Insights into Modulation of Calcium Signaling by Magnesium in Calmodulin, Troponin C and Related EF-hand Proteins

Abstract

The Ca²⁺-binding helix-loop-helix structural motif called EF-hand is a common building block of a large family of proteins that function as intracellular Ca²⁺ receptors. These proteins respond specifically to micromolar concentrations of Ca²⁺ in the presence of ~1000 fold excess of the chemically similar divalent cation Mg²⁺. The intracellular free Mg²⁺ concentration is tightly controlled in a narrow range of 0.5–1.0 mM, which at the resting Ca²⁺ levels is sufficient to fully or partially saturate the Ca²⁺ binding sites of many EF-hand proteins. Thus, to convey Ca²⁺ signals, EF-hand proteins must respond differently to Ca²⁺ than to Mg²⁺. In this review the structural aspects of Mg²⁺ binding to EF-hand proteins are considered and interpreted in light of the recently proposed two-step Ca²⁺-binding mechanism (Grabarek, Z., J. Mol. Biol. 346,1351,2005). It is proposed that, due to stereochemical constraints imposed by the two-EF-hand domain structure, the smaller Mg²⁺ ion cannot engage the ligands of an EF-hand in the same way as Ca²⁺ and defaults to stabilizing the apo-like conformation of the EF-hand. It is proposed that Mg²⁺ plays an active role in the Ca²⁺-dependent regulation of cellular processes by stabilizing the off state of some EF-hand proteins, thereby facilitating switching off their respective target enzymes at the resting Ca²⁺ levels. Therefore, some pathological conditions attributed to Mg²⁺ deficiency might be related to excessive activation of underlying Ca²⁺-regulated cellular processes.

The EF-hand is a Ca²⁺-binding helix-loop-helix structural motif discovered by R.H. Kretsinger in the X-ray structure of parvalbumin [1]. It is a basic building unit of a large family of proteins that serve as intracellular Ca²⁺-receptors. These proteins undergo conformational changes upon binding Ca²⁺, which enables them to reversibly bind to and regulate the activity of various target enzymes. Paradoxically, parvalbumin, the first crystallized member of this family does not change its structure significantly on binding Ca²⁺ and appears to contribute to Ca²⁺-signaling only indirectly by buffering the cytosolic free Ca²⁺. Among the Ca²⁺-sensor proteins, troponin C (TnC) and calmodulin (CaM), each consisting of four EF-hand motifs, have served as models for extensive structure-function studies and our current understanding of the EF-hand is based to a large extent on these studies. The early work in the laboratory of the late Prof. Witold Drabikowski at the Nencki Institute of Experimental Biology in Warsaw has demonstrated that both TnC and CaM can be split by limited proteolysis into two approximately equal size fragments that retain their structure and Ca²⁺-binding properties [2–7]. Thus, this work has defined a pair of EF-hand motifs as a structurally independent, functional domain. Subsequently, the high resolution X-ray and NMR structures of TnC [8–10] and CaM [11–14] provided a vivid demonstration of domain independence in these proteins, the functional significance of which became apparent when the details of TnC and CaM interactions with their target molecules were established [15–20].

A pair of EF-hand motifs forms a structurally stable cooperative domain due to the network of hydrogen bonds that connect the Ca²⁺-binding loops (a short stretch of antiparallel β-sheet) and due to extensive hydrophobic contacts among the helices flanking the Ca²⁺-binding loops [21]. This EF-hand pairing pattern, first defined in the structure of parvalbumin [1], has been reproduced in all high-resolution structures of EF-hand proteins [22], and in model systems utilizing synthetic peptide analogs of EF-hands [23–25]. The bond network directly connecting the two Ca²⁺ ions in the domain, collectively called EF-handβ-scaffold, plays an important role in the Ca²⁺-binding mechanism and, in a recently proposed two-step mechanism, it is viewed as critical for the Ca²⁺-induced conformational transitions, and thus for the Ca²⁺-signaling function of EF-hand proteins [26, 27]. An extensive discussion of the EF-hand proteins structure and function can be found in the recent review by Gifford et al [28].

The focus of the present review is the specificity of the EF-hand proteins, i.e. their ability to recognize Ca²⁺ signals in the cytosol in the presence of ~1000 fold excess of the chemically similar divalent cation Mg²⁺. This remarkable property is discussed in light of the EF-hand-β-scaffold model [26, 27], which implies that, due to specific constraints, Mg²⁺ cannot engage the ligands of the EF-hand in the same way as Ca²⁺ does. It is further proposed that Mg²⁺ may play an active role in the Ca²⁺-dependent regulation of cellular processes by stabilizing the off state of some EF-hand proteins, thereby facilitating switching off their downstream activity at the resting Ca²⁺ levels. Some pathological conditions attributed to Mg²⁺ deficiency might be related to excessive activation of underlying Ca²⁺-regulated cellular processes.

