Command Neurons for Locomotion in Aplysia

JOURNALOF NEUROPHYSIOLOGY Vol. 49, No. 5, May 1983. Printed in U.S.A.

Command Neurons for Locomotion in Aplysia

STEVEN M. FREDMAN AND BEHRUS JAHAN-PARWAR

Worcester Foundation for Experimental Biology, Shrewsbury, Massachusetts 01545

SUMMARY AND CONCLUSIONS

1. Command neurons for the pedal wave motor program (PWMP) that drives locomotion in Aplysia were examined electrophysiologically in the isolated nervous system. The cerebral ganglion contained a minimum of four command (type I) neurons that were excited by stimuli known to trigger locomotion in vivo and evoke its neural correlates in vitro.

2. Type I neurons fired during both spontaneous and sensory evoked pedal wave motor program bursting in pedal nerves and motor neurons. Intracellular stimulation of type I neurons at physiological frequencies (2-8 Hz) was sufficient to initiate the pedal wave motor program in quiescent preparations.

3. Increasing the type I neuron firing frequency decreased both the latency to the first pedal wave burst and the period of ongoing bursts. Driving type I neurons maintained the motor program as long as the stimulation continued. Repeated type I neuron stimulation caused decrement of the evoked motor program.

4. Hyperpolarizing type I neurons blocked both spontaneous and sensory evoked pedal wave motor program bursting. Type I neurons thus appeared to be true command neurons, being both sufficient and probably necessary for pedal wave motor program generation.

5. The command neurons exhibited similar firing patterns due to common synaptic input. They were not monosynaptically or electronically coupled and exhibited reciprocal recurrent inhibition among themselves. Driving one command neuron inhibited the others, indicating that they were individually sufficient and did not function as a mutually

excited network. Thus they constitute a redundant command system.

6. There were no direct synaptic connections between the command neurons and pedal motor neurons. The command neurons did receive excitatory synaptic input during pedal wave bursts, indicating positive feedback from the oscillator.

7. In addition to the type I neurons, other neurons were found that appeared to have a modulatory function. Type II neurons could also initiate weak bursting but did not appear necessary for motor program generation. Types IIIA and IIIB neurons, respec- tively, monosynaptically excited and inhibited pedal motor neurons but could not initiate pedal wave bursting. Type IIIC neurons appeared to inhibit the pedal wave oscillator. The organization of the cerebral command and modulatory neurons in the control of locomotion was considered.

INTRODUCTION

Locomotion is a major component of many behaviors in Aplysia and other animals. As a result, the problem of how locomotion is controlled (initiated, maintained, coordinated) is of considerable interest. Lo- comotion in Aplysia has as its major component pedal waves that pass rostra1 to cau- da1 along the length of the foot, moving the animal forward. Regardless of whether locomotion is initiated by any of several sensory modalities (24) or is spontaneous, pedal waves are present. At any given rate of locomotion, irrespective of how it was triggered, the waves are indistinguishable. Fur- thermore, although the foot of Apfysia is used in behaviors other than locomotion, such as feeding, pedal waves are present only during locomotion. Pedal waves thus serve as an in-

1092 0022-3077/83/0000-OOOO$O 1 SO Copyright 0 1983 The American Physiological Society

PEDAL WAVE COMMAND NEURONS

dicator of locomotor behavior. Like rhythmic behaviors in other animals, locomotion in Aplysia is generated by a central motor program. The analysis of the neural mechanisms underlying locomotion in Aplysia has been facilitated by the large identifiable neurons in its central nervous system (CNS) and the compartmentalization of the different parts of its locomotor control system into separate ganglia. The paired pedal ganglia each contain both the oscillator circuitry and motor neurons that drive the rhythmic pedal waves in the foot (10, 19, 24, 26). The neurons re- sponsible for initiating and maintaining the locomotor program are located in the cerebral ganglion.

In the present study, we have examined the command neurons that initiate pedal waves. While not universally accepted, the term command neuron remains useful if only for its brevity. The concept of command neurons has been the subject of debate since first introduced by Wiersma and Ikeda (41) to describe the properties of certain crustacean interneurons. Operationally defined command neurons have been described for numerous motor systems, particularly in arthropods. These include crustacean locomotion (32) swimmeret beating (4, 5) and escape behavior (44, 45). Recently, Kupfer- mann and Weiss (3 1) have tried to provide a rigorous definition of command neurons to replace the widely varying operational definitions previously used. They proposed that for a neuron to be considered a command neuron, it must be active during a particular behavior. Its firing must be sufficient to evoke that behavior. Additionally, the firing of the neuron must also be necessary for the behavior. Removing the neuron by hyperpolarizing it below threshold or killing it must block the behavior. While neurons in several systems have been shown to meet the sufficiency criterion for being command neurons (4,5, 17,40), only in a few where the number of command neurons is low (1 or 2) such as the tadpole Mauthner cells (35) and the crayfish lateral giant fibers (34) the necessity criterion has been met as well. The primary reason for this is that if the system is redundant and contains several equally sufficient command neurons (a command system), testing the necessity criterion becomes tech- nically difficult if not impossible.

As will be demonstrated below, the cerebral ganglion of Aplysia contains a group of identifiable neurons that meet the criteria for being command neurons. These neurons are excited by stimuli that evoke locomotion in vivo and are active during pedal wave motor program generation. Intracellular stimulation of individual neurons at physiological frequencies initiates and maintains pedal wave bursting. When these neurons are removed from the neuronal circuits by hyperpolarization, the same stimuli fail to initiate the locomotor program. In addition to investigating the properties of the command neurons, the organization of the command system was also examined. A preliminary report of some of these results has already been published (12).

