Ca²⁺/calmodulin-dependent protein kinase II (CaMKII) is activated by calmodulin with two bound calciums

Abstract

Changes in synaptic strength that underlie memory formation in the CNS are initiated by pulses of Ca²⁺ flowing through NMDA-type glutamate receptors into postsynaptic spines. Differences in the duration and size of the pulses determine whether a synapse is potentiated or depressed after repetitive synaptic activity. Calmodulin (CaM) is a major Ca²⁺ effector protein that binds up to four Ca²⁺ ions. CaM with bound Ca²⁺ can activate at least six signaling enzymes in the spine. In fluctuating cytosolic Ca²⁺, a large fraction of free CaM is bound to fewer than four Ca²⁺ ions. Binding to targets increases the affinity of CaM's remaining Ca²⁺-binding sites. Thus, initial binding of CaM to a target may depend on the target's affinity for CaM with only one or two bound Ca²⁺ ions. To study CaM-dependent signaling in the spine, we designed mutant CaMs that bind Ca²⁺ only at the two N-terminal or two C-terminal sites by using computationally designed mutations to stabilize the inactivated Ca²⁺-binding domains in the “closed” Ca²⁺-free conformation. We have measured their interactions with CaMKII, a major Ca²⁺/CaM target that mediates initiation of long-term potentiation. We show that CaM with two Ca²⁺ ions bound in its C-terminal lobe not only binds to CaMKII with low micromolar affinity but also partially activates kinase activity. Our results support the idea that competition for binding of CaM with two bound Ca²⁺ ions may influence significantly the outcome of local Ca²⁺ signaling in spines and, perhaps, in other signaling pathways.

Calmodulin (CaM) is a Ca²⁺ effector protein comprised of two Ca²⁺-binding lobes connected by a short linker region. Each lobe contains two EF hands that bind Ca²⁺ with low micromolar affinity. CaM mediates a host of regulatory effects of Ca²⁺ despite the fact that the Ca²⁺ concentration in the cytosol rarely reaches a level expected to saturate all four Ca²⁺-binding sites on free CaM (1). This sensitivity is possible because association of CaM with its targets stabilizes its Ca²⁺-bound conformation and, therefore, increases its affinity for Ca²⁺ (2–4). When CaM with two Ca²⁺ ions bound to either the N- or the C-terminal lobe associates with a target, the affinity of the remaining lobe for Ca²⁺ increases, often by at least one order of magnitude, causing high positive cooperativity of Ca²⁺ binding (5). Because of this property, we predict that when Ca²⁺ concentrations fluctuate in the cytosol, competition among targets for binding of CaM will depend on their affinities for CaM with one or two bound Ca²⁺ and also on the magnitude of cooperativity induced by binding to each target.

In the CNS, Ca²⁺ influx into postsynaptic spines of excitatory glutamatergic synapses is a key determinant of changes in synaptic strength that underlie memory formation. The direction of the change in strength is determined by the timing between release of glutamate at the synapse and depolarization of the postsynaptic membrane, caused, for example, by a back-propagating action potential. This timing controls the duration and extent of Ca²⁺ influx through NMDA-type glutamate receptors into the spine. Relatively large increases of a few micromolars in Ca²⁺ concentration occurring over 2 to 3 sec induce long-term potentiation of the synapse. In contrast, smaller changes of a few hundred nanomolars, lasting over several seconds to 1 min, induce long-term depression of the synapse (6–10).

How might these slight differences in Ca²⁺ influx produce such dramatic differences in regulation of the synapse? Ca²⁺ that flows into a spine through activated NMDA receptors binds to CaM and to several other Ca²⁺-binding proteins. Ca²⁺ binding to CaM enables its interaction with, and regulation of, several postsynaptic proteins. Two prominent targets of Ca²⁺/CaM are Ca²⁺/calmodulin-dependent protein kinase II (CaMKII) (11, 12) and calcineurin, a protein phosphatase (13, 14). The apparent affinities of these two proteins for Ca²⁺/CaM are quite different when measured at saturating Ca²⁺ concentrations (>200 μM). CaMKII is half maximally activated at ≈40–80 nM CaM (15–17), whereas calcineurin is half maximally activated at 1–10 nM CaM (18, 19). This severalfold difference in affinity for Ca²⁺/CaM has been invoked to explain the different synaptic effects of high and low Ca²⁺ fluxes (20). However, an accurate description of the critical initial steps of Ca²⁺ signaling in spines requires a precise understanding of how different CaM targets in the spine compete for Ca²⁺ under conditions in which quantities of Ca²⁺/CaM are limited and concentrations of Ca²⁺ are fluctuating.

