Calcium homeostasis and signaling in yeast cells and cardiac myocytes

Abstract

Calcium ions are the most ubiquitous and versatile signaling molecules in eukaryotic cells. Calcium homeostasis and signaling systems are crucial for both the normal growth of the budding yeast Saccharomyces cerevisiae and the intricate working of the mammalian heart. In this paper, we make a detailed comparison between the calcium homeostasis/signaling networks in yeast cells and those in mammalian cardiac myocytes. This comparison covers not only the components, structure and function of the networks but also includes existing knowledge on the measured and simulated network dynamics using mathematical models. Surprisingly, most of the factors known in the yeast calcium homeostasis/signaling network are conserved and operate similarly in mammalian cells, including cardiac myocytes. Moreover, the budding yeast S. cerevisiae is a simple organism that affords powerful genetic and genomic tools. Thus, exploring and understanding the calcium homeostasis/signaling system in yeast can provide a shortcut to help understand calcium homeostasis/signaling systems in mammalian cardiac myocytes. In turn, this knowledge can be used to help treat relevant human diseases such as pathological cardiac hypertrophy and heart failure.

calcium homeostasis, calcium signaling, cardiac myocytes, cardiac hypertrophy, calcineurin

Introduction

In eukaryotic cells, the calcium ion (Ca²⁺) functions as a ubiquitous intracellular messengers regulating a myriad of biological processes such as cell proliferation, muscle contraction, programmed cell death, etc. (Berridge, 1998, 2000, 2003; Putney, 2005). Calcium homeostasis systems are highly regulated metabolic pathways used by cells to maintain Ca²⁺ within optimal concentration ranges in the cytosol and other organelles. For example, in normal-growing yeast (Saccharomyces cerevisiae), cytosolic Ca²⁺ is maintained in the range of 50–200 nM, in the presence of environmental Ca²⁺ concentrations ranging from <1 μM to >100 mM, through the functioning of an elaborate calcium homeostasis system as shown in Fig. 1a (Bonilla & Cunningham, 2002; Cui & Kaandorp, 2006; Cui, 2009). Extracellular stimuli cause a sudden increase in the cytosolic Ca²⁺ level, activating the effector proteins to exert a cellular response.

1

(a) Calcium homeostasis system in Saccharomyces cerevisiae. CaM, calmodulin; CaN,calcineurin; ER, endoplasmic reticulum. (b) Calcium homeostasis system in mammalian cardiac myocytes (this graph is modified after Fig. 1 in Bers, 2002). The schematic inset at the down-right corner shows the action potential (solid line) and the bulk cytosolic Ca²⁺ transient (dashed line) (this inset is modified after fig. 1c in Shannon, 2004). LTCC, L-Type Ca²⁺ channel; ATPase (ATP); NCX, Na⁺/Ca²⁺ exchanger; SR, sarcoplasmic reticulum; SERCA, SR Ca²⁺-ATPase; PLB, phospholamban; RyR, ryanodine receptor; mRyR, mitochondrial ryanodine receptor; UP, mitochondrial uniporter; RaM, rapid-mode uptake pathway.

Extracellular Ca²⁺ normally enters the cytosol of yeast cells through an unknown transporter X (Locke, 2000). Cytosolic Ca²⁺ is then transported into large vacuoles through a vacuolar calcium ATPase Pmc1 and a vacuolar Ca²⁺/H⁺ exchanger Vcx1. A calcium ATPase termed Pmr1 is responsible for pumping cytosolic Ca²⁺ into the endoplasmic reticulum and Golgi, and excessive Ca²⁺ in the secretory pathway is then extruded from the cell via exocytosis. Another transporter on the plasma membrane, transporter M, has been detected recently. This transporter transports Ca²⁺ into the yeast cell in response to very high extracellular Ca²⁺ (Cui, 2009). A plasma membrane voltage-gated Ca²⁺ channel, composed of Cch1 and Mid1, becomes activated only under abnormal conditions such as membrane depolarization, depletion of secretory Ca²⁺ (Bonilla & Cunningham, 2003), pheromone stimulation (Muller, 2001; Zhang, 2006) and hypotonic shock (Batiza, 1996). In response to extracellular hypertonic shock, vacuolar Ca²⁺ is released into the cytosol through Yvc1, a distant homolog of mammalian TRPC-type Ca²⁺ channels (Denis & Cyert, 2002). In response to sudden increases in extracellular Ca²⁺, or the sudden removal of extracellular Mg²⁺ (Wiesenberger, 2007), the universal Ca²⁺ sensor protein calmodulin can bind and activate the protein phosphatase calcineurin. In turn, this inhibits the function of Vcx1 and induces the expression of Pmc1 and Pmr1 via activation of the Crz1 transcription factor and its subsequent translocation into the nucleus (Stathopoulos-Gerontides, 1999). In addition to these slow calcineurin-dependent feedback networks, rapid cytosolic Ca²⁺-dependent feedback also inhibits the activities of transporters M and X. However, the details of the constituents and the mechanisms of these rapid feedback inhibition pathways remain to be identified (Cui, 2009). The activity of calcineurin is regulated by Rcn1, whose function is regulated by Mck1, a member of the GSK-3 family of protein kinases (Hilioti, 2004).

