INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Central pattern generators (CPGs) are fundamental neuronal circuits that underlie rhythmic motor behavior in both vertebrate and invertebrate animals (Marder and Calabrese 1996). Understanding how both the endogenous properties of individual interneurons and their synaptic connectivity contribute to the generation of the rhythmic motor output is a major objective in CPG research. Presently, this level of understanding has only been achieved in a limited number of systems, in particular, the crustacean stomatogastric system (Harris-Warrick et al. 1992; Selverston and Moulins 1985). The present study examined the feeding CPG in the snail, Lymnaea. The neuronal elements of this CPG have been identified and their synaptic connectivity is well described at the systems level, but relatively little is known about their endogenous properties. Our present knowledge (Benjamin and Elliott 1989; Brierley et al. 1997b) of how the feeding pattern is generated has been inferred from the firing properties of CPG neurons in the intact CNS, but it is not yet clear to what extent these firing patterns are endogenous or partly/wholly the result of interactions with other neurons.

Establishing the endogenous firing properties of CPG neurons in the intact CNS is difficult against the background of synaptic inputs from other cells and in the presence of complex types of intrinsic or extrinsic chemical modulation (Harris-Warrick and Marder 1991). Electrical properties that initially appear to be purely endogenous, such as bursting, plateau potential generation, or postinhibitory rebound (PIR), are often "conditional" with their occurrence and/or magnitude dependent on the presence of chemical modulators (Kiehn and Katz 1999). One important strategy for identifying truly endogenous properties of neurons is to study them after isolation from the CNS. This approach has been used successfully in a few studies in both vertebrate and invertebrate motor systems (e.g., Dale 1997; Panchin et al. 1993; Syed et al. 1990; Turrigiano and Marder 1993). Here, we adopted this strategy to investigate the Lymnaea feeding circuit. In this system the presence or absence of endogenous properties of CPG interneurons can be confirmed in the intact ganglion, and their function determined in relation to network properties resulting from known synaptic connections.

The Lymnaea feeding CPG interneurons can be divided into three main classes, N1, N2, and N3, according to their phase of activity (protraction, rasp, and swallow) in the feeding pattern. Each of these classes consists of two subtypes: medial N1 (N1M) and lateral N1 (N1L; Yeoman et al. 1995), dorsal N2 (N2d) and ventral N2 (N2v; Brierley et al. 1997a), phasic N3 (N3p) and tonic N3 (N3t; Elliott and Benjamin 1985a). Although several of these cells appear to have important endogenous properties like rhythmic bursting (N1M; Rose and Benjamin 1981b) and plateau potential generation (N2v; Brierley et al. 1997a), these properties have never been explicitly established. Here we have studied these properties in isolated neurons. As pattern generation in the feeding CPG relies strongly on synaptic interactions, we also tested whether synaptic effects could be mimicked by application of acetylcholine (ACh) and glutamate, the main neurotransmitters involved in Lymnaea feeding CPG function (Brierley et al. 1997c; Elliott and Kemenes 1992; Vehovszky and Elliott 1995; Yeoman et al. 1993). For the first time we are now able to present a model for feeding pattern generation in the intact circuit that takes into account the endogenous electrical and pharmacological properties of all feeding CPG interneurons.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Experimental subjects and chemicals

Adult specimens of Lymnaea stagnalis were obtained from Blade Biological (Kent, UK). The animals were kept in large holding tanks containing copper-free water on a 12:12 h light/dark cycle and fed lettuce three times a week.

All chemicals were purchased from Sigma (Poole, UK) unless otherwise stated.

Dissection, intracellular recording techniques, and labeling of interneurons in the intact CNS

All dissections were carried out in HEPES-buffered saline containing (in mM) 50 NaCl, 1.6 KCl, 2 MgCl₂, 3.5 CaCl₂, and 10 HEPES at a pH of 7.9 in distilled water. The CNS, consisting of the circumesophageal ganglionic ring (cerebral, pedal, pleural, parietal, and visceral ganglia) and the buccal ganglia together with a short stretch of esophagus, was isolated from the snail. The preparation was pinned down in a Sylgard-coated dish filled with HEPES-buffered saline with the dorsal surface facing up.

Standard intracellular recording techniques were used to record simultaneously from up to four individual neurons. The recording electrodes were pulled from 2-mm glass capillaries with inner filament (GC200F-15; Clarks Electromedical, Reading, UK) and filled with 4 M potassium acetate (electrode resistance, 30-50 M). Signals were fed into standard intracellular recording amplifiers and visualized on an oscilloscope (5115 Tektronix). Permanent records were obtained using a chart recorder, while the data were also stored on a DAT recorder (Biological DTR-1801; Biological Science Instruments, Claix, France). Penetration of neurons was facilitated by incubation of the isolated CNS in a protease solution (Sigma type XIV; 1 mg/ml in HEPES-buffered saline) for 5 min. The protease treatment was stopped by extensively washing the preparation with HEPES-buffered saline. After electrophysiological identification, interneuronal somata were injected iontophoretically with Fast Green (0.1% in 50 mM KCl), which allowed visual observation of marked neurons during the various stages of dissection. Without labeling, it is impossible to be certain about the identity of small neurons throughout the isolation process prior to cell culture. Fast Green was chosen as a marker dye because it has been used extensively in invertebrate neurophysiology to visualize the successful intracellular injection of material into single cells without affecting their cellular properties (e.g., Lewin and Walters 1999). Furthermore, we have used it previously to label the Lymnaea B7nor neuron for isolation (Park et al. 1998). After isolation, these neurons readily regenerated neuritic processes comparable to other types of Lymnaea buccal motoneurons and their endogenous electrophysiological properties resembled those in the intact CNS.

Identification of SO and CPG interneurons

Extensive work on the feeding system of the pond snail Lymnaea stagnalis has led to the identification of many of its components. This paper concentrates on the endogenous properties of the slow oscillator (SO), an influential modulatory interneuron in the feeding system, and the six previously identified members of the feeding CPG (N1M, N1L, N2d, N2v, N3p, and N3t). The location of these seven cell types in the paired buccal ganglia is shown in Fig. 1A. In the isolated CNS preparation, fictive feeding activity was induced by constant depolarization of the SO (e.g., Figs. 2Ai, 3A, and 4A), and the different types of CPG interneurons were identified by their characteristic firing patterns (Fig. 1C; Brierley et al. 1997b; Elliott and Benjamin 1985a; Rose and Benjamin 1981b; Yeoman et al. 1995). These SO-driven fictive feeding rhythms in isolated preparations are comparable to sucrose-induced feeding activity in the intact animal (Kemenes et al. 1986). The ability of the SO neuron to drive fictive feeding in the isolated CNS is mainly due to its excitatory connection with the N1M interneurons (Rose and Benjamin 1981a), although the SO's biphasic synaptic connection to N2v interneurons is also important in triggering plateau potentials in these cells (Fig. 1B; Brierley et al. 1997b). A complex set of reciprocal synaptic connections occur between the N cells (Fig. 1B) that determine the sequence of firing within the CPG network (Fig. 1C; reviewed in Brierley et al. 1997b).

View larger version (32K): [in this window] [in a new window]   Fig. 1. Location, synaptic connectivity, and firing patterns of Lymnaea feeding central pattern generator (CPG) interneurons. A: map of the Lymnaea buccal ganglia showing the dorsal (left) and ventral (right) surface. Closed circles indicate positions of cell bodies of each interneuron type. For reference purposes, open circles show position of some large, easily identifiable motoneurons on the ventral surface. The slow oscillator (SO) can be located either in the left or right buccal ganglion. l/vbn, lateral/ventral buccal nerve; dbn, dorsal buccal nerve; cbc, cerebral-buccal connective; a, anterior; p, posterior; m, medial; l, lateral. B: simplified schematic diagram summarizing the known synaptic interactions between the SO, a modulatory feeding interneuron, and the various types of feeding CPG interneurons active within the same phase and the overall interactions between the 3 classes of interneurons. Circles indicate inhibitory synapses, bars indicate excitatory synapses, combinations of bars and circles indicate biphasic synapses, and resistor symbols indicate electrotonic coupling. The identified neurotransmitters used by CPG interneurons (Glut, glutamate; ACh, acetylcholine) are also indicated. C: diagram showing the firing patterns of all known feeding CPG interneurons and the SO during 1 cycle of SO-driven fictive feeding activity. The 3 phases of the feeding pattern (P/N1 protraction, R/N2 rasp, and S/N3 swallow) are marked above and below the traces.

