Voltage-gated sodium (NaV) channels play a fundamental role in electrically excitable cells and tissues and are an essential component of the molecular machinery required to generate and propagate the action potential (Box 1) (1). NaV channels are major therapeutic targets in local anesthesia, cardiac arrhythmia, analgesia, epilepsy and bipolar disorder and are also being investigated in conditions such as stroke and Parkinson’s disease (2,3). An increasing number of inherited disorders have been associated with NaV channel mutations (channelopathies) including some forms of periodic paralysis, myotonia, epilepsy and heart arrhythmias (4,5). Studies over the past 20 years have identified many of the molecular details by which NaV channels respond to changes in membrane potential and the basis of their ion selectivity. More recently, attention has focused on how the channels are assembled within the secretory pathway and how they are correctly targeted to distinct subdomains within the plasma membrane. Although in this review we focus on NaV channel targeting, it should be emphasized that many of the same questions also apply to the targeting of other voltage-gated ion channels. For up-to-date reviews of this wider field, see Lai and Jan (6), Arnold (7) and Hedstrom and Rasband (8).
Electrical excitability in cells such as neurons and myocytes depends not only upon the expression of voltage-gated sodium channels but also on their correct targeting within the plasma membrane. Placing sodium channels within a broader cell biological context is beginning to shed new light on a variety of important questions such as the integration of neuronal signaling. Mutations that affect sodium channel trafficking have been shown to underlie several life-threatening conditions including cardiac arrhythmias, revealing an important clinical context to these studies.
Box 1: The neuronal action potential. In the resting state, neurons maintain a negative charge across their plasma membrane that is typically around −80 mV. This is largely because of an ATP-dependent Na/K transporter that exchanges three internal Na+ ions for two external K+ ions. This not only contributes to a negative resting membrane potential but also leads to a higher internal concentration of K+ and a higher external concentration of Na+. Action potentials are propagated by transient reversal of this transmembrane potential because of the opening and closing of NaV and KV channels and subsequent influx and efflux of Na+ and K+ along their electrochemical gradients. A) At the normal resting potential, both NaV and KV channels are largely in the closed state. B) Small, neurotransmitter-triggered depolarizations occur in the somatodendritic membranes of the neuron (e.g. because of the activation of ionotropic neurotransmitter receptors) and spread passively to the AIS, where NaV channels are concentrated (Figure 2A). This causes rapid NaV channel activation and influx of Na+. If the depolarizing stimulus is above a threshold amplitude (e.g. −60 mV in the figure), then an action potential is triggered. This occurs when the influx of Na+ exceeds the background K+ efflux (‘leak’ currents) and the net inward flow of positively charged ions causes further depolarization that drives a positive feedback loop, in which the opening of additional NaV channels causes even greater sodium ion entry and further depolarization. This ‘regenerative’ opening of NaV channels ensures an ‘all or nothing’ response, whereby a full action potential is always generated whenever the initial stimulus is above threshold. The opening of NaV channels in response to depolarization is thought to be triggered by the S4 transmembrane helices present in each of the four channel domains (Figure 1A). These are amphipathic and contain positively charged amino acid residues in every third position within the helix that are proposed to act as voltage sensors. Upon depolarization, electrostatic forces on these amino acids cause movement of the S4 helices triggering a conformational change that opens the channel (84). KV channels are thought to activate in a similar manner, but the kinetics of this process is much slower than for NaV channels. C) Termination of the depolarization phase of the action potential begins within a millisecond when K+ efflux starts to exceed Na+ influx. This occurs as NaV channels begin to inactivate and as the ‘delayed rectifier’ KV channels begin to open. Inactivation of NaV channels is driven by a conformational change in which a hydrophobic motif (IFM) in the cytoplasmic loop between domains III and IV acts as a latch to block the inner mouth of the pore (Figure 1). The outward movement of the S4 helix of domain IV during the activation step is closely coupled to this ‘fast’ inactivation process, and this is thought to allow the inactivation gate to close (85). This inactivation blocks further entry of sodium ions, preventing the neuron from becoming overstimulated and ensuring the action potential can only be propagated in the forward direction. Channels remain in this inactivated state until the membrane is hyperpolarized (the absolute refractory period). D) The exit of potassium ions moving down their electrochemical gradient through activated KV channels returns the membrane to its resting potential. This allows inactivated NaV channels to recover to their resting state and, as the KV channels deactivate, the neuron is reset ready for further stimuli.
