1Voltage-gated Na+ channels and their distribution were studied by the patch-clamp technique in intact dorsal horn neurones in slices of newborn rat spinal cord and in neurones isolated from the slice by slow withdrawal of the recording pipette. This new method of neurone isolation was further used to study the roles of soma and axon in generation of action potentials.
2Tetrodotoxin (TTX)-sensitive Na+ currents in intact neurones consisted of three components. A fast component with an inactivation time constant (τf) of 0.6–2.0 ms formed the major part (80–90%) of the total Na+ current. The remaining parts consisted of a slowly inactivating component (τs of 5–20 ms) and a steady-state component.
3Single fast and slow inactivating Na+ channels with conductances of 11.6 and 15.5 pS, respectively, were identified in the soma of intact neurones in the slice. Steady-state Na+ channels were not found in the soma, suggesting an axonal or dendritic localization of these channels.
4In the whole-cell recording mode, the entire soma of a dorsal horn neurone could be isolated from the slice by slow withdrawal of the recording pipette, leaving all or nearly all of its processes in the slice. The isolated structure was classified as: (1) ‘soma’ if it lost all of its processes, (2) ‘soma+axon’ complex if it preserved one process and at least 85% of its original peak Na+ current or (3) ‘soma+dendrite’ complex if it preserved one process but the remaining Na+ current did not exceed those observed in the isolated ‘somata’.
5The spatial distribution of Na+ channels in the neurone was studied by comparing Na+ currents recorded before and after isolation. The isolated ‘soma’ contained 13.8 ± 1.3% of inactivating Na+ current but no steady-state Na+ current. ‘Soma+axon’ complexes contained 93.6 ± 1.4% of inactivating and 46% of steady-state Na+ current.
6In current-clamp experiments, the intact neurones and isolated ‘soma+axon’ complexes responded with ‘all-or-nothing’ action potentials to current injections. In contrast, isolated ‘somata’ showed only passive or local responses and were unable to generate action potentials.
7It is concluded that dorsal horn neurones of the spinal cord possess three types of TTX-sensitive voltage-gated Na+ channels. The method of entire soma isolation described here shows that the majority of inactivating Na+ channels are localized in the axon hillock and only a small proportion (ca 1/7) are distributed in the soma. Steady-state Na+ channels are most probably expressed in the axonal and dendritic membranes. The soma itself is not able to generate action potentials. The axon or its initial segment plays a crucial role in the generation of action potentials.
Dorsal horn neurones of the spinal cord play a key role in processing the sensory information received from the primary afferent fibres. Several types of dorsal horn neurone have been classified according to their receptive field, the kind of stimulation that leads to neurone excitation and their pattern of firing behaviour (Woolf & Fitzgerald, 1983; Lopez-Garcia & King, 1994). Dorsal horn neurones display a broad range of firing patterns, from tonic firing in response to a sustained depolarization to strongly phasic firing resulting in only one action potential at the beginning of the pulse (Lopez-Garcia & King, 1994). Such complexity in firing behaviour is thought to be determined by the cell-specific expression of ion channels underlying the major membrane conductances. It has been shown that dorsal horn neurones possess inward Na+ as well as inward Ca2+ and outward K+ currents (Huang, 1987), which play an important role in the generation of action potentials (Murase & Randic, 1983).
Recent molecular biology investigations have undoubtedly indicated the expression of several types of Na+ channel α-subunit mRNA in the dorsal horn of the spinal cord (Westenbroek, Merrick & Catterall, 1989; Black, Yokoyama, Higashida, Ransom & Waxman, 1994). Differences in the developmental regulation of Na+ channel mRNA expression in the spinal cord (Gordon et al. 1987; Beckh, Noda, Lübbert & Numa, 1989) and a variable degree of β1-subunit expression amongst different cell types (Oh, Sashihara & Waxman, 1994) could provide the basis for an additional diversity of Na+ channels in dorsal horn neurones. Unfortunately, a comprehensive electrophysiological description of voltage-activated Na+ channels in dorsal horn neurones is lacking at present and the physiological relevance of the molecular biology experiments remains to be proven by direct measurements of Na+ currents in intact neurones.
In the present work we studied Na+ channels in laminae I–III dorsal horn neurones identified in slices of newborn rat spinal cord. It is shown that dorsal horn neurones possess three different types of voltage-gated Na+ channels: two types of inactivating and one type of non-inactivating steady-state channel. By comparing Na+ currents recorded from the neurones within the slice with those recorded from isolated ‘somata’ it was found that the ‘soma’ of the neurone contained only one-seventh of the inactivating Na+ channels. The rest of the inactivating Na+ channels are located on the axon. The steady-state Na+ channels are mostly distributed over the axonal and dendritic, but not the somatic, membranes. Functionally, it is shown that the soma itself cannot generate action potentials; in spinal dorsal horn neurones, an axon is needed for action potential generation.
