Weighing the evidence for a ternary protein complex mediating A-type K⁺ currents in neurons

Abstract

The subthreshold-operating A-type K⁺ current in neurons (I_SA) has important roles in the regulation of neuronal excitability, the timing of action potential firing and synaptic integration and plasticity. The channels mediating this current (Kv4 channels) have been implicated in epilepsy, the control of dopamine release, and the regulation of pain plasticity. It has been proposed that Kv4 channels in neurons are ternary complexes of three types of protein: pore forming subunits of the Kv4 subfamily and two types of auxiliary subunits, the Ca²⁺ binding proteins KChIPs and the dipeptidyl peptidase-like proteins (DPPLs) DPP6 (also known as DPPX) and DPP10 (4 molecules of each per channel for a total of 12 proteins in the complex). Here we consider the evidence supporting this hypothesis. Kv4 channels in many neurons are likely to be ternary complexes of these three types of protein. KChIPs and DPPLs are required to efficiently traffic Kv4 channels to the plasma membrane and regulate the functional properties of the channels. These proteins may also be important in determining the localization of the channels to specific neuronal compartments, their dynamics, and their response to neuromodulators. A surprisingly large number of additional proteins have been shown to modify Kv4 channels in heterologous expression systems, but their association with native Kv4 channels in neurons has not been properly validated. A critical consideration of the evidence suggests that it is unlikely that association of Kv4 channels with these additional proteins is widespread in the CNS. However, we cannot exclude that some of these proteins may associate with the channels transiently or in specific neurons or neuronal compartments, or that they may associate with the channels in other tissues.

A-type K⁺ currents in brain and heart

The fast-transient subthreshold-operating A-type K⁺ current (I_SA) has fascinated elecrophysiologists since it was first characterized in molluscan neurons by Connor & Stevens (1971a). These authors immediately appreciated the computational possibilities of this current, and in an accompanying paper they presented computer simulations showing how it could regulate spike timing and firing frequency (Connor & Stevens, 1971b). Following these findings, evidence has accumulated demonstrating that I_SAs have fundamental roles in neuronal function in invertebrates and vertebrate species. The unique properties of the underlying channels, including rapid, transient activation in the subthreshold range of membrane potentials, fast inactivation, and fast recovery from inactivation, have been shown to cause delayed excitation, and to influence spike repolarization, the delay between depolarization and spiking, and the duration of the interspike interval. As a result, they help pace action potential firing during repetitive firing, thus contributing to the regulation of firing frequency (Llinas, 1988; Rudy, 1988; Baxter & Byrne, 1991; Hille, 2001; Liss et al. 2001), as illustrated by the finding that the pacemaker frequency of individual dopaminergic neurons in the substantia nigra, and hence their levels of dopamine release, is directly correlated with the density of the I_SA (Liss et al. 2001).

Later, starting with the seminal paper by Hoffman et al. (1997), the function of the I_SA in dendritic integration and plasticity has been increasingly appreciated. I_SAs have been implicated in several aspects of signal processing in dendrites, including the modulation of general dendritic excitability, the regulation of action potential back propagation into the dendritic tree, the filtering and integration of synaptic potentials, distance-dependent synaptic scaling, the propagation of dendritic Ca²⁺ plateau potentials, the establishment and expression of long-term potentiation and the branch-specific compartmentalization of excitability and information storage (Hoffman et al. 1997; Schoppa & Westbrook, 1999; Johnston et al. 2000; Ramakers & Storm, 2002; Watanabe et al. 2002; Cai et al. 2004; Frick et al. 2004; Magee & Johnston, 2005; Chen et al. 2006; Losonczy & Magee, 2006; Kim et al. 2007).

More recently, evidence is starting to accumulate that this current is also important in disease. For example, a number of observations have suggested a link between I_SA and epilepsy, both in animal models and in humans (e.g. Beck et al. 1997; Bernard et al. 2004; Ruschenschmidt et al. 2004). Direct evidence was recently found when a truncation in the gene encoding Kv4.2 (KCND2) producing a protein lacking the last 44 amino acids in the carboxyl terminal region was identified in a patient with temporal lobe epilepsy (TLE) (Singh et al. 2006). Studies in mice have also suggested an involvement of I_SA channels in pain mechanisms (Hu et al. 2006).

The physiological significance of I_SA channels in health and disease underlies the effort to elucidate the molecular composition of these channels. A current similar to the I_SA in neurons, known as I_to (for transient outward current) is found in cardiac ventricular myocyes where it has an important role in the repolarization of the cardiac action potential.

Structure of I_SA channels

Voltage-gated K⁺ channels are tetramers of primary, or pore-forming subunits (sometimes called α subunits), arranged with a fourfold symmetry around a central pore (Doyle et al. 1998; Long et al. 2005) (Fig. 1). This tetramer forms the infrastructure of the channel and in most cases is sufficient to form a functional channel. However, in addition, many voltage-gated K⁺ channels contain auxiliary subunits (sometimes referred to as β subunits) that have primary sequences not resembling pore-forming subunits.

Figure 1. Model of a Kv4 channel complex with Kv4 pore-forming subunits, and auxiliary subunits KChIPs and DPP6.

This figure was obtained by adding molecules of DPP6 (based on the crystal structure of the DPP6 extracellular domain obtained by Strop et al. (2004) to the model of the Kv4.3-KChIP1 channel complex by Wang et al. (2007). The four Kv4.3 subunits are labelled in cyan, yellow, red and green. The four KChIP1 proteins interacting with the channel T1 domains, are labelled in blue. The transmembrane domains of DPP6 are shown interacting with the membrane spanning helices of the voltage sensor domains. Only two DPP6 molecules are shown to facilitate viewing, but the channel complex is likely to include four DPP6 proteins (Soh & Goldstein, 2008). Each DPP6 subunit includes an α/β hydrolase domain and a β propeller. The extracellular domains of two DPP6 proteins interact forming a dimer.

