Surface expression and distribution of voltage-gated potassium channels in neurons (Review)

Abstract

The last decade has witnessed an exponential increase in interest in one of the great mysteries of nerve cell biology: Specifically, how do neurons know where to place the ion channels that control their excitability? Many of the most important insights have been gleaned from studies on the voltage-gated potassium channels (Kvs) which underlie the shape, duration and frequency of action potentials. In this review, we gather recent evidence on the expression, trafficking and maintenance mechanisms which control the surface density of Kvs in different subcellular compartments of neurons and how these may be regulated to control cell excitability.

• Voltage-gated potassium channel
• expression
• distribution
• targeting
• trafficking

Introduction

The voltage-gated potassium channels (Kvs) are structurally and functionally diverse integral proteins with representatives in all eukaryotic and some prokaryotic organisms [1]. The Kvs play a central role in determining the excitability of tissues including heart, skeletal muscle to brain [1], [2]. By modulating the membrane potential and the magnitude, shape and frequency of action potentials, Kvs ultimately control phenomena such as neurotransmitter release, muscle contractility and signal integration [3], [4]. Malfunction in Kvs underlies a range of inherited or acquired diseases and also mediate the side-effects of many clinically useful drugs [5]. Consequently, Kvs have received much attention both as therapeutic targets and as models of ion channel function [4], [5]. The culmination of such efforts, and one of the great achievements in ion channel biology, has been the description of crystal structures and models for Kv channels which rationalize decades of painstaking electrophysiology [6], [7].

The contribution of Kvs, or indeed any ion channel, to cell physiology, is not merely a function of their intrinsic biophysical characteristics. Rather, the macroscopic current (I) is a product of the channel's discrete current contribution (i) and open probability (p), and also the number of channels expressed at the cell surface (N) – or, more accurately, their surface density [2], [8], [9]. While i and p are biophysical parameters intrinsic to the channel, about which, for Kvs, much is known [1], less is known about the cell biological mechanisms which determine Kv surface densities. In polarized cells, such as neurons, the picture is even less clear, since Kvs are often found in functionally discrete domains including axons and dendrites and their sub-compartments (e.g., synapses) (Table I). How cells control the surface density and distribution of Kvs, has been a mystery whose secrets are now emerging. In this review, we examine the possible mechanisms underlying the control of Kv surface expression and distribution using mammalian neurons as a model.

Table I.  Properties of neuronal Kvs. Channels are listed according to IUPHAR and HGNC nomenclatures [1]. Biophysical properties of the corresponding channels are derived from Gutman et al. [1] while the distributions are compiled from references given in the text.