In contrast to the calcium homeostasis process in yeast, calcium homeostasis in cardiac myocytes is inseparable from the calcium-signaling process involved in excitation–contraction coupling, the process that links electrical excitation of the myocytes to contraction of the whole heart (Barry & Bridge, 1993; Bers, 2002). Cardiac myocytes must achieve a resting cytosolic Ca²⁺ concentration of <200 nM if the contractile elements in the cells are to relax. As shown in Fig. 1b, when the cardiac myocyte is in a resting relaxed state, extracellular calcium enters the cytosol through a Ca²⁺ leak channel, while the sarcolemmal Ca²⁺-ATPase (PMCA) and a sarcolemmal Na⁺–Ca²⁺ exchanger (NCX) are responsible for cellular Ca²⁺ efflux. During the cardiac action potential, Ca²⁺ enters the cell through L-type Ca²⁺ channels (LTCC), which are located primarily at sarcolemmal–sarcoplasmic reticulum (SR) junctions, where the SR Ca²⁺ release channels [i.e. ryanodine receptors (RyRs)] exist. In addition, a small amount of extracellular Ca²⁺ will enter the cytosol through the NCX (normally, this exchanger functions to pump calcium out of the cell, but it can also function in the reverse direction). These Ca²⁺ entries trigger Ca²⁺ release from SR through RyRs. As a result, the cytosolic free Ca²⁺ concentration is increased, which facilitates the binding of cytosolic Ca²⁺ with the myofilament protein troponin C to switch on the cell's contractile machinery. Later, cytosolic Ca²⁺ is removed by the SR Ca²⁺-ATPase (SERCA; its function is regulated by phospholamban), sarcolemmal NCX, sarcolemmal Ca²⁺-ATPase and the mitochondrial Ca²⁺ uniporter, or other mitochondrial transport pathways [rapid-mode uptake pathway (RaM) and mitochondrial ryanodine receptor (mRyR)], such that its concentration is reduced. In turn, this decreased cytosolic Ca²⁺ concentration leads to the dissociation of Ca²⁺ from troponin C and subsequent cardiac relaxation (Bers, 2002). Mitochondrial RaM and mRyR are thought to underlie this phenomenon of excitation–metabolism coupling (Brookes, 2004).

Comparison of relevant calcium networks

Regulated membrane ion transport constitutes the main scheme used by cells to maintain optimal concentrations of metal ions in their cytosol. The names, types and regulations of the various calcium transport proteins involved in calcium homeostasis/signaling systems in yeast cells and mammalian cardiac myocytes are listed in Table 1. A concise summary of the functional protein counterparts in these two systems is shown in Table 2. As we can see from Tables 1 and 2, most proteins involved in calcium homeostasis and signaling in yeast have functional counterparts in the corresponding system in mammalian cardiac myocytes. In fact, the functional expression of heterologous proteins in yeast calcium homeostasis machinery has been recently shown to be a feasible, and elegant way, to investigate the function of mammalian Ca²⁺-transport proteins. For example, it has been demonstrated that heterologous expression of mammalian PMCA, SERCA and hSPCA1 pumps all fully complemented the BAPTA hypersensitivity of K616 (Δpmr1Δpmc1Δcnb1), a yeast mutant-lacking endogenous Ca²⁺ pumps (Ton & Rao, 2004; Ton, 2002).

