KCNQ channels mediate I_Ks, a slow K⁺ current regulating excitability in the rat node of Ranvier

Abstract

Mutations that reduce the function of KCNQ2 channels cause neuronal hyperexcitability, manifested as epileptic seizures and myokymia. These channels are present in nodes of Ranvier in rat brain and nerve and have been proposed to mediate the slow nodal potassium current I_Ks. We have used immunocytochemistry, electrophysiology and pharmacology to test this hypothesis and to determine the contribution of KCNQ channels to nerve excitability in the rat. When myelinated nerve fibres of the sciatic nerve were examined by immunofluorescence microscopy using antibodies against KCNQ2 and KCNQ3, all nodes showed strong immunoreactivity for KCNQ2. The nodes of about half the small and intermediate sized fibres showed labelling for both KCNQ2 and KCNQ3, but nodes of large fibres were labelled by KCNQ2 antibodies only. In voltage-clamp experiments using large myelinated fibres, the selective KCNQ channel blockers XE991 (IC₅₀ = 2.2 μm) and linopirdine (IC₅₀ = 5.5 μm) completely inhibited I_Ks, as did TEA (IC₅₀ = 0.22 mm). The KCNQ channel opener retigabine (10 μm) shifted the activation curve to more negative membrane potentials by −24 mV, thereby increasing I_Ks. In isotonic KCl 50% of I_Ks was activated at −62 mV. The activation curve shifted to more positive potentials as [K⁺]_o was reduced, so that the pharmacological and biophysical properties of I_Ks were consistent with those of heterologously expressed homomeric KCNQ2 channels. The ability of XE991 to selectively block I_Ks was further exploited to study I_Ks function in vivo. In anaesthetized rats, the excitability of tail motor axons was indicated by the stimulus current required to elicit a 40% of maximal compound muscle action potential. XE991 (2.5 mg kg⁻¹ i.p.) eliminated all nerve excitability functions previously attributed to I_Ks: accommodation to 100 ms subthreshold depolarizing currents, the post-depolarization undershoot in excitability, and the late subexcitability after a single impulse or short trains of impulses. Due to reduced spike-frequency adaptation after XE991 treatment, 100 ms suprathreshold current injections generated long trains of action potentials. We conclude that the nodal I_Ks current is mediated by KCNQ channels, which in large fibres of rat sciatic nerve appear to be KCNQ2 homomers.

In the mammalian node of Ranvier, Na⁺ channels (Na_v1.6) are located at high density in the tiny stretch of the axolemma not covered by myelin, whereas the delayed rectifying K⁺ channels (Kv1.1 and Kv1.2) are restricted to the juxta-paranodes (Poliak & Peles, 2003; Scherer et al. 2004). In addition to the Na⁺ and fast K⁺ current, a more slowly activating and deactivating K⁺ current (I_Ks) has been observed in nodes of Ranvier. This current was first described in frog myelinated nerve fibres (Dubois, 1981). Indications for the existence of I_Ks in mammalian nerve fibres and its function were later found by recording electrotonus in rat axons (Baker et al. 1987) and threshold electrotonus (the threshold analog of electrotonus) in humans (Bostock & Baker, 1988; Baker & Bostock, 1989). Subsequently, voltage-clamp experiments in single myelinated rat nerve fibres performed in isotonic KCl solution showed: that I_Ks, as in the frog node of Ranvier, starts to activate at membrane potentials as negative as −110 mV; that about 30% of I_Ks is already activated at the resting potential; that it does not inactivate; and that it is almost totally blocked by 10 mm TEA (Röper & Schwarz, 1989; Safronov et al. 1993). The slow activation and deactivation kinetics of I_Ks as well as the absence of inactivation were reminiscent of the M-current (Brown, 1988; Jentsch, 2000; Delmas & Brown, 2005), which was later shown to be mediated by heteromeric KCNQ2/KCNQ3 channels (Wang et al. 1998). Selective blockers of the M-current, like linopirdine and XE991 (Wang et al. 1998), and the M-current activator retigabine (Rundfeldt & Netzer, 2000; Wickenden et al. 2000; Tatulian et al. 2001), have greatly facilitated the analysis of M-currents.

Using immunohistochemistry, Devaux et al. (2004) recently demonstrated the existence of KCNQ2 and, in some cases, KCNQ3 subunit proteins at nodes of Ranvier in the rat sciatic nerve, spinal cord, and brain. In sciatic nerve fibres, KCNQ2 co-localized precisely with nodal Na⁺ channels in the narrow unmyelinated part of the node of Ranvier. KCNQ3 proteins were also detected in these peripheral myelinated nerve fibres, but were not seen co-localized at the nodal membrane with KCNQ2 and Na⁺ channels (Devaux et al. 2004). Instead, KCNQ3 antibodies labelled Schmidt-Lantermann incisures and outer mesaxons, portions of Schwann cell non-compact myelin involved in membrane–membrane contact.

The present experiments were performed to answer the question whether I_Ks may be mediated by KCNQ channels and which function I_Ks may have in peripheral nerve fibres. We show that in addition to the KCNQ2 channel protein, which is present in all nodes of Ranvier, the KCNQ3 channel protein can also be present in nodal membranes. However, the distribution and intensity of labelling for KCNQ3 channel proteins was dependent on fibre size; the occurrence of positive KCNQ3 staining drastically decreased in nerve fibres with large diameters. Voltage-clamp experiments performed in large single myelinated nerve fibres showed that I_Ks could be blocked totally by XE991 or linopirdine, whereas it was activated by retigabine. Inhibition of I_Ks by XE991 induced hyperexcitability in the isolated node of Ranvier. Using the threshold tracking method, we demonstrate a selective blockage of that component of threshold electrotonus which has been attributed to activation of I_Ks. In addition, the component induced by I_Ks in the recovery cycle (Bostock et al. 1998; Burke et al. 2001) was blocked. Taken together, our experiments strongly suggest that I_Ks is mediated by KCNQ channels and they demonstrate that selective blockage of I_Ks by the KCNQ channel blocker XE991 induces a drastic increase in the ability of motor axons to fire repetitively.

Methods

Immunocytochemistry and microscopy

Rabbit anti-KCNQ2-N (anti-KCNQ2) antibodies were raised and affinity-purified as previously described (Cooper et al. 2001). These antibodies are directed against residues 13–37 from the intracellular N-terminal region of KCNQ2, which appear to be absolutely conserved in mammals (Singh et al. 1998; Pan et al. 2001). Guinea pig anti-KCNQ3b-N (anti-KCNQ3) antibodies (Devaux et al. 2004) are directed against residues 36–57 from the N-terminal region of splice isoforms of KCNQ3 using a longer first exon (Schroeder et al. 1998); this antibody does not recognize a splice isoform (KCNQ3a) that includes the shorter first exon (Wang et al. 1998). The Na⁺ channel antibody used was PanNa_v (clone K58/35, Sigma), a mouse monoclonal antibody against a peptide conserved in all voltage-gated Na⁺ channel isoforms (Rasband et al. 1998). Secondary antibodies were purchased from Jackson Immunoresearch.

Adult male Wistar rats were killed by CO₂ inhalation and cervical dislocation. This method of killing was approved by the Institutional Animal Care and Use Committee of the University of Pennsylvania. Sciatic nerves were dissected and placed in cold phosphate-buffered saline solution. Fibres were teased out using fine needles, transferred to Superfrost Plus slides (Fisher), and allowed to air dry. Slides were stored at −20°C overnight, or until used for staining with antibodies.

Antibody immunoreactions were performed essentially as previously described (Devaux et al. 2004). Briefly, unfixed nerves were permeabilized and extracted, and non-specific binding sites blocked by incubation with Tris-buffered saline solution containing 0.5% Triton X-100 and 5% fish skin gelatine for 1 h. Nerves were then incubated with primary antibody in blocking buffer (with 0.2% Triton X-100) for 15–18 h, washed, incubated with secondary antibodies for 2 h, washed, and coverslipped using ProLong antifade reagent (Molecular Probes). In all cases, multilabel experiments were performed in parallel with single label and secondary-only controls, which revealed no evidence of antibody crossreactivity.

