Different feeding-related behaviors in Aplysia require substantial variations in the coordination of movements of two separate body parts, the lips and buccal mass. The central pattern generators (CPGs) and motoneurons that control buccal mass movements reside largely in the buccal ganglion. It was previously thought that control of the cerebral neuronal circuitry and motoneurons that generate lip movements was coordinated directly by feedback from buccal interneurons. Here, we describe cerebral lip motoneuron C15, which drives rhythmic activity in the isolated cerebral ganglion. Other lip motoneurons are active during this program, so we define it as a cerebral motor program (CMP). The C15 in each cerebral hemiganglion drives the CMP in ipsilateral neurons only, suggesting there are independent CPGs in each hemiganglion. The cerebral and buccal CPGs interact at several points. For example, cerebral-to-buccal interneurons (CBIs), which can drive the buccal CPG, receive excitatory input when the cerebral CPG is active. Likewise, C15, which can drive the cerebral CPG, is excited when the buccal CPG is active. This excitation is simultaneous in both C15s, coupling the activity in the two hemiganglionic cerebral CPGs. Therefore, there are independent cerebral and buccal CPGs, which can produce distinct rhythms, but which interact at several points. Furthermore, the connections between the cerebral and buccal CPGs alter during different forms of motor program. We suggest that such alterations in the interactions between these CPGs might contribute to the generation of the various forms of coordination of lip and buccal mass movements that are necessary during different feeding-related behaviors.
Most animals need to alter the coordination of different body parts to produce a variety of behaviors. For example, the coordination of different limbs changes during various locomotor gaits. It has been proposed, in several such systems, that there is series of individual central pattern generators (CPGs), one for each limb or joint (vertebrates: Grillner and Wallen, 1985; Rossignol et al., 1993; invertebrates: Mulloney et al., 1993; Ryckebusch and Laurent, 1994). These CPGs could interact in distinct ways to produce the different motor patterns needed to generate different gaits or even separate behaviors such as scratching. It should be noted that this is a different situation than that for chains of oscillators, which are coupled to produce undulatory swimming behaviors, as has been proposed in the leech (Friesen and Pearce, 1993) and lamprey (Sigvardt, 1993). In these systems, individual oscillators are largely identical, whereas in the examples for walking described above, each limb or joint CPG may have quite different intrinsic properties.
Most progress in understanding the mechanisms of rearrangement of oscillatory networks has been accomplished in the crustacean stomatogastric nervous system (for review, see Dickinson and Moulins, 1992). In that system, four CPGs control body parts that are responsible for separate components of the ingestive behavior. Thus, although interactions do occur, the CPGs always produce rhythms with distinct cycle periods. In Aplysia, two body parts, the lips and radula, must be coordinated on a cycle-by-cycle basis during various feeding-related behaviors. In this paper, we show that, as postulated for locomotion, separate CPGs control each body part and that the interactions between these CPGs alter during different motor programs.
The consumatory phase of feeding in Aplysia consists of rhythmic movements of the radula, which are coordinated with appropriate opening and closing of the lips (Morton and Chiel, 1993a). The intrinsic buccal mass muscles that control the movement of the radula are innervated by motoneurons in the buccal ganglion (Cohen et al., 1978; Church and Lloyd, 1994). Firing of these motoneurons during feeding-related behaviors is controlled by CPGs, the neuronal circuitry for which appears to reside largely in the buccal ganglion (Weiss et al., 1982; Susswein and Byrne, 1988; Kirk, 1989). Motoneurons innervating lip muscles are located in the cerebral ganglion (Chiel et al., 1986; Rosen et al., 1989). It was previously considered likely that behaviorally appropriate output from these cerebral motoneurons was coordinated from the buccal ganglion. Each cerebral motoneuron was thought to be directly activated by a number of buccal-to-cerebral interneurons (BCIs) (Chiel et al., 1988; Teyke et al., 1993; Rosen et al., 1991a).
Here, we present evidence that each cerebral hemiganglion contains a CPG that can produce a cerebral motor program (CMP) in isolated cerebral ganglia. The cerebral CPGs provide an alternative means of activating lip motoneurons to the direct control by BCIs. We characterize neuron C15, which drives the CMP, describe basic features of the CMP, including synaptic drive to identified neurons, and demonstrate that different forms of behaviorally relevant motor programs are accompanied by changes in the interactions between cerebral and buccal CPGs.
MATERIALS AND METHODS
Experiments were performed at 15°C on Aplysia californica weighing 100–300 gm. Animals were anaesthetized by an injection of isotonic MgCl2 (50% of body weight) into the body cavity. The cerebral ganglion was dissected out in isolation for some experiments and removed with the buccal ganglion attached for others (see Results for details). The ganglia were pinned onto a SYLGARD-coated dish at room temperature in artificial sea water (ASW). Cerebral ganglia were pinned ventral-surface-up and buccal ganglia rostral-surface-up. The connective tissue overlying the uppermost surface of the ganglia was surgically removed. The cerebral–pedal connective was rotated through 180° and repinned to allow access to the lateral area of the E-cluster. Simultaneous intracellular recordings were made from up to four neurons using single or double-barreled microelectrodes filled with 2 m potassium acetate and beveled to a resistance ranging from 5 to 10 MΩ, depending on the size of the cell to be penetrated. In preparations where the buccal ganglion was attached, buccal motor output was monitored using a polyethylene suction electrode placed over buccal nerve 2 (bn2) (see Morton and Chiel, 1993a) and connected to an AC amplifier. A recording or stimulating suction electrode was placed over the end of the severed cerebral-to-buccal connective (CBC) in experiments involving isolated cerebral ganglia.
