Components of action potential repolarization in cerebellar parallel fibres

Abstract

Repolarization of the presynaptic action potential is essential for transmitter release, excitability and energy expenditure. Little is known about repolarization in thin, unmyelinated axons forming en passant synapses, which represent the most common type of axons in the mammalian brain's grey matter. We used rat cerebellar parallel fibres, an example of typical grey matter axons, to investigate the effects of K⁺ channel blockers on repolarization. We show that repolarization is composed of a fast tetraethylammonium (TEA)-sensitive component, determining the width and amplitude of the spike, and a slow margatoxin (MgTX)-sensitive depolarized after-potential (DAP). These two components could be recorded at the granule cell soma as antidromic action potentials and from the axons with a newly developed miniaturized grease-gap method. A considerable proportion of fast repolarization remained in the presence of TEA, MgTX, or both. This residual was abolished by the addition of quinine. The importance of proper control of fast repolarization was demonstrated by somatic recordings of antidromic action potentials. In these experiments, the relatively broad K⁺ channel blocker 4-aminopyridine reduced the fast repolarization, resulting in bursts of action potentials forming on top of the DAP. We conclude that repolarization of the action potential in parallel fibres is supported by at least three groups of K⁺ channels. Differences in their temporal profiles allow relatively independent control of the spike and the DAP, whereas overlap of their temporal profiles provides robust control of axonal bursting properties.

Introduction

Most synapses in the mammalian brain's grey matter are formed en passant by small varicosities, also called boutons, located on very thin unmyelinated axons. Despite the abundance of such grey matter axons, very little is known about their action potentials (spikes). Typical grey matter axon morphology, particularly a diameter of <0.2 μm, gives an extremely large surface-to-volume ratio, high axial resistance, high intra-axonal noise levels, and a relatively sparse distribution of mitochondria (Palay & Chan-Palay, 1974; Shepherd & Harris, 1998). Adding to these biophysical challenges is the complicated combination of propagating spikes and at the same time controlling transmitter release along most of their path. Because there are few examples of electrical recordings from grey matter axons, we do not know how these functions are solved, but a large number of channel types are thought to participate (Meir et al. 1999; Debanne et al. 2011).

For example, cerebellar granule cells, which give rise to parallel fibres (PFs), may have more than 26 K⁺ channel α-subunits, including six delayed rectifier types (Mathie et al. 2003), and axonal and dendritic compartments of neuropil in hippocampus have mRNA for 33 K⁺ channel subunits (Cajigas et al. 2012). We will focus on functions of voltage-sensitive K⁺ channels (Kvs) because they play an essential role in excitability, transmitter release and bursting, and study PFs because they have dimensions and a transmitter type typical of many cortical grey matter axons (Ramón y Cajal, 1911; Braitenberg & Schüz, 1998).

Previous investigations of PF spikes focused on their fast repolarization. Voltage-sensitive dye recordings showed that TEA-sensitive Kvs (Sabatini & Regehr, 1997), more specifically Kv3.1 and Kv3.3 (Matsukawa et al. 2003), contributed to fast repolarization so that the blocking or lack of these channels increase spike width, Ca²⁺ influx and transmitter release. Interestingly, significant fast repolarization remained in the Kv3.1 and Kv3.3 double knockout mouse and during pharmacological block of Kv3 channels. These authors also pointed out a depolarizing after-potential (DAP), but suggested it was caused by increased extracellular K⁺. By contrast, data in Palani et al. (2012) and in the present article show that the spikes in PFs have a DAP independent of activity in neighbouring axons.

DAPs have long been expected in PFs because of the hyperexcitability that occurs after individual spikes in PFs (Gardner-Medwin, 1972), and may represent a general phenomenon of thin, unmyelinated axons because such hyperexcitability is found in other grey matter axons in both the central (Gardner-Medwin, 1972; Wigström & Gustafsson, 1981) and peripheral (Bostock et al. 2003) nervous systems.

The DAP is interesting because it may help or create bursting. Some myelinated axons generate abnormal bursts unless the proper Kvs are active (Dodson et al. 2003; Ishikawa et al. 2003; Amir et al. 2005) and some help natural bursters induce their bursts (Mathy et al. 2009; Kole, 2011). The DAP and thereby the burst can be modulated by activity (Breton & Stuart, 2009) and glutamate (Park et al. 2010) and may therefore play a role in plasticity and homeostasis.

Methods

Ethical approval

Experiments were conducted according to the National Institutes of Health Guidelines for Animal Research. All animals were treated in accordance with the regulations of the Institutional Animal Care and Use Committee at Emory University, GA, USA.

Slice preparation

Adult Wistar rats of both sexes (3–12 weeks old, n = 111) were used in this study. Rats were exposed to isoflurane and decapitated after they lost consciousness and pain reflexes. The brain was then quickly removed and submerged in ice-cold artificial cerebrospinal fluid (ACSF). The cerebellum was dissected free and cut along the folia, perpendicular to their surface, to obtain long uncut PFs, into 300 μm slices with a VF-200 (Precisionary Instruments, Inc., San Jose, CA, USA) or, in a few cases, a Leica VT1200 (Leica Microsystems GmbH, Wetzlar, Germany) in ice-cold solution bubbled with 5% carbogen. The ACSF used for dissection was composed of glycerol 250 mm, KCl 1.25 mm, KH₂PO₄ 1.25 mm, MgCl₂ 7 mm, NaHCO₃ 25 mm, CaCl₂ 0.5 mm and glucose 16 mm. The slices were stored in an oxygenated chamber for at least 1 h at room temperature before recording. The ionic composition of ACSF used for storage and recording was NaCl 125 mm, KCl 1.25 mm, KH₂PO₄ 1.25 mm, MgCl₂ 1 mm, NaHCO₃ 25 mm, CaCl₂ 2 mm and glucose 16 mm. When Cd²⁺ was used, the extracellular solution consisted of NaCl 150 mm, KCl 2.5 mm, CaCl₂ 2 mm, MgCl₂ 1 mm, glucose 16 mm and HEPES 10 mm. All extracellular solutions were adjusted to pH 7.4. During recording the slices were kept submerged in a 3 ml tissue chamber with oxygenated ACSF flowing at 4.0–7.5 ml min⁻¹. The temperature was controlled by a temperature controller (MTC-20/2SD; npi electronic GmbH, Tamn, Germany) and was kept at 34–36°C or 22–24°C, during grease-gap and intracellular recordings, respectively. Three intracellular recordings were made at 34°C.

Drugs

The study drugs 6,7-dinitroquinoxaline-2,3-dione (DNQX, 30 μm) and dl-amino-5-phosphonovaleric acid (APV, 50 μm), were purchased from Tocris Bioscience (Bristol, UK). Picrotoxin (25 μm), 4-aminopyridine (4-AP), tetraethylammonium (TEA), quinine, margatoxin (MgTX), α-dendrotoxin (DTX) and salts were purchased from Sigma-Aldrich Corp. (St Louis, MO, USA).

Grease-gap recordings

The recording procedure was performed as previously described (Palani et al. 2012). Briefly, presynaptic action potentials were isolated chemically from synaptic activity by adding DNQX (30 μm), APV (50 μm) and picrotoxin (25 μm) to ACSF. Action potentials were elicited with short (50–150 μs) electrical pulses from an ISO-stim 01MD (npi electronic GmbH), delivered through a monopolar glass pipette with a 50 μm tip opening.