Subunit	HGNC	Function	Location
Kv1.1	KCNA1	Delayed rectifier	Axons, terminals, nodes
Kv1.2	KCNA2	Delayed rectifier	Axons, terminals, nodes
Kv1.3	KCNA3	Delayed rectifier	Axons, Somatodendritic, nodes
Kv1.4	KCNA4	A type, rapidly inactivating	Axons, terminals
Kv1.5	KCNA5	Delayed rectifier	Somatodendritic
Kv1.6	KCNA6	Delayed rectifier	Somatodendritic and axons
Kv1.7	KCNA7	Delayed rectifier	Absent or somatodendritic
Kv1.8	KCNA10	Delayed rectifier	ND
Kv2.1	KCNB1	Delayed rectifier	Soma,Proximal dendrites
Kv2.2	KCNB2	Delayed rectifier	Soma,Dendrites
Kv3.1	KCNC1	Fast activating, delayed rectifier	Axons, terminals, 3.1b: nodes
Kv3.2	KCNC2	Fast activating, delayed rectifier	Axons, terminals, somata
Kv3.3	KCNC3	A-type, rapidly inactivating	Axons, terminals, somata, dendrites
Kv3.4	KCNC4	A-type, rapidly inactivating	Axons, terminals, somata
Kv4.1	KCND1	A-type, rapidly inactivating	Somatodendritic, Low levels
Kv4.2	KCND2	A-type, rapidly inactivating	Somatodendritic
Kv4.3	KCND3	A-type, rapidly inactivating	Somatodendritic
Kv5.1	KCNF1	Modifier	ND
Kv6.1	KCNG1	Modifier	ND
Kv6.2	KCNG2	Modifier/silencer	ND
Kv6.3	KCNG3	Modifier/silencer	ND
Kv6.4	KCNG4	Modifier/silencer	ND
Kv7.1	KCNQ1	Slow activating, delayed rectifier	Absent in brain
Kv7.2	KCNQ2	Delayed rectifier, M current	Axons, nodes
Kv7.3	KCNQ3	Delayed rectifier, M current	Axons, Somatodendritic
Kv7.4	KCNQ4	Delayed rectifier, M current	Axons, Somatodendritic
Kv7.5	KCNQ5	Delayed rectifier, M current	Axons, Somatodendritic, terminals
Kv8.1	KCNV1	Modifier/silencer	ND
Kv8.2	KCNV2	Modifier/silencer	Absent in brain
Kv9.1	KCNS1	Modifier/silencer	Somatodendritic
Kv9.2	KCNS2	Modifier/silencer	Somatodendritic
Kv9.3	KCNS3	Modifier/silencer	Somatodendritic
Kv10.1	KCNH1	Delayed rectifier	ND: Somatodendritic?
Kv10.2	KCNH5	Non-inactivating, outward rectifier	ND: Somatodendritic?
Kv11.1	KCNH2	Rapid inactivating, inward rectifier	Somatodendritic, myelinated axons
Kv11.2	KCNH6	Rapid activating, delayed rectifier	Somatodendritic, myelinated axons
Kv11.3	KCNH7	Inward rectifier	Somatodendritic, myelinated axons
Kv12.1	KCNH8	Slowly activating, deactivating	ND
Kv12.2	KCNH3	Slowly activating, deactivating	ND
Kv12.2	KCNH4	Slowly activating, deactivating	ND

The basics: Kv structure, function and assembly

Through sequence comparisons, mammalian Kvs have been classified into 12 subfamilies (Kv1–12) [1] (Table I) of which Kvs 1–4 form functional homo- or hetero-tetramers while Kvs 5–12 are assembled as heteromers [1], [10]. Each Kv monomer – termed the α subunit – has a structure (Figure 1) comprised of an intracellular amino and carboxy- terminus and six transmembrane spanning domains designated as S1–S6. The S1–S4 region, collectively, forms a voltage-sensing domain, while S5–S6 and a re-entrant ‘pore-loop’ region in between them, contain the channel's gating machinery [6], [7] (Figure 1). Assembly of Kvs occurs within the endoplasmic reticulum (ER), primarily, through tetramerization of T1 domains [11], [12] located in the amino terminus of each subunit [9] (Figure 1). By stabilizing interactions between appropriate subunits and destabilizing inappropriate subunit interactions, the T1 domain directs assembly of subunits within, but not between, subunit families. For example, Kv1.4 can heterotetramerize with other Kv1 family subunits, but not with those from the Kv2, Kv3 or Kv4 families. Additional intra-subunit interactions are also likely, since removal of the T1 domains can still effect some functional Kv expression, albeit without the subunit family-specificity [12]. A notable exception to the ‘T1-intra-family’ rule, concerns the Kv5,6,8 and 9 ‘electrically silent’ subunits which fail to form functional channels unless co-assembled with Kv2 and/or Kv3 subunits [10], [13–18]. When expressed alone, the electrically silent subunits are retained in the ER [10], [16], most likely by a signal located in the C-terminal end of S6 [19], but are released from the ER upon complexation with Kv2 or Kv3 subunits. Studies on Kv9.1 and Kv9.3 suggest the heteromerization interactions are subunit specific, involving the amino termini for Kv2.1 or the hydrophobic core for Kv3.4 [17].

Figure 1.  Structure of the Kv channel. (A) Cartoon of single Kv α-subunit, showing transmembrane helices S₁-S₆. The N-linked glycosylation site, amino and carboxy termini are designated NLG, N and C, respectively. Side (B) and top (C) (extracellular) view of the structure of the Kv1.2 homotetramer (open state), coloured as in (A). Extra and intracellular loops are shown in blue or black, respectively. Structures were modelled and coloured using co-ordinates given in Pathak et al. (2007) [7].