MATERIALS AND METHODS

Twenty-three adult Aplysia cal$xmica were used in this study. The isolated CNS tentacle preparation that we have described previously (11, 20, 2 1) was used. Animals were opened with a dorsal incision and the entire CNS was dissected free. Both the anterior and posterior tentacles as well as the eyes were left attached to the CNS by their respective cerebral nerves. Each anterior tentacle was placed in its own sealed compartment of the experimental chamber to allow physiological stimulation. The CNS was pinned to the floor of the central portion of the chamber. Small incisions were made over each cerebropedal (C-P) connective. Both C-P connectives were freed from the connective tissue sheath that surrounded them in order to facilitate extracellular recordings. The connective tissue sheath over the neurons in the cerebral ganglion was then surgically removed. In some preparations the left pedal ganglion was also desheathed to permit intracellular recordings. All preparations were maintained in filtered seawater kept between 14- 15OC by a cold plate under the experimental chamber.

Neural correlates of the pedal wave motor program (PWMP) were monitored using glass-tipped suction electrodes placed on the cut ends of fine branches of several pedal nerves. Bursting previously shown to be a correlate of the PWMP ( 10, 23, 26) was most easily recognized in the parapedal commissure nerve (PPCN) and the third branch of the posterior pedal nerve (P,J. An additional suction electrode was placed on the distal end of P9 for electrical stimulation. The latter permitted triggering of the motor program in quiescent preparations in order to be sure the entire system was functional after desheathing the ganglia. During the actual experiments the PWMP

S. M. FREDMAN AND B. JAHAN-PARWAR

was also initiated by physiological stimulation of the tentacles. En passant suction electrodes were placed on each C-P connective. Two nerves or connectives were monitored simultaneously. These were selected independently using a switch box. Intracellular recordings from up to four neurons simultaneously were made using beveled 2.5 M potassium citrate-filled micropipettes. Neurons were intracellularly stimulated either by passing depolarizing DC currents or by trains of 20-ms pulses, with the stimulus current adjusted to give 1 spike/pulse. The latter permitted precise control of the stimulus frequency. All pedal ganglion recordings were from right and left lower quadrant neurons (sectors IIIb and 111~ of Hening et al., Ref. 19) previously shown to be motor neurons (10, 19). The presence of an axon in Pg was used to identify these neurons. Electrical activity from the nerves and neurons was amplified by conven- tional means and displayed on an oscilloscope and a six-channel chart recorder. Recordings from the latter and data stored on magnetic tape were used for subsequent analysis.

RESULTS

Criteria for pedal wave motor program command neurons

For neurons to be classified as command neurons, they had to meet several criteria. While these were intended to provide an objective basis for judging potential command neurons for Aplysia locomotion, several have applicability to other systems as well. The description of the criteria used follows.

1) Neurons had to conform to known an- atomical constraints. For Aplysia, neurons had to have an axon in one or both cerebropedal (C-P) connectives, as evidenced by an extracellularly recorded axonal spike in the connective following the driven somatic spike 1: 1 with a constant latency. Intact C-P connectives had been shown to be necessary for locomotion (24), suggesting that they were the “command pathways.” This criterion provided an objective means of identifying putative command neurons. It also eliminated from consideration neurons such as those in the cerebral A and B clusters that synapsed with pedal ganglia neurons (33) but did so almost exclusively via the cerebro- pleural connectives.

2) Intracellular stimulation of the putative command neuron had to initiate the pedal wave motor program (PWMP). Ideally, one would like to use the actual behavior. This

was not attempted both for technical reasons (i.e., mechanical stability while recording from small neurons) and that it would not have been conclusive. In behaving preparations it is possible to trigger pedal waves via peripheral reflexes (22). The possibility of coincidence cannot be eliminated. Since the PWMP could be readily recognized in the isolated CNS, the latter was used both for simplicity and to avoid the pitfalls that the intact animal preparation presented. The PWMP was defined as either clear-cut bursting by units in the parapedal commissure nerve (PPCN) or the bursting of intracellularly recorded pedal ganglion motor neurons. Earlier work with semi-intact preparations (10, 22, 26) had shown that these bursts occurred 1: 1 with pedal waves regardless of how the waves were initiated. When pedal waves were absent, so were the bursts. Moreover, pedal waves occurred only during locomotion. Thus the bursting activity was simply a readily recognized part of the total pedal wave motor program. This criterion de- manded that unpatterned (tonic) stimulation had to trigger patterned motor output. This eliminated the possibility that any changes in pedal nerve or neuron activity were merely correlates of a simple reflex. Intracellular recordings from pedal motor neurons also served to distinguish cerebral neurons syn- apsing directly (monosynaptically) with the motor neurons, as compared to those driving the PWMP oscillator circuitry.

3) Putative command neurons had to be excited by stimuli that evoked locomotion in vivo and the PWMP in vitro. For Aplysia locomotion these included food sensory (seaweed extract) and nociceptive (salt) stimulation of the anterior tentacles. Electrical stimulation of the distal end of the posterior pedal nerve (P,) was used to approximate nociceptive (salt or electrical) stimulation of the tail (24, 26, and unpublished observations). These stimuli also help determine the range of firing frequencies for command neurons. This in turn put limits on the stimulus frequencies that could be used to drive the neurons in testing the second criterion.

4) Putative command neurons had to be active during spontaneous PWMP generation. This provided supportive (but not conclusive) evidence that a putative command neuron was necessary for PWMP generation.