(K⁺ channels of the Kv2 subfamily associate with proteins of the Kv5, Kv6, Kv8 and Kv9 subfamilies. The members of these subfamilies have sequence similarity to other Kv proteins but do not express functional channels in the absence of Kv2 proteins. However, they form part of the infrastructure of the pore-forming channel complex and hence we considered them silent pore-forming subunits, rather than typical ‘auxiliary subunits’ (Coetzee et al. 1999; Rudy et al. 2008).)

Auxiliary subunits can modify the properties of the channels, often significantly, and in some cases might be essential for the efficient expression of functional channels in the plasma membrane. Other voltage-gated ion channels, such as sodium and calcium channels also contain auxiliary subunits with crucial roles in determining the localization and function of the channels (Isom, 2001; Tseng et al. 2007).

There is a large number of Kv subunits, the pore-forming subunits of voltage-gated K⁺ channels, which are classified in 12 groups or subfamilies based on sequence similarities. This diversity allows for the generation of many subtypes of voltage-gated K⁺ channels. It is now well established that proteins of the Kv4 subfamily of K⁺ channel proteins are the pore-forming subunits of the channels mediating most of the somato-dendritic I_SA in neurons (Jerng et al. 2004a; Rudy et al. 2008). A total of three genes encoding Kv4 subunits have been identified in mammals: Kv4.1, Kv4.2 and Kv4.3 (Baldwin et al. 1991; Pak et al. 1991; Salkoff et al. 1992; Serodio et al. 1996). The Kv4.3 gene encodes two variants via alternative splicing: Kv4.3S and Kv4.3L (short and long). Two of these genes Kv4.2 and Kv4.3, but not Kv4.1 are prominently expressed in brain (Serodio & Rudy, 1998).

A large number of proteins, probably more than for any other channel group, have been suggested to associate with Kv4 pore forming subunits and function as auxiliary subunits. This work presents a critical review of the evidence supporting a role for these proteins as auxiliary subunits of Kv4 channels.

Auxiliary subunits of Kv4 channels

A major impulse for searching for auxiliary subunits of Kv4 channels was the observation that, while the evidence strongly argued that subunits of the Kv4 family were the pore-forming subunits of I_SA channels in neurons and the I_to in the heart, the currents observed in heterologous cells expressing Kv4 proteins differed substantially from native currents. In particular, inactivation rates were slower, and most importantly, recovery from inactivation, a key property of I_SA channels, was substantially slower (Serodio et al. 1994; Serodio et al. 1996). Fast recovery from inactivation distinguishes Kv4-mediated currents from other transient K⁺ currents and is an essential feature of I_SA channels given their role in repetitively firing cardiac and neuronal cells.

Early efforts to reconstitute the channels mediating the I_SA showed that coexpression in Xenopus oocytes of Kv4 proteins with brain mRNAs lacking Kv4 transcripts could restore many of the properties seen in native A-type K⁺ currents (Rudy, 1988; Chabala et al. 1993; Serodio et al. 1994, 1996). The first Kv4 channel associated proteins found, the so-called K⁺ channel interacting proteins or KChIPs were identified in 2000.

K⁺ channel interacting proteins

K⁺ channel interacting proteins (KChIPs) were identified in yeast-two hybrid screens for Kv4 associated proteins using the Kv4.3 N-terminal region as bait (An et al. 2000). Four KChIP genes (KChIP1–4) are now known, each producing multiple products via alternative splicing. At least 12 different KChIPs, spliced products of the KChIP1–4 genes, have been isolated to date from brain and heart tissue (An et al. 2000; Bahring et al. 2001; Holmqvist et al. 2002; Patel et al. 2002; Takimoto & Ren, 2002; Boland et al. 2003; Van Hoorick et al. 2003; Jerng et al. 2004a; Burgoyne, 2007).

KChIPs are small proteins (200–250 amino acids long), of the recoverin neuronal calcium sensor (NCS-1) family. They consist of a variable N-terminal region and a highly conserved C-terminal ‘core’ with four EF hand-like Ca²⁺-binding motifs (An et al. 2000).

KChIPs modify Kv4 currents in heterologous cells. KChIPs facilitate the trafficking, subunit assembly, stability and surface expression of Kv4 channel complexes (An et al. 2000; Shibata et al. 2003). Hence, coexpression of Kv4 proteins and KChIPs results in large increases (as much as 10-fold) in the amplitude of the expressed Kv4 current. In mammalian heterologous expression systems transfected only with Kv4 cDNAs, the expressed Kv4 proteins are retained in the endoplasmic reticulum and poorly transported to the plasma membrane. Additionally, KChIPs modify the electrophysiological properties of the channels expressed by Kv4 proteins in heterologous expression systems, which in general inactivate more slowly and recover from inactivation with faster kinetics in the presence of KChIPs (An et al. 2000; Rosati et al. 2001; Holmqvist et al. 2002) (Fig. 2). KChIPs also produce small shifts in the conductance–voltage curve, although the reported magnitude and direction of this effect varies between different studies. All KChIP isoforms have similar effects on Kv4 channels except for the alternative spliced version KChIP4a, which does not promote surface expression and prevents channel inactivation (An et al. 2000; Holmqvist et al. 2002).

Figure 2. Reconstitution of the I_SA in HEK 293 cells.

Shown are the currents evoked by voltage steps from –90 to +60 mV in 10 mV increments from a V_H of –90 mV in HEK 293 cells transfected with the indicated cDNAs or in cerebellar granule neurons (CGNs). Cells expressing similar current levels where selected for this figure to emphasize differences in channel kinetics. The red trace corresponds to –30 mV in Kv4.2, –40 mV in Kv4.2 + KChIP1 and –50 mV for the two lower panels. The currents expressed by Kv4.2 proteins in HEK 293 cells in the presence of KChIP1 and DPPX-S resemble closely the I_SA in CGNs. The currents recorded in cells expressing only Kv4.2 or Kv4.2 and KChIP1 activate at more depolarized potentials and have different kinetics. Horizontal scale bar: 50 ms for all traces. Vertical scale bar: 1 nA for HEK cells and 0.5 nA for CGN.

