Encoding properties induced by a persistent voltage-gated muscarinic sodium current in rabbit sympathetic neurones

Abstract

1. A time- and voltage-dependent Na⁺-selective current termed I_Na,M is activated by muscarinic agonists or splanchnic nerve stimulation in sympathetic neurones of rabbit coeliac and superior mesenteric ganglia. The firing patterns induced by I_Na,M were investigated in patch-clamped neurones within intact ganglia, and compared with those generated by a neuronal model including I_Na,M.

2. I_Na,M was characterized by voltage-dependent low-threshold activation and high-threshold inactivation functions. The overlapping functions produced a persistent U-shaped current between −100 and −20 mV, which peaked at the cell resting potential. The activation and inactivation kinetics were fitted to single exponentials with time constants of ≈100 and 400 ms, respectively.

3. Activating I_Na,M with muscarinic agonists or nerve stimulation depolarized and fired the neurones. The depolarization was paralleled by an apparent increase in input membrane resistance. The model showed that this paradox resulted from the turning off of I_Na,M during resistance tests, which also accounted for the all-or-none slow hyperpolarizing responses to current pulses.

4. I_Na,M gave the neurones an N-shaped I-V relationship capable of producing complex firing patterns. Under given conditions, carbachol-treated neurones could either fire regularly or remain silent at ≈-80 mV, i.e. they displayed bistability. Transitions from one state to the other were triggered with short current pulses. The transitions resulted from the turning on and off of I_Na,M.

5. Firing reduced I_Na,M, an effect abolished by blocking Ca²⁺ channels or adding BAPTA (40 mM) to the pipette. The Ca²⁺-related negative regulation of I_Na,M may have mediated endogenous bursting activity. Burst firing was generated by the model upon introducing Ca²⁺ regulation of I_Na,M.

6. The results demonstrate that I_Na,M gives prevertebral sympathetic neurones a wide repertoire of firing patterns: pacemaker-like properties, bistability and burst firing capability. They suggest that the I_Na,M-related encoding properties may provide sympathetic neurotransmission with new potentialities.

Prevertebral sympathetic ganglia are anatomical relays on the efferent pathways controlling gastrointestinal functions. The synaptic inputs of central origin to the ganglionic neurones are mainly mediated by acetylcholine. This mediator triggers both fast and slow depolarizations by acting on nicotinic and muscarinic receptors, respectively. The muscarinic depolarization of prevertebral sympathetic ganglion neurones observed in various species has been mainly ascribed to the inhibition of various potassium currents, such as leak and M-like K⁺ currents (Vanner, Evans, Matsumoto & Surprenant, 1993; Coggan, Putnyn, Knoper & Kreulen, 1994; Delmas, Niel & Gola, 1996) and inwardly rectifying K⁺ current (Wang & McKinnon, 1996). Whole-cell patch-clamp experiments carried out on non-dissociated rabbit coeliac and superior mesenteric ganglion neurones (Gola, Niel, Bessone & Fayolle, 1992) have shown, however, that muscarinic excitation occurring in these neurones resulted mainly from the activation of a cationic current which was termed I_Na,M. This current is characterized by its high Na⁺ selectivity, its U-shaped voltage dependence (Delmas et al. 1996) and its coupling to muscarinic receptors via a BAPTA-resistant and pertussis toxin-insensitive G-protein-mediated pathway (Delmas et al. 1996; Delmas & Gola, 1997). I_Na,M has been involved in the firing discharge induced by muscarinic agonists (Delmas et al. 1996), and is responsible for the muscarinic slow excitatory postsynaptic potentials evoked by stimulating the preganglionic splanchnic nerves (Niel, Delmas & Gola, 1996; Delmas & Gola, 1997). It may also sustain the firing of spontaneously active coeliac neurones (Gola & Niel, 1993), which has been found to result from a tonic presynaptic release of acetylcholine in intact ganglia (Niel et al. 1996).

Receptor-operated inward currents similar to I_Na,M have been described in various neurones of the mammalian central nervous system (for a review, see Delmas, Raggenbass & Gola, 1997). They are induced in response to vasopressin in facial and hypoglossal neurones (Raggenbass, Goumaz, Sermasi, Tribollet & Dreifuss, 1991; Palouzier-Paulignon, Dubois-Dauphin, Tribollet, Dreifuss & Raggenbass, 1994), to oxytocin in rat vagal neurones (Dreifuss, Dubois-Dauphin, Widmer & Raggenbass, 1992; Raggenbass & Dreifuss, 1992), to neurotensin in rat dopaminergic neurones of the mesencephalon (Mercuri, Stratta, Calabresi & Bernadi, 1993) and to acetylcholine (via muscarinic receptors) in rat cortical neurones (Haj-Dahmane & Andrade, 1996). A muscarinic inward current closely resembling I_Na,M is also present in the motoneurones of the lobster cardiac ganglion (Freschi & Livengood, 1989), as well as in those of some insects (Trimmer, 1995). Although they are associated with different neurotransmitter receptors, these currents have several similar properties. They all result from the opening of tetrodotoxin-resistant sodium-selective ion channels and display a characteristic U-shaped steady-state voltage dependence (for review, see Delmas et al. 1997). They are persistently activated in the spike threshold and subthreshold voltage range, and generally do not display any desensitization upon prolonged exposure to agonists. Their activation may therefore result in an N-shaped steady-state current-voltage (I-V) relationship, i.e. part of the I-V curve has a negative slope (Raggenbass et al. 1991; Alberi, Dubois-Dauphin, Dreifuss & Raggenbass 1993; Mercuri et al. 1993; Delmas et al. 1996).

