activation (Fig. 7C). Finally, we asked whether suppression of mTOR attenuated cell injury. The change in cell shape and the cell death induced by Cd²⁺ were markedly prevented by the mTOR inhibitor rapamycin, the AMPK activators AICAR or metformin, and the hemichannel blockers heptanol or lindane (Fig. 7D,E). Collectively, these observations support the hypothesis that involvement of hemichannel-mediated regulation of AMPK is involved in Cd²⁺-induced cell death.

In this study, we reveal a previously unrecognized channel-mediated regulation of AMPK through a purinergic signaling pathway. Furthermore, we define AMPK as a pivotal molecule that underlies the action of hemichannels on cell survival.

Activation of hemichannels has been reported in various pathophysiological conditions, including connexin mutations, depolarization of the membrane potential, hypoxia and changes in intra- and extracellular Ca²⁺, as well as cellular redox status (Anselmi et al., 2008; Sáez et al., 2010). In this study, hemichannels were activated by removal of extracellular Ca²⁺. The opening of hemichannels under Ca²⁺-free conditions has been demonstrated at both structural and functional levels. Using an atomic force microscope, Muller and colleagues (Müller et al., 2002) observed that low extracellular Ca²⁺ increased the outer pore diameter of hemichannels. Consistently, many investigators have found increased hemichannel permeability (Li et al., 2013; Stout et al., 2002). Here, we detected an increased exchange of small molecules across the cell membrane that was disrupted by blockade of Cx43-forming channels, indicating an involvement of hemichannels.

In this investigation, we characterized AMPK as a novel target of extracellular ATP. This is shown by the fact that the prevention of ATP release through blockade of hemichannels, the promotion of ATP degradation with Apyrase or the disruption of purinergic signaling by using receptor antagonists all potentiated AMPK activation. Furthermore, exogenous ATP suppressed AMPK activation that had been initiated by removal of extracellular Ca²⁺. ATP also suppressed AMPK activation triggered by the AMPK activator metformin (data not shown) (Shackelford and Shaw, 2009; Shaw, 2009). As an ultrasensitive sensor that monitors cellular energy status and maintains energy homeostasis, AMPK is under the negative-feedback control of ATP (Hardie, 2011; Mihaylova and Shaw, 2011). In this context, suppression of AMPK by extracellular ATP could be an important and integral part of the multifaceted regulation of AMPK by ATP.

In the presence of the P2 purinoceptor antagonists suramin and PPADS, the hemichannel-mediated suppression of AMPK was abrogated, indicating the involvement of purinergic signaling. Binding of ATP to P2 receptors activates several downstream signaling pathways. Here, it is most probable that signaling is mediated by the P2Y receptor signal transduction pathway. As G-protein-coupled receptors, P2Y family members activate phospholipase C (PLC) to hydrolyze phosphatidylinositol 4,5-bisphosphate to 1,2-diacylglycerol and inositol triphosphate, subsequently causing the release of intracellular Ca²⁺ and the activation of protein kinase C (PKC), as well as downstream signaling cascades such as those mediated by mitogen-activated protein kinase and PI3K (Franke et al., 2009; Thelen and Didichenko, 1997). In support of an involvement of this pathway, we observed that the PLC inhibitor U73122 potentiated, whereas PLC activator m-3M3FBS suppressed AMPK activation (supplementary material Fig. S1). Furthermore, previous studies have demonstrated an involvement of this pathway in the induction of the intracellular release of Ca²⁺ and the regulation of cell proliferation by hemichannel-derived ATP (Belliveau et al., 2006; Weissman et al., 2004). Reports have also implicated G-protein-coupled receptors in growth-factor-induced suppression of AMPK (Ning et al., 2011). Recently, more G-protein-coupled purinergic receptors have been reported to activate the PI3K signaling pathway. These receptors include G_q-coupled P2RY1, P2RY2, P2RY4, P2RY6 and P2RY11 and G_i-coupled P2RY12, P2RY13 and P2RY14 (Burnstock, 2006; Inoue, 2006). The possible implication of these receptors in the mediation of ATP-induced suppression of AMPK cannot be excluded. More detailed analysis of the expression profile and functional roles of these purinergic receptors in NRK cells is required in the future.