In the work reported here, we focus on understanding interactions between CaM and CaMKII when the concentrations of Ca²⁺ or CaM are limiting. CaMKII is a dodecameric oligomer of catalytic subunits, each of which can bind one CaM (21–23). Binding of CaM to an individual subunit activates its kinase activity allowing the subunit to phosphorylate other proteins. In addition, binding of two CaM molecules to two adjacent subunits within a holoenzyme allows autophosphorylation of one or both of the subunits at Thr-286 (24). This autophosphorylation renders the subunit constitutively active and it remains so until the phosphate is removed by a protein phosphatase (25, 26). Activation and autophosphorylation of CaMKII has been shown to be crucial for induction of long-term potentiation (27).

To measure the affinity of CaMKII for CaM with less than four bound Ca²⁺ ions, we designed a mutant CaM that binds Ca²⁺ only at N-terminal sites (CaM-N^WT) and a mutant CaM that binds Ca²⁺ only at C-terminal sites (CaM-C^WT). To closely mimic conformations of CaM that would exist in the cytosol, we used computationally designed mutations to stabilize the inactivated Ca²⁺-binding domains in the “closed,” Ca²⁺-free conformation. We found that, when saturated with Ca²⁺, the mutant CaMs bind to CaMKII with low micromolar affinity and partially activate its kinase activity. Thus, the activity of CaMKII is more sensitive to small increases in cytosolic Ca²⁺ than previously believed. The mutant CaMs described here can be used to study the ability of CaM with two Ca²⁺ ions bound to the C- or N-terminal sites to associate with and activate other CaM targets. Such studies will enable more accurate understanding of the regulatory outcomes of local Ca²⁺ increases in many signaling pathways and cell types.

Results

Activation of CaMKII by Ca²⁺ and CaM Is Highly Cooperative.

Conditions usually used to study activation of CaM targets in vitro are necessarily different from those existing in cells. Most published assays of activation of proteins by CaM employ CaM concentrations in the range of 1 μM and concentrations of targets in the range of 20–50 nM. The actual concentrations of CaM and CaMKII in spines are estimated to be in the range of 10–30 μM (21, 28, 29), whereas fluctuations in the concentration of Ca²⁺ range over 0.1–30 μM near open NMDA-type receptor pores (30). To better approximate conditions under which competition for Ca²⁺/CaM might occur in a spine, we measured activation and autophosphorylation of CaMKII in reactions containing 0.28 μM CaMKII (molarity of catalytic subunits) and 12 μM CaM, concentrations that are 5- to 10-fold higher than those commonly used in enzymatic assays and yet are still practical. We observed a significant rate of autophosphorylation of CaMKII at as low as 350 nM Ca²⁺ (Fig. 1A), a concentration below the K_D of even the highest-affinity Ca²⁺-binding site measured for free CaM (≈1 μM) (1, 31) and two orders of magnitude below the K_D of the lowest affinity site (≈20–30 μM). These results suggest one or both of the following mechanisms. First, the cooperativity of binding among Ca²⁺, CaM, and CaMKII might be extremely high; that is, the presence of CaMKII might increase the apparent affinity of CaM for Ca²⁺ by more than one order of magnitude. Second, CaM might not require occupation of all its high-affinity Ca²⁺-binding sites to initiate autophosphorylation of CaMKII.

Fig. 1.

Cooperative binding of Ca²⁺ to CaM in the presence of CaMKII. (A) Autophosphorylation of CaMKII in the presence of WT CaM at varying Ca²⁺ concentrations. Reactions were performed for 1 min, as described in Methods, with 10 μM CaM, 0.28 μM CaMKII, and the indicated Ca²⁺ concentrations. Autophosphorylation was detected by immunoblotting with the phosphosite-specific antibody 22B1, as described in Methods. The position of the α-subunit of CaMKII is indicated. Note that the band shifts to slightly slower mobility as a second site, Thr-305, becomes autophosphorylated. (B) Increase in affinity of Ca²⁺ for CaM in the presence of CaMKII or CaMKII-cbp. Ca²⁺ binding to 5 μM CaM was measured in a competition assay with a fluorescent Ca²⁺-binding dye, Fluo4FF at 25°C. Ca²⁺ was titrated into solutions containing 5 μM Fluo4FF and WT CaM alone (▴), or in the presence of 5 μM CaMKII (■), or 5 μM CaMKII-cbp (♦). The data were fit with equations for Ca²⁺ binding at equilibrium to CaM alone or to CaM in the presence of CaMKII or CaMKII-cbp as described in Methods.

