Regulator of calcineurin 1 (RCAN1) is a key regulator of the calcineurin–NFAT signaling network in organisms ranging from yeast to human, but its functional role is still under debate because different roles of RCAN1 have been suggested under various experimental conditions. To elucidate the mechanisms underlying the RCAN1 regulatory system, we used a systems approach by combining single-cell experimentation with in silico simulations. In particular, we found that the nuclear export of GSK3β, which switches on the facilitative role of RCAN1 in the calcineurin–NFAT signaling pathway, is promoted by PI3K signaling. Based on this, along with integrated information from previous experiments, we developed a mathematical model in which the functional role of RCAN1 changes in a dose-dependent manner: RCAN1 functions as an inhibitor when its levels are low, but as a facilitator when its levels are high. Furthermore, we identified a hidden incoherent regulation switch that mediates this role change, which entails negative regulation through RCAN1 binding to calcineurin and positive regulation through sequential phosphorylation of RCAN1.
The calcineurin–nuclear factor of activated T-cells (NFAT) signaling pathway has essential roles in many physiological processes, including T-cell activation, memory formation and apoptosis in neurons, and growth and differentiation in skeletal muscle and cardiac cells (Rusnak and Mertz, 2000; Shin et al., 2006; Shin et al., 2008; Tomida et al., 2003). In the calcineurin–NFAT signaling pathway, calcineurin is activated by Ca2+ binding to calmodulin. The activated calcineurin dephosphorylates NFAT, leading to its nuclear localization and subsequent activation of its target genes (Liang et al., 2003; Rusnak and Mertz, 2000; Vega et al., 2003a). Regulator of calcineurin 1 (RCAN1), a protein encoded by one of the target genes of NFAT, binds and inhibits calcineurin (Chan et al., 2005; Vega et al., 2002). However, recent studies have shown that RCAN1 facilitates calcineurin–NFAT signaling under various experimental conditions, and this has led to controversy regarding the functional role of RCAN1 (Fox and Heitman, 2005; Sanna et al., 2006; Vega et al., 2003b) (Table 1). For instance, Vega and colleagues suggested a dual role of RCAN1 in the development of cardiac hypertrophy (Vega et al., 2003b), whereas Sanna and co-workers insisted that RCAN1 functions only as a calcineurin facilitator and reasoned that previous experiments showing RCAN1 to be a calcineurin inhibitor might be due to sequestration effects resulting from an overexpression approach (Sanna et al., 2006). We note, however, that Hoeffer (Hoeffer et al., 2007) and Ryeom and colleagues (Ryeom et al., 2003) demonstrated an inhibitory role of RCAN1 in mouse knockout experiments.
In spite of such diverse and conflicting lines of evidence regarding the regulatory mechanism of calcineurin–NFAT signaling, there has been no attempt to explain the role change of RCAN1 from a system-level perspective by integrating all available experimental evidence. Therefore, we used a systems approach in the present study by combining single-cell experimentation and in silico simulations to elucidate why different roles of RCAN1 have been suggested depending on the experimental conditions. For in silico simulations, we developed a mathematical model based on our own experimentation, as well as integrated information from previous experiments, encompassing calcineurin–NFAT signaling, RCAN1 regulatory processes, crosstalk mediated through extracellular signal-regulated kinase (ERK) and phosphatidylinositol 3-kinase (PI3K) signaling pathways, and the glycogen synthase kinase 3β (GSK3β) translocation mechanism.
As a result, we found that PI3K signaling promotes the nuclear export of GSK3β. Moreover, from extensive in silico simulations, we found that the functional role of RCAN1 changes in a dose-dependent manner: RCAN1 functions as an inhibitor when its levels are low, but as a facilitator when its levels are high. Furthermore, we have identified a hidden incoherent regulation switch that coordinates such a role change via an intricate balance between positive (sequential phosphorylation of RCAN1) and negative (RCAN1 binding to calcineurin) regulation.
A mathematical model of the calcineurin–NFAT signaling network
We developed a mathematical model to describe the calcineurin–NFAT signaling network based on information from previous experiments, including our own. This model encompasses the calcineurin–NFAT signaling pathway, RCAN regulation module and the PI3K signaling pathway (Fig. 1). Below, we describe the major signaling processes that are considered in our model.