Experiments were performed by means of the patch-clamp technique (Hamill, Marty, Neher, Sakmann & Sigworth, 1981) on 200 μm thin slices (Edwards, Konnerth, Sakmann & Takahashi, 1989) prepared from the lumbar enlargement (L3–6) of the spinal cord of 2- to 9-day-old rats. Rats were rapidly decapitated and the spinal cords carefully cut out. The shoes were prepared and kept according to a description given by Takahashi (1990). The study was performed on 8–12 μm dorsal horn neurones identified in laminae I-III of the spinal cord (Fig. 1A). No cleaning procedure was necessary in the present study, since tens of neurones on the surface of each slice were free from connective tissue and directly accessible for the patch pipette.
During experiments the cells were viewed with Nomarski optics. The photographs in Fig. 7B and C were made with a thermal video printer using infrared video microscopy (Stuart, Dodt & Sakmann, 1993). Neurones shown in Fig. 7D and E under fluorescent optics were filled with 0.3% Lucifer Yellow (dilithium salt; Sigma) dissolved in high-Csi+ solution.
Ringer solution for preparing and maintaining the slices contained (mM): NaCl, 115; KCl, 5.6; CaCl2, 2; MgCl2, 1; glucose, 11; NaH2PO4, 1; NaHCO3, 25 (pH 7.4 when bubbled with a 95% O2–5% CO2 mixture). During all experiments the shoes were perfused with low-Ca2+, high-Mg2+ Ringer solution, in order to reduce synaptic activity in neurones. This solution was obtained from the first one by setting the concentrations of Ca2+ and Mg2+ at 0.1 and 5 mM, respectively, and is hence referred to as Ringer solution. External tetraethylammonium (TEA)-containing solution (Ringer-TEA) was made up of (mM): NaCl, 95; KCl, 5.6; CaCl2, 0.1; MgCl2, 5; glucose, 11; NaH2PO4, 1; NaHCO3, 25; and TEA-Cl, 20 (pH 7.4 by bubbling with O2–CO2 as above). External TEA-containing solution used for pipette filling in experiments with inside-out patches (Ringer-TEAo) contained (mM): NaCl, 115.8; KCl, 5.6; CaCl2, 0.1; MgCl2, 5; glucose, 11; TEA-Cl, 36; and Hepes, 10 (pH 7.4 adjusted with NaOH to give a final concentration of 5.2 mM). Na+-free choline-Cl solution contained (mM): choline chloride, 141; KCl, 0.6; CaCl2, 0.1; MgCl2, 5; glucose, 11; and Hepes, 10 (pH adjusted to 7.4 with KOH to give a final concentration of 5 mM). Tetrodotoxin (TTX) was added directly to external solutions.
Standard internal solution (high-Ki+) contained (mM): NaCl, 5; KCl, 144.4; MgCl2, 1; EGTA, 3; and Hepes, 10 (pH adjusted to 7.3 with KOH to give a final concentration of 10.6 mM). Successful isolation of a neurone from a shoe was achieved when the total amount of internal Na+ was increased to 15 mM. The composition of this internal solution (high-Ki+°) was (mM): NaCl, 5; KCl, 144.4; MgCl2, 1; EGTA, 3; and Hepes, 10 (pH adjusted to 7.3 with NaOH giving a final concentration of 10 mM). Internal solution for studying Na+ channels (high-Csi+) contained (mM): NaCl, 5.8; CsCl, 134; MgCl2, 1; EGTA, 3; and Hepes, 10 (pH adjusted to 7.3 with NaOH to give a final concentration of 9.2 mM). Internal solution with 50 mM CsF contained (mM): NaCl, 11.5; CsCl, 87; CsF, 50; MgCl2, 1; and Hepes, 10 (pH adjusted to 7.3 with 3.5 mM NaOH).
The patch pipettes were pulled in two stages from borosilicate glass tubes (GC 150, Clark Electromedical Instruments, Pangbourne, UK). The pipettes used for single-channel recordings were coated with Sylgard 184 (Dow Corning) and had a resistance of 8–22 MΩ. The pipettes for whole-cell recordings had a resistance of 2–7 MΩ. All pipettes were fire polished immediately before the experiments. The patch-clamp amplifier was a List EPC-7 (Darmstadt, Germany) in all voltage- and current-clamp experiments. The effective corner frequency of the low-pass filter, unless otherwise stated, was 3 kHz. The frequency of digitization was 5–10 kHz (except the recording in Fig. 8A). Data were stored and analysed using commercially available software (pCLAMP, Axon Instruments). Transients and leakage currents were digitally subtracted using records with either negative or positive pulses which activated no channels. Offset potentials were nulled directly before formation of a seal. In fifteen experiments we tried to compensate the error produced by the resistance in series (up to 65% compensation). However, smooth activation characteristics of Na+ currents could still not be obtained for neurones in the slices due to incomplete space clamp of remote membrane regions. Therefore no series resistance compensation was employed further in the present study. In the majority of isolated ‘somata’, Na+ currents had smaller amplitudes of 50–500 pA and voltage error due to series resistance became, in most cases, smaller than 5 mV.