Following their assembly, Kv tetramers proceed to the cell surface via the classical secretory pathway, passing through the ER-Golgi Intermediate Compartment (ERGIC), cis, medial and trans-golgi elements and then the Trans Golgi Network [9]. In transit, the Kvs can associate with sub-family specific accessory proteins such as Kvβ subunits [20] and experience post-translational modifications such as glycosylation, which modulate their surface expression and kinetics [9], [21].

From the earliest studies it has been evident that specific Kvs have discrete patterns of subcellular segregation [2], [9], [22], [23] (Table I). In general, Kv1 channels, especially those abundant in brain (Kv1.1, Kv1.2 and Kv1.4), localise to axons and nerve terminals [2], [24]. Although these subunits are often expressed in the same cell, they are thought to form select heterotetramers at specific locales. Thus, Kv1.1 seems to co-localize with Kv1.2 to the exclusion of Kv1.4 in terminals of cerebellar basket cells [25–28] and juxtaparanodal regions of myelinated axons [29], [30], but with Kv1.4 to the exclusion of Kv1.2 in striatal efferents [28]. Some exceptions have been noted, however [28], [31]. The delayed rectifier Kv2 channels are primarily homomers [32] localized to somata and dendrites [33–35] where they are thought to control excitability during periods of high action potential firing [34]. While Kv2.1 channels are often clustered [35], [36], Kv2.2 is distributed more uniformly in dendrites [32]. The distribution of Kv3 channels is complicated by their propensity to form both heteromers and C-terminal splice variants [37]. In general, Kv3 subunits are present in fast-spiking cells, notably interneurons, where they enhance the efficiency of re-polarization [38–40]. Typically, Kv3 channels are found in axon terminals, axons, nodes of Ranvier and somata, but differ in dendritic localization. Thus, the longer splice variant of Kv3.1 – Kv3.1b – is present in proximal dendrites, but its shorter variant – Kv3.1a – is restricted to axons [38], [41]. Interestingly, Kv3.1b seems to be present at the nodes of Ranvier, especially in large myelinated neurons [42]. The Kv3.2 channel is notable by its appearance in subsets of interneurons distinct from those containing Kv3.1 [43]. Both Kv3.3 and Kv3.4 are found in dendrites [44], [45] where they modulate large postsynaptic depolarization events but not the smaller excitatory postsynaptic potentials underlying spatial and temporal summation [44]. The Kv4 channels are invariably expressed in somatodendritic regions [22], [46], [47], especially in distal arbors [47], where they influence dendritic excitability, the back-propagation of action potentials and synaptic plasticity [48], [49]. The Kv7 channels Kv7.2–7.5, underlying the delayed rectifier, M (Muscarinic receptor-modulated)-current have been localized to somatodendrites [50]. However, there is also strong evidence for axonal Kv7 channels especially at the nodes of Ranvier [51] and axon initial segments [52], where they could regulate burst firing. The electrically silent subunits, Kvs 5,6,8 and 9 [1], representing approximately 25% of Kvs, have distributions which generally mirror those of the Kv2 and Kv3 series with whom complexation is obligatory for surface expression [13], [15], [16]. Whether such complexation occurs in neurons awaits confirmation.

Mechanisms for Kv localization

Currently, two major models for neuronal targeting exist (Figure 2) [53], [54]. In the first, targeting occurs via the initial somal sorting of proteins into vesicles which are then trafficked, selectively, to either the axons or dendrites. In the second model, vesicular delivery is not polarized, so proteins are delivered indiscriminately to axons and dendrites and then sorted at the cell surface to retain or remove appropriate or inappropriate components, respectively. Evidence is now accumulating that, the first model controls Kv targeting, especially in axons where, owing to their great lengths, a system of the utmost efficiency is required.

Figure 2.  Proposed models for Kv targeting in neurons. In the polarized vesicular trafficking model (A), cargo is sorted at the level of the trans-Golgi network (TGN) into vesicles which are then delivered to axons or dendrites, as appropriate. In the selective retention model (B), vesicular trafficking is not polarised, rather Kvs are trafficked into axons or dendrites, where they are retained in, or removed (large arrows) from, appropriate or inappropriate regions, respectively. Sorting, polarized trafficking and maintenance/retention steps are indicated as 1–3, respectively. Axonal membrane and Kvs are in white, dendritic membrane and Kvs are shown in black.