PEDAL WAVE COMMAND NEURONS 1095

5) Changes in the firing frequency of a command neuron had to be correlated with appropriate changes in the PWMP. Increased command neuron firing frequencies had to result in decreases in the PWMP period (i.e., increased PWMP bursting). A decreased firing frequency had to be associated with increases in the PWMP period (decreased PWMP bursting). This criterion was important for distinguishing whether the command neuron firing was required for the maintenance of the PWMP, as opposed to only triggering the PWMP.

6) The putative command neuron had to be necessary for the initiation of the PWMP. This was tested by hyperpolarizing the neuron below threshold while trying to trigger the PWMP by sensory stimulation. The failure to elicit the sensory evoked PWMP in the absence of the command neuron’s firing would show that its activity was necessary for PWMP initiation. This helped distinguish neurons whose firing, while sufficient to initiate the PWMP, was not actually necessary as well as those whose firing served a modulatory role in the maintenance of the PWMP but were not required for its initiation.

As will be shown below, identifiable cerebral ganglion neurons met all the above criteria and thus were considered to be true command neurons. These were termed type I neurons. Since there were a minimum of four such neurons (see below), together they constituted a command system. In all, 18 type I neurons were recorded for extended periods (up to 5 h). Other neurons were also found that fulfilled some but not all of the command neuron criteria. They have been termed type II and type III neurons. How their physiological properties differed from those of the type I neurons will also be considered.

Localization of neurons

As noted above, previous work (24) had indicated that information from the cerebral ganglion necessary for initiating pedal locomotion in Aplysia reached the pattern-generating circuitry in the pedal ganglia via the C-P connectives. Neurons of potential interest in the cerebral ganglion were visualized by backfilling the C-P connectives with dye (not shown). The results of unilateral con-

nective fills revealed the somata of stained neurons, which were located primarily in the ipsilateral C, D, and E clusters (2 1). There were also stained neurons in the contralateral C cluster. This indicated that some C-neurons either had bilateral projections into both C-P connectives or decussated with projections only in the contralateral connective. Further localization of putative command neurons was obtained in electrophysiological studies. E cluster neurons were soon eliminated from further consideration since 1) they appeared to be primarily involved in feeding (28) and 2) with the exception of having an axon in the ipsilateral C-P connective, none met any of the criteria for being command neurons. Stable recordings were obtained from 114 neurons that had detect- able axons in the C-P connectives. Of these, 103 were in the C clusters, with the remain- der (n = 11) in the D clusters. Contrary to expectations, axons were detected in only one connective or the other. None of the neurons sampled had axons in both C-P connectives. Of the recorded C cluster neurons, 83.5% had an axon in the contralateral connective (Table 1). None of the left D- (LD) neurons and only 50% of the right D- (RD) neurons examined exhibited this apparent decussation. None of the D cluster neurons met the other criteria for being command neurons. On the other hand, C cluster neurons that met the criteria were reliably found. As a result, attention was focused on the C- neurons. Foremost among these were the type I neurons, which constituted 17% of the total C-neurons examined in detail.

TABLE 1. Decussation of neurons with axons in cerebropedal connectives

Axon in R+L Cluster n RC-P LCP C-P

RC 59 9 (15.3) 50 (84.7) 0 LC 44 36 (81.8) 8 (18.2) 0 RD 8 4 (50) 4 WV 0 LD 3 0 0 3 (100) 0

Total 114 Total

C-neurons 103 (83.5)

Values are from electrophysiological recordings. Numbers in parentheses are percentages.

1096 S. M. FREDMAN AND B. JAHAN-PARWAR

Initiation and maintenance of PWMP by type I neurons

The neurons classified as type I neurons met all the above criteria for being command neurons. There were a minimum of four type I neurons, two in each cerebral C cluster. This is an estimate based on simultaneously recording bilateral and ipsilateral pairs as well as recording different type I neurons (as indicated by location within the cluster) in sequence. We assume here that the cerebral C clusters are symmetrically organized. All had, as far as it was possible to test, the same physiological properties (see below). This suggested that the type I neurons formed a command system made from redundant elements. All the type I neurons had axons in the contralateral C-P connectives, with the amplitude of the extracellular action potential typically ranging from 10 to 20 pV. This was confirmed by intracellular dye injection (not shown). Dye injection also revealed a

very fine ipsilateral axon, although this axon was never detected electrophysiologically. Type I neurons thus appear to project to both pedal ganglia.

Intracellular stimulation of type I neurons was highly effective in initiating the PWMP in quiescent preparations. This is shown in Fig. 1. In this and five other experiments, 5- Hz stimulation of type I neurons initiated bursting in both the PPCN and a posterior pedal nerve branch, P9,. Figure 1 also shows the largely inhibitory spontaneous synaptic input that was characteristic of type I neurons. In addition to initiating the PWMP, type I neuron stimulation also caused increased firing by other neurons in the cerebral C clusters (Fig. 1). The latency of this excitation was about 10 s, approximately the same as the increased unit activity leading into the first PPCN burst of the PWMP. The excitation of the C-neuron was polysynaptically mediated. As will be discussed in more detail below, no monosynaptic connections

PPCN

L Ppc

FIG. 1. Initiation of the pedal wave motor program by type I neuron stimulation. Intracellular stimulation of a cerebral ganglion type I neuron (RCi) at 5 Hz triggered pedal wave bursting in the parapedal commissure nerve (PPCN) and a branch of the left posterior pedal nerve (LPs,). The stimulation also evoked increased firing in another unidentified cerebral neuron (RC). The latency of the onset of increased firing in the RC neuron was approximately the same as the increased unit activity preceding the first PPCN burst. The PWMP stopped shortly after the command neuron stimulation ceased. Note the increased IPSP frequency in the type I neuron following the stimulation.


were found between type I neurons and other neurons (cerebral and pedal). These results suggest that stimulating type I neurons not only activates the PWMP generating circuitry but also other pathways that were presumably secondarily related to it.