The grease-gap pipette was made from pipette tips with a wall thickness of 0.8 mm, (usually Gilson DIAMOND D200, 200 μl; Gilson, Inc., Middleton, WI, USA) and cut to an outer diameter of 4 mm. An arch, 0.8 mm high and 1.5 mm wide, was carved out at the bottom (widest end) of the pipette wall. Grease (petroleum jelly, Vaseline; Chesebrough-Pond's, Inc., Englewood Cliffs, NJ, USA) was applied to fill in the arch and along the remaining cut surface of the pipette. The bottom of the plastic recording chamber was covered with Sylgard 184 (Dow Corning Corp., Midland, MI, USA). By positioning the arched window over a bundle of axons, a monophasic signal with distinct fast [compound action potential (cAP)] and slow [compound depolarized after-potential (cDAP)] components could be recorded. Signals were amplified 10× at the head-stage and 100× at the DPA-2FL amplifier (npi electronic GmbH), low pass-filtered at 1 kHz, digitized at 10 kHz with an NI USB 6343 device (National Instruments, Inc., Austin, TX, USA), and stored on computer hard disk.

The amplitude of the initial fast part (cAP) was measured as the difference between the first positive peak after the stimulus and the baseline before stimulus. The amplitude of the early part of the depolarized after-potential (cDAP₁) was measured as a mean over 5 ms starting at the transition between the fast and slow repolarization. The amplitude of the late part of the depolarized after-potential (cDAP₂) was measured as a mean over 50 ms, starting just after cDAP₁. Changes in the parameters were measured relative to their average over a 3–12 min control period.

Whole-cell patch clamp recordings

Granule cells were recorded using the blind approach (Blanton et al. 1989). Pipettes were pulled from borosilicate glass (OD 1.5 mm, internal diameter 0.86 mm; BioMedical Instruments, Zöllnitz, Germany) and had a tip resistance of 5–8 MΩ. The intracellular solution contained K-gluconate 120 mm, KCl 20 mm, MgCl₂ 2 mm, Na₂ATP 2 mm and HEPES 10 mm, at pH 7.2. Intracellular patch clamp recordings were amplified by an SEC-10LX (npi electronic GmbH), voltage clamped in linear mode and current clamped in switching mode. The electrode's access resistance, of 6–20 MΩ, was not compensated in voltage clamp experiments. Axons were electrically activated by a 10–20 μm-wide monopolar glass electrode.

The somatic spike threshold was determined by injecting square steps of depolarizing current, 200–500 ms long, through the somatic electrode. The current step was increased until it gave a potential that triggered spikes; this potential was defined as the somatic threshold.

Spike widths were measured at 70% of their peak amplitude, from intracellular somatic recordings at 27–34°C and from grease-gap recordings at 34–36°C.

Statistics

Measurements are given as the mean ± s.e.m. unless otherwise specified. Differences between three or more groups of means were tested with ANOVA, with a post hoc Tukey's test when relevant. To determine significance in experiments in which baseline was normalized to 1.0, we compared the deviation from 1.0 using a one-sample Wilcoxon's signed-rank test. Effects between two groups were compared using a related-samples Wilcoxon signed-rank test or Mann–Whitney U test for paired and unpaired data, respectively. Statistical analyses were performed using IBM spss Statistics for Windows Version 20.0 (IBM Corp., Armonk, NY, USA). P-values of <0.05 were considered to indicate statistical significance.

Results

4-AP induces axonal bursts primarily by increasing DAP amplitude

To identify processes that control excitability and bursting behaviour of PFs, we first recorded antidromic action potentials from granule cell somata in cerebellar slices from rats aged 20–31 days while activating their axons (PFs) (Fig.1A). Fast synaptic transmission was blocked in all experiments. The cells were identified as granule cells by their location in stratum granulosum, the fact that they could be activated by stimulating the stratum moleculare and their high resistance to somatic current injection (mean ± s.d. 1.43 ± 0.67 GΩ), which is similar to findings reported by Diwakar et al. (2009).

Figure 1. Axonal bursts detected at the granule cell soma.

A, schematic representation of a granule cell with axon [parallel fibre (PF)] showing the position of stimulation (Stim, red) and intracellular recording (Rec, blue) electrodes. B, five somatically recorded action potentials in response to axonal activation, with slow depolarized after-potential (DAP). C, one of the 19 neurons that burst in response to axonal activation, showing typical small humps or spikelets intermingled with larger spikes. D, antidromic somatic spikes recorded from the soma at potentials of −67 mV (black), −71 mV (red) and −82 mV (blue). The fast part of the spike failed at −71 mV and −82 mV. The stimulus artefacts here, in D–G and in I were truncated. E, at the same stimulation strength the attenuated spike (blue in D) could flip between failure and its full amplitude (here illustrated by three traces of each). F, the threshold for the somatic spike (Somatic T) was determined by depolarizing current injection (rightmost trace). Granule cell recording during electrical activation of PFs (as in A) showed a combination of fast spikes and spikelets, interpreted as intermittent failures (arrow) of the spike to propagate to the soma because it was hyperpolarized by current injection. The middle trace occurred spontaneously without electrical activation. G, antidromic somatic spikes before (black) and during 0.25 mm 4-AP (red). 4-AP converted a single spike response to a combination of spikes and spikelets taking off from potentials (arrow) well below Somatic T. H, at somatic potentials below Somatic T the first spike could be blocked (red trace, arrow), but the second spike in the burst remained. I, membrane potentials for the somatic thresholds (determined by injecting current at the soma) which can be compared with the membrane potentials from which the second spike or spikelet took off (meaning the first spike that was not directly activated by the stimulation electrode). These ‘second spikes’ are separated into two groups (with and without a failure of the first stimulated spike, respectively) and show that spikes could be detected at the soma at potentials well below the somatic threshold. J, measurements of resting membrane potentials (mpot, blue), peak action potential (peak AP, red) and DAP (yellow), during the transition from antidromic single spikes to bursts, during wash-in of 0.25 mm 4-AP. The first three bursts are marked red. K, the most obvious change just before the first burst occurred (dotted line) was an increase in DAP potential (yellow). Inset: after 5–10 min of 4-AP wash-in (black horizontal bar above the plot), the spikes in the burst appeared on top of a large depolarizing wave. L, the change of spike peak amplitude, DAP and membrane potential from control (before 4-AP) to the last response before the first burst. Spike peak amplitude and DAP increased significantly. **P < 0.01; ns, not significant.

Most cells responded with a single antidromic spike with a slowly decaying DAP (Fig.1B), but some showed bursts (19 of 61 neurons). The bursts were composed of discrete spike-like events usually intermingled with smaller humps, riding on a depolarizing wave (Fig.1C). We hypothesized that the bursts were triggered in the axon by a DAP, and that the smaller humps represented spikes that failed to propagate all the way to the soma. We found this to be the most likely explanation because the spike could be blocked by hyperpolarizing current injection at the soma, often with several intermediate spike amplitudes (red and blue traces from different pre-stimulus potentials in Fig.1D), as expected by antidromic spikes blocked at different distances from the soma. Furthermore, the depolarizing potential was present in the axon because it persisted after the soma-near spike was blocked by hyperpolarization (Fig.1D, red and blue traces). It was often difficult to abolish the remaining slow potential by hyperpolarization, but by also reducing stimulation strength at the axonal electrode, this attenuated spike could be made to flip between all or none at one threshold stimulation strength (Fig.1E), as expected by an action potential. Similar results were reported and comprehensively analysed in Palani et al. (2012), and are discussed at the end of the present article.