The EF-hand calcium switch

The mechanism of the Ca²⁺-induced conformational changes in EF-hand proteins has been the subject of extensive studies for several decades. Although numerous techniques have provided evidence for the Ca²⁺-dependent structural changes in TnC, CaM and other EF-hand proteins, the first comprehensive description of the Ca²⁺ switch came from Herzberg, Moult and James [29] on the basis of the crystal structure of TnC [8, 9]. The switch occurs in a two-EF-hand domain and involves a change in the relative orientation of the helices flanking the Ca²⁺-binding loops, thus causing a transition from a closed to an open domain conformation. This off to on transition leads to the exposure of a hydrophobic pocket that serves as a target interaction site (reviewed in [21, 22, 28]).

The HMJ-model provided a description of the Ca²⁺ switch, but has not explained the underlying mechanism. An important insight in this respect came from the work of Sykes and colleagues who pointed to the critical role of the invariant Glu residue in the 12^th position of the Ca²⁺-binding loop [30, 31]. This residue contributes two oxygen atoms of its γ-carboxyl group to the coordination sphere of the Ca²⁺ ion (a bidentate ligand). The importance of this residue was also evident from the earlier experiments on CaM showing that substitutions of Glu₁₂ with Gln or Lys in the four Ca²⁺-binding sites of CaM had deleterious effects on the Ca²⁺-binding affinity and target enzyme activation [32–34]. More detailed information about the Ca²⁺-binding mechanism came from the X-ray structure of a CaM mutant, which had intact Ca²⁺-binding loops, but was locked with a disulfide bond in the closed, apo-like, conformation [26]. In that structure, the Glu₁₂ ligand was found at some distance from the Ca²⁺ ion, thus providing evidence for direct coupling between that ligand and the movement of the C-terminal (exiting) helix of an EF-hand and opening of the domain. In that structure the relative position of the two Ca²⁺ ions with respect to each other was found to be the same as in the open-domain conformation. The inference from this observation was that the structure connecting the two bound Ca²⁺, which includes the invariant carbonyl oxygen ligand in the center of the loop (the Y metal coordinating position) and the short stretch of the β-sheet connecting the two Ca²⁺-binding loops provide a stable, conformation independent scaffold for Ca²⁺ binding [26]. Since each EF-hand can bind Ca²⁺ only in the space very narrowly defined by the β-scaffold, the helices flanking the Ca²⁺-binding loop have to move in order for the other ligands to reach the metal ion (Figure 1). This is consistent with the requirement for pairing of the EF-hand motifs in a domain. In a single EF-hand the Ca²⁺ ion would most likely affect the local loop conformation without any long-range energetically unfavorable helix movement. The proposed model referred to as the EF-hand-β-scaffold model (EFBS) emphasizes the importance of the structure connecting the two bound Ca²⁺ ions. The invariant backbone carbonyl oxygen ligand in the center of the loop (cf. Figure 1) defines the position of the bound Ca²⁺ because unlike other ligands it is virtually immobilized by the β-scaffold [26]. The hydrogen bonds of the β-scaffold are highly polarized [35, 36]; a property that imparts the carbonyl Y ligand with a partial negative charge, thus further facilitating Ca²⁺ coordination. According to the EFBS model, Ca²⁺ binding occurs in two steps. Initially the N-terminal flexible part of the Ca²⁺-binding loop coordinates the Ca²⁺ ion. In the next step, bond rotation in the β-scaffold, specifically a change in the dihedral ϕ,ψ angles of the residue in the 8^th position of the loop, enables the side chain of the Glu residue in the 12^th position to approach the Ca²⁺ ion [26]. The interaction of that last bidentate ligand with Ca²⁺ closes the Ca²⁺-binding loop and causes the exiting helix to move, which results in domain opening.

Figure 1.

The Ca²⁺-induced conformational change in site I of CaM. The structures of the N-terminal domain of human recombinant CaM in the Ca²⁺-bound form (PDB entry 1CLL, [112]) and apo form of rat recombinant CaM (PDB entry 1QX5, [113]) have been superimposed using the backbone atoms of the β-scaffold (see ref [27] for details). In this conformation independent frame of reference the movement of the helices and closing of the loop associated with Ca²⁺ coordination can be discerned. Note the large shift of the C-terminal bidentate ligand (Glu31). This shift is enabled by the backbone dihedral angle changes in the branched hydrophobic residues of the β-scaffold (Ile27 in site I of CaM), which work as a hinge [26].

