Human axons contain at least five types of voltage-dependent potassium channel
Sobell Department of Neurophysiology, Institute of Neurology, Queen Square, London WC1N 3BG, UK
Corresponding author G. Reid: Department of Animal Physiology and Biophysics, Faculty of Biology, University of Bucharest, Splaiul Independentei 91-95, Bucharest 76201, Romania. Email: email@example.com
1We investigated voltage-gated potassium channels in human peripheral myelinated axons; apart from the I, S and F channels already described in amphibian and rat axons, we identified at least two other channel types.
2The I channel activated between -70 and -40 mV, and inactivated very slowly (time constant 13.1 s at -40 mV). It had two gating modes: the dominant (‘noisy’) mode had a conductance of 30 pS (inward current, symmetrical 155 mM K+) and a deactivation time constant (τ) of 25 ms (-80 mV); it accounted for most (≈50-75 %) of the macroscopic K+ current in large patches. The secondary (‘flickery’) gating mode had a conductance of 22 pS, and showed bi-exponential deactivation (τ= 16 and 102 ms; -80 mV); it contributed part of the slow macroscopic K+ current.
3The I channel current was blocked by 1 μM α-dendrotoxin (DTX); we also observed two other DTX-sensitive K+ channel types (40 pS and 25 pS). The S and F channels were not blocked by 1 μM DTX.
4The conductance of the S channel was 7-10 pS, and it activated at slightly more negative potentials than the I channel; its deactivation was slow (τ= 41.7 ms at -100 mV). It contributed a second component of the slow macroscopic K+ current.
5The F channel had a conductance of 50 pS; it activated at potentials between -40 and +40 mV, deactivated very rapidly (τ= 1.4 ms at -100 mV), and inactivated rapidly (τ= 62 ms at +80 mV). It accounted for the fast-deactivating macroscopic K+ current and partly for fast K+ current inactivation.
6We conclude that human and rat axonal K+ channels are closely similar, but that the correspondence between K+ channel types and the macroscopic currents usually attributed to them is only partial. At least five channel types exist, and their characteristics overlap to a considerable extent.
Myelinated axons have been shown to contain at least three types of potassium current activated by depolarization: two fast currents, labelled IKf1 and IKf2, and one slow current, labelled IKs. These have been identified in frog nodes of Ranvier (Dubois, 1981) and internodes (Grissmer, 1986), and similar currents appear to be present in rat axons, albeit with a different spatial distribution (Röper & Schwarz, 1989; Corrette et al. 1991). Single-channel recordings in myelinated axons from Xenopus (Jonas et al. 1989) and rat (Safronov et al. 1993) reveal three types of voltage-dependent K+ channel: the F (fast) channel, which activates rapidly over a broad range of relatively positive membrane potentials, and deactivates rapidly; the S (slow) channel, which activates at more negative potentials and has slower kinetics; and the I (intermediate) channel, which activates at similar potentials to the S channel, and has kinetics intermediate between those of the other two types. It has been suggested that these channels underlie the IKf2, IKs and IKf1 currents, respectively (Jonas et al. 1989). Experiments in undissected axons allow conclusions to be drawn about the likely functions of these currents: the fast K+ currents IKf1 and IKf2 appear to limit repetitive activity after a single impulse; accommodation is influenced by IKf1 and IKs; and IKs is also important for spike-frequency adaptation (Baker et al. 1987; Poulter et al. 1989). The resting potential is probably determined in large part by the internodal IKs and IKf1 conductances, which are active near the resting potential (Grafe et al. 1994). In mammals, the nodal fast K+ conductances are small and their activation plays no significant part in repolarization after the action potential (Chiu et al. 1979). The field has been reviewed by Vogel & Schwarz (1995).
Although human axons generate action potentials in a similar way to those of other species, and human nodes of Ranvier contain ionic currents with similar properties to those in rat and amphibian axons (Schwarz et al. 1995), in vivo recordings reveal electrophysiological differences between human and rat axons which suggest different populations of voltage-gated K+ channels (Bostock & Baker, 1988). Such species differences would have important consequences, because axonal K+ channels are one target for symptomatic treatment of demyelinating diseases like multiple sclerosis (Davis et al. 1990; Schwid et al. 1997). We therefore investigated human axonal ion channels with the patch-clamp technique, and our initial impression was that channels in rat and human axons are similar (Scholz et al. 1993). We have now characterized in detail the properties of single K+ channels in human axons, and find close quantitative similarities with those in rat and amphibian axons; the species differences observed in vivo therefore do not result from substantial differences in the single-channel properties. Our results do not support a simple classification of axonal K+ channels into three distinct types responsible for the three macroscopic currents described above: at least five types are present, and their kinetic properties show a substantial degree of overlap.