The voltage region corresponding to the negative slope in N-shaped I-V curves is unstable when the cells are observed under current-clamp conditions, which may result in complex firing behaviour. Indeed, it has been amply established both experimentally (review by Benson & Adams, 1987) and in modelling studies (e.g. Epstein & Marder, 1990) that the negative slope, when located in the spike subthreshold region, is potentially able to result in spike burst patterns. Depending on its amplitude and its location relative to the voltage and current axes, the N shape may also result in transient or sustained spike discharges following a brief depolarizing current pulse or even in bistability, i.e. under given conditions the cell may exist in either of two states, silent or firing (Steinberg, 1988). Previous examples of such behaviours have been described in the cat spinal motoneurone (Schwindt & Crill, 1980; Crone, Hultborn, Kiehn, Mazieres & Wigström, 1988; Hounsgaard, Hultborn, Jespersen & Kiehn, 1988) and in some molluscan neurones (Gola, 1976) in which the activation of a persistent or slowly inactivating inward current gives the steady-state I-V curve a negative slope from about −50 to −25 mV.

Although the main effect of I_Na,M activation is that it drives the membrane potential to the spike threshold, it is not yet clear whether it may also give the cell additional encoding properties such as those listed above. It was therefore of interest to investigate this possibility in the framework of the muscarinic transmission in the prevertebral sympathetic ganglia. Since several currents in these neurones are regulated by muscarinic receptors, the task of sorting out the contribution of a particular current was not straightforward. The experimental behaviour of the neurones was therefore compared with that generated by a neuronal model into which I_Na,M was incorporated. It turned out that most of the firing patterns observed in carbachol-treated coeliac neurones can be accounted for by the properties of the Na⁺-selective muscarinic current, which appears to play a major role in the encoding capability of the prevertebral sympathetic neurones.

METHODS

Electrophysiological recordings

Patch-clamp experiments were carried out on non-dissociated sympathetic neurones within the coeliac and superior mesenteric ganglia of adult rabbits (1-1.5 kg). The animals were stunned and killed by exsanguination. All procedures were approved by the French Ministry of Agriculture and were in agreement with the European Communities Council Directive (86/609/EEC). The coeliac tract was quickly dissected out together with the preganglionic splanchnic nerves, and pinned in a Sylgard-coated Petri dish continuously superfused with Krebs solution. The preganglionic nerves were placed on bipolar stimulating electrodes. The neurones were recorded using the whole-cell configuration of the patch-clamp technique. The cleaning procedure that made it possible to apply these techniques to intact ganglion neurones was that previously described (Gola et al. 1992; Gola & Niel, 1993) with the improvements reported by Delmas & Gola (1997).

Patch electrodes were pulled from 1.5 mm diameter borosilicate glass capillaries using a P87 Brown-Flaming-type puller (P87, Sutter Instruments, Novado, CA, USA) and fire-polished to give a tip diameter of 1.5–2 μm. They had a tip resistance of 2–4 MΩ when filled with the following intracellular saline (mM): 140 KCl, 4 NaCl, 2 MgCl₂, 1 CaCl₂, 2 Na₂ATP, 10 BAPTA; 10 Hepes (pH 7.3; osmolarity 320–325 mosmol l⁻¹). In some experiments, CsCl was isosmotically substituted for KCl. The standard external saline had the following composition (mM): 140 NaCl, 5 KCl, 1 MgCl₂, 2.5 CaCl₂, 10 Hepes and 10 D-glucose (pH 7.4; osmolarity 320 mosmol l⁻¹). The properties of I_Na,M were determined in the presence of CsCl (1–2 mM) in order to block an inwardly rectifying K⁺ current present in our preparation (Gola & Niel, 1993) which would have contaminated I_Na,M at voltages negative to −90 mV. In a few experiments where CsCl was omitted, we found no evidence that this inward rectifier was sensitive to muscarinic agonists as was reported to be the case in rat coeliac neurones (Wang & McKinnon, 1996). Any changes in the composition of the solutions are indicated in the text. The cells were continuously superfused with the bath saline (2 ml min⁻¹). All the experiments were performed at room temperature in the presence of the nicotinic receptor antagonists hexamethonium (50 μM) and d-tubocurarine (50 μM). Carbachol (Cch, Sigma) was used as muscarinic agonist. The splanchnic nerves were stimulated with 0.5 ms, 5–10 V pulses.

Membrane current and voltage were measured with a List EPC7 patch clamp amplifier. Except for ramp-generated currents (slope of the voltage ramp, 10–20 mV s⁻¹), leak and capacitive currents (measured using subepisode protocols) were subtracted after appropriate scaling. Voltages were corrected for a 9 mV junction potential (Gola & Niel, 1993).

Modelling the neuronal behaviour

The single compartment model used was based on a Hodgkin- Huxley type spike-generating system. The sodium and potassium currents were classically modelled as follows:

where the equilibrium potentials for Na⁺ and K⁺, E_Na and E_K, are +40 and −90 mV, respectively, and the maximum cell conductances to Na⁺ and K⁺, g*_Na and g*_K, are 10–12 and 5 μS, respectively. The state parameters m, n and h change with time according to the differential equation:

where x stands for m, n and h, and α and β are voltage-dependent rate parameters. The functions describing the voltage dependence of the three sets of rate parameters were strictly identical to those modelled by McCormick & Huguenard (1992).