There are still very few EF-hand proteins for which both the apo and the Ca²⁺-bound high-resolution X-ray structures are available. It is clear, however, that the large conformational changes characteristic of TnC and CaM are neither unique, nor common to all EF-hand proteins. In fact, there is a great variety of possible structural responses to Ca²⁺ [22, 27, 28, 37]. However, irrespective of the extent of the Ca²⁺-induced hydrophobic site exposure, or other more subtle structural changes, the bidentate Ca²⁺ coordination by the Glu residue in the last position of the Ca²⁺-binding loop is preserved. A comparison of the high resolution X-ray structures of various EF-hand proteins suggests also that a mechanism similar to that outlined by the EFBS model may operate not only in the canonical EF-hands, i.e. those containing 12 amino acids in the Ca²⁺-binding loop, but also in those that have 11, 13 or 14 amino acids in the loop [27]. In these unconventional EF-hands variations in the amino acid sequence, composition and ultimately conformation are limited to the N-terminal part of the Ca²⁺-binding loop, while the β-scaffold structure and the bidentate Ca²⁺-coordination by the C-terminal Glu are preserved. While several facets of the EFBS model are hypothetical at present and require experimental verification, the model appears to offer attractive possibilities for design of new experiments, which might lead to a better understanding of the EF-hand proteins function.

Calcium signaling in the presence of magnesium

Mg²⁺ is one of the three physiologically relevant metal ions that are known to bind to EF-hand proteins and to modulate their Ca²⁺binding properties. The other two are Zn²⁺ and Cu²⁺, which bind to some members of the S100 protein family [38]. The binding site for Zn²⁺ and Cu²⁺ is located at the interface between two S100 monomers, outside of the N-terminal Ca²⁺-binding loop, and involves His and Asp or Cys side chains in a tetrahedral coordination geometry [39, 40]. Apparently, different preferences for ligands and different coordination geometry effectively exclude Zn²⁺ and Cu²⁺ from competing with Ca²⁺ for the EF-hand Ca²⁺-binding sites.

The intracellular free Ca²⁺ concentrations fluctuate between 10⁻⁷ M at rest and 10⁻⁵ M during activation [41, 42]. In contrast, the free Mg²⁺ level is relatively constant and falls in a narrow range of 0.5–1.0 mM in most cells [43]. Interestingly, hormonal stimulation can induce large fluxes of Mg²⁺ through the cell membrane, which suggests that the total Mg²⁺ content in cells and, indirectly, the free Mg²⁺ concentrations are tightly regulated [44, 45]. Many EF-hand Ca²⁺-binding sites have Mg²⁺ dissociation constants in submillimolar range, thus can be expected to bind Mg²⁺ in the absence of competing Ca²⁺. This observation alone implies that the Mg²⁺-bound structure should be different from the Ca²⁺-bound structure. One of the two structures must represent the activated form of the protein (the on state) while the other must correspond to the off state, otherwise, for those EF-hand proteins that bind Mg²⁺ with appreciable affinity, no relaxation and therefore no Ca²⁺-signaling would be possible. Indeed, there is a large body of published data documenting differences between Ca²⁺ and Mg²⁺ binding and the resultant structural and functional effects for various EF-hand proteins. Some relevant observations for CaM and TnC are reviewed below and interpreted in light of the EF-hand-β-scaffold model in anticipation of arriving at some more general conclusions potentially applicable to other EF-hand proteins.

Mg²⁺ binding to calmodulin

Calmodulin is a ubiquitously expressed Ca²⁺ sensor protein responsible for the regulation of many crucial cellular processes [46]. Therefore, the Mg²⁺-binding properties of CaM have been the subject of great interest. A number of early studies with the use of diverse techniques provided evidence for Mg²⁺ binding to CaM and for significant structural differences between the Mg²⁺-bound and the Ca²⁺-bund forms of this protein [7, 47–53]. There are large differences in the Mg²⁺-CaM binding constants reported by various authors (~10²–10⁴ M⁻¹), however there is a consensus that Mg²⁺ has the opposite preference for the binding sites relative to Ca²⁺, i.e. site I in the N-terminal domain has the highest affinity for Mg²⁺ [53–56]. For example, Tsai et al. have reported K_a=3.5×10³ M⁻¹ for the N-terminal domain and K_a=2–3×10² M⁻¹ for the C-terminal domain [53]. On the basis of their NMR studies Malmendal et al. have concluded that nearly half of the sites in the N-terminal domain of CaM would be occupied by Mg²⁺ in resting eukaryotic cells [55]. They also suggested that Mg²⁺ binding might occur without ligation to the residue in the 12^th loop position, consequently causing smaller structural changes. In another work from the same group, the apparent affinity for Ca²⁺ of the N-terminal domain of CaM was found to be significantly decreased at physiological Mg²⁺ levels, consistent with a model in which each site can bind either Ca²⁺ or Mg²⁺ [57]. Under specific conditions a significant population of a mixed Mg²⁺/Ca²⁺ state was found and the Ca²⁺ off-rate from this state was at least one order of magnitude faster than from the 2Ca²⁺ state. The mixed state was found to have a conformation more similar to the "closed" apo and 2Mg²⁺ states than to the "open" 2Ca²⁺ state [57]. A more complex view of the Ca²⁺/Mg²⁺ competition emerged from microcalorimetric studies [51, 58–60]. Gilli et at. have found the thermodynamic parameters of Mg²⁺ and Ca²⁺ binding to CaM inconsistent with a simple competition mechanism and proposed that Mg²⁺ is an allosteric effector with respect to Ca²⁺ binding and that the four main Mg²⁺ binding sites are distinct from the EF-hand Ca²⁺ binding sites [60]. In the absence of high resolution X-ray structure of Mg²⁺-bound CaM it is not clear what exactly the distinction between the Mg²⁺ and Ca²⁺ binding sites might be. Perhaps the two different views could be reconciled by a model in which Mg²⁺ engages only a subset of the ligands that interact with Ca²⁺ and, as a result, stabilizes domain conformation that is very different from that induced by Ca²⁺.