The effect of intracellular stimulation of type I neurons on the PWMP was highly reproducible from preparation to preparation. This is shown in Fig. 2. The PWMP began with an accelerating burst by the pedal motor neurons that was terminated by inhibitory synaptic input (10). The termination of the motor neuron burst provided an objective point for measuring the PWMP. The program period was defined as the time from the termination of one motor neuron burst to the termination of the next burst. As can be seen in Fig. 2, the effect of stimulating three different type I neurons in three different preparations was virtually the same. There was a roughly fivefold reduction in the program period during the stimulation. The motor program was maintained until the

160

stimulation was terminated, following which the program period again increased. In some preparations the PWMP continued for up to five cycles before either returning to prestimulus durations or stopping altogether. More typically, the PWMP ceased within two to three cycles. None of the preparations exhibited any long-term PWMP bursting in the absence of type I neuron firing. It is important to note that different type I neurons in the same preparation were equally effective (see Figs. 7 and 12) and evoked the same motor program both qualitatively and quantitatively. In addition, simultaneously depolarizing two type I neurons was more effective than driving either one alone. This summation has not yet been quantitatively examined in detail however.

In three preparations, the relationship between the type I neuron firing frequency and the PWMP was examined. The results of a typical experiment are shown in Fig. 3. Driv- ing a type I neuron at 2 Hz was nearly as effective in initiating the PWMP as higher

Neuron 1 P

Neuron 2 A

I

fl

Neuron 3

, I I I I 1 I I I 1 1 2 3 4 5 6 7 8 9

Program Cycle Number

FIG. 2. The effect of type I neuron stimulation on the pedal wave motor program was highly reproducible. The effect of stimulating three different type I neurons on the program period of the PWMP in three different preparations is shown. The stimulation frequency was 5 Hz in each case. The onset of the stimulation (arrow) caused a sharp reduction in t-he program period. The PWMP ran with reduced program periods until stimulation was terminated, typically after six or seven cycles. In all three preparations, type I neuron stimulation produced nearly the same effect.


(5 and 8 Hz) frequencies. The minimum program period was approximately the same over a fourfold range of type I firing frequencies. While there was some preparation- to-preparation variation, the threshold for initiating the PWMP by stimulating a single type I neuron appeared to be about 1.5 Hz. Stimulating the same neuron (as in Fig. 3) at 1 Hz for nearly 3 min failed to initiate the PWMP (Fig. 4). The PWMP was turned on after about 20 s when the type I neuron frequency was increased to 2 Hz. Once initiated, 1 -Hz stimulation was capable of maintaining the PWMP. Further increases in stimulus frequency transiently increased the burst frequency (Fig. 4).

While there was little difference in the minimum evoked program period for type I neuron firing frequencies between 2 and 8 Hz, there was a definite effect on the latency with which the PWMP was initiated. Figure 5 shows that with increasing type I firing frequency there was a linear decrease in the latency to the start of the first PWMP burst.

300 t

The degree to which the PWMP was active had a significant influence on the latency to burst. As seen in Fig. 5, the latency at a given frequency was longer when the PWMP was quiescent than when the PWMP was spontaneously active. Spontaneous activity was defined as any bursting occurring 1 min or less prior to stimulation. If extended, both lines would reach zero latency at about 8 Hz. In reality the line tailed off at 8 Hz. This is probably the result of there being a minimum time in which the PWMP can be initiated. Driving the command neurons at progres- sively higher frequencies (above 8 Hz) would presumably be unable to reduce that minimum processing time.

Repetitive stimulation of the command neurons affected the timing of the PWMP. This can be seen in Fig. 3 where the PWMP period gradually increased during ~-HZ stimulation. The ~-HZ stimulation was the eighth stimulus presented. Prior stimulations lasted from 3 to 15 min. This may account for some of the reduced effectiveness of the ~-HZ stim-

Type I Neuron Frequency - 5Hz

b--- * 2Hz

C- --I 8Hz

1 2 3 4 5 6 7 a 9 Program Cycle Number

FIG. 3. The effect of command neuron firing frequency on the-pedal wave motor program. A single type I neuron was stimulated (bar) at three different frequencies in the same preparation. The ~-HZ stimulation (triangles) was equally as effective in initiatingthe PWMP as was 5- (circles) and ~-HZ (squares) stimulation. With a spontaneous period of 300 s, the PWMP prior to the ~-HZ stimulation was essentially not running. Note that the ~-HZ stimulation was less effective than lower frequency stimulation in maintaining the PWMP at short program periods, as evidenced by the upward shift of the curve. See text for details.