Our results identify Akt as a key molecule that mediates the suppressive effect of hemichannels and ATP on AMPK activation. This is shown by the fact that blockade of Akt or its upstream kinase potentiated AMPK activation, whereas overexpression of Akt suppressed it. The involvement of Akt in this study is not surprising because Akt has been reported to repress AMPK activities through several pathways (Ning et al., 2011). Akt regulates energy homeostasis by increasing glycolysis and oxidative phosphorylation, causing an increased level of intracellular ATP and a concomitant reduction in the AMP to ATP ratio (Hahn-Windgassen et al., 2005). Akt activates mTOR through phosphorylating TSC2 and inhibiting TSC1–TSC2 dimer formation (Hahn-Windgassen et al., 2005). Akt also suppresses AMPK activation by inhibiting AMPK kinases, such as LKB1 (Liu et al., 2012; Soltys et al., 2006). In addition, Akt affects AMPK activity by phosphorylating AMPK itself at serine residue 485 (Ning et al., 2011). Consistent with the previous reports, we observed that inhibition of Akt enhanced LKB1 phosphorylation but suppressed mTOR activities. In addition, transfection of cells with constitutively active Akt reduced the AMPK activation that was caused by the exposure of cells to several different stimuli, including low glucose, Hank's balanced salt solution (HBSS) and hypoxia (supplementary material Fig. S2). Thus, activation of the PI3K–Akt signaling pathway is responsible for the suppressive effects of extracellular ATP on AMPK activity.

The activity of AMPK is modulated both by the cellular AMP to ATP ratio and upstream kinases, such as LKB1 and CAMKKβ (also known as CAMKK2) (Hardie, 2011; Hurley et al., 2005; Mihaylova and Shaw, 2011); however, the question arises as to how AMPK is activated under Ca²⁺-free conditions. Prevention of the loss of ATP, by blocking hemichannels, enhanced AMPK activation, suggesting that AMPK activation is not due to the loss of ATP through hemichannels. It is also unlikely that the change in hemichannel activities affected the cellular AMP to ATP ratio because both AMP and ATP are small molecules that can freely pass through hemichannels. One recent report has indicated that Ca²⁺ is an indispensable factor for the maintenance of the normal function of mitochondria; removal of extracellular Ca²⁺ interferes with mitochondrial biogenesis and activates AMPK (Cárdenas et al., 2010; Green and Wang, 2010). In line with this report, we detected a time-dependent decrease in the total amount of ATP following removal of extracellular Ca²⁺ (supplementary material Fig. S3), which could be one of the mechanisms that contributes to the observed AMPK activation. We also found that Ca²⁺ deprivation was associated with an increase in LKB1 phosphorylation, the extent of which was equivalent to that of AMPK and subject to the regulation of purinergic signaling. Thus, activation of LKB1 could be crucially involved in AMPK activation. Consistent with this notion, we observed that exogenous ATP strongly suppressed metformin-induced AMPK activation (data not shown), which activates AMPK through a LKB1-dependent mechanism (Towler and Hardie, 2007). LKB1 was also activated by Cd²⁺, which was potentiated by hemichannel blockers (data not shown). CAMKKβ is less likely to be involved in our experimental setting because previous studies have shown that activation of CAMKKβ requires the presence of extracellular Ca²⁺ and that depletion of extracellular Ca²⁺ abolished AMPK activation induced by CAMKKβ activators (Chi et al., 2011; Ghislat et al., 2012).

ATP is released into the extracellular space in response to various stimuli through several different routes (Gordon, 1986; Lazarowski, 2012; Schwiebert and Zsembery, 2003). In addition to hemichannels, ATP is released by ATP-binding cassette proteins, exocytosis, anion channels, pannexin channels and the P2X7 receptor. It is conceivable that the observed effects of the released ATP on AMPK could be independent of the type of stimulant or route of secretion. In this context, our findings could have a wide application. It is especially noteworthy that channels composed of pannexin subunits have been reported to share many properties with connexin channels and have been shown to play a key role in many pathological situations (Qu et al., 2011; Silverman et al., 2009). However, a role for pannexin in this study is unlikely because NRK cells have been previously reported to be pannexin deficient (Penuela et al., 2007). Indeed, blockade of pannexin channels with a specific blocker, probenecid, did not affect AMPK activation (supplementary material Fig. S4).

Of note, the treatment of cells with hemichannel blockers or siRNA against Cx43 failed to completely abolish the release of ATP caused by Ca²⁺ deprivation. These observations imply that removal of extracellular Ca²⁺ also activates other ATP-releasing pathways. The exact properties of these pathways remain to be characterized. Intriguingly, the extracellular ATP that is released following cell exposure to Ca²⁺-free medium did not accumulate. Instead, it decreased in a time-dependent manner after reaching a peak within 5 min. Further analysis of hemichannel activity using ethidium bromide uptake revealed that the reduction of ATP was not associated with a corresponding decrease in the intracellular influx of ethidium bromide (supplementary material Fig. S5), suggesting that hemichannels are still active at these later timepoints. The decrease in the level of ATP could be attributed to the rapid degradation of ATP by extracellular nucleotidases (Komlosi et al., 2005).