The calcineurin–NFAT signaling pathway is activated by an increase in the cytosolic Ca2+ level. Ca2+ binds to the catalytic subunit of calcineurin through calmodulin, which leads to dephosphorylation of cytosolic NFAT and promotion of its nuclear translocation (Rusnak and Mertz, 2000; Vega et al., 2003a). Although NFAT has several serine phosphorylation sites (Chow and Davis, 2000), in this model these were represented by a single, virtual phosphorylation site. Nuclear NFAT, in cooperation with GATA4 (Tokudome et al., 2005), activates transcription of various hypertrophy-related genes, including those encoding atrial natriuretic peptide (ANP), brain natriuretic peptide (BNP) and RCAN1; however, only RCAN1 was considered in this model because the other proteins do not regulate calcineurin–NFAT signaling. It should be noted that RCAN1 has several splice variants and, among these, RCAN1.1 and RCAN1.4 are most abundant in cardiac cells (Dey et al., 2007). Only RCAN1.4 is induced by NFAT (Dey et al., 2007). Thus, hereafter RCAN1 refers to RCAN1.4, unless otherwise mentioned. RCAN1 competitively binds to and inhibits calcineurin activity through the region encoded by exon 7, which results in negative feedback regulation (Chan et al., 2005). In this model, RCAN1 is considered to be a multi-phosphoprotein that is sequentially phosphorylated by ERKs and GSK3β (Hilioti et al., 2004; Vega et al., 2002). The functional impact of RCAN1 on induction of nuclear NFAT varies depending on the phosphorylation state of RCAN1 (Abbasi et al., 2006). For instance, Abbasi and colleagues showed that RCAN1 dissociates from calcineurin if RCAN1 is phosphorylated at Ser112 by ERK5, and that calcineurin–NFAT signaling is then re-established (Abbasi et al., 2006). However, the effect of RCAN1 phosphorylation by ERK1 and ERK2 (ERK1/2) on calcineurin–NFAT signaling has not yet been described. From our experiments, we confirmed that the MEK1 and MEK2 (MEK1/2) inhibitor U0126 also suppresses the nuclear localization of NFAT in H9C2 cells, after treatment of the cells with the Ca2+ activator ionomycin (supplementary material Fig. S1A,B). Taken together, these results imply that the effects of ERK1/2 and ERK5 are functionally the same. Thus, in this model we denote ERK1/2 and ERK5 by ERK, without any discrimination. RCAN1 phosphorylated at Ser112 by ERK is further phosphorylated at Ser108 by GSK3β, but calcineurin selectively removes the phosphate at Ser108 of RCAN1 (Vega et al., 2002).
GSK3β can be localized to the cytosol or nucleus (Haq et al., 2000; Morisco et al., 2001), but its functional effect on calcineurin–NFAT signaling changes depending on its localization. In the cytosol, GSK3β can re-establish calcineurin–NFAT signaling through the sequential phosphorylation of RCAN1 in cooperation with ERK, but in the nucleus, calcineurin–NFAT signaling is inhibited through NFAT phosphorylation by GSK3β. In addition, NFAT phosphorylated by GSK3β in the cytosol binds to 14-3-3 protein, which enhances inhibition of the nuclear localization of NFAT (Chow and Davis, 2000; Du et al., 2005; Liao et al., 2005). In this model, we have simplified the sequential multi-step processes of NFAT translocation mediated by GSK3β to a single, semi-mechanistic step. Although localization of GSK3β has an important role in calcineurin–NFAT signaling, the underlying regulatory mechanism is not fully understood. In this study, we show that the nuclear export of GSK3β is promoted by the PI3K signaling pathway (see the following section for details), and this finding was incorporated into the model.