The magnitude of the increase in affinity of CaM for Ca²⁺ when CaM binds to CaMKII has not been measured previously under conditions approximating those in spines. We therefore compared the affinity of free CaM for Ca²⁺ to that of CaM in the presence of CaMKII and in the presence of a peptide having the sequence of the CaM-binding domain of CaMKII (CaMKII-cbp), at concentrations of 5 μM for each protein (Fig. 1B). We found that the concentrations of Ca²⁺ at which half of the Ca²⁺-binding sites on CaM are occupied are ≈15 μM for free CaM, ≈3–4 μM in the presence of CaMKII, and ≈1–2 μM in the presence of CaMKII-cbp. These data show that the presence of CaMKII increases the affinity of CaM for Ca²⁺ ≈5-fold under conditions approximating those in a postsynaptic spine. This effect undoubtedly contributes to the ability of Ca²⁺ concentrations <1 μM to activate CaMKII in our experiments; however, it is not high enough to account for it completely.

Computationally Designed Mutant CaMs That Bind Two Ca²⁺ ions at Either the C-Terminal or the N-Terminal Sites.

One traditional view held that occupation of all four Ca²⁺-binding sites on CaM is required to change the activity of most targets and, perhaps, also for high-affinity binding of CaM to targets. Discoveries of alternative modes of regulation by CaM, including some involving CaM with Ca²⁺ bound to only one of its two lobes, indicate that CaM is, in fact, more versatile in its regulatory mechanisms (32). CaM consists of N- and a C-terminal globular helical domains connected by a central flexible helix (Fig. 2A and B). Each globular domain contains two Ca²⁺-binding sites of the EF-hand variety. The C-terminal sites have ≈6-fold higher intrinsic affinity for Ca²⁺ than the N-terminal sites (31, 36); however, the on rate for Ca²⁺ binding to the N-terminal sites is ≈10-fold faster. In the absence of Ca²⁺, CaM adopts a closed conformation (33, 37, 38) in which it is not able to interact with most of its targets (Fig. 2A). Binding of Ca²⁺ to CaM causes a reorientation of helices inside the globular domains and a consequent exposure of patches of hydrophobic residues important for target recognition (34) (Fig. 2B). A shift toward this “open” CaM conformation is believed to be a prerequisite for interactions with many CaM-regulated proteins (1, 32, 39). Binding to most targets, including the CaM-binding domain of CaMKII (CaMKII-cbp), induces a conformation in which CaM wraps around the target domain, resulting in a coil-to-helix transition in the target (35, 40) (Fig. 2C).

Fig. 2.

Conformations of WT CaM and the designed mutant CaMs. (A) NMR structure of Ca²⁺-free CaM (PDB ID code 1CFD) (33). (B) X-ray crystal structure of Ca²⁺-bound CaM (PDB ID code 3CLN) (34). The hydrophobic patches important for target recognition are shown in red. Ca²⁺ atoms are shown as yellow spheres. (C) X-ray crystal structure of Ca²⁺/CaM bound to CaMKII-cbp, shown in pink (PDB ID code 1CM1) (35). (D and E) Models of the mutant CaMs designed to bind Ca²⁺ only in the C-terminal domain, CaM-C^WT (D), and only in the N-terminal domain, CaM-N^WT (E). These models were generated by combining the structure of the WT Ca²⁺/CaM domain with the structure of Ca²⁺-free CaM representing the mutated CaM domain. Mutant residues predicted by the computation to produce the best stabilized structure (Table 1) are shown in red. The figure was generated with PyMOL (DeLano Scientific, South San Francisco, CA).

To test whether binding of Ca²⁺ to both the C-terminal and the N-terminal lobes of CaM is necessary for activation of CaMKII under conditions that approximate those in the spine, we designed mutant CaMs that bind Ca²⁺ only in the C- or N-terminal lobes: CaM-C^WT and CaM-N^WT, respectively (Fig. 2 D and E and Table 1). Unlike previously reported mutant CaMs that are deficient in Ca²⁺ binding because of point mutations of Ca²⁺-ligating residues (36, 41, 42), our mutant CaMs were designed to mimic intermediate conformations that CaM might adopt in the cytosol by specifically stabilizing the closed conformation of the mutant Ca²⁺-binding loops. We used the protein design program ORBIT (43), which employs an empirical atom-based force field and a fast side-chain selection algorithm to search through different protein sequences to find the lowest energy sequence for a given protein structure. To abolish Ca²⁺ binding to a particular CaM domain, we redesigned the amino acid sequence of either the N-terminal or the C-terminal domain, leaving the other domain intact. For each calculation, a sequence of CaM residues forming the mutant Ca²⁺-binding site in the N- or C-terminal domain was optimized by using the Ca²⁺-free CaM structure as the template (ref. 33; Table 1). For each of these calculated sequences, we verified that the redesigned Ca²⁺ site had fewer residues with Ca²⁺ coordinating side chains.

Table 1.