The source of human nerve was excess donor material from nerve-graft operations, which would otherwise have been discarded. The procedure was approved by the ethical committee of the Institute of Neurology and the National Hospital for Neurology and Neurosurgery. The age of the patients was 17-48 years. Short pieces (< 5 cm) of nerve were transferred directly from the patient into sterile tissue culture medium (Dulbecco's modified Eagle's medium (DMEM)-Ham's F12; Gibco, Paisley, Scotland), which had been saturated with 95 % O2-5 % CO2, and were transported to the laboratory at the Institute of Neurology by courier; nerves arrived within 2 h of removal and were thereafter stored at 10°C, under which conditions they remained excitable for up to 3 days. No differences were noted in this study between channel properties in patches from fresh nerve and from nerve which had been stored for 3 days. Most of the nerves which were studied were cutaneous, either the sural nerve or the medial cutaneous nerve of the forearm, and in these cases the axons are identified as sensory in the figure legends. Axons not so identified were from mixed nerves, either ulnar, median or brachial plexus.
Individual fascicles were separated from the nerve by gripping the endoneurium with fine forceps and pulling. Surprisingly little force is required to do this, only slightly more than is required to cause the diagonal striations (p. 39 in Sunderland, 1968) to disappear, indicating straightening of the axons from their usual undulating course. We are therefore confident that the axons were not unduly stretched by this procedure. Fascicles were tied at the ends with sewing thread, pinned onto the base of a 1 ml chamber (diameter 25 mm) moulded from Sylgard 184 resin (Dow Corning, Seneffe, Belgium), and incubated in the presence of the following enzymes: first stage - 2 mg ml−1 collagenase (Sigma, Type XI), 0.1 mg ml−1 protease (Sigma, Type X), 1-2 h; second stage - 1 mg ml−1 protease (Sigma, Type XXIV), 30-40 min. Enzyme mixtures were made up in standard extracellular solution (see below) to which 10 mM D-glucose had been added. Incubations took place at 37°C, with linear agitation (50 strokes min−1, stroke length ∼6 cm). We found it to be important for good dissociation that the fascicles were not pulled taut when pinned to the Sylgard base, so that they could move during agitation, which perhaps allowed easier access of the enzymes. After enzyme treatment, fascicles were cut into short (∼2 mm) sections with fine scissors, and shaken gently to allow the nerve fibres to separate. This created a suspension of single fibres, which were transferred to 35 mm tissue culture dishes (type 25000 or 29000; Corning Glass Works, Corning, NY, USA) and allowed to settle. To improve adhesion, the dishes had already been coated with a layer of highly viscous silicone fluid (DC200/100 000 cs; Dow Corning) applied on top of a very thin layer of cured Sylgard 184.
Patch-clamp recordings (Hamill et al. 1981) were made at room temperature (22-27°C) using an Axopatch 200 amplifier (Axon Instruments, Foster City, CA, USA). Pipettes were pulled from borosilicate glass (GC150F; Clark Electromedical Instruments, Pangbourne, UK) on a Narishige PP-83 puller (Narishige Ltd, Tokyo, Japan) which had been modified by the addition of a laboratory-made constant-current circuit (± 0.5 % at 13 A). They were coated with Sylgard 184 and heat-polished; bubble numbers were 2-3 (Corey & Stevens, 1983) and resistances 15-30 MΩ. The current was filtered at 10 kHz (-3 dB) using the 4-pole Bessel filter of the Axopatch, and this signal was led to a digital audiotape recorder (DTR-1202; Biologic, Claix, France) and to a 4-pole Butterworth filter (Neurolog NL125; Digitimer, Welwyn Garden City, UK) which was set to a corner frequency of 1 kHz in all the recordings presented here. The output of this filter was led to the input of an A/D converter (MC-DAS 1512; Scientific Solutions, Mentor, OH, USA), mounted in an IBM PS/2 computer. Data acquisition and control of the pulse protocols was done using pCLAMP software, version 5.6 (Axon Instruments). Data analysis was done with pCLAMP and with software written by ourselves; least-squares fits were performed using FigP software (Biosoft, Cambridge, UK). Leakage and capacity currents were compensated digitally after storage of the raw data, by subtracting currents in response to either inverted pulses, pulses that produced no channel activity, or pulses to 0 mV in symmetrical 155 mM K+. Figures have been corrected where appropriate for a liquid junction potential of 3 mV between 155 mM KCl solution and standard extracellular solution, with the KCl solution negative (Marty & Neher, 1995).