1

Calcium transport proteins that affect calcium homeostasis in yeast cells and cardiac myocytes

	Protein name	Type	Regulation	Other notes
Budding yeast	Transporter X	Transporter	Regulated by the extracellular Mg2+ level and by the rapid cytosolic Ca2+-dependent feedback inhibition pathway independent of calcineurin (Cui, 2009)
Transporter M	Transporter	Regulated by the extracellular Mg2+ level and by the rapid cytosolic Ca2+-dependent feedback inhibition pathway independent of calcineurin (Cui, 2009)	Opens in case of high extracellular Ca2+
Cch1-Mid1	VGCC-type channel	Regulated by plasma membrane potential and some other unidentified factors (Locke, 2000; Muller, 2001). Its activity is partially inhibited by L-type VGCC blockers nifedipine and verapamil and increased by diltiazem (Teng, 2008)	Activated only under some abnormal conditions
Pmc1	ATPase	Regulated by a slow calcineurin-dependent gene expression feedback control pathway through Crz1 (Kingsbury & Cunningham, 2000; Stathopoulos-Gerontides, 1999)	Critical for long-term calcium homeostasis
Vcx1	Ca2+/H+ exchanger	Regulated by vacuolar pH and by calcineurin through a presumably post-translational mechanism (Bonilla & Cunningham, 2002; Kingsbury & Cunningham, 2000)	Critical for short-term calcium homeostasis
Pmr1	P-type ATPase	Regulated by a slow calcineurin-dependent gene expression feedback control pathway through Crz1 (Kingsbury & Cunningham, 2000)	Mediating the high-affinity transport of Ca2+ and Mn2+
Yvc1	Channel	Regulated by cytosolic calcium level, calmodulin and cytosolic pH (Bertl & Slayman 1992; Su, 2009)
Mammalian Cardiac Myocyte	Sarcolemmal Leak	Channel	Regulated by sarcolemmal membrane potential (Shannon, 2004)
LTCC	Channel	Regulated by sarcolemmal membrane potential and by CaMKII and PKA through phosphorylation (Bers, 2002; Sipido & Eisner, 2005)
Sarcolemmal NCX	3Na+/Ca2+ exchanger	Regulated by sarcolemmal membrane potential and by both extracellular and the cytosolic Na+ and Ca2+ concentrations and by cytosolic Ca2+-dependent gene expression feedback control pathway (Bers, 2002, 2008; Shannon, 2004; Prasad & Inesi, 2009)	Rapid removal of cytoplasmic Ca2+(contribution: 28%); reversible
Sarcolemmal Ca2+-ATPase	ATPase	Regulated trivially by cytosolic Ca2+ (Shannon, 2004)	Slow removal of cytoplasmic Ca2+ (1%)
RyR	Channel	Regulated by cytosolic calcium through a CICR mechanism and by numerous proteins (e.g. calmodulin, CaMKII, FK-506 binding protein, PKA, phosphatases 1 and 2A and sorcin) and by a cytosolic Ca2+-dependent gene expression feedback control pathway (Bers, 2002, 2008)	A total of more than 100 RyRs are arranged in large organized arrays
SERCA	ATPase	Regulated by PLB, whose phosphorylation state is regulated by PKA and CaMKII. The expressions of both SERCA and PLB may be regulated by cytosolic Ca2+-dependent gene expression feedback control pathways (Bers, 2002; Sipido & Eisner, 2005; Prasad & Inesi, 2009)	Rapid removal of cytoplasmic Ca2+ (70%)
Mitochondrial Uniporter	Transporter	Regulated by mitochondrial membrane potential and by a p38 MAP kinase-dependent pathway (Brookes, 2004)	Slow removal of cytoplasmic Ca2+ (<1%)
Mitochondrial NCX	2Na+/Ca2+ exchanger	Regulated by mitochondrial membrane potential and by both mitochondrial and cytosolic Na+ and Ca2+ concentrations (Brookes, 2004)
mRyR	Channel	Regulated by cytosolic calcium through a calcium-induced Ca2+-uptake mechanism (Brookes, 2004)	Rapid removal of cytoplasmic Ca2+ (<1%)
RaM	Possibly channel	Regulated by cytosolic Ca2+ concentration, strongly inhibited by AMP and affected by the cytosolic ADP level in a biphasic way (Buntinas, 2001)	Quick removal of cytoplasmic Ca2+ (<1%)

The quantitative importance of various routes (SERCA, sarcolemmal NCX, sarcolemmal Ca²⁺-ATPase, mitochondrial Ca²⁺-sequestering pathways) for the removal of cytoplasmic Ca²⁺ varies between species. The proportions (70%, 28%, 1%, 1%) shown in Table 1 are for rabbit ventricular myocytes (Bers, 2002).