Computationally designed N- and C-terminal mutant CaMs

CaMs	Designed positions
	Site 1					Site 2					Site 3					Site 4
	20	22	24	27	31	56	58	60	62	67	93	95	97	100	104	129	131	133	135	140
CaM WT	D	D	D	I	E	D	D	N	T	E	D	D	N	I	E	D	D	D	Q	E
CaM-CWT	T	K	Q	I	E	S	D	E	I	K	–	–	–	–	–	–	–	–	–	–
CaM-NWT	–	–	–	–	–	–	–	–	–	–	A	R	D	M	E	R	D	N	E	N

All CaM residues that directly coordinate Ca²⁺ ions with their side chains (positions 20, 22, 24, 31, 56, 58, 60, and 67 for the N-terminal domain and 93, 95, 97, 104, 129, 131, 133, and 140 for the C-terminal domain) were included in the optimization. Residues 62 and 135 also were included because they coordinate Ca²⁺ ions with their backbone carbonyls. Residues 27 and 100, which do not coordinate Ca²⁺ ions, but are in close proximity to the ion and help define the Ca²⁺-binding loop, also were optimized. Bold lettering shows mutations with respect to the WT CaM sequence. –, positions were not included in the calculation for the mutant CaM indicated.

Assessment of Folding and Ca²⁺ Binding by Mutant CaMs.

The designed mutant CaMs were expressed in Escherichia coli and assessed by far-UV circular dichroism for proper folding and Ca²⁺ binding. In the absence of Ca²⁺, both mutant CaMs retain a predominantly α-helical structure essentially identical to that of WT CaM. Upon addition of Ca²⁺, both exhibit a small change in the circular dichroism signal, as does WT CaM, reflecting Ca²⁺-induced rearrangements of the EF-hand domains (Fig. 6, which is published as supporting information on the PNAS web site). Binding of Ca²⁺ to the C-terminal sites has been shown to be responsible for most of these changes in circular dichroism spectrum (44). Consistent with that finding, the structural transition of our N-terminal mutant (CaM-C^WT) is nearly identical to that of WT CaM (Fig. 6 A and B).

We compared the Ca²⁺ binding stoichiometry of WT and mutant CaMs by electrospray mass spectrometry as described in ref. 45. In the presence of EGTA, the spectra of WT CaM and mutant CaMs contained a major peak corresponding to the molecular weight of the protein in the Ca²⁺-free state (data not shown). In the presence of saturating Ca²⁺, the WT CaM spectrum comprised a peak corresponding to Ca²⁺-free CaM and four additional peaks reflecting incremental additions of one Ca²⁺ ion. The major peak corresponds to four bound Ca²⁺ ions (Fig. 3A). For CaM-C^WT and CaM-N^WT, the major peaks correspond to one and two bound Ca²⁺ ions, as expected (Fig. 3 B and C).

Fig. 3.

High-affinity Ca²⁺-binding sites on WT and mutant CaMs. (A–C) Stoichiometry of Ca²⁺ binding to WT and mutant CaMs measured by electrospray mass spectrometry at 10 μM CaM and 200 μM Ca²⁺. (A) WT CaM. (B) CaM-C^WT. (C) CaM-N^WT. (D) Ca²⁺ binding to WT and mutant CaMs measured in a competition assay with the fluorescent Ca²⁺-binding dye, Fluo4FF. Ca²⁺ was titrated into a solution of Fluo4FF (10 μM) alone (●), Fluo4FF plus 10 μM CaM-C^WT (▿),10 μM CaM-N^WT (▴), or 10 μM WT CaM (■). The data were fit with a model assuming four high-affinity Ca²⁺-binding sites for WT CaM, two for CaM-C^WT and CaM-N^WT, and a single Ca²⁺-binding site for Fluo4FF (Table 2).

We measured the Ca²⁺-binding affinities of the mutant CaMs employing the fluorescent calcium sensor Fluo4FF. When CaM competes with Fluo4FF for available Ca²⁺, the fluorescence signal is reduced (Fig. 3D). The concentrations of Ca²⁺ and CaM over which the competition occurs reflect the affinity of CaM for Ca²⁺. The competition curves were fit with a four Ca²⁺ site model for WT CaM and a two Ca²⁺ site model for the mutant CaMs (46). Ca²⁺-binding affinities of WT and mutant CaMs range from 1 to 40 μM (Table 2) and are in approximate agreement with previously published values for WT CaM (1, 36). The data reveal positive cooperativity between the two neighboring Ca²⁺-binding sites in both mutant CaMs, as has been reported for WT CaM (1, 31). Thus, by several criteria, the two mutants behave as desired. In the presence of a saturating concentration of Ca²⁺, CaM-C^WT binds two Ca²⁺ ions in the C-terminal lobe, whereas CaM-N^WT binds two Ca²⁺ ions in the N-terminal lobe.

Table 2.