To help identify previously described cells and to reveal the morphology of new ones, microelectrodes, beveled to a resistance of 10–15 MΩ, were filled with a 3% solution of 5(6)-carboxyfluorescin in 0.1 m potassium citrate, titrated to pH 8.0 with KOH (see Rao et al., 1986). At the end of each experiment, previously described cells were identified on the basis of their morphology and known synaptic connections. In some experiments, polysynaptic pathways were suppressed using solutions containing elevated levels of divalent cations (with both Ca2+ and Mg2+ at three times the normal concentration) to raise neuronal firing thresholds.
To determine possible peripheral functions of C15, experiments were performed on preparations in which the cerebral ganglion was removed with the lips and anterior tentacles still attached by their peripheral nerves (cf. Rosen et al., 1979). The ganglion was pinned on an elevated SYLGARD platform, ventral-surface-uppermost, and desheathed as described above. The head structures were pinned, ventral-surface-uppermost, to the bottom of the SYLGARD-coated dish in such a way as to reveal the inner portions of the lips and perioral zone (Rosen et al., 1982). The anterior aorta was cannulated and perfused with ASW.
Salines, all at pH 7.6, were composed as follows. ASW: NaCl, 460 mm; KCl, 10 mm; CaCl2, 11 mm; MgCl2, 55 mm; NaHCO3, 5 mm. High divalent saline (3× Mg2+, 3× Ca2+): NaCl, 460 mm; KCl, 10 mm; CaCl2, 33 mm; MgCl2, 165 mm; NaHCO3, 5 mm. For experiments in which chemical transmission was blocked, the Ca2+ was omitted and replaced with 10 mm Co2+. Experiments were performed at least six times, unless otherwise stated. Figures quoted are means ± SE.
Morphology and synaptic connections of C15
We have discovered a neuron, C15, which was uniquely identified on the basis of its physiological and morphological properties. There was a single C15 in each half of the cerebral ganglion. It was a medium-sized oval neuron (∼100–150 μm in diameter) situated medially on the ventro-lateral surface of the E-cluster (Jahan-Parwar and Fredman, 1976). The approximate positions of C15 and other cerebral neurons used in this study are shown schematically in Figure1 A. Injections of 5(6)-carboxyfluorescin into C15 revealed two peripheral axons, leaving the upper labial (ULAB) and anterior tentacular (AT) nerves (Fig.1 B). Dendritic arborisations were present in the E-cluster at the base of the CBC and in the M-cluster (Ono and McCaman, 1980). Processes were limited to the ipsilateral half of the ganglion.
In preparations in which the head structures were left connected to the cerebral ganglion, firing C15 at frequencies above 8 Hz produced a radial contraction in the tissue immediately adjacent to the anterior perioral zone of the ipsilateral inner lips. This resulted in an opening of the anterior portion of the jaws (n = 5). This contraction was observed before the occurrence of large IPSPs in C12, which would indicate the start CMP (see below). Firing C15 at these frequencies also readily produced the contraction while the whole preparation was bathed in solution containing 3× normal Ca2+ and Mg2+ (to suppress polysynaptic pathways), conditions under which the CMP was completely abolished. This means that the contraction was likely to be a direct effect of firing C15, which can therefore be described as a probable lip motoneuron. Mechanical or chemical (seaweed; Laver, Vega trading company, NY) stimuli applied to the lips, rhinophores, or tentacles had no effect on the membrane potential of C15, so we have no evidence for any sensory functions.
Two neighboring E-cluster neurons, C16 and C17, which are smaller and generally more rostral than C15, were also characterized in this study. C16 and C17 had physiological properties and synaptic inputs (see below) which were indistinguishable from C15. These neurons produced radial contractions in the tissue of the ipsilateral inner lips that persisted in high divalent solutions, like C15, so are also probable lip motoneurons. While C15 opened the anterior part of the jaws, C16 opened the medial region, and C17 the posterior. Firing all of these motoneurons together in the semi-intact preparation resulted in full opening of the ipsilateral jaws along their entire length. The motoneurons could also be distinguished from each other by the ability of C15 to reliably drive the CMP and, more important, by their morphology, revealed by injection of 5(6)-carboxyfluorescin. Not only were C16 and C17 generally smaller than C15, but they sent axons to the periphery via different combinations of nerves. C16 had axons in the AT and lower labial (LLAB) nerves and C17 had a single axon in the LLAB (Fig. 1 C,D). These three motoneurons were electrically coupled to each other. Either a hyperpolarizing or a depolarizing current injected into C15, C16, or C17 resulted in a change of the same polarity but reduced amplitude in the membrane potential of the other two neurons (Fig. 2 A). The coupling ratios (the voltage change in the postsynaptic neuron divided by the voltage change in the neuron into which current was injected) were between 0.1 and 0.2. Apparent coupling was higher when current was injected into C15 (Fig. 2 A), possibly because of its larger size. Because C2 produced an IPSP in these neurons (see below), C15, C16, and C17 probably correspond to the Ea cluster of neurons of Chiel et al. (1986). These were described as inhibitory followers of C2, putative motoneurons for the lips that were in approximately the same position in the E-cluster.