VGCC, voltage-gated Ca²⁺ channel.

2

Functional counterparts of proteins in calcium homeostasis and signaling systems in yeast and cardiac myocytes

Yeast protein	Cardiac protein in human	Similarity		Equivalence and complementation
Functional similarity	Gene sequence similarity (%)
Transporter X	Sarcolemmal leak	Both mediate constitutive Ca2+ influx	Not known	Not known
Cch1	Ca(v)1.2 (α 1C subunit of LTCC)	Both mediate additional Ca2+ influx in most cases of calcium signaling and are inhibited by L-type VGCC blockers nifedipine and verapamil (Teng, 2008)	45	Equivalent
Vcx1	NCX1	Exchangers mediating the removal of cytoplasmic Ca2+	31	No
Pmc1	PMCA1	ATPases mediating the removal of cytoplasmic Ca2+	50	Equivalent, complementation carried out (Ton & Rao, 2004)
Pmr1	SERCA2	Both mediate the sequestration of cytoplasmic Ca2+ into ER/SR and are regulated by cytosolic Ca2+-dependent gene expression feedback control pathways (Locke, 2000; Cui & Kaandorp, 2006; Prasad & Inesi, 2009)	48	Partially equivalent, complementation carried out. Mammalian SERCA complements Pmr1 on its Ca2+ transport activity, but fails to complement its Mn2+ transport activity (Ton & Rao, 2004; Ton et al., 2002)
Cmd1	CAMI, CAMII	Calmodulin (Ca2+ sensor proteins)	66 for Cmd1 with CAMI; 63 for Cmd1 with CAMII	Equivalent, complementation carried out (Cyert, 2001)
Cna1	CNA1	Catalytic subunit of calcineurin	54 (Cyert, 1991)	Equivalent
Cnb1	CNBII	Regulatory subunit of calcineurin	56 (Cyert & Thorner, 1992)	Equivalent
Rcn1	RCAN1	Regulator of calcineurin	43	Equivalent, complementation carried out (Kingsbury & Cunningham, 2000)
Mck1	Gsk3β	Kinases mediating the phosphorylation of Rcn1 and RCAN1 proteins	49	Equivalent
Crz1	NFATc4	Calcineurin-dependent transcription factor	40	No
Yvc1	TRPC1	Transient receptor potential (TRP) Ca2+ channels	50	Equivalent

The gene sequence similarity is analyzed using the global pairwise alignment tool named as stretcher available from the website http://mobyle.pasteur.fr/cgi-bin/portal.py. A detailed introduction of this popular tool can be found at http://bioweb2.pasteur.fr/docs/EMBOSS/stretcher.html. The sequence similarity score calculated using stretcher are in accordance with the reported values in the literature (for example, the similarity score for Cna1 and CNA1 is 54%; the similarity score for Cnb1 and CNBII is 56%).

Another important observation from Table 1 is that the phosphorylation of several key proteins (phospholamban, LTCC and RyR), by kinases such as protein kinase A (PKA) and calcium/calmodulin-dependent protein kinase II (CaMKII), plays a critical role in the calcium homeostasis process in cardiac myocytes (Bers, 2002; Sipido & Eisner, 2005). This is because the short duration of a heart beat (typically <1 s) requires very rapid regulation processes such as protein phosphorylation, whose speed can never be achieved by an expression control process, which usually requires >3 min (Bers, 2002; Cui, 2009). This also explains why yeast cells use slow calcineurin-dependent expression control pathways (see Fig. 1a) to regulate only the long-term rather than the short-term calcium homeostasis process (Cui & Kaandorp, 2006; Cui, 2009). From this comparison, we can speculate that, besides the possibility of involvement of direct inhibition by calmodulin, the rapid Ca²⁺-dependent feedback inhibition of transporters M and X may also involve kinase-dependent phosphorylation (Cui, 2009).