Van Hoorick et al. (2003) reported that the KChIP1 isoform KChIP1b had the opposite effect of KChIP1a on the kinetics of recovery from inactivation (significantly slowing inactivation recovery). Although this could have interesting functional consequences in cells expressing this isoform, Boland et al. (2003) found that both KChIP1 variants produce similar recovery rates that were faster that those of Kv4.2 alone.

Although other than these differences all KChIP proteins appear to have similar effects on Kv4 channels in heterologous cells, each is expressed in specific neuronal populations in the CNS, suggesting specific, yet to be determined requirements for Kv4-KChIP combinations in different neurons (Rhodes et al. 2004). Many neuronal populations express two or more KChIPs; however, it is not known whether the same channel complex includes more than one type of KChIP.

Evidence that KChIPs are components of native Kv4 channels

A large and diverse series of experimental observations strongly support the notion that KChIP proteins are associated with Kv4 proteins in brain and heart and function as auxiliary subunits of native Kv4 channels in these tissues. These include the following.

1. KChIPs and Kv4 proteins can be reciprocally coimmunoprecipitated from non-denaturing brain extracts, suggesting they are part of the same molecular complex (An et al. 2000; Ohya et al. 2001; Holmqvist et al. 2002; Rhodes et al. 2004).

2. KChIPs colocalize with Kv4.2 and Kv4.3 proteins in brain. They are expressed in the same neuronal populations and have the same subcellular distribution (An et al. 2000; Rosati et al. 2001; Rhodes et al. 2004).

3. As described further below, KChIPs are one of the main components of Kv4 channel complexes purified from brain (Nadal et al. 2003).

4. In the ventricular wall of the heart, KChIP2 (the isoform present in cardiac tissue) displays an expression gradient that parallels the gradient in I_to (Rosati et al. 2001; Zicha et al. 2004) Furthermore, I_to is nearly completely abolished in ventricular myocytes from mice lacking KChIP2 (Kuo et al. 2001).

5. KChIPs are necessary (but not sufficient, as discussed later) to reconstitute Kv4 channels with properties resembling neuronal I_SA in heterologous cells. In particular, KChIPs are necessary to explain the inactivation mechanisms of native I_SA in some neuronal populations (see below).

6. Arachidonic acid has kinetic effects on neuronal I_SA that are not seen on the currents expressed by Kv4s alone, but are reproduced in heterologous cells expressing Kv4 proteins in the presence of KChIPs (Holmqvist et al. 2001). The ability of protein kinase A (PKA) to modulate the conductance–voltage curve of native Kv4.2 channels was also suggested to require coexpression of KChIPs in heterologous cells (Schrader et al. 2002).

7. There is a large regional and cell-specific decline in the levels of expression of KChIPs in Kv4.2 knockout mice, indicating that the expression of the two types of protein is coregulated (Menegola & Trimmer, 2006). This suggests that the predominant role for neuronal KChIPs is as auxiliary subunits of Kv4 channels.

Mechanism of KChIP modulation of Kv4 channels

The primary effects of KChIPs on Kv4 channels in heterologous cells are to facilitate trafficking of Kv4 channels to the plasma membrane, to slow macroscopic inactivation and to accelerate the recovery from inactivation of Kv4 channels. Wang et al. (2007) and Pioletti et al. (2006) recently described a cocrystal structure of the complex formed between the Kv4.3 N-terminus and KChIP1. The structure shows that each KChIP1 molecule (4 in the complex) interacts with two neighbouring Kv4.3 N-termini in a 4: 4 manner, forming a cross-shaped octamer. The proximal N-terminal peptide of one Kv4.3 N-terminus is sequestered (‘clamped’) by its binding to an elongated hydrophobic groove on the surface of KChIP1. At the same time, each KChIP1 binds to the T1 domain of an adjacent Kv4.3 subunit to stabilize the tetrameric Kv4.3 complex. T1 or tetramerization domains are present in the N-terminal region of Kv proteins and determine subfamily-specific association of pore-forming subunits. The stabilization of the tetrameric Kv4 complex by KChIPs is likely to contribute to facilitating the trafficking of channels to the plasma membrane (Cui et al. 2008). In addition, Shibata et al. (2003) have suggested that by binding to the N-terminus of Kv4 proteins, KChIPs may mask a cytoplasmic retention/solubility motif.

The sequestration of the N-terminus is also likely to explain how KChIPs slow down inactivation (Pioletti et al. 2006; Wang et al. 2007), by preventing a fast open-state inactivation resembling the N-type inactivation in other transient K⁺ channels (Gebauer et al. 2004). As a result, the channels inactivate more slowly, and predominantly from closed states (Jerng et al. 2004a). The mechanism by which KChIPs accelerate inactivation recovery is less clear.

Dipeptidyl peptidase like proteins

DPP6 and DPP10 – new auxiliary subunits of Kv4 channels: discovering a new function for dipeptidyl peptidase like proteins

Dipeptidyl peptidases (DPPs) are an important family of proteolytic enzymes that hydrolyse dipeptides from a number of protein targets resulting in the activation or inactivation of the substrate (Lambeir et al. 2003; Gorrell et al. 2006; Maes et al. 2007). Unexpectedly, purification of Kv4 channel complexes from rat cerebellar membranes suggested that a member of this protein family known as DPP6 or DPPX was an integral component of Kv4 channels in neurons (Nadal et al. 2003).