It is also worth mentioning that both heptanol and lindane potentiated the low-Ca²⁺-initiated AMPK activation; however, the increase in AMPK phosphorylation did not appear to be associated with ATP release upon treatment with lindane. In contrast with heptanol, the timecourse over which the lindane-induced AMPK phosphorylation decreased did not correlate with that of the reduction in the release of ATP. In fact, there was a statistically significant difference between the potentiation of AMPK activation by heptanol and lindane at the 30-min timepoint. Additional experiments revealed that heptanol and lindane additively stimulated AMPK activation without further suppressing ATP release (supplementary material Fig. S6); thus it is probable that lindane potentiates AMPK activation through other mechanisms, additional to its inhibition of hemichannels. Heptanol and lindane are known to inhibit connexin channels through completely different mechanisms. Heptanol blocks connexin channel formation through decreasing the open probability of the channels (Takens-Kwak et al., 1992), whereas lindane works by regulating multiple intracellular signaling molecules, including PKC, ERK (also known as Mapk1), PLC, cyclic AMP, arachidonic acid and glutathione (Caruso et al., 2005; Criswell and Loch-Caruso, 1999; Krieger and Loch-Caruso, 2001; Mograbi et al., 2003; Wang and Loch-Caruso, 2002). The sustained activation of AMPK by lindane might be related to its actions on these signaling molecules. It might also be related to the reported actions of lindane on mitochondrial disorder and oxidative stress (Bagchi and Stohs, 1993; Benarbia et al., 2013), both of which are known to be able to activate AMPK (Cárdenas et al., 2010; Pung et al., 2013). This unexpected action of lindane on AMPK should not greatly affect our conclusion because a similar potentiating effect on AMPK activation was achieved by downregulation of Cx43 with siRNA or by using other hemichannel blockers.

The importance of our findings is further exemplified by the pivotal role of the signaling cascade in Cd²⁺-induced cell death. As a negative regulator of mTOR, AMPK has been previously reported to be able to inhibit Cd²⁺-induced generation of reactive oxygen species, mTOR activation and cell death (Chen et al., 2011). In this context, suppression of AMPK by hemichannel-initiated activation of purinergic signaling could be an important mechanism underlying Cd²⁺-induced cell death.

These data reveal a previously unrecognized hemichannel-mediated regulation of AMPK and implication of the mechanisms in the biological action of hemichannels. Our findings, thus, provide novel and important mechanistic insights into the functions of extracellular ATP, as well as channels that release ATP.

Normal rat kidney (NRK) renal epithelial cell line NRK-E52 and porcine kidney epithelial cell line LLC-PK1 cells were purchased from the American Type Culture Collection (Rockville, MD). Mesangial cells were cultured as previously described (Johno et al., 2012). Cells were maintained in Dulbecco's modified Eagle's medium (DMEM, Gibco-BRL, Gaithersburg, MD) supplemented with 5% fetal bovine serum (FBS). For comparison of cell responses between normal Ca²⁺ and Ca²⁺-free conditions, cells were exposed to Ca²⁺-free DMEM (Gibco-BRL, catalogue number 21068) with or without supplementation of 1.8 mM Ca²⁺.

FBS, trypsin with EDTA, antibiotics, cadmium chloride (CdCl₂), heptanol, lindane, Lucifer Yellow, ethidium bromide, lanthanum chloride (La³⁺), gadolinium chloride (Gd³⁺), suramin, pyridoxalphosphate-6-azophenyl-2′,4′-disulfonic acid (PPADS) and all other chemicals were obtained from Sigma (Tokyo, Japan). Antibodies against AMPK, Akt and β-actin were obtained from Cell Signaling (Beverly, MA).

The presence of functional hemichannels was evaluated by the cellular uptake of Lucifer Yellow or ethidium bromide, as described previously (Fang et al., 2011; Garré et al., 2010). NRK cells were exposed to Ca²⁺-free medium in the presence or absence of 0.1% Lucifer Yellow or 10 µM ethidium bromide for 15 min. The cells were then rinsed and fixed with 3% paraformaldehyde. Immunofluorescence images were captured using a CCD camera attached to an Olympus BX50 microscope.

Total cellular protein was extracted by suspending the pre-washed cells in sodium dodecyl sulphate (SDS) lysis buffer (62.5 mM Tris-HCl, 2% SDS, 10% glycerol) together with freshly added proteinase inhibitor cocktail (Nacalai tesque, Kyoto, Japan). Lysates were incubated on ice for 30 min with intermittent mixing and then centrifuged at 12,000 r.p.m. for 10 min at 4°C in an Eppendorf centrifuge 5415R with a F45-24-11 rotor. The supernatant was recovered and the protein concentration was determined using the Micro BCA Protein Assay Kit (Pierce, Rockford, IL).

