Binding of Gd³⁺ to the neuronal signalling protein calexcitin identifies an exchangeable Ca²⁺-binding site

Calexcitin is a two-domain, neuron-specific signalling protein which possesses four EF-hand motifs, of which only the first three are functional metal-binding sites. Recombinant calexcitin from L. pealei has been crystallized with an equimolar concentration of Gd³⁺ in the presence of trace Ca²⁺, and it was demonstrated that it is the Ca²⁺ ion bound to the N-terminal EF-hand which becomes replaced by the exogenous heavy-metal ions.

Abstract

Calexcitin was first identified in the marine snail Hermissenda crassicornis as a neuronal-specific protein that becomes upregulated and phosphorylated in associative learning. Calexcitin possesses four EF-hand motifs, but only the first three (EF-1 to EF-3) are involved in binding metal ions. Past work has indicated that under physiological conditions EF-1 and EF-2 bind Mg²⁺ and Ca²⁺, while EF-3 is likely to bind only Ca²⁺. The fourth EF-hand is nonfunctional owing to a lack of key metal-binding residues. The aim of this study was to use a crystallographic approach to determine which of the three metal-binding sites of calexcitin is most readily replaced by exogenous metal ions, potentially shedding light on which of the EF-hands play a ‘sensory’ role in neuronal calcium signalling. By co-crystallizing recombinant calexcitin with equimolar Gd³⁺ in the presence of trace Ca²⁺, EF-1 was shown to become fully occupied by Gd³⁺ ions, while the other two sites remain fully occupied by Ca²⁺. The structure of the Gd³⁺–calexcitin complex has been refined to an R factor of 21.5% and an R _free of 30.4% at 2.2 Å resolution. These findings suggest that EF-1 of calexcitin is the Ca²⁺-binding site with the lowest selectivity for Ca²⁺, and the implications of this finding for calcium sensing in neuronal signalling pathways are discussed.

1. Introduction  

A central endeavour in the field of neuroscience is to understand the complex information processing performed by biological neural networks that underlies animal behaviour. Paramount to this pursuit is the investigation of the processes of learning and memory, as well as the biophysical and molecular mechanisms that make such processes possible (Giese et al., 2001 ▸). Owing to the complexity of vertebrate nervous systems, initial insights into the molecular basis of memory have been obtained from more amenable experimental invertebrate models with far simpler neural networks, containing many orders of magnitude fewer neurons (Alkon, 1983 ▸).

In much of the seminal work of Alkon and coworkers, memory-specific biochemical changes were investigated using a classical Pavlovian conditioning paradigm in Hermissenda crassicornis. In this nudibranch mollusc, pairing of light and rotation creates an association between the usually attractive stimulus of light and foot retraction, which is usually an aversive response to movement. Studies of changes in neuronal physiology and molecular composition of photoreceptor cells that correlated specifically with the learned behaviour were undertaken (Nelson & Alkon, 1988 ▸, 1990 ▸, 1997 ▸; Alkon, 1979 ▸; Alkon et al., 1990 ▸; Kawai et al., 2002 ▸). This work demonstrated the upregulation and phosphorylation of a neuron-specific ∼20 kDa protein (Neary et al., 1981 ▸), which was subsequently termed calexcitin. This protein was found to have GTPase activity and could reproduce the electrophysiological effects of conditioning when injected into photoreceptor cells (Nelson et al., 1990 ▸). Furthermore, it was found that phosphorylation of calexcitin, which is mediated by protein kinase C, caused its translocation from cytosolic to membrane fractions (Nelson & Alkon, 1995 ▸).

Cloning and sequencing of a squid homologue of calexcitin from the optic lobe of Loligo pealei (Nelson et al., 1996 ▸) showed a number of interesting features such as the presence of three functional EF-hand binding motifs which form Ca²⁺- and Mg²⁺-binding sites (Gombos et al., 2001 ▸). Electrophysiological studies using recombinant squid calexcitin and human fibroblast inside-out patches also showed that the protein is capable of directly regulating K⁺-channel mean open time and probability in the presence of Ca²⁺ (Nelson et al., 1996 ▸). Furthermore, investigation using sensory and chemical blocks during learning in Hermissenda established calexcitin as being necessary to establish and consolidate long-term memory (Kuzirian et al., 2001 ▸, 2003 ▸; Borley et al., 2002 ▸; Child et al., 2003 ▸; Epstein et al., 2003 ▸). Additional experimental evidence has also implicated calexcitin in several mechanisms that could enable it to functionally alter neuronal physiology. Pre­synaptically, calexcitin is thought to be a key constituent of terminal nerve endings (Eyman et al., 2003 ▸) and to regulate retrograde axonal transport (Moshiach et al., 1993 ▸). Furthermore, postsynaptically it may alter dendritic morphology upon phosphorylation (Lederhendler et al., 1990 ▸) and can transform synaptic outputs from inhibitory to excitatory potentials (Sun et al., 1999 ▸). Noteworthy evidence has furthermore suggested that its postsynaptic effects could result from the modulation of intracellular Ca²⁺ levels by direct interaction with the ryanodine receptor, facilitating Ca²⁺ release from intracellular stores (Nelson et al., 1999 ▸). Although calexcitin shows the highest homology to the sarcoplasmic calcium-binding protein (SCP) family, past work identified protein features consistent with a calcium-sensor function.