The leak Na⁺ and K⁺ currents flowed through the respective leak conductances g_Na,leak and g_K,leak; g_Na,leak= 0.2–0.4 nS and g_K,leak= 1.5–2 nS. These values were chosen to obtain a cell input resistance of 400–500 MΩ and a resting potential of −60 mV (Gola & Niel, 1993). In all the simulations illustrated, g_Na,leak= 0.4 nS and g_K,leak= 1.6 nS. The membrane capacitance was set at C_m= 0.22 nF, the value measured in rabbit coeliac neurones (Gola & Niel, 1993). The resulting membrane time constant ranged between 90 and 110 ms.

This minimal model generated train of spikes when stimulated with outward currents larger than 20 pA.

I_Na,M was introduced as a component of the sodium current:

where g*_Na,M is the maximal conductance and m_Na,M and h_Na,M are the steady-state activation and inactivation variables. These are defined in the Results section. In the presence of 25 μM Cch, g*_Na,M ranged between 2 and 3 nS. It is worth noting that g_Na,M was of the same order of magnitude as the leak conductance.

The Ca²⁺ channels of rat and rabbit coeliac neurones are mainly of the N type (Carrier & Ikeda, 1992; Delmas & Gola, 1997). The calcium current was modelled accordingly, taking the constant field equation. In the absence of specific analytical data on the Ca²⁺ channel dynamics in prevertebral ganglionic neurones, we used the rate parameters of Kay & Wong (1987):

This model generated a high-voltage (∼-40 mV) Ca²⁺ current, similar to that of mammalian prevertebral neurones (see Carrier & Ikeda, 1992). The maximal calcium permeability ranged between 5 and 7 cm³ s⁻¹, which induced a maximal Ca²⁺ current of 600–850 pA, as previously determined in rabbit coeliac neurones (Delmas & Gola, 1997). We assumed intracellular calcium to accumulate in a narrow space just beneath the membrane and to diffuse away with a time constant of 1–2 s (see McCormick & Huguenard, 1992).

The numerical solution of the differential equations was achieved using a two-step Euler integration method for preliminary results and then the Runge-Kutta method.

RESULTS

Properties of I_Na,M

The properties of the carbachol-induced and synaptically evoked I_Na,M in rabbit prevertebral sympathetic neurones have been previously described (Delmas et al. 1996; Niel et al. 1996). The recordings in Fig. 1 show the criteria we used to identify I_Na,M. The present experiments were performed on voltage-clamped non-dissociated coeliac neurones in the presence of nicotinic receptor antagonists (see Methods). In Fig. 1A, the cell was held at −70 mV and subjected to either 20 mV hyperpolarizing pulses or slow voltage ramps. At this holding potential, bath-applied carbachol induced an inward current (i.e. I_Na,M) and reversed the voltage pulse-induced current from inward to outward. These characteristic features of I_Na,M are due to its voltage dependence; once induced by muscarinic agonists, I_Na,M recorded at −70 mV was partly turned off by applying hyperpolarizing steps. The quasi steady-state I-V relationship determined with slowly rising voltage ramps therefore became N-shaped in the presence of carbachol. All the carbachol-induced effects were abolished by the muscarinic antagonists atropine and pirenzepine (Delmas et al. 1996; Niel et al. 1996).

Figure 1. Main properties of I_Na,M.

Whole-cell recording from a non-dissociated voltage-clamped sympathetic neurone bathed in physiological saline. A, continuous recording. Lower trace: voltage protocol. Upper trace: current recording. The cell was held at −70 mV and subjected to either slow depolarizing ramps (from −100 to −40 mV in 10 s) or 20 mV, 0.5 s hyperpolarizing pulses. Carbachol (Cch, 25 μM) induced a slowly developing inward current. I_Na,M was voltage dependent, which resulted in (1) an N-shaped I-V relationship (second ramp) and (2) a pulse-induced time-dependent outward-going current jump (shown in right-hand inset) instead of the inward-going leak current in control saline (left-hand inset). B, quasi-steady-state I-V relationships (voltage-ramp method) in a voltage-clamped neurone before (Control) and during Cch application (20 μM). Lower trace: difference current. The Cch-induced current peaked at −55 mV, and vanished between −20 and −30 mV.

The voltage dependence of I_Na,M was determined over a wide voltage range using the voltage-ramp method in neurones lacking M-like K⁺ currents (∼70 % of the neurones; Wang & McKinnon, 1996; Delmas et al. 1996). I_Na,M (bottom trace in Fig. 1B) had a peak value at voltages ranging between −50 and −60 mV. It decreased at more negative potentials, which accounted for the pulse-induced outward current jump shown in Fig. 1A. At more positive potentials, I_Na,M decreased sharply. It tended to vanish between −20 and −30 mV, but failed to reverse. This point has been thoroughly investigated in the study by Delmas et al. (1996), where it emerged that I_Na,M is almost completely carried by Na⁺ ions and that the current decrease in the −50 to −20 mV range might result from an inactivating process. The properties of the inactivation are characterized in the next section.

Steady-state and dynamic properties of I_Na,M

We first examined the characteristics of the inactivation which occurs at voltages positive to −50 mV. The inactivation was not related to the opening of calcium channels and hence to the subsequent increase in the cytoplasmic calcium level, since it remained unchanged in the presence of the calcium channel blockers nifedipine (20 μM) and ω-conotoxin MVIIA (0.2–1 μM) and when high concentrations of BAPTA (up to 40 mM) were present in the pipette saline (Delmas et al. 1996). The inactivation was therefore assumed to be voltage dependent. The steady-state inactivation function was determined in voltage-clamped carbachol-treated cells by recording a test I_Na,M after applying a 2 s conditioning depolarization with an increasing amplitude (Fig. 2A). In these experiments, outward K⁺ currents were blocked by using Cs⁺ (75 mM)-filled patch electrodes. The test current decreased when the conditioning prepulse was in the −50 to 0 mV range. The relationship between the normalized test I_Na,M and the prepulse level could be described by the following Boltzmann function:

Figure 2. Voltage-dependent I_Na,M inactivation.