The mathematical model comprises 24 ordinary differential equations and 57 kinetic parameters (supplementary material Tables S1–S7). The kinetic parameter values were estimated using both our own experimental data and those previously obtained by others, as a training data set. The kinetic parameter values of the calcineurin–NFAT pathway were estimated based on the time course data of nuclear NFAT induction after platelet-derived growth factor (PDGF) treatment of H9C2 cells (supplementary material Fig. S1C). The transcriptional time delay – the time taken for Rcan1 induction up to the level of statistical significance (P<0.05) after ionomycin (IM) treatment – was estimated to be about 30 minutes based on real-time PCR data (supplementary material Fig. S1D,E). The kinetic parameters of the ERK and PI3K signaling pathways were fitted to previous experimental data (Aoki et al., 2000; Clerk et al., 2006; Markou et al., 2008; Morisco et al., 2008; Morisco et al., 2000; Morris et al., 2005; Tian et al., 2003; Yin et al., 2006), as well as to our own data (supplementary material Fig. S2 and Fig. S3A,B). We note that the time profiles of phosphorylated ERK, Akt and GSK3β are quite different depending on the hypertrophic stimulants used, such as PDGF, angiotensin II (AngII), phenylephrine (PE), isoproterenol (ISO) and endothelin-1 (ET-1) (supplementary material Fig. S2), because these stimulants all have different input points (i.e. receptors) and intercellular mediators through which the signals pass to the ERK and PI3K signaling pathways. In the model, we did not take into account all the detailed variations from the input to the major signaling pathways involving the various mediators that differed according to the hypertrophic stimulant used. Instead, we constructed a common minimal model of the major ERK and PI3K signaling pathways and estimated the condition-specific kinetic parameter values based on the experimental data obtained for each hypertrophic stimulant. The kinetic parameters of the RCAN1 regulation module were fitted to the experimental data obtained from rat neonatal cardiac myocytes (Abbasi et al., 2005) (supplementary material Fig. S3C,D). We confirmed that all the simulation results were quantitatively consistent with the training data set (supplementary material Fig. S1C–E, Fig. S2 and Fig. S3C–E).
To validate the responses of the calcineurin–NFAT signaling network to typical hypertrophic stimulants such as norepinephrine (NE), AngII, ET-1 and PDGF, we determined the induction levels of cytosolic Ca2+ and NFAT translocation to the nucleus using H9C2 cells, where IM was used as the positive control. All the examined hypertrophic stimulants increased the Ca2+ levels, which were measured using the Ca2+-chelating dye Fluo-4 (Fig. 1B,C). To investigate the dynamics of NFAT translocation to the nucleus, the cells were transfected with NFATc1–YFP. After treatment with the various hypertrophic stimulants, NFATc1–YFP was found to be imported to the nucleus (Fig. 1B,D and supplementary material Movie 1). The translocation of NFATc1–YFP corresponded to the increasing cytosolic Ca2+ levels. Taken together, these results confirm that in H9C2 cells, these typical hypertrophic stimulants induce an increase in cytosolic Ca2+ levels and thereby nuclear NFAT localization.
PI3K induces the nuclear export of GSK3β
The translocation mechanism of GSK3β in cardiac cells is not yet fully understood. Several conflicting experimental results have been reported: for example, ISO induces nuclear export of GSK3β (Morisco et al., 2001), whereas ET-1 induces pronounced translocation of GSK3β to the nucleus (Haq et al., 2000). Because GSK3β is regulated by PI3K signaling and is exported from the nucleus to the cytosol by nuclear export signal (NES)-containing proteins such as FRAT1 and Axin2 (Franca-Koh et al., 2002; Yook et al., 2006), we hypothesize that the localization of GSK3β is regulated by PI3K signaling and NES-containing proteins. To investigate this possibility, cells were stained with GSK3β antibody and transfected with YFP–GSK3β and YFP–GSK3βS9A (a form insensitive to PI3K). Without serum deprivation, both endogenous and overexpressed GSK3β were localized to the cytosol, whereas endogenous GSK3β was accumulated in the nucleus following pre-incubation for 1 hour with the PI3K-specific inhibitor LY294002 (LY29) or the Crm1-specific inhibitor leptomycin B (LMB) (Fig. 2A,B).
To investigate whether translocation of GSK3β is directly regulated by PI3K through phosphorylation of GSK3β at Ser9, cells co-expressing constitutively active PI3K (p110-CAAX), the PH domain of Akt1 (PIP3 biosensor, PHAkt1), and YFP–GSK3β or YFP–GSK3βS9A were treated with LY29 or LMB (Fig. 2C–E). Interestingly, both YFP–GSK3β and YFP–GSK3βS9A were translocated to the nucleus by LY29, indicating that PI3K signaling regulates GSK3β translocation and that the phosphorylation of GSK3β at Ser9 is not essential for this translocation (Fig. 2C,D and supplementary material Movies 2 and 3).