Kinetic properties of WT and mutant CaMs

CaMs	KD values for Ca2+, μM				Kact for CaMKII, μM2	kp of CaMKII, (sec−1)3
	KD1	KD2	KD3	KD4
WT CaM	7.9	1.7	35	8.9	0.04	0.96
CaM-CWT	25	1.6	–	–	5	0.064
CaM-NWT	28	3.3	–	–	20	0.12

K_D values for Ca²⁺ were calculated as described in Methods from the data in Fig. 3D with the CaLigator program (47) downloaded from www.bpc.lu.se/research/caligator. A model containing four Ca²⁺ sites was used for WT CaM, and one containing two Ca²⁺ sites was used for CaM-C^WT and CaM-N^WT. The concentration of WT CaM required for half-maximal rate of CaMKII autophosphorylation, K_act, is taken from Meyer et al. (16). K_act values for mutant CaMs were determined from the data in Fig. 4B. The intrinsic turnover rate of CaMKII autophosphorylation, k_p, was determined from data in Fig. 5. Confidence intervals (95%) for k_p values were 0.67–1.25 for WT CaM, 0.044–0.084 for CaM-C^WT, and 0.079–0.16 for CaM-N^WT.

We compared binding of WT and mutant CaMs to CaMKII-cbp at a saturating Ca²⁺ concentration by using circular dichroism to measure the coil-to-helix transition in the peptide upon binding, as previously described (ref. 48; Fig. 7, which is published as supporting information on the PNAS web site). Dissociation constants calculated by fitting the data to a 1:1 binding model show that WT CaM binds to CaMKII-cbp with a K_D of 5 nM, similar to values reported for other CaM-binding peptide sequences (48). In contrast, CaM-C^WT binds with a K_D of 70 nM, and CaM-N^WT binds with a much lower affinity K_D of 6 μM.

Activation of CaMKII by Mutant CaMs.

We tested whether the mutant CaMs support activation of CaMKII by measuring their ability to activate autophosphorylation at Thr-286. Unexpectedly, we found that, at saturating Ca²⁺ concentration, either of the mutant CaMs (10 μM) supports substantial autophosphorylation of CaMKII in a 1-min reaction (Fig. 4A). This result means that CaM does not require Ca²⁺ bound to both of its lobes to activate CaMKII. The K_act values for CaM-C^WT and CaM-N^WT under these conditions were 5 and 20 μM, respectively (Fig. 4B), compared with a K_act value of ≈50 nM for WT CaM with four bound Ca²⁺ (Table 2). These results are consistent with the higher affinity of CaM-C^WT for CaMKII-cbp and indicate that the binding of 2 Ca²⁺ ions to the C-terminal lobe of WT CaM results in more effective activation of CaMKII than binding of two Ca²⁺ ions to the N-terminal lobe.

Fig. 4.

Activation of autophosphorylation of CaMKII by WT and mutant CaMs. (A) Reactions were performed with 10 μM CaM and 0.28 μM CaMKII subunits for 1 min in the absence (−) or presence (+) of 300 μM Ca²⁺, as described in Methods. The positions of α and β subunits of CaMKII are indicated. (B) Dependence of autophosphorylation on the concentration of mutant CaMs. Autophosphorylation was performed as in A, except reactions were performed for 30 sec. To determine the K_act values reported in Table 2, the data were normalized to the level of autophosphorylation in saturating Ca²⁺/CaM for 1 min and fit with a model assuming binding of 1 CaM to 1 catalytic subunit. (C) Dependence of autophosphorylation on Ca²⁺ in the presence of WT and mutant CaMs. Reactions were performed as in A, except with 5 μM CaM, and quantified as in B. (D) Dependence of autophosphorylation on CaM concentration in the presence of WT and mutant CaMs. Reactions were performed as in A, except with 12 μM Ca²⁺, and quantified as in B. Mean ± SD of four experiments. WT (▾), CaM-C^WT (●), and CaM-N^WT (■).

To explore the ability of the mutant CaMs to activate CaMKII under conditions closer to those believed to exist in postsynaptic spines during synaptic activity, we measured autophosphorylation of CaMKII while varying Ca²⁺ concentration from 0 to 10 μM in the presence of 5 μM CaM. Autophosphorylation of CaMKII in the presence of CaM-C^WT was first detectable at ≈1 μM Ca²⁺ and in the presence of CaM-N^WT at ≈5 μM Ca²⁺ (Fig. 4C). The rate of autophosphorylation was half-maximal at 2 μM Ca²⁺ for WT CaM, at 6 μM Ca²⁺ for CaM-C^WT, and at >32 μM Ca²⁺ for CaM-N^WT. We then measured activation of CaMKII by each mutant CaM while varying their concentrations from 0.5 to 20 μM in 12 μM Ca²⁺. Under these conditions, WT CaM had a K_act of <0.4 μM, CaM-C^WT ≈7 μM, and CaM-N^WT >30 μM (Fig. 4D). These results show that, in the presence of Ca²⁺, both CaM-C^WT, which mimics WT CaM with two Ca²⁺ ions bound at its C terminus (CaM2C), and CaM-N^WT, which mimics WT CaM with two Ca²⁺ ions bound at the N terminus (CaM2N), can bind to and partially activate CaMKII in vitro under conditions, approximating those in stimulated postsynaptic spines. The results suggest that CaM2C is far more effective than CaM2N at binding and activating CaMKII in low concentrations of Ca²⁺.