Measured dynamics and mathematical modeling

Protein-based indicators such as aequorin have been widely used to monitor the cytosolic-free calcium concentration in populations of yeast cells due to the ease of recombinant expression in these unicellular organisms (Miseta, 1999; Denis & Cyert, 2002; Cui, 2009). A simple algebraic relationship has been experimentally established between the light emission of aequorin and the calcium concentration (Allen, 1977). For example, Fig. 2a describes cytosolic calcium transients that were measured using aequorin for the yvc1 cch1 mutant under conditions of a very strong extracellular calcium shock (800 mM), together with an extracellular Mg²⁺ challenge (Cui, 2009). In cardiac myocytes, nonratiometric dyes, such as fluo and rhod dyes, are used widely to determine the calcium concentration (Cheng, 1996; Weber, 2002; Rudolf, 2003). Compared with the rapid change in the cardiac calcium concentration shown in the inset of Fig. 1b (the dashed line), the change of yeast cytosolic calcium concentration (see Fig. 2a) seems to be much slower.

2

(a) Measured aequorin luminescence curves for yeast yvc1 cch1 mutant cells in response to the sudden step-like increase of the extracellular calcium concentration (800 mM), together with the step-like increase of extracellular Mg²⁺ to various concentrations (0, 0.1, 0.3, 1 and 3 mM). (b) The corresponding simulated response curves.

A mathematical model has been constructed to reproduce these measured calcium transients in yeast cells, and the corresponding simulated response curves are shown in Fig. 2b. In the model (Cui, 2009), standard Michaelis–Menten kinetics were used to describe the uptake behavior of cytosolic Ca²⁺-sequestering proteins (Pmc1, Pmr1 and Vcx1), while the experimentally noted Mg²⁺-sensitive Ca²⁺ influx was modeled using Michaelis–Menten kinetics with Mg²⁺ competitive inhibition. The necessity of having two Mg²⁺-sensitive influx pathways (transporters M and X), rather than having only one Mg²⁺-sensitive influx pathway (i.e. transporter X) was demonstrated by both theoretical analysis and optimal fitting to the experimental data using a hybrid optimization algorithm. The validity of the model was further confirmed by its ability to reproduce the experimentally determined effects of Vcx1 on cytosolic-free Ca²⁺ dynamics and to simulate the effects of extracellular Mg²⁺ removal (Wiesenberger, 2007; Cui, 2009).

In yeast cells, calcium signals are usually in the form of transients (see Fig. 2a) whereas in cardiac myocytes various forms exist, including calcium oscillations, sparks and transients (Cheng, 1996). In part, this difference can be explained by the relatively large size of cardiac myocytes relative to yeast cells. In turn, spatial effects such as diffusion play an important role in cardiac myocytes. For example, it has been experimentally noted that the calcium concentration in sarcolemmal–SR junctions controlling RyR and LTCC differs considerably from the local submembrane calcium concentration that controls sarcolemmal NCX (Bers, 2002; Weber, 2002). In addition, the spatial organization of RyRs in cardiac myocytes and their regulatory mechanisms involving a combination of positive [i.e. calcium-induced Ca²⁺ release (CICR)] and negative feedback controls by cytosolic Ca²⁺ (see Fig. 1b and Table 1) further account for the difference between the cells (Bers, 2002; Kim, 2008). These structural and functional differences between the cells also translate to the mathematical models of calcium dynamics in yeast and cardiac myocytes. Because of the biological complexity of cardiac myocytes, the mathematical models describing calcium homeostasis/signaling in these cells (Sobie, 2002; Shannon, 2004; Niederer, 2006; Cui & Kaandorp, 2008; Groff & Smith, 2008) are usually much more complicated than those for yeast cells (Cui & Kaandorp, 2006; Cui, 2009). Besides ordinary differential equations (ODEs), partial differential equations are frequently used to model the calcium homeostasis and signaling processes in cardiac myocytes. Moreover, in order to approximate microdomain and stochastic calcium signaling events in cardiac myocytes, more sophisticated and novel methods such as molecular dynamics simulations and stochastic simulation methods are required (Bers, 2002; Beckstein, 2003; Groff & Smith, 2008).

As an example of the complexity of calcium systems in cardiac myocytes, the calcium–calcineurin–regulator of calcineurin (RCAN)–nuclear factor of activated T cells (NFAT) signaling network is shown in Fig. 3. This describes the experimentally found calcineurin-dependent signaling pathways controlling heart growth under conditions of hypertrophic stress, such as pressure overload (PO) and active calcineurin (CaN^*) overexpression (Cui & Kaandorp, 2008). This network is a part of the recently characterized group of signal-transduction pathways implicated in the regulation of cardiac hypertrophy (Heineke & Molkentin, 2006). The activation of different pathways seems to be stimulus specific (Sipido & Eisner, 2005).