There is evidence for a strong stabilization of both domains of CaM in the presence of Ca²⁺ or Mg²⁺, as reflected in a large increase in the unfolding temperature [61, 62]. For each metal, however, the stabilized conformation is different. Indeed, a simple test with a hydrophobic fluorescent probe bisANS can be used to demonstrate that Mg²⁺ binding does not induce domain opening in CaM, whereas Ca²⁺ does (Figure 2, cf. also ref. [50]). This highly specific effect of Ca²⁺ is apparently attributable to the ionic radius of the metal ion, i.e. those metals whose ionic radii are closest to Ca²⁺ (~1.0 Å) are able to substitute for Ca²⁺ in the activation of CaM target enzymes [63]. The final point that needs to be considered with respect to the potential Mg²⁺ effect on Ca²⁺-CaM signaling is the role of the target enzyme. There is ample evidence that Mg²⁺ decreases CaM s affinity for various model peptide targets, but perhaps more importantly, the extent of the effect depends on the type of the target [54, 56]. According to Martin et al. Mg²⁺ has a role in directing the mode of initial target binding preferentially to the C-domain of CaM, due to the opposite relative affinities for binding of Ca²⁺ and Mg²⁺ to the two domains. Thus, they concluded, Mg²⁺ amplifies the intrinsic differences of the domains in a target specific manner [56].

Figure 2.

Mg²⁺ does not induce domain opening in CaM and TnC. Fluorescence emission spectra of bisANS (4,4'-dianilino-1,1'-binaphthyl-5,5'-disulfonic acid) alone and after addition of CaM (left) or C-TnC, the C-terminal domain fragment of rabbit skeletal TnC, in the presence of 1 mM CaCl₂ or 10 mM MgCl₂ and 0.5 mM EGTA were recorded at 20°C upon excitation at 394 nm in solution containing 0.1 M NaCl, 20 mM HEPES pH 7.0. Note a large increase in fluorescence of bisANS if the protein is added in the presence of CaCl₂ but not in the presence of MgCl₂

