Neuronal Calcium Sensor Proteins: Generating Diversity in Neuronal Ca²⁺ Signalling

Changes in Ca²⁺ concentration

Neurons maintain a resting intracellular Ca²⁺ concentration in the range of 40-100 nM ⁶. The receipt of a neurotransmitter signal from another neuron, or some other form of stimulation, can lead to changes in Ca²⁺ concentration ranging from highly localized (see Fig. 1) and transient Ca²⁺ elevations to longer lasting and global changes throughout the neuron⁷. Ca²⁺ can enter cells through plasma membrane channels such as voltage-gated Ca²⁺ channels or NMDA (N-methyl-D-aspartate) receptors, or be released from intracellular stores in response to messengers such as inositol-1,4,5-trisphosphate ⁸. The existence of sites of local Ca²⁺ entry into the cytoplasm (from external or internal sources) has resulted in the concept of Ca²⁺ nanodomains and microdomains ⁷, in which the effective intracellular Ca²⁺ concentration could transiently reach up to hundreds of micromolar near the mouths of the Ca²⁺ channels.

In presynaptic nerve terminals, local elevations in Ca²⁺ levels are involved, for example, in triggering neurotransmitter release through exocytosis of synaptic vesicles less than 100 μsec after Ca²⁺ entry⁹. Post-synaptic Ca²⁺ entry following neurotransmission can subsequently generate localized Ca²⁺ elevations in dendritic spines. Ca²⁺ concentration can be increased in dendritic spines without changes in the adjacent dendrite due to limited diffusion of Ca²⁺ through the narrow necks of the spines ⁶ . More extensive or sustained synaptic input to a neuron can result in Ca²⁺ waves in the dendrites or global Ca²⁺ elevations throughout the neuron (Fig. 1).

Different signals, diverse outcomes

The significance of the differing types of Ca²⁺ signals in neurons is that they result in a wide range of different physiological outcomes. Indeed, even quite similar Ca²⁺ signals can have distinct or even opposite results. Examples of this are the opposing phenomena of long-term potentiation (LTP) and long-term depression (LTD), either of which can be brought about in the same neurons by transient Ca²⁺ signals depending on their exact timing and magnitude. LTP is triggered by larger (micromolar) and shorter-lived Ca²⁺ elevations (a few seconds) and LTD by a Ca²⁺ elevation to only a few hundred nM but for a longer duration ¹⁰ . Another example is the ability of Ca²⁺ signals of differing magnitude in neuronal growth cones to stimulate either attraction or repulsion ¹¹ . In addition to these effects are long-term changes that can lead to potentially permanent changes in gene expression ¹² that, for example, underlie learning and memory. Also, global Ca²⁺ signals can lead to neuronal cell death ¹³.

Ca²⁺ binding and membrane localization

A common feature of the NCS proteins is that their first EF-hand cannot bind Ca²⁺ owing to inactivating substitutions in the EF-hand loop in all the family In structural studies, human and yeast NCS-1 ^{27, 28} , bovine neurocalcin δ ²⁹ bovine GCAP2 ³⁰ and human GCAP3 ³¹ were found to have three bound Ca²⁺ ions. By contrast, recoverin ³² and KChIP1 ^{33, 34} were isolated with only two bound Ca²⁺ ions as a result of an inactivating substitution of residues required for Ca²⁺ coordination in another one of their EF hands.

The NCS proteins all have very similar structures which, in contrast to the dumb-bell shape of Ca²⁺-loaded calmodulin, are more compact and globular even when Ca²⁺ is bound (Fig.2). All mammalian NCS proteins of classes A-D, as well as KChIP1, are N-terminally myristoylated ⁴ . Some KChIP2 and 3 isoforms instead have putative palmitoylation sites³⁵ . Both myristoylation and palmitoylation would be predicted to allow membrane association of these NCS proteins. In fact, as discussed below, some NCS proteins such as recoverin and the class B NCS proteins only expose their myristoyl group when Ca²⁺ is bound to them ^{32, 36, 37} . These NCS proteins are cytosolic at resting Ca²⁺ and reversibly associate with the plasma membrane and the Golgi complex when Ca²⁺is elevated. By contrast, others such NCS-1 are always associated with the plasma membrane and Golgi complex ³⁸ and KChIP1 with distinct vesicular structures ^{39, 40} . The distinct localisations of these myristoylated proteins are in part determined by interactions of their N-terminal amino acids with specific phosphoinositides ^{39, 41}.