Fluorescence microscopy was performed using a Nikon TE2000 inverted microscope equipped with Chroma (31000, 41004 and 41017) filter sets and a 60 × 1.4 NA oil immersion objective. Monochrome images were acquired using a SPOT KE Slider cooled digital camera (Diagnostic Instruments). Colour superimpositions were performed and minimum and maximum intensity levels were adjusted using Photoshop (Adobe). Superimposition of fluorescence images and differential interference contrast (DIC) images were performed with the Photoshop add image function. Fibre diameters were determined using the measurement tool in SPOT image analysis software (Diagnostic Instruments), after calibration using a stage micrometer.

To quantify the percentage of nodes with and without associated KCNQ3 immunoreactivity, teased fibres were prepared from nerves of seven rats, and double-stained using mouse anti-Na_v antibodies/FITC-conjugated secondary antibodies, and guinea pig anti-KCNQ3 antibodies/Cy3 conjugated secondary antibodies. For ≥ 100 nodes per animal, the following procedure was performed: a node was selected at random on the basis of strong Na_v (i.e. FITC) staining, and photographed. Then the filters of the microscope were changed and an image acquired of KCNQ3 (i.e. Cy3) stain. The images were superimposed in Adobe Photoshop, and the nodal KCNQ3 staining was classified as moderate or strong (e.g. Figs 2I and L, and 3D), weak/diffuse/glial (e.g. Figs 2O and 3I), or absent (e.g. Fig. 2C and F).

Figure 2. KCNQ2 is strongly localized to nodes of rat sciatic nerve fibres, but KCNQ3 exhibits a heterogeneous distribution.

A–F, fibres show strong KCNQ2 or Na⁺ channel labelling at nodes, but absence of nodal KCNQ3 immunoreactivity. Fibres were labelled with rabbit anti-KCNQ2-N or mouse panNa_V, and guinea pig anti-KCNQ3b-N antibodies, and primary antibodies were detected using FITC-donkey anti-rabbit or anti-mouse and Cy3 donkey anti-guinea pig secondary antibodies. A, lower power image showing KCNQ3 (red) labelling of myelin and KCNQ2 (green) labelling of node. B and C, nodal region at higher power, with strong KCNQ2 staining of the node, but no co-labelling for KCNQ3. D–F, similar images as A–C, but double labelled for Na⁺ channels and KCNQ3. G–L, representative images, acquired as described for A–F, illustrating nodes apparently positive for KCNQ3b-N. In some examples (i.e. panel I), the KCNQ3b-N immunoreactivity is not only highly concentrated at the node, but exhibits a parallel line distribution suggestive of membrane labelling. M–O, some fibres show a combination of strong myelin labelling and weak nodal labelling for KCNQ3. M, two fibres conspicuously labelled by KCNQ3 antibodies at outer mesaxons (arrowheads) and incisures (arrows), as described by Devaux et al. (2004). A node is strongly labelled by panNa_V antibodies only (*). N and O, higher power images showing that strong panNa_V staining of the node is associated with strong punctate KCNQ3b-N staining around the node, and weaker but detectable KCNQ3b-N staining at the node itself. Scale bars, 10 μm (A, D, G, J and M) and 5 μm (B, C, E, F, H, I, J, K, L, N and O).

Figure 3. Larger diameter fibres are less likely to exhibit nodal KCNQ3 immunostaining.

A–E, a 3 μm diameter fibre exhibits strong nodal membrane co-labelling by panNa_V and KCNQ3b-N antibodies. A, lower magnification DIC image of several myelinated fibres, with superimposed panNa_V (green) and KCNQ3b-N (red) immunostaining at a node. B, DIC image of the nodal region from panel A, showing the bleb-like appearance of myelin membranes and demyelination near the node. C, immunofluorescence image of identical region to B, showing colabelling of nodal membrane by Na⁺ channel and KCNQ3b-N antibodies. D, E, pattern of KCNQ3b-N (D, red) and Na⁺ channel (E, green) immunoreactivity at the same node. F–J, a 10 μm fibre showing conspicuous labelling of glial membranes and puncta near the node, but weak nodal labelling. F, lower magnification image showing fibre with puncta of KCNQ3 staining along its length (white arrowheads), and near the node (yellow arrowhead), and strong Na channel labelling of the node. Inset shows KCNQ3 immunostaining along axon in grey scale; the arrowheads indicate an apparent labelled outer mesaxon. G, H, DIC and immunofluorescence images of nodal regions reveal strong nodal panNa_V stain and puncta of KCNQ3b-N staining near the node. Several KCNQ3-immunoreactive puncta appear to be localized at tips of glial blebs near the node (yellow arrowheads). I, J, individual colour channels showing pattern of KCNQ3 (I, red) and Na⁺ channel (J, green) immunoreactivity at node. Scale bars, 10 μm (A, F), 5 μm (B, C, G, H), and 2.5 μm (D, E, I, J).

Voltage- and current-clamp experiments

Male Wistar rats (250–350 g) were killed by exposure to CO₂ or by exposure to isoflurane with subsequent decapitation. These procedures were approved by the animal welfare committee of the University Hospital Hamburg-Eppendorf. Single myelinated nerve fibres were isolated from the sciatic nerve over a distance of two internodes by using fine needles, forceps (Dumont no. 5) and iridectomy scissors. The isolated nerve fibre was placed on top of the separating walls of the nerve chamber (see Fig. 1A and B) and fixed to the ridges with thin Vaseline seals. In addition to the electrical resistance provided by the Vaseline seals, the external resistance between pool B (connected to ground potential) and pool C was further increased by introducing an air gap (pool F in Fig. 1B). The side pools of the experimental chamber (Nonner, 1969) were filled with axoplasmic solution (see below) and both internodes were cut. A manually operated pulse generator and an oscilloscope were used for eliciting and monitoring of action potentials and membrane currents. After obtaining stable conditions, membrane currents and potentials were recorded by using the PULSE stimulation and data acquisition software (HEKA, Lambrecht) in conjunction with an EPC-10 patch-clamp amplifier. The measurements were performed at room temperature (20–23°C).

Figure 1. Nerve chamber and schematic of voltage clamping the node of Ranvier.

A, nerve chamber for recording of action potentials and membrane currents of an isolated node of Ranvier. The nerve chamber is made of Perspex and consists of four compartments separated by thin walls. B, central part of the nerve chamber at higher magnification. A single nerve fibre is shown placed on top of the compartment ridges with the node of Ranvier located in the middle of compartment A. C, diagram demonstrating the principle of the voltage clamp (Nonner, 1969). E, A, B and C denote electrodes (designated corresponding to the compartments shown in panel A) in the fluid surrounding the nerve fibre. D, point in the axoplasm at the node. Vaseline seals (v) were placed on top of the ridges to increase the electrical resistance between the chamber compartments and to hold the fibre in position. F (in B) denotes an air gap to further increase the external resistance of the nerve fibre. After exchanging the solution in the side pools C and E for axoplasmic solution, the internodes of the nerve fibre under investigation were cut in both side pools. Therefore the electrodes in C and E were in direct contact with the axoplasm of the nerve fibre. PG, pulse generator. Input of the feedback amplifier connected with electrode C, its output connected with electrode E. For recording action potentials (current-clamp mode), the output of the amplifier was connected with electrode A and the pulse generator with electrode E. Modified from Stämpfli & Hille (1976).