Structure of NaV Channels
The NaV channels consist of a highly glycosylated pore-forming α-subunit (260 kDa), together with associated auxiliary β-subunits (33–36 kDa). The α-subunit alone is capable of forming a functional ion-selective channel. Ten α-subunit genes have been identified in mammals, designated NaV1.1–NaV1.9 and an atypical Nax. These different genes show distinct expression patterns, and their protein products have subtle differences in gating properties (3,9). The α-subunit is a single polypeptide chain consisting of four homologous domains (I–IV), each of which contains six membrane-spanning alpha-helices (S1–S6) (Figure 1A). Each of the four domains forms a complex with pseudo fourfold symmetry around the ion-selective pore (Figure 1B), a model that is supported by single-particle image analysis obtained from cryo-electron microscopy (10).
In mammals, four accessory β-subunit genes (β1–4) have been identified (11–14). The β-subunits affect gating kinetics in complex ways that depend on the particular α/β-subunits expressed and the host expression system used (15). In mammalian host cells, β-subunits can modulate the voltage dependence of activation and inactivation, the time–course of inactivation and recovery from inactivation. These effects are likely to be functionally important as they will influence the delicate balance between Na+ and K+ fluxes that fine-tune the initiation, duration and kinetics of the action potential (Box 1). A good example is the modulation of NaV channels in Purkinje neurons by the β4-subunit that binds to, and blocks, open channels upon depolarization and then unbinds during repolarization unblocking them again. This bypasses the classical pathway of inactivation and causes a ‘resurgent current’ during repolarization, thus shortening refractory periods and contributing to the ability of Purkinje neurons to carry out repetitive high-frequency firings (16).
Based on their sequence similarity, the β-subunits can be divided into two subgroups (β1/3 and β2/4). The β2/4-subunits are covalently linked to the α-subunit through a disulfide bridge in the extracellular domain. The β1/3-subunits are bound non-covalently (1). The β-subunits are all type I membrane proteins with a single-pass transmembrane domain and an extracellular V-type immunoglobulin domain that is most similar in sequence to the L1 family of cell adhesion molecules (CAMs) (12,17). Indeed, the β-subunits show a close functional association with these molecules and can act as CAMs in their own right, independent of their role in sodium channel association (18) (see below). Consistent with this view, the β3-subunit can traverse the secretory pathway independent of the α-subunit, and the intracellular domain is essential for correct subunit targeting (19).
Trafficking and Selective Localization of Neuronal NaV Channels
As with most membrane-bound proteins, the NaV channels undergo extensive glycosylation in the endoplasmic reticulum (ER) with further modifications in the Golgi (20). Some neuronal NaV isoforms require additional proteins to increase the efficiency of traffic through the secretory pathway. A striking example is the peripheral sensory neuronal channel NaV1.8. The binding of annexin II light chain (p11) to this channel enhances its export to the plasma membrane (21). The p11 subunit is implicated in the facilitated export of other ion channels (22), but among NaV channels, only NaV1.8 interacts in this way (23). The upregulation of NaV1.8 expression in sensory neurons is correlated with reduced pain thresholds (24). Hence, there is interest in exploiting this association pharmacologically to modulate NaV1.8 expression (21). The covalent association of neuronal NaV channels with β2-subunit is critical for high-level surface expression and takes place late in secretion in a post-Golgi compartment. In neurons, some 70% of NaV channels appear to lack associated β2-subunit and account for a substantial intracellular pool, exit from which is likely to be the rate-limiting step in NaV channel trafficking (20). Contactin, a glycosyl-phosphatidylinositol-anchored CAM protein, binds the NaVβ1-subunit and the α-subunits of NaV1.3 and NaV1.9 and likewise increases the expression of channels at the plasma membrane (25–27).
