Elucidating the mechanisms of cooperative calcium-calmodulin interactions: a structural systems biology approach
© Valeyev et al. 2008
Received: 30 July 2007
Accepted: 02 June 2008
Published: 02 June 2008
Skip to main content
© Valeyev et al. 2008
Received: 30 July 2007
Accepted: 02 June 2008
Published: 02 June 2008
Calmodulin is an important multifunctional molecule that regulates the activities of a large number of proteins in the cell. Calcium binding induces conformational transitions in calmodulin that make it specifically active to particular target proteins. The precise mechanisms underlying calcium binding to calmodulin are still, however, quite poorly understood.
In this study, we adopt a structural systems biology approach and develop a mathematical model to investigate various types of cooperative calcium-calmodulin interactions. We compare the predictions of our analysis with physiological dose-response curves taken from the literature, in order to provide a quantitative comparison of the effects of different mechanisms of cooperativity on calcium-calmodulin interactions. The results of our analysis reduce the gap between current understanding of intracellular calmodulin function at the structural level and physiological calcium-dependent calmodulin target activation experiments.
Our model predicts that the specificity and selectivity of CaM target regulation is likely to be due to the following factors: variations in the target-specific Ca2+ dissociation and cooperatively effected dissociation constants, and variations in the number of Ca2+ ions required to bind CaM for target activation.
Calmodulin (CaM) is a multisite and multifunctional protein that contains four EF-hand Ca2+ binding sites , and is involved in a wide variety of cellular functions . For example, it regulates the concentration of intracellular cAMP concentration in a very complex manner by regulating activities of cAMP producing adenylate cyclases (AC) and cAMP hydrolysing enzyme phosphodiesterase (PDE). CaM also regulates a large number of kinases and phosphatases as well as other enzymes with opposing cellular effects. Despite a large number of experimental studies [3–15], the detailed mechanisms underlying CaM-dependent intracellular regulation of such a large variety of target proteins are still not fully understood.
Recent experimental studies have attempted to elucidate the mechanisms underlying Ca2+/CaM-dependent target regulation by measuring the kinetics and steady-state levels of CaM-target binding [3, 11, 31–33] as well as by analysing the mechanisms of Ca2+-CaM interactions [9, 12, 34–37]. Ca2+ ion binding to EF-hand sites was shown to lead to CaM conformational alterations [1, 38–41]. In the modified conformational state, CaM is likely to alter its affinity to different targets by increasing and decreasing its affinity to certain proteins. Ca2+ ion binding to CaM is also argued to positively modulate the affinity of other Ca2+ binding sites of the molecule. There is still, however, an ongoing debate about the existence, the mechanisms and the degree of cooperativity in Ca2+-CaM interactions. In some studies, Ca2+ binding to CaM has been reported to be independent [15, 37]. On the other hand, other studies have reported cooperative interactions between the neighbouring EF-hand binding sites [12, 35] or cooperativity linking all sites of the CaM molecule .
A number of previous studies have attempted to use mathematical modelling to obtain a quantitative understanding of the mechanisms involved in Ca2+-CaM interactions. Different mathematical models, including the well known Hill , Adair  and Monod-Wyman-Changeux (MWC)  models, have been used in the literature to describe the cooperativity of ligand binding to a multisite protein. The Hill equation is frequently used to qualitatively measure the degree of cooperativity in multisite binding. It describes the simultaneous binding of n ligand molecules to a protein where the parameter n can be interpreted as the number of bound molecules. The Adair model represents ligand-protein interactions in terms of successive binding steps. The MWC model is based on two conformations that are in equilibrium and have different affinities for a ligand. To date, the Hill and Adair models have been most frequently used to investigate Ca2+ binding to CaM [4, 9–12, 14, 33, 45]. While these studies have provided much useful information, the use of the classical models mentioned above also introduces some limitations in the analysis – see  for a full discussion of this issue. In particular, the detailed analysis of ligand-protein interactions which are unique to CaM requires the development of a model that captures multiple functionally important intermediate conformations of the protein. A steady-state solution to the cooperativity problem for Ca2+ binding has been analysed in . In this paper, we develop a new model, based on the assumption that the specificity in CaM target regulation arises from the Ca2+-CaM complex specific target interactions with variable numbers of bound Ca2+ ions. In this approach, Ca2+ binding to each EF-hand sites causes conformational transitions in the CaM molecule leading to a model that has multiple conformational states in complex with variable numbers of Ca2+ ions. In the proposed model, CaM may regulate its targets with one, two or three Ca2+ ions as well as in the apo- or fully bound states. In particular, we address the Ca2+-CaM interaction in significant detail, although we do not incorporate detailed Ca2+-CaM species interactions with target proteins. This approach is in agreement with recent experimental evidence that the concentration-dependent profiles for several Ca2+-CaM-dependent protein targets exhibit quite a diverse range of behaviour. PMCA and PDE protein concentrations in the active state, for example, reveal "Hill-shape"-like curves, whereas the ACII isoform is inhibited by increasing Ca2+ concentration. The ACVI isoform exhibits inhibition with an interesting plateau feature on the Ca2+-dependent profile. Yet ACI isoforms have bell-shaped concentration-dependent profiles . It has also been shown that CaMPKII  as well as the K+ channel from Paramecium  are activated by CaM with two bound Ca2+ ions.