Kv4.2–KChIP complex produces incomplete I_SA channels

The association of DPP6 with Kv4 proteins was discovered as a result of continuing efforts to reconstitute the I_SA in heterologous expression systems. Evidence from studies in neurons and cardiac tissue accumulated supporting the conclusion that KChIPs were a component of native Kv4 channels functioning as auxiliary subunits. KChIPs significantly facilitated Kv4 channel expression at the plasma membrane in heterologous cells. Furthermore, one of the effects of KChIPs on Kv4 currents in heterologous cells, the acceleration of the kinetics of recovery from inactivation, was in the right direction to increase resemblance with native currents (An et al. 2000; Jerng et al. 2004a).

However, in the presence of KChIPs other properties become less similar to those of native currents. In particular, the kinetics of inactivation are slowed by KChIP coexpression whereas the kinetics of inactivation of native I_SA channels in many populations of neurons is faster, not slower, than the inactivation of the channels produced by Kv4 expressed alone (Figs 2 and 3). This raised the possibility that native I_SA channels may include additional factors that were missing in the heterologously expressed channels.

Figure 3. Kv4 channel properties depend on subunit composition.

Voltage dependence of activation, steady-state inactivation and kinetic properties of Kv4 channels composed of Kv4.2, Kv4.2 + DPP6-S, Kv4.2 + KChIP1, or Kv4.2 + KChIP1 + DPP6-S in CHO cells at room temperature.

Speeding up inactivation: the search for a K⁺ channel accelerating factor (KAF)

Further evidence supporting the existence of a third protein came from experiments showing that a specific fraction of cerebellar mRNA of large molecular size, when coexpressed with Kv4 proteins in heterologous cells produced currents with faster kinetics. The unidentified component in the mRNA responsible for accelerating the current was dubbed the K⁺ channel accelerating factor (KAF) (Nadal et al. 2001).

Nadal et al. (2003) used classical biochemical techniques to immunopurify Kv4 channel complexes from solubilized rat cerebellar membranes. The complex consisted of three major polypetides, including Kv4.2, KChIP and an unidentified component of ∼115 kDa present in the complex in apparent similar abundance. Sequencing revealed that this polypeptide corresponded to a previously identified integral membrane protein for which no function was previously known, the DPP-like protein DPP6.

DPP6 is a member of the dipeptidyl aminopeptidase family of proteins, with approximately 30% identity and 50% similarity to a prominent member of this family known as DPPIV or CD26 (Nadal et al. 2003). However, in contrast to DPPIV, and all other members of the family known at the time, DPP6 has mutations in the active site and lacks enzymatic activity (Wada et al. 1992; Kin et al. 2001). The effects of DPP6 on the currents expressed by Kv4.2 in heterologous cells and knockdown experiments demonstrated that DPP6 was KAF, the unidentified factor that accelerated Kv4 channel kinetics.

Effects of DPP6 on Kv4 channels in heterologous expression systems

Like KChIPs, DPP6 facilitates the trafficking of Kv4 channels to the plasma membrane in heterologous cells, resulting in up to 20-fold increases in current density (Nadal et al. 2003, 2006). This dramatic effect is largely due to an increase in the trafficking of channel complexes to the plasma membrane. Also like KChIPs, DPP6 accelerates the rate of recovery from inactivation, even to a larger degree than KChIP (Fig. 3). DPP6 also produces substantial hyperpolarizing shifts in the voltage dependence of activation and inactivation of Kv4 channels, whether they include KChIPs or not, and increases the rate of activation and inactivation (Nadal et al. 2003, 2006; Zagha et al. 2005; Amarillo et al. 2008) (Fig. 3). All these are effects that can profoundly affect neuronal properties, particularly at subthreshold membrane potentials.

DPP10: another DPP that lacks enzymatic activity and modifies Kv4 channels

Following the report that DPP6 associates with Kv4 channels, a new member of the DPP family of proteins, named DPP10, was identified (Qi et al. 2003). DPP10 was found to be prominently expressed in brain and like DPP6 to have mutations in the catalytic triad and lack DPP activity. Phylogenetic analysis showed that DPP10 and DPP6, which we collectively call DPP-like proteins or DPPLs, form an evolutionarily divergent subfamily within the extended DPPIV-like family, sharing 51% amino acid sequence identity, compared to 32% identity between DPP10 and DPPIV. DPP6 and DPP10 have nearly identical transmembrane and juxtamembrane domains, sharing 92% similarity compared with 41% similarity between DPP10 and DPPIV. The extracellular domains of the three proteins also have similar structure. However, DPP6 and DPP10 are more similar to each other than either is to DPPIV (Zagha et al. 2005).

Jerng et al. (2004b), Zagha et al. (2005) and Ren et al. (2005) showed that DPP10, like DPP6, facilitated Kv4 protein trafficking to the plasma membrane and had similar effects on the voltage dependence and kinetic properties of Kv4 currents, which were not shared by DPPIV. Similar to KChIPs, there might be a certain degree of division of labour between DPP6 and DPP10 in brain. Zagha et al. (2005) observed that DPP6 mRNAs were preferentially expressed in neurons that contain predominantly Kv4.2 (hippocampal pyramidal neurons, striatal medium spiny neurons and cerebellar granule cells) while neurons that express predominantly Kv4.3 (for example Purkinje cells and hippocampal interneurons) preferentially expressed DPP10. Some neurons may express both genes.

Evidence for a ternary Kv4–KChIP–DPPL complex in neurons: critical evaluation of DPPLs as auxiliary subunits of neuronal Kv4 channels

Nadal et al. (2003) found that DPP6 proteins copurified with Kv4.2 proteins from rat cerebellar membranes, suggesting that they are associated in the same molecular complex. However, at least in detergent solution this interaction seemed to be weak, requiring cross-linking prior to solubilization in order to optimize recovery of a stable complex in detergent solution. Furthermore, the finding that DPP6 was an associated protein of Kv4 channels was without precedent in the DPP family of proteins, none of which had been shown to function as an associated protein of other membrane proteins. It has therefore been critical to validate the association between DPPLs and native Kv4 channels and their role as auxiliary subunits.