Structural insights from recoverin

Extensive structural information is available for recoverin, which was the first of the NCS proteins to be discovered ⁴² . Recoverin inhibits rhodopsin kinase when in its Ca²⁺-bound form ⁴³ , and is therefore important in light adaptation in photoreceptors ⁴⁴ . In its Ca²⁺-free form, the myristoyl group of recoverin is sequestered in a hydrophobic pocket within the protein⁴⁵ . Sequential Ca²⁺ binding to EF hand 3 followed by EF hand 2 ^46-48 results in a significant conformational change that leads to extrusion of the myristoyl group ³² . This ‘Ca²⁺/myristoyl switch’ enables recoverin to associate with membranes through its myristoyl group in a Ca²⁺-dependent manner (see the Ca²⁺/myristoyl switch movie in the Online links box). As the hydrophobic residues of recoverin that are involved in the interaction with the myristoyl group are conserved in the other members of the NCS protein family, it was thought that all myristoylated NCS proteins would have a similar Ca²⁺/myristoyl switch³² . However, in addition to recoverin, only the class B proteins show this switch ^{36-38, 49} . Other members of the family instead use their myristoyl group for constitutive membrane association ^{38, 39, 50}.

The Ca²⁺-induced conformational change in recoverin also exposes a hydrophobic groove, which forms the binding site for its only known target, rhodopsin kinase ⁵¹ . A similar hydrophobic surface is exposed in the structure of Ca²⁺-bound NCS-1 ^{27, 28} and KChIP1 ^{33, 52} , and so the same mechanism could be used by other NCS proteins for some of their Ca²⁺-dependent target interactions. This does not, however, explain all of the interactions as, for example, GCAPs do not require Ca²⁺ for binding to their target guanylyl cyclases and remain bound in the presence or absence of Ca^{2+ 50, 53} and binding sites for target proteins on NCS-1 have been mapped to domains other than the exposed hydrophobic surface ⁵⁴ . In addition, early structural studies implied that KChIP1 interacted with Kv4 channels through the hydrophobic groove ³³ but more recent structural analyses ^{52, 55} have shown that two distinct sites within the N-terminal domain of Kv4 channels can bind independently to different regions of KChIP1, and mutagenesis has demonstrated a third direct interaction with KChIPs involving the C terminus of the channel ^{56, 57} . All three sites of interaction affect the traffic of Kv4 channels to the plasma membrane and their regulation by KChIPs, but in distinct ways ^{52, 55, 57} . The differential ability of the NCS proteins to interact with specific target proteins is undoubtedly the result of differences in surface residues that are involved in the protein-protein interactions within an overall conserved structure².

NCS-1: one protein, multiple functions

NCS-1 is most probably derived from the primordial NCS precursor and is expressed from yeast to man. Studies in mammalian cells have implicated NCS-1 in several functions ⁵ , although the only data using genetic manipulations come from lower organisms, in which very diverse phenotypes have resulted (Table 2). The S. cerevisiae orthologue frequenin (Frq1) is essential for survival because it is involved in the activation of one of the two phosphatidylinositol 4-kinases, Pik1 ²¹ . Knockout of NCS-1 is not lethal in other organisms but leads to developmental phenotypes, including changes in the rate of development, in Dictyostelium discoideum ⁷⁶ . Knockdown of ncs-1, one of the two very closely related NCS-1 genes in zebrafish, abolishes formation of the semicircular canals of the inner ear ²² . Changes in neuronal functions are shown in D. melanogaster, in which overexpression of the NCS-1 orthologue frequenin resulted in increased facilitation of neurotransmission ¹⁶ , and in C. elegans, in which knockout of NCS-1 impaired learning and memory ⁷⁷ . How can a single Ca²⁺ sensor be involved in so many different activities? NCS-1 is expressed highly in all brain regions ^{20, 25} and also in many non-neuronal cell types ^{18, 20} . In mammalian cells, NCS-1 regulates the orthologue of the yeast Pik1, phosphatidylinositol 4-kinase IIIβ ^{78, 79} , which could explain some of its functional roles, but a key aspect of NCS-1 is its ability to bind to, and regulate, many target proteins that are unrelated to each other ⁸⁰ (Figure 3). The interactions of NCS-1 with different targets proteins can involve distinct regions of NCS-1 ^{54, 73} , and some interactions can be independent of Ca²⁺ binding ^{73, 80} . The link between these target interactions and the varied physiological roles of NCS-1 still remain to be fully established.