We then confirmed our results using the opposite approach. After serum deprivation, PI3K signaling was inactivated (supplementary material Fig. S4A). Under these conditions, GSK3b was distributed evenly throughout the cells and nuclear GSK3β was exported following treatment with PDGF, a well-known PI3K activator (supplementary material Fig. S4B,C). After preincubation with LY29 for 1 hour, overexpressed YFP–GSK3β or YFP–GSK3βS9A were evenly distributed in the cells coexpressing p110-CAAX and PHAkt1. After LY29 wash out, nuclear YFP–GSK3β and YFP–GSK3βS9A were translocated to the cytosol (supplementary material Fig. S4D), confirming that export of nuclear GSK3β is mediated by PI3K signaling.
The functional role of RCAN1 changes in a dose-dependent manner
To quantitatively analyze the effect of various hypertrophic stimulants on the calcineurin–NFAT signaling network, we simulated nuclear NFAT induction levels, phosphorylated ERK levels and cytosolic GSK3β levels after treatment with AngII, PE, ET-1, PDGF and ISO, respectively. Nuclear NFAT levels were increased prominently following PDGF and AngII treatment (Fig. 3A), and phosphorylated ERK levels were markedly increased by PDGF treatment. Cytosolic GSK3β levels were increased remarkably by AngII, PE and ISO treatment. It can be seen, however, that PE failed to increase phosphorylated ERK levels, and that ET-1 did not increase cytosolic GSK3β levels, which is related to the low level of induction of nuclear NFAT by PE and ET-1. Taken together, these results suggest that the two crosstalk signal pathways mediated by ERK and GSK3β are simultaneously required for full activation of calcineurin–NFAT signaling.
In this model, the calcineurin–NFAT signaling pathway and the RCAN1 regulation module together constitute a core circuit of the signaling network that is most important in merging and processing the hypertrophic signals mediated by cytosolic Ca2+ levels and the ERK and PI3K pathways. Thus, our analysis focuses on the functional role of signaling components of the core circuit in manipulation of responses to hypertrophic stimulants. We have simulated the output profiles of phosphorylated RCAN1 (at Ser108) and nuclear NFAT by changing the kinetic parameter values over a wide range. Of the 19 total kinetic parameters, eight parameters (P7–P10, P12–P14 and P19) showed nonlinear (‘curved’) steady-state trajectories of the outputs in the phase plane (supplementary material Fig. S5). In particular, the fold increase of the four parameters (P7–P10) related to RCAN1 production induced a biphasic response in nuclear NFAT induction (supplementary material Fig. S6), which suggests that depending on its expression level, RCAN1 might have different roles in the regulation of calcineurin–NFAT signaling.
To further investigate the functional role of RCAN1, we simulated the temporal concentration profiles of nuclear NFAT relative to the fold increase of RCAN1. The early responses (within about 200 minutes) of nuclear NFAT after AngII treatment showed a monotonic decrease, but the late responses showed a biphasic curve where the nuclear NFAT increases beyond a certain threshold level of RCAN1 (Fig. 3B). These results further suggest that the role of RCAN1 changes in a dose-dependent manner. To examine whether other hypertrophic stimulants also generate the same result, we simulated the induction levels of nuclear NFAT by gradually increasing the concentration of PE, PDGF, ISO and ET-1, respectively. Under these conditions, the biphasic response of nuclear NFAT was observed only after stimulation with PDGF or ISO, and not after stimulation with PE or ET-1 (Fig. 3C, left). Together with the preceding result indicating that full activation of nuclear NFAT requires crosstalk signals from both ERK and GSK3β (Fig. 3A), this suggests that such crosstalk signals are also related to the induction of the biphasic response of nuclear NFAT. To investigate this possibility, we simulated the nuclear NFAT level changes after site mutation at Ser112 (phosphorylated by ERK) and at Ser108 (phosphorylated by GSK3β), respectively. Site mutation of RCAN1 at Ser112 or Ser108 abolished the increase in nuclear NFAT levels observed at high RCAN1 concentrations (Fig. 3C, right).