3

A schematic graph depicting the calcium–calcineurin–RCAN–NFAT signaling network in cardiac myocytes (this figure is modified after the graph in the cellml version of the model developed by Cui and Kaandorp, see http://models.cellml.org/exposure/8e4d38d104176cb9a0d2e126ea23941c/cui_kaandorp_2008_c.cellml.index.html (Lloyd, 2008). As shown in the top left corner, the increased level of cytosolic calcium incurred by stress is sensed by calmodulin. Ca²⁺-bound calmodulin binds to CaN to activate it. CaN^* can bind to RCAN to form Complex1 (Vega, 2003a, b) and can also dephosphorylate NFAT^P to produce NFAT. GSK3β catalyzes the phosphorylation of NFAT to produce NFAT^P, which can bind to protein 14-3-3 to form Complex3 (Liao, 2005). Such phosphorylation of NFAT and dephosphorylation of NFAT^P also occur in the nucleus (Hallhuber, 2006). Cytosolic NFAT is imported into the nucleus and nuclear NFAT^P is exported into the cytosol. Nuclear NFAT is responsible for initiating the transcription of hypertrophic genes and also the gene encoding RCAN (more precisely, RCAN1, a form of RCAN). Both GSK3β and CaN^* are shuttled between the nucleus and the cytosol. As shown in the top right corner, certain stresses, such as PO, can activate BMK1 (Takeishi, 2001), which phosphorylates RCAN on serine 112 to produce RCAN^P. RCAN^P can be further phosphorylated by GSK3β to produce RCAN^PP. The dephosphorylation of RCAN^PP into RCAN^P is again mediated by CaN^*. Protein 14-3-3 can bind RCAN^PP to form Complex2 (Abbasi, 2006). CaN, calcineurin; CaN^*, active calcineurin; NFAT, nuclear factor of activated T-cells; NFAT^P, phosphorylated NFAT (please note that NFAT has multiple phosphorylation sites. Here, for simplicity, we denote phosphorylated NFAT as NFAT^P. This is similar in the case of Crz1); RCAN, regulator of calcineurin (also named as calcipressin, down syndrome critical region or modulatory calcineurin-interacting protein) (Davies, 2007); RCAN^P, phosphorylated RCAN on serine 112; RCAN^PP, phosphorylated RCAN on both serine 112 and serine 108; BMK1 (also called ERK5), big mitogen-activated protein kinase 1; GSK3β, glycogen synthase 3β; Complex1, the complex formed by RCAN and calcineurin; Complex2, the complex formed by RCAN^PP and 14-3-3; Complex3, the complex formed by NFAT^P and 14-3-3; PO, pressure overload; stress, hypertrophic stimuli.

In order to accurately simulate the calcium dynamics using a mathematical model, the whole network was broken down into its composite reactions, which were automatically transformed into 28 equations using cellerator software (Cui & Kaandorp, 2008). A similar procedure can be generally applied to the ODE modeling of various signal transduction networks (Shapiro, 2003; Hilioti, 2004; Cui, 2008).

The model simulation results demonstrate that the inhibiting effects of RCAN, under conditions of CaN^* overexpression, are due to the buffering effects of RCAN on calcineurin. Meanwhile, the seemingly positive influence of RCAN on calcineurin signaling, in the presence of PO, can be explained by the second compensatory buffering effect of phosphorylated RCAN (i.e. RCAN^PP) on protein 14-3-3, which otherwise would buffer NFAT^P and prevent recognition by calcineurin (Vega, 2003b; Cui & Kaandorp, 2008).

Besides this mathematical model, there exist several other published computational studies, that focus on the paradoxically dual role of RCAN. For example, the model by Shin (2006) assumes the existence of a threshold concentration of nuclear NFAT, above which RCAN is switched on, and the apparent facilitating role of RCAN under PO is explained as the effect of the release of CaN from the CaN/RCAN complex (i.e. Complex1) by some protein kinase. Compared with the work by Shin (2006), the current model integrates more recent and extensive experimental findings such as the function of active calcineurin in the nucleus and the buffering effect of 14-3-3 on NFAT^P (Liao, 2005; Hallhuber, 2006).