A schematic drawing of the method of Nonner (1969) of voltage clamping the node of Ranvier is shown in Fig. 1C. After cutting the internodes in both side pools the axoplasm of both internodes was in direct contact with the axoplasmic solution and the electrodes C and E. The feedback amplifier kept the potential at point C at ground potential by adjusting the potential at E. Any internodal current between C and D would also flow across the seal resistance between C and B and would produce a potential drop. The feedback amplifier prevented such a potential drop and thus eliminated flow of current between D and C. These two points were effectively kept at ground potential. The potential at A with respect to ground corresponded to the absolute membrane potential of the node under investigation. It could be changed by means of a pulse generator. By passing current through the resistance between E and D, the external potential in A was forced to follow the command waveform generated by the pulse generator. The current flowing between E and D was identical to the membrane current of the node under investigation and was recorded as the voltage drop across the resistance of the internode ED assuming an internodal resistance of 15 MΩ (for details see Nonner, 1969; Stämpfli & Hille, 1976; Neumcke et al. 1987; Röper & Schwarz, 1989). Membrane currents were low-pass filtered at 3 kHz. Capacity currents and leakage currents were not subtracted. At the beginning of each experiment, the holding potential (E_H) was adjusted so that about 30% of the Na⁺ current was inactivated (h_∞ = 0.7). In most experiments the node was superfused with isotonic KCl, so that E_H was determined from the potential step required to reach the reversal potential for K⁺, assuming E_K = 0 mV. The value of E_H measured in this way was −75.1 ± 1.3 mV (mean ± s.e.m.; n = 18). A similar value for the resting potential was previously reported for rat nerve fibres: −80 mV (Brismar, 1980), −78 mV (Neumcke & Stämpfli, 1982) and −77 mV (Röper & Schwarz, 1989). A value of −75 mV was used to determine the absolute membrane potential in those fibres in which E_K was unknown.

The following solutions (mm) were used: Ringer solution: NaCl 154, KCl 5.6, CaCl₂ 2.2; isotonic KCl: KCl 160, CaCl₂ 2.2; axoplasmic solution: KCl 160. The pH was adjusted to 7.4 with Tris-HCl. All chemicals were purchased from Sigma, and XE991 from Tocris, Ellisville. Additional batches of XE991 as well as retigabine were a kind gift of Dr D. Isbrandt (ZMNH, University of Hamburg).

In vivo experiments

Male Wistar rats (250–300 g) were anaesthetized with ketamine (30 mg kg⁻¹i.p.) and xylazine (10 mg kg⁻¹ i.p.). In addition, diazepam (2 mg kg⁻¹ i.p.) was administered to prevent convulsions occurring after administration of XE991 (1–5 mg kg⁻¹ i.p.). The animal welfare committee of the University of Tokushima agreed to this procedure. The anaesthetized rats were placed on a warming mat to maintain a temperature between 32 and 35°C. Ag–AgCl disk electrodes (15 mm in diameter) were used for stimulation. The cathode was attached to the lateral aspect of the tail at 15 mm from the base. The anode was placed on the skin of the hip where the hair was removed. Stainless needle electrodes were used for recording compound muscle action potentials and for ground. The recording electrode was inserted into the ipsilateral aspect of the tail muscle 60 mm distal to the stimulating electrode and the reference electrode was placed 20 mm distal to the recording electrode. A ground electrode was placed in between the stimulating and recording electrodes, 20 mm proximal to the recording electrode (for details see Yang et al. 2000). Stimulation was controlled by a PC running Qtrac software (Institute of Neurology, London), connected via a Micro 1401 data acquisition unit (Cambridge Electronic Design, Cambridge, UK) to the electromyogram preamplifier (Nihon Kohden MEG-1200) and stimulator (purpose-built ±50 mA bipolar voltage–current converter).

Using 1 ms rectangular stimuli and recording the negative peak of the compound muscle action potential, the excitability tests entailed first recording the stimulus–response curve, and then using the slope of the curve to track the stimulus required to evoke a 40% of maximal response, which was defined as the ‘threshold’ stimulus. To record threshold electrotonus, one Qtrac channel tracked the unconditioned threshold, while four other channels determined the threshold at discrete points during and after 100 ms polarizing currents, set to ±20% and ±40% of the unconditioned threshold. The five Qtrac channels were activated in turn, one every 0.8 s (see Kiernan et al. (2000) for a more complete description of the method). To record the recovery cycle, the unconditioned threshold was again tracked continuously on one Qtrac channel, while on another channel a supramaximal conditioning stimulus was given at 18 delays from 200 ms to 2 ms before the test stimulus. To separate the responses to conditioning and test stimuli when they overlapped, the response to the conditioning stimulus alone was recorded on a separate channel and subtracted from the response to the double stimulus. To track the recovery cycle following a short train of impulses, two more channels were used to record the effects of a train of seven supramaximal conditioning stimuli at 4 ms intervals, with the last conditioning stimulus delivered between 200 and 2 ms before the test stimulus. In some experiments, an additional Qtrac channel was used throughout the recording period to deliver a 100 ms current pulse equal in amplitude to the (1 ms) threshold current. For this stimulus, all the muscle action potentials above the noise level were recorded, to provide a measure of the tendency of the depolarized motor axons to fire repetitively.

Results

Immunocytochemistry

KCNQ2 and KCNQ3 subunit proteins were localized in teased fibres from sciatic nerve using polyclonal antibodies against the two channel proteins. Nodes were identified by co-labelling fibres with a mouse monoclonal antibody against voltage-gated Na⁺ channels, PanNa_v. Strong KCNQ2 and Na⁺ channel immunoreactivity were always co-localized in narrow ∼1 μm patches along the fibres, i.e. the nodes, and were absent from internodes and myelin (Fig. 2A and B). At some nodes, labelling for KCNQ2 and Na⁺ channels appeared as parallel lines, suggestive of membrane localization.

KCNQ3 immunoreactivity was much more widely distributed along the nerve fibres and, unlike immunoreactivity for KCNQ2 and Na⁺ channels, was also detected at outer mesaxons and incisures, which are locations representing sites of membrane–membrane adhesion and contact in the myelin sheath (as previously noted, Devaux et al. 2004). However, node-like clusters of KCNQ3 immunoreactivity were detected in teased nerve samples singly labelled with anti-KCNQ3 antibodies. To determine if these were nodes, we performed double label experiments using KCNQ3 antibodies and either KCNQ2 (Fig. 2A–C and G–I) or Na⁺ channel (Fig. 2D–F and J–O) antibodies. The results were strikingly variable. Some nodes exhibited a complete absence of KCNQ3 immunoreactivity at characteristic KCNQ2 or Na⁺ channel node-like clusters (Fig. 2A–F). Others exhibited strong KCNQ3 immunoreactivity that was well co-localized with the nodal markers (Fig. 2G–L). A third category of fibres exhibited an intermediate intensity of nodal staining, that was often accompanied by more intense labelling of outer mesaxons, incisures and punctate staining near, but not at the node itself (Fig. 2M–O). We quantified these results on teased fibres prepared from seven animals. For each set of fibres, KCNQ2 or Na⁺ channel node-like clusters were selected randomly and photographed using a high-power objective, and then the pattern of KCNQ3 nodal immunoreactivity was determined. KCNQ3 immunoreactivity was strong in 23 ± 14%, weak or most conspicuously glial in 27 ± 9%, and absent in 50 ± 16% of the nodes (n = 702, mean ± s.d.).

The fibre composition of rat sciatic nerve is heterogeneous, including about 8000 myelinated motor and sensory fibres that vary in size, pharmacological sensitivity, and such physiological properties as conduction velocity and frequency adaptation (Schmalbruch, 1986; Gokin et al. 2001; Vleggeert-Lankamp et al. 2004). To learn whether the variability in KCNQ3 staining pattern we observed might be correlated with functional differences among nerve types, we performed an additional analysis of KCNQ3 localization as a function of myelinated fibre diameter. We visualized teased, immunostained fibres with a combination of DIC and fluorescence optics (Fig. 3). DIC images revealed bleb-like distortion of the myelin structure in these unfixed detergent-treated fibres, most clearly near the nodes, which in many cases appeared partially demyelinated for a distance of about 5 μm (Fig. 3B and G). We measured the cross-sectional diameter of well-isolated single fibres in DIC images, and classified the pattern of KCNQ3 immunostaining as previously. Again, the fibres exhibited a mixture of absent (data not shown), strong and often nodal membrane associated (Fig. 3A–E), and weak/diffuse/glial (Fig. 3F–H) KCNQ3 patterns. However, smaller fibres (2–6 μm) were more likely to exhibit strong nodal KCNQ3 staining than medium fibres (7–9 μm), and this sampling revealed no large fibres of 10 μm or greater with strong KCNQ3 immunoreactivity at the node (Fig. 4).