Evidence that DPPLs are auxiliary subunits of Kv4 channels in neurons and that these channels are likely to be ternary complexes including KChIPs and DPPLs is now very strong and includes the following.

1. All three types of protein, Kv4s, KChIPs and DPPLs, can be immunoprecipitated from brain extracts with antibodies to any of the three components (Nadal et al. 2003; Jerng et al. 2005; Zagha et al. 2005).

2. DPPLs meet clear, predetermined functional criteria when coexpressed in heterologous expression systems. Kv4 channels containing KChIPs and DPPLs resemble more closely I_SA channels in neurons (Fig. 2) (Nadal et al. 2003; Jerng et al. 2005; Jerng et al. 2007; Amarillo et al. 2008). In the absence of DPPLs, the voltage range of activation and inactivation of Kv4 channels in heterologous cells is not as negative as in native channels. Furthermore, without DPP6 the channels inactivate more slowly and recover more slowly from inactivation than I_SA channels in neurons (see summary in Fig. 3). After 20 years of effort we can now with confidence say that we can reconstitute the neuronal I_SA in heterologous cells (Fig. 2).

   The similarities in inactivation kinetics of ternary and native channels are particularly striking. The kinetics of inactivation of ternary Kv4 channels expressed in heterologous expression systems has an unusual voltage dependence: the rate of inactivation slows down with increasing depolarization. This unusual behaviour is caused by KChIPs and is also observed in cells expressing Kv4 and KChIP proteins but not in cells expressing only Kv4 proteins or Kv4 and DPPLs without KChIP (Jerng et al. 2007; Amarillo et al. 2008).

   The precise profile of the voltage dependence of the inactivation rates observed in ternary channels depends in a complex fashion on the closed state inactivation promoted by KChIPs and the hyperpolarizing shift in voltage dependence and acceleration of the inactivation rate produced by DPPLs. As described earlier, KChIPs bind and immobilize the N-terminus of Kv4 proteins, preventing a fast open-state N-type inactivation. As a result, the channels inactivate predominantly from closed states (Beck et al. 2002; Jerng et al. 2004a, 2007; Amarillo et al. 2008). Closed-state inactivation slows down during large depolarizations due to increased occupancy of open states and hence reduced residency in the inactivation-prone pre-open closed state. On the other hand, DPPLs shift the channel's voltage dependence, shifting the voltage range in which the inactivation rate slows down with depolarization towards more negative potentials, and produce an overall acceleration of the inactivation rate. Therefore, the relation between inactivation rate and membrane potential is distinct for channels only containing KChIP or DPPLs and channels containing both auxiliary subunits (Amarillo et al. 2008) (Fig. 4).

   Amarillo et al. (2008) found that the profile of the voltage dependence of the inactivation rate of the I_SA in cerebellar granule neurons was remarkably similar to that observed for the currents expressed by ternary channel complexes (Figs 2 and 4). A similar profile has also been seen in other neurons, including hippocampal pyramidal cells, neurons in which the I_SA has been extensively studied (Klee et al. 1995; Hoffman et al. 1997; Martina et al. 1998; Lien et al. 2002). The finding that this complex behaviour, not seen in other fast transient voltage-gated K⁺ channels, is reproduced by ternary channels containing Kv4, DPPLs and KChIP proteins provides particularly strong evidence that the somatically recorded I_SA in neurons is mediated by Kv4 channels that include the two types of auxiliary subunit (Jerng et al. 2007; Amarillo et al. 2008).

   The effect of DPP6-S on the unitary conductance of Kv4 channels in heterologous cells is also unusual for an auxiliary subunit and supports the hypothesis that these proteins are part of the native channel complex. DPP6-S has been found to increase the unitary conductance of Kv4 channels in heterologous cells by ∼70% from ∼4.0 to ∼7.0 pS (Rocha et al. 2004; Kaulin, Santiago-Custillo, Rocha, Nadal, Rudy and Covarrubias unpublished observations). This results in a unitary conductance close to that reported (6–8.5 pS) for I_SA channels in hippocampal CA1 and neocortical pyramidal neurons and cerebellar granule cells (Hoffman et al. 1997; Bekkers, 2000; Chen & Johnston, 2004). These observations suggest that DPPX is necessary and sufficient to set the unitary conductance of neuronal Kv4 channels.

3. DPP6 and Kv4.2 proteins are colocalized in the same neuronal populations in brain, and within these precisely in the same neuronal compartments (Clark et al. 2008) (Fig. 5).

4. Knockdown of DPP6 in CA1 hippocampal pyramidal neurons using small interfering RNAs (siRNAs) or in cerebellar granule neurons from DPP6—/– mice produced changes in the properties of the I_SA consistent with the effects of DPP6 on Kv4 currents in heterologous expression systems (Kim et al. 2008; Zagha et al. 2008).

Figure 4. Voltage dependence of the inactivation rate of Kv4 channels.

A, the half-inactivation time (t_1/2) as a function of membrane potential of the currents recorded in HEK cells expressing Kv4.2 alone (•), Kv4.2 and DPP6-S (○), Kv4.2 and KChIP1 (▵) and Kv4.2 + KChIP1 and DPP6-S (▴). Note that the relation between inactivation rate and membrane potential is distinct for channels only containing KChIP or DPP6 and channels containing both auxiliary subunits. B, half-inactivation time (t_1/2) as a function of membrane potential of I_SA isolated from cerebellar granule cells in Kv4.2-containing anterior cerebellar lobules (black) and Kv4.3-containing posterior lobules (red). Note that the profile of the voltage dependence of the inactivation rate of the I_SA in CGNs is remarkably similar to that observed for the currents expressed by ternary channel complexes containing Kv4.2, KChIP1 and DPP6-S. From Amarillo et al. (2008).