Similarly, in yeast, the paradoxical role of Rcn1 in calcineurin activity has been both reported experimentally and modeled computationally. Using this combination of techniques, Hilioti (2004) demonstrated that the facilitating role of phosphorylated Rcn1 in calcineurin signaling involves the phosphorylation of a conserved serine residue of Rcn1 by Mck1. Later, the experimental work by Kishi (2007) suggested that this Mck1-dependent phosphorylation targets Rcn1 for degradation through the SCF^Cdc4 complex, which relieves the inhibition of calcineurin, thereby stimulating calcineurin. Lastly, Mehta (2009) showed that a low expression of Rcn1 stimulates calcineurin signaling in a docking-dependent and proteolysis-dependent fashion, which strongly suggests that calcineurin signaling benefits from a transient interaction with Rcn1. Based on this observation, Mehta (2009) argued that Rcn1 may function as a suicidal molecular chaperone of calcineurin and could regulate its activity in a manner similar to the regulation of type-1 protein phosphatase (PP1) by inhibitor-2 (I-2).

Specificity encoding in calcium signaling

The central question in the field of calcium signaling is related to its specificity. How can a single intracellular messenger (Ca²⁺) link different extracellular signals with strict specificity to the corresponding cellular responses? The calcium signature hypothesis attempts to explain this phenomenon by assuming that signals are encoded in the temporal and spatial nature, and the amplitude of the cytosolic calcium concentration changes (the calcium signature), which are later decoded by the effectors (Berridge, 1998; Thomine, 2001; Petersen, 2005). Experimental evidence to support this hypothesis has been found in many different types of mammalian cells as well as in Arabidopsis guard cells (Dolmetsch, 1997; Li, 1998; Allen, 2000; Mackenzie, 2004). For example, it has been demonstrated experimentally that Ca²⁺/calmodulin-dependent protein kinase II (CaMKII) and PKC are able to decode frequency-modulated Ca²⁺ signals (Koninck & Schulman, 1998; Oancea & Meyer, 1998; Allen, 2000). In yeast cells, although the cytosolic calcium concentration hardly oscillates, it has been reported recently that the nuclear localization of the downstream transcription factor Crz1 (see Fig. 1a) exhibits a frequency-modulated burst behavior to coordinate gene regulation under conditions of calcium stress (Cai, 2008).

In contrast, Scrase-Field & Knight (2003) argued that calcium may merely be an essential chemical switch and that the calcium signature hypothesis describes the exception rather than the rule. Their arguments for the calcium switch hypothesis include experimentally documented phenomena in plant calcium signaling, such as similar calcium signatures with different end responses, as well as similar end responses with different calcium signatures. Similar phenomena have also been found experimentally in yeast cells. For example, Wiesenberger (2007) reported that Mg²⁺ starvation and Ca²⁺ stress can induce very similar cytosolic calcium transients in yeast. In addition, this experimental observation has been supported by the simulation results of a mathematical model (Cui, 2009). Further, a striking portion of the genes that are upregulated under conditions of Mg²⁺ depletion, including ENA1 (encoding a P-type ATPase sodium pump) and PHO89 (encoding a sodium/phosphate cotransporter), are also induced by Ca²⁺ stress. However, there are many other genes (such as SUL1, ARA1, MDH2 and STF2) that are upregulated under conditions of Mg²⁺ depletion, but are not induced by Ca²⁺ stress (Wiesenberger, 2007).

Because both the calcium signature hypothesis and the calcium switch hypothesis are supported by experimental evidence, it will be important to discriminate between the cases in which calcium is responsible for encoding specificity, from those in which it functions only as a chemical switch. If we consider this viewpoint: the end response is usually the result of an interaction between multiple signaling pathways, we can more easily understand why the signal specificity can be encoded by signaling components other than calcium. Indeed, the cardiac calcium signaling network presented in the previous section illustrates how different stimuli can trigger their respective end response through the complex cross-interaction of different signaling pathways. This is because PO induces cardiac hypertrophy by activating both the calcium–calcineurin–NFAT pathway and the BMK1/ERK5 signaling pathway whereas the BMK1/ERK5 signaling pathway, is not activated in the case of CaN^* overexpression (Takeishi, 2001; Hayashi & Lee, 2004). We predict that accurate mathematical model descriptions of the calcium-signaling systems will help us to uncover the mystery of specificity encoding in calcium signaling networks.