The structural systems biology approach, , employed in this paper provides new insights into the Ca2+-CaM-target binding dose-response curves which have been derived experimentally, and allows us to advance testable hypotheses about the nature of cooperative mechanisms unique to calcium-CaM interactions. The resulting analysis further bridges the gap between our understanding of CaM structural properties and intracellular Ca2+-CaM-dependent target regulation.
In order to investigate the dependence of one Ca2+ binding site on another in both the N- and C-terminal domains, we assumed the alteration of a dissociation constant when a neighboring site is occupied as illustrated schematically in Figure 3 (compare the Model 1 and Model 2 descriptions in the Materials and Methods section). This approximation allows the derivation of a model that has an analytical solution in the form of conditional probabilities (Equation 12). It provides a quantitative comparison of the concentration of Ca2+ bound to CaM in the presence and absence of Ca2+ binding site cooperative interactions (Equations 13–14). Figure 5A shows the model predictions in the case where CaM is assumed to have two pairs of independent EF-hand globular domains. Within these domains, one Ca2+ binding site influences the other. In the N-terminal domain, the affinity of the second site depends on the state of the first and changes from K 2 = 0.9 to = 0.2 (μM l -1) when a Ca2+ ion occupies the first centre. In the C-terminal, the affinity of the fourth site depends on the state of the third and changes from K 4 = 0.8 to = 0.1 (μM l -1) when a Ca2+ ion occupies the third centre. The model (Model 2 in Materials and Methods) predicts that such changes will mainly influence the "amplitudes" of the intermediate conformations of the concentration-dependent profiles while leaving the ligand concentrations that produce their maximum values largely unchanged. Figure 5B shows the difference in the total amount of bound ligand with and without the type of cooperativity described above. Since the amount of bound Ca2+ is frequently measured in Ca2+-CaM or other ligand-multisite protein interaction experiments, these results allow a direct quantitative comparison to be made of the binding reactions with and without the presence of cooperative binding.
A more realistic description for EF-hand Ca2+ binding sites would involve the incorporation of the influence of both Ca2+ binding sites on each other within the CaM globular domains as schematically illustrated on Figure 5 (Model 3 in Materials and Methods). The resulting system of differential equations describing the Ca2+-CaM interactions is given by (Equation 15) in Materials and Methods. The model based on these assumptions leads to some interesting predictions regarding Ca2+-dependent CaM interactions with various CaM target protein peptides [33, 49] and reveals a complex story of specificity in CaM regulation. While it is well established that Ca2+ ions are required to modulate the CaM-target protein interactions, the mechanism of Ca2+-induced CaM conformational transitions that allow selective interactions with a particular target protein is still unclear. The presented model for Ca2+ binding to CaM provides new insights into how the cooperative interactions between EF-hand binding sites contribute to the mechanism of selective target regulation by CaM, as described below.