Figure 4. KCNQ3 staining pattern depends on fibre diameter.

Bar graph showing the dependence of KCNQ3 staining pattern on fibre diameter. Smaller fibres had a higher likelihood of showing a strong KCNQ3 nodal staining; the largest fibres most often show glial staining, accompanied by weak, diffuse or absent KCNQ3 staining at the node.

Voltage- and current-clamp experiments

Membrane currents

Large single myelinated nerve fibres were isolated from the rat sciatic nerve, mounted in the experimental chamber, and action potentials and membrane currents were recorded using the experimental set-up described in Methods (Nonner, 1969; Stämpfli & Hille, 1976; Schwarz et al. 1995). The holding potential was adjusted to the membrane potential at which 30% of the Na⁺ channels were inactivated (h_∞ = 0.7), i.e. about −75 mV (see Methods). Since in Ringer solution the amplitude of I_Ks currents was very small and I_Ks had to be isolated from the other currents with XE991, it was not possible to measure accurately in Ringer solution properties of the current such as the time course of activation and deactivation. Therefore, as was previously done by Dubois (1981) and Röper & Schwarz (1989), most of the voltage-clamp measurements were performed in isotonic KCl solution to increase the I_Ks amplitude. Na⁺ currents were not detected under these recording conditions (external and axoplasmic solutions did not contain Na⁺), therefore tetrodotoxin was not added to the test solutions. Upon exchanging Ringer to isotonic KCl solution a steady inward current appeared at the holding potential of −70 mV (Fig. 5A). Upon a potential step to −130 mV a tail current was elicited which, after a short delay, slowly deactivated to reach a small constant inward current. Upon repolarization to the holding potential, the initial current amplitude was almost zero; thereafter the inward current slowly increased to reach the level of the holding current which was the same as before applying the negative potential step. Figure 5A shows superimposed membrane currents recorded with 80 ms potential steps varying between −130 and 50 mV. Upon application of 100 μm XE991, the holding current as well as a large part of the currents recorded upon the hyperpolarizing test pulses were blocked. The currents elicited upon depolarizing pulses were also reduced. The XE991-sensitive current (obtained by subtracting the current traces recorded in the presence of the inhibiting substance from the corresponding control current traces) clearly showed that XE991 totally blocked I_Ks. The membrane currents remaining unblocked by XE991 consisted of XE991-insensitive K⁺ currents and an unspecific leakage current. The leakage current in the node of Ranvier is that current component exhibiting a linear voltage dependence (see Fig. 2 in Röper & Schwarz, 1989; see also Stämpfli & Hille, 1976). The XE991-sensitive current (Fig. 5C) had the same properties as I_Ks described by Dubois (1981) in frog and by Röper & Schwarz (1989) in rat nerve fibres: it was already activated at the holding potential of −70 mV, upon negative and positive potential steps it slowly deactivated and activated in a potential-dependent way, and it did not inactivate. The voltage dependence of I_Ks activation was determined from the tail current amplitudes of the XE991-sensitive current recorded upon repolarization to the holding potential (Fig. 5C). The tail current amplitudes were normalized to the maximal tail current and plotted versus test pulse potential. Fifty per cent of I_Ks was activated at E_0.5 = −61.7 ± 1.4 mV (mean ± s.e.m.; n = 11), the slope factor being 12.3 ± 0.8 mV. This value was similar to the value of −60 mV measured by Röper & Schwarz (1989) in the rat and by Dubois (1981) in the frog, and similar to the values for E_0.5 obtained with the patch-clamp technique in rat (−76 mV, Safronov et al. 1993) and human nerve fibres (−68 mV, Reid et al. 1999). The latter experiments have also been done in isotonic KCl solution. As shown below, the voltage dependence of I_Ks activation was remarkably sensitive to the external K⁺ concentration ([K⁺]_o).

Figure 5. XE991-sensitive current measured in isotonic KCl.

A–C, membrane currents recorded in isotonic KCl. Holding potential, −70 mV. Hyperpolarizing and depolarizing potential steps (duration, 80 ms) were applied before (A, Control) and after application of XE991 (100 μm; B) in steps of 20 mV. Subtraction of the membrane current traces shown in B from those in A yielded the XE991-sensitive current (C). D, activation curve, determined from the normalized amplitude of the tail currents recorded upon repolarization to the holding potential plotted against the amplitude of the preceding test pulses. A Boltzmann equation was fitted to the data points yielding E_0.5 = −72 mV (slope factor k = 12.5 mV).

Pharmacology of I_Ks

It has been reported that 50% of the native M-current recorded in supracervical ganglion cells was blocked by 0.98 μm XE991 and 50% of the current mediated by homomeric KCNQ2 or heteromeric KCNQ2/3 channels after heterologous expression in CHO cells was blocked by 0.71 and 0.6 μm XE991, respectively (Wang et al. 1998). To determine the concentration-dependent inhibition by XE991 of the nodal I_Ks, the current recorded in the presence of 100 μm XE991 was assumed to represent 100%I_Ks inhibition. This current was subtracted from the current traces recorded at lower XE991 concentrations. The amplitudes of these tail currents were normalized and plotted against the XE991 concentration. A Hill function fitted to the data points yielded a 50% inhibition of I_Ks by 2.2 μm XE991, the Hill coefficient being 1.2 (Fig. 6A). Similar experiments were performed with linopirdine, another substance used to block KCNQ-mediated currents. Linopirdine at 100 μm totally blocked I_Ks. There was no additional block of the remaining membrane current by subsequent application of 100 μm XE991. Fifty per cent of I_Ks was blocked by 5.5 μm linopirdine, the Hill coefficient being 1.0 (Fig. 6B).

Figure 6. Dose–response curves of XE991, linopirdine and TEA.

A, dose–response curve of XE991. The inhibition of I_Ks in the presence of XE991 at concentrations between 1 and 100 μm was determined from the reduction of the XE991-sensitive tail currents recorded upon repolarization to a holding potential of −75 mV from a 400 ms test pulse to 35 mV. A Hill function was fitted to the data points yielding an IC₅₀ = 2.2 μm; Hill coefficient, 1.2; n = 5. B, dose–response curve of linopirdine. The inhibition of I_Ks in the presence of linopirdine at concentrations between 0.1 and 100 μm was determined from the reduction of the linopirdine-sensitive tail currents recorded upon repolarization to the holding potential of −75 mV from a 400 ms test pulse to 35 mV. A Hill function was fitted to the data points yielding an IC₅₀ = 5.5 μm; Hill coefficient, 1.0; n = 7. C, dose–response curve of TEA. The inhibition of the nodal K⁺ current was determined from the reduction of the tail currents recorded upon repolarization to the holding potential of −75 mV from a 400 ms test pulse to 35 mV. The effect of 10 mm TEA was assumed to be maximal. The membrane current recorded in the presence of 10 mm TEA was subtracted from the tail currents recorded at lower TEA concentrations. The difference currents were normalized to the control tail current amplitude recorded in the absence of TEA. The means ± s.e.m. were plotted against the TEA concentration. A Hill function was fitted to the data points yielding an IC₅₀ = 0.22 mm; Hill coefficient, 0.75; n = 5.