Conclusions

Both yeast cells and mammalian cardiac myocytes are equipped with calcium-signaling tool kits composed of channels, pumps, exchangers and other relevant components (such as sensors like calmodulin, and effectors like calcineurin). These function to maintain calcium homeostasis and also perform the task of calcium signaling. Compared with the same systems in mammalian cardiac myocytes, calcium homeostasis and signaling systems in yeast cells are relatively simple and easy to approximate. The extreme complexity of calcium homeostasis/signaling processes in cardiac myocytes arises from their coupling with other ion homeostasis processes, important spatial and stochastic effects (e.g. the local and microdomain calcium signals), the huge number of factors involved in the systems and their sophisticated regulatory mechanisms (Bers, 2002, 2008; Heineke & Molkentin, 2006).

Surprisingly, most of the factors known in the yeast calcium homeostasis/signaling network are conserved and operate in a similar manner in mammalian cells, including cardiac myocytes (see Table 2 and Fig. 4; Ton & Rao, 2004; Dolinski & Botstein, 2007). For example, as shown in Fig. 4, NFAT translocation in cardiac myocytes is strikingly similar to the process of Crz1 translocation in yeast (Stathopoulos-Gerontides, 1999). Further, RCAN regulation on calcineurin signaling in cardiac myocytes is similar to Rcn1 regulation on calcineurin signaling in yeast (Hilioti, 2004). Thus, our knowledge of the calcium homeostasis/signaling system in a simple organism, such as the budding yeast (S. cerevisiae) can help us to better understand the calcium homeostasis/signaling systems in mammalian cardiac myocytes, and treat relevant human diseases such as pathological cardiac hypertrophy and heart failure.

4

Highlighting the striking similarity between the Ca²⁺ homeostasis/signaling systems of yeast cells and cardiac myocytes. In yeast, Ca²⁺ induces the expression of RCN1 (the gene encoding Rcn1) through the calcineurin-dependent activation of Crz1. After the phosphorylation of serine 117 by an unknown priming kinase, Rcn1^P can be further phosphorylated by MCK1 on serine 113. The dephosphorylation of Rcn1^PP into Rcn1^P is mediated by calcineurin (Kishi, 2007). All these components and mechanisms are conserved in the corresponding system in mammalian cardiac myocytes (Cui & Kaandorp, 2008). Crz1^P, phosphorylated Crz1; Rcn1^P, phosphorylated Rcn1 on serine 117; Rcn1^PP, phosphorylated Rcn1 on both serine 117 and serine 113.

In order to accelerate our progress in the field of yeast calcium homeostasis/signaling systems, MS-based proteomics, complemented with yeast two-hybrid assays, can provide a very powerful technique for searching for the missing components in the networks, detecting and determining the protein interactions and quantifying the concentrations of proteins (Causier, 2004; Olsen, 2006; Cox & Mann, 2007). Functional assays based on gene-knockout techniques can provide a useful method for checking the validity of mathematical models. In order to determine the rate constants for such models, various methods exist, such as surface plasmon resonance analysis (Slepak, 2000; Portmann, 2004). Electrophysiological recordings of ion channel activity in the plasma membrane of live yeast cells have been proven to be difficult to obtain (Putney, 2005). An alternative method to achieve such measurements is to express the corresponding genes in mammalian cells, where electrophysiological recordings are much easier to obtain. Effective collaborations among scientists who are proficient in genetics, proteomics and computational science will result in the eventual completion of the whole-yeast calcium homeostasis/signaling system and an improved understanding of its dynamics. This will be achieved via high-throughput experimental and computational methods, and then back through the iterative systems biology procedure (i.e. experiment→model→experiment) (Finkelstein, 2004).

Acknowledgements

We would like to thank Prof. Kyle W. Cunningham (Johns Hopkins University) for his valuable data and stimulating discussions. We sincerely acknowledge all the other colleagues and collaborators who were involved in the work presented in this paper, including Y. Fomekong Nanfack, Olufisayo O. Ositelu, Veronica Beaudry and Alicia Knight. We thank two anonymous referees for their valuable comments and suggestions. J.C. was firstly funded by the Dutch Science Foundation on his PhD research, later supported by two grants from the EC on MORPHEX (NEST contract no. 043322) and QosCos projects and lastly by an NUS grant (R-252-000-350-112) for his postdoctoral project ‘Decomposition and composition of large signaling pathway models with emphasis on parameter estimation.’

Author notes

Editor: Teun Boekhout