The role of GCAPs in light adaptation

There are three genes that encode GCAPs in the human genome ²³ . GCAPs are present only in the retina and have a specific role in light adaptation during phototransduction through their ability to regulate retinal guanylyl cyclases 1 and 2. GCAPs interact with and activate these two proteins at low concentrations of Ca²⁺ in the light, but at higher Ca²⁺ levels in the dark these GCAPs undergo an ‘activator-inhibitor’ transition ⁸¹ to become inhibitors of guanylyl cyclases 1 and 2 (Figure 2). GCAP1 and GCAP2 are present in both rod and cone photoreceptors in human retina; GCAP3 is expressed only in cones, but is not expressed in the mouse, which implies that this protein is not essential for normal vision ⁸² . GCAP1 and GCAP2 are present in the retina at similar concentrations and regulate both of the guanylyl cyclases⁸³ . Expressing GCAP1 in mice that lack both GCAP1 and GCAP2 recovers almost all normal function in rods and cones ^{59, 60}.

Different GCAPs, same targets

As the two forms of GCAP regulate the same targets in the same cell type, why is more than one GCAP required? GCAP1 and GCAP2 are only 49% identical and appear to interact with guanylyl cyclases in different ways ^{83, 84} . A significant aspect of GCAP function is their high Ca²⁺ affinity. Analysis of the Ca²⁺ affinities of GCAP1 and GCAP2 and the Ca²⁺ requirement of the GCAPs for regulation of retinal guanylyl cyclases has revealed two important issues. First, the affinities of the GCAPs for Ca²⁺ is set to the relevant physiological range (25-250 nM in mouse rods, for example 85) at intracellular concentrations of Mg^{2+ 86} The second issue is that GCAP1 and GCAP2 differ in their Ca²⁺-binding affinities for retinal guanylyl cyclase activation (Fig 2) by about 7-fold ^{83, 86} . This has led to the proposal of a ‘Ca²⁺ relay’ model for retinal guanylyl cyclase activation ⁸⁴ in which both GCAPs would be needed to function over the physiological range of Ca²⁺ concentration in the retina. The fine-tuning of these different Ca²⁺ affinities would increase the overall sensitivity of the regulatory mechanisms, as their joint action would expand the dynamic range of responses to Ca²⁺ without losing the sensitivity that is contributed by the cooperative effect of Ca²⁺ on retinal guanylyl cylase regulation by each GCAP.

KChIPs: a large subfamily with distinct properties

Fast-inactivating A-type K currents are known to be important in controlling crucial aspects of neuronal excitability, including inhibition of the back-propagation of dendritic action potentials. These currents are also important for synaptic plasticity, and for regulation of excitability in the heart ^{87, 88} . Kv4 channel α-subunits are essential pore-forming determinants of A-type currents in somato-dendritic compartments such as in hippocampal pyramidal neurons ⁸⁹ . Changes in the expression of Kv4 channels have been implicated in conditions such as epilepsy ⁹⁰ and Kv4.2 channels in dorsal horn neurons of the spinal cord are targets for regulation of pain sensitivity ⁹¹ . The regulation of Kv4 subunits by KChIPs has therefore been studied in detail and provides clues as to why this class of NCS proteins is so diverse. The KChIP class has become the most diversified of the NCS proteins during evolution, from one KChIP in D. melanogaster, which is equally similar to all mammalian KChIPs, to mammals, which have four KChIP genes and a large number of potential splice variants ²⁴ . Analysis in human tissues indicates that at least 16 KChIP isoforms are detectably expressed (Fig 5) although some of the KChIP2 isoforms are expressed only in the heart ⁹² , where KChIP2 has an established role in the regulation of Kv4 channels ⁹³ . The conserved C-terminal EF hand-containing domain of the human isoforms are at least 70% identical to each other and the key difference between these KChIP isoforms lies in their N-terminal domains. The literature on these splice variants is confusing as the variants have been given multiple names ^{88, 94} and additional variants that have been studied might be expressed only in certain mammalian species⁹².