Figure 5. Enrichment of Kv4.2 and DPP6 in the dendrites of hippocampal neurons and cerebellar granule cells.

A–C, confocal image of the CA1 field of the hippocampus in mouse double immunostained for Kv4.2 (A, red) and DPP6 (B, green) and the overlay of the two channels (C). Strong staining with both antibodies is observed in pyramidal cells’ apical and basal dendrites, while pyramidal cell somata are only weakly stained. Scale bar A–C, 20 μm. D and E, confocal fluorescence images of the cerebellar cortex single labelled with antibodies to Kv4.2 (D, red) or DPP6 (E, green). Immunostaining for both Kv4.2 and DPP6 is prominent in the granule cell layer (G), while the molecular layer (M) is weakly stained in both instances. Also, neither antibody labelled Purkinje cells (P). The insets show high magnifications of the granule cell layer demonstrating high density labelling of the cerebellar glomeruli (arrowheads) where granule cell dendrites contact the unstained cerebellar mossy fibre axons. Granule cell somata are more weakly stained, and the label is restricted to the periphery of the cell. Scale bars: 100 μm for main panels in D and E; 10 μm for the insets in D and E. (Addapted from Clark et al. 2008)

DPPLs are key determinants of the diversity of A-type currents in neurons

Somato-dendritic I_SAs vary across a number of properties in different neuronal populations. The molecular basis of this diversity is not well understood.

Alternative splicing of the N-terminus of DPP6 and DPP10 produces a large number of distinct isoforms which differ only in short intracellular domains following common extracellular, transmembrane and juxtamembrane domains. Five DPP6 isoforms (DPP6-S, DPP6-L, DPP6-E, DPP6-D, and DPP6-K) and four DPP10 splice variants (DPP10-A to DPP10-D) have been described (Nadal et al. 2006; Takimoto et al. 2006; Jerng et al. 2007). DPP6-S, L and K have similar effects on Kv4 channel properties, small differences were observed on the effects of these isoforms on the rate of recovery from inactivation and on the shift in the voltage dependence of inactivation (Nadal et al. 2006). On the other hand, Jerng et al. (2007) found that DPP10-A produced currents that inactivated much faster than those produced by other DPP10 isoforms and lacked the slow down of the inactivation rate during large depolarizations induced by KChIPs. The N-terminus of DPP6-E shows sequence similarity to the N-terminus of DPP10-A and was found to induce similar effects as DPP10-A (Maffie et al. 2008).

In situ hybridization with spliced version-specific probes revealed that distinct neuronal populations express combinations of different DPP6 and DPP10 isoforms, suggesting that DPPLs can contribute to the diversity of I_SA in different neuronal populations (Nadal et al. 2006; Jerng et al. 2007). Strongly supporting this hypothesis, DPP6-E was found to be expressed in neurons in intermediary layers of the superior colliculus, which have I_SAs that inactivate much faster than in other neurons and lack the slowing of inactivation with increasing depolarization (Maffie et al. 2008). DPPLs form dimers and each channel complex is likely to include two DPPL dimers (Strop et al. 2004; Soh & Goldstein, 2008) (Fig. 1). Since different DPPLs can form heterodimers (Strop et al. 2004) there are a large number of possible DPPL combinations which can contribute to I_SA diversity.

Mechanisms of action of DPPLs

One the most fundamental effects of DPPLs appears to be facilitating the voltage-dependent gating transitions that activate the channels, resulting in large hyperpolarizing shifts in voltage dependence. Other effects of DPPLs might be consequences, at least in part, of this shift in voltage dependence. Recently Dougherty & Covarrubias (2006) found that DPP6-S causes a –26 mV shift in the gating charge–voltage (Q–V) relation, suggesting that DPP6 remodels the channel's voltage sensor. This is interesting given evidence suggesting transmembrane interactions between DPPL proteins and Kv4 subunits, possibly directly with the voltage sensor.

Zagha et al. (2005) found that chimeric DPPL proteins with extracellular domains replaced by a series of Myc tags were able to reproduce many of the effects of DPP10 and DPP6 on Kv4 channels, suggesting a transmembrane interaction between DPPLs and Kv4 proteins. This is supported by mutagenesis studies by Ren et al. (2005) suggesting that the S1 and S2 domains in the Kv4.3 voltage sensor determine association with DPP10.

Kv4-unrelated functions of DPPLs and KChIPs?

KChIP3 has been independently identified three times and assigned two additional roles besides its function as an auxiliary subunit of Kv4 channels. It was identified as DREAM (downstream regulatory element antagonist modulator), a Ca²⁺ regulated transcription repressor shown to bind the dynorphin response element (Cheng et al. 2002), and as calsenilin, a protein that binds and modulates some of the effects of the Alzheimer's disease-related protein presenilin. Observations with KChIP3 knockouts have provided evidence supporting these Kv4-unrelated functions. Knockouts of the KChIP3 locus produce animals with decreased pain sensitivity as a result of enhanced trascription of dynorphin, an opiate neuropeptide (Cheng et al. 2002). Another study found that KChIP3 knockouts had reduced Aβ levels and enhanced long-term potentiation consistent with its presenilin interaction and channel modulatory roles (Lilliehook et al. 2003).

DPP6 may also have Kv4-unrelated functions. It has been shown, as described earlier, that overall, DPP6 proteins are expressed in the same neuronal populations, and within these in the same neuronal compartments as Kv4 proteins. However, in one brain structure, the hippocampal mossy fibre axons, which lack Kv4.2 or Kv4.3 proteins, there is prominent expression of DPP6. The function of DPP6 in these axons remains to be investigated; however, the finding suggests that DPP6 may also have Kv4-unrelated functions in brain (Clark et al. 2008).