To confirm the biphasic response of the nuclear NFAT, we transfected RCAN1 transiently and measured the expression level of RCAN1 through the intensity of CFP-conjugated RCAN1. Hoechst 33342 (nucleus staining in live cells) was used for normalization. The cells expressing NFAT–YFP and RCAN1 were randomly selected and tracked for 30 minutes after ionomycin treatment. NFAT was clearly translocated to the nucleus in the cells that do not express RCAN1 but the NFAT translocation was blocked in the cells that weakly express RCAN1 after ionomycin treatment. However, the cells that highly express RCAN1 showed the nuclear localization of NFAT after ionomycin treatment (upper panel of Fig. 3D and supplementary material Fig. S7). To ensure consistency of simulations under steady-state conditions, we measured the nuclear NFAT level for 30 minutes after ionomycin treatment (we confirmed that the time trajectories of the nuclear NFAT were stabilized after 30 minutes of ionomycin treatment). We found that the experimental results are consistent with the simulation results shown in Fig. 3C, which confirms the biphasic response of the nuclear NFAT. To validate the crosstalk effect from ERK and GSK3β, we investigated the induction of the nuclear NFAT by a point mutation of RCAN1. The nuclear NFAT level was prominently increased in the cells expressing NFATc1–YFP and constitutively active calcineurin (ΔCnA) as shown in Fig. 3D (bottom panel). However, when the cells were further co-transfected with CFP–RCAN1S108/112A in which both ERK and GSK3β phosphorylation sites were mutated to Ala, nuclear NFAT was exported to the cytosol. This is consistent with the simulation result (Fig. 3C, right). To confirm the facilitative role of RCAN1 in the nuclear NFAT induction, we further co-transfected the cells with CFP–RCAN1S108/112D in which both ERK and GSK3β phosphorylation sites were mutated to Asp. A large amount, if not all, of the cytosolic NFAT was imported to the nucleus (note that such a mutation effect is always less prominent than biochemical phosphorylation). To further confirm the facilitative mechanism of RCAN1 using hypertrophic ligands, we increased the expression level of RCAN1 through incubation of cells with ionomycin for 1 hour (we confirmed that RCAN1 expression level was elevated within 30 minutes and sustained for 1 hour). After ionomycin was washed out, the cells were treated with PDGF and NE, respectively. The nuclear NFAT levels were increased in cells treated with PDGF, but not in the cells treated with NE (supplementary material Fig. S8). This is because PDGF effectively activates ERK and cytosolic GSK3β whereas NE activates only GSKβ, as shown in Fig. 3A. Taken together with the simulation results (Fig. 3C, right), these suggest that both ERK and GSK3β are simultaneously required to induce the biphasic response of the nuclear NFAT.
To compare the relative effects of these two crosstalk signals, we further investigated the induction levels of nuclear NFAT relative to the fold change of RCAN1 for each crosstalk signal (Fig. 3E). The nuclear NFAT levels for a fixed high concentration of RCAN1 increased along with the fold increase of ERK or PI3K, but the threshold at which nuclear NFAT levels started to increase was much higher for ERK than for PI3K. Such different threshold levels might arise because of the distinct roles of ERK and PI3K in calcineurin–NFAT signaling: ERK re-establishes calcineurin activity by releasing RCAN1 from the calcineurin complex, whereas GSK3β exported to the cytosol by PI3K directly facilitates calcineurin–NFAT signaling in cooperation with 14-3-3 (Abbasi et al., 2006).
The incoherent regulation switch coordinates the role change of RCAN1
In the above sections, the functional role of RCAN1 was found to switch from inhibition to facilitation of calcineurin–NFAT signaling in a dose-dependent manner. The question then arises as to whether there is a core circuit in the calcineurin–NFAT signaling network that coordinates such a role change of RCAN1. To answer this question, we further simplified the signaling network model by considering only three major signaling proteins that are directly related to RCAN1: calcineurin, NFAT and RCAN1. In this simplified model, we assumed that the crosstalk signals mediated through ERK and GSK3β are constitutively active and that there is no transcriptional time delay for RCAN1 (supplementary material Table S4). To validate whether the response profiles of the simplified model are similar enough to those of the full model, we compared the response curves of the nuclear NFAT and the active calcineurin along with the gradual increase of RCAN1 and confirmed that the simplified model represents the dynamic behavior of the full model in a consistent way (supplementary material Fig. S9). By dissecting this simplified model, we identified a central incoherent regulation circuit where the nonphosphorylated RCAN1 binds to calcineurin and inhibits its activity (negative regulation), whereas the sequentially phosphorylated RCAN1 facilitates calcineurin–NFAT signaling (positive regulation), as shown on the left in Fig. 4A. This incoherent regulation circuit conceptually functions as a regulation switch in which the crosstalk signals mediated through ERK and GSK3β divert the negative regulation by RCAN1 to a positive regulation (Fig. 4A, right). The incoherent regulation switch is structurally similar to the incoherent feed-forward loop that is widely known to induce a biphasic response (Kim et al., 2008; Mangan et al., 2006). This provides us with a hint on the hidden regulatory mechanism underlying the role change of RCAN1 in calcineurin–NFAT signaling.