FIG. 4. Maintenance of the pedal wave motor program by type I neuron stimulation. Prior to stimulating the type I neuron (RC,) two pedal motor neurons (LP MNI, LP,& fired tonically. There was only irregular unit activity in the PPCN. Driving the type I neuron at 1 Hz did not initiate the PWMP. When the frequency was increased to 2 Hz, the PWMP was triggered, as evidenced by bursting in the motor neurons and the PPCN. Once initiated, l-Hz stimulation was able to maintain the PWMP. Further increases in driving frequency to 2 and 3 Hz increased PWMP activity as evidenced by accelerated motor neuron bursts. PWMP activity ceased several cycles after the type I neuron stimulation was terminated, with the neurons returning to tonic firing. Note the absence of any indication of monosynaptic connections between the type I neuron and the motor neurons. The PWMP ran for almost 15 min.

ulation. Although not systematically inves- The effect of repetitive type I neuron stim- tigated, it appeared that the increases in pro- ulation on the PWMP was examined in order gram period during a stimulation were ac- to determine if the cumulative increases in centuated by high-frequency stimulation. program period observed were due to dete- While ~-HZ stimulation had the shortest la- rioration of the preparation or synaptic dec- tency for initiating PWMP bursting, it was rement. In these experiments, the stimulus also less effective in maintaining the PWMP frequency (5 Hz), duration (3 min), and in- than lower frequency stimulation. terval between stimulations ( 10 min) were

25-

O-----4 Progrom Quiet

Active

.

1 2 3 4 5 6 7 8 9 Type1 Neuron Firing Frequency Hz

FIG. 5. The effect of command neuron frequency on the latency of pedal wave motor program initiation. Increasing the frequency of type I neuron stimulation in a quiescent preparation (circles) caused a linear decrease in the latency from the onset of the stimulation to the start of the first PWMP burst. Type I neuron stimulation when there was spontaneous PWMP activity (squares; typically from prior command neuron stimulations) caused a shift to shqrter latencies. If extrapolated, both lines would reach zero latency at about 8 Hz. In reality there was a tailing off at 8 Hz. Only data from a single command neuron was used here in order to eliminate preparation- to-preparation variation.


held constant. As seen in Fig. 6, repetitive had been stimulated 7 times over a period stimulations caused a progressive shift to- of 3 h. That the two curves are virtually the ward longer program periods. This was more same indicates that there was no significant pronounced in later program cycles than for overall deterioration in the oscillator-motor the initial reduced period cycle. Thus by the neuron circuitry. This argues that the changes fourth program cycle of the fifth stimulation, in the PWMP seen in Fig. 6 were the result the period was nearly twice that of the fourth of decrement, presumably at the type I neu- program cycle of the first stimulation (Fig. ron-oscillator synapse(s). The results in Fig. 6). These results are in good agreement with 3 suggest that this decrement occurs more those in Fig. 3 and indicate that the shifts in 2 rapidly at higher type I neuron firing fre- the curves seen in Fig. 3 cannot be accounted quencies. for solely by the differences in stimulus frequency. Experiments where two type I neu- Maintenance of PWMP by type I neurons

rons were recorded stimultaneously served to distinguish fatigue (or deterioration) from a decremental process. Figure 7 shows that stimulating two different type I neurons in the same preparation produced nearly identical changes in PWMP period. The data were taken from the first ~-HZ (3 min duration) stimulation of each neuron. The type I neuron labeled B in Fig. 7 was stimulated for the first time after the neuron labeled A

FIG. 6. The effect of repetitive single t .ype I neuron at 5 Hz for

160’

$ 120- .- L 2

20-

The above results show that intracellular stimulation of type I neurons was sufficient to initiate the PWMP. Type I neurons could also maintain the PWMP. As noted above, the threshold frequency in most preparations for initiating the PWMP was about 1.5 Hz. As seen in Fig. 4, l-Hz stimulation could maintain the PWMP once initiated. Driving a type I neuron at 1 Hz was adequate to maintain the PWMP for 5 min. This did not

Stimulus Presentation

- 1 st

&-- +3rd

w.-. -6th

1 2 3 4 5 6 7 8


command neuron stimulation on the 3 min (bar) at lo-min intervals caused a

periods. This was particularly evident for the appeared to be due to synaptic decrement in the

fourth comm

and fifth program cycles of the fifth stimulation. The shift and neuron-program generating pathway. See text for details.

pedal wave motor program. Stimulating a progressive shift toward longer PWMP


. r 120 Type I Neuron -A

1 2 3 4 5 6 7 a 9 Program Cycle Number

FIG. 7. The effect of stimulating two type I neurons on the pedal wave motor program. Two type I neurons (A, B) in the same preparation were stimulated at 5 Hz for 3 min. They produced nearly identical changes in the period of the PWMP. The first stimulations of both neurons are shown. Command neuron A was stimulated early in the experiment. Command neuron B was penetrated and stimulated about 3 h later. During that interval, command neuron A had been stimulated 7 times. That the two stimulations produced similar changes in the PWMP program period shows that neurons were equally effective and that there was little if any change in the efficacy of PWMP-generating circuitry due to repeated command neuron stimulation.

appear to be “running on” by the program. The PWMP ended after three cycles following termination of ~-HZ stimulation (Fig. 4). The ability for l-Hz firing to maintain the PWMP was confirmed in two other experiments. Further increases in type I neuron frequency resulted in transiently shorter motor neuron bursts and reduced program periods. Note that after the termination of the type I neuron stimulation, the motor neurons reverted to the low-frequency tonic firing they displayed prior to the onset of stimulation. These results show that low-frequency (l-2 Hz) driving of type I neurons could maintain the PWMP for many minutes. This is in reasonable agreement with type I neuron activity seen in preparations where the PWMP was spontaneously active. While irregular, in these preparations type I neurons typically fired at an average frequency of 0.6 Hz.