Other putative Kv4 channel-associated proteins

A surprisingly large number of additional proteins have been suggested to associate with Kv4 subunits. Most of these have not been pursued by workers in the field as enthusiastically as KChIPs and DPPLs, and additional validation that they interact and modulate Kv4 channels in native tissue, and therefore that the association is physiologically relevant is required. The list includes the following.

1. All three Kvβ subunits (Yang et al. 2001; Deschenes & Tomaselli, 2002), small cytoplasmic proteins known to associate with subunits of the Kv1 subfamily facilitating the trafficking of Kv1 channels to the cell surface and modulating their inactivation properties (Pongs et al. 1999; Long et al. 2005; Rudy et al. 2008).

2. The K⁺ channel accessory protein (KChAP), a member of a family of transcription factor-binding proteins that has been suggested to enhance Kv2, Kv1 and Kv4 subunit surface expression (Wible et al. 1998; Kuryshev et al. 2000) via unclear mechanisms since KChAP was not detected at the cell surface of transfected cells.

3. Several of the products of KCNE1–5 genes, MinK and MinK-related peptides (MiRPs) (Zhang et al. 2001; Deschenes & Tomaselli, 2002; McCrossan & Abbott, 2004; Liu et al. 2006; Radicke et al. 2006; Abbott et al. 2007; Roepke et al. 2008). These are short integral membrane proteins with a single membrane spanning domain flanked by an extracellular N-terminus and a cytosolic C-terminus. MinK is known to associate with the Kv7.1 (KCNQ1) subunit to produce the channels mediating the cardiac current known as I_Ks (Barhanin et al. 1996; Sanguinetti et al. 1996); MiRP1 is believed to associate with Erg1 (Kv11.1 or HERG) to produce the channels that mediate the cardiac current known as I_Kr (Abbott et al. 1999) and MiRP2 with Kv3.4 to form a potassium channel that regulates the resting potential of skeletal muscle (Abbott et al. 2001). However, these proteins are ubiquitously expressed and reportedly have a promiscuous association with pore-forming subunits of different Kv subfamilies, including Kv4s, in heterologous expression systems (McCrossan & Abbott, 2004; Abbott et al. 2007). Although these additional associations need to be validated to conclude that they are physiologically significant, they are potentially interesting, at least in the cardiac field, because inherited mutations and polymorphisms in most KCNE genes have been associated with cardiac disease. MinK is not expressed in brain, but all MiRPs apparently are (McCrossan & Abbott, 2004), where their role still needs to be investigated.

Validating channel modulation by auxiliary proteins

Essentially two experimental approaches have been used to suggest that Kvβs, KChAP and KCNE proteins are auxiliary subunits of Kv4 channels, coexpression with Kv4 proteins in heterologous expression systems and coimmunoprecipitation from solubilized extracts of native tissue.

Changes in channel properties produced by the presence of the putative associated subunits suggests that the two types of protein interact, and raises the possibility that the same interaction occurs in native cells. However, in over-expressing heterologous cells, interactions could take place between proteins that never see each other in native cells or at similar concentrations. For example Kvβs are typically found in axons in many neurons (Rhodes et al. 1997; Monaghan et al. 2001), but Kv4 proteins are found in dendrites. Moreover, the putative auxiliary protein could have effects on channel function that are not the result of physical interaction with the pore forming subunits. When the putative auxiliary protein changes channel properties in such a way that the reconstituted channels better resemble native channels, the heterologous expression experiment becomes a good argument in support of a physiological association. Yet, even for a physiologically relevant interaction the reconstituted channels may not resemble native channels due to the existence of additional binding partners or other factors. However, when this is the case, obtaining independent evidence that the association takes place in native systems becomes particularly critical. This is the case with the proteins discussed in this section. They all enhance Kv4 current levels, albeit to a much smaller degree than KChIPs and DPPLs, but most tend to produce currents that increasingly depart from native currents in heart or brain, by slowing the kinetics of activation, inactivation and/or recovery from inactivation and by producing depolarizing shifts in voltage dependence.

The observations suggest that is easy to modify Kv4 channel expression in heterologous cells. This might be related to the fact, discussed earlier, that Kv4 proteins express poorly at the cell surface. A number of proteins, especially when they are over-expressed, may non-specifically affect Kv4 channel structure or the mechanisms that retain channels in intracellular compartments. Proteins with a single transmembrane sequence are of particular concern given the evidence that the membrane spanning domain of DPPLs interacts with the Kv4 voltage sensor. This indicates that the voltage sensor of Kv4 proteins has a sufficiently ‘flexible’ or ‘open’ structure that easily accommodates an extrinsic hydrophobic sequence. It is worth mentioning in this regard that Deschenes & Tomaselli (2002) found that the Na⁺ channel β1 subunit, a protein with a single membrane spanning domain as in KCNE and DPPL proteins, had effects on Kv4.3 channels in heterologous cells that resembled in magnitude the effects of many of the putative Kv4 channel auxiliary proteins discussed in this section.

Co-immunoprecipitation (co-IP) from native membranes can in principle provide a very powerful argument supporting association in vivo. However, the uncertainties regarding which of the associations discussed in this section are physiologically relevant suggest to us that, in the absence of independent evidence of the association, a more strict approach to co-IP is necessary.

Typically, investigators first immunoprecipitate tissue extracts solubilized with a non-denaturing detergent using an antibody to the pore-forming subunit (sometimes of dubious specificity; Rhodes & Trimmer 2006). A denaturing detergent is then added to the precipitate to separate the polypeptides present in the immunoprecipitated complex and these are separated by electrophoresis. Immunoblots of these polypeptides are then used to show that the immunoprecipitate not only contains the pore-forming subunit but also the putative auxiliary protein(s). Often no controls are performed, and rarely a quantification of the immunoprecipitated proteins relative to their abundance in the extract is presented.