Three aspects of the KChIPs might be important for understanding their diversity: differential cell-type expression of the proteins; differences in their intracellular targeting and localization; and the different effects that they have on the traffic and gating properties of Kv4 channels. It is also possible that the existence of multiple genes might allow cell-type specific regulation at the level of transcription through their distinct promoters.

Different KChIPs, different intracellular localizations

When expressed in heterologous cell types the KChIPs show differences in their intrinsic targeting and localization, which might be important for their efficiency in stimulating traffic of Kv4 channels to the plasma membrane ^{39, 95} . KChIP1.2 is myristoylated and this allows its targeting to a population of intracellular vesicles, which increases the efficiency of trafficking of Kv4 channels to the plasma membrane ^{39, 40} . KChIPs 2.1, 2.2, 2.3 and 3.1 possess potential sites for palmitoylation (Fig 4) that is predicted to allow their membrane association. Palmitoylation of KChIPs 2 and 3 is required for efficient trafficking of Kv4 channels ³⁵ . Surprisingly, however, the intrinsic targeting of KChIPs 2 and 3 differs: only palmitoylated KChIP2 associates with the plasma membrane when expressed alone ⁹⁵ (Fig 4). It seems that KChIP3 only becomes membrane associated when it interacts with Kv4 channels. The other KChIPs do not possess any obvious membrane localization signals and when expressed alone are usually cytosolic ⁹⁵ (Fig 4) and become membrane-associated only through their binding to Kv4 channels.

Different effects of KChIPs on Kv4 channels

The first study of the effects of KChIP on Kv4 channels compared KChIPs 1.2, 2.3, 3.1⁶¹ , which were all found to have the same effects (with some quantitative differences): increasing cell surface expression of Kv4, shifting the voltage-dependency of activation, slowing of inactivation and more rapid recovery from inactivation. The N terminus of the KChIPs is not required for its interaction with Kv4 or the basic effects on channel traffic and gating properties ⁶¹ , but it has subsequently become clear that the variability in the N-terminal domains of the KChIP isoforms generates the variable effects on Kv4 channels. A comparison of the effects of KChIP1 splice variants on the rate of recovery from inactivation of Kv4 channels has shown differences in the effects of KChIP1.1 and 1.2 isoforms ^{96, 97} . Several studies on KChIP2 isoforms have shown differences in the efficiency of stimulation of traffic of Kv4 channels to the cell surface and in the modulation of gating properties ^{35, 94, 98, 99} . In fact, one KChIP2 isoform inhibits Kv4 surface expression, although this isoform is expressed only in the heart ^{92, 94} .One KChIP4 isoform, KChIP4.1, has similar effects to those originally reported in increasing traffic to the cell surface and slowing channel inactivation ^{65, 100} . By contrast, the KChIP4.4 isoform does not stimulate traffic of Kv4 channels and leads to the almost complete abolition of the fast inactivation of Kv4 channels. This effect has been attributed to the presence of a K-channel inactivation suppressor (KIS) domain in the N terminus of this isoform ^{65, 100}.

The majority of studies on the properties of KChIPs and their influence on Kv4 channels have looked at the effect of single KChIPs. It has become apparent, however, that interaction of KChIPs with Kv4 channels results in the formation of an octomeric structure containing four KChIPs and four Kv4 channel subunits ^{52, 101} . As KChIPs can form homo- or hetero-oligomers ^{102, 103} , it raises the possible additional complexity that functional KChIP-Kv4 complexes could contain more than one KChIP isoform leading to further degrees of subtle variation in channel function.

Differential expression for different proteins

The possibility of KChIPs being components of hetero-oligomers makes consideration of the expression patterns of the KChIP proteins of considerable importance. Analysis of the expression of the four KChIP genes through the use of PCR after reverse transcription of RNA (RT-PCR) and in situ hybridization showed major regional differences in their expression in the brain ²⁴ . The use of antisera specific for KChIP1-4, which are believed to be able to recognise all of the splice variants of each gene, indicated a marked cell-type specific expression of the KChIPs in the hippocampus, cortex striatum and cerebellum ^{95, 104, 105} . It would clearly be of interest to have more detailed mapping of the expression of each of the KChIP splice variants, but the information already available shows that each neuronal cell type might express only a limited repertoire of the possible set of KChIPs along with specific Kv4 channel isoforms. The detailed properties of A-type K currents are very variable between neuronal cell types or even in different regions of the same neuron ^{87, 88} , and it is easy to imagine that this could result in part from cell-type specific expression or localization of the KChIP isoforms to fine-tune the channels for their physiological function.