Mg²⁺ binding to troponin C

In contrast to CaM, which has multiple targets and many functions, troponin C is specifically expressed in striated muscles where, as a part of the F-actin bound regulatory troponin-tropomyosin complex, it functions as a Ca²⁺-activated trigger of muscle contraction [64–66]. Although TnC, like CaM, is comprised of two independent domains, the difference in Ca²⁺ affinity between the domains is larger, which appears to provide a clear definition of their function [5, 67]. The lower affinity sites in the N-domain are considered Ca²⁺ specific and are directly involved in the regulatory function of TnC. There are conflicting reports on the effect of Mg²⁺ on the Ca²⁺-sensitivity of myofibrils and reconstituted thin filaments [67, 68]. In more recent model studies Mg²⁺ was found to compete with Ca²⁺ for the regulatory domain of skeletal TnC with the apparent dissociation constant K_d=2.2 mM [69]. The C-terminal domain contains the high affinity Ca²⁺-binding sites that also bind Mg²⁺, thus are called the Ca²⁺ /Mg²⁺ sites and are believed to play a structural role by providing stability to the troponin complex. [67, 70–72]. The potential role of the Ca²⁺/Mg²⁺ sites in regulation of contraction was originally rejected due to the slow kinetics of Mg²⁺ dissociation from these sites (~8 s⁻¹) which would be a rate limiting step for Ca²⁺ binding at these sites during activation [70]. The regulatory role of the N-domain has been confirmed by site directed mutagenesis [72, 73] however various observations suggest that the Ca²⁺ /Mg²⁺ sites also contribute to the mechanism of regulation. For example, in a reconstituted system the C-terminal domain fragment of TnC retains partial regulatory activity [74–76]. There is evidence for a Ca²⁺-dependent communication between the TnC domains in the troponin complex [77] which most likely involves the interaction of the inhibitory segment of troponin I (a twelve-residue stretch that binds actin and inhibits actin-myosin ATPase) with helix E, the N-terminal helix of the Ca²⁺ /Mg²⁺ site III of TnC [20]. It is not clear if that interaction is affected in any way by the Mg²⁺ /Ca²⁺ exchange at the adjacent site. The apparent equilibrium binding constant for Mg²⁺ binding to the C-domain of TnC has been estimated at 5×10³ M⁻¹ in free TnC and 5×10⁴ M⁻¹ in the troponin complex [67], which is sufficiently high for these sites to be fully occupied by Mg²⁺ at the resting Ca²⁺ concentrations. Due to slow Mg²⁺ dissociation rate little Mg²⁺ /Ca²⁺ exchange can be expected during a single activation-relaxation cycle, however full exchange can occur during prolonged or repetitive contraction [70]. The C-domain of TnC has little structure in the apo form, however both Ca²⁺ and Mg²⁺ induce folding [5, 78–80] and strong stabilization of that domain [61, 62]. Again, like for CaM, the Mg²⁺ bound conformation of the C-domain of TnC is closed, while the Ca²⁺-bound conformation is open (Figure 2). The open conformation of the C-domain of TnC, as detected by bisANS, is consistent with the crystal structure of TnC [8, 9] and is clearly required for the interaction of this domain with the N-terminal α-helical segment of troponin I in the presence of Ca²⁺ [19, 20, 81]. However, the closed conformation of that domain in +Mg²⁺ state poses a problem, because it would require dissociation (or a different type of interaction) of the troponin I N-terminal segment. No such dissociation is evident in the currently available low-resolution structure (7.0 Å) of the apo form of skeletal troponin core domain [20]. Also the NMR studies of cardiac C-domain complex with TnI peptide revealed a monodentate Mg²⁺ ligation by the conserved Glu₁₂ and partial closure of the hydrophobic binding cleft around Ca²⁺-binding site IV [82]. These data might suggest that the closed domain Mg²⁺-bound structure detected in solution is sufficiently flexible to open and accommodate the interacting TnI helix. Alternatively, a high-resolution structure of a complete regulatory complex may be required to reveal the effects of Mg²⁺ /Ca²⁺ exchange at this domain. In favor of the latter interpretation are data on F1 isoform of TnC from stretch activated insect flight muscle, in which only site IV in the C-domain is functional, and Ca²⁺-binding to that site is required for the function of this protein [83–85].

Structure of the Mg²⁺-bound EF-hand

Clearly, there is strong evidence for conformational differences between the Ca²⁺ and Mg²⁺ bound forms of CaM and TnC. Unfortunately these differences are difficult to interpret in structural terms because no X-ray structures of the Mg²⁺ bound forms of these proteins are available. In fact, currently there are only three X-ray crystallographic structures of the Mg²⁺-bound EF-hand proteins in the Protein Data Bank: parvalbumin, calbindin D9k (S100G, according to the recent update on nomenclature [86]) and the scallop myosin regulatory light chain (RLC). In the structure of S100G, Mg²⁺ is bound to the C-terminal Ca²⁺-specific EF-loop. The loop has extended conformation similar to that in the apo form and Mg²⁺ interacts only with the N-terminal ligands of the loop. The Glu₁₂ is too far from the bound cation for its side chain to contribute to Mg²⁺ coordination. Instead, a water molecule is found at that position (PDB entry 1IG5, [87], cf. Figure 3). The helices are more tightly packed than in the Ca²⁺-bound form. Overall, the Mg²⁺ bound EF-hand conformation of S100G is highly reminiscent of the Ca²⁺ bound EF-hands in the disulfide locked CaM [26]. In the Mg²⁺ bound loop of parvalbumin (PDB entry 4PAL, [88], cf. Figure 3) the side chain of Glu₁₂ is rotated 120 degrees to provide only one oxygen ligand for Mg²⁺ coordination as compared to the Ca²⁺-bound structure (PDB code 4CPV, [89], cf. Figure 3). The EF-hand conformation is essentially the same as that of the Ca²⁺-bound form. The transition from bidentate Ca²⁺ coordination to monodentate Mg²⁺ coordination by Glu₁₂ is apparently related to the difference in ionic radii (1.06 Å for Ca²⁺ vs 0.72 Å for Mg²⁺ [90]) as demonstrated by molecular dynamics simulations [91]. These structures point to two very distinctive features of the Mg²⁺-bound vs. Ca²⁺-bound EF-hand conformations: i) The geometry of the Mg²⁺ coordination in both parvalbumin and S100G is octahedral, even though the two proteins accomplish this geometry in different ways. ii) The bidentate interaction with the Glu₁₂ side chain, which is a defining characteristic of the Ca²⁺ complex, is absent in both Mg²⁺-protein complexes.

Figure 3.

Geometry of the equatorial plane of Ca²⁺ and Mg²⁺ complexes with an EF-hand.