In myelinated neurons of the vertebrate central and peripheral nervous system, the myelin sheath surrounds and insulates the axon except at the nodes of Ranvier. This ensures that when an action potential fires at the axon initial segment (AIS), it moves down the axon by saltatory conduction from node to node far more rapidly than is possible for an unmyelinated axon (Figure 2A). For this arrangement to be effective, all the molecules that co-ordinate the action potential must be selectively concentrated both at the AIS and at the nodes of Ranvier. This includes not just the NaV channels but also voltage-gated potassium (KV) channels, together with the sodium potassium adenosine triphosphatase that is required to maintain the resting membrane potential (6). The major neuronal NaV channels clustered at the AIS and nodes of Ranvier are NaV1.2 and NaV1.6, but these channels are initially inserted uniformly into the plasma membrane (28). Fluorescence photobleach recovery experiments conducted on both cortical and spinal cord neurons established that the sodium channels are free to diffuse within the plasma membrane of the cell body, but diffusion is tightly restricted in the AIS, suggesting a role for tethering within this subdomain (29). The motif responsible for tethering has been identified as a short consensus sequence [V/A]P[I/L]AxxE[S/D]D within the intracellular loop connecting domains II and III (Figure 1A) (28,30,31). This is conserved in most NaV channels (30), and the glutamate residue within this sequence is particularly crucial for tethering (32). A similar sequence is also found in the KV channels KCNQ2 and KCNQ3, which are also clustered at the AIS and nodes of Ranvier (33). A phylogenetic analysis suggests that this motif is unique to vertebrates, evolved independently in sodium and potassium channels and may have emerged concomitantly with saltatory conduction (33). Using coimmunoprecipitation assays, the motif was shown to bind the cytoskeletal protein ankyrinG(30,31). In mice lacking cerebral ankyrinG, NaV channels no longer cluster (34), suggesting a central organizing role for this protein. AnkyrinG is bound to spectrin βIV that in turn connects to the underlying actin cytoskeleton (35), and both ankyrinG and spectrin βIV copurify in tight association with NaV channels (36–39).
AnkyrinG may also play a role modulating NaV channel gating. Many neuronal types express low levels of non-inactivating or persistent Na+ current that is thought be important in regulating their excitability. Under certain conditions, a similar persistent current is observed when NaV channel α-subunits are expressed in non-neuronal cells (3). In the case of NaV1.6, coexpression of ankyrinG (but not its close homolog ankyrinB) has been shown to suppress this persistent current (40), suggesting that ankyrinG regulates channel behavior not just by promoting clustering but also by directly modifying the channel inactivation kinetics. Indeed ankyrinG has also been reported to bind the III–IV loop that regulates inactivation gating (Figure 1A), independent of its binding to the II–III loop, albeit with weaker affinity (41).
AnkyrinG also binds the NaVβ1-subunit and a variety of CAMs of the L1 family such as NrCAM and neurofascin (Nf) (18,42) through a common tyrosine-containing motif present in the intracellular domains of these molecules. The tyrosine can be phosphorylated in a signal-dependent manner with the loss of ankyrinG binding (43,44). The functional significance of this phosphorylation-dependent binding is not understood, but could indicate the presence of a dynamic signaling pathway, perhaps important for the co-ordinated regulation of NaV channel assembly during development (15). The NaVβ1- and β3-subunits bind NrCAM, Nf and contactin through their extracellular immunoglobulin domains (27,45,46). Hence, the β1- and β3-subunits act as adaptors between a number of different CAMs and the NaV channels. As the CAMs contain multiple extracellular immunoglobulin domains separated from fibronectin-like domains by a flexible linker (47), the NaV channels can readily assemble into extended and stable clusters tethered both intracellularly and extracellularly by multiple protein–protein interactions. Work using single-molecule tracking techniques has shown that this clustering is so effective that at the AIS, even the diffusion of lipid molecules in the membrane is restricted (48).