View larger version (34K): [in this window] [in a new window]   Fig. 2. Plateauing properties of medial N1 (N1M) protraction phase CPG interneurons. Ai: simultaneous recording from the N1M, lateral N1 (N1L), and SO interneurons in the intact CNS. The fictive feeding pattern was driven by constant injection of a depolarizing current into the SO for the duration of the record. The 3 phases of the fictive feeding pattern (P/N1 protraction, R/N2 rasp, and S/N3 swallow) are marked above and below the traces. Aii: 1st cycle of N1M protraction phase activity shown in Ai (shaded gray) at faster time scale. Inset: synaptic connectivity between protraction phase interneurons in intact CNS. Bars indicate excitatory synapses and resistor symbols indicate electrotonic coupling. Bi: isolated N1M interneuron in cell culture generates sustained plateau potentials in response to short supra-threshold current pulses (2 s). Sub-threshold current pulses (left) only evoked depolarizations for the duration of the current injection, which was followed by a short depolarizing afterpotential. Bii: plateau potentials triggered in the same isolated N1M as in Bi could be terminated prematurely by the injection of short hyperpolarizing current pulses. Note that the potential changes recorded during current injections are not accurate representations of the membrane potential since the recording electrode was not properly balanced. Ci: spontaneous plateau potential in an isolated N1M interneuron. Note the gradual depolarization (arrow) that triggers N1M activity and leads to the generation of the plateau potential. Cii: spontaneous N1M plateau potential in the intact CNS. As in the isolated N1M shown in Ci, the N1M activity is initiated by a gradual depolarization of the membrane potential. N1M plateau potential generation in the intact CNS leads to the activation of rasp phase interneurons, which in turn inhibit protraction phase activity and prematurely terminate the N1M plateau potential (arrow). D: record from an isolated N1L interneuron in cell culture that generated a burst of action potentials that had the same duration as the depolarizing current pulse. E: record from an isolated SO interneuron in cell culture showed a similar burst of action potentials in response to current injection.

View larger version (31K): [in this window] [in a new window]   Fig. 3. Absence of plateau potential generation in cultured ventral N2 (N2v) and dorsal N2 (N2d) rasp phase CPG interneurons. A: simultaneous recording from N2v, N2d, and SO interneurons in the intact CNS. The fictive feeding pattern was driven by constant injection of a depolarizing current into the SO for the duration of the record. The 3 phases of the fictive feeding pattern (P/N1 protraction, R/N2 rasp, and S/N3 swallow) are marked above and below the traces. Inset: synaptic inputs from protraction to rasp phase interneurons and electrotonic coupling between rasp phase interneurons in intact CNS. Circles indicate inhibitory synapses, bars indicate excitatory synapses, combinations of bars and circles indicate biphasic synapses, and resistor symbols indicate electrotonic coupling. B: membrane potential changes in an isolated N2v interneuron that were evoked by positive current pulses of various length and amplitude. The current injections evoked a truncated spike followed by a steady depolarization for the duration of the pulses. Ci: an isolated N2d interneuron generated a burst of action potentials in response to the injection of a positive current pulse lasting for the duration of the current pulse. Cii: start of current pulse-induced N2d activity (Ci) shown at a faster time scale.

View larger version (24K): [in this window] [in a new window]   Fig. 4. Cultured N3p interneurons show no postinhibitory rebound (PIR) property. A: simultaneous recording from N3p and SO interneurons in the intact CNS. The fictive feeding pattern was driven by constant injection of a depolarizing current into the SO for the duration of the record. The 3 phases of the fictive feeding pattern (P/N1 protraction, R/N2 rasp, and S/N3 swallow) are marked above and below the traces. Inset: synaptic inputs from protraction and rasp phase interneurons to swallow phase interneurons in intact CNS. Circles indicate inhibitory synapses. B: responses of an isolated phasic N3 (N3p) interneuron to various current pulses. A positive current pulse triggered a short burst of N3p activity that ceased before the end of the current pulse. Negative current pulses caused hyperpolarizations of the membrane potential that were followed by some small hyperpolarizing afterpotentials (arrows). Note the lack of any depolarizing afterpotentials as would be expected for a neuron with an endogenous PIR property. C: response of an N3p interneuron to positive and negative current pulses in the intact CNS. As in the isolated N3p interneuron, a positive current pulse triggered a brief burst of action potentials, while the negative pulse caused a hyperpolarization of the membrane potential, but again no PIR-induced depolarization was present.

Isolation and culture of identified neurons

Our cell culture procedure has been described in detail elsewhere (Straub and Benjamin 2001). Briefly, Fast Green-labeled identified interneurons were individually isolated from the CNS after the isolated CNS had received one of two alternative enzymatic treatments designed to soften the inner connective tissue. The first type of enzyme treatment consisted of incubation in trypsin and collagenase/dispase followed by incubation in soybean trypsin inhibitor. For the second type of enzyme treatment, the isolated CNS was incubated in a protease solution (Sigma type VIII) and washed. In our experience, this treatment facilitated the isolation of individual neurons with long lengths of primary neurites.

The cell bodies of Fast Green-labeled interneurons were exposed by mechanically disrupting the inner connective tissue and then removed, together with their main processes, by gentle suction with a fire-polished micropipette (tip diameter, 20-80 µm) prepared from 1.5-mm glass tubing (GC150T-10; Clark Electromedical Instruments) that had been coated with Sigmacote. After isolation, neurons were transferred onto poly-L-lysine-coated culture dishes containing conditioned culture medium and cultured at 20°C for 1-7 days.

Electrophysiological and pharmacological studies on cultured neuron

For intracellular recordings from isolated neurons, culture dishes were placed on the stage of an inverted microscope (Nikon Diaphot) that was equipped with a custom-built, gravity-fed perfusion system. The culture dishes were perfused with HEPES-buffered saline at a flow rate of 1-2 ml/min throughout the experiment. Cell bodies were impaled with microelectrodes pulled from 1-mm capillaries (GC100F-10; Clark Electromedical Instruments) and filled with saturated potassium sulfate (tip resistance, 20-30 M). All electrodes were routinely bridge-balanced using 1-nA, 1-s square wave current pulses prior to impalement of any neuron. During the recordings, the bridge balance was not readjusted. The electrodes were manipulated with hydraulic manipulators (MW-3 and MO-300; Narishige, Tokyo, Japan) mounted to the stage of the microscope that enabled the independent control of up to three recording and/or perfusion pipettes. The intracellular signals were amplified using an AXOCLAMP-2B amplifier (Axon Instruments), output to a storage oscilloscope (5115 Tektronix), and stored on a DAT recorder (Biologic DTR-1801). Amplified signals were also digitized using a DigiData 1200 interface (Axon Instruments) and stored on a PC.

The effects of ACh and glutamate were tested by focal application to the cell body of isolated neurons from micropipettes using 6-12 psi pressure pulses generated by a picospritzer (General Valve, Fairfield, NJ). We concentrated on the effects of these transmitters on each interneuron type at membrane potential levels comparable to those seen in the intact CNS when ACh- and glutamate-mediated inputs are expected to occur in a given cell. Therefore we recorded the resting membrane potentials of the various feeding CPG interneuron types in the intact CNS (SO: 61 ± 2 mV, n = 9; N1M: 60 ± 2 mV, n = 14; N1L: 63 ± 5 mV, n = 5; N2d: 66 ± 3 mV, n = 4; N2v: 64 ± 4 mV, n = 5; N3p: 65 ± 1 mV, n = 8; N3t: 54 ± 3 mV, n = 4). Protraction phase interneurons are considerably depolarized at the end of the protraction phase when they receive glutamatergic N2 inputs (V_m for N1M: 31 ± 1 mV, n = 12; N1L: 43 ± 2 mV, n = 2; SO: 52 ± 1 mV, n = 7), so the effects of glutamate on isolated protraction phase interneurons were also tested at these more depolarized levels.