In six of the 19 bursting cells, we were able to block the first spike in the burst by hyperpolarizing the soma with current injection. This usually left an attenuated spike (spikelet) at the place of the full spike, and the response could at certain membrane potentials flip between spike and spikelet (Fig.1F, arrow). We interpret the spikelets as axonal spikes that failed before they arrived at the soma. Both the spikes and spikelets could take off from membrane potentials well below the somatic spike threshold and therefore could not be triggered at the soma. The somatic spike threshold (defined in Methods) was determined by depolarizing current injection through the somatic electrode (Fig.1F, extreme right trace).

To study the role of the DAP in burst generation, we used 4-aminopyridine (4-AP) in the bath because this compound increases post-spike hyperexcitability (Palani et al. 2010) and induces bursting in several preparations. 4-AP (0.25 mm) converted the single spike responses to bursts in 11 of 11 cells (Fig.1G, black and red, respectively). To distinguish between somatic and axonal burst mechanisms, we hyperpolarized the soma until some spikes were blocked or attenuated (Fig.1G, arrow). In all 11 cases, we observed attenuated spikes, probably blocked in the axon at some distance from the soma. Additionally, in some recordings (five of 11), the first, electrically activated spike failed at hyperpolarized potentials, but the later spikes in the burst remained (Fig.1H, red). This suggests that the second spike was not triggered by somatically depolarizing currents.

We measured the most negative membrane potential from which a spike, attenuated or not, took off (Fig.1I) and found that the first spike after the initial, electrically activated spike could take off from a potential 11.8 ± 0.7 mV (n = 6; P = 0.03) more negative than the somatic threshold, and from a potential 15.8 ± 1.5 mV (n = 5; P = 0.04) more negative if the first, electrically activated spike was blocked. These findings support the hypothesis that bursts were generated in the axon, although we cannot exclude the possibility that some (but not all) spikes in the burst were triggered at locations near the soma.

While 4-AP washed into the tissue we followed the membrane potential, the voltage at the peak of the spike, and the voltage at which spike repolarization changed from fast to slow repolarizing potential (Fig.1J). The transition to bursting was followed every 20 s and was characterized by very little change in action potential shape (Fig.1J, black traces) until the first burst occurred [Fig.1J (red traces) and K (dotted line)].

The most obvious change, measured at the last trace before the first burst occurred, was in the early part of the DAP (Fig.1K). This could be confirmed by measurements of voltage changes in all 11 cells (Fig.1L), where the peak of the spike increased by 3.17 ± 0.96 mV (P = 0.007), DAP increased by 4.20 ± 0.92 mV (P = 0.004), and the membrane potential did not change significantly (by – 0.15 ± 0.66 mV; P = 0.50). After the first burst occurred, DAP continued to increase to a large depolarizing wave with spikes riding on top of it (inset in Fig.1K).

These observations suggest that the membrane potential of DAP controls the burst behaviour. Therefore, we wanted to investigate pharmacologically the control of the DAP amplitude. However, the shape and properties of the axonal spikes are difficult to interpret from somatic recordings because of passive and possibly active filtering by the highly specialized and channel-rich axonal initial segment. For this reason, we explored the control of DAP amplitude and other spike parameters using an axonal recording method.

4-AP changed the grease-gap recorded DAP similarly to the intrasomatic

We utilized a grease-gap technique to record changes in axonal spike shape in populations of PFs several millimetres away from the soma (Fig.2A). In all experiments fast synaptic transmission was blocked (see Methods). Several lines of evidence (Palani et al. 2012; Discussion herein) showed that the grease-gap signal is formed by axonal action potentials with slow DAPs.

Figure 2. Grease-gap recordings showed changes similar to those in intrasomatic recordings.

A, schematic representation of a cerebellar slice with PFs showing the position of stimulation (Stim, red) and grease-gap recording (Rec, blue) electrodes. B, mean ± s.e.m. (shaded area) of the signals from 13 grease-gap experiments (black) and five granule cell somatic recordings with antidromic spikes (blue). The signals were temporally aligned to the peak of the spike before averaging. The grease-gap experiments were additionally normalized to peak cAP amplitude before averaging, and the average was scaled to overlap the transition point between fast and slow repolarization (arrow) in the intracellular recordings. Note the similar time courses of slow decays in the intracellular and grease-gap recordings. The stimulus artefacts were truncated. C, single traces of the grease-gap signal in response to electrical activation of PFs before (black) and during (blue) 0.05 mm 4-AP, and after wash-out (red). D, a higher concentration of 4-AP (0.3 mm) gave larger effects on cAP and cDAP and added a hump on cDAP. Insets in B, C and D represent signals at a higher time resolution. E, the amplitudes of cAP and cDAP (marked with an arrow and horizontal bar in C) followed during wash-in and wash-out of 0.05 mm 4-AP. F, the same parameters as in E followed during wash-in and wash-out of 0.3 mm 4-AP. G, the amplitudes of cAP and cDAP at different concentrations of 4-AP show that the effect saturated around 0.3 mm, and that cAP increased less than cDAP at concentrations of ≥0.3 mm. Note the logarithmic scale on the vertical axis. H, at high 4-AP concentrations (≥0.6 mm) the increases of cAP and cDAP were often transient. Inset with grey curves represents averages of traces during periods indicated by grey horizontal bars. I, initial increases of cAP (left) and cDAP (right) followed the same pattern as the maximal effects displayed in G, with saturation of the effects at 4-AP concentrations of ≥0.3 mm and larger increases, at the highest concentrations, of cDAP compared with cAP. J, mean ± s.e.m. (shaded) waveform of nine grease-gap experiments before (black line) and during (red line) 0.3 mm 4-AP, normalized to peak cAP amplitude during the control period. To allow for comparisons with the intrasomatic waveform during 4-AP-induced bursting (blue), we made averages from nine somatic recordings temporally shifted to align the peak of their electrically triggered spike. The averaging procedure smoothened out the intrasomatic spikes during the burst because of their variable latencies. Qualitative similarities between intrasomatic and grease-gap recordings refer to increases in amplitude in response to 4-AP of the fast depolarization (spike and cAP) as well as the slow depolarizing component (DAP and cDAP).

The grease-gap recordings showed a cAP with a shape similar to that of the somatically recorded antidromic spikes (Fig.2B). Relative to their depolarized after-potentials (scaled to the same amplitude in Fig.2B), the peak amplitude of the grease-gap signal was smaller than the intracellular spike and tended to be broader (intracellular 0.77 ± 0.18 ms, n = 7; grease-gap 1.70 ± 0.28 ms, n = 10; P < 0.01), possibly as a result of the summation of imperfectly synchronized action potentials and attenuation of the high-frequency components across the grease-gap barrier. Despite the likely distortion of the shape of the spikes, the grease-gap recordings give the opportunity to distinguish between changes in axonal spike potentials during the first few milliseconds of the cAP and the much longer-lasting cDAP.

The amplitudes of both cAP and cDAP increased when 0.05 mm or 0.3 mm 4-AP was added to the extracellular solution (Fig.2C and D), although the higher concentrations increased cDAP more than cAP and often added a hump on the slow decaying phase (Fig.2D). These effects were reversible with washout of 4-AP (Fig.2E and F).