However, proteins other than those truly associated with the pore forming subunits in the native membranes prior to solubilization may appear in the immunoprecipitate. Non-specific interactions may take place once the membranes are solubilized. Hydrophobic proteins might be particularly ‘sticky’. The auxiliary proteins may not even interact with the pore-forming subunits in the solubilized material, but may bind non-specifically to the reagents used for immunoprecipitation or to the beads typically used to bring down the immunoprecipitate. A small amount of the protein might just be ‘trapped’ in the immunoprecipitate. This is why quantification could be very useful. Controls for these and other possible artifacts of the procedure will go a long way to increase the value of coimmunoprecipitation assays.

Another protein that has been suggested to associate with Kv4 channels in cardiac and nervous tissue is the neuronal calcium sensor protein-1 (NCS-1 or frequenin), a member of the EF-hand family of Ca²⁺-sensing proteins that includes KChIPs. NCS-1 was reported to have effects on Kv4 channels in heterologous cells similar to (but apparently less effective than) those of KChIPs (Nakamura et al. 2001; Guo et al. 2002). The evidence that has been used to support NCS-1 association to Kv4 subunits is similar to that used initially to support KChIP association. However, this evidence has not been corroborated by many research groups or strengthened by independent evidence as in the case of KChIPs. Moreover, it is not clear whether NCS-1 associates with the channels in the presence of KChIPs, since both bind to the same site.

Interactions of Kv4 subunits with cytoskeletal proteins

K⁺ channel proteins also interact with cytoskeletal proteins. These associations are important in the localization of channels to specific neuronal compartments and in the formation of microdomains with signalling molecules. The C terminus of Kv4.2 was found in a yeast two-hybrid screen to interact with filamin, an actin-binding protein (Petrecca et al. 2000). Filamin was also shown to coimmunoprecipitate with Kv4.2 from brain extracts. To investigate possible functional consequences of this interaction, Kv4.2 was expressed in filamin-containing and filamin-lacking cells. Kv4.2 was found to colocalize with filamin at filopodial roots in filamin-containing cells, but had a uniform expression pattern in cells lacking filamin or in filamin-containing cells transfected with a Kv4.2 mutant lacking the filamin-binding region. Petrecca et al. (2000) proposed that filamin functions as a scaffold protein in the postsynaptic density, mediating a direct link between Kv4 proteins and the actin cytoskeleton. While this hypothesis still needs to be confirmed it is of interest in that it may contribute to the highly specific and dynamic localization of Kv4 channels within dendritic spines (Burkhalter et al. 2006; Kim et al. 2007).

Kv4.2 was also found to interact with PSD-95 via Kv4.2's C-terminal putative PDZ-binding domain (VSAL) (Wong et al. 2002). PSD-95, another scaffolding protein present in the postsynaptic density, is involved in the localization of other K⁺ channels and glutamate receptors to synapses.

Summary and perspectives: the impact of DPPLs and KChIPs on Kv4 channel modulation

The bulk of the evidence strongly supports the view that Kv4 channels in many CNS neurons are ternary complexes that include, in addition to Kv4 pore-forming subunits, KChIP and DPPL auxiliary subunits. With four copies per channel of each protein, this is a large complex of ∼750 kDa, probably one of the largest K⁺ channel complexes known. Research can now move from validating the associations to investigating the molecular mechanisms by which these auxiliary subunits regulate channel function, which is necessary to understand how Kv4 channels work. Future research can now also address how the association of the channels with these proteins affects neuronal function and how mutations in the proteins produce disease. Given unexplored and intriguing structural features of these proteins, including Ca²⁺ binding motifs in KChIPs and a DPP-like extracellular domain in DPPLs, this research is likely to lead to the discovery of novel aspects of Kv4 channel function in neurons. Of interest for example are the contributions of these proteins to the localization of Kv4 channels in specific microdomains (Burkhalter et al. 2006), as well as their modulation by extracellular and intracellular signals and their activity-dependent dynamic expression at the plasma membrane (Kim et al. 2007).

The extracellular domain of DPPLs constitutes most of the mass of these proteins (Fig. 1). As in other DPPs the extracellular domain consists of a β propeller and an α/β hydrolase domain (Strop et al. 2004). β propellers provide excellent platforms for protein–protein interactions, and in DPPIV the extracellular domain binds to components of the extracellular matrix and mediates roles in cell adhesion, cellular trafficking and T cell activation (De Meester et al. 1999; Hildebrandt et al. 2000; Gorrell et al. 2001). DPPLs may confer targeting or cell adhesion properties to Kv4 channels through its homologous extracellular domain (Fig. 1). Protein–protein associations mediated by the DPPL extracellular domain could be important in determining the organization of Kv4 channels in the dendritic plasma membrane. Alternatively, cell–cell interactions or interactions with extracellular matrix components may modulate the function of I_SA channels.

The other putative associated proteins of Kv4 channels that were evaluated here are unlikely to be part of the channel complex throughout the CNS given that they produce changes in channel properties not seen in neurons. Furthermore, it is unlikely, if not physically impossible, that Kv4 channels are interacting with all these proteins simultaneously. However some of these may associate with channels in specific neurons or specific neuronal structures, where the channels could have biophysical properties that are different from those known so far. However, convincing evidence that this occurs still needs to be obtained. Furthermore, the available evidence does not exclude the possibility that the interactions of Kv4 channels with KChIPs and DPPLs are dynamic, or that in some neuronal populations or in specific neuronal compartments channels may exist associated with only one of these proteins.

Less is known about the composition of Kv4 channels in other tissues. In ventricular myocytes, where Kv4 channels mediate the current known as I_to (or sometimes fast I_to), the channels contain KChIPs, and specifically KChIP2. However, these cells seem to lack DPPLs (at least in rodents, see Radicke et al. 2005). Data suggesting that MiRPs are components of cardiac Kv4 channels were discussed; however, these don't seem to reconstitute the properties of native I_to.