Previously, 10 mm TEA was used to block I_Ks in voltage-clamped nodes of Ranvier (Röper & Schwarz, 1989) or 1 mm TEA to block I_Ks in rat dorsal root fibres (Baker et al. 1987). Since it has been shown that the current mediated by homomeric KCNQ2 channels is much more sensitive to TEA (IC₅₀ = 0.16 mm; Wang et al. 1998) than that mediated by heteromeric KCNQ2/3 channels (IC₅₀ = 3.5 mm; Wang et al. 1998) we determined the dose–response curve of TEA. The reduction of the slow component of the tail current recorded upon repolarization to −70 mV after superfusion with increasing concentrations of TEA was determined, normalized to the maximal tail current amplitude and plotted versus the TEA concentration. The Hill function fitted to the data yielded an IC₅₀ of 0.22 mm TEA (Fig. 6C). This value is much closer to the TEA concentration needed to block homomeric KCNQ2 channels than to that needed to block heteromeric KCNQ2/3 channels (Wang et al. 1998). Since it is much more difficult to dissect rat nerve fibres of small diameter as compared with those of large diameter, only large diameter nerve fibres were investigated. The high sensitivity of these nodes to TEA nicely corresponds to the immunohistochemical finding that large fibres predominantly stain only for KCNQ2 channel proteins. These results together with the E_0.5 value of the activation curve suggested that in the rat node of Ranvier of large nerve fibres the channels mediating I_Ks may predominantly be formed by KCNQ2 channel subunits only.

Retigabine has been shown to increase the amplitude of the M-current as well as heterologously expressed KCNQ currents by shifting the activation curve to more negative membrane potentials (Rundfeldt & Netzer, 2000; Wickenden et al. 2000; Tatulian et al. 2001). We were therefore interested to investigate whether retigabine has similar effects on I_Ks. In the presence of 10 μm retigabine, the current at the holding potential as well as the amplitude of the inward tail currents recorded upon negative potential steps were increased (Fig. 7). In contrast, the amplitude of the outward currents remained almost unchanged. After recording the retigabine-induced effects, application of 100 μm XE991 totally blocked I_Ks. Before application of retigabine, 50% of the XE991-sensitive current was activated at −61.8 ± 1.0 mV (k = 12.4 ± 1.1 mV), whereas in the presence of retigabine, 50% of the XE991-sensitive current was already activated at −85.4 ± 2.8 mV (k = 17.3 ± 2.5 mV; mean ± s.e.m.; n = 4). This showed that retigabine shifted the activation curve to more negative membrane potentials by −23.6 mV and induced an increase in the slope factor. Retigabine also induced changes in the time course of deactivation and activation. In the presence of retigabine deactivation became slower, the time constant of deactivation of the current elicited with a negative potential step to −130 mV increased by a factor of 2.7 from τ = 30.7 ± 4.0 ms to τ = 83.4 ± 5.7 ms (mean ± s.e.m.; n = 6), at −110 mV the time constant of deactivation increased by a factor of 2.9 from 58.5 ± 6.8 ms (control) to 168.4 ± 26.5 ms. In contrast, the time course of activation was accelerated by retigabine; at the holding potential of −70 mV the time constant of current activation following a negative pulse to −130 mV decreased from τ = 24.9 ± 4.8 ms to τ = 4.7 ± 2.3 ms (mean ± s.e.m.; n = 6).

Figure 7. Retigabine activates I_Ks.

A and B, membrane currents recorded in isotonic KCl. Holding potential, −60 mV. Hyperpolarizing and depolarizing pulses were applied in steps of 20 mV in the absence (A, Control) and presence of retigabine (10 μm; B). C, membrane currents recorded after application of XE991 (100 μm). D, XE991-sensitive current obtained by subtracting the membrane current traces shown in C from those in A. E, XE991-sensitive currents in the presence of retigabine obtained by subtracting the membrane currents shown in C from those in B. F, activation curves, determined from the amplitudes of the tail currents recorded upon repolarization to the holding potential shown in D (Control) and E (XE991-sensitive current in the presence of retigabine). The tail current amplitudes were normalized and plotted against test potentials. Boltzmann functions were fitted to the data points yielding V_0.5 = −61 mV (slope factor = 15.8 mV) for the XE991-sensitive current (Control) and V_0.5 = −85.6 mV (slope factor = 15.5 mV) for the activation curve of the XE991-sensitive current in the presence of retigabine.

Voltage dependence of I_Ks activation depends on [K⁺]_o

The data described so far showed that 50% of I_Ks is activated at about −60 mV if the measurements were made in high [K⁺]_o. This voltage dependence of I_Ks is far more negative than that reported for currents mediated by heterologously expressed KCNQ channels. For KCNQ2 E_0.5 values of −14 and −18 mV have been reported if CHO cells were used as an expression system (Selyanko et al. 2000, 2001) or −37 mV if KCNQ2 channels were expressed in oocytes of Xenopus laevis (Biervert et al. 1998). Since no data are available about the effect of [K⁺]_o on the voltage dependence of I_Ks activation, we performed experiments in solutions containing different K⁺ concentrations (Fig. 8). From a holding potential of −70 mV test pulses of 150 ms duration were followed by a potential step to −100 mV to record tail currents. All experiments were performed in the presence of 1 mm 4-AP to block K⁺ currents different from I_Ks (Dubois, 1981). Figure 8 shows that with increasing [K⁺]_o the holding current increased as well as the amplitudes of the tail currents. The tail currents recorded in isotonic KCl were totally blocked by linopirdine (Fig. 8Af). This result allowed us to determine the activation curves from the tail current amplitudes. The main result of this and three other experiments was that upon increasing [K⁺]_o the activation curves shifted to more negative membrane potentials and their slope factors decreased. Mean values (± s.e.m.; n = 4) of E_0.5 and slope factors of the activation curves were: E_0.5 = −40.5 ± 2.8 mV, k = 21.7 ± 1.3 mV (10 mm KCl); E_0.5 = −57.0 ± 2.6 mV, k = 15.9 ± 2.0 mV (80 mm KCl); E_0.5 = −64.4 ± 2.4 mV, k = 15.8 ± 1.4 mV (160 mm KCl). E_0.5 shifted by about 25 mV upon a change from 10 mm to 160 mm[K⁺]_o. Figure 8Aa shows that the tail current amplitudes recorded in Ringer solution were small. In other fibres we found that these small tail currents were totally blocked by XE991 or linopirdine. A tentative evaluation of these tail currents recorded in Ringer solution showed that 50% of I_Ks was activated near −25 mV and that the slope of the activation curve was flatter than that in 10 mm KCl. The activation curve obtained from the data in Fig. 8Aa was included in the graph of Fig. 8B, but different symbols were used to indicate that these results are less certain, on account of the small amplitude of the currents. Nevertheless these data indicated that in Ringer solution part of I_Ks is already activated at the normal resting potential of these fibres and that activation of I_Ks and that of heterologously expressed KCNQ2 currents takes place within a similar potential range.

Figure 8. I_Ks activation curve is shifted to more negative membrane potentials upon increasing [K⁺]_o.

Aa, and c–e, membrane currents recorded in Ringer solution (5.6 mm), and in solutions with increasing [K⁺]_o (10, 80 and 160 mm K⁺). Af, family of membrane currents recorded in 160 mm KCl containing 100 μm linopirdine. Ab, pulse protocol. Holding potential, −70 mV. Na⁺ inward currents were cut off. B, activation curves. Normalized tail current amplitudes of the currents shown in panels Aa–e were plotted against test pulse potentials. Boltzmann functions were fitted to the data points. Values for 50% activation (E_0.5) and the slope factor k: Ringer (^; E_0.5 = −22.8 mV; k = 30.2 mV); 10 mm KCl (▴; E_0.5 = −39.1 mV; k = 18.1 mV); 80 mm KCl (▪; E_0.5 = −51.4 mV; k = 14.7 mV); isotonic KCL (•; E_0.5 = −61.9 mV; k = 14.0 mV).