Western blot was performed using the enhanced chemiluminescence system (Chi et al., 2011; Fang et al., 2011). Briefly, extracted cellular proteins were separated by 10% SDS polyacrylamide gels and electrotransferred onto polyvinylidine difluoride membranes. After blocking with 3% bovine serum albumin in PBS, the membranes were incubated with primary antibody for 1.5 h at room temperature or at 4°C overnight. After washing, the membranes were probed with horseradish-peroxidase-conjugated anti-rabbit IgG (Cell Signaling, Beverly, MA), and the bands were visualized by using the enhanced chemiluminescence system (Amersham Biosciences, Buckinghamshire, UK). The chemiluminescent signal was captured with a Fujifilm luminescent image LAS-1000 analyzer (Fujifilm, Tokyo, Japan) and quantified with the ImageJ software (http://rsb.info.nih.gov/ij). The results of quantification were expressed as optical density (OD). To confirm equal loading of proteins, the membranes were stripped with 62.5 mM Tris-HCl (pH 6.8) containing 2% SDS and 100 mM 2-mercaptoethanol for 30 min at 60°C and probed for total AMPK protein or β-actin.

NRK cells were transiently transfected with siRNA specifically targeting Cx43 (Mm_Gja1_2 HP siRNA, catalogue number SI00191625, Qiagen, Japan) or a negative control siRNA (AllStars Negative Control siRNA) at a final concentration of 20 nM using HiPerfect transfection reagent for 48 h (Chi et al., 2011; Fang et al., 2011). Cells were then either left untreated or exposed to Ca²⁺-free medium for the indicated intervals. Cellular lysates were harvested and assayed for Cx43 expression and AMPK phosphorylation using western blot analysis.

Vectors encoding wild-type Cx43 tagged with EGFP (wild-type Cx43-pEGFP1) or mutant Cx43 tagged with EGFP (mutant Cx43-pEGFP1) were a kind gift from Oyamada (Department of Pathology, Kyoto Prefectural University of Medicine, Kyoto, Japan). These vectors were transfected into LLC-PK1 cells by using Lipofectamine Plus reagent (Invitrogen, Carlsbad) according to the manufacturer's instructions (Fang et al., 2011; Li et al., 2013). Clones with high levels of GFP fluorescence were used for this study, which were selected by using the fluorescence microscopy. The mutant Cx43-pEGFP1 vector, which lacks 24 bases that correspond to amino acid residues 130 to 137 of rat Cx43 (Δ130–137 Cx43), was designed according to the initial report of Krutovskikh et al. (Krutovskikh et al., 1998; Oyamada et al., 2002). Previous studies have demonstrated that cells carrying the mutant vector cannot form gap junctions that are capable of intercellular communication (Krutovskikh et al., 1998; Oyamada et al., 2002). Stable transfection of mesangial cells with myristoylated Akt was performed, as previously reported (Johno et al., 2012). pcDNA3-myrHA-Akt1 encoding constitutively active Akt was kindly provided by Kenneth Walsh (Boston University School of Medicine, Boston, MA). Mock-transfected cells were also established by stable transfection with pcDNA3.1 (Invitrogen, Life Technologies, Carlsbad, CA).

Cytotoxicity was evaluated by the release of LDH using an LDH cytotoxicity detection kit (Takara Bio, Otsu, Shiga, Japan), as described previously (Fang et al., 2011). Briefly, cells in a 96-well culture plate were incubated with various chemicals for the described time intervals. Culture medium was collected and measured for LDH activity. LDH release was calculated and expressed as percentage of total release. Culture medium was used as a background control, whereas cells treated with 2% Triton X-100 were used as 100% release.

The cell viability was assessed by formazan assay using Cell Counting Kit-8 (Dojindo Laboratory, Kumamoto, Japan), as described previously (Fang et al., 2011).

ATP was measured using a luciferin and luciferase bioluminescence assay kit (Molecular Probes). The intensity of the chemiluminescent signal was determined by using a luminometer (Gene Light 55; Microtech Nition, Chiba, Japan), as described previously (Fang et al., 2011; Yao et al., 2003).

Values are expressed as mean±s.e.m. Comparison of two populations was performed by using Student's t-test. For multiple comparisons, one way analysis of variance (ANOVA), followed by Dunett's test, was employed. Both analyses were performed by using the SigmaStat statistical software (Jandel Scientific, CA). P<0.05 was considered to be a statistically significant difference.

We thank Hisashi Johno for providing the glomerular mesangial cells. We thank Masanori Kitamura and Hironori Kato (Department of Molecular Signaling, University of Yamanashi, Japan) for antibodies, technical support and advice on the experiments.