Many of the molecules known to be important in NaV clustering at the AIS are also present at the nodes of Ranvier. One clear difference, however, is the requirement for myelinating glial cells for clustering at the nodes. For example, the nodal assembly of NaV channels is severely disrupted in Shiverer mice. These mutant mice lack myelin basic protein (49,50). In the developing optic nerve of the rat, NaV1.2 is first expressed on unmyelinated neurons. As the compact mature myelin assembles, the NaV1.2 isoform is selectively replaced with NaV1.6 at the nodes of Ranvier. Neither the targeting of NaV1.6 nor its upregulation occurs in the Shiverer mouse, indicating that myelination controls both the expression and the targeting of distinct NaV channel isoforms (49). Axonal myelination is formed by oligodendrocytes in the central nervous system and Schwann cells in the peripheral nervous system. In both cases, molecular interactions between proteins on the axon and proteins on or released by the myelinating cell are required for clustering of NaV channels at the nodes of Ranvier (51,52). In the peripheral nervous system, for dorsal root ganglia cocultured with Schwann cells, the axonal proteins NrCAM and the Nf isoform Nf186 mediate this process (53–55). A leading candidate for the presumed NrCAM and Nf186 receptor is gliomedin, a type II membrane protein synthesized by the Schwann cells (56). Gliomedin exists as both a membrane-bound form and a secreted form generated by proteolysis. Recent work suggests that it is the secreted fragment of gliomedin, rather than the membrane-bound form that is responsible for clustering (57). The extracellular domain of gliomedin is a trimer associated through collagen-like triple helix repeats and contains N-terminal olfactomedin-like domains that contain the CAM-binding sites. The collagen helix of the released gliomedin binds to heparin sulfate proteoglycans of the Schwann cell extracellular matrix (57). This leads to a very high local concentration of gliomedin enmeshed at the myelinating edges of the Schwann cell and provides multi-point high-avidity binding sites for the axonal CAMs that together with their associated NaV channels become ‘squeezed’ into a tight nodal location (Figure 2B). Additional heat-sensitive factors further facilitate sodium channel clustering (51). For example, the extracellular matrix-binding proteins tenascin-C secreted by Schwann cells and tenascin-R secreted by oligonucleotides bind to the NaVβ2-subunit with high affinity (58,59) (Figure 2B). NaVβ-subunits are known to be expressed on both axons and glia. They may play a role in further stabilizing node formation acting as trans-binding homophilic CAMs through their extracellular immunoglobulin domains (15).
Selective tethering at the AIS and axon alone does not explain the relative absence of NaV channels from the rest of the neuronal plasma membrane. Here, selective endocytosis most likely through clathrin-coated pits is also required. Two separate endocytic motifs have been identified in NaV1.2 that facilitate channel endocytosis. One occurs close to but distinct from the AIS retention motif in the domain II–III loop (Figure 1A). Chimeric marker proteins containing this loop as a cytoplasmic domain are inserted uniformly into the plasma membrane, but become internalized predominantly from the dendrites and soma (32). However, the sequence responsible for endocytosis does not fit a consensus di-leucine or Yxxф motif commonly recognized by AP2 adaptors. This could indicate the existence of novel adaptors for this motif. For example, clathrin-associated adaptor 1 (CAP-1) has recently been shown to act in such a capacity for the case of NaV1.8, although CAP-1 binds to a different sequence to the one identified in the II–III loop of NaV1.2 (60). In NaV1.2, a second endocytic motif has been localized at the C-terminal region. In this case, the sequence fits the conventional di-leucine sequence recognized by clathrin-coated pit AP2 adaptors (61) (Figure 1A). In addition, most sodium channels contain a conserved PY motif at the C-terminus that is a binding site for Nedd4 and Nedd4-2 ubiquitin protein ligases, implicated in the endocytic removal from the plasma membrane (62).