C15 received chemical input from several cerebral and buccal neurons. The histaminergic mechanoafferent C2, which may play an important role in several feeding-related behaviors (Chiel et al., 1986), inhibited C15 (Fig. 2 B) as well as C16 and C17. C15 also received input from several buccal-to-cerebral interneurons (BCIs: B18, B19, and B24), which are responsible for direct feedback to cerebral neurons during BMPs (Chiel et al., 1988; Rosen et al., 1990;Rosen et al., 1991a; Teyke et al., 1993). Spikes in B18 were followed one-for-one by fast EPSPs in C15, whereas firing in B19 produced a rapid inhibition in C15 followed by a slow excitation and B24 produced just an inhibitory response (Fig. 2 C–E). All of these responses persisted in salines containing 3× normal Ca2+and Mg2+. Inputs from B18, B19, and B24 onto C16 and C17 were similar to those onto C15. There appeared to be no monosynaptic connections onto C15, C16, and C17 from another BCI, B20 (Teyke et al., 1993), although they all received excitatory input during the BMP driven by that cell.
C15 can drive a motor program in the isolated cerebral ganglion
At its resting potential (−44.4 ± 3.8 mV, n = 15) C15 usually received no large spontaneous synaptic inputs and was silent (e.g., Figs. 3 A, 4, 8C) or weakly tonically active (e.g., Figs. 3 B,8 A,B). Some low-amplitude spontaneous input was present, because the resting membrane potential showed less short-term variations in the presence of 3× Ca2+, 3× Mg2+, which presumably reduced presynaptic spiking activity (Fig. 2 D,E). When C15 was fired at frequencies above 11–12 Hz, rhythmic synaptic activity with a cycle period of between 9 and 50 sec was recorded in a variety of E- and M-cluster neurons (Figs.3, 4, 8). In 25 out of 30 preparations, C15 produced repeating cycles of the program. In the remaining cases, only a single cycle was initiated, despite continued high-frequency firing of C15. Fast PSPs observed in some neurons (see below) were not one-for-one with spikes in C15 and they were completely abolished by 3× Ca2+, 3× Mg2+. This indicates that the responses detailed below were not monosynaptic from C15, but arose through a polysynaptic pathway, after the activation of a cerebral CPG. The rhythmic activity could be induced in the isolated cerebral ganglion and, because it incorporated firing of lip motoneurons (see below) at frequencies that produce lip movements, it can be defined as a cerebral motor program (CMP). The CMP stopped within 100 msec, after cessation of current injection in C15, even if this occurred in the middle of a cycle. Spontaneous CMPs were never observed. C15 could reliably drive the CMP in preparations in which the cerebral ganglion had been isolated for up to 36 hr. C16 and C17 could occasionally drive a single cycle of the CMP (4 of 25 preparations), but never repeating cycles, unless both were depolarized together, and interactions of this sort were not thoroughly investigated.
To characterize more fully the CMP, we investigated the input received by a variety of identified cerebral neurons. Lip motoneurons C11 and C12 were active during the CMP (Fig. 3 A). Firing in C11 was caused by barrages of EPSPs, whereas firing in C12 relied, in part at least, on postinhibitory rebound from a barrage of IPSPs, because the cell could also fire a burst of spikes at the end of a negative current pulse. Lip motoneurons C16 and C17 received large, fast EPSPs that drove high-frequency firing (Fig. 3 B). Because of the electrical coupling of these neurons to C15, they were also tonically depolarized throughout the period of current injection in C15. C15 itself also received rhythmic input (open bar in Fig.3 B), which caused higher-frequency spiking (15–20 Hz) during those periods. In most records, the exact nature of this input was obscured by the current-induced spikes but, in cases where the CMP continued for a few tens of milliseconds after the current pulse, the input could be seen to consist of fast EPSPs, similar to those seen in C16 and C17.