A range of 4-AP concentrations (0.05–1.0 mm) showed differential effects on cAP and cDAP. An ANOVA with five groups of concentrations, individually tested for cAP (P = 0.003) and cDAP (P = 0.002; n = 37 experiments), showed post hoc tests with P > 0.5 at concentrations of ≥0.3 mm (Fig.2G), suggesting that 0.3 mm 4-AP was close to maximal blocking effect. At these high concentrations cDAP increased more than cAP. This was tested by combining the amplitude changes from 0.3 mm, 0.6 mm and 1.0 mm 4-AP. cAP increased by 140.0 ± 19.8% and cDAP by 384.2 ± 69.7% (n = 19; P = 0.009).

At concentrations of ≥0.6 mm the effects on cAP and cDAP usually declined spontaneously (Fig.2H). This may have reflected the depolarization of the axons, or the result of intrinsic homeostatic mechanisms. If the decline of cAP and cDAP amplitudes started before full effect was reached, the 4-AP effects may have been underestimated, particularly for the higher concentrations. We therefore measured the initial trajectory of the amplitude changes (Fig.2I), which showed the same pattern as in Fig.2G, with a saturation of the effects at 0.3 mm 4-AP, and a larger increase of the cDAP than the cAP at high concentrations.

The larger effects of 4-AP at higher concentrations on cDAP compared with cAP suggests that different channels with different 4-AP affinities were involved in the control of cAP and cDAP amplitudes, respectively. Alternatively, the DAP was amplified by non-linear processes involving, for example, voltage-sensitive Ca²⁺ or Na⁺ channels.

If the action potential shape and 4-AP-induced changes that occurred close to the soma (as recorded from the soma with antidromic action potentials) were the same as those further out in the axon (as recorded by the grease-gap), similarities between the average shapes from those recordings might be expected (Fig.2J). The fast spike components increased with 4-AP in both the intracellular and grease-gap recordings, although mostly in the latter, which may reflect the more efficient summation of possibly broader spikes in the grease-gap recording. The slow components (DAP and cDAP) also increased in both recordings, although mostly in the intrasomatic recording. This suggests higher 4-AP sensitivity closer to the soma than in the more distal axon and/or a higher density of channels that boost the DAP, such as voltage-sensitive Na⁺ or Ca²⁺ channels.

K⁺-induced depolarization did not increase cAP amplitude

As the amplitudes of both cAP and cDAP increased in response to 4-AP, it is possible that axons depolarized and thereby reduced the threshold for electrical activation, resulting in the activation of more axons. Although 4-AP did not change resting membrane potential at the granule cells somata (Fig.1), axons could theoretically have specific mechanisms to maintain resting membrane potential. To distinguish between depolarization and other amplitude-increasing effects, we manipulated the axonal resting membrane potentials by using different concentrations of K⁺ in the bath solution.

As expected with moderate depolarization of the axonal membrane, the latency to the cAP peak decreased during 4 mm and 6 mm K⁺ [by 4.6 ± 1.8% (P = 0.075) and 16.1 ± 7.7% (P = 0.046), respectively, in six experiments] (Fig.3). At these K⁺ concentrations there was no increase in amplitude, but rather a decrease in most experiments, like that displayed in Fig.3C. On average in six experiments (Fig.3D), the cAP amplitude decreased by 2.3 ± 3.3% (P = 0.46) and 10.3 ± 5.0% (P = 0.12) at 4 mm and 6 mm K⁺, respectively. At 10 mm K⁺ the latency first decreased, similarly to effects with lower K⁺ concentrations, but then increased (Fig.3C and D, left) as is expected if the membrane potential is too depolarized to reactivate Na⁺ channels (Kocsis et al. 1983). While this happened the amplitude dropped dramatically (Fig.3C and D, right).

Figure 3. Effects of different extracellular K⁺ concentrations.

A, 4 mM K⁺ (blue) reduced the latency to the peak of cAP but had little effect on its amplitude compared with control at 2.5 mm K⁺ (black). B, 6 mM K⁺ (red) reduced cAP latency and reduced the amplitudes of cAP and cDAP more than 4 mm K⁺. C, cAP latency and amplitude followed during wash-in and wash-out of different K⁺ concentrations: 4 mm (blue) and 6 mm (red) K⁺ reduced latency and cAP amplitude. The small, initial decrease of latency observed during the 10 mm K⁺ (grey) application was followed by an increase, as expected with depolarization large enough to inactivate voltage-sensitive Na⁺ channels. D, on average in six experiments latency to cAP decreased during 4 mm and 6 mm K⁺ as expected with moderate depolarization of membrane potential, whereas cAP amplitude declined only slightly. cAP amplitude dropped towards zero during 10 mm K⁺.

These findings show that the amplitude increases observed with 4-AP were probably not caused by threshold reduction because of depolarization. The more likely alternative is that the cAP amplitude increased as a result of the reduction in fast repolarizing K⁺ currents, allowing the action potentials to reach a higher amplitude, as in the intrasomatic recordings, and possibly summate more because of broadening.

TEA influenced cAP

To test if cAP and cDAP were controlled by separate processes, as suggested by the 4-AP experiments, we utilized more specific K⁺ channel blockers. We also wanted to quantify the changes of the time course of the after-potential. Therefore, we divided the cDAP estimator in two, cDAP₁ and cDAP₂, to measure the early and late parts of the cDAP (Fig.4A) (see also Methods). TEA (1 mm) increased the amplitude of cAP much more than that of cDAP₁ and cDAP₂ (Fig.4A). On average in 17 experiments (Fig.4B), cAP increased by 29.3 ± 5.4% (P < 0.001), whereas cDAP₁ and cDAP₂ increased by 10.2 ± 4.7% (P = 0.076) and 14.6 ± 4.7% (P < 0.01), respectively. The width of cAP, measured at 70% of cAP control amplitude, increased from 1.69 ± 0.28 ms to 2.09 ± 0.28 ms (i.e. by 24%) after 1 mm TEA (n = 10; P = 0.005). Latency to cAP did not change significantly (from 3.56 ± 0.32 ms to 3.53 ± 0.33 ms, n = 16; P = 0.92, data not shown), suggesting that TEA did not systematically change the excitability of the axons. Therefore, it is unlikely that the amplitude increases were caused by the activation of more fibres as a result of a lower activation threshold. There was no evidence of change in DAP shape because cDAP₁ and cDAP₂ increased similarly (n = 17; P = 0.28) and therefore it may be that only fast repolarization was reduced, stopping at a more positive membrane potential when TEA-sensitive channels were blocked, whereas the slow decay followed the same time course as it did without TEA.

Figure 4. Effects of tetraethylammonium (TEA).

A, the grease-gap signal from one experiment before (black) and during (red) 1 mm TEA. Each curve is the average of nine individual traces. The inset in the left panel represents traces at a higher time resolution, for better visualization of changes in cAP amplitude. Right panel: TEA increased cAP (blue) amplitude more than cDAP₁ (black) and cDAP₂ (red) amplitudes. B, in 17 experiments the average amplitude of cAP increased more than those of cDAP₁ and cDAP₂ in response to 1 mm TEA. C, the parameters expressed as a fraction of control cAP amplitude (by normalizing to the peak of control cAP, black) showed that the amplitude of cAP changed much more than cDAP₁ and cDAP₂ in response to 1 mm TEA (red, mean ± s.e.m. of 13 experiments). D, a comparison of the effects of 1 mm and 5 mm TEA shows that most of the amplitude-increasing effect of TEA was present at 1 mm. E, the grease-gap signal from one experiment shows a train of three stimuli at 100 Hz in control conditions (black) and after 1 mm TEA (red), illustrating that TEA did not change the frequency-following ability at this moderate frequency.