The standard extracellular solution contained (mM): NaCl, 154; KCl, 5.6; CaCl2, 2.2; MgCl2, 1.2; Hepes, 5.46; Na-Hepes (Hepes, sodium salt), 4.54 (pH 7.4 at 25°C). The high-[K+] extracellular solution contained: NaCl, 6.5; KCl, 155; CaCl2, 2.2; MgCl2, 1.2; Hepes, 5.46; Na-Hepes, 4.54 (pH 7.4 at 25°C). The internal solution contained: KCl, 149; NaCl, 5; CaCl2, 1.21; MgCl2, 1.21; Hepes, 3.96; Na-Hepes, 6.04; EGTA, 3; KOH, 6 (pH 7.2 at 25°C). The final concentration of Ca2+ in this solution was 10−7 M and of Mg2+ was 1.1 mM. These concentrations were calculated according to Durham (1983), with binding constants for EGTA from Harrison & Bers (1989) and Durham (1983); corrections for temperature, ionic strength, and EGTA purity were made according to Harrison & Bers (1989) and Miller & Smith (1984). External solutions contained 300 nM tetrodotoxin (TTX, Sigma) to block Na+ currents. The K+ channel blocker α-dendrotoxin (DTX; Latoxan, Rosans, France) was dissolved at 100 μM in distilled water, and stored at -18°C in 10 μl aliquots; when required, the stock solution was added directly to the external recording solution, diluting that solution by up to 1 % (at 1 μM DTX). Solutions were applied through a multi-barrel application system, described previously (Reid, 1993). Hepes and EGTA were from Sigma, and salts were from Merck.
Verification of patch origin
The origin of 20 outside-out patches was tested by including Lucifer Yellow CH (dilithium salt, 1 mg ml−1; Sigma) in the pipette solution (Wilson & Chiu, 1990). The pattern of dye diffusion indicated that all 20 were from axonal membrane (Fig. 1B). We therefore consider the probability to be high that other patches were also from axonal and not Schwann cell membrane. Patches were made at nodal, paranodal and internodal sites, the maximum distance from the node being about 100 μm. The node is usually clearly distinguishable as a dark line or small bulge in the middle of the constricted axon segment (Fig. 1A); our use of the terms ‘paranodal’ and ‘internodal’ is based on axon diameter, with ‘paranodal’ referring to the constricted segment on either side of the node and the regions of the axon tapering towards it from the internode, and ‘internodal’ referring to the regions where the full internodal diameter has been reached.
Potassium channels activated by depolarization were present at moderately high density at all patch sites, whether nodal, paranodal or internodal. We were not able to reach firm conclusions on the density or distribution of the different channel types described below: each of them was found in patches from nodal, paranodal and internodal membrane, and no clear systematic differences were noted in patches from different sites. The classification of Jonas et al. (1989), described in the Introduction, is adequate for a large part of our observations, and we will use it as a starting point in the presentation of our results; the properties of the channels we have observed are summarized in Table 1. Those of our observations which are not consistent with this scheme will be evident as they arise, and will be treated further in the Discussion.
Table 1. Properties of the five identified K+ channel types in human axons
Block by 1 μM DTX
aInward current in symmetrical 155 mM K+. bHalf-maximal activation and cslope obtained from Boltzmann equation (see legend to Fig. 2D) in symmetrical 155 mM K+. dVoltage dependent (range −40 to +80 mV). n.d., not determined (not distinguishable from I channel).
I (noisy mode)
25 (at −80 mV)
1.4 (at −100 mV)
62 ms (at +80 mV)
42 (at −100 mV)
No inactivation observed
activated at −40 mV
Probably within 1–2 s
Single I channels and other types of dendrotoxin-sensitive K+ channel
We have already published a brief description of the human I channel (Scholz et al. 1993). Virtually all patches contained K+ channels of this type. Figure 2A shows a typical recording from an outside-out patch which contained two I channels. These channels were first activated by depolarization to around -70 mV, and fully activated at around -40 mV, reaching a maximum open probability (Popen) of ∼0.5. The time to half-maximal activation was steeply voltage dependent in the range between -60 and -10 mV, reaching an asymptote of about 1 ms at potentials positive to +10 mV (Fig. 2E). The deactivation of I channels could be fitted with a single exponential, with a mean time constant of 22.5 ms at -80 mV (range 21.2-24.9 ms, n= 3). In their dominant gating mode (see below), their conductance in symmetrical 155 mM K+ was about 30 pS for inward currents, and about 17 pS for outward currents; in standard extracellular solution their conductance was ohmic and around 12 pS (Fig. 2B).
The I channel is blocked almost completely by 1 μM α-dendrotoxin (DTX; Scholz et al. 1993). Some patches contained other types of DTX-sensitive K+ channel, with conductances for inward current of about 40 pS and about 25 pS in symmetrical 155 mM K+. The 40 pS channel showed brief re-openings on returning to the holding potential after depolarizing pulses (Fig. 2C), a feature never observed in the I channel. The re-openings were most frequent immediately after the end of the depolarizing pulse, and declined over a period of some hundreds of milliseconds thereafter. Their presence depended on channel activity during the depolarizing pulse: pulses which elicited no channel activity were not followed by re-openings, and their frequency increased with the potential of the depolarizing pulse between -70 and -40 mV. These characteristics of the re-openings are consistent with recovery from N-type inactivation (see Discussion).
Two modes of I channel gating
In several patches, the I channel displayed two different conductance levels in tail events, which are visible in Fig. 2A. The most frequent type, visible in the second and bottom traces in this figure, had a ‘noisy’ appearance, characterized by long openings and very brief, incompletely resolved closings, and had a conductance of about 30 pS for inward currents in symmetrical 155 mM K+. The other type, seen in the top and third traces, showed bursts of shorter openings, with longer, well-resolved closings, giving these events a ‘flickery’ appearance. The conductance of these openings was about 22 pS. About 10 % of tail events in this patch were in ‘flickery’ mode, and about 90 % in ‘noisy’ mode.