Statistical analysis

Standard statistical analysis of the data were performed using the software package SPSS. The potential association between the length of time neurons spend in cell culture and their ability to generate action potentials, or plateau potentials in the case of the N1M, was analyzed using the Exact analysis tool for 2 × 7 tables (SISA Tables software), an extension of the Fisher Exact Probability test for 2 × 2 tables.

Validity of cell culture approach

Isolating neurons from the intact CNS and culturing them for periods of 7 days to study their basic intrinsic properties is obviously a highly artificial situation that raises two major questions. First, does isolation change the overall properties of a neuron? Second, do the properties of the isolated neurons change with time when they are cultured for any length of time?

Extensive literature on the electrical properties of molluscan and other invertebrate neurons in cell culture (reviewed in Bulloch and Syed 1992; Saver et al. 1999), including our own studies on Lymnaea feeding neurons (Straub and Benjamin 2001; Park et al. 1998), emphasize the overall similarities in the gross electrical properties of isolated neurons in cell culture and the intact CNS. Isolated molluscan neurons have been used successfully to study for example their endogenous properties (e.g., Arshavsky et al. 1986, 1988) and to examine plastic changes in synaptic interactions (e.g., Bailey et al. 2000; Lin and Glanzman 1997; Schacher et al. 1997). In one study, isolated molluscan neurons were also used to reconstruct a three cell network that is believed to resemble the respiratory CPG of Lymnaea (Syed et al. 1990). Overall, these previous studies clearly illustrate the usefulness of studying isolated molluscan neurons in cell culture to characterize their basic endogenous properties without the interference of synaptic inputs.

Furthermore, while time-dependent changes in the endogenous properties have been described in stomatogastric neurons (Turrigiano and Marder 1993; Turrigiano et al. 1995), we are not aware of any evidence to indicate similar changes in isolated molluscan neurons. However, to further substantiate the usefulness of the cell culture method for Lymnaea neurons, we have analyzed a number of basic electrical parameters of the feeding CPG interneurons for changes in their electrophysiological properties over the 1-7 days that the neurons were kept in cell culture. Statistical analysis revealed no evidence for a correlation between the length of time in cell culture and either the resting membrane potential (Spearman rank correlation: P values between 0.1 and 0.97, n values between 8 and 45) or the input resistance (Spearman rank correlation: P values between 0.06 and 0.97, n values between 4 and 17) for SO, N1M, N1L, N2d, N2v, and N3p interneurons. Further analysis also revealed no evidence for an association between the length of time in cell culture and the ability of isolated SO, N1M, N1L, N2d, N2v, and N3p neurons to generate action potentials (Exact analysis for 2 × 7 tables: P values between 0.07 and 1).

The majority of the isolated interneurons (75%, 100 of 134) retracted their secondary processes and some or all of the primary axons within a few hours of isolation. Only a small proportion of these cells (25%, 34 of 134) generated growth cones and showed any sign of neuritic regeneration. We analyzed whether regeneration or the lack of it affected the electrical properties of the cultured neurons. The resting membrane potential of regenerating neurons tended to be slightly more hyperpolarized than the resting membrane potential of nonregenerating neurons. This difference was only significant for N1M interneurons (t-test: P < 0.01), but not for the other interneuron types tested (t-test: P > 0.2; see Table 1). However, we found no evidence for significant differences in the input resistance of regenerating and nonregenerating interneurons, including N1M interneurons (t-test: P > 0.4; see Table 1). Further analysis of the regenerative status of isolated neurons and their ability to generate action potentials only revealed a statistically significant correlation for the SO interneurons (Fisher exact probability test: P = 0.02) but for none of the other interneuron types tested (Fisher exact probability test: P values between 0.42 and 1). Since regenerating and nonregenerating N1M and SO interneurons showed some differences in their basic properties, we analyzed whether these differences had any effect on the other endogenous and pharmacological properties we describe here, but found no supporting evidence. In N1M interneurons, the growth status did not affect their ability to generate plateau potentials (see RESULTS). Furthermore, the SO responses to ACh and the N1M and SO responses to glutamate in regenerating and nonregenerating interneurons showed no apparent differences (N1M responses to ACh were only tested in nonregenerating neurons).

Since there was no consistent correlation between the length of time in cell culture, the regenerative status of the isolated cells, their ability to generate action potentials, their resting membrane potential, and their input resistance beyond those described above, results for each individual cell type were pooled and analyzed together. We are not excluding the possibility that there might be quantitative or qualitative differences in individual currents expressed in isolated neurons that depend on their growth status or time in cell culture, but that study was beyond the scope of the current investigation.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

N1M, but not N1L or SO, protraction phase interneurons generate plateau potentials

In terms of the endogenous properties of the protraction phase interneurons, the N1Ms were of particular interest. One feature of their activity, which can be observed in fictive feeding rhythms in the intact CNS (Fig. 2Ai), is plateau-like depolarizations. This membrane potential "plateau" follows a gradual depolarization of the N1M during which the spike frequency accelerates and the spike amplitude reduces (Fig. 2Aii). The current model for feeding rhythm generation has assumed that this characteristic build-up of N1M activity is due to positive feedback effects within a network of electrically coupled N1M interneurons (Fig. 2A; Benjamin and Elliott 1989). However, the possibility that these plateau-like potentials are an endogenous property of this cell type has never been tested. Here we examined this issue in isolated cells in culture where all network influences are removed.

In isolated N1M neurons, plateau potentials could be triggered by brief depolarizing square current pulses from membrane potential levels more negative than 50 mV in about 50% of all neurons tested (23 of 45). The ability of isolated N1M interneurons to generate plateau potentials was not dependent on the length of time in cell culture (exact analysis for 2 × 7 table: P = 0.95). Furthermore, the growth status of isolated N1M interneurons (regenerating or nonregenerating) had no effect on their ability to generate plateau potentials (Fisher exact probability test: P = 0.53).

The amplitude of depolarizing current pulses had to be above a certain threshold of depolarization before sustained activity could be evoked (Fig. 2Bi). Sub-threshold current pulses caused a depolarization of the membrane potential that was followed by a transient afterpotential (Fig. 2Bi). Furthermore, the plateau potentials could be terminated prematurely by the injection of hyperpolarizing current pulses (Fig. 2Bii), supporting the hypothesis that N1M interneurons are true plateauing neurons, based on the criteria of Hartline and Graubard (1992). Additional studies showed that plateau potentials could also be triggered by positive current pulses and prematurely terminated by negative current pulses in N1M interneurons in the intact CNS (n = 3, data not shown). Interestingly, isolated N1M cells were also capable of producing plateau potentials in the absence of a depolarizing prepulse (Fig. 2Ci). Similar "spontaneous" plateau potentials have been observed in N1Ms in quiescent intact CNS preparations (Fig. 2Cii).

Both spontaneous and depolarization-evoked plateau potentials in isolated N1M cells had many features in common with spontaneous or protraction phase N1M activity seen in the intact CNS (Fig. 2, Ai and Cii). Spontaneous N1M plateau activity, both in culture and in the intact CNS, started with a gradual depolarization of the membrane potential that could trigger the generation of action potentials. As in the intact CNS, the progressive depolarization of the membrane potential (Fig. 2Ci, arrow) led to an increase in the N1M firing rate and a decrease in action potential size until the membrane potential reached a plateau 34 ± 2 mV (n = 23) more depolarized than the resting potential. This plateau potential amplitude is not significantly different from the depolarization of the N1M membrane potential at the end of the protraction phase in the intact CNS (30 ± 2 mV; n = 12; t-test: P = 0.24). The main difference between the plateau potentials seen in isolated N1M interneurons and in the intact CNS was their duration. In isolated N1Ms, this duration (approximately 20-90 s; Fig. 2, Bi and Ci) was considerably longer than in the intact ganglia (approximately 1-10 s; Elliott and Andrew 1991; Yeoman et al. 1996; also see Fig. 2, Aii and Cii).