The results of the present study, when combined with previous experimental data from the literature, suggest that CaM interacts with phosphorylase kinase, CaATPase and skMLCK in the apo state, but activates these proteins only when Ca2+ ions bind to a CaM-target protein complex. CaMKII kinase, on the other hand, binds to the Ca2+-CaM complex rather than apo-CaM. Each kinase has a unique combination of K 1 and K 2 dissociation constants. The half length peptides of CaATPase and skMLCK appear to have an even more complex mechanism of binding. As mentioned earlier, each Ca2+-CaM complex (with variable numbers of Ca2+ ions bound) may have a unique set of K 1 and K 2 constants for a target protein or peptide. For simplicity, all possible CaM species were divided into two groups: i) those with less than 3 Ca2+ ions bound and ii) those with three and four Ca2+ ions bound to CaM. A comparison of the model predictions with the Ca2+-CaM-parts of skMLCK and CaATPase peptide binding data (Figure 7B and 7C) suggests that both the K 1 and the cooperatively influenced K 2 dissociation constants are different when peptides are bound to CaM species with less than 3 Ca2+ ions or to CaM species with 3 or 4 ions. The conclusion from this observation is that the specificity in Ca2+-CaM-dependent regulation arises from a combination of the target specific affinity between Ca2+ and CaM, target specific cooperative constants, the order of the Ca2+-CaM-target complex assembly, as well as the number of Ca2+ ions bound to CaM. All these factors contribute to the mechanism of selective Ca2+-CaM dependent regulation in addition to the diversity of CaM-target interfaces .
Ca2+-CaM dissociation constants derived by the different mathematical models.
KD1 = 0.24, KD2 = 13
K1 = 1, Kc1 = 1, K2 = 1, Kc2 = 1
KD1 = 0.02, KD2 = 0.08
K1 = 0.04, Kc1 = 0.02, K2 = 0.04, Kc2 = 0.02
KD1 = 0.26, KD2 = 6
K1 = 1.2, K2 = 0.5 K1 = 5, K2 = 1
KD1 = 3.4, KD2 = 4
K1 = 0.06, Kc1 = 0.02, K2 = 0.06, Kc2 = 0.02
KD1 = 0.09, KD2 = 0.2
K1 = 0.15, Kc1 = 0.05, K2 = 0.15, Kc2 = 0.05
KD1 = 0.12, KD2 = 3.9
K1 = 2, Kc1 = 1, K2 = 2, Kc2 = 1
KD1 = 0.66, KD2 = 2.4
K1 = 0.4, Kc1 = 0.2 K2 = 2, Kc2 = 1
K1 = 0.5, Kc1 = 0.5, K2 = 0.5, Kc2 = 0.5
K1 = 5, Kc1 = 5, K2 = 5, Kc2 = 5
CaM pH = 7.2
K1 = 0.34, K2 = 0.36, K3 = 0.13, K4 = 0.06
K1 = 17, Kc1 = 7, K2 = 20, Kc2 = 0.5
K1 = 0.142, K2 = 0.062
K1 = 17, Kc1 = 7, K2 = 17, Kc2 = 7
K3 = 0.0543, K4 = 1.82
K1 = 20, Kc1 = 0.5, K2 = 20, Kc2 = 0.5
CaM pH = 6
K1 = 10, Kc1 = 5, K2 = 10, Kc2 = 5
CaM pH = 10.1
K1 = 2, Kc1 = 1.8, K2 = 2, Kc2 = 1.8
In this paper a structural model of CaM interactions with and without cooperativity has been used to elucidate the mechanisms of Ca2+-CaM-target complex assembly. The differences seen in dose-response curves for proteins activated by Ca2+-CaM pairs were explained in terms of cooperative interactions between the EF-hand pairs [38, 40, 41] in both CaM domains. This study predicts that the specific interaction interface between CaM and CaM-regulated proteins [8, 10, 39] is complemented by a number of additional factors influencing the Ca2+-CaM-target complex assembly. By comparing our model predictions with experimentally measured dose-response curves from the literature, we propose that some proteins bind CaM without Ca2+ ions and only become activated when Ca2+ ions interact with the CaM-target complex, whereas others are activated by CaM molecules with already bound Ca2+ ions. The Ca2+-CaM interaction properties are tuned by the target proteins and characterized not only by the macroscopic dissociation constant set for Ca2+ sites, but also by the macroscopic cooperatively altered dissociation constants that are also unique to the CaM binding proteins. In other words, the order of Ca2+-CaM-target complex assembly, the number of bound Ca2+ ions, target specific Ca2+-CaM cooperative affinities, in addition to unique CaM-target interaction interfaces, all allow CaM to achieve its highly versatile intracellular multifunctionality. This proposition also explains the effects of pH on the considered dose-response curves by allowing for the modulation of the cooperatively effected dissociation constants. We would also like to point out that while we addressed the Ca2+-CaM interactions in great detail, the model could still be developed further by incorporating detailed dissociation constants between the intermediate Ca2+-CaM complexes and target proteins in a similar way to how it has been done for the Ca2+-CaM interactions.