In cultured fetal neurons, sustained activation of NaV channels leads to their enhanced endocytosis and degradation in the lysosome (63). The mechanism underlying the stimulated downregulation of NaV channels is not clear, but in cultured adrenal chromaffin cells, endocytosis of NaV channels is enhanced by sustained (although not transient) elevation of intracellular calcium concentration. This calcium-induced endocytosis is reduced by inhibitors of the calcium-dependent protease calpain and inhibitors of the calcium-dependent phosphatase calcineurin and is stimulated by activation of protein kinase C (PKC)-α(64,65). Calpain binds clathrin-coated vesicles in a calcium-dependent manner and can stimulate the local proteolysis of spectrin, perhaps thereby allowing coated pits to bud. PKC-α and calcineurin could stimulate endocytosis by, respectively, phosphorylating and dephosphorylating dynamin I (64,65). In addition to induction of endocytosis, elevated intracellular calcium also activates PKC-ɛ that leads to the destabilization of NaV1.7 messenger RNA and reduced sodium current density (65). Spatio–temporal changes in intracellular calcium concentrations are important and ubiquitous in electrically excitable cells, and the coupling of calcium flux with the regulation of NaV channel trafficking could be crucial to the understanding of neuronal excitability and plasticity.
Distribution in NaV Channels Cardiac Myocytes
The major heart sodium channel is NaV1.5 (1). In transfected mammalian cells, NaV1.5 α-subunit accumulates within the ER, where it assembles with β-subunits (66). Failure to exit the ER may contribute to some cases of inherited cardiac channelopathies. Brugada syndrome is a serious life-threatening condition characterized by an increased susceptibility to sudden and unpredictable ventricular arrhythmias. It is inherited in an autosomal dominant manner and caused by a heterogeneous range of mutations in the NaV1.5 α-subunit gene (67). Many Brugada syndrome mutations directly affect the gating kinetics of the channel, but several have now been described that reduce surface expression because the channel does not fold correctly and is retained within the ER (68). In some cases, the presence of pharmacological NaV channel blockers induces the mutant channel to fold correctly and be exported (68). Whether this approach can be adapted for therapeutic use remains to be seen. In one remarkable case, an individual within a Brugada syndrome family expressed an NaV1.5 misfolding mutation but was nonetheless asymptomatic because this heterozygous individual also expressed an otherwise neutral NaV1.5 polymorphism that suppressed the Brugada mutation (69). This implies that in the ER, nascent NaV1.5 channels can interact and the Brugada mutant channel can be induced to fold correctly by the presence of the other allelic variant.
Pharmacological agents that stimulate protein kinase A (PKA) activity can enhance sodium current in cardiac myocytes, suggesting an increase in sodium channel expression at the plasma membrane (70), and activation of PKA stimulates the trafficking of NaV1.5 to the cell surface (71). In the NaV1.5 α-subunit, the intracellular loop connecting domains I and II contains three consensus PKA phosphorylation sites close to the sequence RXR (Figure 1A). This motif has been shown to act as an ER retention signal in other membrane proteins (72), and its deletion in NaV1.5 abolishes the PKA-dependent trafficking to the cell surface (73).
Although the details are not so well understood for cardiac cells as they are for neurons, a similar picture of differential tethering is also emerging to explain the selective localization of NaV channels in cardiac cells. In cardiac ventricular myocytes, the majority of NaV1.5 is localized to the intercalated discs (74). These specialized membrane domains separate adjacent cells in the cardiac muscle fiber and contain gap junctions that electrically connect the cells in the fiber. Hence, they allow the rapid conduction of electrochemical potentials and the depolarization wave directly between the cytoplasm of adjacent cells. Cardiac ventricular myocytes also express the neuronal NaV channel isoforms, particularly NaV1.1, 1.3 and 1.6, but these channels are predominantly expressed on the T-tubules (74) (Figure 2C). This suggests that NaV1.5 is specialized for the initiation of the cardiac action potential as it jumps from one cell to another in the fiber, while neuronal NaV channels conduct the action potential rapidly and synchronously into the interior of the cell to activate the contractile machinery.