The next question that arises concerns which particular molecular interactions dominantly contribute to the induction of the biphasic response of nuclear NFAT. To address this question, we generated 10,000 sets of random parameter values selected from a log-uniform distribution over a range of the order of 102 and simulated the corresponding profiles of the nuclear NFAT activity with respect to the fold increase of RCAN1 (supplementary material Table S4). These profiles were then clustered into two groups: ‘monotonic increase’ and ‘concave-up’. By comparing the parameter values of these two profile groups, we found that the sequential phosphorylation of RCAN1 (k6 and k7) and the inhibition of calcineurin by RCAN1 binding (k3) primarily contribute to induction of the biphasic response (Fig. 4B). Taken together, these results show that the biphasic response of the nuclear NFAT is implemented via an intricate balance between the positive and negative regulation pathways in the core regulation switch.
The functional role of RCAN1 has been under debate since it became clear that its function changes depending on experimental conditions. In this study, we have unraveled the mechanisms and principles underlying regulation by RCAN1 using a combined approach of single-cell experimentation and in silico simulations. In particular, we revealed that PI3K signaling prompts the nuclear export of GSK3β, a translocation mechanism that had been unclear in cardiac cells. Based on this, together with integrated information from previous experiments, we developed a mathematical model of the calcineurin–NFAT signaling network. Through extensive in silico simulations using this model, we found that RCAN1 has different roles in a dose-dependent manner: as an inhibitor of calcineurin–NFAT signaling when RCAN1 levels are low, and as a facilitator when RCAN1 levels are high. Furthermore, we have identified an incoherent regulation switch that coordinates this role change of RCAN1.
Our finding that a gradual increase in RCAN1 levels induces a biphasic response (‘concave-up’) of nuclear NFAT elucidates why the role of RCAN1 changes under different experimental conditions. Let us consider the case in which the induction level of RCAN1 lies below a critical value (Rc) under a certain set of biological conditions (Fig. 4C, left). In this case, nuclear NFAT levels would decrease along with the increase in RCAN1. If RCAN1 levels were suddenly reduced using knockout methods, nuclear NFAT levels would increase, leading to the interpretation that RCAN1 is inhibitory under this condition. However, if RCAN1 levels were strongly increased over the critical value (RCAN1>>Rc) using overexpression methods, nuclear NFAT contents would also be increased, leading to the interpretation that RCAN1 is a facilitator of NFAT levels under this condition (Fig. 4C, left). If the initial induction level of RCAN1 were already beyond the critical value, nuclear NFAT levels would increase along with those of RCAN1 (Fig. 4C, middle). In this case, nuclear NFAT levels would decrease or increase in knockout or overexpression models of RCAN1, suggesting that RCAN1 has a consistent role in facilitating NFAT levels. However, if the crosstalk signals mediated by ERK and GSK3β were absent, the biphasic response of nuclear NFAT would disappear, and nuclear NFAT levels would monotonically decrease along with those of RCAN1, suggesting an inhibitory role of RCAN1 (Fig. 4C, right).