At first sight it appeared that this patch contained channels of two types: I channels, producing the ‘noisy’ events, and channels of a different type, responsible for the ‘flickery’ events. However, if this were the case, the patch would have had to contain at least four channels, because some tails contained two ‘noisy’ channels and others two ‘flickery’ channels (see Fig. 2A, top and second traces). We estimated the channel number using binomial statistics (Colquhoun & Hawkes, 1983): in tail currents following 284 depolarizing pulses to -8 mV, over a 20 min period, about half contained a maximum of one or two simultaneous openings, with none containing more than two. This is consistent with binomial statistics for a patch containing two channels, but if the patch had contained four independent channels, we should have observed three or four simultaneous openings in 14 % (40/284) of the tail currents. We therefore conclude that the patch contained only two channels, which alternated between the ‘flickery’ and ‘noisy’ modes of gating. Switching between these two gating modes, with no intervening closure that could be resolved, was occasionally observed directly during single bursts of I channel openings in other patches (not shown).
When the distribution of times until the last closing of the channels during tail currents was plotted for ‘noisy’ and ‘flickery’ tail events, the ‘noisy’ mode of gating produced a distribution with a single time constant of 24.9 ms at -80 mV (Fig. 3B, top), close to that of the averaged I channel tail current (see above). The ‘flickery’ mode of gating produced an additional slower component, with a time constant of 102 ms (Fig. 3B, bottom), similar to the deactivation time constant of the slow component of macroscopic K+ current in some patches (see below).
The ‘flickery’ mode activated at more negative potentials than the ‘noisy’ mode: in some patches, sporadic openings in the ‘flickery’ mode were already apparent at -80 mV, and constituted the whole of the channel activity at this potential. The activation of ‘flickery’ activity at -80 mV, from a holding potential of -120 mV, was slow and could be described by a single exponential, with a time constant of 50 ms (not shown).
Macroscopic currents with I-like kinetics
In large tail currents (50-250 pA) recorded in symmetrical 155 mM K+, the deactivation process could be described by the sum of two or three exponential functions. In all of 30 large patches, the largest component of macroscopic K+ current (about 50-75 % of the current in most patches, and virtually all of the current in some patches) had a voltage dependence and time course similar to that of the I channel in its ‘noisy’ gating mode (see above). At a holding potential of -100 mV, the mean time constant was 12.1 ms (range 7.6-15.5 ms, n= 5), and at -80 mV, it was 17.5 ms (range 10.7-29.2 ms, n= 3). In all of seven patches, this current component was blocked completely by 1 μM DTX (Fig. 3A).
The voltage dependence of activation of this component is shown in Fig. 2D, and the time course of its activation in Fig. 2E. In both of these respects, as well as its deactivation kinetics and its DTX sensitivity, its behaviour is accounted for satisfactorily by the properties of single I channels in the ‘noisy’ gating mode.
Inactivation of the I channel and of the macroscopic I-like current
In patches with large K+ currents, the inactivation process at -40 mV had both fast and slow components; the slower component could be fitted with a single exponential, with a mean time constant of 13.1 s (range 7.2-20.7 s, n= 3; Fig. 4A), while the faster component had a time constant of about 2 s. Both were blocked by 1 μM DTX. The slower of the two processes can be accounted for by the inactivation of the I channel: it could be followed at potentials between -60 and +60 mV in one patch, where the currents were large enough to allow measurement of the time constant, but small enough for single I channels to be visible during much of the inactivation process (Fig. 4B, bottom trace). Inactivation became slower at more positive potentials (Fig. 4C), and it was incomplete at potentials positive to +30 mV (Fig. 4B, top two current traces): at these potentials, I channels remained active throughout pulses of several minutes duration. A related observation is that complete inactivation of the I channel at -40 mV could be overcome by strong depolarizing pulses (Fig. 4D). Strong depolarizing pulses could also relieve DTX block: in the presence of 1 μM DTX, single I channel activity was visible during pulses positive to +30 mV, although the block was complete at more negative potentials (not shown). The voltage dependence of the current escaping from DTX block was similar to that escaping from inactivation. The implications of these observations for the mechanism of inactivation are considered in the Discussion.