In addition to motoneurons, we also recorded from a sample of previously identified sensory and modulatory cerebral neurons that participate in feeding motor programs (Chiel et al., 1986). The E-cluster neuron C4 can modulate contractions of extrinsic buccal mass muscles produced by cerebral motoneurons (Chiel et al., 1986). During the CMP, C4 received large EPSPs that drove it above threshold (Fig.4 A). C5, which is neighboring C4 but whose function is unclear, received relatively small EPSPs, one-for-one with those in C4, but these had little influence on its firing rate (Fig. 4 A). The giant serotonergic metacerebral cell (MCC), which exerts modulatory influences both centrally and peripherally during feeding (Weiss et al., 1978; Rosen et al., 1989), received rhythmic weak, slow excitation but did not fire (Fig.4 B). The histaminergic mechanosensory neuron C2, which has a powerful influence over many neurons involved in the cerebral feeding motor circuitry (Chiel et al., 1986), received fast IPSPs (Fig. 4 B).
Pairwise recordings revealed that the fast PSPs described above occur synchronously in all neurons, suggesting a common source (see expanded time-base records in Figs. 3, 4 A). Occasionally, there were also fast EPSPs in the MCC, associated with the peak of the slow depolarization, but these were not one-for-one with the PSPs in the other cells (Fig. 4 B). Only cells in the ipsilateral half of the ganglion appear to receive synaptic input during the CMP, because no rhythmic activity could be recorded contralaterally (Fig. 11 A). There were only two discernible phases during the CMP, one being characterized by the presence of fast PSPs in the neurons described above and the other by firing in C12 and no observable input to any other neurons. The CMP, therefore, appears to be a less complex rhythm than the usually multiphase BMPs.
Input to buccal motoneurons during the CMP
In preparations with the buccal ganglion attached, a CMP that appeared identical to that produced in the isolated cerebral ganglion could still be driven by C15. Several identified cerebral-to-buccal interneurons (CBIs) fire during the CMP (see below) and are known to make synapses onto buccal motoneurons (Rosen et al., 1991b), so one might expect feedforward synaptic inputs from these neurons. Spiking of cerebral-to-buccal neurons was also monitored in isolated cerebral ganglia by an extracellular suction electrode placed on the CBC. An increase in activity recorded in this electrode was always observed during the phase of the CMP in which fast PSPs occurred in C2, C4, C5, C11, C12, C16, and C17. This might be explained, in part, by the increase in spiking activity observed in CBI-2, CBI-3, and CBI-4 (described below). However, there was also a series of spikes in the CBC that were one-for-one with the fast PSPs in C12 (Fig.5 A). Firing in the identified CBIs occurred at a later phase than these spikes and was never one-for-one with the individual fast PSPs seen in other neurons (Figs. 8,9 B2,C2). This means there exists an as yet unidentified cerebral-to-buccal neuron that is responsible for this activity. Intracellular recordings from the buccal ganglion revealed that a number of neurons in the ventral motoneuron cluster received fast IPSPs during the CMP that were one-for-one with the fast PSPs in the cerebral ganglion (Fig. 5 B). This suggests that the unidentified cerebral-to-buccal neuron makes synapses within the buccal ganglion, providing another source of input to buccal neurons, in addition to those arising via identified CBIs.
The contribution of C15 to the CMP
We have shown that C15 is sufficient to generate the CMP. To assess more fully the contribution of C15 to the operation of the cerebral CPG, we investigated if C15 firing was also necessary for the CMP (Kupfermann and Weiss, 1978). We therefore searched for other ways of driving the CMP to examine this possibility. InPleurobranchaea, it has been shown that high-frequency stimulation of the CBC can drive a rhythm in the isolated cerebral ganglion (Davis et al., 1973; Cohan and Mpitsos, 1983), so we attempted the equivalent experiment in Aplysia. We found that stimulating the CBC at frequencies above 10 Hz resulted in rhythmic input to cerebral neurons (Fig. 6). Because lip motoneurons C11, C15, C16, and C17 were active during this rhythm, this can also be classified as a CMP. All of the recorded neurons had the same category of input during both C15- and CBC-driven CMPs. Thus, C4, C5, C11, C15, C16, and C17 all received fast EPSPs, whereas C2 and C12 received fast IPSPs. The cycle period of the CBC-driven CMP was 20–60 sec, which overlaps the values for the C15-driven rhythm. As with the C15-driven CMP, the source of these PSPs appeared to be an unidentified cerebral-to-buccal neuron, because the PSPs followed CBC stimulation with a fixed latency, regardless of stimulation intensity or frequency (Fig. 7 A). Identified CBIs were ruled out as candidates because either no antidromic spikes were observed at the stimulus intensity used (CBI-2 and CBI-4) or blocking these spikes by hyperpolarization had no effect on the rhythm (CBI-1 and CBI-3). Because of these similarities, the CMPs driven by C15 depolarization or CBC stimulation appear to be very closely related and are likely to be produced by the same CPG. To test whether spiking in C15 is necessary for the occurrence of the CMP, we hyperpolarized it during the CBC-driven CMP to a level at which it no longer spiked. Under these conditions, the CMP could still be driven, suggesting that spiking in C15 is not necessary for the rhythm (n = 6; Fig.6 B).