Up to now, many attempts have been made to explain the contentious role of RCAN1 in calcineurin–NFAT signaling. For instance, Molkentin and colleagues reported that TAK1 phosphorylates RCAN1 at Ser94 and Ser136, and that this converts RCAN1 from an inhibitor to a facilitator of calcineurin–NFAT signaling (Liu et al., 2009); however, the molecular mechanisms responsible for this facilitation were not fully addressed. Kishi and co-workers reported that phosphorylation of yeast RCAN1 (Rcn1) triggers its degradation through the SCFCdc4 ubiquitin ligase complex, which mitigates inhibition of calcineurin by Rcn1 (Tokudome et al., 2005). Cunningham and colleagues reported through structure–function analysis that the inhibitory effect of RCAN1 requires a degenerate PxIxIT-like motif and an LxxP motif, whereas its facilitatory effect requires a TxxP motif, an ExxP motif, the GSK3β phosphorylation site SCFCdc4, and the proteasome (Mehta et al., 2009). In this study, we have shown that the change in the role of RCAN1 is coordinated by the incoherent regulation switch of the calcineurin–NFAT signaling network. However, the detailed mechanism by which calcineurin–NFAT signaling is facilitated after phosphorylation of RCAN1 by GSK3β remains unclear. In the mathematical model, we highlighted only the fact that the sequentially phosphorylated RCAN1 directly contributes to calcineurin dephosphorylation of NFAT, as reported in a previous experiment (Abbasi et al., 2006). By contrast, other authors (Mehta et al., 2009; Tokudome et al., 2005) suggested that calcineurin–NFAT signaling increases through degradation of RCAN1 by SCFCdc4 ubiquitin ligase after GSK3β phosphorylation of RCAN1. However, this can explain only the re-establishment (‘restoration’) of calcineurin–NFAT signaling and not the facilitation (‘increase’). In addressing this problem, we might consider the 14-3-3 protein, which has been reported to bind to over 200 substrates including NFAT, and also to recognize specifically the phosphoserine/threonine-containing sequence motifs in target proteins (Liao et al., 2005). In particular, 14-3-3 inhibits the nuclear import of NFAT by binding to phosphorylated NFAT (Chow and Davis, 2000; Liao et al., 2005). Phosphorylated RCAN1 is another target substrate of 14-3-3, and therefore, if phosphorylated RCAN1 levels dramatically increase, the levels of the complex of NFAT and 14-3-3 would also decrease. This means that highly increased phosphorylated RCAN1 levels might result in competitive inhibition of the binding between NFAT and 14-3-3, which would lead to facilitation of the nuclear import of NFAT. This hypothesis could explain the facilitative role of RCAN1 when its levels are high, as shown in our simulations.
The incoherent regulation switch that coordinates the role change of RCAN1 is structurally similar to the incoherent feed-forward loop, which is one of the most common network motifs in transcription regulatory networks and protein-protein interaction networks (Kim et al., 2008; Mangan and Alon, 2003). This network motif is known to induce a fast activation followed by a delayed inhibition of the output response to a given input stimulation (Mangan and Alon, 2003), and thereby results in time-dependent biphasic responses (Kim et al., 2008). The incoherent feed-forward regulation can also induce dose-dependent biphasic responses (Kim et al., 2008), and such response profiles can be further shaped as either ‘concave-up’ or ‘concave-down’ depending on the intricate balance between the negative and positive regulations that comprise the incoherent feed-forward circuit. For instance, if the positive regulation is stronger than the negative regulation for a low-level stimulation, then a concave-down-shaped curve will be generated; otherwise, a concave-up-shaped curve will be generated. In this study, we found that the incoherent regulation switch can also generate such a biphasic response. This means that a biological system containing an incoherent regulation switch can induce apparently opposite responses depending on biological conditions, which explains why different roles of RCAN1 in calcineurin–NFAT signaling have been suggested under various experimental conditions.
Materials and Methods
Endothelin-1 was purchased from the Peptide Institute, PDGF-BB was purchased from PeproTech and leptomycin B and norepinephrine were purchased from Sigma. Ionomycin (Sigma), LY294002 (LC Laboratories) and U0126 (Promega) were dissolved in dimethyl sulphoxide (DMSO; Sigma) and used at the indicated concentrations.
GSK3β and GSK3βS9A were generated using RT-PCR cloning with gene-specific primers containing attB sequences to create entry clones for the Gateway cloning system® (Invitrogen). Expression constructs were made using LR clonase (Invitrogen) to transfer the coding regions into YFP expression vectors. The NFATc1–YFP construct was made by exchange of the sequence encoding fluorescent protein of NFATc1–GFP, which was kindly provided by Colleen M. Sheridan (Sheridan et al., 2002). CFP and GFP-PHAKT1 were gifts from Tobias Myer (Stanford University, Palo Alto, CA). RCAN1 and truncation of Calcineurin (1–398) were amplified by PCR and subcloned into CFP-C1 and mCherry-C1 vectors (Clontech), respectively. RCAN1S108/112A and RCAN1S108/112D were generated using the QuikChange site-directed mutagenesis kit (Stratagene).