The voltage dependence of steady-state inactivation of the macroscopic I-like current was followed in one patch in which the currents were dominated by this component. Holding potentials between -100 and -40 mV were maintained for at least 3 min, and then 250 ms depolarizing pulses were applied at 2 s intervals to potentials between the holding potential and +60 mV. Steady-state inactivation at a given holding potential was estimated by measuring the current amplitude during each depolarizing pulse (at 25 ms), and expressing this as a fraction of the current amplitude during the same depolarizing pulse from a holding potential of -100 mV; this fraction was then averaged across all the depolarizing pulses at each holding potential. The voltage dependence of inactivation is shown in Fig. 4E. The maximum of the fitted Boltzmann function is around 5 % greater than the current at -100 mV, implying that only a small fraction of the K+ current is inactivated at -100 mV. The fit is subject to some uncertainty because the patches often became unstable at potentials negative to -100 mV, making measurements in this region difficult. About half of it is inactivated at -80 mV. To eliminate the possibility that the small rapidly inactivating current in this patch affected the measurement, we estimated the voltage dependence of inactivation from currents near the beginning (at 25 ms) and at the end of the 250 ms depolarizing pulses. The measurements were identical, implying either that the fast-inactivating component had a negligible effect on the measurement, or that the voltage dependence of inactivation of the two components is similar; in either case, the measurements are a good estimate of the steady-state voltage dependence of I channel inactivation.
Single S channels
In six of seven patches, block of the I channel current with 1 μM DTX (Fig. 3A) revealed identifiable single channels of other types. In five of these patches, we observed a channel with small conductance (Fig. 5A), which became active at around -80 mV, and deactivated more slowly at -100 mV than the I channel. It appears to be the same as the S channel in Xenopus and rat axons (Jonas et al. 1989; Safronov et al. 1993), and will also be called the S channel in this study.
The amplitudes of single S channel currents are shown in Fig. 5B. In symmetrical 155 mM K+, the conductance of this channel type for inward current was usually around 10 pS. Two S channels in one patch had a distinctly smaller current than the rest, and these currents have been fitted separately in this figure to reveal a conductance of ∼7 pS. We could not measure the conductance of the S channel for outward current because its activation was rapid at positive potentials, and it usually remained open once activated, without the brief closings during depolarizing pulses which we observed in I channels.
In the patch shown in Fig. 5A, the number of S channels was estimated to be three using binomial statistics. The voltage dependence of their open probability is shown in Fig. 5C. Like I channels, the maximum open probability of S channels was substantially less than 1: in the patch shown in Fig. 5A it was 0.62. Their activation was steeply voltage dependent over the range of potentials from -80 to -50 mV, slightly more negative than the range of potentials over which the I channel is activated. The deactivation of the S channels in this patch was fitted well by a single exponential with a time constant of 41.7 ms at -100 mV.
Macroscopic currents with slow kinetics
In almost all patches with large K+ currents, a fraction of the tail current (up to 25 %) deactivated more slowly than the component associated with the I channel. In some cases, we could account for this component by the properties of single S channels. In other cases, as described below, its origin is different.
In four of five patches, the time constant of deactivation was around 40 ms at -100 mV, similar to that of the S channel. The voltage dependence of activation of the slow component in these four patches, measured as described above for the intermediate component, is shown in Fig. 5C. In two of these four patches, the voltage dependence of activation was indistinguishable from that of the S channel (Fig. 5C a), while in the other two, it was much less steep (Fig. 5C b). In only two of these four patches, therefore, can the slow component of macroscopic K+ current be accounted for by the properties of single S channels. We did not observe any single channels whose voltage dependence could account for the macroscopic currents in Fig. 5C b.
In the fifth of these patches, deactivation was much slower, with a time constant of about 90 ms at -100 mV and a little over 100 ms at -80 mV. In addition, in this patch, no S channels were visible in tail currents; instead, the channel activity resembled the I channel in its ‘flickery’ mode (see above). The deactivation time constant was also consistent with ‘flickery’ gating of I channels, and the voltage dependence of activation of the slow K+ current component in this patch (not shown) was indistinguishable from that of the I channel. We therefore conclude that the slow component in this patch was satisfactorily accounted for by the I channel in its ‘flickery’ mode.
To summarize, the slow component in patches with large K+ currents can only partially be accounted for by the S channel; part of it is produced by the I channel in its ‘flickery’ gating mode, and the remainder does not correspond to any identified type of single channel.
Single F channels
In two of seven patches, application of 1 μM DTX revealed single channels with a large conductance, activating at more positive potentials than the I channel (Fig. 6A). They deactivated very rapidly (τ= 1.4 ms at -100 mV) and inactivated over a period of a few hundred milliseconds. They showed occasional re-openings on returning to the holding potential after depolarizing pulses. This channel type appears to be the F channel described in Xenopus and rat (Jonas et al. 1989; Safronov et al. 1993), and will also be called the F channel here. F channels were rarely encountered in human axons: in only two of six patches where 1 μM DTX exposed single channels were they F channels, and we never saw them in single-channel patches without DTX. This is in apparent contrast to rat axons (see Discussion).
The conductance of the F channel was about 50 pS for inward current and about 30 pS for outward current in symmetrical 155 mM K+ (Fig. 6B). The voltage dependence of activation could be estimated in only a single patch, which contained at least three F channels (the data were not sufficient to allow us to estimate the channel number using binomial statistics). The voltage dependence of activation is shown in Fig. 6C; the potential for half-maximal activation is close to 0 mV, more than 40 mV positive to that of the I channel, and the slope is substantially less steep. The maximum Popen, like that of the I and S channels, is less than 1: the maximum of a Boltzmann function fitted to the number of open channels corresponded to 2.4 open channels at potentials positive to +40 mV, implying a maximum Popen of not more than 0.8. The F channel activated at a similar rate to the I channel, but at potentials about 20 mV more positive (Fig. 6D).