X-ray and NMR structural studies ultimately demonstrated that calexcitin is primarily an α-helical protein organized into two separate domains, each with two EF-hand motifs, although the last one is nonfunctional (Erskine et al., 2006 ▸; Gombos et al., 2006 ▸). The two domains appear to adopt a closed conformation with a pronounced hydrophobic core which is similar to that found in calmodulin, suggesting that domain movements allow calexcitin to interact with binding-partner proteins (Erskine et al., 2006 ▸; Gombos et al., 2003 ▸). The most recent data from NMR and molecular-simulation studies suggested that the domains have sufficient flexibility in solution to allow opening of the inter-domain hydrophobic cleft (Erskine et al., 2015 ▸).

Titration studies have indicated that calexcitin has different affinities for Ca²⁺ ions depending on the Mg²⁺ concentration (Gombos et al., 2003 ▸). In the absence of Mg²⁺ the apparent K _d for Ca²⁺ is 0.77 µM (range 0.37–3.3 µM), which changes to 2.58 µM (range 1.5–4.8 µM) in the presence of 2 mM Mg²⁺. In quiescent neurons the Ca²⁺ concentation is in the 10–100 nM range and the presence of Mg²⁺ means that even the high-affinity site(s) will probably not be occupied by Ca²⁺. During cellular activation, the intracellular level of Ca²⁺ increases to the micromolar range and can be further elevated in particular cellular compartments. Thus, it is likely that all three sites will be occupied by Ca²⁺ during neuronal activation.

There is uncertainty as to which of the metal-binding sites of calexcitin has a sensory role in the neuron and also as to how the sites give rise to the different measurable Ca²⁺ affinities (Gombos et al., 2003 ▸). Our aim is to define how the metal-binding sites of this protein allow it to act as a calcium sensor in a similar manner to calmodulin, which is known to adopt a significantly different structure at high Ca²⁺ concentration. Understanding the structural basis of calcium sensing by EF-hand proteins is an important objective in defining how intracellular signalling pathways operate. To address some of these questions, we have crystallized recombinant calexcitin in calcium-free buffer in the presence of an equimolar concentration of the heavy atom gadolinium and only trace Ca²⁺. Lanthanide ions such as Gd³⁺ are well known for binding at Ca²⁺-binding sites in proteins owing to their unique chemical property known as lanthanide contraction. This arises from poorer shielding of valence electrons by the 4f electrons in the lanthanide elements, such that the increase in nuclear charge across this series of the periodic table causes an enhanced reduction in ionic radius. We have co-crystallized calexcitin with equimolar Gd³⁺ in the presence of only trace Ca²⁺ and determined the structure at 2.2 Å resolution, demonstrating that Gd³⁺ binds at EF-1, the first EF-hand in the N-terminal domain of the protein, and suggesting that this is the site with the lowest selectivity for Ca²⁺.