Differing Ca²⁺ affinities

Calmodulin interacts with a large number of different target proteins to regulate a multitude of cellular processes, so the question arises as to why NCS proteins are also required to transduce Ca²⁺ signals in neurons. One key factor is the difference in affinity for Ca²⁺ shown by the NCS proteins compared to calmodulin (Fig 2). Calmodulin binds Ca²⁺ in vitro with a dissociation constant (K_D) of ∼5-10 μM. Its affinity is increased, however, following binding of target proteins, which can reduce the off-rate for Ca^{2+ 106, 107} , leading to prolonged activation of target proteins and an increase in the apparent Ca²⁺ affinity within cells ¹⁰⁸ . It was also found ¹⁰⁸ .that most of the Ca²⁺-bound calmodulin in cells is bound to target proteins which are i excess and therefore only target proteins with high affinities for calmodulin are likely to be regulated under physiological conditions. This could explain why additional specific Ca²⁺ sensors are required.

The measured Ca²⁺-binding affinities of each of the NCS proteins varies ⁴ and the exact Ca²⁺ -binding affinities of each of the NCS proteins seems to be fine-tuned by conserved variation in the sequences of the EF-hand domains and by effects on protein structure of the N-terminal myristoyl group ⁴⁸ and even the C-terminal residues of the proteins ¹⁰⁹ so as to match their sensitivity to physiological requirements. Nevertheless, the general picture is that they are expected to be in a Ca²⁺ -free state under resting conditions and all have a higher affinity for Ca²⁺ than calmodulin does, which allows them to bind Ca²⁺ following small increases in concentration above resting levels. In addition, all of the NCS proteins show intrinsic cooperativity in Ca²⁺ binding when assayed in vitro or in assays that measure activation of their target proteins. This means that the NCS proteins usually have restricted dynamic ranges over which they can respond to changes in intracellular Ca²⁺ concentration to regulate their target proteins. This is exemplified by experiments on hippocalcin: analysis of its Ca²⁺/myristoyl switch in living cells indicated a value for half-maximal activation of 300 nM free Ca²⁺ and is fully activated when intracellular free Ca²⁺ concentration reaches 800nM Ca^{2+ 37} . The small concentration ranges over which the NCS proteins can respond to Ca²⁺ is also an explanation for the need for more than one GCAP as discussed above.

Specific subcellular and cellular localizations

As noted above, not all NCS proteins have a Ca²⁺/myristoyl switch and some of them, such as NCS-1 and KChIP1, are constitutively membrane associated through their myristoyl group. Despite this membrane association, NCS-1 and KChIP1 target to distinct organelles where their presence enables them to respond to very transient, local elevations in intracellular Ca^{2+ 39} . By contrast, NCS proteins that translocate to membranes using the Ca²⁺/myristoyl switch might require longer lasting and more global Ca²⁺ elevations for their activation and interaction with targets or could require repetitive Ca²⁺ signals to allow them to accumulate on membranes. This already gives some indication of how NCS proteins could mediate distinct cellular changes in response to different Ca²⁺ signals.

In addition to their distinct subcellular localization patterns, many NCS proteins are also expressed in a cell-type specific manner. NCS-1 seems to be expressed in most, if not all, neurons whereas other NCS proteins have distinct and more limited patterns of cellular expression in the brain ^{25, 104} and, as described above, the multiple KChIPs isoforms have highly cell-type specific expression patterns.

Specific target proteins

Initial work on NCS-1 showed that in vitro it could regulate the same targets as calmodulin, including calcineurin and phosphodiesterase ¹¹⁰ . These are not likely to be physiological targets for NCS-1, as calmodulin is present at much higher concentrations than NCS-1 in neurons and calcineurin and phosphodiesterase have a high affinity for calmodulin which is therefore much more likely to regulate these targets. It cannot be ruled out, however, that these proteins might be regulated by NCS-1 at concentrations of Ca²⁺ that are too low to activate calmodulin. It is now clear that NCS-1 (Figure 3) and other members of the NCS family have distinct target proteins that they regulate⁸⁰ — many of these do not overlap with calmodulin and so are more likely to be the true physiological targets that contribute to the specificity of the NCS protein function.