Using CBC stimulation, we were also able to investigate the nature of the possible connection between C15 and the unidentified cerebral-to-buccal neuron. For these experiments, the CBC was stimulated at low frequency (1–2 Hz, insufficient to induce the CMP) at an intensity just suprathreshold for the PSPs in C15, C12, etc. Under these conditions, hyperpolarizing C15 increased the amplitude of the EPSP in that cell, without affecting the amplitude of PSPs in other cells (Fig. 7 B1). However, we found that by further hyperpolarizing C15, we were able to block the PSPs in C15 as well as the PSPs in other neurons (Fig. 7 B2). This suggests that C15 is electrically coupled to the unidentified cerebral-to-buccal neuron so that the hyperpolarization in C15 is transferred electrotonically, sufficiently hyperpolarizing the terminals of the other neuron to prevent transmitter release. It should be noted that a chemical synapse from the cerebral-to-buccal neuron onto C15 also exists, producing the EPSPs during the CMP, and that a chemical synapse in the other direction cannot be ruled out yet. The electrical connection between C15 and the cerebral-to-buccal neuron means that although spiking in C15 is not necessary for the CMP, a certain level of membrane potential is necessary. Thus, if C15 is strongly hyperpolarized (to presumably nonphysiological levels), it blocks the PSPs that follow CBC stimulation and, therefore, no program can be generated.
Connections between cerebral and buccal CPGs
Cerebral motoneurons receive input during BMPs (Rosen et al., 1991b) (see Figs. 10, 11), and we have shown above that buccal motoneurons receive input during the CMP (Fig. 5). We also wished to characterize possible interactions between the cerebral and buccal CPGs themselves. To this end, we investigated whether neurons that are important during the operation of one type of program (either the CMP or various BMPs) received input during the other program. InAplysia, we found the following. (1) Four identified cerebral-to-buccal interneurons (CBIs) are important in the control of BMPs (Rosen et al., 1991b) (see Fig. 10). We therefore investigated input received by these neurons during the CMP. (2) C15 is important in the control of the CMP (because it is sufficient to drive it and receives excitatory input during the program). We therefore investigated input received by this neuron during BMPs.
Inputs to CBIs during the CMP
CBI-1 received weak, slow inhibition during the CMP (Fig.8 A), which was not studied further. CBI-2 and CBI-4 both received slow excitation, which was usually sufficient to produce spiking (Fig. 8 B, 9B,C). Because firing C15 produced this excitation via a polysynaptic pathway (because the CMP was blocked in high divalent solutions), a detailed examination of the excitation in CBI-2 and CBI-4 was not attempted. However, it was apparent that in both cases the amplitude of the excitation was decreased by hyperpolarization of the CBIs (see Fig. 9 B,C). CBI-3 was also excited during the CMP (Fig. 8 C). We found that CBI-3 was strongly electrically coupled to C4. Injecting either a positive or a negative current into either CBI-3 or C4 altered the membrane potential of the other (Fig.8 D), the coupling ratio being 0.20–0.25. Spikes in one cell produced transient depolarizations of the other with zero latency. Bathing the ganglion in ASW with 10 mmCo2+ had no effect on these potentials, so there appears to be no chemical connection. The excitation of CBI-3 during the CMP could be effectively mimicked by positive current injection into C4 (Fig.8 E). This suggests that the excitation of CBI-3 during the CMP was largely attributable to electrotonic transfer of the chemical EPSPs and spiking in C4 described above. The input to all of the CBIs occurred during the same phase as the fast PSPs in the cerebral neurons described previously, although the firing in CBI-2 and CBI-4 was somewhat delayed and lasted longer, presumably because of the slow nature of the excitation.
In preparations with cerebral and buccal ganglia connected, the same inputs were observed for each of the CBIs during the CMP. Inducing the CMP by exciting C15 most often had little effect on buccal motor output, monitored via the extracellular recording electrode on bn2 (Figs. 9 B,C, 14). However, in a significant number of preparations (5 of 23) single cycles of a BMP (defined as strong, multiphasic activity in bn2 and high-frequency firing in B4) were associated with the first cycle of the CMP (Fig. 9 A). In only one case did consecutive cycles of the CMP produce repeating cycles of a BMP. Both CBI-2 and CBI-4 can be excited beyond threshold during the CMP. Because spiking in either CBI-2 or CBI-4 can drive BMPs (Rosen et al., 1991b) (see Fig. 10), this is a possible pathway for the generation of BMPs. This was tested by altering the membrane potential of CBI-2 and CBI-4 in preparations with the buccal ganglion attached. In the six preparations investigated in this manner, BMPs were not induced by C15 stimulation with all other neurons at their resting potentials. However, a full cycle of a BMP could be generated by slightly depolarizing CBI-2 or CBI-4, thus increasing the frequency at which they fired during the CMP (Fig. 9 B,C). These results are not direct evidence that the CMP occasionally generates BMPs via excitation of CBI-2 and CBI-4 (or the combined action of both), but do show that this pathway is feasible.