Type I neurons appear necessary for initiation and maintenance of PWMP

Firing by type I neurons appears to be sufficient to initiate and maintain the PWMP. Of all the criteria for command neurons, that

of necessity is the most difficult to satisfy. That type I neurons were necessary for PWMP generation is suggested by the observation that there were no spontaneous PWMP bursts in the absence of spontaneous type I neuron firing and that when PWMP bursts were present, at least one type I neuron was spontaneously active. Such coactivity suggests but does not prove necessity. Further evidence that type I neurons were necessary was obtained in experiments where the PWMP was initiated by electrically stimulating nerve Pg. The resultant PWMP was compared with that triggered when a single (and in one experiment two) type I neuron was held hyperpolarized below threshold. With the type I neuron hyperpolarized, the PWMP was reduced, producing fewer cycles at longer program periods. It still ran, however, and the differences seen were not dra- matic. There appear to be two reasons for this. Since there are at least four (and possibly more) type I neurons, hyperpolarizing one or two still leaves two (or more) type I neurons functioning. As is clear from the results shown above, the firing of a single type I neuron is adequate to evoke the PWMP. Second,


electrical stimulation of the pedal nerves not palement. A stable recording was made from only activates sensory pathways but anti- a single type I neuron. With the exception dromically stimulates motor neurons as well. that it exhibited irregular spontaneous firing There is evidence ( 10, 26, unpublished ob- (< 1 Hz averaged frequency), the neuron ap- servations) that there is feedback from the peared otherwise normal. The PWMP was motor neurons to the oscillator circuitry, also spotaneously active. Hyperpolarizing the which is independent of the command neu- type I neuron below threshold caused an rons. This would further minimize the effec- abrupt increase in program period (Fig. 8). tiveness of eliminating a single or pair of The program period became so long that the command neurons. That hyperpolarizing a PWMP was effectively shut off. When the single type I neuron had any effect suggests neuron was released from hyperpolarization, that if one could hyperpolarize all the type the resultant rebound firing was sufficient to I neurons, the PWMP might indeed be restart the PWMP. The program period re- blocked. This is just within the realm of fea- turned to prehyperpolarization levels after sibility if there are only four type I neurons. two cycles. Thus in this preparation at least, If there are more, then such a test becomes the firing of a type I neuron was necessary all but impossible. to maintain the PWMP. Given the redun-

In one experiment, it was fortuitously pos- dancy of the system, it would seem that the sible to provide additional evidence that type other type I neurons may have been damaged I neurons were indeed necessary for both ini- either during desheathing the ganglion or tiating and maintaining the PWMP (Figs. 8 attempted penetration, with the result that and 9). In this preparation many of the ce- the recorded neuron was the only one still rebral neurons had either suffered damage functioning. An alternative explanation for during desheathing or were injured on im- these results is that while all the type I neu-

Type I Neuron Hyperpolarized

1 I I 1 I I I I 1 1 2 3 4 5 6 7 8 9


FIG. 8. Type I neuron firing was necessary for maintaining the spontaneous pedal wave motor program. In this preparation, the recorded type’1 neuron was firing spontaneously. The PWMP was also spontaneously active. Hyperpolarizing the command neuron (bar) effectively shut off the PWMP. The PWMP resumed when the type I neuron was released from hyperpolarization. This indicated that type I neuron firing was necessary for maintaining the PWMP.


FIG. 9. Type I neuron firing was necessary for sensory-evoked initiation of the pedal wave motor program. A type I neuron (LC,), the parapedal commissure nerve (PPCN), and posterior pedal nerve (LPgc) are shown. Stim- ulating one anterior tentacle with salt (first arrow) caused accelerated type I neuron firing and increased PWMP activity, as evidenced by the bursts in the PPCN and LP sc. Hyperpolarizing the type I neuron below threshold (downward arrowhead) caused the PWMP to slow down. Stimulating the same tentacle (second arrow) and the other tentacle (third arrow) caused minimal command neuron firing due to the hyperpolarizing current. The stimulations failed to retrigger the PWMP. Bursting in the PPCN and LPg, continued to decline. When the type I neuron was released from hyperpolarization (upward arrowhead), PWMP bursting in both nerves resumed. This indicated that the firing of the type I neuron was necessary for sensory initiation of the PWMP. Note: both tentacles were tested prior to and immediately after obtaining the above records and were found equally effective in triggering the PWMP. Stimulating the second tentacle served as a control for any sensory adaptation and habituation that stimulation of the first tentacle at short intervals might have produced.

rons are sufficient, some, and the one shown in Fig. 9 in particular, are more important than others. The results obtained in other experiments do not support this, however.

In the same preparation it was also possible to show that the firing of this type I neuron was necessary for sensory initiation of the PWMP. Figure 9 demonstrates that presenting a noxious stimulus (salt) to one of the anterior tentacles triggered increased type I neuron firing and the PWMP. Stim- ulating the other tentacle was equally effective. Seaweed extract stimulation (not shown) also caused increased firing, but was less effective (in this preparation) than was the salt. Similarly, in every preparation, tactile stimulation of the tentacles and electrical stimulation of the posterior pedal nerve either initiated or increased type I neuron spiking. Following the initial salt stimulation, the type I neuron was then hyperpolarized below threshold (Fig. 9). As in Fig. 8, this shut off the PWMP. Presenting the salt stimulus first to one tentacle and then the other failed to turn on the PWMP. Very strong current was required to prevent the neuron from firing during stimulus presentation. This resulted in the noisy base line seen in Fig. 9. The stimulus paradigm eliminated the possibility that failure to evoke the PWMP might have

been due to sensory adaptation or habituation along the sensory pathway. When the type I neuron was released from hyperpolarization it fired strongly with a peak frequency of 8 Hz for 3 s. The firing frequency then declined to 3 Hz after 10 s. The PWMP bursting resumed. Stimulating the tentacles with salt with and without the type I neuron being hyperpolarized was repeated 3 times. Each time when the type I neuron was permitted to fire, the stimulus triggered the PWMP. When the neuron was strongly hyperpolarized, it did not. While not ideal, these results indicate that collectively the type I neurons probably are necessary for PWMP generation as well as being individually sufficient.