A, Cs⁺ (75 mM)-loaded neurone bathed with carbachol (20 μM) and subjected to the two-step voltage protocol shown in the top row. A test pulse to −60 mV was preceded (100 ms delay) by a 2 s conditioning prepulse to various levels. The prepulse level is indicated on the test current traces. Leak and capacitive currents have been subtracted. B, relative I_Na,M amplitude (±s.e.m.) as a function of the prepulse level. Data pooled from 8 experiments similar to A. Continuous curve: steady-state inactivation curve h_Na,M (see text). C, slow I_Na,M relaxation during pulse depolarization at voltages positive to −50 mV. Cs⁺-loaded neurone. D, time constant of prepulse-induced inactivation. Relative I_Na,M amplitude (±s.e.m.) as a function of the prepulse duration (prepulse level: 0 mV). Data fitted to a single exponential with time constant τ= 350 ms.

The steady-state inactivation function h_Na,M is displayed as a continuous curve in Fig. 2B.

The inactivation kinetics were obtained with the above voltage protocol by varying the prepulse duration. The inactivation was fitted to single exponential functions (Fig. 2D) with time constants of 300–500 ms. This was confirmed by directly observing the inactivating process in Cs⁺-loaded cells during depolarizations at voltages positive to −50 mV (Fig. 2C). At −70 mV, the recovery from inactivation proceeded exponentially with a ∼400 ms time constant. Although we did not systematically study this point, it turned out that the rates of inactivation and recovery from inactivation were poorly voltage dependent. In the simulation, the time constant of the inactivating process was therefore set at 350–400 ms.

The steady-state activation function m_Na,M was determined from ramp-generated I_Na,M similar to that shown in Fig. 1B. Pooled data obtained on fourteen carbachol (25 μM)-treated neurones are given in Fig. 3A. The continuous curve through experimental points was drawn according to:

Figure 3. Steady-state and kinetics of I_Na,M activation.

A, I_Na,M-V relationship obtained with the slow voltage-ramp protocol as in Fig. 1B. •, mean I_Na,M (±s.e.m.) from 14 neurones. The I_Na,M was obtained by digitally subtracting the I-V curve in the presence of Cch plus atropine (1 μM) from that recorded in the presence of Cch alone. Continuous curve: theoretical I_Na,M calculated from the steady-state curves shown in B. B, steady-state activation (m_Na,M) and inactivation (h_Na,M) curves. Dotted curve: m_Na,Mh_Na,M product. C, Cch-treated neurone, subjected to 500 ms depolarization to −70 mV (upper current trace) and −60 mV (lower current trace). Current activation during the pulse and current relaxation after the pulse were fitted to single exponentials (superimposed dashed curves). Leak current subtracted. D, time constant of activation (^) and deactivation (•) as a function of the membrane potential. The continuous curve was drawn based on the time constant-voltage relationship given in the text. Data pooled from 6 experiments performed as in C.

where m_Na,M is a Boltzmann function describing the voltage-dependent steady-state activation. A least-squares program gave the m_Na,M(V) function and the conductance g*_Na,M:

and

The synaptically evoked I_Na,M had a mean peak amplitude amounting to about one-third of the current generated by 25 μM carbachol (Delmas & Gola, 1997). The corresponding g*_Na,M was therefore set at 0.8–1 nS.

The activation and inactivation functions are displayed in Fig. 3B along with the product m_Na,Mh_Na,M which shapes the steady-state I_Na,M.

The activation and deactivation time course of I_Na,M was fitted to single exponentials (Fig. 3C). The activation time constant tended to decrease at voltages above and below −70 to −80 mV (Fig. 3D). It was modelled as a bell-shaped function (continuous curve in Fig. 3D) according to:

Efficiency of I_Na,M

Activating I_Na,M with carbachol (10–25 μM) fired 93 % of the neurones (n = 331; Gola & Niel, 1993; Delmas et al. 1996). When preganglionic nerve stimulations were applied, slow atropine-sensitive depolarizations were induced in 91 % of the neurones tested (n = 72; Delmas & Gola, 1997). Prior to I_Na,M activation, these neurones had a resting potential of −57 ± 1 mV. Due to the negative location of the activation function h_Na,M, we hypothesized that I_Na,M-mediated depolarizations might still effectively fire much more polarized silent neurones. Large resting potentials of up to −100 mV were observed when the pipette solution contained 20–30 mM Na⁺ (substituted for K⁺). This large polarization resulted presumably from the activation of the electrogenic Na⁺,K⁺-ATPase, since normal resting potentials in the −50 to −60 mV range were then obtained by adding ouabain (50 μM) or bathing the cells in a K⁺-free saline (data not shown). The results presented in Fig. 4A and B show that activating I_Na,M with either splanchnic nerve stimulation (Fig. 4A) or carbachol (Fig. 4B) still triggered spike discharges in these highly polarized neurones (initial resting potential, −90 to −95 mV, n = 7). Similar results were obtained when the resting potential was negatively shifted to the same voltage range by lowering the bath K⁺ from 5 to 0.5 mM (n = 5, Fig. 4C). With the standard values defined in the Methods section, the neuronal model generated persistent spike discharges in cells polarized near −100 mV, provided that g*_Na,M was larger than 1 nS (not illustrated).

Figure 4. Efficiency of I_Na,M in firing hyperpolarized neurones.