2. Methods  

2.1. Expression and crystallization of the calexcitin–Gd³⁺ complex  

His-tagged L. pealei calexcitin was expressed in Escherichia coli and purified as described previously (Beaven et al., 2005 ▸; Erskine et al., 2006 ▸), except that protease inhibitors were not used during purification. A 1 ml aliquot of purified calexcitin at a concentration of 4 mg ml⁻¹ was dialyzed overnight against a 1 l volume of 50 mM Tris pH 7.5, 100 mM NaCl, 1 mM β-mercaptoethanol that was made without adding any CaCl₂, which is normally present in the crystallization buffer at a concentration of 1 mM. After dialysis, the protein at a concentration of 2.6 mg ml⁻¹ (60 µM) was mixed with GdCl₃ to give an equimolar ratio and was subjected to the Morpheus HT-96 (Gorrec, 2009 ▸) and PACT premier HT-96 (Newman et al., 2005 ▸) crystal screens using a Mosquito robot (TTP Labtech) to mix 400 nl well solution and 400 nl protein solution. Molecular Dimensions Structure Screens 1 + 2 (Jancarik & Kim, 1991 ▸; Wooh et al., 2003 ▸) were also used in 24-well plate format by mixing 3 µl well solution and 3 µl protein solution. Crystals suitable for diffraction experiments were obtained from two Structure Screen 1 conditions (0.2 M sodium acetate trihydrate, 0.1 M sodium cacodylate pH 6.5, 30% PEG 8000 and 0.2 M sodium acetate trihydrate, 0.1 M Tris pH 8.5, 30% PEG 4000). Crystals from the latter condition were cryoprotected by the addition of glycerol to a final concentration of 30%, mounted in 0.2 mm LithoLoops (Molecular Dimensions) on 18 mm CrystalCap pins (Hampton Research) and finally flashed-cooled in a nitrogen-gas cryostream (Oxford Cryosystems) at a temperature of 100 K for storage in liquid nitrogen.

2.2. X-ray data collection and processing  

Synchrotron X-ray data for the calexcitin co-crystals were collected on the I04 beamline at the Diamond Light Source facility (DLS), Didcot, England using a Pilatus 6M-F detector. A total of 190 images, each with an oscillation angle of 1° and an exposure time of 3 s, were recorded with a beam size of 20 × 20 µm and a transmission of 33%. The data were processed with MOSFLM (Leslie & Powell, 2007 ▸) and scaled using SCALA (Evans, 2006 ▸) in the CCP4 program suite (Winn et al., 2011 ▸), keeping the Friedel mates separate to allow analysis of anomalous differences. The scaling statistics for alternative point groups and inspection of systematic absences indicated that the space group was either P4₁ or P4₃, with unit-cell parameters a = b = 77.0, c = 29.3 Å. Use of the MATTPROB server (Kantardjieff & Rupp, 2003 ▸) indicated that the crystals contained one molecule per asymmetric unit, with a solvent content of 38%. The structure was determined by molecular replacement with a monomer of the native protein (PDB entry 2ccm; Erskine et al., 2006 ▸) as the search model. Phaser (McCoy et al., 2007 ▸) found a solution in space group P4₃, and the resulting maps were interpreted using Coot (Emsley & Cowtan, 2004 ▸; Emsley et al., 2010 ▸) prior to refinement of the model using REFMAC5 (Murshudov et al., 1997 ▸, 2011 ▸) and PHENIX (Adams et al., 2010 ▸). The latter program was used for the refinement of group TLS parameters, and the metal anomalous scattering factors were refined using the I(+) and I(−) data. The resulting structure and data have been deposited in the Protein Data Bank with identifier 5f6t.

3. Results  

Data-processing and refinement statistics for the gadolinium complex are shown in Table 1 ▸, where it can be seen that the R-factor values are slightly worse than would be expected for a structure of this resolution; we attribute this to crystal quality since the diffraction spots displayed some splitting. However, radiation damage did not seem to be significant during data collection.

Table 1. X-ray data-processing and refinement statistics for the calexcitin–Gd³⁺ complex.

Values in parentheses are for the outer resolution shell.

Beamline	I04, DLS
Wavelength (Å)	0.9173
Space group	P43
Unit-cell parameters
a = b (Å)	77.0
c (Å)	29.3
Solvent content (%)	37.7
No. of monomers per asymmetric unit	1
Matthews coefficient (Å3 Da−1)	2.0
Mosaic spread (°)	0.8
Resolution (Å)	29.3–2.2 (2.3–2.2)
R merge † (%)	13.7 (139.0)
R meas ‡ (%)	15.2 (156.4)
CC1/2 § (%)	99.6 (34.8)
Completeness (%)	100.0 (100.0)
Average I/σ(I)	7.5 (1.5)
Multiplicity	6.1 (6.0)
No. of observed reflections	54653
No. of unique reflections	9020
Wilson plot B factor (Å2)	66.2
R factor (%)	21.5 (39.6)
Free R factor (%)	30.4 (46.8)
R.m.s.d., bond lengths (Å)	0.009
R.m.s.d., bond angles (°)	1.058
No. of I(+) and I(−) in working set	16100
No. of reflections in test set	870
Mean protein B factor (Å2)	61.5

^† R _merge = .

^‡ R _meas = .

^§Half-set correlation coefficients of Karplus & Diederichs (2012 ▸).