A Schematic representation of an idealized metal oxygen complex. The radius of the sphere in the center (r_Me) is related to the number (n) and the radius (r_O) of the surrounding spheres by: r_O/(r_Me+r_O)=sin(360/2n). For r_O=1.52 Å (van der Waals radius of the oxygen atom) we obtain r_Me=1.06 Å for n=5 and r_Me=0.63 Å for n=4, which correlate well with the ionic radii for Ca²⁺ and Mg²⁺ obtained by X-ray crystallography (1.06 Å and 0.72 Å respectively [90]). Note that the relative orientation of the Glu₁₂ ligand with respect to the invariant carbonyl oxygen of the EF-hand-β-scaffold (assigned here as number 1 at the top) depends on the coordination number. For and the two oxygen atoms of the bidentate Glu₁₂ ligand (atoms numbered 2 and 3) occupy adjacent positions in the equatorial plane of the complex. This is the key feature of the complex that sets the direction of approach for the side chain of Glu₁₂ and consequently the position of the exiting helix with respect to the β-scaffold (approximately 108° as marked by the arrows). A similar bidentate coordination of Mg²⁺ by Glu₁₂ would not be possible due to a larger angle of approach (135°) which would require the exiting helix to move much farther away from the β-scaffold and possibly clash with the incoming helix. For the Ca²⁺ complex the O-Ca²⁺-O bond angle is 72°, as compared to 90° for the Mg²⁺ complex.

B The EF-loop of carp parvalbumin in the Ca²⁺-bound form (PDB entry 4CPV, [89] beta isoform pI=4.25) and pike parvalbumin in Mg²⁺-bound form (PDB entry 4PAL, [88] beta isoform pI=4.1) conformation. The equatorial plane of each complex is shown in a similar orientation as that in (A) with the carbonyl oxygen ligand of the β-scaffold (the Y coordinating position) at the top. The yellow and cyan spheres represent Ca²⁺ and Mg²⁺, respectively, and the red and magenta spheres are the oxygen atoms in direct contact with the metal ion. The backbone atoms of the 12 amino acids comprising the Ca²⁺-binding loop are shown in stick representation. For clarity, sidechain atoms (in stick representation) are shown only for Glu₁₂. The ionic radii were set to 1.06 Å for Ca²⁺ (sevenfold coordination) and 0.72 Å for Mg²⁺ (sixfold coordination) according to Shannon [90]. For oxygen atoms the radius is set to 1.52 Å, which represents the van der Waals volume [114].

C The C-terminal Ca²⁺-binding site of bovine recombinant S100G (calbindin D9k) in the Ca²⁺-bound (PDB entry 4ICB, [115]) and Mg²⁺-bound (PDB entry 1IG5, [87]) conformation. The pink sphere in the +Mg²⁺ structure is the oxygen atom of a water molecule that substitutes for the Glu₁₂ oxygen atoms, which are too far. These figures were generated with PyMol [116].

The Mg²⁺-bound structure of scallop myosin RLC (PDB code 1WDC [92]) shows a different and less frequent variant of the EF-hand motif featuring an Asp residue at position 12 instead of Glu. Paradoxically, this substitution makes an EF-hand specific for Mg²⁺ [93], or perhaps more aptly, the affinity for Ca²⁺ decreases to the extent that it becomes irrelevant in the physiological range of cellular free Ca²⁺ concentrations. The preference of the Mg²⁺ ion for a shorter side chain (Asp) in the last position of the loop is somewhat counterintuitive considering that the ionic radius of Mg²⁺ is ~30% smaller than Ca²⁺. One might expect that an Asp containing loop would have to compress significantly more, thus causing a larger shift of the exiting helix, as compared to that featuring Glu in the last position. This is not the case, however, since the Asp side chain only contributes a single oxygen atom to the octahedral coordination of Mg²⁺, therefore does not constrain significantly the loop conformation. The bidentate coordination that is required for Ca²⁺ is apparently not possible with the shorter Asp side chain. This subtle difference accounts for the reversed Mg²⁺/Ca²⁺ specificity of the RLC. Support for this interpretation comes from the work of Phillips and colleagues who have shown that substitution of Asp for Glu in the EF loop of parvalbumin causes a 100 fold decrease in Ca²⁺ affinity, a 10 fold increase in Mg²⁺ affinity, and that both complexes adopt octahedral coordination [94].

The two Mg²⁺-EF-hand complexes shown in Fig. 3 represent different domain structures in terms of the relative orientation of the helices flanking the Ca²⁺-binding loop. The S100G-Mg²⁺ complex resembles the closed domain, whereas the parvalbumin-Mg²⁺ complex is similar to the open domain conformation of CaM and TnC. Unfortunately, neither S100G, nor parvalbumin undergoes a large structural change upon Ca²⁺ binding. Thus, it is not clear if the conclusions based on these structures can be extrapolated to TnC, CaM and other regulatory EF-hand proteins. In particular, we are left with the following questions: 1) Why is there a change from bidentate to monodentate metal ion coordination by the carboxyl group in the 12^th position? In principle, eliminating an oxygen ligand at any other coordinating position of the loop could fulfill the Mg²⁺ requirement for hexa-coordination. 2) What are the functional consequences of the transition from bidentate to monodentate Glu₁₂ interaction with the metal ion?