A and B, these neurones had a high resting potential (-90 to −95 mV) due to the activation of the electrogenic Na⁺, K⁺-ATPase. A ≈40 mV depolarization was required to reach the spike threshold (left-hand recording in A). Preganglionic splanchnic nerve stimulation (Spl.st., single shock, 0.5 ms, 10 V, A) and Cch (B) were still able to fire the hyperpolarized neurones. C, pacemaker-like activity of a Cch (20 μM)-treated neurone. Lowering the bath K⁺ concentration from 5 to 0.5 mM failed to stop the spike discharge. Note the resulting large post-spike hyperpolarization (-90 mV) shown in the right-hand expanded time scale. In this and subsequent figures, dashed lines are zero voltage and current levels.

Apparent input resistance changes

The carbachol-induced depolarization was paralleled by an apparent transient increase in the cell input resistance, as evaluated under current-clamp conditions from the increase in the voltage change in response to current pulses (Fig. 5A; also Delmas et al. 1996). The voltage response increase could be interpreted as an increase in membrane resistance, which appeared not to be consistent with the activation of a Na⁺ current. This point was examined by gradually introducing g*_Na,M into the neuronal model. With g*_Na,M= 2–2.5 nS (mimicking the effect of 25 μM Cch), the neuronal model depolarized and fired at 5–7 Hz (Fig. 5B). Applying a current pulse to test the input resistance resulted in a similar pattern to that observed experimentally (Fig. 5A): the apparent input resistance first increased during the depolarization and then decreased when the cell reached the spike threshold (Fig. 5C). Upon examining the changes in g*_Na,M during these periods, it was observed that the apparent increase in the resistance was actually due to the voltage-dependent deactivation of the g*_Na,M, which tended to restore the pulse-induced voltage to its control level. The duration of the period of apparent resistance increase was dependent on the amplitude and duration of the current pulse. With large, long-lasting (1–2 s) current pulses, this phenomenon persisted during the firing period. Under these conditions, the neurones (both experimental and model) might generate slowly developing hyperpolarizing responses.

Figure 5. Input resistance changes induced by I_Na,M.

A, sustained spike discharge in a coeliac neurone in response to 10 μM bath-applied Cch. The Cch-induced depolarization was associated with an apparent increase in the input resistance as assessed from the voltage change induced by 50 pA inward current pulses (lower trace). B, sustained spike discharge in the neuronal model resulting from the gradual incorporation of the conductance g*_Na,M; final value of g*_Na,M= 2.5 nS (lower traces). The trace labelled g_Na,M is the instantaneous activation level of the conductance. Inset: steady-state I-V curve before and after introducing the Na⁺-selective muscarinic conductance (trace labelled I_Na,M). A permanent inward current (35 pA) was applied to set the initial membrane potential at −80 mV in order to prolong the depolarizing period preceding the tonic firing. C, same parameters as in B. Superimposed test current pulses: 80 pA (upper trace). The apparent increase in the input resistance prior to the spike discharge actually resulted from the hyperpolarization-induced deactivation of g_Na,M (lower trace).

Hyperpolarizing responses

I_Na,M-related hyperpolarizing responses to inward current pulses have been observed in both spontaneously firing (Gola & Niel, 1993; Niel et al. 1996) and carbachol-treated coeliac neurones (Delmas et al. 1996). These responses are characterized by a triggering current (or voltage) threshold (Fig. 6Aa), and by their slow time course (Fig. 6B). The hyperpolarizing response became faster when the current pulse amplitude was increased. Following the current pulse, the cell returned to its initial state after a variable delay. Depending on the amplitude of I_Na,M, this delay could last for several seconds, during which the cell was in a metastable state. The model neurone behaved in the same way (Fig. 6C). With g*_Na,M= 2.5 nS, the hyperpolarizing response occurred at a critical current pulse of 170 pA (70 pA with g*_Na,M= 1 pS). This current threshold corresponded to the peak inward current in the steady-state I-V curve (Fig. 6Ab) generated by introducing g*_Na,M (inset in Fig. 5B). The slow time course of the response was not directly correlated with the rate of deactivation of I_Na,M. In the voltage range where the response developed, the current deactivation rate (50–100 ms) was faster than the cell time constant (90–110 ms). The slowness of the response actually resulted from the dynamics of the voltage and conductance changes which occurred in current-clamped cells. Owing to its moderate sensitivity to voltage changes, the current-induced hyperpolarization deactivated a small part of the active g*_Na,M, which slowly accelerated the cell hyperpolarization. The process continued in a self-reinforcing manner until I_Na,M was completely deactivated. This gave the hyperpolarizing response its all-or-none behaviour and slow sigmoidal time course, which differed from the graded exponential response of silent cells to current pulses (see Figs 1 in Delmas et al. 1996 and Niel et al. 1996).

Figure 6. Hyperpolarizing responses.

A and B, Cch-treated (10 μM) neurones. Aa, step increase in injected hyperpolarizing current (upper trace) gradually slowed down the firing rate, up to a critical value at which a large hyperpolarization resulted. Ab, same neurone as in Aa. Current (upper trace) induced by a slowly rising (10 mV s⁻¹) voltage ramp from −120 to −40 mV (lower trace). The two arrows indicate the relationship between the hyperpolarizing response and the I-V curve. B, typical slow time course of the hyperpolarizing response. C, neuronal model. g*_Na,M= 2.5 nS. Current pulses (upper trace) 150, 200 and 250 pA. Note the similarity between the time course of the hyperpolarizing responses induced by the two largest current pulses and that shown in B.