Input to C15 during BMPs
In cerebral–buccal ganglia preparations, we also investigated the input to C15 during BMPs evoked by a variety of means. During the ingestive-like BMP driven by positive current injection into CBI-2 (cf.Rosen et al., 1991b), C15 received fast EPSPs that drove spiking at high frequencies (maximum frequency sustained over 1 sec = 24.5 ± 1.3 Hz, n = 6; Figs.10 A, 11B, 12A). C15 was also excited, although more weakly (maximum firing frequency, 14.3 ± 2.3 Hz, n = 5) during egestive-like BMPs (Fig. 10 B,C). These were driven by a brief burst of high-frequency (10 Hz) electrical shocks to the radula nerve or by constant stimulation of the esophageal nerve at 2–3 Hz (Susswein and Byrne, 1988). C15 was also excited during BMPs driven by both CBI-1 (maximum firing frequency of C15 = 19.4 ± 2.7 Hz, n = 5; Fig. 13 A) and CBI-4 (maximum firing frequency = 16.0 ± 1.5 Hz, n = 6; Figs. 10 D, 13 B). During BMPs initiated by any means, C15 was also occasionally inhibited between bursts (arrowheads, Figs. 10 D,11 B). The CMP was strictly limited to one-half of the cerebral ganglion (Fig. 11 A), but the general timing of excitatory inputs to both right and left C15s was highly synchronized during BMPs (Fig. 11 B), although individual EPSPs were not always one-for-one.
Although qualitatively similar (consisting of fast EPSPs), the input to C15 was not quantitatively the same for each variety of BMP. The firing rate of C15 was significantly higher (p < 0.01) during the BMP induced by CBI-2 (mean firing rate measured over the entire burst was 19.3 ± 0.7 Hz, n = 8) than that during CBI-1 or CBI-4 induced BMPs (9.9 ± 1.6 Hz,n = 5 and 9.5 ± 0.9 Hz, n = 6, respectively). Both of these firing rates were again significantly higher (p < 0.01) than during egestive-like BMPs (4.3 ± 1.0 Hz, n = 6). Note that the maximum firing rates described above were also different between the various BMPs.
C15 was active during protraction for each type of BMP. Thus, during ingestive BMPs (Morton and Chiel, 1993b), such as that driven by CBI-2, buccal motoneurons B8 and B4 are active at the same time (Rosen et al., 1991b; Church and Lloyd, 1994). During this program, C15 was active before B8/B4 (Fig. 10 A), during protraction, and at the same time as “p-group” of neurons of Church and Lloyd (1994). During egestive-like motor programs, B8 and B4 are active out of phase (Morton and Chiel, 1993b; Church and Lloyd, 1994). In this program, C15 fires at the same time as B8 and stops before firing in B4 (Fig.10 B,C). This is also during protraction, but in this case at the same time as “c-group” neurons of Church and Lloyd (1994). C16 and C17 received similar input and were active at similar frequencies and with the same timing as C15. The timing of the activity in these lip motoneurons relative to buccal motoneurons is consistent with the timing of jaw opening compared to radula movements during the two behaviors. During ingestion, the jaws are open while the radula is protracting and open, whereas during egestion the jaws are open while the radula is protracting and closed (Morton and Chiel, 1993a). It is possible that the BMP driven by CBI-4 underlies a swallowing behavior, so the second burst of spikes in B8, which coincides with inhibition of C15, during this motor program (Fig. 10 D) may occur during hyper-retraction of the radula.
The interactions between the cerebral and buccal CPGs change during different feeding-related motor programs
Cerebral neurons receive input during BMPs partly as a result of direct, monosynaptic feedback from buccal-to-cerebral neurons, via axons in the CBC (Chiel et al., 1988; Teyke et al., 1993). However, during BMPs, C15 fired at frequencies that could drive the CMP. Because the cycle periods of BMPs were close to those for the CMP, C15 only fired at high frequencies long enough to drive part or all of a single CMP cycle for each BMP cycle. Nevertheless, this may be an important means of amplifying other synaptic inputs to some cerebral neurons during BMPs. We therefore investigated the effect of altering the membrane potential of C15 on input to cerebral neurons during BMPs. We generally used the large IPSPs observed in C12 during the CMP as a convenient monitor, but the results were also confirmed using the large EPSPs in C4.
For BMPs induced by CBI-2, large IPSPs were always observed in C12. These IPSPs were in phase with the excitation in C15, suggesting that the feedback loop through C15 to the cerebral CPG could contribute to their formation. If this were the case, then hyperpolarizing C15 might prevent the activation of the cerebral CPG, and so block some of these IPSPs. However, no reduction in the frequency of these IPSPs was observed even after quite extreme (and probably nonphysiological) hyperpolarization of C15 (Fig. 12).
In BMPs driven by both CBI-1 and CBI-4, large IPSPs in C12 were only sometimes observed (4 of 10 preparations), but when present were again in phase with the peak excitation in C15. In these types of BMP, however, the presence of IPSPs in C12 was critically dependent on the membrane potential of C15 (Fig. 13 A,B). If IPSPs were present, they could be blocked by hyperpolarizing C15 (n = 4; Fig. 13 A). Conversely, if no IPSPs were present, they appeared if C15 was depolarized by 4–10 mV, below the threshold for driving the CMP in the quiescent state (n = 6; Fig. 13 B). This suggests that the IPSPs in C12 during these BMPs result from activity in the cerebral CPG, driven at least partly by the phasic excitation in C15.