F channel inactivation was recorded at +80 mV (Fig. 7A). The time course of inactivation at this potential can be fitted with a single exponential with a time constant of 62 ms, but inactivation was not complete: although the 250 ms pulses were sufficiently long for this process to have reached a steady state, some openings of single F channels remained visible throughout the depolarizing pulses.
Macroscopic currents with fast deactivation and fast inactivation
Some patches with large K+ currents displayed two features reminiscent of F channels: a fast component of deactivation, and a fast phase of inactivation. We will deal with deactivation first.
After short depolarizing pulses, a fast component of deactivation with a time constant of 1-2 ms was revealed by fitting sums of exponential functions to macroscopic tail currents. Its amplitude was usually too small to allow its characteristics to be measured. However, one patch had a fast component making up almost half of the total current, and we analysed this in detail. The fast component was present after pulses to potentials between -40 and +70 mV. In order to test whether its voltage dependence of activation was consistent with that of the F channel, we subtracted the tail currents after pulses to -40 mV (this potential is sufficient to fully activate the I and S channels, but only minimally activates the F channel) from those after pulses to more positive potentials. The current resulting from this subtraction had a time constant of deactivation of 2.4 ms at -80 mV (compared with 1.4 ms for the F channel; see above). Its amplitude was measured by fitting a 2.4 ms exponential component to the subtracted tail currents, and the resulting voltage dependence of activation, shown in Fig. 6C, was indistinguishable from that of the F channel. We therefore conclude that the fast-deactivating K+ current in this patch is adequately accounted for by the F channel.
A more noticeable feature in macroscopic currents was the fast component of inactivation. A fast-inactivating component of the K+ current was visible in about half of all large patches, and in three patches it was large enough to measure its amplitude and time course. Fits of a single exponential function to the fast inactivation process in these patches produced a time constant which was steeply voltage dependent, from around 500 ms at -40 mV to around 100 ms at +60 mV. However, the quality of the fit was poor at potentials positive to +10 mV, and a better fit was obtained with two time constants (Fig. 7B). The faster of these two time constants is similar to that obtained from recordings of single F channels. The slower component is present at potentials as negative as -40 mV (see also Fig. 4A), at which few F channels are active (see Fig. 6C). Its amplitude did not increase at potentials positive to -40 mV, suggesting that it is due to channels which are already fully active at this potential and not to the F channel (see Discussion). It is about 10 times faster than the I channel inactivation process described above.
These observations suggest that there are two components of fast inactivation, one about an order of magnitude faster than the other, and that the apparent voltage dependence of the rate of inactivation at potentials positive to +10 mV (becoming faster at more positive potentials) is actually due to the voltage dependence of the amplitude of the faster component (which becomes larger at more positive potentials). Accordingly, the currents in these three patches at -40 mV were fitted with a single exponential function, producing a time constant of around 0.5-1 s, and this time constant was fixed when subsequently fitting the sum of two exponential functions to currents at more positive potentials (Fig. 7B). The resulting fits were much improved over those with only a single exponential, and the faster time constant was insensitive to changing the slower one between 0.5 and 2 s. The faster time constant of inactivation in four patches is plotted against voltage in Fig. 7C. It is close to that measured for the F channel (see above), and does not appear to be voltage dependent.
We therefore conclude that the fast components of macroscopic K+ current inactivation result from two channel types. The faster component of inactivation, with a time constant around 50 ms, can be accounted for by the F channel. The slower, with a time constant of 0.5-2 s, is associated with a conductance which is already fully activated at -40 mV and is blocked by 1 μM DTX. Its origin is considered further in the Discussion.
Similarities and differences between human and rat axonal potassium channels
Our results can be compared directly in most cases with measurements in rat axons (Safronov et al. 1993), and in some cases with those in Xenopus axons (Schwarz & Vogel, 1971; Koh & Vogel, 1996) where the rat data are unavailable. Figure 8 shows the voltage dependence of activation of the human I-like macroscopic current and the human S and F channels, compared with the corresponding measurements in rat axons reported by Safronov et al. (1993). Measurements in the two species are closely similar; the difference is no greater than the variation observed within one species, in single Xenopus I channels (Jonas et al. 1989; Koh & Vogel, 1996). The time to half-maximal activation of the human I channel is comparable to that in Xenopus (Koh & Vogel, 1996), when allowance is made for the difference in temperature, assuming a Q10 (temperature coefficient over a 10°C temperature range) of about 3. The time constants of deactivation of the I and F channels are almost identical to the corresponding measurements in the rat, but the human S channel deactivates about three times faster than that in the rat; a similar species difference is evident in the slow macroscopic K+ current in voltage-clamped nodes (Röper & Schwarz, 1989; Schwarz et al. 1995). The I channel in excised patches deactivates much more slowly than the current in intact nodes which it probably underlies (Schwarz et al. 1995); this discrepancy has already been commented on in rat and Xenopus axons (Jonas et al. 1989; Safronov et al. 1993), and the cause remains to be explained. Inactivation of the human I channel appears to be somewhat slower than that in the rat at +40 mV, and that of the human F channel more rapid. The voltage dependence of I channel inactivation is similar to that of the slow component of K+ current inactivation in Xenopus (Schwarz & Vogel, 1971) and, as in that study, inactivation of the human I channel becomes slower and remains incomplete at more positive potentials.