Previous work has also suggested that the N1M cell type might have an endogenous bursting property that could play an important role in driving the feeding CPG to produce a rhythmic motor output (Benjamin and Elliott 1989), but this has not been tested explicitly. Evidence for the presence of an endogenous bursting or pacemaker property in isolated N1M interneurons was not seen. Although spontaneous plateau potentials did occur, no intrinsic rhythmicity in these depolarizations was ever observed.

Another question relates to whether the other protraction phase CPG interneuron N1L or the protraction phase modulatory interneuron SO have any endogenous properties that could contribute to pattern generation. Again, these issues were examined by studying the electrical properties of these cells in culture. In contrast to N1M, N1L neurons (n = 8) showed no plateau potential generation in cell culture and lacked any spontaneous bursting activity. This was also true of the protraction phase modulatory interneuron SO (n = 38). In both cell types, artificial supra-threshold depolarization of their membrane potential by current injection led only to the generation of action potentials (Fig. 2, D and E) with no evidence for the progressive depolarization leading to plateau-like depolarizations seen in the N1M interneurons. Similarly, hyperpolarization of the membrane potential failed to reveal the presence of any endogenous properties that might be relevant to burst generation (data not shown).

Lack of plateau potential generation in isolated rasp phase interneurons N2v and N2d

The activation of N2v and N2d rasp phase CPG interneurons during fictive feeding in the intact CNS is the result of complex synaptic inputs from protraction phase interneurons like N1M and N1L, as well as the modulatory SO interneuron (Fig. 3A). In an SO-driven rhythm in the intact ganglion, both N2 type interneurons show a characteristic "plateau-like" depolarization with superimposed action potentials (Fig. 3A). For N2v interneurons, these depolarizations fulfilled all the criteria for plateau potentials, but in the case of N2d interneurons, they were shown to be driven electrotonically by N2v activity (Brierley et al. 1997b). Whether plateau potential generation is an endogenous property of N2v interneurons was the most important question in this set of experiments. However, we also wanted to confirm that N2d neurons have no endogenous plateauing properties.

Despite the fact that plateau potentials were recorded from all N2v cells before isolation, no plateau-like potentials were seen in the same cells in culture (n = 12), either occurring spontaneously or evoked by depolarizing current steps. In the example shown in Fig. 3B, current steps of increasing amplitude and duration produced truncated spike-like events at the start of the step, but no maintained depolarizing waveform persisted after the end of the current injection. Similar spike-like events, comparable in amplitude to those superimposed on N2v plateau potentials in the intact ganglion (Fig. 3A), were observed in three of the isolated N2v interneurons. However, this is different from the situation in the intact CNS, where N2v interneurons generate action potentials for the duration of plateau potential. This difference between isolated N2v interneurons and cells in the intact CNS could indicate that N2v somata are not able to generate repetitive action potentials and that the site for action potential generation in the intact CNS is located at some distance from the N2v soma, which might be lost during the isolation procedure. Spike initiation zones at some distance from the cell body are not uncommon and have been described for example in some stomatogastric neurons where the site of spike initiation has been identified as near the point of exit of axons from the ganglion (Hartline and Graubard 1992).

All N2d interneurons that were recorded in cell culture (n = 10) were also silent at resting potential levels and never showed any plateau potentials, even when depolarized. Current-induced responses ranged from single action potentials at the start of the depolarization (not shown), to tonic spike activity that lasted throughout a current pulse (Fig. 3, Ci and Cii). The amplitude of the individual action potentials was similar to those observed in the intact CNS riding on the plateau-like N2d depolarizations during fictive feeding (Fig. 3A). Depolarization-induced spike activity in the isolated cells always ceased at the end of the pulse.

Lack of PIR in the N3p swallow phase interneurons

Previous work has shown that the swallow phase interneurons N3p and N3t fire mainly in the swallow phase of fictive feeding activity (Fig. 4A) following a period of synaptic inhibition arising from the N1 and N2 cells (summarized in Fig. 4A). In the absence of known excitatory synaptic inputs from the other CPG interneurons, it was hypothesized that both cell types fire by PIR. Endogenous PIR was directly demonstrated in the N3t cells in the intact system by Elliott and Benjamin (1985a) by showing that the N3t neurons fired on the recovery from artificial hyperpolarizing pulses that were similar in size and duration to the natural synaptic inhibitory input. However, this PIR property had not been investigated in the N3p interneurons and this was the main aim of the present study.

The majority of isolated N3p cells (18/19) generated single or short bursts of action potentials in response to the injection of depolarizing current pulses (Fig. 4B, left) but maintained spike activity in response to prolonged constant depolarizing current pulses was never observed. This is similar to the intact CNS where the N3p cells usually only generate brief bursts of activity during the swallow phase (Fig. 4A) or during depolarizing current steps (Fig. 4C).

Isolated N3p interneurons did not show the sort of after-depolarization following termination of negative current pulses that would be consistent with an endogenous PIR mechanism. In the majority of cases, the return to the resting membrane potential level was actually delayed by a brief after-hyperpolarization (Fig. 4B, arrows). The absence of PIR in the isolated cells could have been due to the removal of modulatory influences that were present in the intact ganglia. However, in the intact CNS a similar absence of PIR was observed (Fig. 4C, n = 6), suggesting that PIR is not occurring in these neurons either in culture, or in the intact CNS.

A study of the PIR properties of N3t interneurons in culture was not possible because recordings of these neurons, the smallest (<10 µm) and most fragile of the CPG neuron types, were too unstable to enable an accurate assessment of the effects of hyperpolarizing current pulses.

ACh and glutamate responses of protraction phase interneurons N1M, N1L, and SO mimic synaptic effects in intact CNS

All three protraction phase neurons considered in this paper, N1M, N1L, and SO, have previously been shown to be cholinergic, and the effects of ACh have been examined in the intact CNS (Elliott and Kemenes 1992; Vehovszky and Elliott 1995; Yeoman et al. 1993). In this paper, we have investigated the pharmacological responses to ACh in isolated neurons, but we confined this to N1M and SO cells, since N1L interneurons receive no known cholinergic synaptic inputs (Yeoman et al. 1995).

Isolated N1M interneurons responded to focal application of ACh (pipette concentration: 10 µM) with a transient depolarization of the membrane potential in all cells tested (Fig. 5Ai, n = 9). The average amplitude of the depolarization was 16 ± 2 mV (n = 5) at a membrane potential of 63 ± 2 mV. This is consistent with the excitatory synaptic effects that the cholinergic SO and N1L cells have on N1M interneurons in the intact CNS (Yeoman et al. 1993, 1995; see Fig. 5A). Furthermore, brief application of ACh could trigger plateau potentials in four of the cells tested (Fig. 5Aii), with similar characteristics to plateau potentials evoked by depolarization or occurring spontaneously (see Fig. 2, B and Ci). Similar depolarizing effects have been seen in the intact buccal ganglia with focal application of ACh to the N1M interneurons (Yeoman et al. 1993), and these could also lead to plateau-like sustained responses as reported by Elliott et al. (1992).

View larger version (22K): [in this window] [in a new window]   Fig. 5. Acetylcholine (ACh) triggers plateau potentials in isolated N1M interneurons. Ai: record of strong depolarizing response in isolated N1M interneurons to focal application of ACh. Aii: identical ACh pulse triggered a full plateau potential in another isolated N1M interneuron. Inset: diagram showing cholinergic synaptic inputs to N1M interneuron. Bars indicate excitatory synapses. Bi: isolated SO interneuron that responded with a strong depolarization followed by a delayed hyperpolarization (arrow) to focal ACh application. Bii: 10 times lower ACh concentrations evoked exclusively hyperpolarizing responses in isolated SO interneurons. Inset: diagram showing cholinergic synaptic inputs to SO interneuron. Bars indicate excitatory synapses.