A third situation was observed for the egestive-like BMP driven by stimulation of the esophageal nerve. In this BMP, no large IPSPs were observed in C12 at any phase of the rhythm. Additionally, large IPSPs could not be induced by strong depolarization in C15 during the peak of its synaptic excitation (Fig. 14). During the egestive-like BMP, C12 received PSPs not correlated with the current pulse in C15, presumably attributable to direct input from BCIs. These PSPs caused a drop of ∼15% in the input resistance of C12 measured by constant hyperpolarizing current pulses. This relatively small change in the input resistance of C12 is unlikely to reduce the IPSPs during the CMP below detectable amplitude. For comparison, the decrease in input resistance in C12 caused by the IPSPs of the CMP itself was as much as 50%. Even current levels well above the threshold for driving the CMP in the quiescent state could not drive the CMP during this BMP. Thus, not only does C15 fire at a lower frequency during the egestive-like BMP, but it is rendered incapable of driving the CMP .
A separate cerebral CPG
Our findings indicate that there is a CPG within the cerebral ganglion of Aplysia. In most other molluscan feeding systems, there is little evidence for separate cerebral CPGs, possibly because studies have concentrated on buccal motor circuitry. InLymnaea, rhythmic synaptic inputs were recorded in some neurons in isolated cerebral ganglia (McCrohan, 1984), but the nature of this “oscillator” was not investigated further. InPleurobranchaea, cerebral interneurons that drive buccal feeding rhythms [the phasic paracerebral (PCp) cells] also drive rhythmic motor output in isolated cerebral ganglia (Davis et al., 1984). The cerebral rhythm in Pleurobranchaea could only be obtained for about an hour after CBC transection. InAplysia, however, the CMP could be activated in cerebral ganglia 36 hr after CBC transection, showing that the cerebral CPG does not require short- or long-term feedback from the buccal ganglion to maintain its rhythm-generating ability.
It should be noted that there are two separate cerebral CPGs, one in each hemiganglion, and each can be active independently. This feature may be important in producing asymmetrical lip movements during feeding behaviors if the exposure of seaweed to the lips is unilateral. Once the seaweed is positioned centrally and ingestion has begun, activity in the two cerebral CPGs will be synchronized because of simultaneous excitation of both C15s by the buccal CPG (Fig. 11), resulting in symmetrical lip movements.
The fast PSPs observed in cerebral neurons during the CMP do not represent monosynaptic connections from C15. The fact that the PSPs are synchronous in all neurons suggests a common source, but this remains to be identified. It is likely that this common source is the unidentified cerebral-to-buccal neuron that produces (1) the spikes in the CBC that are one-for-one with PSPs in cerebral neurons during the C15-induced CMP, and (2) The constant-latency PSPs after CBC stimulation. The discovery of this neuron will be an important step in elucidating how the CMP is generated.
Because C15 is a lip motoneuron, driving the CMP via intracellular excitation of C15 caused strong tonic contraction of the inner lips that masked any superimposed rhythmic movements. However, during the CMP other cerebral motoneurons (such as C16 and C17) fire at frequencies that result in muscle contraction in the semi-intact preparation. Therefore, the cerebral CPG may be capable of producing rhythmic lip movements. Supporting evidence for such a role is provided by studies showing that lip movements can occur after CBC transection (Kupfermann, 1974), or in the absence of ingestive movements of the buccal mass (during feeding: Hurwitz et al., 1996; during egg laying:Arch and Smock, 1977). Even if no behavioral correlate for the CMP is discovered, the changing interactions of cerebral and buccal CPGs are likely to play a role in shaping the coordination of lips and radula, as discussed below.
The functional significance of C15
C15, C16, and C17 cause contraction of the inner lips, resulting in jaw opening. These motoneurons are active at phases of ingestive and egestive BMPs when the jaws are opening and fire at a high enough frequency during these BMPs to cause muscle contraction, so are likely to contribute to jaw opening during feeding behaviors. The higher firing rate of C15 during ingestive rather than egestive motor programs agrees with behavioral observations that the lips fully open during ingestion, but remain mostly closed during egestion (Morton and Chiel, 1993a).
C15 is a motoneuron capable of driving a CPG. There is evidence that motoneurons supply excitation to the vertebrate spinal locomotor pattern generator (Perrins and Roberts, 1995) and in several invertebrate systems motoneurons supply feedback to CPGs (Kristan and Calabrese, 1976; Heitler, 1978; Simmers and Bush, 1983) or can drive a repeating rhythm (buccal CPGs of Tritonia: Willows, 1980;Aplysia: Hurwitz et al., 1994). C15 is also an important contact point between the cerebral and buccal CPGs, as discussed below.