Bistable states

The hyperpolarizing response showed that current-clamped carbachol-treated neurones could not settle in the voltage range (∼-80 to −50 mV) corresponding to the negative slope in the I-V curve. Depending on the experimental conditions, the membrane potential of these neurones might lie either above (positive to −50 mV) or below (negative to −80 mV) the unstable voltage region. Since the first level was positive to the spike threshold, it drove the pacemaker-like activity.

These findings suggest the possibility that carbachol-treated cells may exhibit bistable behaviour, i.e. might have two different states under the same experimental conditions. The only prerequisite for this behaviour is that the N-shaped I-V curve must intersect the current baseline at three points. This was obtained by injecting a moderate (30–60 pA) hyperpolarizing current. An example of these experiments is given in Fig. 7A relative to a carbachol-treated neurone permanently injected with an inward current. With a holding current of 20 pA, the cell continued to fire. The hyperpolarizing response induced by a short current pulse was followed by a transient silent period. The cell could stay in this metastable state, at about −80 mV, for several tens of seconds before spontaneously returning to the firing state. Upon slightly increasing the holding current to 40 pA, the cell either fired continuously or stayed quiescent at −80 to −90 mV (Fig. 7A). Shifts from the one state to the other were induced by applying short current pulses with appropriate amplitudes and directions. This behaviour was mimicked by the neuronal model (Fig. 7C), which showed that the current pulse had to be large enough to allow the passage over (silent to pacemaker transition) or below (reverse transition) the negative slope region. The model also showed that the shift to the hyperpolarized state had the properties of the hyperpolarizing response illustrated in Fig. 6.

Figure 7. Bistability.

A, carbachol-treated (10 μM) neurone permanently injected with 40 pA inward current. Under these conditions, the neurone either fired regularly or remained silent at ≈-85 mV. Transitions between these two states were induced by applying short hyperpolarizing or depolarizing pulses. B, neurone model; g*_Na,M= 2.5 nS. Permanent inward current: 80 pA.

Burst firing

In several instances, setting the holding current around the value generating hyperpolarizing responses led the regular firing discharge of carbachol-treated neurones to develop into burst firing patterns (Fig. 8A). These patterns were characterized by successive bursts of action potentials and silent intervals, each lasting for several (5-15) seconds. Contrary to what was observed in the case of the bistable behaviour, the transitions between the firing and silent states occurred spontaneously. The transitions may have been triggered by synaptic inputs of unknown origin. Alternatively, they may also have resulted from intrinsic membrane properties. This means that the shape of the steady-state I-V curve might depend on the state of the cell in such a way that the firing suppresses I_Na,M, whereas the current recovers during the rest period (Epstein & Marder, 1990). Since it has been established that I_Na,M is negatively regulated by an increase in intracellular calcium (Delmas et al. 1996), we tested whether a firing-associated Ca²⁺ influx might therefore play a role in shaping the I-V relationship. Hybrid current- and voltage-clamp experiments showed that the negative slope in the I-V curve was consistently attenuated when a voltage ramp was applied just after a 20 s spiking period (Fig. 8B). This effect was abolished by voltage clamping the cell at −80 mV for 20–30 s. The current suppressed by the firing activity had the U-shaped voltage dependence expected from the Ca²⁺ sensitivity of the I_Na,M. The same procedure had no effect on the N-shaped I-V curve in cells heavily loaded with the calcium chelators EGTA or BAPTA (up to 40 mM; Delmas et al. 1996) or in the presence of the N- and L-type calcium channel blockers nifedipine (20 μM) and ω-conotoxin MVII A (1 μM) (data not shown).

Figure 8. Burst firing.

A, carbachol-treated neurone. Permanent inward current: 60 pA. Spontaneous transitions between the two states resulted in burst firing. B, currents induced by a voltage ramp from −100 to −20 mV (slope, 20 mV s⁻¹) before (Control) and after adding 25 μM Cch. The two current traces in Cch were drawn immediately after either a 20 s spiking period (Cch + spiking, freely firing unclamped cell) or a 30 s voltage clamp period at −80 mV (Cch + rest). C, neuronal model. An N-type calcium current and a negative feedback of intracellular calcium ([Ca²⁺]_i trace) to g_Na,M were added to the model (see text for details). Regular spike bursts were induced upon injecting 120 pA inward current.

We therefore assumed that calcium flowing through N-type Ca²⁺ channels blocked I_Na,M. The Ca²⁺ channels were incorporated in the model (see Methods). The calcium unbinding rate, k_off, was set at 0.2 s⁻¹ (corresponding time constant of 5 s), which accounted for the 20–30 s recovery period mentioned above. Upon adjusting the dissociation constant, K_D, between 0.5 and 1 μM and the level of the permanently injected current (80–130 pA), the model generated repetitive bursts of spikes (Fig. 8C). The burst period was dependent on both the rate of calcium removal and the holding current.

DISCUSSION

In the present study, we examined the contribution of the muscarinic inward current, I_Na,M, to the signalling capacity of the rabbit prevertebral sympathetic neurones. The muscarinic responses observed were compared with those of a spike-generating model into which I_Na,M was incorporated. In this way, the effects of I_Na,M on neuronal behaviour were specifically characterized by manipulating its properties. The results show that muscarinic activation of the low-threshold I_Na,M gives sympathetic neurones a wide repertoire of firing patterns, including pacemaker-like properties, bistable behaviour and burst firing capability. These potentialities are mainly due to the voltage-dependent gating of the muscarinic Na⁺-dependent (Na,M) channel which occurs in the spiking subthreshold voltage range.