The responses of cultured SO cells to ACh (n = 10) were more complex. They showed a strong depolarization of 27 ± 3 mV at a membrane potential of 66 ± 3 mV in response to focal application of 10 µM ACh (n = 6). However, at 51 ± 2 mV (n = 6), identical ACh pulses caused a biphasic response that consisted of an initial depolarization of 12 ± 2 mV that was followed by a hyperpolarization of 2 ± 0 mV. (Fig. 5Bi, arrow). This second component of the ACh response could be isolated by reducing the ACh pipette concentration to 1 µM. This lower concentration evoked only hyperpolarizing responses (average amplitude 14 ± 5 mV at a membrane potential of 60 ± 1 mV, n = 4) in all the SO interneurons tested (Fig. 5Bii).

We also examined glutamate responses in isolated protraction phase interneurons, since previous work has shown that the N2v interneurons make strong inhibitory glutamatergic synapses onto N1M, N1L, and SO interneurons (Brierley et al. 1997b,c). It was important to establish here whether isolated neurons showed the same pharmacological responses.

In isolated N1M interneurons, focal application of glutamate (pipette concentration: 0.1 and 1 mM) could terminate spontaneous or triggered plateau potentials (Fig. 6Ai), consistent with the inhibitory synaptic effect of the glutamatergic N2v cells on the N1M reported in the intact CNS (Brierley et al. 1997b; Fig. 6Ai). The detailed nature of this response was further unmasked by examining N1M neurons, which were not generating plateau potentials. In these neurons, the average amplitude of the hyperpolarizing effect was 16 ± 2 mV at membrane potentials (33 ± 1 mV, n = 13; Fig. 6Aii) that are comparable to N1M membrane potentials at the end of the protraction phase in the intact CNS. In five cases, the predominant hyperpolarization was preceded by a weak depolarization of 3 ± 1 mV (n = 5) that was only observed at the higher glutamate concentration (1 mM, Fig. 6Aiii). A statistical analysis revealed no significant difference in the peak amplitude of the hyperpolarization caused by 1 mM glutamate, either as part of the biphasic response or a purely hyperpolarizing response, and by 0.1 mM glutamate (ANOVA; F(2, 10) = 0.55, P = 0.6). However, the presence of the depolarizing component significantly delayed the occurrence of the peak hyperpolarization from 2.8 ± 0.5 s (n = 8) to 10.6 ± 1.9 s (n = 5) after the onset of the glutamate application (t-test: P < 0.01). When the membrane potential was hyperpolarized to 69 ± 3 mV (n = 7), as expected, the depolarizing effect increased in amplitude to 29 ± 8 mV (n = 4), while the glutamate induced hyperpolarization reduced in amplitude to 5 ± 2 mV (n = 7).

View larger version (39K): [in this window] [in a new window]   Fig. 6. Glutamate terminates firing activity in isolated protraction phase interneurons. A: focal application of glutamate to isolated N1M interneurons caused strong hyperpolarizing responses. Ai: record from an isolated N1M interneuron where glutamate application terminated an endogenously generated plateau potential that was triggered by a brief depolarizing current pulse. Aii: in another isolated N1M interneuron that did not generate plateau potentials, glutamate application caused a strong hyperpolarization (i) when the membrane potential was adjusted to 30 mV. Aiii: at slightly more negative membrane potentials (50 mV), glutamate evoked biphasic responses that consisted of an initial brief depolarization (e) followed by a strong prolonged hyperpolarization (i). Aiv: recording from an isolated N1M interneuron whose membrane potential was adjusted from about 80 mV to about 60 mV by constant current injection. At the more depolarized membrane potential, the neuron started to generate a plateau potential that was terminated by the application of a glutamate pulse. This mimicked the N2v inhibitory synaptic input in the intact CNS. After some time, the membrane potential recovered from the glutamate-induced hyperpolarization, and a new plateau potential was initiated. Varying the interval between glutamate pulses (Aiv,right) clearly showed that the termination of the N1M plateau potentials was a direct effect of the transmitter application and not the result of an endogenous bursting property. B: biphasic response to glutamate application recorded from an isolated N1L interneuron. The response consisted of a fast, brief depolarization (e) followed by a delayed, more pronounced hyperpolarization (i). C: effect of glutamate on isolated SO interneuron. The SO interneuron showed a biphasic response (fast depolarization, e; delayed hyperpolarization, i) to a pulse of glutamate at membrane potentials around 60 mV. When the cell was driven to fire action potentials by constant current injection, glutamate caused a strong hyperpolarization that transiently stopped spiking activity, mimicking the effect of the N2v input in the intact CNS. Inset (Ai, B, andC): diagrams showing glutamatergic inhibitory synaptic connection in the intact CNS.

In Fig. 6Aiv, we show an example of an isolated N1M neuron, where we mimicked chemosensory synaptic inputs that can drive fictive feeding activity in semi-intact preparations by constantly depolarizing the membrane potential sufficiently to trigger plateauing activity (see start of record). In this situation, rasp phase synaptic inputs could be mimicked by focal application of glutamate, which caused termination of the on-going plateau potential. After some time, the N1M membrane potential recovered from the hyperpolarizing effect of glutamate, and the constant depolarization triggered a new plateau potential that could again be terminated by glutamate application. Varying the interval between glutamate pulses (Fig. 6Aiv, right) clearly showed that the termination of the N1M plateau potentials was a direct effect of the transmitter application and not the result of an endogenous bursting property. So, the interaction of constant depolarization and phasic application of glutamate could induce a rhythmic activity pattern in an isolated N1M interneuron, which resembled its activity pattern during fictive feeding in the CNS.

Like the N1M cells, the large majority of the N1L CPG interneurons (3/4, Fig. 6B), as well as all the SO modulatory interneurons (n = 10, Fig. 6C), showed predominantly hyperpolarizing responses to glutamate (pipette concentration: 0.1 and 1 mM), consistent with the known effects of glutamate and/or the glutamatergic N2v cells in the intact ganglion (Brierley et al. 1997b; panels in Fig. 6, B and C). At membrane potentials comparable to those at the end of the protraction phase, when N1L and SO interneurons receive inhibitory synaptic inputs from rasp phase interneurons, the amplitude of the glutamate induced hyperpolarizations was 5 ± 3 mV for N1L interneurons (V_m 44 ± 7 mV, n = 3) and 14 ± 3 mV for SO interneurons (V_m 50 ± 2 mV, n = 7). In addition, all the N1L cells and a number of SO cells (4/10) also showed a short-lived depolarizing component before the longer-lasting hyperpolarization (Fig. 6, B and C). In SO interneurons, this additional depolarizing response only became conspicuous when the membrane potential was adjusted to values in the range of the SO resting potential in the intact CNS. Within that range, the average depolarization was 5 ± 2 mV (V_m 65 ± 2 mV, n = 4). However, unlike in N1M interneurons where only the higher glutamate concentration (1 mM) elicited depolarizing responses, both concentrations of glutamate (0.1 and 1 mM) could induce depolarizing responses in N1L and SO interneurons. The glutamate-induced depolarizations in SO interneurons were of short duration and did not significantly delay the peak of the hyperpolarizing response (2.7 ± 0.4 s, n = 4 vs. 3.3 ± 1.4 s, n = 3, t-test: P = 0.69). As with the N1M, the hyperpolarizing effect of glutamate could terminate a burst of action potentials in isolated SO neurons (Fig. 6C, right).

Acetylcholine causes complex responses in rasp phase interneurons N2v and N2d

In the intact CNS, the N2d and N2v interneurons receive various types of direct synaptic inputs from all the cholinergic protraction phase interneurons (Fig. 7, Ai and Bi). These were excitatory (N1M  N2d/N2v, N1L  N2d; Brierley et al. 1997b; Elliott and Benjamin 1985a) inhibitory (N1L  N2v interneurons; Brierley et al. 1997b), or biphasic (inhibitory-excitatory, SO  N2d/N2v; Brierley at al. 1997b; Yeoman et al. 1993).