Interpretation of the Ca²⁺/Mg²⁺ specificity based on the EF-hand-β-scaffold model

Differences in ionic radii, hydration and coordination geometry are generally listed among those properties that enable EF-hand proteins to discriminate between Ca²⁺ and Mg²⁺. The Mg²⁺ ion has ~30% smaller ionic radius as compared to Ca²⁺. However, owing to the tight ordering of water molecules the hydrated Mg²⁺ ion is approximately 400× larger in volume than its dehydrated form. In contrast, the hydrated Ca²⁺ ion is only ~25× larger [95]. The difference in hydration is important for ion transport [96] and for the kinetics of binding to the EF-hand. The exchange rate of solvent water around the hydrated Mg²⁺ is 3–4 orders of magnitude slower than for Ca²⁺ [95]. Thus, the Mg²⁺ binding on-rate is limited by the dehydration rate, which accounts for the generally weak binding of Mg²⁺ to proteins and small molecules. As discussed earlier, many EF-hand Ca²⁺-binding sites have sufficiently high affinity for Mg²⁺ to be partially or fully saturated with Mg²⁺ at the resting Ca²⁺ concentrations. The key factor for effective Ca²⁺ signaling is the difference in the resultant EF-hand conformation, which in turn must be related to the ionic radius and the coordination geometry of the metal ion. The EFBS model offers a plausible explanation of the relation between these factors and why Mg²⁺ is generally unable to substitute for Ca²⁺.

Figure 3A illustrates the distribution of atoms in the equatorial plane of an idealized metal ion-oxygen complex. The larger Ca²⁺ ion can accommodate 5 oxygen atoms whereas only four oxygen atoms can fit in a plane around the smaller Mg²⁺. This difference can be rationalized most simply in the following way: The radius of the sphere in the center (r_Me) is related to the number (n) and the radius (r_O) of the surrounding spheres by: r_O/(r_Me+r_O)=sin(360/2n). For r_O=1.52 Å (van der Waals radius of the oxygen atom) we obtain r_Me=1.06 Å for n=5 and r_Me=0.63 Å for n=4, which correlate well with the ionic radii for Ca²⁺ and Mg²⁺ obtained by X-ray crystallography (1.06 Å and 0.72 Å respectively [90]). Clearly, upon substitution of Ca²⁺ with Mg²⁺ one of the oxygen ligands has to be removed. There appear to be two reasons why it is the bidentate Glu₁₂ ligand that is reduced to monodentate interaction.

The first reason is related to the fixed position of the carbonyl oxygen ligand of the β-scaffold (assigned arbitrarily as number 1 in Figure 3A). The two oxygen atoms of the bidentate Glu₁₂ ligand (atoms numbered 2 and 3 in Figure 3A) occupy adjacent positions in the equatorial plane of the complex. This is the key feature of the complex that sets the direction of approach for the side chain of Glu₁₂ and consequently the position of the exiting helix with respect to the β-scaffold (approximately 108° as marked by the arrows). A similar bidentate coordination of Mg²⁺ by Glu₁₂ would not be possible because it would require the exiting helix to move much farther away from the β-scaffold (marked with a dashed arrow in Figure 3A) and possibly clash with the incoming helix. A change to a monodentate interaction of Glu₁₂ positions that residue at ~90° with respect to the carbonyl oxygen ligand and also permits more flexibility to the Glu₁₂ side chain and ultimately to the exiting helix.

The second reason for the switch from bidentate for Ca²⁺ to monodentate for Mg²⁺ Glu₁₂ -metal ion interaction may be related to the difference in the Me-O bond lengths and, consequently, Me-O-C bond angles. Analysis of the Ca²⁺-carboxylate interactions in crystal structures of calcium complexes indicates a relatively broad distribution of bond lengths (2.3–2.7 Å) and angles (110° 170°) for monodentate Ca²⁺-carboxylate interactions. In contrast, a bidentate Ca²⁺-carboxylate interaction sets the Ca²⁺-O-C bond angle at 90° [97]. Due to the smaller ionic radius, thus a shorter O-Me distance for Mg²⁺ the Mg²⁺-O-C angle would have to be significantly less than 90° for bidentate interaction, which apparently is unfavorable. Analysis of high-resolution X-ray crystallographic structures of small molecules containing carboxyl-metal ion complexes indicates that bidentate interaction is virtually excluded for carboxyl-Mg²⁺ interaction. In contrast, Ca²⁺ frequently forms bidentate complexes with carboxylates [98].