Selective targeting of Kvs into axons

Seminal work demonstrated that the A-type Kvs, Kv1.4 and Kv4.2, are segregated into axons and dendrites, respectively [22]. Although these Kvs represent good models for targeting, the mechanisms remained elusive until 2002, when sequences within the highly conserved T1 tetramerization domain of Kv1 channels were shown to specify axonal targeting [55]. The T1 domain is also responsible for interaction with the auxiliary Kvβ1 and Kvβ2 subunits [20] and it is these subunits which appear to be both necessary and sufficient for axonal targeting [55], [56]. Significantly, fusion of the T1 domain of Kv1.3 to CD8, a protein normally distributed throughout axons and dendrites causes expression of the T1-CD8 chimera at the surface and within axons, but never in dendrites. Conversely, fusion of the T1 domain to the transferrin receptor (TfR), a protein exclusive to dendrites, localizes the T1-TfR chimera to axons, in accord with the polarized delivery model [57].

Quite how Kvs experience polarized delivery is not fully understood but seems to involve the microtubule (MT)-based cytoskeleton. In neurons MTs have a uni-directional, plus- (distal growing-) end out orientation in nascent axons and dendrites, but, as the neuron acquires its polarized characteristics, the dendrites also, and quite specifically, acquire MTs with a reverse (minus end out) orientation [58]. Thus, it was proposed that proteins might be delivered to axons or dendrites as cargo in vesicles bearing plus or minus end-directed motor proteins. This notion was given considerable support by the recognition that several of the 45 different isoforms of the MT-dependent motor protein, kinesin, were able to direct vesicular trafficking into dendrites [59]. Thus, vectorial delivery of Kvs can be envisaged as reflecting attachment to so-called ‘smart’, destination-specific, MT-dependent motor proteins [53], [59]. The demonstration that the postsynaptic, and hence dendritic, NMDA-subtype of glutamate receptors, associates, via its NR2B subunits, with the vesicular motor protein Kif17, lend powerful support for the smart motor hypothesis and have spurred efforts to identify potential interactions of Kvs with components of the MT-dependent trafficking machinery [60]. In another tour de force of nerve cell biology, Jan's laboratory reported, recently, that axonal trafficking of Kvs requires the docking of Kvβ subunits with the +TIP protein EB1 a member of the TIP family of proteins associated with the growing tips of microtubules [56]. This observation is especially significant as it is now thought that MTs exhibit dynamic growth throughout the axons [56], [61]. Does this preclude a role for vesicular trafficking? It seems not, since these workers also showed that axonal targeting of Kv1 channels also requires an association with KIF3/kinesin II [56]. Thus, it seems likely that Kv1 channels are targeted to axons by selective MT motor-dependent trafficking machines, whose activity also requires or is modulated by complexation with EB1 and, perhaps, other MT binding proteins.

Selective targeting of Kvs into dendrites

Recent experiments show that targeting of the somatodendritic Kv4.2 channel also relies upon MT-based polarized vesicular trafficking. Thus, Kv4.2 is found within and at the surface of dendrites but not axons and, like NMDARs, is trafficked to dendrites via complexation with the MT-motor, Kif17 [62]. Unlike Kv1 channels, however, interaction with Kif17 involves the distal carboxy terminus of Kv4.2 rather than its amino-terminal T1 domains. The underlying mechanism is more complex than this simple interaction suggests, however, since mere Kif17-Kv4.2 association is insufficient to mediate dendritic targeting. Rather, dendritic targeting depends upon an evolutionarily conserved non-endocytic 16 amino acid, dileucine-containing motif within the carboxy terminus of Kv4.2 [63]. Surprisingly, the di-leucine motif is necessary and sufficient for dendritic targeting, but does not interact with Kif17, directly [62]. Consequently, targeting specificity is more likely to reflect regulation of Kif17 by proteins bound to the dileucine motif, than smart motor activity. Indeed, Kif17 mutants, incapable of binding vesicular cargo, are found in both axons and dendrites [64]. Moreover, association with specific cargo-attached proteins can define the motor directionality of kinesins [65].