Action potentials

Since the time course of I_Ks activation is slow, it barely influences the shape of an action potential (Reid et al. 1993). In contrast, it has been assumed that I_Ks plays an important role in accommodation to subthreshold depolarizing current, in spike-frequency adaptation and in stabilizing the resting membrane potential (Dubois, 1981; Baker et al. 1987; Reid et al. 1993; Schwarz et al. 1995). In the past, these presumed functions of I_Ks in the node of Ranvier could not be tested directly, because a selective inhibitor of I_Ks was not available. Since the voltage-clamp experiments described above show that XE991 selectively blocks I_Ks, we used this substance to investigate the function of I_Ks in current-clamp experiments. Figure 9A shows that 100 μm XE991 neither influenced the amplitude nor the duration of action potentials elicited with short depolarizing currents. In five fibres the action potential amplitude was 109.6 ± 2.5 mV (mean ± s.e.m.) and its duration near the threshold potential was 1.25 ± 0.07 ms in Ringer solution. In the presence of 100 μm XE991 these values did not change significantly; the action potential amplitude was 110.0 ± 2.3 mV and its duration near the threshold potential 1.28 ± 0.10 ms. Nevertheless, application of XE991 induced a small depolarization of the resting membrane potential (4.2 ± 2.0 mV) and slightly decreased the threshold potential for eliciting an action potential, from 22.6 ± 0.9 to 19.0 ± 1.8 mV. Figure 8B shows current-clamp recordings in a nerve fibre which responded with repetitive activity to a long lasting depolarizing current stimulus in normal Ringer solution. In the control (Fig. 8Ba), the amplitude of the stimulating current was adjusted to elicit just one action potential. In the presence of XE991, the same depolarizing current amplitude elicited three successive action potentials (Fig. 9Bb). A similar observation was made in one other nerve fibre which exhibited repetitive activity before application of the drugs. The other three fibres investigated did not exhibit repetitive activity. In the rat, motor as well as sensory nerve fibres are able to fire repetitively, therefore this property cannot be used to distinguish between the two fibre types (reviewed by Vogel & Schwarz, 1995).

Figure 9. XE991 reduces threshold potential.

A, local responses and action potentials elicited with 0.1 ms depolarizing sub- and suprathreshold currents, respectively, in the absence (Aa, control) and presence of XE991 (100 μm; Ab), recorded from a node of Ranvier of a single myelinated nerve fibre. B, action potentials recorded with 45 ms depolarizing pulses in normal Ringer solution (Ba), and in Ringer solution containing 100 μm XE991 (Bb). All nerve fibre responses were elicited with the same depolarizing current amplitude from a single myelinated nerve fibre.

The effects of XE991 on threshold potential and frequency adaptation in the isolated rat nerve fibre were rather small compared with the drastic effects of XE991 in the in vivo experiments described below. This is most likely caused by the unspecific damage of the nerve fibres during the dissection procedure. We therefore studied the function of I_Ks on the excitability properties of peripheral motor nerve fibres in experiments performed in vivo in which the damage of nerve fibres due to mechanical manipulation was avoided.

In vivo experiments

Threshold electrotonus

Bostock & Baker (1988) showed that it is possible to detect the effects of I_Ks on the response of axons to subthreshold polarizing currents (i.e. electrotonus) by threshold tracking (i.e. threshold electrotonus) and that this method can be applied in vivo. The accommodative sag in the response to depolarizing currents was prevented in vitro by TEA and attributed to increased activation of I_Ks (Bostock & Baker, 1988). Figure 10 illustrates the effects of XE991 on this manifestation of I_Ksin vivo. Figure 10A shows a family of threshold electrotonus curves for motor axons in the tail of a rat, for 100 ms currents set to ±20 and ±40% of resting threshold. There is a rapid phase of threshold change at current onset and offset, due to the change in nodal membrane potential, followed by slower changes due to the change in potential of the internodal axolemma (most pronounced with the 40% hyperpolarizing current) and the activation or deactivation of slow K⁺ currents (most pronounced with the 40% depolarizing current). These recordings resemble those in humans (e.g. Kiernan et al. 2000), although the accommodative sag in the responses to the depolarizing currents is less pronounced in the rat. After injection of XE991 (2.5 mg kg⁻¹ i.p.) the accommodative sag to depolarizing currents was abolished altogether (Fig. 10B). This effect was highly repeatable, and Fig. 10C shows the mean ± s.e.m. of the waveforms obtained from five such experiments (40% depolarizing current only). XE991 abolished both the slow accommodation to the depolarization and the post-depolarization undershoot in threshold reduction, which has also been used as a measure of I_Ks (e.g. Kiernan et al. 2001). Since I_Ks is already partly activated at the resting potential, there is also a small effect on the responses to hyperpolarizing currents, and abolition of the post-hyperpolarization (or ‘anode break’) increase in excitability (Fig. 10A and B).

Figure 10. Effects of XE991 on threshold electrotonus.

A, threshold changes produced in motor axons innervating the tail muscles of an anaesthetized rat by 100 ms rectangular current pulses set to ±20 and ±40% of control threshold. Threshold reduction, or increase in excitability, plotted upwards. B, threshold electrotonus recorded as in A after injection of XE991 (2.5 mg kg⁻¹ i.p.). C, mean ± s.e.m. of 5 such recordings (+40% of control threshold) from different rats, showing excellent repeatability of these excitability changes (thick continuous line, control, thin line, post-XE991).

Recovery cycles for one and seven conditioning stimuli

The filled circles in Fig. 11Aa show the recovery cycle for motor axons in the tail of a rat after a single supramaximal stimulus. The excitability changes qualitatively resemble those in human median nerve axons (Kiernan et al. 2000), with successive phases of relative refractoriness (up to 3 ms), superexcitability (between 3 and about 30 ms) and late subexcitability (between 30 and 200 ms). However, the late subexcitability, attributed to I_Ks, is considerably smaller in the rat. To amplify this manifestation of I_Ks, we substituted a train of seven conditioning stimuli at 250 Hz (open circles in Fig. 11Aa). Bergmans (1970) used a similar protocol to enhance the subexcitability, which he called H₁ (i.e. first hyperpolarizing afterpotential), to distinguish it from H₂, the second hyperpolarizing afterpotential due to activation of the Na⁺ pump by long trains of impulses. The change in the recovery cycle with the number of conditioning impulses in Fig. 11Aa is qualitatively similar to that seen in human median motor axons (Fig. 22 in Bergmans, 1970). After injection of XE991, the late subexcitability, and the difference between the effects of one and seven conditioning stimuli largely disappeared, indicating that KCNQ channels are also responsible for these manifestations of I_Ks (Fig. 11Ab and c).

Figure 11. Effects of XE991 on recovery cycles and spike-frequency adaptation.

A, recovery cycles recorded with 1 (•) and 7 (^) supramaximal conditioning stimuli in rat tail motor axons. Multiple conditioning stimuli at short (4 ms) intervals enhance the long-lasting subexcitability, due primarily to increased activation of I_Ks. The letters a, b and c refer to recordings made at the times shown in D, i.e. control, post-1.25 mg kg⁻¹ XE991 and post-2.5 mg kg⁻¹ XE991. B, electromyograms recorded from rat tail muscles to 100 ms suprathreshold rectangular current pulses at the times indicated in D. C, raster plot of latencies of all peaks in the electromyograms for responses to the 100 ms current pulses as in B plotted against elapsed time (axis as in D). D, numbers of peaks in the responses to 100 ms current pulses as in B plotted against elapsed time (axis as in D). D, numbers of peaks in the responses to 100 ms current pulese in B and C. Letters indicate times of recordings in A and B. Arrows indicate times of i.p. drug injections: Diaz, 2 mg kg⁻¹ diazepam; XE991, 1.25 mg kg⁻¹ XE991.