View larger version (31K): [in this window] [in a new window]   Fig. 7. Multiphasic cholinergic responses of isolated rasp phase interneurons. A: isolated N2v interneurons show 2 types of responses to focal ACh application. Ai: isolated N2v interneuron that responded with a simple depolarization of the membrane potential to ACh application (1 s). Aii: complex response to ACh in another N2v interneuron. Here, ACh (0.5 s, left) caused an initial fast hyperpolarization (i), which was followed by a delayed depolarization (e) with a superimposed spike and "plateau-like" shoulder. Shorter ACh pulses (0.2 s, right) also evoked a biphasic response, but failed to trigger the spike and "plateau-like" depolarization. Bi: one-half of the N2d interneurons responded with a dual hyperpolarization (i1, i2) of the membrane potential to ACh application. Bii: however, in the other one-half, ACh evoked a simple hyperpolarization. Insets: summary diagrams showing the cholinergic protraction phase synaptic inputs to N2v and N2d interneurons in the intact CNS. Circles indicate inhibitory synapses, bars indicate excitatory synapses, and combinations of bars and circles indicate biphasic synapses.

ACh (10 µM) had a predominantly depolarizing effect on N2v neurons in culture, causing an average depolarization of 11 ± 2 mV at a membrane potential of 62 ± 2 mV (n = 6, Fig. 7Ai). This could account for the depolarizing effect of the cholinergic N1M  N2v synaptic input in the intact CNS (Brierley et al. 1997b). In three N2v cells, the longer depolarizing response was preceded by a much shorter hyperpolarizing component (average amplitude 13 ± 9 mV, n = 3, Fig. 7Aii). Dual hyperpolarizing-depolarizing responses to ACh were also seen in the intact ganglion (Brierley et al. 1997b) and can account for the biphasic synaptic effects of SO and N1L activity on N2v cells (Brierley et al. 1997b). Significantly, in two experiments (one of which is shown in Fig. 7Aii, left), the depolarizing component of the ACh response was plateau-like in appearance, with a single truncated spike riding on a waveform that far outlasted the duration of ACh application. In the intact ganglion, the delayed depolarizing component of the ACh response in N2v could also trigger plateau potentials (Brierley et al. 1997b). A shorter pulse of ACh to the N2v in the experiment shown in Fig. 7Aii (right) could still trigger a biphasic response, but the delayed depolarization was not sufficient to evoke a plateau-like waveform.

Isolated N2d interneurons (n = 4) were hyperpolarized by ACh pulses (pipette concentration: 10 µM; Fig. 7B). In one-half of the isolated N2d neurons, the hyperpolarization appeared to consist of two components, a fast initial hyperpolarization followed by a secondary delayed hyperpolarization (Fig. 7Bi), while the other half showed only a simple hyperpolarizing response (Fig. 7Bii). In N2d interneurons that clearly showed two components (n = 2), the peak amplitude of the initial hyperpolarization was 33 ± 8 mV and 27 ± 7 mV for the secondary component (V_m 63 ± 5 mV). In the rest of the N2d interneurons, the peak amplitude of the simple ACh-induced hyperpolarization was 22 ± 7 mV (V_m 62 ± 4 mV, n = 2). However, unlike N2v interneurons, ACh could never elicit "plateau-like" or even smaller depolarizing events in N2d cells.

Acetylcholine and glutamate produce inhibitory responses on swallow phase interneurons N3p and N3t

In the intact CNS, N3p and N3t interneurons receive inhibitory synaptic inputs from the cholinergic protraction phase interneurons N1M and N1L that are probably all monosynaptic (Fig. 8, A and B; Elliott and Benjamin 1985a; Yeoman et al. 1995). In isolated N3 cells, we found that these hyperpolarizing responses could be mimicked by focal application of ACh (pipette concentrations: 1 and 10 µM; N3p, n = 5, Fig. 8A; N3t, n = 2, Fig. 8B). In N3p interneurons, the average peak amplitude of the ACh-induced hyperpolarization was 14 ± 3 mV (V_m 58 ± 2 mV, n = 5).

View larger version (31K): [in this window] [in a new window]   Fig. 8. Cholinergic and glutamatergic responses of swallow phase interneurons are all inhibitory. A: ACh caused a strong hyperpolarization of the membrane potential in an isolated N3p interneuron. B: ACh had a similar, but considerably weaker, effect on an isolated tonic N3 (N3t) interneuron. C: record from an isolated N3p interneuron that responded with a strong hyperpolarization to the focal application of glutamate. D: similarly, an isolated N3t interneuron showed a hyperpolarizing response to an identical glutamate pulse. Insets: diagrams summarizing the cholinergic and glutamatergic synaptic inputs from protraction and rasp phase interneurons to swallow phase interneurons. The presence of glutamate in N2d interneurons has not been confirmed, but as glutamate mimics the synaptic effect of N2d interneurons on known follower neurons (Brierley et al. 1997c), the N2d synaptic connections were labeled as putatively glutamatergic (?).

We also examined the effect of glutamate on N3 interneurons, since both subtypes are thought to receive inhibitory inputs from rasp phase interneurons (Fig. 8, C and D; Brierley et al. 1997b). Focal glutamate application (pipette concentration: 1 mM) to isolated N3p interneurons mimicked the inhibitory effect of rasp phase synaptic inputs in 4 of 10 of the neurons tested (Fig. 8C, average peak amplitude 6 ± 4 mV at V_m 56 ± 2 mV, n = 4). The remaining six neurons failed to show any response. Pulses of glutamate applied to a single N3t also caused a hyperpolarizing response (Fig. 8D).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

In the current study we set out to characterize the basic endogenous electrical and pharmacological properties of the different Lymnaea feeding CPG interneurons in an attempt to test whether their endogenous properties contribute to the firing patterns seen in fictive feeding in the intact CNS. Considering the expression of endogenous electrical properties in the isolated neurons, it appears that the Lymnaea feeding CPG can be classified into three types: 1) neurons that constitutively express endogenous properties in isolation, 2) neurons that possess conditional endogenous properties, which require the presence of a modulator, and 3) neurons that lack any endogenous properties and are completely synaptically driven. Here, we have provided evidence that plateau potential generation in N1M interneurons is a basic property of these neurons. In contrast, plateau potential generation in N2v interneurons appears to be a conditional property, while isolated SO, N1L, N2d, and N3p interneurons apparently lacked any basic endogenous properties in cell culture. However, we cannot rule out the possibility that any of these neuron types show new or additional conditional properties under certain conditions that might be present in the intact CNS.

What makes the Lymnaea CPG interneurons fire?

Here we discuss how the basic endogenous properties of the isolated CPG interneurons might contribute to the firing patterns of these neurons seen in fictive feeding rhythms in the intact CNS. Of particular significance was the finding that of the two protraction phase CPG interneuron types investigated, only the isolated N1M cells have endogenous plateauing properties. In the intact system, this property allows N1M interneurons to fire strong bursts in response to depolarizing inputs from sensory pathways at the onset of feeding (Kemenes et al. 2001) and from other interneurons during ongoing patterns (Elliott and Benjamin 1985a). The N1M interneurons form strong excitatory monosynaptic connections with other CPG and modulatory interneurons (Brierley et al. 1997b; Elliott and Benjamin 1985a,b; Yeoman et al. 1995), so that once such a burst is triggered in this cell, it will have a profound influence on the rest of the feeding network. Indeed, previous work has demonstrated a pivotal role for N1M activation in both the initiation and maintenance of rhythmic activity in the whole CPG (Kemenes and Elliott 1994; Kemenes et al. 2001).