Maintenance of Kv distributions in axons and dendrites

Irrespective of the precise targeting modality, segregation of Kvs inserted into the surface of axons and dendrites demands mechanisms for their retention in those membrane regions. In epithelial cells, proteins resident in the apical or basolateral surface membrane compartments are prevented from co-mingling, by virtue of the interfacial cuff of cytoskeletal and membrane proteins comprising the tight junctions [66]. Single particle tracking studies suggest a similar cuff exists at the axon hillock forming the boundary of the soma and axon. Thus, the axon hillock is envisaged to act as a picket fence preventing axonal membrane proteins, generally, from diffusing into the somatodendritic regions and vice versa [67]. As discussed below, epigenetic cues may be used to provide additional, more refined, specification of select Kv surface distributions at the nerve surface.

Dynamic regulation of Kv surface expression

A growing body of evidence indicates that Kv surface expression is controlled at the level of both the ER and the plasma membrane. Elegant studies by Deutsch indicate that Kvs attain their overall topology through sequential assembly of biogenic units and association of T1 domains during their co-translational insertion into the ER membrane [12], [68]. Any misfolded Kvs are, in most cases, probably degraded. However, procession of correctly folded Kvs through the secretory pathway does not always occur by default. Thus, the efficiency of ER export of Kv1 channels depends upon the presence of four specific pore-mouth residues, three of which also confer sensitivity to Kv blockade by dendrotoxin (DTX) [69]. Indeed, DTX-sensitive Kvs (Kv1.1, Kv1.2, and Kv1.6) are retained in the ER, while DTX-insensitive ones (Kv1.3, Kv1.4 and Kv1.5) are exported. Expression of DTX in the ER lumen also enhances the cell surface expression of Kv1.1, in homomers or heterotetramers [69] suggesting a DTX-like protein in the ER lumen regulates export of this Kv. Glycosylation also affects the stability and surface expression [9], [21], [70] of some Kvs (Kv1.4) but not others (Kv1.1) and also involves pore determinants [70]. Additional post-translational modifications may also be significant. For example, disruption of thioacylation seems to enhance the degradation of Kv1.5 in its early biogenesis [71].

In addition to ER lumenal mechanisms, export from the ER also seems to require interactions with Kv motifs accessible from the cytoplasm. Thus, mutation of a carboxy terminal VXXSL motif localizes Kv1.4 to the ER [72] by a mechanism which seems to require the presence of residues deep within the Kv pore [73]. In many cases, association of Kvs with modulatory auxiliary subunits also enhances their forward trafficking. Thus, early interactions of Kv1.2 with Kvβ2 and other Kvβ subunits promote N-linked glycosylation and enhance the stability and surface expression of Kv1.2-Kvβ complexes [20], [74].

Forward trafficking is also modulated by the interaction of a subset of Kvs with a second type of auxiliary subunit termed KChap – a member of the Protein Inhibitor of Activated Signal transducer STAT 3 (PIAS3) gene family. In co-expression studies, KChap enhances the functional expression of Kv1.3, Kv2.1, Kv2.2 and Kv4.3 [75], [76] without affecting their kinetics and gating. KChap also increases the total and surface levels of Kv2.1, Kv2.2 and Kv4.3 in a transcription-independent manner [75], [76]. While these data argue that KChap manifests ‘chaperone-like’ properties, the mechanisms are unclear. KChAP, is a cytosolic protein which binds transiently to the cytoplasmic amino termini of its target channels, but, unlike Kvβ subunits is not found at the membrane [75], [76]. Significantly, however, co-expression with Kvβ1.2 subunits reduces the functional effects of KChap, even on those Kvs (Kv2.1 and Kv4.3) which do not bind Kvβ subunits [77]. Conversely, KChap, blocks the actions of Kvβ1.2 on Kv1 subunits, including those (Kv1.4 and Kv1.5) whose functional expression is not enhanced. Since KChap binds Kvβ, it seems likely that Kvβ-KChap complexation blocks either subunit's activity. If this model is correct, the relative levels of Kvβ and KChap, may endow cells with an elegant mechanism for fine-tuning the physiological contribution of its Kvs [77].