The NaVβ1-subunit associated with NaV1.5 at the intercalated disc is predominantly phosphorylated on its intracellular tyrosine. The major intercalated disc protein N-cadherin interacts with the phosphorylated β1-subunit, and the major gap junction protein connexin 43 in turn interacts with this complex, thus ensuring a tight structural and most likely functional connection between NaV1.5 and gap junctions (75) (Figure 2C). AnkyrinG is also found at the intercalated disc (76). As the phosphorylated β1-subunit does not bind ankyrinG (see above), NaV1.5 is unlikely to attach through this subunit. However, like the neuronal channels, the NaV1.5 α-subunit contains the conserved ankyrinG-binding motif between domains II and III (Figure 1A). Recently, a patient with a distinct form of Brugada syndrome has been identified in which the glutamic acid residue known to be particularly crucial for ankyrinG tethering (Figure 1A) is mutated to lysine. The ability to bind ankyrinG is completely abolished in this NaV channel mutant, and it is no longer retained at the intercalated discs. Furthermore, the mutant NaV1.5 channel displays a faster onset of inactivation and slower recovery from inactivation than normal, again emphasizing the important interplay between the channel targeting machinery and electrophysiological properties (76).
NaV1.5 also copurifies with cytoskeletal proteins of the dystrophin-associated protein complex (DAPC) such as syntrophins (77). Syntrophins contain PDZ domains that bind the carboxy-terminus of NaV1.5 and also bind similar PDZ-binding sites on cardiac KV(78). Dystrophin-deficient mice, in which the DAPC is disrupted, show a significantly reduced surface level of NaV1.5 (79). This suggests an important role for the DAPC to co-ordinate and to regulate cardiac excitability and conduction.
Non-phosphorylated β1-subunit is largely associated with neuronal sodium channels on the T-tubules, where it binds the predominantly T-tubule protein ankyrinB(75) (Figure 2C). This is a striking example of how posttranslational modification can affect channel subunit targeting and raises the question of how this phosphorylation is regulated and by what signaling pathways.
Expression and Targeting of NaV Channels in Non-Excitable Cells
Recent work has established the surprising fact that NaV channels are also expressed in some non-excitable cells. For example, the neuronal channel NaV1.6 and the cardiac channel NaV1.5 are expressed in primary human monocyte-derived macrophages. Moreover, these two isoforms are localized in distinct intracellular compartments, NaV1.5 in the late endosome and NaV1.6 in the ER (80). The inhibition of NaV1.5 activity prevented the lipopolysaccharide-stimulated acidification of endosomes and reduced bacterial phagocytosis in these macrophages. This suggests that NaV1.5 plays a novel role in fine-tuning endosome acidification, perhaps by acting as a charge sink to allow positive charge from the interior of the endosome to the cytosol and thus facilitate hyperacidification. The mechanism by which this NaV isoform is selectively targeted to endosomes is unknown.
Expression of NaV1.5 has also been reported in some human T-cell tumor lines (81). Similarly, enhanced NaV1.7 channel expression has been detected in prostate cancer cell lines (82,83). In both cases, NaV channel activity is correlated with tumor invasiveness. In the rat prostate cancer cell line Mat-LyLu, NaV1.7 expression at the cell surface is stimulated by nerve growth factor acting through PKA (83). Sodium ion influx can also stimulate adenylate cyclase that in turn activates PKA, suggesting that the high level of NaV1.7 expression at the cell surface of this tumor cell line may be driven by a positive feedback effect (82).
Figure 3 provides a summary of the major trafficking steps implicated in modulating the steady-state distribution of NaV channel isoforms discussed in this review: controlled export from the ER, uniform integration into the plasma membrane, followed by selective retention on a localized cytoskeletal and extracellular scaffold and regulated endocytosis act together to sculpt the distinct cellular distribution of NaV channels. Initial characterization of NaV channels necessarily focused on simple heterologous expression systems in which channel behavior could be easily examined. In such studies, differences in the electrophysiological properties of NaV channels have been found to be relatively small (3,9). Increasingly, however, it is becoming clear that the channels function as part of an extended complex of interacting proteins whose correct subcellular location and targeting is critical to its normal physiological properties. This may provide a more plausible explanation for the variety of different NaV channel isoforms typically expressed in cells. There is now a need to understand the connection between location and electrophysiological function in greater quantitative detail, how these complexes self-assemble during development to reliably generate functionally specialized membrane subdomains and how disruption of this targeting may lead to disease.
We thank Tim Dale and Sarah Lummis for helpful comments on the article. This work was supported by a Biotechnology and Biological Sciences Research Council CASE studentship to F. S. C.