We performed quench flow experiments to measure the turnover rates of autophosphorylation supported by WT and mutant CaMs at their saturating concentrations. The turnover rate supported by either mutant was ≈10% of that supported by WT CaM (Table 2 and Fig. 5), suggesting that binding of the mutant CaMs cannot produce the optimal conformational change required for a maximal rate of autophosphorylation. We conclude that both mutant CaMs support activation and autophosphorylation of CaMKII; however, their affinity for CaMKII is lower than that of WT CaM, and they support a lower enzymatic turnover number.

Fig. 5.

Determination of turnover number for autophosphorylation of CaMKII in the presence of WT and mutant CaMs. The time course of autophosphorylation of CaMKII was determined in saturating Ca²⁺/CaM. Data were gathered with a Quench Flow apparatus and normalized as described in Methods, then fit with the equation: A(1 − e^{−k_p(t − t₀)}), where k_p is the turnover rate (unknown), t₀ (a negative term) is the time at which the extrapolated autophosphorylation curve crosses the x axis (Supporting Text), and A is a scaling factor. A, k_p, and t₀ were fit simultaneously by nonlinear regression performed with the Levenberg–Marquardt method in Mathematica. The k_p values supported by each form of CaM are summarized in Table 2. Data were obtained from five experiments with WT CaM, five with CaM-C^WT, and three with CaM-N^WT. WT CaM (●), CaM-C^WT (▿), and CaM-N^WT (▴).

Discussion

We used a computational method to design mutant CaM proteins that bind two Ca²⁺ ions at either the N- or C terminus. At the inactivated terminus, we designed mutations that stabilize the closed structure of the Ca²⁺-binding loop favored when Ca²⁺ is not bound (33, 37, 38). We used these mutants to study how CaM with Ca²⁺ bound to only one of its two lobes (CaM2C or CaM2N) interacts with its target protein CaMKII. These interactions are of interest because the direction of the changes in synaptic strength that occur during storage of information in neural circuits is exquisitely sensitive to small changes in the concentration of postsynaptic Ca²⁺ that occur during repetitive electrical activity (6, 8). This sensitivity to Ca²⁺ occurs at concentrations at which the equilibrium concentration of CaM2C (reached in ≈50 msec) is the same or higher than the concentration of WT CaM with Ca²⁺ ions bound at all four sites (CaM4), as determined by simulations of Ca²⁺ binding to free CaM under conditions present in spines (ref. 49 and unpublished observations). Furthermore, because the N-terminal lobe of CaM binds Ca²⁺ with ≈10- to 100-fold higher k_on and k_off than the C-terminal lobe, as much as 20% of free CaM may be in the CaM2N conformation 5 msec after the onset of Ca²⁺ influx through NMDA receptors. Thus, competition for CaM2C or CaM2N within a synaptic spine may determine whether the synapse will be potentiated or depressed.

Using the values reported in this study we can estimate the relative concentrations of CaM2C bound to CaMKII (CaM2C·CaMKII), CaM2N·CaMKII, and CaM4·CaMKII when the Ca²⁺ concentration in the postsynaptic density is 0.5 μM or 2 μM Ca²⁺ (S.M. and M.B.K., unpublished data). We assumed concentrations of 30 μM CaMKII subunits and 10 μM free CaM and a K_D of 50 nM for binding of CaM4 to CaMKII. At 0.5 μM Ca²⁺, the concentration of CaM2C·CaMKII is ≈0.8 μM, which is ≈10 times more than the concentration of CaM4·CaMKII. Because of the low affinity of CaM2N for CaMKII, its concentration is ≈80 times lower than the concentration of CaM2C·CaMKII. At 2 μM Ca²⁺, the concentrations of CaM2C·CaMKII (2.3 μM) and CaM4·CaMKII (3.2 μM) are roughly equal; both still exceed CaM2N·CaMKII by a factor of 80. These numbers show that the enhanced affinity for the third and fourth Ca²⁺ ions bound to CaM when CaMKII is present results in formation of a significant amount of CaM4·CaMKII at 2 μM Ca²⁺. However, this mechanism doesn't appear to account quantitatively for the substantial activation of autophosphorylation at concentrations of Ca²⁺ as low as 0.35 μM (Fig. 1). We conclude that activation of CaMKII by binding of CaM2C alone likely contributes to activation of CaMKII during small increases in Ca²⁺ in the spine above the basal concentration of 80 nM such as might occur during low-frequency stimulation. This finding is consistent with the recently reported structural arrangement of the holoenzyme of CaMKII (22) (see Supporting Text and Fig. 8, which are published as supporting information on the PNAS web site). Our findings demonstrate the importance of understanding the kinetics of interactions of CaM2C and CaM2N with their various targets to correctly simulate the orchestration of biochemical responses to Ca²⁺ influx into the spine.