A third type of auxiliary subunit – termed KChip – has been identified that, specifically, seems to regulate the forward trafficking of Kv4 channels [78]. The association of KChip proteins, via N-terminal residues and T1 domains in adjacent Kv4.3 α subunits [79], is thought to promote surface trafficking by masking intrinsic ER retention determinants [80]. However, KChip-Kv4 interactions may occur en route to, or within, the golgi apparatus and involve myristoylation [81]. Significantly, the Kv4-KChip interaction site seems to serve as a locus for several proteins which either impair (syntaxin 1A [82]), enhance (frequenin [83]) or modify (long pentraxin PTTX [84]) Kv4 expression or distribution. Enhancement of surface expression is also seen when Kv4 channels interact with certain integral proteins including DPPX (dipeptidyl-aminopeptidase-like protein X, DPP6) a CD26-related homologue of the membrane-bound DPP family of ectopeptidases and also DPPY (DPP10, DRP3) [85]. The DPP-Kv4 interaction appears to involve association of the transmembrane domain of DPP with S1–S2 in Kv4 and can yield ternary complexes with KChip [86]. Another group of single transmembrane auxiliary subunits proteins, termed KCNEs (mink and MiRPs1-3), also promote the surface trafficking of Kv3.1, Kv3.2, Kv4.2, Kv4.3 and Kv11.1 and in at least one case, do so by an early interaction in the secretory pathway [87].

The surface expression of Kvs is also controlled at the end-stages of their trafficking itineraries through endocytic mechanisms. Endocytic removal of Kvs represents a powerful and rapid means to regulate electrical excitability, but, with just a few exceptions [40], [88–91], has not been explored. Most Kvs have motifs conforming to canonical internalization signals (notably di-leucines and NPXY sequences) [92]) recognized by the clathrin-dependent internalization machinery, although other signals and internalization mechanisms probably exist. In HEK293 cells, Kv1.4 internalizes with a half-life of around 90 min [88] and likely involves carboxy-terminal di-leucine motifs [92]. Following internalization, Kv1.5 can experience Rab-GTPase-dependent endocytic recycling, via early endosomes, back to the cell surface, adding an additional level of control to surface expression [89]. Precisely, how Kvs become demarcated for internalization is poorly understood. Phosphorylation appears to be important (see below), but other mechanisms seem likely. Of these, the attachment of ubiquitin to lysine residues – ubiquitylation – is emerging as a key player. Ubiquitylation is known to direct many membrane proteins to the multi-vesicular body/lysosomal degradative pathway [93]. Thus, it is notable that the ubiquitin ligase NEDD-4-2 is reported to down-regulate Kv1.3 [94] and, just recently, Kv7.1-7.5 (KCNQs1-5) [95], [96]. Whether ubiquitylation occurs directly upon binding of NEDD-4-2 [95], [96] and/or involves other ubiquitin ligases or accessory proteins, e.g., KChap awaits clarification [94–96].

Dynamic regulation of Kv surface distributions

Although vectorial trafficking seems to specify axonal or dendritic targeting, Kvs are clearly not uniformly distributed in axons or dendrites. The simplest explanation is that Kv distributions are determined hierarchically (Figure 3): Initially, intrinsic programmes specify the vectorial delivery of Kvs into axons or dendrites, and then secondary, epigenetic mechanisms sculpt their distributions according to local need. What might these epigenetic mechanisms be and how would they operate to control local channel topography? The answer seems to lie in an intersection of the mechanisms which stabilize Kvs at the cell surface.

Figure 3.  Proposed model for Kv targeting in neurons. Here the overall surface density in specific regions of the nerve membrane is a product of Kv availability and distribution. Biogenic mechanisms determine Kv availability while Kv distributions are determined by sorting, trafficking and maintenance events in two stages. First, Kvs are delivered to axons or dendrites, where their distributions are sculpted according to local need.