Biophysical properties of I_Na,M

The muscarinic Na⁺-dependent current (I_Na,M) recorded in neurones of intact coeliac and superior mesenteric ganglia exhibits a U-shaped voltage dependence, peaking at the resting membrane potential and disappearing on each side at about −100 and −20 mV (Delmas et al. 1996; see also Fig. 1B). We attempted here to characterize further the voltage dependence of this current under conditions (internal and external Cs⁺) which minimized the involvement of K⁺ channels in the muscarinic modulation. This procedure did not affect the properties of I_Na,M, since Na,M channels are neither sensitive nor permeant to caesium (Delmas et al. 1996).

Although the decrease in I_Na,M at very negative membrane potentials resulted simply from the voltage-dependent turning off of the current, the nature of the null-current voltage in the suprathreshold voltage range still remained to be determined. Since this null-current voltage was insensitive to changes in [Na⁺]_o and [K⁺]_o (Delmas et al. 1996), it did not reflect the actual equilibrium potential of I_Na,M. In fact, I_Na,M failed to reverse from inward to outward at more depolarized levels (see Fig. 1B). We have previously proposed that it might result from an inactivating mechanism. To settle this question, I_Na,M was evoked by a test pulse applied after a conditioning depolarizing step to various levels. I_Na,M was gradually reduced in a time- and voltage-dependent manner when the conditioning depolarization entered the −50 to 0 mV range. The voltage-dependent decrease in I_Na,M did not depend on Ca²⁺ influx, since blocking N- and L-type Ca²⁺ channels (which carry I_Ca in prevertebral sympathetic ganglion neurones; Delmas & Gola, 1997) had no effect on this process. Likewise, the inactivation persisted when [Ca²⁺]_i changes were limited using high BAPTA concentrations in the patch electrode. These results clearly indicate that the decrease in I_Na,M observed at suprathreshold voltages resulted from a voltage-dependent inactivation, which reinforces our previous hypothesis that I_Na,M might be an almost pure Na⁺ current. The possibility that a minor contribution from K⁺ ions might be involved could not, however, be ruled out, since technical limitations inherent to intact voltage-clamped neurones made it impossible to evaluate the I_Na,M reversal potential in the positive voltage range (Delmas et al. 1996).

I_Na,M was therefore modelled by assuming that it was purely a Na⁺ current, and that, once activated by muscarinic agonists, the Na,M channel may be in one of the three closed, open and inactivated states, depending upon the membrane potential. The rate transitions between these three states were found to obey first order processes. The steady-state voltage dependence of the activation (closed-open equilibrium) and inactivation (open-inactivated equilibrium) were well-fitted by Boltzmann equations. It should be noted that although overestimating the value of the I_Na,M reversal potential slightly affected the Boltzmann activation parameters (but not the inactivation ones), it did not qualitatively affect the outcome of the simulation. The respective locations of the activation and inactivation curves defined a large current window (steady-state activation) around the resting potential of the neurone, which placed I_Na,M in a key position to control the neuronal signalling (see below).

I_Na,M-induced electrophysiological behaviours

Increase in input membrane resistance and hyperpolarizing responses

The depolarization induced by activating I_Na,M was associated with an apparent increase in membrane resistance, as evaluated by injecting inward current pulses (Fig. 5). The simulation showed that the paradox arises entirely from the voltage dependence of I_Na,M which turns off within tens of milliseconds during the test pulse, increasing the efficiency of the effect of the current pulse on the membrane potential. Although I_Na,M activation was associated with an increase in the chord conductance, it therefore resulted in a decrease in the slope conductance. A similar apparent increase in membrane resistance resulting from activation of a voltage-dependent cation current closely resembling I_Na,M has been described in rat cortical neurones exposed to muscarinic agonists (Haj-Dahmane & Andrade, 1996). Other examples are to be found in rat facial motoneurones (Raggenbass et al. 1991) and in neurones of various molluscan species (see e.g. Matsumoto, Sasaki, Sato, Shozushima & Takashima, 1988), in which vasopressin- and cAMP-activated cationic currents display the same voltage dependence as I_Na,M. Thus, when observed under current-clamp conditions, the activation of I_Na,M and related currents resulted in the same membrane impedance change as would be expected from the decrease in potassium conductance. In addition, since these currents vanish at voltages close to E_K, it is possible that depolarizations driven by these currents might have been mistakenly thought to be mediated by a decrease in potassium conductance (see also, Haj-Dahmane & Andrade, 1996). Similar misinterpretations were also made in respect of the initial work on NMDA currents (Engberg, Flatman & Lambert, 1979) which generate a ‘negative resistance’ region in the cell's I-V relationship (Nowak, Bregestovski, Ascher, Herbet & Prochiantz, 1984) due to the voltage-dependent current blockage by Mg²⁺ ions.

The hyperpolarization-induced shut off of I_Na,M also explains the genesis of slowly developing all-or-none hyperpolarizations observed in response to inward current pulses in carbachol-treated neurones. The hyperpolarizing response unmasked the unstable voltage region, corresponding to the I-V negative slope, in which unclamped cells cannot settle. As shown by the simulation (see Fig. 6C), an appropriate current that brings the membrane potential to the level (∼-60 mV) of the inward peak in the I-V curve deactivates part of I_Na,M and this in turn hyperpolarizes the cell in a self-reinforcing manner until I_Na,M is completely deactivated. Similar slowly developing hyperpolarizing responses have been observed in molluscan (Gola, 1976) and mammalian neurones (Schwindt & Crill, 1980) in which a persistent inward current results in an N-shaped I-V curve. In the present study, hyperpolarizing responses were obtained by injecting inward current pulses of a few tens of picoamperes. It is conceivable that inhibitory synaptic inputs, which have been recorded in prevertebral sympathetic neurones in response to peripheral nerve stimulation (Ma, Dun & Jiang, 1983; Hankins & Dray, 1988), may be large enough to trigger responses of this kind. If so, this mechanism would provide an ingenious system for amplifying the inputs, switching the neuronal activity from a regularly discharging mode to a highly hyperpolarized silent state.