Previous studies have shown that the individual N- or C-terminal lobes of CaM with bound Ca²⁺ can activate certain effector proteins, phosphorylase kinase (50), myosin light chain kinase, neuronal nitric oxide synthase (51), and adenylate cyclase (52). These studies relied either on isolated N-terminal and C-terminal fragments of CaM generated by proteolysis (50, 51) or on point mutants of CaM that have lost one or more Ca²⁺-binding sites (52). More recent studies have made use of CaM with single-point mutations in each of the two N-terminal sites (CaM₁₂) or the two C-terminal sites (CaM₃₄) to show that the individual lobes of CaM have distinct functional effects on voltage-activated Ca²⁺ channels (53, 54). Although these mutants have been useful for understanding the qualitative ability of individual CaM domains to interact with target proteins, they are not ideal for determining the actual binding constants of such interactions as they occur in vivo. The reason that single-point mutations or proteolytic fragmentation may introduce structural changes in CaM that do not represent its intermediate conformations in the cytosol. These changes may influence their rates of binding to targets (55). By stabilizing the mutated Ca²⁺-binding loops of CaM in the structure assumed by apo-CaM, we have attempted to mimic as closely as possible the transient forms of CaM loaded with Ca²⁺ ions at either the N or C terminus. These mutant CaMs will be useful for understanding how Ca²⁺ controls mechanisms of synaptic plasticity that underlie learning and memory. They also may be useful for studying Ca²⁺-signaling events in other cell types.

Methods

Computational Methods.

The sequences for the mutant CaMs were designed by using the protein design program ORBIT (ref. 43; see Supporting Text). The sequences of the N- and C-terminal CaM domains were optimized with the Ca²⁺-free CaM structure as a template (33) to produce CaM-C^WT and CaM-N^WT, respectively.

Expression of CaM Proteins.

cDNAs encoding CaM-C^WT and CaM-N^WT mutants were constructed with an inverse PCR procedure beginning with the WT CaM cDNA cloned into pET-15 (Novagen, San Diego, CA). WT and mutant CaMs were expressed and purified as described in ref. 56.

Peptide Synthesis.

The CaMKII-cbp peptide corresponding to the CaM-binding domain of CaMKII (sequence -AKSKWKQAFNATAVVRHMRKLQ-) was synthesized and purified in the Caltech Peptide Synthesis Facility as described in ref. 56.

Electrospray Mass Spectrometry.

Electrospray ionization mass spectrometry was performed on an API365 LC/MS/MS system (PE SCIEX) with a nanospray ion source (Protana, Toronto, ON, Canada) operated at an infusion mode of 0.2 μl/min and a spray voltage of −1.4 kV. The data were collected in the negative ion mode. The Ca²⁺-bound samples (10 μM CaM) were diluted into 4 mM NH₄HCO₃/200 μM CaCH₃CO₂/15% CH₃OH, pH 8.0. For the Ca²⁺-free samples, CaCH₃CO₂ was replaced with 500 μM EGTA.

Ca²⁺-Binding Affinities.

Ca²⁺ binding to WT and mutant CaMs was measured by competition with the Ca²⁺-binding fluorescent dye Fluo4FF (Molecular Probes, Eugene, OR). We determined the intrinsic binding affinity of Fluo4FF for Ca²⁺ (K_D ≈ 13 μM) in the absence of CaM by titrating concentrated CaCl₂ into a 10 μM solution of Fluo4FF in 50 mM Tris/100 mM NaCl/1 mM MgCl₂, pH 7.2 at 25°C and recording fluorescence after addition of each aliquot of Ca²⁺ on a Hitachi (Tokyo, Japan) F-4500 fluorescence spectrophotometer (excitation at 488 nm and emission recorded at 516 nm). The measured fluorescence signal F was used to calculate the free Ca²⁺ concentration [Ca²⁺]_free, the concentration of Ca²⁺ bound to the dye [Ca²⁺·Dye], and the affinity of CaM for Ca²⁺ in the presence of targets as described in Supporting Text.