In the absence of any restraining factors, membrane proteins are free to diffuse in the plane of the lipid bilayer. In the few cases examined, fluorescence photobleaching studies indicate that Kvs diffuse at rates typical of membrane proteins [97], [98]. However, most Kvs are not freely diffusible, but are immobilized by protein-protein interactions which promote their clustering and attachment to the underlying cytoskeleton [97]. A direct interaction of Kv4.2 with the cytoskeletal protein filamin has been described [99], but less direct interactions through intermediaries such as the actin-binding protein cortactin, and ankyrin have been reported [2], [91]. Many Kvs, including Kv1.4 [88], [100] and Kv4.2 [101], also dock via carboxy terminal motifs to PDZ (Post-Synaptic Density protein 95-Discs Large-ZO1) modules in membrane-associated guanylate kinase (MAGUK) scaffolding proteins [102]. Such interactions control Kv distributions on multiple levels including their incorporation into kinesin-based MT-motor complexes [103], cytoskeletal attachment [102] and clustering [88], [100], [101]. Significantly, the clustering interactions of Kv1.4 with the MAGUK, PSD-95, or cytoskeletal attachment, also prevent Kv internalization [88] and stabilize their axonal expression [104]. Consequently, Kvs probably exist in either free or clustered/immobilized pools, of which only the former are internalization-competent. Thus, the interactions of Kvs with MAGUKs, and/or with specialized regions of the underlying cytoskeleton, permit the mechanisms controlling lateral segregation and surface expression to intersect and effect a site-specific generation of Kv complexes. Precisely where such complexes are formed is unclear, but studies on the concentration of Kvs at the nodes of Ranvier have highlighted a major role for cell contact.

A role for cell contact?

In adult myelinated neurons, Kv1.1, Kv1.2 and Kvβ2 subunits are clustered under the myelin sheath in juxtaparanodal regions (JXPRs) through PDZ-mediated interactions which may involve PSD95 or, more likely, the transmembrane cell adhesion molecule (CAM) Caspr2 [105]. In development, Kv clusters appear concomitantly with myelination [106], [107], while hypomyelination and chronic demyelination induce a more diffuse Kv distribution [106], [107]. In rat sciatic nerve preparations, removal of the myelin sheath by intraneural application of lysolecithin, causes Kv1.1, Kv1.2 and Kvβ2 to re-distribute from the JXPRs and disappear [30]. However, during subsequent re-myelination, the subunits appear transiently at the nodes and then paranodes, after which they re-acquire their adult JXPR distribution. Taken together, these data provide compelling evidence that myelination is required for both the nucleation and maintenance of Kvs at JXPRs. Coordination of Kv distributions in JXPRs is, thus, envisaged to follow a sequence where glial cell adhesion molecules, such as TAG-1, on the surface of oligodendrocytes (CNS) or Schwann cells (periphery), interact with and concentrate, cognate axonal CAMs, notably Caspr2, at sites of axon-glial apposition [108]. The axonal CAMs then serve as sites for the nucleation of complexes of specialized cytoskeletal and scaffolding proteins which stabilize the CAM complexes and allow Kv accumulation. The organization of Kvs by extracellular contact may not be restricted to nodes of Ranvier. For example, Kv2.1, but not Kv2.2, is localized to proximal dendrites at sites close to contact by astrocytic processes [34]. Many Kvs are also expressed at sites of synaptic contact (Table I) [46]. Thus, in the visual cortex, Kv4.2 and Kv4.3 are found postsynaptically, opposite GABA terminals in excitatory neurons and interneurons, but at extra-synaptic locations, or peri-synaptically, at sites of excitatory input onto dendrites and spines [47].