Type II neurons In addition to the type I command neu-

rons, the cerebral C clusters contain at least two type II neurons that met several but not all the command neuron criteria. Type II neurons were encountered less frequently than type I neurons, being recorded in only 57% of the preparations (compared to 70% for type I neurons). They had axons in the contralateral C-P connective and were characterized by primarily excitatory spontaneous synaptic input. Type II neurons were excited by stimuli that evoked the PWMP.

1104 S. M. PREDMAN AND B. JAHAN-PARWAR

PPCN

FIG. 10. The initiation of the pedal wave motor program by type I and type II neurons. A: intracellular stimulation of a type I neuron (RCr) at 5 Hz triggered bursting in pedal motor neurons (LPhlNlr LPN& and the PPCN. B: stimulating a type II neuron (LCrr) earlier in the same experiment also evoked the PWMP. The latency for the onset of motor neuron bursting was longer, the motor neuron bursts were weaker (lower firing frequency), and fewer cycles were triggered as compared to the type I neuron shown in A. While sufficient to initiate the PWMP, type II neurons were less effective in doing so than type I neurons.


Like type I neurons, intracellular stimulation of type II neurons was sufficient to turn on pedal wave bursting in quiescent preparations (Fig. 10). Type II neurons were much less effective than type I neurons, however. Type II neurons had to be driven at higher frequencies than type I neurons to initiate the PWMP. The threshold frequency for type II neurons was about 3 Hz, as compared to 1.5 Hz for type I neurons. As can be seen in Fig. 10, the latency to the first burst was also much longer for type II neurons. The latency for a type II neuron driven at 5 Hz was approximately 18 s, as compared to about 10 s for a type I neuron driven at 5 Hz in the same preparation. As can be seen in Fig. 11, the overall PWMP output was also less. Not only were the program periods longer, but the PWMP ran for fewer cycles as well, even though the duration of the stimulus was approximately the same. Although not inves- tigated in detail, it appeared that the abilitv

of type II neurons to drive the PWMP declined much more rapidly with repetition than did type I neurons. The difference in efficacy between type I and type II neurons shown in Fig. 10 was not due to deterioration of the preparation. The type II neuron stimulation occurred several hours before that of the type I neuron. Nor was this difference due to neuron-to-neuron variation, since Fig. 7 showed that driving two type I neurons in the same preparation had virtually the same effect on the PWMP. In fact, comparing Fig. 11 and Fig. 2, it is clear that the preparation- to-preparation variation in PWMP activity evoked by driving type I neurons was less than the variation between driving type I and type II neurons in the same preparation. Thus, type II neurons were sufficient to initiate pedal wave bursts but were physiologically distinct from type I command neurons. Unlike type I neurons, type II neurons did not appear to be necessary for the initiation

rTYPell / Neuron

1 2 3 4 5 6 7 8 9

Progrom Cycle Number

FIG. 11. The effect of type I neuron versus type II neuron stimulation on the pedal wave motor program. Both the type I neuron and the type II neuron were stimulated (arrow) at 5 Hz for about the same duration (approximately 4 min). The first stimulation of both neurons is shown. While both neurons initiated PWMP bursting, the type II neuron evoked- fewer program bursts. The PWMP ran with much longer periods. The difference in effectiveness between type I and type II neurons in a given preparation was greater than the variation between type I neurons in different preparations.


and maintenance of the PWMP in that the PWMP (both spontaneous and evoked) occurred in the absence of their firing.

Type I and type II neurons differed in other respects in addition to their efficacy in driving the PWMP. Type II neurons had different (compared to type I neurons) synaptic input and synaptic interactions both with themselves and with type I neurons. This will be discussed in more detail below. Some results suggest that type II neurons interact differently with the PWMP generating circuitry than type I neurons. Brief ( 10 s) stimulation of type I neurons usually evoked firing in both the pedal nerves and PPCN. Brief stimulation of type II neurons rarely triggered PPCN spikes but increased the firing frequency of several pedal nerve units at short latency. These units differed in amplitude from those evoked by type I neuron stimulation (not shown). PPCN spikes frequently occurred after type II neuron stimulation ended. Since neither type I nor type II neurons appeared to synapse directly on pedal motor neurons (Fig. 9) these results raise the possibility that type I and type II neurons make different connections with the interneurons that drive the pedal motor neurons.

Synaptic interactions command neurons

among

Type I command neurons were characterized by common synaptic input and firing patterns. This was true for both ipsilaterally and bilaterally recorded pairs. Figure 12 shows two simultaneously recorded ipsilat- era1 type I neurons. They had nearly identical firing patterns and synaptic input. Such similar firing might be due to either electrotonic coupling or reciprocal excitatory synapses. No electrotonic coupling between type I neurons was observed. When one type I neuron was depolarized, both ipsilateral and contralateral type I neurons were hyperpolarized due to increased inhibitory synaptic input (Figs. 12 and 13). Hyperpolarizing a type I neuron had no effect on the membrane potential of other simultaneously recorded type I neurons. Type I neurons thus had neither electrotonic nor chemical monosynaptic connections among themselves. Their very similar firing patterns were due to common

synaptic input, such as that evoked by sensory stimuli.