Interactions between cerebral and buccal CPGs and their functional significance
It was previously thought that input to cerebral motoneurons was controlled by direct feedback from buccal circuitry (Fig. 15 A). Theoretically, the different coordination of lips and buccal mass seen during various feeding-related behaviors could be generated by a single buccal CPG if various motoneuron pools were decoupled during different behaviors. However, in this study we show the existence of a CPG in each cerebral hemiganglion that can also activate cerebral neurons and that interacts with buccal motor circuitry in various ways (Fig. 15 B). Both buccal and cerebral CPGs have direct inputs to motoneurons in the other ganglion and, as previously thought, changes in the input from buccal circuitry to cerebral motoneurons will be vital in altering motor output during different behaviors. However, there are also other potential sites of modification.
The cerebral and buccal CPGs themselves are connected, because neurons that drive one program receive input during the other program. Thus, CBI-2 and CBI-4 (which drive BMPs) were excited during the CMP and, in turn, C15 (which drives the CMP) was excited during BMPs. It seems likely that the CPGs also interact in other ways. Rather than attempt to characterize all possible connections, we tried to ascertain whether such interactions alter during different motor programs. We show that for each variety of BMP (i.e., each BMP driven by a different means, such as intracellular stimulation of CBIs or esophageal nerve stimulation) the excitation in C15 changes both its amplitude and its duration. Because C15 is a motoneuron, it could be argued that these changes reflect a need to change the extent and timing of jaw opening and have no functional relevance for interactions of buccal to cerebral CPG.
Evidence that interactions between the CPGs can be changed comes from experiments in which C15 alters the input received by other cerebral neurons (monitored using the large IPSPs in C12) during BMPs driven by some methods, but not others. We investigated BMPs driven by three means, as follows. (1) Depolarization of CBI-1 or CBI-4. Depolarizing C15 during these BMPs increased the number of IPSPs in C12, whereas hyperpolarizing C15 prevented them altogether. This suggests that the synaptic excitation received by C15 during these BMPs is sometimes sufficient to drive part of a cycle of activity in the cerebral CPG. Altering the excitation received by C15 by injecting positive or negative current either increases or suppresses this phasic activation of the cerebral CPG. The firing rate of C15 during naturally evoked behaviors could be altered by the various inhibitory and excitatory connections onto it, as shown in this study. For example, the mechanosensory neuron C2 inhibits C15 and is active during feeding (Chiel et al., 1986; Weiss et al., 1986). (2) Depolarization of CBI-2 (an ingestive-like BMP). In this program, large IPSPs were present in C12, regardless of the membrane potential of C15. There are three possible explanations for this. First, the IPSPs in C12 during this program may arise entirely from direct feedback from BCIs (connection1a in Fig. 15 B). Second, the cerebral CPG may be so strongly excited by direct feedback from the buccal CPGs (connection2a) that even extreme hyperpolarization of C15 cannot prevent its activation. Third, the connection from C15 to the cerebral CPG may be reduced in strength. (3) Stimulation of the esophageal nerve (an egestive-like BMP). Here, large IPSPs were not observed in C12 and could not generated by depolarization of C15 at levels suprathreshold for the CMP in the quiescent state. The inability of C15 to drive the CMP during this BMP might be attributable to decoupling at any of several sites. For example, the rhythm-generating ability of the cerebral CPG might be suppressed, the excitatory connection from C15 to the cerebral CPG might be reduced, or the synaptic outputs of the cerebral CPG might be inhibited.
How might activity in the cerebral CPGs contribute to specific aspects of feeding behavior? A possible role in generating asymmetrical lip movements was addressed earlier. Additionally, we can speculate how specific cerebral neurons may be influenced during protraction, the time when C15 receives excitation, thus possibly activating the CMP. For example, the mechanoafferent C2 is inhibited during the CMP and appears to be least active during protraction in the semi-intact preparation (Weiss et al., 1986). The CMP, therefore, might contribute to a relative enhancement of proprioceptive input from the lips and peri-oral zone during retraction, when seaweed is being ingested.
In conclusion, there are distinct buccal and cerebral CPGs, these CPGs interact at several points, and the interactions between these CPGs vary during different motor programs. Although the mechanisms underlying the changes in the interactions are not yet known, such variation does exist. Most obviously, C15 can drive the cerebral CPG during CBI-1 and CBI-4 driven BMPs but cannot during egestive-like BMPs. These results lend support to two hypotheses: first, that there is often a separate CPG for each individual body part; second, that the interconnections between such CPGs alter during different motor programs. Future studies will need to determine whether such changes represent a mechanism for altering the coordination of separate body parts on a cycle-by-cycle basis.
This work was supported by Human Frontiers Science Program Grant LT-561/95 and National Institutes of Health Grants GM 32009 and MH 50235. We thank Drs. E. C. Cropper and C. G. Evans for comments on earlier versions of this manuscript.
Correspondence should be addressed to Ray Perrins, Department of Physiology and Biophysics, Mount Sinai School of Medicine, Mount Sinai Medical Center, One Gustave L. Levy Place, New York, NY 10029-6574.