Activity-dependent regulation of Kv expression, localization and maintenance

An exciting development has been the discovery that the density and distribution of Kvs which dynamically regulates excitability can in turn be regulated by neuronal activity. Here an intersection of all the machineries controlling Kv surface density is likely. In hippocampal neurons, suppression of synaptic activity, prior to the normal appearance of Kv1.1, Kv1.2, and Kv1.4 in development, appears to inhibit expression of these, but not other, Kv1 channels [109]. Unfortunately, the mechanisms are not well-defined. Indeed, the transcriptional and translational control mechanisms underlying Kv gene expression represent a fertile area, requiring much research. In contrast, much information is beginning to emerge on post-translational control of Kv expression. Thus, the reduction in Kv1.2 currents seen upon phosphorylation of a carboxy-terminal tyrosine residue seems to be due to a weakening of the Kv1.2-cortactin interaction and an enhancement in endocytosis [91]. Neuromodulation by second messenger systems appears to be critical. For example, at basal levels, cyclicAMP (cAMP) controls steady-state Kv1.2 turnover via a PKA-dependent pathway. However, elevations in cAMP increase the surface expression of Kv1.2 via a PKA-independent pathway that blocks endocytosis [110]. Growth factor signalling pathways may also be important as the growth factor receptor TrKB appears to up-regulate Kv1.3 by reducing its rate of internalization [111]. Significantly, the activation of synapses by glutamate [112] or muscarinic [113] receptor stimulation of rat hippocampal neurons, in culture, is reported to cause a rapid dispersal of Kv2.1 from somatodendritic clusters and a shift in its voltage-dependence of activation. The underlying mechanism appears to involve the rapid, calcineurin-mediated dephosphorylation of the Kv2.1 carboxy terminus, triggered by calcium influx from the plasma membrane or IP3 receptor-mediated release from intracellular stores [113]. Here, the Kv2.1 C-terminus seems to be both necessary and sufficient for the phosphorylation-dependent clustering, gating and modulation properties of Kv2.1 and can be transplanted to other Kvs. Activity-dependent regulation of Kv trafficking may also occur through control of ubiquitylation [94]. Recent studies on Kv1.5, however, have revealed another intriguing possibility involving covalent modification with a polypeptide homologous to ubiquitin – the Small Ubiquitin-related Modifier protein, SUMO [114]. Like ubiquitylation, SUMOylation is influenced by, and may be the point of convergence of, diverse signalling pathways on Kvs, many of which contain consensus SUMOylation sequences [114]. In Kv1.5, SUMOylation seems to modulate channel kinetics rather than trafficking. However, the presence of SUMO-specific E3 ligase activity in PIAS proteins, of which KChap is a member [75–77], raises the intriguing possibility that SUMOylation underlies the ability of KChap to enhance trafficking of some Kvs.

Over a decade ago, experimental and computational evidence began to emerge which showed that neurons could modulate channel conductances through feedback mechanisms, requiring sensors and effectors, designed to sustain their firing ‘identities’ [115]. Since the primary means by which voltage changes and neuromodulatory transmission are ‘sensed’ involves Ca^2 + entry through voltage-gated Ca^2 + channels or release from intracellular stores, respectively, cytosolic Ca^2 + may control feedback through trafficking ‘effectors’. Indeed, activation of CamKinase II seems to enhance Kv4.2 surface expression and, thereby, decrease the excitability of hippocampal neurons [116]. Growing evidence suggests the Kv4.2-KChip interaction is a major focal point for activity-dependent regulation of Kv4 trafficking. For example, Ca^2 +-dependent conformational changes in KChips, or their phosphorylation by the G-protein receptor kinase, GRK2, may promote forward trafficking of Kv4.2 [81], [117]. As KChip de-phosphorylation seems to involve the Ca^2 +-dependent phosphatase, calcineurin, it seems likely that coupled phosphorylation and Ca^2 +-dependent dephosphorylation steps control Kv4 trafficking to the cell surface. The forward trafficking of Kv4.2 into dendrites and spines may also be enhanced by a separate mechanism involving the Ca^2 +-dependent phosphorylation of the MAGUK, SAP97 [103]. Perhaps the ultimate denouement of how activity may affect Kv trafficking, and its functional significance, has come from very recent evidence showing that dendritic excitability, and long-term potentiation, are enhanced by activity- and Ca^2 +-dependent, trafficking of Kv4.2 in hippocampal neurons [45].

Conclusions

In this review we have gathered together evidence, much very recent, describing how the Kvs attain their regionalized subcellular distributions in neurons. We argue that the regionalized expression of Kvs is controlled, hierarchically, through motif-based mechanisms which control the assembly, vectorial delivery and stabilization of Kvs at the cell surface. Regulation of these events occurs through quality control mechanisms within the ER, and an intersection of the mechanisms which relay extracellular cues from synaptic activity and cell contact, with those that define assembly, trafficking and maintenance. Many questions remain unanswered. For example is this model correct? What are all the protein interactions and signalling mechanisms involved and how specific and robust are they? Do neurons acquire their identities by controlling Kv expression and distributions independently or in concert with other channels? Based on precedent, it seems the answers to these questions will be not long coming.

Acknowledgements

This work was supported by funds from the Biotechnology and Biological Sciences Research Council UK (34/C15752). Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.