Although the comparison with the control period allowed for the sensitive detection of the drug effects on individual parameters (Fig.4B), a better way to quantify the relative changes between the parameters is to normalize to the pre-drug period of the cAP (Fig.4C) and estimate the changes relative to this value. The cDAP₁ was 49.2 ± 3.6% of the normalized cAP in control conditions, and increased to 54.3 ± 3.8% with TEA (n = 13; P = 0.016). This 5% change is small compared with the 29.7 ± 5.3% increase in cAP reported above. Similarly, cDAP₂ increased by only 3% (from 26.7 ± 2.4% to 29.7 ± 2.6%, n = 13; P = 0.007). We conclude that TEA had its main effect during the first few milliseconds of the action potential, with a small but significant amplitude-increasing effect on cDAP₁ and cDAP₂.

Increasing the TEA concentration to 5 mm did not add much amplitude to cAP or cDAP (Fig.4D). In seven experiments in which we compared the effects of 1 mm and 5 mm TEA, cAP increased by 20.5 ± 0.05% with 1 mm TEA and 29.4 ± 0.08% with 5 mmTEA. Similarly, 1 mm TEA increased cDAP₁ and cDAP₂ by 8.3 ± 2.2% and 8.5 ± 3.3%, respectively, and 5 mm TEA increased them by 9.0 ± 3.0% and 14.2 ± 5.0%, respectively.

Two observations suggested that TEA-insensitive channels contributed significantly to spike repolarization when TEA-sensitive channels were blocked. First, the main part of the fast repolarization was intact when TEA-sensitive channels were blocked (Fig.4C). Secondly, the ability to fire three spikes over a 30 ms period, a common event in vivo (Eccles et al. 1966; Chadderton et al. 2004; Jorntell & Ekerot, 2006), was not changed by 1 mm TEA (Fig.4E). We quantified the ability to follow this train of stimuli as the ratio between the last and first cAP amplitudes in control conditions (n = 14) and in the presence of TEA (n = 7) in recording solution. This ratio did not change significantly [from 0.91 ± 0.04 to 0.99 ± 0.05; P = 0.28 (summarized later in Fig.7C)]. We therefore conclude that the ability to follow 100 Hz for 30 ms was not influenced by 1 mm TEA.

Figure 7. Summary of effects of Kv channel blockers on fast and slow components of repolarization in parallel fibres.

A, summary of averaged shapes (those in Figs4C, 5D and 6E) showing that TEA mostly increased fast repolarization (red) compared with control (black). The maximal TEA effect on cAP was similar when combined with margatoxin (MgTX) and with quinine, but because the amplitude of all components of the recorded signal decreased over time when MgTX, quinine and TEA were combined (Fig.6C) the maximal TEA effect on cAP is marked by a red open circle (rightmost panel), and the red trace is taken when the cDAP₁/cAP ratio was close to 1.0. B, summary of the relationships between cAP and cDAP₁ (upper panel), and cDAP₂ (lower panel). Mean ± s.e.m. relative changes in response to TEA 1 mm, MgTx and the combinations of MgTx + TEA and MgTX + quinine + TEA are presented in detail in Figs4 and 5. C, TEA, MgTx or quinine applied alone did not change the axon's ability to follow three stimuli with 10 ms intervals, measured as the ratio of the amplitudes of the third and first cAP. These data were presented in detail earlier in this article.

MgTX influenced cDAP

Margatoxin (MgTX), a peptide with high affinity for Kv1.3, but also other Kv1-family channels (Garcia-Calvo et al. 1993), was tested because immunostaining studies have shown strong labelling for Kv1.3 in PFs (Vacher et al. 2008). By contrast with TEA, MgTX (10 nm) had little or no effect on cAP but increased the amplitude and changed the time course of cDAP (Fig.5A). The average changes of 10 experiments, relative to the control period, showed that cDAP₁ increased by 27.5 ± 6.7%, whereas cDAP₂ increased much more, by 74.3 ± 12.9% (P = 0.008, comparing cDAP₁ and cDAP₂), meaning that there was a significant change in the shape of the after-potential. cAP changed only by 10.7 ± 5.5%, not significantly compared with the pre-drug period (P = 0.11) (Fig.5, column B). Similarly to TEA, MgTX alone did not affect the ability to follow 100 Hz for 30 ms. The ratio of CAP3/CAP1 during MgTx block was not significantly different from control [changed from 0.91 ± 0.04 (n = 14) to 0.93 ± 0.09 (n = 7); P = 0.55 (summarized later in Fig.7C)].

Figure 5. Effects of margatoxin (MgTX).

A, single traces of the grease-gap signal in response to electrical activation of PFs before (black) and during (red and blue) 10 nm MgTX. The amplitude of cAP changed little, whereas that of cDAP₂ increased considerably during 30 min (blue). B, mean data across 10 experiments show that MgTX had little effect on cAP amplitude (blue), but increased cDAP₁ and cDAP₂ (black and red, respectively). The bottom panel shows the mean ± s.e.m. (shaded area) shape of the recorded potentials during control (black) and the last 3 min with MgTX (red). The stimulus artefacts were truncated. C, the introduction of 10 nm MgTX when 1 mm TEA was already present in the bath showed that cDAP₁ and cDAP₂ were still increasing after 25 min of MgTX treatment. The average shapes before and during MgTX (bottom panel) show changes similar to those with MgTX alone. D, the introduction of 1 mm TEA when slices had been incubated in the holding chamber with 10 nm MgTX resulted in an increase of cAP similar to that with TEA alone (Fig.4B). However, cDAP₁ and cDAP₂ increased faster and to a greater extent than in the experiments in B and C, as is also apparent in the average shapes (bottom panel; the stimulus artefacts were truncated). E, summary of amplitude increases of cAP, cDAP₁ and cDAP₂ with different combinations of blockers. (*P < 0.05; **P < 0.01; ***P ≤ 0.001; ns, not significant.) F, when cDAP₂ was first increased by the combination of MgTX (10 nm) and TEA (1 mm), the amplitude could be reduced by the addition of 0.05 mm Cd²⁺ to the bath. (*P < 0.05.) G, combining TEA (1 mm) with MgTX (10 nm) (n = 2) or MgTx (10 nm) + DTX (100 nm) (n = 2), called peptide TX in the figure, increased the amplitude of the depolarizing after-potential also when Cd²⁺ (0.1 or 0.2 mm) was present in the bath from the beginning of the experiment. The stimulus artefacts were truncated.

The waveforms of 10 experiments (Fig.5, bottom of column B) individually normalized to the peak of cAP before drug application (similarly to Fig.4C) showed that MgTX increased cDAP₁ by 10.9% relative to the control cAP amplitude (from 48.1 ± 5.3% to 59.0 ± 7.7%; P = 0.005). Because a large fraction of the fast repolarization remained both with TEA and MgTX, one possibility is that MgTX-sensitive channels accounted for the repolarization when TEA-sensitive channels were blocked, and vice versa. We tested this hypothesis by using both blockers.