An interesting feature of the evoked plateau potentials observed in isolated N1Ms is that they are much longer in duration than those seen during fictive feeding in the intact CNS. This raises an important question about the mechanism that normally terminates plateau potentials during rhythmic feeding patterns in the intact circuit. Previous work suggested that early termination of the N1M plateau potentials during a feeding rhythm is likely to be due to recurrent inhibitory feedback originating from the N2d/N2v cells (Brierley et al. 1997b). Our results support this hypothesis because we were able to show that glutamate can terminate plateau potentials in isolated N1M interneurons. The timing of this N2 inhibitory feedback reduces the duration of the plateau potentials to a few seconds compared with the 20- to 90-s duration of the endogenous plateau potentials seen in isolated cells. It is of interest to note that on the occasions when N1M plateau potentials in the intact CNS apparently fail to trigger N2 activity and the expression of the full fictive feeding pattern, the duration of N1M plateau potentials is comparable to the duration in isolated N1M interneurons.

A second issue relates to the repetitive burst firing in N1M interneurons observed during sustained fictive feeding rhythms in the intact CNS (Elliott and Benjamin 1985a; Rose and Benjamin 1981b). Previously it has been proposed that this might be due to a pacemaker mechanism intrinsic to N1M interneurons (Rose and Benjamin 1981b), but there was no conclusive evidence for this in the intact CNS. Moreover, the N1Ms we isolated in culture never showed repetitive burst firing, suggesting that this is not a basic endogenous property of these cells. It appears most likely that for a plateau potential to be triggered in each successive feeding cycle, the N1M cell must receive excitatory inputs, which depolarize the neuron. In a naturally evoked feeding rhythm, these may arise from maintained activity in chemosensory pathways, supported by phasic excitatory synaptic inputs from other protraction-phase interneurons (Yeoman et al. 1995). This, together with identified phasic inhibitory inputs from rasp and swallow phase interneurons (Brierley et al. 1997b; Elliott and Benjamin 1985a), phase-locks the plateau-driven N1M activity to the rest of the CPG network. As N1M/protraction phase activity is the important first step in the activation of each feeding cycle, triggering N1M activity in this way would ensure that feeding activity is limited to periods of sufficient excitatory drive from chemosensory inputs without the requirement of an independent mechanism to terminate feeding at the end of a feeding episode. However, this does not imply that feeding stops immediately after removal of the sensory stimulation as sensory activation of feeding probably involves the activation of higher order integrating centers whose activity might persist for some time after removal of the stimulus (Kemenes et al. 2001).

No evidence was obtained that the other type of protraction phase CPG neuron, N1L, showed an endogenous property that contributed to patterning in the feeding system. In this respect N1L neurons were more similar to the SO, a modulatory feeding interneuron active in the protraction phase that was included in these experiments. Therefore the previously described activation of SO and N1L interneurons in the intact CNS during fictive feeding rhythms (Kemenes et al. 2001; Yeoman et al. 1995) should be fully accounted for by excitatory synaptic inputs. In the case of N1L interneurons, chemosensory inputs from the lips will contribute to their activation (Yeoman et al. 1995). The SO, in addition to chemosensory inputs (Kemenes et al. 1986, 2001), also receives excitatory inputs from the N1M interneurons (Elliott and Benjamin 1985b). N1L and SO neurons can also excite each other through their electrotonic coupling (Yeoman et al. 1995). As with the N1M interneurons, termination of protraction-phase bursts in N1L and SO neurons arises from identified inhibitory synaptic inputs from N2 phase interneurons (Brierley et al. 1997b). This means that bursting activity of the N1L and SO cells in a food-driven fictive feeding pattern is most probably due to a combination of chemosensory inputs from the lips and the synaptic connections within the feeding circuit described above. This distinguishes N1L from N1M interneurons making them more similar to the modulatory SO neuron. We cannot exclude the possibility that N1M interneurons (or any of the other protraction phase interneurons) possess conditional bursting/pacemaker properties that require the presence of a modulator, which is absent from the culture system but present in the intact CNS. However, the existing experimental data make this less likely than the input-dependent plateau potential generation outlined above.

Unlike N1Ms, N2v interneurons did not show depolarization-driven endogenous plateau potentials in cell culture, although they generate plateau potentials in the intact CNS (Brierley et al. 1997a). Considering that N1M interneurons retain their plateauing property under identical conditions, the apparent loss in isolated N2v interneurons appears unlikely to be due to the isolation procedure. Therefore the N2v plateau potentials seen in the intact CNS either reflect the waveform of a so far unidentified, closely coupled neuron, or they are a conditional endogenous property of N2v interneurons requiring the presence of some modulatory factors. The N2d neurons never showed endogenous plateau potentials in isolation and this confirmed the previous finding in the intact CNS that their excitation arises from purely synaptic sources (Brierley et al. 1997a,b).

In fictive feeding rhythms, the N3p and N3t interneurons receive consecutive phases of inhibitory inputs from N1M and N2d/N2v cells, respectively, and appear to fire on the rebound from the second phase of inhibition (Elliott and Benjamin 1985a; Rose and Benjamin 1981b). It had been proposed that a PIR mechanism is involved in activating spikes in both cell types. The presence of such a mechanism has been demonstrated in N3t interneurons in the intact CNS (Elliott and Benjamin 1985a). However, in the present work, no evidence was found that N3p cells also showed PIR either in the intact CNS or in isolation. An alternative explanation for their spike activation during feeding rhythms comes from recent work on a subset of swallow phase motoneuron types that fire in an overlapping sequence in the rasp-swallow phase of the feeding cycle at the same time as the N3p cells. They provide an excitatory synaptic drive to N3p interneurons via electrotonic connections (Staras et al. 1998), which in the absence of PIR or any other excitatory input to the N3p cells, is likely to provide the main mechanism for driving spike activity in these neurons.

Our current model for rhythm generation in the feeding CPG can be summarized as follows. Chemosensory inputs from the lips strongly depolarize the N1M protraction phase interneurons triggering the generation of endogenous plateau potentials that maintain protraction phase activity. Activation of the N1M interneurons is further supported by direct excitatory synapses from N1L interneurons and the SO onto the N1M interneuron. The combined N1M and N1L activity has a biphasic effect on rasp phase interneurons that consists of an initial hyperpolarization followed by a slow depolarization that activates N2v interneurons and triggers plateau potential generation in these interneurons. N2v activity also spreads to N2d interneurons via their electrotonic connection. The activation of these rasp phase interneurons in turn terminates the N1M plateau potential and the protraction phase activity by reciprocal inhibitory synaptic connections from N2v and N2d interneurons onto all protraction phase interneurons and the SO. Rasp phase activity ceases when N2v plateau potentials terminate spontaneously. The end of the rasp phase activity releases swallow phase interneurons N3t and N3p from their dual protraction-rasp phase inhibition. This leads to the direct activation of N3t interneurons by an endogenous PIR mechanism, while N3p interneurons are activated indirectly due to their electrotonic connection to swallow phase motoneurons, which fire, like N3t interneurons, on the rebound from a rasp phase inhibition. The swallow phase interneurons possess direct inhibitory connections with protraction phase interneurons that delay their recovery from the rasp phase inhibition. Once swallow phase activity ceases, sustained sensory activation can initiate a new cycle of feeding activity.

Does the synaptic connectivity of Lymnaea feeding interneurons predict their pharmacological sensitivity?

Two neurotransmitters, ACh and glutamate, have been identified in the intact CNS as being important in mediating many of the feeding CPG interactions (e.g., Brierley et al. 1997c; Elliott et al. 1992; Yeoman et al. 1993). In this paper, we have shown that the basic responses to ACh and glutamate in most of the isolated cells resembled those predicted on the basis of known synaptic connectivity and pharmacological responses in the intact system, except that these responses often revealed a further level of complexity not previously reported in the intact CNS. These additional responses were generally weak at the recorded membrane potential, which might explain why they were previously overlooked in studies in the intact CNS. It is also possible that they represent receptor types located at synapses different from those that mediate the interactions between the CPG interneurons that are not usually activated during feeding rhythm generation. However, we also cannot rule out the possibility that some of these responses were due to receptor types not usually present in these cell types, but whose expression was induced by the isolation procedure.