Although the presented model predicts similar curves to the ones already used to approximate the experimental Ca2+-CaM-target dose-response data using the Hill and Adair models, it allows a far more detailed interpretation of the Ca2+-CaM dependent interactions involved. In particular, it distinguishes the structure-dependent properties of CaM molecules and suggests potential scenarios for Ca2+-CaM-target complex assembly. Importantly, it reveals the CaM specific type of cooperativity involved in this process and helps to explain the contribution of cooperativity in the specificity of CaM-dependent regulation.
We propose a number of conclusions from this study which, we believe, contribute to improving our understanding of intracellular CaM regulation and provide useful hypotheses for further experimental validation.
(1) Mathematical models for protein interactions are usually derived according to a number of assumptions which will inevitably be more or less applicable to each particular protein. The structure of CaM suggests that this molecule is very likely to have non-sequential Ca2+ access to EF-hand binding sites. The results of our analysis support the theory of non-sequential cooperative access of Ca2+ to CaM binding sites, and also allow the derivation of cooperatively effected dissociation constants, thus providing a more realistic tool for fitting experimental dose-response curves.
(2) Our model suggests that the structural data alone cannot provide the required level of information and comparisons with the dose-response data are required. Predictions from the mathematical model used in this study were compared with the dose-response curves for Ca2+ binding to CaM and Ca2+-CaM-target peptides. This analysis allowed us to distinguish between (a) proteins that form a complex with CaM in its Ca2+ free state and then interact with Ca2+ ions and (b) other proteins which interact with Ca2+ bound CaM with variable numbers of Ca2+ ions. However, the transient kinetics has not been addressed in this study.
(3) In addition to the diversity of interaction interfaces, the specificity and selectivity of CaM target activation may be achieved by variations in the target-specific dissociation and cooperatively effected dissociation constants, the order of Ca2+-CaM-target complex assembly and the number of Ca2+ ions required to bind CaM for target activation.
Here, we present the mathematical equations used to describe Ca2+-CaM interactions in our study. To clarify how various factors contribute to the CaM-dependent regulation, we describe three models for CaM starting from a very basic approximation of completely independent Ca2+ binding sites and gradually progressing to more realistic models that take account of cooperativity mechanisms and the Ca2+-CaM-target peptide complex assembly.
where cm0 is the total concentration of CaM, u is normalised Ca2+ concentration, is the probability of binding site i not being occupied and is the probability of binding site i being bound. c i equals 1 if a binding site is occupied and 0 if it is not. The probability of CaM being in a particular bound state is equal to the product of the probabilities of each individual binding site.
where the K and u are the microscopic equilibrium dissociation constant and the ligand concentration, respectively. Effectively these are Michaelis-Menten equations for a protein in a complex with and without a ligand molecule, but normalized by the total protein concentration.
where the K and u are the equilibrium dissociation constant and the ligand concentration, respectively.
A complete mathematical description of the relationship between macroscopic constants derived from the Adair equation and the proposed model is provided in the Supplementary Materials Section (Additional File 1).
where L0 is the total protein concentration and L i is the concentration of conformation i, K j are the equilibrium dissociation constants of each binding site, and u is the ligand concentration.
The conformations L 1(u), ..., L n-1(u) of a multisite protein are all bell-shaped curves, the conformation L 0(u) is the apo state of a multisite protein, whereas the L n (u) is the fully bound multisite protein. If equation (5) is divided by L0, then instead of predicting protein concentrations in specific ligand-bound conformations, it predicts the probability of a particular conformation to be in that state as a function of ligand concentration.
where cm i (u) is substituted from equation 5, and K d is the dissociation constant for CaM-target interactions, N0 is the total concentration of target protein.
While the previous model provides predictions for the number of Ca2+-CaM complexes as a function of Ca2+ concentration with a reasonable accuracy, it does not capture the effects of the cooperative influence of Ca2+ binding sites. There are several possible ways to incorporate these cooperative mechanisms into the model. In order to derive a model that illustrates what contribution cooperativity makes to the distribution of concentration profiles of Ca2+-CaM complexes, we assume that in the N-terminal domain, the first centre is cooperatively bound to the second, and in the C-terminal, the third is cooperatively bound to the fourth. In this case, we will define the dissociation constants as K 1, K 2, , K 3, K 4, , where and are the cooperatively influenced dissociation constants for the second and fourth centres when the ligand is bound to the first and third binding sites, correspondingly.