TEA combined with MgTX or DTX

Blocking TEA-sensitive channels before MgTX (10 nM) was added (Fig.5C) tended to enhance the effect of MgTX on cDAP amplitude, whereas the effect on cAP was small and indistinguishable from the effect of MgTX alone. The amplitude of cDAP continued to increase even 30 min after the addition of MgTX, suggesting that the full effect was not reached (Fig.5C, red and black data points). However, when the axons were exposed to both TEA and MgTX over a long time, the signal spontaneously declined in some experiments, increasing the variance between experiments and preventing a conclusion. To avoid these problems and to maximize the MgTX block, we incubated slices with MgTX in the holding chamber for 1–6 h before transferring them to the recording chamber. The recording chamber did not contain MgTX, but based on the large effects described below, and support in the literature for the very slow wash-out of MgTX (Ishikawa et al. 2003), we assume that MgTX-sensitive channels were blocked during these experiments (20–30 min).

Under those conditions the addition of TEA to the solution gave a fast and large increase of cDAP₁ and cDAP₂ relative to their baseline values, shown by the average of seven experiments in Fig.5D. cDAP₁ increased by 75.4 ± 24.4% and cDAP₂ rose by 168.0 ± 45.0%, which differ from responses to TEA alone (P = 0.001 for both parameters). All P-values refer to ANOVA tests (one for each parameter) or post hoc tests if the ANOVA was significant (summarized in Fig.5E). The most likely explanation for the large TEA effects on cDAP₁ and cDAP₂ is that MgTX-sensitive channels normally upheld most of the fast repolarization when TEA-sensitive channels were blocked, but were now prevented from contributing because they were blocked at the beginning of the experiment. The amplitude of cAP increased by 42.2 ± 7.8%, which is not significantly different from the increase with TEA alone (P = 0.44). This supports the hypothesis that cAP in these experiments increased as a result of the block of TEA-sensitive channels, and that MgTX-sensitive channels did not contribute much to cAP amplitude.

We also tested dendrotoxin (DTX + TEA) in a way similar to the tests of MgTX + TEA described above. Although 100 nm DTX alone did not have detectable effects (data not shown), possibly as a result of very slow diffusion, incubation with DTX and the subsequent application of 1 mm TEA in six experiments elicited large increases of cDAP₁ (of 60.5 ± 8.5%; P = 0.025 in comparison with TEA alone) and cDAP₂ (of 98.0 ± 42.0%; P = 0.139 in comparison with TEA alone). cAP increased by 41.7 ± 4.5%, which did not reflect any significant difference with the effect of TEA alone (P = 0.44).

Incubation with DTX (100 nm) together with MgTX (10 nm) and subsequent application of TEA (1 mm, n = 4) increased cDAP₁ by 73.1 ± 17.8% and cDAP₂ by 186.2 ± 59.5%. These changes were significantly different from responses to TEA alone (P = 0.015 for cDAP₁; P = 0.003 for cDAP₂). cAP increased by 40.1 ± 14.7%, which did not reflect any significant difference with the effect of TEA alone (P = 0.44).

None of the changes induced in cDAP₁ or cDAP₂ by combining TEA with MgTX, DTX or MgTX + DTX differed significantly from one another (ANOVA, all post hoc tests, P > 0.30), which suggests that MgTX and DTX blocked the same set of channels. These results are summarized in Fig.5E.

All of the combinations of peptide blockers (MgTX and DTX) and TEA resulted in a hump on the slow depolarization, similar to that pointed out under the description of the effect of MgTX alone. When the hump was present the cDAP was sensitive to Cd²⁺, a broad Ca²⁺ channel blocker, and may have been a Ca²⁺ current. On average in six experiments, cDAP₂ increased to 321.4% ± 70.9% (where a value of 100% represents no change) in response to 10 nm MgTX and 1 mm TEA, but decreased to 201.9% ± 20.1% of the control value after the addition of 50 μm Cd²⁺ in the bath solution (P = 0.028) (Fig.5F). Importantly, the effects of the drugs on cAP and cDAP were still apparent in the presence of Cd²⁺ (0.05–0.2 mm) (Fig.5F and G), confirming that these effects were not only the result of Ca²⁺ currents or changes in synaptic transmission.

Quinine-sensitive channels can contribute to fast repolarization

Even with MgTX, TEA and in some cases DTX in the bath, significant fast repolarization always remained. To focus on this remaining fast repolarization, we measured cDAP₁ as a fraction of cAP. If cDAP₁/cAP is 1.0, there will be no fast repolarization and the signal will return to baseline following the slow time course of the cDAP. Figure6A shows that the combination of TEA and MgTX (same data as in Fig.5D) did not change the cDAP₁/cAP fraction much (from 0.39 ± 0.06 to 0.49 ± 0.08, n = 7; P = 0.176) and thus most of the fast repolarization remained.

Figure 6. Effects of quinine (0.1 mM) on slices incubated with margatoxin (MgTX) (10 nM).

A, most of the fast repolarization remained even when 1 mm TEA was added to slices incubated with MgTX, as shown by only modest changes in ratio between cDAP₁ and cAP (n = 7). B, quinine (yellow horizontal bar) applied to PFs incubated with MgTX did not change cAP, cDAP₁ or cDAP₂. The shape of the grease-gap signal is the average of nine traces before (black) and during quinine (red) from a single experiment (left). The parameters were followed over 25 min and displayed as the average of five experiments (right). C, the addition of quinine (yellow horizontal bar) to slices incubated with MgTX did not change the cDAP₁/cAP ratio, cAP, cDAP₁ or cDAP₂, but the subsequent addition of TEA (black horizontal bar) increased all of these parameters, and most importantly brought the cDAP₁/cAP ratio close to 1.0, meaning fast repolarization was abolished. The amplitudes of cAP, cDAP₁ and cDAP₂ decreased a few minutes after TEA was introduced, probably as a result of membrane depolarization, but was largely restored in solution without Kv blockers. D, representative traces from the period with only MgTX, MgTX + quinine + TEA, and washout, from the experiment used in C. Note the smaller cAP peak in the middle trace, probably caused by membrane depolarization, but at the same time a greatly reduced fast repolarization that was never observed in experiments with K⁺-induced depolarization as in Fig.3. E, the mean ± s.e.m. (shaded area) shape of grease-gap recorded potential from five experiments in which only MgTX-sensitive channels were blocked (black) and after subsequent addition of quinine and TEA (red). The individual experiments were normalized to the peak cAP before quinine and TEA were added. The stimulus artefacts were truncated. F, when 1 mm TEA was added after quinine (MgTX-incubated slices) the fast repolarization was almost completely blocked as shown by the cDAP₁/cAP ratio approaching 1.0 (n = 5).

As TEA and MgTX preferentially block Kv3 and Kv1 channels, respectively, we used quinine to reduce the possible contribution from Kv2 channels (Schmalz et al. 1998; Johnston et al. 2008). Quinine blocks other delayed rectifier channels and tandem-pore domain K⁺ channels (Kim, 2005) and hence the main aim was to abolish the remaining fast repolarization, not to determine the genetic identity of the channels.

Quinine (0.1 mm) applied to slices incubated with MgTX (10 nm) in five experiments did not have detectable effects (average changes in cAP, cDAP₁ and cDAP₂ were all <3.4%; P > 0.35) measured as the difference over 17 min (Fig.6B). Quinine alone (without MgTX) did not reduce the ability to fire three spikes over 30 ms either [the CAP3/CAP1 ratio changed from 0.91 ± 0.04 (n = 14) to 0.95 ± 0.11 (n = 4); P = 0.92 (summarized later in Fig.7C)].