The inside-out patches were studied in an additional small chamber as described elsewhere (Safronov & Vogel, 1995) in order to avoid the destruction of the whole slice during its superfusion with internal solutions. The present study is based on recordings from 268 intact neurones in the shoe, 119 isolated ‘somata’, ‘soma+axon’ and ‘soma+dendrite’ structures (see Results) and from twenty-nine excised membrane patches. All experiments were carried out at a room temperature of 21–24 °C. Data are given as the means ±s.e.m.
Identification of dorsal horn neurones
Several criteria were used in the present study for distinguishing between dorsal horn neurones and glial cells. It has been shown that some types of glial cells in the grey matter of rat spinal cord slices possess voltage-gated Na+ channels (Chvátal, Pastor, Mauch, Syková & Kettenmann, 1995). However, the amplitude of Na+ currents was at least one order of magnitude smaller than that of K+ currents and no glial cells in the spinal cord slice were able to generate action potentials. Unfortunately, the major part of the present experiments was performed using pipettes filled with high-Csi+ solution and no action potentials could be recorded. Therefore, in these experiments dorsal horn neurones were separated from glial cells on the basis of the amplitudes of their peak Na+ currents.
A total of sixty-three neurones and glial cells were first studied in Ringer solution using pipettes filled with standard high-Ki solution. Under voltage-clamp conditions, the maximum Na+ current activated by depolarizing voltage steps from a holding level of –70 mV was determined. This was usually observed at –40 or –30 mV, less frequently at –20mV. Thereafter, the cell's ability to generate action potentials was tested in current-clamp mode. Figure 1B shows typical recordings obtained from neurones and glial cells. Cells which responded to depolarizing voltage pulses with large 0.5–4 nA Na+ currents under voltage-clamp conditions (upper trace) and generated short action potentials in current-clamp mode (lower trace) were considered as neurones. In contrast, cells which showed either none or only very small 0.01–0.3 nA Na+ currents (upper traces) and could not generate action potentials (lower traces) were assumed to be glial cells. One of the reasons why glial cells with Na+ channels were not able to generate action potentials was the presence of relatively large leakage currents seen in our experiments. Therefore, the ratio of Na+ currents to leakage currents could be used as a further criterion for the separation of neurones from glial cells.
The distribution of maximum Na+ currents is given in Fig. 1C. The cells which generated no action potentials were considered as glial cells and are shown by filled bars. It could be seen that the cells with peak Na+ current exceeding 0.4 nA were most probably neurones.
The ratio of Na+ currents to leakage currents was defined as the ratio between the maximum Na+ current activated from a holding potential of –70 mV and the leakage current produced by a hyperpolarizing 40 mV voltage step. This ratio was between 0 and 4 for glial cells and between 10 and 600 for neurones (Fig. 1D).
From our data, the following criteria for neurone identification in voltage-clamp experiments were used. Cells with Na+ currents exceeding 1 nA were considered as neurones. Cells with Na+ current amplitudes between 0.6 and 1.0 nA (indicated by the dashed lines in Fig. 1C) were considered as neurones if the ratio of Na+ current to leakage current exceeded 15 (dashed line in Fig. 1D). All cells with Na+ current amplitudes less than 0.6 nA were discarded regardless of their Na+ current to leakage current ratio. The neurone identification was often confirmed by spontaneous synaptic currents observed during experiments. No further identification of different types of spinal dorsal horn neurones was undertaken in the present work.
The neurones investigated had a high input resistance of 3.7 ± 0.5 GΩ (104 neurones). The neurones investigated with standard high-Ki+° solution showed resting potentials between –50 and –85 mV under the present experimental conditions.
Na+ currents in dorsal horn neurones of rat
Voltage-activated Na+ currents were mainly recorded in external Ringer-TEA solution using pipettes filled with high-Csi+ solution for the suppression of voltage-activated K+ currents. A typical whole-cell Na+ current activated by a voltage step to [–30 mV in a dorsal horn neurone in the spinal cord slice is shown in Fig. 2. Three components of whole-cell Na+ current could be identified on the basis of their inactivation kinetics. A fast component inactivated with a time constant of 0.6–2.0 ms (τf) and made up 80–90% of the total Na+ current. A slow component inactivated with a time constant (τs) of 5–20 ms and produced 5–20% of the total Na+ current. The third component was a steady-state one, which showed no inactivation and an amplitude of 0.5–6% of the total Na+ current (mean, 2.1 ± 0.2%, 60 neurones). Both the slow and the steady-state components began to activate at –60 to –50 mV. The fast component recorded from neurones in the slice activated abruptly at –50 to –40mV, indicating insufficient space clamp of the neuronal membrane under these experimental conditions (see also Fig. 10A). For the same reason, no reversal potential could be measured for the peak Na+ currents, since they remained inwardly directed even at nominative potentials as positive as +80 mV (the calculated equilibrium potential for Na+ (ENa) was +53 mV). In contrast, the steady-state current changed its polarity at potentials of +20 to +40 mV, presumably due to activation at these potentials of non-specific outward Cs+ current through delayed-rectifier K+ channels which could not be completely blocked by 20 mM external TEA. There are several indications that none of these components resulted from activation of voltage-gated Ca2+ channels. All three were recorded in external Ringer-TEA solution containing 0.1 mM Ca2+ and 5 mM Mg2+. They were also recorded when the internal solution contained 50 mM F− (17 neurones), known to produce an irreversible block of Ca2+ channels. All three components disappeared in external Na+-free choline-Cl solution (5 neurones) and were blocked by 1 μM external TTX (15 neurones).