Spike frequency adaptation

In response to a prolonged suprathreshold stimulus, axons may produce a train of impulses, but there is a progressive reduction in impulse frequency, usually leading to cessation of repetitive activity. This spike-frequency adaptation is particularly pronounced in motor fibres, which normally produce two or at most three impulses, however strong the depolarizing current (Bostock, 1995). In rat motor fibres the adaptation is reduced by TEA, and has been related to the subthreshold accommodation and also attributed to activation of I_Ks, which increases in a stepwise fashion with successive nerve impulses (Baker et al. 1987; Schwarz et al. 1995). We tested for a contribution of KCNQ channels to spike-frequency adaptation in vivo, by alternating the tests of excitability with a suprathreshold 100 ms stimulus, equal in amplitude to the 1 ms test stimulus that produced a compound muscle action potential (CMAP) that was 40% of maximal. Figure 11B–D illustrates the result of one such experiment. Initially spike-frequency adaptation was pronounced, and there were usually only two peaks in the electromyogram (Fig. 11Ba), indicating that a significant proportion of fibres were firing twice, but triple discharges were rare. Injection of diazepam (2 mg kg⁻¹) had no discernible effect (Fig. 11C and Da). However, XE991 (1.25 mg kg⁻¹) produced a slow build-up of repetitive activity, which appeared to reach a plateau after about 45 min (Fig. 11Bb and Db), but after a second injection of XE991 (1.25 mg kg⁻¹) increased still further. Similar inhibition of spike-frequency adaptation was seen in two further experiments with long suprathreshold current pulses. It is clear that KCNQ channels make a major contribution to spike-frequency adaptation in these rat motor axons.

Discussion

Our immunostaining results show that KCNQ2 channel proteins are present in all nodes of Ranvier in peripheral myelinated nerve fibres, whereas strong labelling for KCNQ3 channel proteins was likely to be found in nodes of small and medium sized nerve fibres, but was absent in large myelinated nerve fibres. The slowly activating K⁺ current (I_Ks) recorded in large myelinated nerve fibres was totally inhibited by the selective KCNQ channel blockers XE991 and linopirdine, and it was activated by retigabine, a selective KCNQ channel opener. Voltage-dependent activation of I_Ks was shown to be dependent on extracellular [K⁺]. In vivo, XE991 induced a drastic reduction of both subthreshold accommodation (as measured by threshold electrotonus) and spike-frequency adaptation to suprathreshold depolarizing currents. Taken together, these immunocytochemical, biophysical and pharmacological data strongly indicate that I_Ks is mediated by KCNQ channels, and provide new evidence that I_Ks serves to stabilize the nodal membrane potential. Furthermore, I_Ks in the large rat sciatic nerve fibres studied here appears to be mediated by homomeric KCNQ2 channels.

Indications that I_Ks is mediated by KCNQ channels

The pharmacological experiments we performed provide compelling evidence that I_Ks is mediated by KCNQ channels. Linopirdine and XE991 are selective KCNQ channel blockers (Lamas et al. 1997; Wang et al. 1998) that inhibit I_Ks, I_M and heterologously expressed KCNQ currents with similar potency. Thus, XE991 blocks I_Ks with an IC₅₀ (2.2 μm) only slightly higher than that reported by Wang et al. (1998) for blocking currents mediated by heterologously expressed KCNQ2 (0.71 μm), KCNQ2/3 (0.6 μm) and M-channels in sympathetic neurons (0.98 μm). Linopirdine blocks I_Ks with an IC₅₀ (5.5 μm) very similar to the values reported for blockage of KCNQ2 (4.8 μm), KCNQ3 (4.8 μm), KCNQ2/3 (4.0 μm) and M-channels (7.0 μm; Hadley et al. 2000; Wang et al. 1998).

Although KCNQ2 and KCNQ3 are similar in their sensitivity to XE991 and linopirdine, they have a very different sensitivity to TEA. The higher sensitivity of KCNQ2 to TEA results from the presence of a single pore tyrosine residue located just downstream of the canonical GYG sequence (Hadley et al. 2000). In Xenopus laevis oocytes, KCNQ2 homomers and KCNQ2/3 heteromers exhibit TEA IC₅₀ values of 0.16 and 3.5 mm, respectively (Wang et al. 1998). Values obtained in CHO cells are similar (KCNQ2: IC₅₀ = 0.3 mm; KCNQ2/3: 3.8 mm; Hadley et al. 2000). KCNQ3 homomeric channels are very insensitive to TEA, with even high TEA concentrations (> 30 mm) showing only minimal effects (Hadley et al. 2000). The IC₅₀ for TEA block of I_Ks (0.22 mm) found in the present experiments is nearly the same as for homomeric KCNQ2 channels; this IC₅₀ is at least 15-fold less for KCNQ2/3 channels or native M-currents (Wang et al. 1998; Hadley et al. 2000). Data for TEA inhibition of I_Ks was well fitted assuming a single homogeneous population of binding sites. These results are unlike those obtained in neurons from dorsal root ganglion (Passmore et al. 2003) and neurons of the superior sympathetic ganglion of immature animals (Hadley et al. 2003), where data suggest that a mixture of KCNQ2 and KCNQ2/3 channels are expressed.

I_Ks is also activated by the selective KCNQ channel opener retigabine (Rundfeldt, 1997; Rundfeldt & Netzer, 2000; Wickenden et al. 2000; Tatulian et al. 2001). Retigabine (10 μm) shifts the I_Ks activation curve to more negative membrane potentials by −24 mV. This shift is the same as the shift in KCNQ2 currents (−24 mV), but less than shifts measured for KCNQ3 (−43 mV) and KCNQ2/3 (−30 mV) currents (Tatulian et al. 2001). In addition, retigabine accelerates activation and slows deactivation kinetics of I_Ks, changes which have also been reported in KCNQ-mediated currents (Wickenden et al. 2000).

In addition to these pharmacological properties, we discovered a novel and, potentially, physiologically significant shared property of KCNQ2 channels and I_Ks, the sensitivity of voltage-dependent activation to extracellular [K⁺]. Owing to the small size of I_Ks and its gating in the voltage range near physiological E_K, previous recordings at both macroscopic and single channel levels have been performed in the presence of elevated extracellular [K⁺]. The new availability of the selective I_Ks blocking substances, XE991 and linopirdine, prompted us to study I_Ks also at lower concentrations of extracellular [K⁺]. We found that the voltage dependence of I_Ks activation is very sensitive to [K⁺]_o. In addition to a shift to more positive potentials, the activation curves become flatter upon decreasing [K⁺]_o, suggesting that part of I_Ks is already activated at the normal resting potential, despite the shift to more positive membrane potentials. We are not aware of any previous studies investigating the dependence of the biophysical properties of KCNQ channels or of the M-current on [K⁺]_o. The cellular or molecular mechanisms underlying the large shift of the activation curve to more negative membrane potentials by 30–40 mV upon changing the external K⁺ concentration from 5.6 mm in Ringer to isotonic KCl solution are unknown. Previous voltage-clamp measurements of the nodal I_Ks in frog (Dubois, 1981) and rat nerve fibres (Röper & Schwarz, 1989) were performed in isotonic KCl and yielded activation curves with 50% activation at −60 mV. A similar voltage dependence of the activation curve of nodal S channels presumably mediating I_Ks was derived from single channel recordings in isotonic KCl solution by using the patch-clamp method in demyelinated nerve fibres. These studies show that 50% of the S channels are activated at −76 mV in rat (Safronov et al. 1993) and at −68 mV in human nerve fibres (Reid et al. 1999). Preliminary experiments in CHO cells expressing KCNQ2 channels have shown that the voltage dependence of activation of the KCNQ2-mediated current was shifted by about 20 mV to more negative membrane potentials upon a change from Ringer to isotonic KCl solution (G. Glassmeier & JR Schwarz, unpublished data). Further experiments have to be done to explain the high sensitivity of I_Ks activation to [K⁺]_o. Regardless of the underlying mechanism, this novel feature further supports the idea that KCNQ subunits (KCNQ2 by other criteria noted above) mediate I_Ks.