The electron-density map obtained for the gadolinium co-crystal showed a large peak of positive difference density (18.3 r.m.s.) at EF-1 (Fig. 1 ▸). Since all metal ions were omitted from the model at this stage, positive difference density was expected at the remaining two calcium-binding sites, and accordingly peaks of 7.6 r.m.s. and 6.7 r.m.s. were found at EF-2 and EF-3, respectively, suggesting that these are still occupied by Ca²⁺ rather than Gd³⁺. The binding of Gd³⁺ to EF-1 was corroborated by the calculation of an anomalous difference map in which the highest peak (6.7 r.m.s.) was found to occur at this site, whereas no significant peaks were found at EF-2 or EF-3. In the complex Gd³⁺ is the only significant anomalous scatterer, with a theoretical f′′ of 6 e at the wavelength used for data collection (0.92 Å). The final model of the crystal structure contains a Gd³⁺ ion bound at EF-1, with Ca²⁺ ions bound at EF-2 and EF-3 and 30 water molecules. This model was refined using data within a resolution range of 27.4–2.2 Å to an R factor of 21.50% and an R _free of 30.44%. Structure validation by MolProbity (Davis et al., 2007 ▸; Chen et al., 2010 ▸) showed that 88% of residues were within the most favoured region of the Ramachandran plot and no residues were classed as outliers. The model has a mean isotropic B-factor value of 61.10 Ų for all atoms in the model, which is a rather high value, but this again is probably a reflection of crystal quality. Validation of the metal-binding site geometry with CheckMyMetal (Zheng et al., 2014 ▸) indicated that the model of the Gd³⁺-binding site is plausible.

Figure 1.

The difference electron density for the three metal-binding sites of calexcitin co-crystallized with equimolar Gd³⁺ in the presence of trace Ca²⁺. The Ca²⁺-binding sites formed by EF-hands 1–3 are shown in (a), (b) and (c), respectively, with the final refined position of each metal ion shown by a grey sphere. The OMIT electron density is contoured at 5 r.m.s. (blue) and at 15 r.m.s. (dark blue). A peak in EF-1 exceeding 18 r.m.s. indicated the presence of Gd³⁺, whereas EF-2 and EF-3 had peaks of around 7 r.m.s., indicating that these retain Ca²⁺. Accordingly, significant anomalous difference density (contoured at 5 r.m.s. and shown in brown) was only found at EF-1. Each view is oriented such that the bidentate ligand (Asp34, Glu85 and Glu130) is on the left.

The structure of the Gd³⁺ complex superimposes with the native structure of calexcitin with a C^α r.m.s. deviation of 0.66 Å (Fig. 2 ▸). This value is slightly higher than might be expected, but is best explained by the different crystal-packing environment of the complex, which is non-isomorphous with the native structure. Accordingly, the regions with the largest deviations are the non-EF-hand loops, which interact with symmetry-related molecules in the two crystal forms and lack the stabilizing influence of a bound metal ion. The metal-binding geometry of the Gd³⁺ site is very similar to that of the native Ca²⁺-bound protein, emphasizing that Gd³⁺ is a judicious choice to replace Ca²⁺. The structures of the other metal-binding sites (EF-2 and EF-3) are very similar to those of the native protein, as would be expected since these are most likely to have retained Ca²⁺. However, the metal-bound water molecules, which are visible in the higher resolution electron density for the native protein, do not appear to be resolved in the map for the Gd³⁺ complex and this is likely to be owing to the higher mean B factor of the Gd³⁺ complex as well as the lower crystal and data quality, which are typical of heavy-atom-substituted proteins.

Figure 2.

Superposition of the structure of the calexcitin–Gd³⁺ complex with the native protein. The folds of the Gd³⁺ complex and the native protein are coloured gold and cream, respectively, and the metal-binding sites, numbered 1–3, are shown as grey spheres.