Kinetics of Na+ current blockade by TTX
In the following experiments the time course of Na+ current reduction during perfusion of the slice with TTX was studied (Fig. 3). The time elapsed from the start of TTX perfusion is indicated above the corresponding traces. TTX first suppressed the fast component of Na+ current completely and only several minutes later the block of the slow and steady-state components developed. The effect could be seen more clearly when the current traces were superimposed (Fig. 3, middle). The same time course of Na+ current block by TTX was observed in ten out of ten dorsal horn neurones. Such an effect could be explained by assuming that the connective tissue in the spinal cord slice impedes the diffusion of the blocker molecules and, therefore, different regions of the neuronal membrane experience different gradients of TTX concentration.
It was concluded that the channels underlying the fast and the slow components of whole-cell Na+ current differ from one another either in their sensitivity to TTX or in their spatial distribution across the neuronal membrane.
Because of the problems arising from insufficient space clamp, no further investigations of Na+ currents in neurones in the spinal cord slice were performed.
Outside-out membrane patches obtained from the soma of dorsal horn neurones with 9–18 MΩ pipettes contained between two and fifteen Na+ channels of both fast and slow types. The proportion of fast to slow components in averaged currents (approximately 5–10:1) and their inactivation kinetics (τf of 0.5– 1.5 ms and τs of 5–15 ms) were very similar to those described for whole-cell currents. In contrast to Na+ currents recorded from neurones in slices, a steady-state component was never seen in recordings from ten outside-out patches.
Patches containing only a few Na+ channels could seldom be obtained in the outside-out recording mode. Therefore, the study of the single-channel properties was mostly performed by using inside-out patches. The pipettes filled with Ringer-TEAo solution had a resistance of 8–22 MΩ. The internal surface of the patch was exposed to high-Csi+ bath solution. Approximately every second inside-out patch contained active Na+ channels, indicating their relatively low density in the somatic membrane of dorsal horn neurones. Twelve of nineteen inside-out patches with active Na+ channels contained only the fast-type channels. Figure 4A shows single-channel recordings and an averaged current (lowermost trace) from one such inside-out patch. The channels opened quickly and inactivated completely within 3–4 ms. The time of the last channel closing seen in this patch is indicated by the asterisk above the averaged current. The inactivation kinetics of the averaged fast current could be satisfactorily fitted by a monoexponential with a τf of 1.2 ms. The mean value of τf at -20 mV was 1.1 ± 0.1 ms (12 patches).
Seven of nineteen inside-out patches contained a mixture of one slow and between one and four fast Na+ channels. Figure 4B shows openings of fast (first and second trace) and slow channels (third and fourth trace) recorded from the same patch. The slow channels inactivated within 10–30 ms. The time of the last channel closing is shown by the asterisk above the averaged current (lowermost trace). Two exponentials, fast and slow, were needed for fitting the inactivation kinetics of the averaged current, revealing a τf of 1.5 ms (62%) and a τs of 11.9 ms (38%). The mean values of τf and τs obtained for these seven patches at -20 mV were 1.1 ± 0.1 ms and 7.2 ± 1.1 ms, respectively.
In further experiments the kinetics of the fast Na+ channel inactivation were studied at different potentials. The averaged currents from one inside-out patch containing four fast Na+ channels are shown at -40, -20 and 0 mV (Fig. 5). The kinetics of channel inactivation depended on voltage but remained monoexponential at all potentials investigated (4 patches).
It can be seen from the recordings shown in Fig. 4B that the currents through slow Na+ channels were slightly larger than those through fast channels. In order to verify this observation, we constructed point-amplitude histograms for both channel currents recorded at a potential of -20 mV (Fig. 6). Twenty-eight fast channel openings were selected from eight inside-out patches containing only fast Na+ channels. Ten slow channel openings were selected from three inside-out patches which contained a mixture of both channels. In such patches the averaged currents were first constructed and the τf and τs values determined by biexponential fitting of the inactivation kinetics. Each single episode was then analysed. Parts of long openings 3 ×τf ms after the moment when the averaged Na+ current reached its peak were taken for the histogram. The fitting of the point-amplitude histograms with two Gaussian functions gave the values of -0.85 pA (variance, δ= 0.20 pA) for the fast channel and -1.13 pA (δ= 0.25 pA) for the slow channel at a potential of -20 mV. Assuming an ENa value of +53 mV, the corresponding conductances were calculated to be 11.6 pS for the fast channel and 15.5 pS for the slow channel.
The channels responsible for the steady-state component of the whole-cell Na+ current were not observed in inside-out patches originating from the soma membrane.