In isolated N1M interneurons, ACh reliably evoked a strong depolarizing response, which could trigger full plateau potentials and is consistent with the synaptic input N1M interneurons receive from the cholinergic SO (Yeoman et al. 1993). In contrast, ACh evoked biphasic responses (e-i) in isolated SO cells. The fast depolarizing component was consistent with a reciprocal excitatory cholinergic input from N1M interneurons (Elliott and Benjamin 1985b), while the delayed hyperpolarization may underlie the often observed reduction in SO spike frequency during the late protraction phase of fictive feeding rhythms, which was attributed previously to an inhibitory input from an unknown source (Elliott and Benjamin 1985b).

When the isolated protraction phase interneurons were depolarized to membrane potentials comparable to values at the end of the protraction phase in the intact CNS, glutamate application caused a significant hyperpolarization. This was sufficient to terminate activity, like endogenously generated N1M plateau potentials, underlining its functional importance in the fictive feeding rhythm. An underlying biphasic (e-i) effect was observed when these cells were not firing. The excitatory component was generally short-lived and the hyperpolarization was still the predominant glutamate effect on all three cell types, consistent with the effect of glutamate-mediated N2v synaptic inputs to these cells in the intact CNS (Brierley et al. 1997b).

In all isolated N2v neurons, ACh caused a depolarization, mirroring excitatory cholinergic N1M inputs. In one-half of the cells, a dual hyperpolarizing-depolarizing ACh response was seen, resembling the biphasic inputs caused by SO and N1L activity in the intact CNS (Brierley et al. 1997a,b). The differences in the ACh responses are unlikely to represent heterogeneity in the population of N2v interneurons, since there is thought to be only a single N2v cell per ganglion. One possibility is that a proportion of isolated N2v interneurons lacked the receptor mediating the hyperpolarizing component. This could be due to differences in the spatial distribution of the ACh receptor types, with receptors mediating the depolarizing response being expressed on the cell body or proximal axon branches, while receptors mediating the hyperpolarization being expressed on more distal processes. Distal processes are more likely to be lost during cell isolation, which could easily lead to the selective loss of one receptor type. Interestingly, the pharmacological experiments with isolated N2v interneurons also provided some evidence for conditional endogenous plateau potential generation in this cell type. Although plateau potentials could not be triggered by depolarization alone, the depolarizing component of the biphasic ACh response triggered a "plateau-like" waveform in some cells. This suggests that plateau potential generation in N2v interneurons may be conditional on the presence of ACh, a criterion always fulfilled in the intact CNS during feeding patterns.

The exclusively hyperpolarizing ACh responses in all isolated N2d neurons were somewhat surprising because previous studies in the intact CNS have shown either biphasic (hyperpolarizing-depolarizing, SO  N2d, Yeoman et al. 1993) or depolarizing effects (N1M/N1L  N2d, Elliott and Benjamin 1985a; Yeoman et al. 1995) of cholinergic synaptic inputs to N2d interneurons. Furthermore, focal application of ACh to N2d interneurons in the intact CNS also caused dual hyperpolarizing-depolarizing effects (Yeoman et al. 1993). However, it is possible that the depolarizing component observed in the intact CNS is in fact an indirect effect that arises from N2v interneurons and spreads to N2d interneurons via their electrotonic connection.

In comparison to the other interneuron types, ACh and glutamate evoked rather more simple responses in isolated N3p and N3t cells that were consistent with the synaptic inputs they receive in the protraction and rasp phase, respectively. The lack of any response to glutamate in a large proportion of isolated N3p interneurons was somewhat surprising. However, it is possible that glutamate receptors are located on distant N3p processes that are regularly lost during the isolation process. This is consistent with the observation that the recorded hyperpolarization of the membrane potential in the N3p soma during the rasp phase is usually very weak (e.g., Fig. 4A). Alternatively, the population of N3p interneurons might consist of various subtypes, some of which lack glutamate receptors. Inhibition of these subtypes during the rasp phase could be an indirect effect due to their electrotonic coupling to B4/B4CL neurons that are inhibited by rasp phase interneurons. Finally, we cannot rule out the possibility that a different transmitter type mediates the inhibitory effect of rasp phase interneurons on N3p interneurons. The lack of any depolarizing responses in isolated swallow phase interneurons to ACh and glutamate emphasized the importance of alternative mechanisms for their activation during feeding patterns (N3t: PIR, N3p: electrotonic coupling to swallow phase motoneurons).

Comparison with other systems

Despite the obvious advantages of the cell culture approach to study endogenous properties of CPG interneurons, it has been used in only a minority of CPGs studied such as the crustacean stomatogastric and cardiac CPGs and the gastropod respiratory CPG.

Cultured stomatogastric neurons can generate plateau potentials (Panchin et al. 1993) comparable to those observed in the current study in isolated N1M interneurons. Furthermore, many of the isolated stomatogastric neurons show endogenous bursting and PIR properties (Panchin et al. 1993; Turrigiano and Marder 1993). However, the expression of endogenous properties in many of the cell types studied is modulated by, or conditional on, the presence of neuromodulators in the intact CNS (Harris-Warrick et al. 1997), where many of these properties are believed to play important roles in the function of the pyloric and gastric CPGs (Selverston et al. 1997). Similarly, the so-called "large cells" of the crustacean cardiac CPG show apparent endogenous bursting properties in cell culture, which resemble their behavior in the cardiac ganglion, suggesting that they are important in generating their rhythmic activity patterns (Saver et al. 1999). In contrast, interneurons from the Lymnaea respiratory CPG show neither endogenous bursting nor plateauing properties in cell culture, but they were reported to possess PIR properties (Syed et al. 1990). It is not clear from the data presented whether the PIR properties in these examples are due to the electrical membrane properties or whether they are in fact caused by biphasic (i-e) synaptic interactions. Either way, co-culturing of the three neuron types that are believed to comprise the respiratory CPG was sufficient to reconstruct a neuronal network, which showed rhythmic activity patterns resembling those of the CPG in the CNS (Syed et al. 1990). This result illustrates that the isolated neurons retained all the relevant properties to recreate a functional network.

In some additional studies, spinal neurons from Xenopus embryos (Dale 1997), feeding CPG interneurons from the gastropod Planorbis (Arshavsky et al. 1988), and swim CPG interneurons from the gastropod Clione (Arshavsky et al. 1986) were also studied after complete isolation from the CNS, but they were not maintained in cell culture. Spinal Xenopus embryo neurons possess rather simple electrical properties lacking any obvious endogenous pattern generating properties. In contrast, acutely isolated Planorbis group 1 interneurons, which are comparable to Lymnaea N1M interneurons, were able to generate endogenous plateau potentials with features similar to those observed in cultured N1M interneurons. Furthermore, both Planorbis group 1 and group 2 interneurons showed endogenous bursting properties. Similarly, endogenous bursting properties were also found in acutely isolated Clione swim interneurons 7 and 8 that are believed to play an important role in the generation of the swim pattern. However, since the Planorbis and Clione neurons were acutely isolated and kept in the same recording chamber as the rest of the CNS, it is possible that these neurons were still under the influence of long-lasting modulatory effects or that they were affected by neuromodulators released into the medium by the CNS. Therefore it is not possible to assess whether the observed endogenous properties are constitutively expressed in these cell types or whether they are conditional properties that require the presence of specific neuromodulators.

Of the isolated CPG interneurons discussed above, most seem to retain properties that are believed to be important in the generation of rhythmic motor patterns. In crustacean CPGs, the Clione swim CPG and the Planorbis feeding CPG, pattern generation appears to rely on endogenous bursting properties in certain cell types. In contrast, in the Lymnaea respiratory CPG, pattern generation is mainly determined by synaptic interactions. Our work on the Lymnaea feeding CPG suggests that this system may generate rhythmic motor patterns through the interaction of endogenous conditional and unconditional properties and synaptic connectivity in the absence of any endogenous bursting neurons. However, we do not exclude the possibility that as yet undiscovered modulatory effects are present, which could influence this system and would need to be incorporated into any future model.