One of the aims of this study was to try to explain the differences in electrotonus measured in vivo by Bostock & Baker (1988). In that study, electrotonus in rat axons was made more similar to that in human axons by application of low concentrations of 4-aminopyridine (4-AP) and tetraethylammonium, which block fast and slow K+ conductances, respectively (Baker et al. 1987). This suggests that in rat axons, more K+ channels are active near the resting potential than in human axons, which could be the result of a higher density, or a different voltage dependence of activation. Patch-clamp studies give us no reliable information about channel density; as far as the voltage dependence is concerned, we found no evidence of a species difference between rat and human axons for any of the three major K+ channel types. None of the other minor species differences in K+ channel properties mentioned above is capable of explaining the electrophysiological differences observed in vivo between rat and human axons by Bostock & Baker (1988).
The possibility remains, however, that the quantitative species differences we have observed could have physiological implications. The faster rate of deactivation of the human S channel than that of the rat would produce different behaviour during trains of action potentials at moderate frequencies, within the range commonly encountered in motor fibres and large afferents: assuming a Q10 of about 3, at a frequency of 20 Hz in vivo the human S channel would be expected to deactivate completely between impulses, while the rat S channel would remain substantially open, lowering the safety factor for conduction of sustained trains of impulses. The recently described frequency-dependent conduction block in single peripheral nerve fibres of human subjects (Inglis et al. 1998) could also be due to the rate of S channel deactivation. Here, conduction block was observed at frequencies of 20 Hz and above, and conduction was restored at lower frequencies. At frequencies above 20 Hz, we would expect the S channel to deactivate incompletely between impulses, leading to a raised threshold for the generation of subsequent action potentials. We suggest that conduction failure during naturally evoked activity might be explained on the basis of the K+ channels described in this study.
There is circumstantial evidence that F channels are less frequent in human than in rat axons: block of I channels by DTX in rat axons leaves a relatively homogeneous population of F channels (Safronov et al. 1993), whereas in human axons we observed S channels much more commonly than F channels in 1 μM DTX (see Results). Similarly, patches from rat axons are occasionally dominated by an F channel component (Schneider et al. 1993), whereas we found that all of 30 large patches from human axons were dominated by the I channel component. However, since F channels are expected to be closed near the resting potential, this difference is not likely to be responsible for the differences in electrotonus.
Which types of potassium channel are present in human axons?
While our results are in accordance with the idea that the three components of macroscopic K+ current first identified by Dubois (1981) are largely contributed by the three types of K+ channel described by Jonas et al. (1989) and Safronov et al. (1993), some of our observations indicate that the picture is more complex.
A single channel type can produce behaviour associated with more than one macroscopic current component. The I channel has two modes of gating, named here ‘noisy’ and ‘flickery’; two different gating modes have also been mentioned in the Xenopus I channel (Koh & Vogel, 1996). The ‘noisy’ mode is the more frequent, and underlies the largest component of macroscopic K+ current (see Results). However, deactivation in the ‘flickery’ mode has an additional, slower time constant, and we have shown that the ‘flickery’ mode of the I channel can account for part of the slow macroscopic current. It is interesting in this connection that Dubois (1981) observed a bi-exponential time course for the deactivation of the IKf1 component of macroscopic K+ current in intact frog axons, with one time constant about four times longer than the other. This behaviour is expected to produce a DTX-sensitive IKs-like current with a predominantly paranodal and internodal location (Röper & Schwarz, 1989), which would explain some of the effects of DTX on the slow process of accommodation (Poulter et al. 1989).
The slow macroscopic current is therefore contributed by several channel types: we have shown that in some patches, the slow component of macroscopic current can be accounted for adequately by the S channel, and in others, as discussed above, by the ‘flickery’ mode of the I channel; however, in other patches, the slow K+ current behaves differently from either of these two channel types. The voltage dependence of activation of the I and S channels is quite different from that of IKs in intact nodes: the I and S channels activate steeply at potentials between -70 and -40 mV, while the nodal IKs activates much less steeply, over a range from -100 to 0 mV (Schwarz et al. 1995). The same discrepancy is evident in the rat (Röper & Schwarz, 1989; Safronov et al. 1993). In two patches in this study (Fig. 5C b), the voltage dependence of the slow macroscopic current was more similar to that of the macroscopic IKs current in the node of Ranvier than to that of the S channel. It is possible that the slow macroscopic current is made up of the S channel, the I channel in its ‘flickery’ mode, and another channel type, as yet unidentified; another possibility is that a cytoplasmic component, missing in excised patches, modifies the voltage dependence of the S channel to make it more similar to IKs in intact nodes.