On the basis of whole-cell and single-channel experiments it was possible to conclude that the TTX-sensitive Na+ conductance in spinal dorsal horn neurones consisted of three components. The channels underlying inactivating fast and slow current components could be identified in patches originating from the soma membrane. Insufficient space clamp during whole-cell recording in the spinal cord slice and a low frequency of Na+ channel appearance in inside-out patches from the soma membrane could indicate that the majority of Na+ channels are located on the axonal or dendritic membrane. The axons and dendrites of dorsal horn neurones with diameters smaller than 0.5 μm could barely be resolved even with infrared optics and, therefore, axonal or dendritic localization of Na+ channels could not be directly shown by single-channel recording.
In order to solve this problem we developed a method for investigating the functional distribution of channels in small (ca 10 μm) cells with fine processes.
The method of entire soma isolation
In the whole-cell recording mode, entire somata of dorsal horn neurones could easily be isolated from the slice by slow withdrawal of the recording pipette, leaving all or nearly all of their processes in the slice (Fig. 7). The channel distribution could then be studied by comparing the macroscopic Na+ currents recorded in neurones before and after their isolation.
When recording from the neurone in the slice was completed, a slight suction was applied to the recording pipette and it was gently withdrawn until the connection between the soma and the slice was broken (Fig. 7A). The suction applied was similar to that needed to break the membrane during formation of whole-cell mode. The suction was immediately released after the isolation was completed. The isolated structure was classified as a ‘soma’ if adjacent processes could not be seen either during the experiment or after, when the recording pipette was turned over. The isolated structure was classified as a ‘soma+axon’ if it contained one 10–100 μm process and preserved more than 85% of the original Na+ current recorded in the slice before isolation. The structure was considered as a ‘soma+dendrite’ if one adjacent process was observed but the amplitude of Na+ current was in the range typical for isolated ‘somata’. A total of ninety-four isolations could be classified into seventy-six ‘somata’, twelve ‘soma+axon’ and six ‘soma+dendrite’ complexes. Stable recordings lasting for 1 h and more could be obtained from each of the three isolated structures. In general, the probability of obtaining a certain configuration depended on the age of the animal, the time elapsed since the slice preparation, the duration of the whole-cell recording from the neurone in the slice before isolation, and the type of pipette solution.
The isolation of the ‘soma’ from a dorsal horn neurone in a spinal cord slice is shown under infrared optics (Fig. 7B) and fluorescent optics when the neurone was filled with Lucifer Yellow dye (Fig. 7D). The dendritic tree remaining in the slice after ‘soma’ isolation can be seen in Fig. 7D (middle). Isolated ‘soma+axon’ complexes obtained from other neurones are shown in Fig. 7C and E.
Changes in leakage current and membrane resting potential monitored during isolation of the cell in Ringer solution are shown in Fig. 8. The leakage currents were studied in voltage-clamp mode using pipettes filled with high-Csi+ or high-Ki+° solutions. Leakage current was monitored every second by 100ms voltage steps from -80 to -120mV (Fig. 8A). A successful isolation was always accompanied by a considerable reduction in the leakage current, probably due to the loss of the greater part of membrane area. (Compare the area of the isolated ‘soma’ with that of the processes remaining in the slice in Fig. 7D.) In ninety-three neurones the leakage current was reduced to 50 ± 3% after isolation. The loss of membrane area led to an essential reduction in electrical noise as well as to an acceleration in the transients, which could no longer be resolved under the recording conditions of Fig. 8A. The change of resting potential during isolation was studied in current-clamp mode using pipettes filled with high-Ki+° pipette solution. Neurones with relatively low resting potentials of -50 to -60 mV hyperpolarized during isolation (Fig. 8B, upper trace), probably due to a decrease in non-specific leakage conductance. Neurones with resting potentials arround -70 mV did not show considerable changes in membrane potential during isolation (Fig. 8B, lower trace).
Distribution of Na+ channels in soma, axon and dendrites
The following experiments were performed in external Ringer-TEA solution using pipettes filled with high-Csi+ solution. Na+ currents recorded from neurones before and after isolation were compared. A typical isolated ‘soma’ lost the majority of inactivating and all steady-state Na+ channels (Fig. 9A). The mean inactivating current measured in the ‘soma’ was 13.8 ± 1.3% (52 neurones) of that recorded from the neurones in the slice before isolation. The steady-state currents disappeared completely in forty-six of fifty-two ‘somata’. Openings of non-inactivating Na+ channels could not be revealed even when the ‘soma’ currents were recorded at the high amplification normally used in single-channel experiments.
In the remaining six ‘somata’ the mean steady-state current was 5.8 ± 1.1 pA and openings of the steady-state channels could be seen. It should be noted that these ‘somata’ preserved relatively large peak Na+ currents of 570 ± 72 pA. Therefore, it could not be excluded that these six ‘somata’ may have contained parts of the axon hillock.