In summary, our nodal recordings strongly suggest that, in the largest fibres of sciatic nerve of the rat, I_Ks is mediated by homomeric KCNQ2 channels. The immunocytochemistry also supports this, but conclusions drawn from this approach alone were limited by the current lack of antibodies against the alternatively spliced KCNQ3 isoform, i.e. KCNQ3a. Since KCNQ3a is quite TEA insensitive (Wang et al. 1998), it is unlikely that this subunit makes a significant contribution to I_Ks in these large fibres.

Of native neuronal currents so far attributed to KCNQ subunits, I_Ks is the first current apparently mediated exclusively by homomeric KCNQ2 channels. The somatic M-current of sympathetic and dorsal root ganglia, hippocampus, and striatal medium spiny neurons are all primarily mediated by heteromeric KCNQ2/3 channels (Wang et al. 1998; Passmore et al. 2003; Hadley et al. 2003; Shen et al. 2005). I_Ks activation and deactivation kinetics are considerably faster than those of KCNQ currents or native neuronal M-currents. We cannot explain these differences, although interaction with as yet unidentified proteins in the node or post-translational modifications are potential mechanisms. The basis for this must be explored in future work.

Distinctive distribution of ion channels in the node of Ranvier

Mammalian nodes of Ranvier exhibit a complex distribution of ion channels: Na_v1.6 channels are located at a very high density in the nodal axolemma, whereas the delayed rectifier K⁺ channels K_v1.1 and K_v1.2 are predominantly confined to the juxtaparanodes (reviewed in Vogel & Schwarz, 1995; Poliak & Peles, 2003; Scherer et al. 2004). Devaux et al. (2004) showed that KCNQ2 channel proteins are present in the node of Ranvier and that they are co-localized with Na⁺ channel proteins and not with the juxtaparanodal Kv1.1 and Kv1.2 channels. These findings nicely correspond to the previous electrophysiological finding that the nodal I_Ks density is about 30 times larger than the internodal I_Ks (Röper & Schwarz, 1989). KCNQ3 channel proteins are also present in peripheral nerve fibres, but they seem to be located predominantly in the myelin; staining of KCNQ3 proteins in the nodal axolemma of peripheral nerve fibres was scarcely detected (Devaux et al. 2004). Our present data extend these previous results. In addition, we observed a more wide-spread staining of nodes for KCNQ3. While strong nodal staining for the KCNQ3 channel protein was found in up to 50% of small and medium-sized nerve fibres, such strong KCNQ3 staining was absent in nodes of large-diameter fibres. Since only the largest fibres were dissected for voltage-clamp experiments, the I_Ks data presented in this paper were obtained from nodes of Ranvier presumably containing only KCNQ2 channels. Since an antibody against the alternatively spliced form of KCNQ3 is not available, we could not exclude the possibility that this isoform was also present in nodal membranes by immunohistochemistry. The co-localization of KCNQ2 and KCNQ3 channels in the nodes of Ranvier in about 50% of the sampled small and medium-sized nerve fibres suggests that heteromeric KCNQ2/3 channels may exist in these fibres. However, at present there are no data available about the properties and function of I_Ks in these fibres with smaller diameters. We also do not know whether the nerve fibres which contain nodal KCNQ3 channels are correlated with a certain function. One possibility could be that some types of sensory nerve fibres are equipped with a different set of ion channels than motor nerve fibres. Although in amphibian nerve sensory and motor axons can be distinguished by differences in their biophysical properties (summarized in Vogel & Schwarz, 1995), these differences are more subtle in mammalian nerve fibres. Compared with sensory fibres, motor fibres probably express a smaller persistent Na⁺ conductance (Bostock & Rothwell, 1997) and less inward rectification mediated by the hyperpolarization-activated I_h (Bostock et al. 1994), but these differences are too small to be detected in recordings of macroscopic currents measured with the Nonner clamp method used in the present experiments.

Function of I_Ks

Since the voltage dependence of I_Ks activation in normal extracellular [K⁺] is relatively shallow, a small proportion of channels is expected to be activated at the resting potential. This explains how inhibition of I_Ks by XE991 induced a depolarization of a few millivolts and a slight reduction in the threshold potential in our experiments. Because of its slow activation kinetics and its small amplitude, I_Ks may exert little influence on repolarization of the action potential, and, consistent with this, we did not observe a significant change in the shape of the action potential after application of XE991. Due to the slow time course of deactivation, the amplitude of I_Ks successively increases during repetitive activity since part of I_Ks is activated by each action potential (Reid et al. 1993). This accumulation of I_Ks induces frequency adaptation. Since repetitive action potential firing may lead to a local increase in the K⁺ concentration in the nodal gap, this may shift the activation curve to more negative membrane potentials, thereby effectively increasing the amplitude of activated I_Ks. In our current-clamp experiments, total inhibition of I_Ks by XE991 decreased frequency adaptation and in the in vivo experiments it was virtually abolished. This accommodative function of I_Ks was first proposed by Dubois (1981, 1983). Later, the stabilizing function on membrane excitability and its role for frequency adaptation were clearly demonstrated experimentally in rat ventral root fibres by application of TEA (Baker et al. 1987). However, the present study is the first to demonstrate the function of I_Ks by using a selective KCNQ blocker as a pharmacological tool. Patients with a point mutation in the S4 domain of KCNQ2 (R207W) suffer from myokymia and exhibit a loss of frequency adaptation (Dedek et al. 2001) very similar to that observed in the anaesthetized rat after pharmacological blockage of I_Ks.

The separate contributions of fast and slow K⁺ channels to the electrical properties of myelinated axons were shown by studies of electrotonus and afterpotentials in rat spinal roots (Baker et al. 1987), using 4-aminopyridine and TEA, respectively, as selective channel blockers. After the inital fast change in potential (F component) on applying a 100 ms depolarizing current, there was a further slow depolarization (S1 component), attributed to depolarization of the internodal axolemma, followed by a slow repolarization (S2 component), that was blocked by TEA and attributed to I_Ks. Similar components can be seen in the accompanying changes in threshold (i.e. threshold electrotonus), where the S2 component appears as an accommodative increase in threshold. Excitability testing therefore enables these K⁺ channel functions to be investigated in vivo (Bostock & Baker, 1988; Yang et al. 2000). In our experiments i.p. injection of XE991 abolished all nerve fibre excitability functions attributed to I_Ks, i.e. accommodation to 100 ms subthreshold depolarizing currents as well as the post-depolarization undershoot in excitability. In addition, the late subexcitability after a single impulse or a short train of seven impulses was totally inhibited by XE991.

The in vivo experiments clearly show that blockage of I_Ks induces a drastic increase in the excitability of motor nerve fibres. Such a degree of hyperexcitability was only obtained in current-clamp experiments done on single nerve fibres isolated from the sciatic nerve under extreme artificial conditions, such as a total blockage of K⁺ currents by TEA in addition to full removal of external Ca²⁺ (Bergmann et al. 1968). The nerve fibres isolated from the sciatic nerve for voltage clamping may have been damaged during the dissection procedure, thereby increasing, for example, the unspecific leakage current. This could explain why these nerve fibres did not exhibit such a degree of repetitive activity as in the in vivo experiments testing the recovery cycle. These observations are in line with the finding that a mutation (R207W) in the voltage sensor of KCNQ2 channels causes myokymia and neonatal epilepsy (see above; Dedek et al. 2001).

In conclusion, we have shown beyond reasonable doubt that the slow potassium current at rat nodes of Ranvier, designated I_Ks, is generated by XE991-sensitive KCNQ channels. In the largest fibres, these channels appear to be KCNQ2 homomers. This has been shown at the level of the membrane currents in isolated nodes of Ranvier by their distinct pharmacology, and the in vivo recordings have confirmed that all the nerve excitability functions previously attributed to I_Ks, such as accommodation, spike-frequency adaptation and late subexcitability in the recovery cycle, are indeed caused by this current and generated by KCNQ channels. The functional contribution of the KCNQ3 subunits detected at nodes of smaller diameter fibres, and in myelin sheaths, remains obscure.