4. Discussion  

We have determined the structure of the neuronal signalling protein calexcitin crystallized in the presence of equimolar Gd³⁺ and only trace levels of Ca²⁺. Since Gd³⁺ was found to bind exclusively at EF-1, this site may be the most prone to binding exogenous metal ions when the protein is in a low Ca²⁺ concentration. The conditions of our experiment were such that the protein was partially demetallated by dialysis prior to the addition of Gd³⁺. Our finding that Gd³⁺ binds exclusively to EF-1 suggests that this is the first site to lose its Ca²⁺ ion as the Ca²⁺ concentration decreases. Indeed, earlier studies suggested that the N-terminal domain has mixed Ca²⁺/Mg²⁺-binding sites and that the C-terminal site has a higher specific affinity for Ca²⁺ (Gombos et al., 2003 ▸). Thus, we propose that EF-2 and EF-3 retained a bound Ca²⁺ ion following the dialysis step. It is interesting that the second site has a number of acidic residues nearby such that a Ca²⁺ ion attempting to dissociate might be rapidly pulled back by electrostatic forces. The unique structural nature of EF-3, owing to the lack of a functional partner EF-hand in the C-terminal domain and its proximity to the unusual C-terminal 3₁₀-helix which is implicated in GTPase activity, make it difficult to speculate on its relative affinity for calcium ions, but our results suggest that it is also a high-affinity Ca²⁺-binding site. Our suggestion that EF-1 is the first to lose its bound Ca²⁺ as the buffer Ca²⁺ concentration decreases is consistent with the occurrence of an aspartate residue (Asp34) as the bidentate metal ligand (Fig. 1 ▸). In contrast, EF-2 and EF-3 have glutamate residues at this position (Glu85 and Glu130) and this is associated with high selectivity for Ca²⁺. Indeed, mutational studies of other EF-hand proteins have shown that the presence of aspartate rather than glutamate as the bidentate metal ligand decreases the selectivity for Ca²⁺ (see, for example, Cox et al., 1979 ▸; Da Silva et al., 1995 ▸; Cates et al., 1999 ▸). In addition, this finding is in agreement with studies on calmodulin, which showed that Mg²⁺ preferentially binds to EF-1 and EF-4 in the absence of Ca²⁺ and that EF-3 has the lowest affinity for Mg²⁺ (Ohki et al., 1997 ▸). It is worth remembering that the fourth EF-hand of calexcitin has lost its ability to bind metal ions, which might be owing to its putative GTPase activity.

Our evidence that EF-1 of calexcitin has the lowest Ca²⁺-affinity site is significant since it opens an interesting line of enquiry into how metal binding triggers structural transitions that enable calexcitin to functionally interact with protein partners and mediate its putative GTPase activity. An important observation in this respect is that of all of the metal-binding sites of calexcitin, EF-1 is the closest to the hydrophobic domain interface, which is known to be considerably affected by flexible protein motion (Erskine et al., 2015 ▸). Earlier studies suggested that the binding of metal ions to the N-terminal domain reduces its molten-globule nature, thereby stabilizing the protein (Gombos et al., 2003 ▸). During neuronal activation, the binding of Ca²⁺ to EF-1 as the Ca²⁺ concentration increases might cause an ordering of the hydrophobic residues which form the domain interface. Indeed, it is the domain interface which plays a key role in the interactions of EF-hand proteins with partner molecules in signalling cascades. It is important to bear in mind that both Ca²⁺ and Mg²⁺ are likely to bind to the protein, even though the overall affinity for Ca²⁺ is several orders of magnitude greater than that for Mg²⁺. However, the intracellular concentration of Mg²⁺ is at least several orders of magnitude greater than the concentration of Ca²⁺, and Mg²⁺ can thus compete very effectively with Ca²⁺ for the binding sites, especially when the intracellular Ca²⁺ concentration is low, for example in quiescent neurons. Gombos et al. (2003 ▸) argue that calcium signalling involves a constant competition between Ca²⁺ and Mg²⁺ for binding sites in signalling proteins and their experimental studies indicate that the N-terminal domain of calexcitin possesses two mixed Ca²⁺/Mg²⁺-binding sites, whereas the C-terminal domain probably has a more specific Ca²⁺-binding site. Fully Ca²⁺-bound calexcitin has been shown by solution NMR to undergo quite a drastic domain movement that is most likely to expose the hydrophobic inter-domain cleft and could then allow it to interact with partner proteins in signalling cascades (Erskine et al., 2015 ▸). Indeed, these studies indicate that the fully Ca²⁺-occupied form of calexcitin is predominantly in a domain-opened state in solution and thus the binding of Ca²⁺ to all three sites of calexcitin must trigger it to adopt a conformation which is capable of interacting with partner proteins. This is in agreement with studies performed on calmodulin, which have shown that the full conformational change upon calcium binding does not occur until all sites are occupied by Ca²⁺ following the displacement of Mg²⁺ from the protein (Malmendal et al., 1999 ▸).

Thus, our results have corroborated previous findings on the Ca²⁺ affinities of the different metal-binding sites of calexcitin (Gombos et al., 2003 ▸) and have provided some perspective on the previous conclusions that the C-terminal binding site is the sensor site (Erskine et al., 2015 ▸). Furthermore, our new evidence that the N-terminal site is the lowest-affinity Ca²⁺ site opens possible avenues for research into how the N-terminal domain of calexcitin plays a functionally relevant role in the signalling events of learning and memory.

Supplementary Material

PDB reference: calexcitin–Gd³⁺ complex, 5f6t