Isolated ‘soma+axon’ complexes preserved almost the whole inactivating current (93.6 ± 1.4%, 6 neurones) and a large part of the steady-state current (Fig. 9B). Unfortunately, it was not possible to estimate the proportion of steady-state current remaining in a ‘soma+axon’ complex by direct comparison from the same cell. Adequate measurements of the steady-state currents in the neurone within the slice could be achieved no earlier than 3–5 min after formation of the whole-cell recording mode, when the original intracellular K+ in the small-diameter processes had been replaced by Cs+ and outward K+ currents no longer masked the steady-state Na+ current. The isolation of such neurones nearly always resulted in formation of a ‘soma’ configuration. The probability of getting a ‘soma+axon’ complex was much higher if isolation was completed not later than 1–1.5 min after breaking through the membrane. We therefore compared the mean amplitude of the steady-state current measured in six isolated ‘soma+axon’ complexes with that measured in twenty other intact neurones in the slice. The six isolated ‘soma+axon’ complexes with a mean peak Na+ current of 1.6 ± 0.4 nA had a mean steady-state current of 23.2 ± 6.5 pA. The reference group of twenty intact neurones, selected for a similar mean peak Na+ current (1.6 ± 0.1 nA), showed a mean steady-state current of 50.2 ± 5.4 pA. Thus, isolated ‘soma+axon’ complexes preserved 46% of the steady-state Na+ currents.
It could be supposed that the rest of the steady-state Na+ channels are located on dendrites. Indeed, small remaining steady-state components (5–25 pA) were seen in two ‘soma+dendrite’ complexes. Larger currents were not obtained, since only a small part of the dendritic tree could be isolated (see Fig. 7D).
Macroscopic Na+ currents in isolated ‘somata’ or ‘soma+axon’ complexes activated and inactivated faster than those recorded from neurones in the slice (Fig. 9). The activation of Na+ currents in the ‘soma’ took place over a broader voltage range, indicating improved space clamp conditions. The reversal potential for Na+ currents became equal to the equilibrium potential for Na+ ions (ENa) (Fig. 10A). However, decay kinetics of Na+ current remained biexponential (Fig. 10B) and the ratio between the amplitudes of fast and slow components was similar to that observed for neurones in the slices.
In five isolated ‘somata’ we tested the sensitivity of the fast and slow inactivating components of Na+ current to TTX. The range of TTX concentrations tested was 3–100 nM. In isolated structures, 100 nM TTX blocked both components completely. At 3, 10 and 30 nM TTX, a progressive block of both components was seen. The fast component was only slightly more sensitive to TTX, but at no concentration was a selective block of only one component observed. Therefore, the different block of the fast and slow components of Na+ current during slice perfusion with 1 μM TTX (Fig. 3) most probably resulted from a different spatial distribution of fast and slow Na+ channels across the neuronal membrane.
Functional role of soma and axon in action potential generation
The role of the soma and axon in the generation of action potential was studied under current-clamp conditions. Dorsal horn neurones before and after isolation were kept at a potential of -80 or -70 mV by injection of steady-state currents through the recording pipette. Action potentials were evoked by 10 ms depolarizing current pulses of increasing strength. A typical neurone in the slice generated an ‘all-or-nothing’ action potential in response to a current injection (Fig. 11). Smaller currents were needed for action potential activation in isolated ‘soma+axon’ complexes (Fig. 11A) probably due to an increase in input resistance during isolation. Injection of larger currents produced almost no changes in the shape of the action potential, in agreement with the all-or-nothing principle of action potential generation. In general, the shape of the action potentials recorded from neurones in slices and from ‘soma+axon’ complexes were almost identical (Fig. 11A, superimposed traces).
Another type of membrane response to current stimulation was observed for isolated ‘somata’ (Fig. 11B). The cell shown preserved about 10% of peak Na+ current after isolation and its input resistance was increased by a factor of 4.2. Much smaller current pulses depolarized the membrane to levels beyond the threshold of the action potential seen in this neurone before isolation (indicated by arrows). Further increases in the amplitude of injected current produced even stronger depolarization but all-or-nothing action potentials were no longer observed. This phenomenon became more evident when the action potential of the intact neurone in the slice was superimposed with the membrane responses of the isolated ‘soma’ (Fig. 11B, right).
In the following experiments the contribution of voltage-gated Na+ and passive membrane conductances to the ‘soma’ responses during current stimulation was studied. The passive membrane responses were obtained by injecting hyperpolarizing -10 pA currents both in neurones in slices and in isolated ‘soma’ (Fig. 11C). The active membrane responses were evoked by injection of depolarizing currents of different strength. The passive responses multiplied by the corresponding factors are shown as dashed lines. It can be seen that the response of the intact neurone was passive until the threshold level was reached and the activating Na+ conductance induced a steep membrane depolarization, i.e. the action potential (Fig. 11C, left). In contrast, the ‘soma’ responses remained passive and only a small deflection of membrane potential in the depolarizing direction could indicate the activation of a small Na+ conductance (Fig. 11C, middle). The ‘soma’ response to stronger current injection remained passive until the membrane potential deflected in the hyperpolarizing direction (indicated by the asterisk), presumably due to an activation of K+ conductance (Fig. 11C, right).
Isolated ‘somata’ with a higher percentage of remaining Na+ channels showed larger local Na+ responses but they were still not able to generate an all-or-nothing action potential. Another reason why the ‘soma’ could not generate action potentials was the presence of a large proportion of voltage-gated K+ channels (30–50%) remaining after ‘soma’ isolation. Detailed investigation of K+ channel distribution was beyond the scope of the present study.