The DTX-sensitive channel family and K+ channel inactivation
Along with the rat I channel, Safronov et al. (1993) described a channel with a conductance of 40 pS, which they suggested was a subtype of the I channel. The I channel and the 40 pS channel are also present in human axons, as well as an apparently I-like channel with a conductance of about 25 pS, and all three are blocked by 1 μM DTX. This finding is in accord with immunocytochemical evidence in central myelinated axons: both Kv1.1 and Kv1.2 α-subunits are present, and they form heteromultimeric channels in vivo (Wang et al. 1993), suggesting that (if the situation in peripheral nerve is similar) we ought to find at least three distinct but related DTX-sensitive K+ channels in axons. More recent studies in rat dorsal root ganglion cells (Gold et al. 1996; Safronov et al. 1996) show a similar variety of delayed rectifier K+ channels. However, the observation that most of the K+ current in human axons is blocked by 1 μM DTX (Fig. 3A) implies that a limited set of α-subunits may contribute the major part of the K+ current, since only three types of α-subunit are known to be DTX sensitive (Kv1.1, Kv1.2 and Kv1.6; Robertson, 1997).
The observation that the 40 pS channel re-opens on returning to the holding potential after a depolarizing pulse (Fig. 2C) suggests that it may inactivate by an N-type (‘ball-and-chain’) mechanism. N-type inactivation has features in common with intracellular open-channel block (Demo & Yellen, 1991), including the presence of re-openings during recovery from inactivation: the ‘ball’ is considered to block the channel in the open state, and must unbind from its site before the channel can close. The time course of the re-openings we observed in the 40 pS channel, and their dependence on previous channel activity, are exactly what would be predicted from a mechanism like that described by Demo & Yellen (1991) in the Shaker H4 channel. If this is the case, it would imply that the inactivation of the 40 pS channel is rapid, because the 1 s pulses preceding the tail currents in Fig. 2C must already have induced substantial inactivation; this makes the 40 pS channel a likely candidate for the channel underlying the 0.5-2 s component of inactivation. This suggestion is supported by recent work in rat axons, showing that the 0.5-2 s component of inactivation is contributed by a channel which is distinct from the I channel and which has two mechanisms of inactivation, one fast and one slow (G. Reid, V. Ristoiu, A. Babes & T. Gliga, unpublished).
In contrast, the I channel in human axons does not appear to inactivate by an N-type mechanism; since the molecular mechanisms of slow inactivation have recently been studied intensively in cloned K+ channels, it was of interest to us to compare the features of slow inactivation in this native K+ channel with those described for cloned channels. Two observations we made in this study - the slower and incomplete inactivation in patches held at positive potentials, and the ‘knockout’ of inactivation by brief pulses to positive potentials - suggest that outward K+ current may antagonize I channel inactivation, as has been reported for C-type inactivation (Baukrowitz & Yellen, 1995). However, the voltage dependence of the rate of C-type inactivation in a Kv1.4 channel (Rasmusson et al. 1995) is clearly different from that in the I channel reported here. It is therefore not safe to suppose that the slow inactivation of the I channel is C-type.
In summary, we have shown that the three major types of voltage-gated K+ channel identified in human axons are quantitatively very similar to the corresponding channel types in rat axons, apart from minor differences in their kinetics. In addition, there are two other types of DTX-sensitive channel with some similarities to and some differences from the I channel, and there may be one more K+ channel with slow kinetics. It seems puzzling that the axon expresses several channel types whose macroscopic behaviour in excised patches is so difficult to distinguish. One obvious possibility is that their behaviour is not always similar in the intact axon, because they are modulated in different ways (Chung & Kaczmarek, 1995). Our results and those from previous patch-clamp studies in axons suggest that the I channel (and perhaps the S channel too) requires some cytoplasmic factor, missing in excised patches, for its normal function. In recent years, the study of ion channel modulation has revealed a fascinating complexity, yet we know little of how these processes operate in axons, and this area appears ripe for investigation.
One situation in which small molecular differences between apparently similar K+ channels may become practically important is when it becomes desirable to block them, for instance in the symptomatic treatment of multiple sclerosis. Some success in this direction has been achieved with 4-AP (Davis et al. 1990; Schwid et al. 1997), but its side effects can be distressing because it blocks a variety of K+ channel types. A more targeted approach, directed towards one subtype of axonal ion channel (especially if it is specific to axons), would be desirable here, and a clearer understanding of the human axonal K+ channel population is a necessary step in this direction.
We are grateful to Mr Rolfe Birch, Consultant Orthopaedic Surgeon, Royal National Orthopaedic Hospital for the human nerve graft material and to Pfizer UK Ltd for paying for its delivery by courier. This study was supported financially by the Medical Research Council, the British Council and the Deutsche Akademische Austauschdienst.