Bistability and bursting

The above reasoning also applies to the bistable behaviour we observed in the presence of carbachol. Bistability requires that the holding current intersects the negative slope of the I-V curve (Schwindt & Crill, 1980). This was achieved experimentally by injecting a sustained hyperpolarizing current. Under these conditions, the neurone fired regularly with a relatively low frequency and could be silenced by a brief ‘off’ stimulus as long as an ‘on’ stimulus induced the reverse transition and restored the firing. The model showed that the transition between the two states resulted from the voltage-dependent deactivation/reactivation of g_Na,M. Owing to the interplay between the cell polarization and I_Na,M, changes in membrane potential induced by transient events are therefore capable of durably influencing the cell's behaviour in a similar way to that described by Crone et al. (1988) in cat spinal motoneurones. This raises the possibility that fast excitatory (e.g. cholinergic) and inhibitory inputs occurring while I_Na,M is activated may play a switching role of this kind.

Bursting is an extreme example of the effects that can arise from the interactions between the intrinsic properties with which neurones are endowed by the synaptically activated I_Na,M and additional incoming synaptic inputs. Alternatively, self-sustained bursting may also result from the negative regulation that intracellular Ca²⁺ exerts on I_Na,M (Delmas et al. 1996). We observed that firing reduced g_Na,M and that this effect was dependent on Ca²⁺ influx and on the subsequent increase in the intracellular Ca²⁺ level. Incorporating this negative feedback loop into the neuronal model made it possible to generate regular burst firing patterns. This mechanism resembles others described in molluscan nerve cells in which bursting activity resulted from the interplay between Ca²⁺ and a cAMP-activated Na⁺ current reminiscent of I_Na,M (Huang & Gillette, 1993). Although the added parameters had realistic values, this simulation was not intended to describe precisely the events occurring between Ca²⁺ and Na,M channels. Our main aim was to emphasize that the bursting ability given by the N-shaped I-V curve (Benson & Adams, 1987) was backed up by I_Na,M regulation exerted by intracellular Ca²⁺(Nowak et al. 1982). In this respect, the burst generation is phenomenologically similar to that occurring in a number of NMDA-treated neurones, although the basic ionic mechanisms differ. In both cases, the burst generation is favoured by setting the cells at hyperpolarized levels (see e.g. Durand, 1993) within the ‘negative resistance’ region.

I_Na,M and muscarinic inhibition of K⁺ currents

Although the activation of I_Na,M appeared to be the main mechanism responsible for the slow muscarinic excitation occurring in intact rabbit coeliac neurones, several other mechanisms have been postulated in prevertebral neurones of many species, particularly those involving inhibition of leak, M-like and inwardly rectifying K⁺ currents (see introduction). The exact participation of these various mechanisms in muscarinic depolarization of prevertebral neurones still remains to be determined and there exist discrepancies in this respect among the few available studies. These discrepancies are probably due to differences between the animal species studied and the recording techniques and methods of cell preparation used.

In some rabbit coeliac ganglion neurones (∼30 %; Delmas et al. 1996; see also Wang & McKinnon, 1996) M-current inhibition and I_Na,M activation contribute together to the muscarinic responses. Here the intriguing question arises as to why these neurones express two conductances which perform overlapping functions. The answer probably relates to the voltage range in which these currents are present, and to their efficiency in depolarizing the neurones. The M-current is a non-inactivating voltage-dependent K⁺ current active at voltages positive to ∼-65 mV, which stabilizes voltage around the resting potential and brakes the cell excitability (Brown & Adams 1980; Adams, Brown & Constanti, 1982). Its inhibition by various neurotransmitters (reviewed by Brown, 1988; Brown et al. 1997) consequently makes the cell more likely to fire. However, due to its modest activation near rest, its inhibition is unlikely to cause large depolarizations (Lamas, Selyanko & Brown, 1997) and to drive sustained action potential discharges as does I_Na,M (which peaks at rest), but rather provides a switch between phasic and tonic firing properties (Wang & McKinnon, 1996). In addition, the M-current cannot operate in cells polarized beyond −60 mV, a voltage range where I_Na,M is able to depolarize and fire neurones. It is therefore suggested that muscarinic activation of I_Na,M and blockade of the M-current act sequentially and synergistically in depolarizing and firing prevertebral sympathetic neurones. Other possible explanations of the presence of these two currents in the same cell might relate to the existence of specialized synapses, differences in sensitivity between the currents to neurotransmitters, or the involvement of distinct intracellular regulatory pathways (Delmas & Gola, 1997).

Conclusion

The present findings show that the Na⁺-selective muscarinic current dominates the neuronal behaviour of rabbit prevertebral ganglia. This conclusion was based on the results of in vitro experiments performed on non-dissociated ganglia in which the synaptic contacts were preserved, so that I_Na,M could be activated by either adding muscarinic agonists or stimulating the preganglionic nerves. In both cases, the ganglionic neurones developed the same wide repertoire of activities. The efficiency of the synaptically evoked I_Na.M suggests that the encoding properties with which the neurones are provided by this current may endow the ganglionic transmission with considerable new potentialities. The muscarinic regulation of mammalian prevertebral neurones in situ is still poorly documented so far, and much more research needs to be carried out on the natural patterns of impulse flow (Jänig & McLachlan, 1992) before we will be able to understand the functional significance of the findings reported here.