We studied voltage-gated Na+ channels in dorsal horn neurones visually identified in laminae I–III of the rat spinal cord. Lamina I–III neurones are known to receive most of their information from myelinated Aδ and unmyelinated C fibres of primary afferents (Light & Perl, 1979; Ralston & Ralston, 1979; Woolf & Fitzgerald, 1983) which convey information about pain and thermoreception. Neurones were separated from glial cells on the basis of the amplitudes of their maximum Na+ currents and the ratio of Na+ current to leakage current.
Three types of Na+ channel
Voltage-activated Na+ currents in dorsal horn neurones consisted of fast, slow and steady-state components. All of these were also seen in isolated structures and both fast and slow components were additionally observed in excised patches, indicating that the three components did not result from insufficient voltage clamp of remote membrane regions. Our experiments with TTX have further shown that Na+ channels underlying fast and slow components most probably differ in their spatial distribution over the neuronal membrane. Therefore, it seems to be unlikely that the bi-exponential inactivation kinetics of the macroscopic currents was produced by changes in the gating mode of one type of channel. Further evidence for the existence of two types of inactivating channels was provided in experiments with inside-out patches. In the majority of patches, only fast inactivating Na+ channels were seen and the kinetics of channel inactivation remained monoexponential at all potentials investigated. In a few inside-out patches with one fast and one slow channel, the inactivation kinetics were bi-exponential and the contribution of the slow component was about 40%. Furthermore, the single-channel conductance of the slow channel was 25% larger than that of the fast channel. The fast Na+ channel underlying the major part of Na+ conductance in dorsal horn neurones had very similar properties to a Na+ channel type described in the soma of spinal motoneurones (Safronov & Vogel, 1995).
The steady-state component made up 0.5–6% of the total Na+ conductance. The amplitude of the steady-state current was 20–160 pA and it could therefore produce a strong depolarization in dorsal horn neurones with a large input resistance of several gigaohms. We investigated the possibility that the steady-state component could be a ‘window’ current resulting from an overlap of steady-state inactivation and activation characteristics of transient Na+ currents (Atwell, Cohen, Eisner, Ohba & Ojeda, 1979). Our calculations showed that the maximum amplitude of the window current would be 0.5% of the peak Na+ current at potentials of -50 to -40 mV and almost no window current would be seen at potentials positive to -10 mV. In contrast, the steady-state currents measured in the present experiments were much larger, reached their maximum at between -30 and 0 mV becoming negligible at potentials of between +30 and +40 mV because of a decrease in the driving force for Na+ ions and the activation of non-specific outward currents through delayed-rectifier K+ channels. Differential block of the fast and steady-state currents during slice perfusion with TTX is a further indication that the steady-state current is not a window current and could not result from incomplete inactivation of the fast channels reported for pyramidal neurones from rat sensorimotor cortex (Alzheimer, Schwindt & Crill, 1993). Furthermore, the steady-state current was considerably reduced in ‘soma+axon’ complexes, whereas the fast and slow components remained unchanged. The steady-state current disappeared completely in the majority of isolated ‘somata’ although the inactivating components were only reduced by a factor between 3 and 50. The steady-state current was seen in some ‘soma+dendrite’ complexes in which the inactivating components became very small.
Further evidence for Na+ channel diversity conies from molecular biology studies which showed the expression of several types of Na+ channel α-subunit mRNA in the dorsal horn of the spinal cord (Westenbroek et al. 1989; Black et al. 1994) and revealed a different degree of β1-subunit expression between different cell types within the spinal cord (Oh et al. 1994). A direct comparison between wild-type and cloned Na+ channels is unfortunately not possible at the moment, but it could be of importance for understanding the functional expression of Na+ channels in dorsal horn neurones.
The present results allow us to suggest that Na+ conductance in spinal dorsal horn neurones is based on three different types of Na+ channels. However, direct single-channel recordings from the steady-state Na+ channel still remain to be made.
Method of ‘soma’ and ‘soma+axon’ isolation
In order to study the channel distribution over the neuronal membrane, we developed a method which allowed the direct comparison of currents recorded from the neurone in the slice with those from isolated ‘somata’ and ‘soma+axon’ complexes. The method could be applied to small 8–12 μm neurones with very fine processes which could not be subjected to direct patch-clamp investigation. In comparison with the well-known outside-out (Hamill et al. 1981) and nucleated patch techniques (Sather, Dieudonné, MacDonald & Ascher, 1992) the present method allows the isolation of the entire neuronal soma or the entire soma with an attached axon or dendrite. The good physiological state of the isolated ‘somata’ or ‘soma+axon’ complexes was confirmed by a decrease in membrane leakage conductance, by stable or even improved membrane resting potentials and by the ability of ‘soma+axon’ complexes to generate action potentials.
Improved space clamp conditions of isolated ‘soma’ in comparison with the neurone in the slice make the present method useful also for studying the kinetics and activation characteristics of fast transient Na+ as well as K+ currents. Furthermore, ‘somata’ and ‘soma+axon’ complexes isolated from surrounding connective tissue could be successfully used in pharmacological studies of different types of identified neurones.