where K 1 and u are the equilibrium dissociation constant for the first centre and the ligand concentration, respectively.
where is the probability for both the first and the second centres to be free, is the probability for the first site to be free and the second to be occupied, is the probability for the first site to be bound and the second to be free, and is the probability for both sites to be ligand bound. K 2 and are the dissociation and the cooperatively modified dissociation constants for the second Ca2+ binding site.
where K 3 is a dissociation of the third binding site.
where is the probability for both the third and the fourth centres to be free, is the probability for the third site to be free and the fourth to be occupied, is the probability for the third site to be bound and the fourth to be free, and is the probability for both sites to be ligand bound. K 4 and are the dissociation and the cooperatively modified dissociation constants for the second Ca2+ binding site.
where k i = 0, if the binding site i is not occupied and k i = 1 if the centre i is occupied by a ligand molecule.
The distribution of intermediate Ca2+-CaM complexes with 1, 2 and 3 Ca2+ ions with and without cooperativity is shown in Figure 5A.
Figure 5B shows the total amount of ligand bound to CaM in the presence (1) and absence (2) of cooperative binding. The line (3) shows the difference in the level of bound ligand between the two types of binding mechanisms.
where cm 00, cm 01, cm 10, cm 11 are CaM molecules without Ca2+ ions, with one Ca2+ ion bound to the N-terminal domain, with one Ca2+ ion bound to the N-terminal domain, and CaM species with two bound Ca2+ ions at each terminal domain, respectively. k 1 and k 2 are the association constants and kc 1 and kc 2 are the cooperatively modified association constants for the N- terminal binding sites of CaM, respectively. Similarly, d 1, d 2, dc 1, dc 2 are the dissociation and cooperatively modified dissociation constants for the N- and terminal binding sites, respectively. Note that a similar system of differential equations can be developed for the C-terminal.
The conservation law gives:
cm00 +cm10 + cm01 + cm11 = 1
where , . k 1 = k 2 = kc 1 = kc 2 is a simplifying assumption that allows analytical solution of the current system.
where N 00, N 01, N 10 and N 11 correspond to the N-terminal domain without any Ca2+ ions bound, with one Ca2+ ion bound to one or another binding site and to the fully bound state, respectively.
where is the probability for the N-terminal to be in a Ca2+ state, and describe the probabilities for a complex with one Ca2+ ion, and is the probability function for the N-terminal to be occupied by Ca2+.
where is the probability for the N-terminal to be in a Ca2+ state, and describe the probabilities for a complex with one Ca2+ ion, and is the probability function for the N-terminal occupied by Ca2+.
where are CaM species in apo state, is fully bound CaM, and and are CaM species with fully bound N- and C-terminals, respectively. and are CaM species with one Ca2+ ion bound to N- and C-terminals, respectively. Comparison of numerical solutions of this system with the available experimental data [9, 15, 33] allows us to propose that CaM molecules can be represented as a pair of two independent N- and C-terminal globular domains, each containing two symmetrical and cooperatively bound EF-hand Ca2+ binding sites.
The equations (28) are specific to Ca2+-CaM interactions and incorporate the pairwise cooperative interactions between the EF-hand binding sites within the N- and C-terminals, whereas the N- and C-terminal domains are considered to be independent of each other. Note that the equations developed here for the Ca2+-CaM complexes are essentially different from the Michelis-Menten, Hill and Adair models, and also differ from models with independent binding sites (5) or with limited amounts of cooperativity (11) and (12).
where T0 is the total concentration of a target protein or a peptide, K d is the equilibrium dissociation constant between CaM and a target protein, and is a single Ca2+ complex or the sum of several CaM complexes.
The combinations of Ca2+-CaM complexes have been varied simultaneously with the dissociation constants to fit the experimental data. This analysis allows us to predict the Ca2+-CaM complexes required for activation of specific protein targets. The fitting of dissociation constants of Ca2+ binding sites on CaM molecules to the experimental dose-response curves reveals the impact of the target protein on the Ca2+-CaM interactions. The dissociation constants calculated based on the cooperative Model 3, which also takes into account the impact of target proteins, are compared with the dissociation constants calculated using the Hill and modified Adair equation in the original experimental publications for the same data in Table 1.
This work was carried out under EPSRC platform grant (EP/D029937/1), BBSRC grant (BB/D015340/1) and by RFBR grant.
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.