However, when TEA was applied to slices incubated with MgTX for >2 h and exposed to quinine for >17 min, the amplitude of cDAP₁ increased dramatically to an amplitude similar to that of cAP (Fig.6C–F). After TEA was added, cAP, cDAP₁ and cDAP₂ increased initially, but started to decline after 4–5 min, most likely as a result of the depolarization of resting membrane potential. Washing out with solution without blockers restored the signal (Fig.6C and D), which suggests that the blockers did not cause permanent damage to mechanisms that maintain membrane potential.

At the maximal effect of the three blockers on the cDAP₁/cAP ratio, which occurred slightly after the maximal effect on cAP (Fig.6C), almost no fast repolarization was left, changing the cDAP₁/cAP fraction from 0.51 ± 0.06 to 0.93 ± 0.06 (n = 5; P = 0.043) (Fig.6E and F). At this time the cAP was slightly reduced, probably as a result of membrane depolarization, but the large cDAP₁ amplitude at this time was not attributable to depolarization because such an increase in cDAP₁ amplitude was never observed in experiments in which depolarization was induced by increasing K⁺ concentration (Fig.3). The requirement for the combination of quinine, MgTX and TEA to block fast repolarization suggests that channels from three pharmacologically distinguishable groups participated in the fast repolarization.

Summary

Figure7 summarizes the main effects of TEA on cAP and cDAP and uses data presented elsewhere in this article. The small effect of TEA alone on cDAP (Fig.7A, left) suggests that other channels could compensate and repolarize the spike relatively well when TEA-sensitive channels were blocked. The strong effect of TEA when MgTX (Fig.7A, middle) and MgTX + quinine-sensitive channels were blocked (Fig.7A, right) suggests that those were the channels that compensated. In all of these three groups of experiments (Fig.7A), TEA had similar effects on the fast action potential component (cAP), although with the combination of TEA, MgTX and quinine, cAP amplitude dropped from its maximal value (marked by a red circle).

The relative changes of cAP compared with relative changes of cDAP (scatterplot in Fig.7B) showed that TEA effects were clearly above the identity line, meaning that its effect on cAP exceeded that on cDAP, whereas MgTX alone had a lesser effect on cAP than on cDAP. MgTX and quinine added to the amplitude-increasing effect of TEA on cDAP₁ and even more so on cDAP₂.

Figure7C summarizes data showing that TEA, MgTX or quinine given alone did not reduce the ability to follow moderate spike frequencies over a short time. Higher axonal performance (higher frequencies, longer spike trains and accurate spike width control) may have been influenced by these blockers, but was not investigated.

Discussion

Main findings

By using a miniaturized grease-gap method on rat cerebellar PFs, we were able to distinguish between a fast TEA-sensitive spike repolarization and a much slower MgTX- and DTX-sensitive repolarizing phase. The effects of these drugs overlapped where the fast repolarization stopped and the slow decay started, captured by our parameter cDAP₁. The membrane potential at cDAP₁ was of particular interest because the corresponding intrasomatic DAP depolarized just before the axons started to burst during wash-in of 4-AP.

This suggests that control of the initial amplitude of the DAP is important for the burst behaviour of PFs. Because a significant fraction of the fast repolarization resisted TEA and MgTX block, and a third blocker (quinine) was needed to abolish the remaining fast repolarization, it is likely that the cDAP₁ was controlled by at least three pharmacologically distinguishable groups of Kv channels.

Interpretations of the grease-gap signal

Two critical questions about the grease-gap signal stand out. Is it axonal? Is the DAP caused by a rapid action potential-induced change of extracellular K⁺ concentration? As the answers to these questions are essential for the interpretation of the grease-gap experiments, we will refer to some of the control experiments described in Palani et al. (2012), and add figure references from the present article when relevant.

The DAP was of axonal origin because the DAP recorded at the soma when the axon was electrically activated persisted even when no spikes invaded the soma at hyperpolarized membrane potentials (Fig.1D and E). It did not reflect potassium efflux from neighbouring cells for three reasons: (i) a comparison of somatic voltage and current clamp recordings shows no detectable current underlying the somatically recorded DAP; (ii) at hyperpolarized potentials blocking the somatic spike the remaining DAP could flip between failure and its full amplitude at one constant stimulation strength (Fig.1D), and during stochastic failures no depolarizing wave was seen, but would have been expected if extracellular K⁺ from other cells caused the DAP; and (iii) after a single axon was activated by intrasomatic current injection, limiting the K⁺ efflux to that which naturally occurs from a single axon, the threshold for activation of that axon was clearly reduced for 10–50 ms. Furthermore, the time course of the threshold reduction resembled the time course of the decay of the DAP, and that of the grease-gap axonal recordings. Finally, the DAP was not synaptic because it resisted 50–200 μm Cd²⁺ (Fig.5F).

The qualitative similarity between 4-AP effects on somatically recorded antidromic APs and 4-AP effects on the grease-gap signal supports the interpretation that changes of cAP, cDAP₁ and cDAP₂ measured by the grease-gap method reflect changes in the axon. However, the increases in the amplitude of cAP and cDAP may be attributable to hyperpolarization of the membrane potential, shifting the grease-gap baseline down, while the membrane potentials of the spike peak and its after-potential remain unchanged. Although absolute membrane potentials cannot be recorded by our implementation of the grease-gap method, we consider this possibility unlikely because the K⁺ channel blockers we used would have a depolarizing effect, if any, on the resting membrane potential and therefore also the grease-gap baseline. This conclusion is supported by the minimal shift in membrane potential with 4-AP during intrasomatic recordings. Furthermore, depolarization induced by increased K⁺ concentration in the bath solution did not increase the amplitude of cAP or cDAP. This observation also makes it unlikely that the amplitude increases were caused by the activation of more axons in response to K⁺ channel blockers.

Control of fast repolarization

We found that TEA-sensitive channels in PFs contributed to the control of fast repolarization, similarly to findings in investigations using voltage-sensitive dyes (Sabatini & Regehr, 1997; Matsukawa et al. 2003). We complement the conclusions of these earlier studies with the finding that TEA-sensitive channels also control fast repolarization at 35°C (in addition to 20–24°C), and that TEA had minimal effect on potentials following the fast repolarization.

The channels blocked are probably Kv3.1b and Kv3.3 (Puente et al. 2010) because TEA of <1 mm had small effects on PFs in Kv3.1 and Kv3.3 double knockout mice (Matsukawa et al. 2003). By contrast with TEA's narrow time window of action, we found that 4-AP increased both cAP and cDAP and therefore probably blocks more channel types in the PFs than TEA.

Increases in both individual spike amplitudes and their width may have contributed to the observed increases in cAP. The reason is that the average conduction distance was approximately 1 mm (based on the average latency to cAP, which was 3.56 ± 0.32 ms). Combined with an increase in the width of the arrival time distribution of 0.46 ms · mm⁻¹ along PFs (Soleng et al. 2003), this gives an average spread of the spike arrivals over 0.46 ms at the recording electrode, and a widening of the action potential would give more efficient summation and increase the cAP amplitude.

Control of slow repolarization

By contrast with the fast repolarization, control of the DAP has not been investigated in small-bouton grey matter axons. Sabatini & Regehr (1997) and Matsukawa et al. (2003) observed a slow after-potential in PFs using voltage dyes, but explained it with extracellular K⁺ accumulation. Based on the arguments given above under ‘Interpretations of the grease-gap signal’, we consider that explanation unlikely.