Functional implications

There are several ways in which Mg²⁺ may affect the function of EF-hand proteins. The simplest effect is a decrease in the apparent binding constant for Ca²⁺ due to Mg²⁺ competition for the binding sites, which in turn will cause a shift of the target enzyme activation curve to higher Ca²⁺ concentrations often referred to as a decrease in Ca²⁺ sensitivity . The extent of the effect depends on the Mg²⁺ binding constant and the Mg²⁺ concentration and can be calculated from the relation: K _Ca=K_Ca/(1+K_Mg.[Mg²⁺]), where K_Ca and K _Ca are the binding constants for Ca²⁺ in the absence and presence of Mg²⁺, and K_Mg is the binding constant for Mg²⁺. This relation holds true for a single, or independent site. In reality, the relation is likely to be more complex due to the fact that Ca²⁺ binds cooperatively to CaM, TnC and other EF-hand proteins. Furthermore there is evidence for negative allostery between Mg²⁺ and Ca²⁺ [57, 87], as reflected in the faster dissociation of Ca²⁺ from the mixed Ca²⁺ /Mg²⁺ species. The problem of predicting Mg²⁺ effects is further compounded by the various preferences of the interacting target molecules (eg. ref. [56]). In view of the data in Fig. 2 and other published observations, Mg²⁺ clearly stabilizes the closed domain conformation of CaM. This apparently is the preferred conformation for the group of targets that utilize the so-called IQ domain for CaM binding [99]. Recent studies by Persechini and colleagues suggest a switching mechanism in which each domain of CaM plays a specific and different role in each of its conformations [100]. It is likely that Mg²⁺ may further diversify the Ca²⁺ signaling in these proteins due to specific modulation of the Ca²⁺ response of each domain [56]. Moreover, one may expect rather complex responses for various EF-hand protein-target systems in environments where Mg²⁺ concentrations change.

While it is important to consider the effects of Mg²⁺ on the EF-hand target interaction in order to interpret the in vitro binding parameters, the absolute effect of Mg²⁺ is perhaps of lesser importance from the physiological point of view. The important question is if small changes in intracellular Mg²⁺ concentrations might be expected to alter in any physiologically relevant manner the intracellular Ca²⁺-signaling. In other words how big an effect has to be in order to be physiologically relevant? Candidates include systems requiring very tight regulation of cellular responses. For example, in familial hypertrophic cardiomyopathy a small (pCa<0.1) increase in Ca²⁺ sensitivity of TnC regulated actin filaments leads to diminished quality of life and premature death [101]. Indeed, the function of the heart is not only regulated directly by precise Ca²⁺ transients, but also modulated in many ways by diverse signaling pathways, some of which are sensitive to Ca²⁺ [102]. Small effects at the many checking points of this complex system caused by changes in Mg²⁺ concentration may have a significant cumulative effect on the function of the heart. Other systems that seem likely to be affected by changing Mg²⁺ concentrations are those involving Ca²⁺-regulated cascades such as the CaM-kinase cascade [103], or the integrative systems that respond to cumulative effects of multiple small Ca²⁺ transients.

The total magnesium concentration in mammalian cells is 15–20 mM, thus making Mg²⁺ the second most abundant cellular cation after potassium. However, only a small fraction (~1 mM) of the total Mg²⁺ in cells is in the form of free ion [95]. Mg²⁺ is a cofactor in many intracellular processes by virtue of binding to ATP and other nucleotides. It is also involved in the catalytic function of many enzymes, contributes to the stability of proteins and mediates some specific protein-protein interactions. The intracellular Mg²⁺ concentrations are hormonally regulated [45, 104, 105], suggesting that Mg²⁺ homeostasis is of vital importance. There is extensive literature documenting a correlation between cellular or dietary Mg²⁺-deficiency and a number of pathological conditions [106]. In particular, Mg²⁺ deficiency has been correlated with cardiovascular diseases including heart attack, specific forms of cardiac arrhythmias, angina and congestive heart failure, as well as hypertension, inflammation, asthma, metabolic syndrome and many others [107–110]. In some studies the protective effect of Mg²⁺ on cardiovascular system has been compared favorably to those of statin pharmaceuticals [111]. Although the important role of intracellular Mg²⁺ is recognized, the underlying biochemical and molecular mechanisms are not well understood.

Conclusion

The structural analysis presented in this review indicates that Mg²⁺ and Ca²⁺ have opposing effects on the structure and function of calmodulin, troponin C and related EF-hand Ca²⁺-binding proteins. Therefore Mg²⁺ may be an important factor in modulating the Ca²⁺-sensitivity and, in particular, switching off some Ca²⁺-regulated systems at the resting cellular Ca²⁺ levels. Although at this point it would be premature to make any specific connections, it is likely that some pathological conditions attributed to Mg²⁺ deficiency are related to excessive activation of underlying Ca²⁺-regulated cellular processes.