Distribution of Na+ channels between soma, axon and dendrites
The present results show that the soma of dorsal horn neurones contained only a small proportion (ca 1/7) of inactivating Na+ channels and almost no steady-state Na+ channels. The axon contained the rest of the inactivating and only about 50% of the total steady-state Na+ channels. Another 50% of the steady-state Na+ channels are probably distributed amongst the dendrites of the dorsal horn neurones.
The density of inactivating Na+ channels in the soma membrane could be estimated in two different ways. The mean peak Na+ current measured in isolated ‘soma’ was 306 ± 27 pA (52 ‘somata’) at -30 to -10 mV. This gives a mean channel density of 1.0 per μm2 under the assumption that the mean single Na+ channel current is 1 pA and the mean soma diameter is 10 μm. In our single-channel experiments, one inside-out patch obtained with 8–22 MΩ pipettes (mean, 11.8 ± 0.8 MΩ; 19 pipettes) contained two active Na+ channels on average. According to Sakmann & Neher (1995) the area of the patch obtained with a 12 MΩ pipette is 1.3 μm2, giving a channel density of 1.5 per μm2. Thus, the data obtained for isolated ‘somata’ are in good agreement with our single-channel experiments.
It should be mentioned that the number of active Na+ channels observed in outside-out patches was usually several times greater than in inside-out patches. One possible explanation for this is that the membrane area of the outside-out patches formed from the neuronal somata in the slice preparation is several times larger than the area predicted by Sakmann & Neher (1995). Additionally, we did not observe a correlation between the pipette resistance and the number of Na+ channels in outside-out patches. Even the patches obtained with very narrow pipettes contained several Na+ channels. In contrast, the probability of Na+ channel appearance in inside-out patches depended strongly on the pipette resistance. Therefore, the data from experiments with inside-out patches seems to be more appropriate for estimation of channel densities in the neuronal somata.
Immunochemical staining of the neurones with antibodies against Na+ channels from rat brain showed a localization of Na+ channels in the axon hillock and initial segment of retinal ganglion cells (Wollner & Caterall, 1986). Microscopic autoradiography studies of the specific binding of scorpion toxin to Na+ channels in cultured neurones indicated an approximately seven times higher density of Na+ channels on one neurite initial segment than in the adjacent cell body (Catterall, 1981; Boudier et al. 1985). Here we show that the ratio of somatic to axonal Na+ currents is about 1:6 for dorsal horn neurones in the spinal cord slice. This ratio is the mean value for 3- to 9-day-old rats. In general, there was a tendency for the soma of younger animals to have a higher percentage of inactivating Na+ channels than that of older animals (16% at day 3–4 versus 10% at day 9). Thus, there appears to be a redistribution of Na+ channels between soma and axon hillock during postnatal development of these neurones.
Indirect evidence for the dendritic location of persistent Na+ channels can be found in the paper by French et al. (1990) which shows that the amplitude of persistent Na+ current in enzymatically dissociated guinea-pig hippocampal neurones is approximately 50% of that recorded from intact neurones within slices. The dissociated neurones seemed to have preserved their axon hillocks, since (1) long (<l00 μm) processes remained with the somata after dissociation and (2) inactivating Na+ currents showed large amplitudes exceeding 10 nA (Sah, Gibb & Gage, 1988). It could therefore be assumed that the loss of about 50% of persistent Na+ channels resulted from the loss of the major part of the dendritic tree during enzymatic isolation.
The presence of steady-state Na+ channels in fine dendrites and the axon of dorsal horn neurones can play an important role in the excitability of these cells. Non-inactivating Na+ channels could amplify dendritic excitatory postsynaptic potentials and therefore improve their transmission to the cell soma, as has been suggested for apical dendrites of neocortical neurones (Schwindt & Grill, 1995; Stuart & Sakmann, 1995).
Functional role of soma and axon in action potential generation
It is generally accepted that the axon hillock is the point of initiation of action potentials in neuronal membranes (Eccles, 1964). However, it still remains unclear whether the neuronal soma itself (without axon) is able to generate full action potentials. The present results allow us to conclude that the soma of dorsal horn neurones cannot itself generate action potentials. It contains only a small portion (13.8%) of inactivating Na+ channels but a much larger population of inactivating as well as delayed-rectifier K+ channels. Even in those ‘somata’ in which the leakage current after isolation was reduced more than the Na+ current, no all-or-nothing action potentials were generated, probably due to a shunt of membrane depolarization by activating K+ channels. In the ‘somata’ with a larger percentage of inactivating Na+ channels (15–20%), some local responses could be observed, suggesting that somatic Na+ channels may only support propagation of an action potential from the axon to the cell soma and dendrites (Stuart & Sakmann, 1994). Therefore, the axon or its initial segment, with 6/7 of the total inactivating Na+ current, is necessary for generation of all-or-nothing action potentials in dorsal horn neurones.
We would like to thank Dr M. E. Bräu for critically reading the manuscript and Mrs B. Agari and Mr O. Becker for excellent technical assistance. The work was supported by the Deutsche Forschungsgemeinschaft (DFG Vol88/16).