We found that MgTX and DTX increased the amplitude of the cDAP, but did not change the cAP. MgTX and DTX did not add their effects and may therefore block the same channel or set of channels. In particular the observation that cDAP₂ increased more than cDAP₁, giving the cDAP a hump, suggests that depolarizing currents, most likely Na⁺ or Ca²⁺, were able to contribute to the DAP when MgTX-sensitive channels were blocked. The involvement of Ca²⁺ currents is supported by the finding that Cd²⁺ reduced the MgTX-enhanced cDAP₂. This makes the control of DAP amplitude very important to axons because Ca²⁺ activates many biochemical pathways, including those that are apoptotic (Giorgi et al. 2008; Nakamura & Lipton, 2010). The involvement of Ca²⁺ currents does not rule out a contribution from Na⁺ currents, such as reported at the calyx of Held (Kim et al. 2010) and the granule cell's axon initial segment (Osorio et al. 2010).

Although TEA alone had little effect on cDAP₁ and cDAP₂, its combination with MgTX increased these amplitudes considerably. Our interpretation of these data is that, normally, TEA- and MgTX-sensitive channels collaborate to bring the membrane potential down to a value sufficiently negative to inactivate voltage-sensitive channels (Na⁺ and Ca²⁺), but that MgTX-sensitive channels are particularly important to keeping the potential at that level until the inactivation has taken place, reducing the effect of Ca²⁺ tail currents (Jung et al. 2001). A similar explanation was proposed for the calyx of Held, where the contribution from MgTX-sensitive channels was seen only when TEA-sensitive channels were blocked (Ishikawa et al. 2003), although the relative contribution from TEA- and MgTX-sensitive channels may vary among types of terminal, with activity and with experimental conditions (Steinert et al. 2011).

Like those of Steinert et al. (2011) our data support powerful and flexible roles of TEA-sensitive channels. Those features are illustrated (Fig.7A) by TEA's dramatic effect on fast repolarization when MgTX- and quinine-sensitive channels were blocked, meaning that TEA-sensitive channels accounted for virtually all fast repolarization in that case. However, when MgTX- and quinine-sensitive channels were not blocked, TEA had virtually no effect on fast repolarization, which implies that MgTX- and quinine-sensitive channels must have been the main participants.

Control of DAP and bursting

Our data provide evidence that PFs have an intrinsic ability to burst. The mechanism for the burst generation observed without K⁺ channel blockers is unknown, and may theoretically be attributable to pathology induced by our experimental procedure. The findings that 4-AP always induced bursts in previously non-bursting axons, and that DAP amplitude increased just before bursting occurred, suggest that proper control of DAP amplitude by 4-AP-sensitive channels usually prevents bursting. This may give the impression that bursting is easily achieved in PFs; we will discuss this below.

In apparent support of this hypothesis is the fact that 4-AP gave only 4.2 mV average depolarization of the early part of DAP before additional spikes appeared during the slow decay of the after-potential. This could mean that the early part of DAP was very close to the spike threshold (without 4-AP). However, the finding that somatic hyperpolarization, on average by 15.8 mV, did not block the bursts suggests that it was initiated electrotonically far from the soma, possibly allowing 4-AP to depolarize DAP more at the axonal burst initiation site than we were able to measure at the soma. Exactly how far out in the axon bursts could be generated we cannot know because the grease-gap technique does not detect individual action potentials.

Although grease-gap recordings showed that 4-AP increased DAP amplitude in distal parts of the PFs, we cannot conclude that this led to bursts originating in the distal axon. This is because several conditions in addition to depolarization are needed for a burst to occur; for example, voltage-sensitive Na⁺ channels must reactivate and membrane resistance must be sufficiently high to allow integration of the Na⁺ current. Direct recordings from individual PFs may solve this question, similarly to bursts recorded from individual axons in stratum radiatum of the hippocampus (Palani et al. 2010).

However, several data favour the alternative hypothesis, namely, that the amplitude of the DAP, and therefore bursting, is strictly controlled in PFs. Most importantly, as mentioned above, three blockers, MgTX, TEA and quinine, were required to abolish fast repolarization, suggesting that at least three types of Kv channel may support fast repolarization and the control of DAP amplitude. Furthermore, the finding that none of the blockers TEA, MgTX or quinine alone interfered with the ability to fire three action potentials with 10 ms intervals suggests strict control of the DAP. This interval is in the middle of a typical burst, and under the reasonable assumption that the axon's ability to respond to electrical activation is changed during a burst, it is unlikely that any of those blockers induced bursting in these axons.

The lack of burst-inducing effect of TEA is supported by others (Sabatini & Regehr, 1997; Matsukawa et al. 2003). We must emphasize, however, that we have not investigated the fine-tuning of spike width, nor the ability to follow high frequencies over long periods, in which TEA-sensitive channels are important (Sabatini & Regehr, 1997; Matsukawa et al. 2003).

Functional consequences

With the complexity implicit in the combination of action potential propagation and transmitter releasing functions, it is not surprising to find several components of spike repolarization. The finding that PFs have MgTX-sensitive channels that can fine-tune DAP amplitude independently of fast repolarization may give neurons the ability to influence the TEA-sensitive synchronous release (Sabatini & Regehr, 1997) independently from the asynchronous release (Atluri & Regehr, 1998). By contrast with these theoretically separable controls, the initial amplitude of the DAP, which seems to be the most important point for the suppression of bursting, is regulated by at least three groups of channels (TEA-, MgTX- and quinine-sensitive), and may therefore be particularly resistant to modulation.

Glossary

4-AP

4-aminopyridine

ACSF

artificial cerebrospinal fluid

APV

dl-amino-5-phosphonovaleric acid

cAP

compound action potential

cDAP

compound depolarized after-potential

DAP

depolarized after-potential

DNQX

6,7-dinitroquinoxaline-2,3-dione

DTX

dendrotoxin

Kvs

voltage-sensitive K⁺ channels

MgTX

margatoxin

PFs

parallel fibres

TEA

tetraethylammonium

Key points

• The presynaptic action potential waveform was investigated in cerebellar parallel fibres from rats.

• The spike repolarization was composed of a fast tetraethylammonium-sensitive component and a slow margatoxin-sensitive depolarized after-potential (DAP). These components could be manipulated relatively independently, possibly offering independent control of synchronous and asynchronous transmitter release.

• Axonal electrical activation sometimes gave bursts of action potentials at the soma; these bursts were created by the axon because they invaded the soma at membrane potentials well below the somatic spike threshold.

• Axonal bursts could reliably be induced by increasing the DAP pharmacologically, suggesting that proper control of DAP amplitude is necessary to suppress axonal bursting.

• The fast repolarization was particularly well controlled because blockers of three groups of K⁺ channels (tetraethylammonium, margatoxin and quinine) were needed to abolish it.

Additional information

Competing interests

None declared.

Author contributions

D.P. and M.R. contributed to the conception and design of the study. D.P., H.J.S. and A.B. performed the experiments. All authors analysed the data. D.P. and M.R. wrote the manuscript. All authors contributed to the critical revision of the paper and approved the final manuscript for publication.

Funding

This work was supported by the National Institutes of Health, Bethesda, MD, USA (grant no. 1R21NS082680-01A1).

Author's present address

H. J. Szkudlarek: Institute of Physiology I, Westfaelische Wilhelms University, Münster, Germany.