Cell culture and electroporation
H9C2 cells were purchased from the American Type Culture Collection (ATCC). Cells were cultured in Dulbecco's modified Eagles medium (DMEM; Invitrogen) supplemented with 10% FBS, 2 mM L-glutamine, and penicillin-streptomycin at 37°C in 10% CO2. Electroporation was performed using a Microporator™ (MP-100; Digital Bio Technology), following the manufacturer's instructions. For cell imaging, shocked cells were aliquotted into 96-well black plates with glass-bottoms (Metrical) and, after 24 hours, cells were used for imaging.
Imaging and microscopes
Before imaging, the medium was replaced with phosphate-buffered saline (PBS) containing glucose (1 g/l, Invitrogen). All images were taken using a confocal microscope (A1R, Nikon) with a CFI plan Apochromat VC objective lens (60×/1.40 oil) at a resolution of 512×512. Images were analyzed using MetaMorph software (Molecular Devices).
Measurement of cytosolic Ca2+ levels
H9C2 cells were incubated with serum-free DMEM containing 1 μM Fluo-4 and AM ester (Invitrogen) at 37°C. Cells were incubated for 1 hour and washed twice with PBS before imaging.
For analysis of endogenous GSK-3β localization, cells were exposed to LY29 and LMB for 1 hour without serum starvation, or PDGF for 10 minutes with serum starvation for 6 hours, in serum-free DMEM media. Cells were fixed in 4% paraformaldehyde for 15 minutes, rinsed in Dulbecco's phosphate-buffered saline solution (DPBS; Invitrogen), permeabilized in ice-cold 100% methanol for 10 minutes, rinsed three times in DPBS, and blocked in DPBS containing 0.3% Triton X-100 (Sigma) and 5% fetal bovine serum (FBS) for 1 hour at room temperature. Primary antibody solutions were made up at a 1:100 dilution in DPBS containing 0.3% Triton X-100, and cells were incubated overnight at 4°C. Cells were rinsed three times for 5 minutes each in DPBS, and then incubated for 1 hour at room temperature in the appropriate secondary antibody at a 1:1000 dilution in blocking solution. After washing three times in DPBS for 5 minutes per wash, cells were incubated briefly in 4′-6-diamidino-2-phenylindole (DAPI; Invitrogen). Fixed cells were stored at 4°C for up to 1 week. Primary antibodies were anti-GSK3β (BD Transduction Laboratories) and anti-phosphorylated GSK3β (Ser9, Cell Signaling Technology). Alexa Fluor secondary antibodies from Molecular Probes were used in all cases.
RNA extraction and real-time quantitative PCR
After harvesting trypsinized H9C2 cells, total RNA was extracted using an easy-BLUE RNA extraction kit (iNtRON) and complementary DNA was synthesized with a cDNA synthesis kit (Invitrogen). Real-time quantitative PCR was performed using the IQ5 real-time PCR detection system (BioRad) with 2× prime Q-master mix (Genet Bio, dye SybrGreen I). The primers used were (5′→3′) RCAN forward, GCCCAATCCAGACAAACAGT and reverse, TCCTCTTCCTCCTCCTCCTC; and internal control GAPDH forward, CTCATGACCACAGTCCATGC and reverse, TTCAGCTCTGGGATGACCTT. PCR product quality was monitored with post-PCR melt curve analysis. Fold inductions were calculated using the comparative Ct (cycle threshold) method (also known as ΔΔCt method) (Livak and Schmittgen, 2001).
Mathematical modeling and analysis
The ordinary differential equations for the mathematical model of the calcineurin–NFAT signaling network and the simplified model are presented in supplementary material Tables S1–S4. The numerical integration was performed using Matlab R2009a software. The kinetic parameters were fitted to the previous experimental data as well as to our own experimental data obtained from H9C2 cells. The parameter estimates are summarized in supplementary material Tables S5–S7. To examine the robustness of our mathematical model to parameter variations, we carried out extensive simulations repetitively (n=50) over 50% random variation of parameter values selected from a log-uniform distribution and confirmed that the responses of the nuclear NFAT and the active calcineurin of the full and simplified models with respect to the gradual increase of RCAN1 are well preserved to such parameter variations (supplementary material Fig. S9).
The authors are grateful to Ki-Sun Kwon for providing the GSK3β and GSK3βS9A constructs, and thank Je-Hoon Song for helpful discussion. This work was supported by the National Research Foundation of Korea (NRF) grants funded by the Korea Government, the Ministry of Education, Science & Technology (MEST) (2009-0086964, 2010-0017662, 2010-0002160 and 2010-0000617).