CaMKII Autophosphorylation Assay.

Reactions were initiated by addition of purified CaMKII (0.28 μM) to a final volume of 100 μl of 50 mM Tris/0.1 M NaCl₂/0.9 mg/ml BSA/50 μM ATP/5 mM DTE, pH 7.2, the desired concentration of WT or mutant CaM, and desired Ca²⁺ concentration determined by mixing appropriate ratios of 100 mM CaCl₂ and 100 mM EGTA as described in instructions supplied with the Ca²⁺ buffer kit (Molecular Probes, Carlsbad, CA). The buffer was warmed to 30°C for 1 min, reactions were initiated, proceeded for 1 min, and were stopped by the addition of 3% SDS/2% 2-mercaptoethanol. Samples containing 0.2 μg of CaMKII were fractionated by SDS/PAGE, transferred by electrophoresis onto a nitrocellulose membrane for 1 h at 350 mA in 25 mM Tris base/0.2 M glycine/20% methanol. Membranes were immunoblotted with 7.5 μg antibody 22B1 in 15 ml of 50 mM Tris, pH 7.4/0.5 M NaCl/0.05% Tween 20 (TTBS) plus 2% normal goat serum for 3 h at 4°C. Monoclonal antibody 22B1 (anti-phospho-CaMKII; Affinity Bioreagents, Golden, CO) is specific for α and β subunits of CaMKII only when they are phosphorylated at Thr-286/287 (in β). Bound antibodies were visualized by measuring fluorescence of a secondary antibody, (IRDye800-Anti-mouse; Rockland, Gilbertsville, PA; 1.5 μg in 30 ml of TTBS) with an Odyssey Imaging System (Li-Cor Biosciences, Lincoln, NE). Conditions were adjusted so that fluorescence intensity was linear over the range of CaMKII measured. The amount of autophosphorylated CaMKII was calculated by quantifying the intensity of each band with the Odyssey software and normalizing it to the intensity of a band corresponding to standard fully autophosphorylated CaMKII prepared by incubating in the assay solution for 1 min with 300 μM Ca²⁺ and 10 μM WT CaM.

Measurement of Turnover Numbers for Autophosphorylation of CaMKII.

Initial rates of autophosphorylation were measured at saturating levels of Ca²⁺/CaM and ATP with the use of a temperature-controlled Kintek Model RQF-3 Quench Flow apparatus (Kintek Corp., Austin, TX). Reactions were initiated at 100 ms, 300 ms, 1, 3, 10, and 30 sec by rapid mixing of solution 1 (25 mM Tris·HCl, pH 7.2/0.1 M NaCl/1 mM MgCl₂/2 mM Ca²⁺/0.9 mg/ml BSA/5 mM DTE/1.4 μM CaMKII catalytic subunits, and either 12 μM WT CaM, 30 μM CaM-C^WT, or 60 μM CaM-N^WT) and solution 2 (identical to solution 1 except that CaMKII and CaM were replaced by 200 μM ATP). The two solutions were kept at 4°C until they were transferred to the quench flow apparatus and warmed for 1 min to 30°C. Autophosphorylation was initiated by rapid mixing of 16 μl of each solution and terminated by rapid addition of a final concentration of 1% SDS, 3.3 mM glycine·HCl, pH 2.9 (see Supporting Text) After quenching, all reactions were brought to a final concentration of 3% SDS, 66 mM Tris·HCl (pH 7.2)/2% 2-mercaptoethaol/5% glycerol/40 μg/ml Bromophenol Blue and subjected to SDS/PAGE followed by immunoblotting with antibody 22B1 as described for the autophosphorylation assay. Intensities of autophosphorylated bands were quantified with the Odyssey Imaging System. The levels of autophosphorylation were normalized to that of fully autophosphorylated CaMKII incubated in the assay for 30 sec (WT CaM) or 100 sec (mutant CaMs). The amount of autophosphorylated CaMKII in each sample lane was obtained from a standard curve built for each set of experiments by measuring the intensity of 0.06, 0.12, and 0.23 μg of fully autophosphorylated CaMKII and fitting the standard data with a power function.

Abbreviations

CaMKII

Ca²⁺/calmodulin-dependent protein kinase II

CaM

calmodulin

CaM-N^WT

mutant CaM that binds Ca²⁺ only at N-terminal sites

CaM-C^WT

mutant CaM that binds Ca²⁺ only at C-terminal sites

CaMKII-cbp

peptide having the sequence of the CaM-binding domain of CaMKII

CaM2C

WT CaM with two Ca²⁺ ions bound at its C terminus

CaM2N

WT CaM with two Ca²⁺ ions bound at its N terminus

CaM4

WT CaM with Ca²⁺ ions bound at all four sites.