Type I neurons exhibited two significant sources of synaptic input in addition to that mediated by sensory pathways. The first of these was recurrent inhibition. As seen in Figs. 12 and 13, stimulating one command neuron caused short-latency (< 1 s) inhibitory input in the other command n eurons. Al- though it is largely obscured by spiking, intracellular stimulation also caused increased inhibitory input in the stimulated neuron (Fig. 13). This was only reasonable because the inhibitory input was common to all the type I neurons. Since the inhibitory postsynaptic potentials (IPSPs) were not 1: 1 with the type I neuron spikes, the inhibition was not monosynaptic. The inhibitory input, while recurrent, was of the Renshaw type being mediated by an interneuron rather than col- laterals from the type I neurons themselves. A second source of input to the command neurons was weak excitatory feedback. This can be seen in the initial part of Fig. 12A. The command neurons exhibited increased firing during each pedal wave burst. Since all experiments were done in an isolated nervous system, the excitation was not due to sensory feedback from the foot. Nor were there any connections from the recorded motor neurons to the type I neurons (Figs. 4, 10, 12). Although connections with other motor neurons cannot be excluded, these results suggest that the command neurons received excitatory feedback from the PWMP oscillator circuitry. The lack of mutual excitatory connections among type I neurons and the short-latency inhibitory feedback was significant for another reason. Since driving one type I neuron inhibited the other type I neurons, the effect of such driving on the PWMP was due to the stimulated type I neuron alone. As is clear in Fig. 12, driving a single type I neuron for 3 min did not turn on the entire comma nd neuron population. Each type I neuron was indivi .dually sufficient, making the command system truly redundant.

Type II neurons were characterized by common excitatory synaptic input. As with the type I command neurons, type II neurons were not interconnected by either electrical or monosynaptic connections. However,


FIG. 12. Interactions between command neurons. A: two command neurons (RC,) fired in near synchrony during spontaneous PWMP generation. Note the increased firing during each pedal motor neuron (LP,,) burst. Intracellular stimulation of one command neuron (top trace) increased the motor neuron bursting but hyperpolarized the second command neuron (middle trace) due to increased IPSPs. B: a continuation of records in A. Stimulating the second command neuron also increased PWMP bursting and inhibited the first command neuron. This shows that the two neurons were not electrically coupled and that the inhibitory connections were reciprocal. The common inhibitory input to both neurons can be seen in the latter part of the record.

there were polysynaptic interactions between this was due to increased IPSP frequencies. type I and type II neurons. Stimulating type The results indicate that while the effect of I neurons excited type II neurons (Fig. 13). type II neuron stimulation on the PWMP On the other hand, stimulating type II neu- cannot be accounted for by weak synaptic rons caused inhibition in type I neurons (not activation of type I neurons, some of the ef- shown). As with type I-type I interactions, ficacy of type I neurons may have been due


&A RCIB

&B

Rf,A

A,B 10.2

FIG. 13. Reciprocal recurrent inhibition among command neurons. A, B: two ipsilateral type I neurons (RCIA, RCia) are shown. A 10-s depolarizing pulse injected into one command neuron caused it to fire and hyperpolarized the other with a latency of <l s, due to increased IPSPs. The firing of the depolarized type I neuron was slowed due to increased IPSPs in it as well. This input was particularly evident in A. This indicates that the inhibition was recurrent as well as reciprocal. The connection was not monosynaptic, since the IPSPs did not follow the presynaptic spike 1: 1. There was common inhibitory synaptic input to both command neurons. C, D: a bilateral pair of type I neurons (RC,, LC,) are shown. Here too, depolarizing one hyperpolarized the other. As with the ipsilateral pair, interactions were reciprocal but not monosynaptic. E: depolarizing a type I neuron (LC,) polysynaptically triggered spiking in a contralateral type II neuron (RC,,). Note that the type II neuron received primarily excitatory synaptic input, compared to the inhibitory input in the type I neuron.

to the increased type II neuron firing they evoked.

Type III neurons

Neurons that had axons in the C-P connectives and altered pedal motor activity but did not initiate the PWMP were considered to be type III neurons. This classification included several different types of neurons. Various type III neurons were found in 83% of the experiments. Foremost among these were neurons that made monosynaptic connections with pedal motor neurons. Eighty percent of the type III neurons indirectly “modulated” pedal motor neuron firing.

When they were stimulated, the frequency of motor neuron firing typically increased; when they were hyperpolarized, it decreased. Other type III neurons altered the timing of the PWMP (see below).

While no synaptic connections between type I and type II neurons and pedal motor neurons were ever observed, type III neurons made apparently monosynaptic connections with pedal motor neurons. Both excitatory (Fig. 14) and inhibitory (Fig. 15) connections were observed. These connections appeared to be monosynaptic in that the PSP followed the driven spike in the type III neuron 1: 1 at a constant latency. Figure 14 shows a neu-


FIG. 14. Type IRA neurons made excitatory monosynaptic connections with pedal motor neurons. A: intracellular stimulation of a type IRA neuron (RCuIA) evoked 1:l constant-latency EPSPs in a pedal motor neuron (LPMNI). The motor neuron was hyperpolarized to prevent its firing. The spike in the type IIIA neuron was largely obscured by the artifact caused by the failure of the bridge to remain balanced. Note the polysynaptic EPSP that followed the monosynaptic EPSP in the motor neuron. This is an example of feedforward summation. A second motor neuron (LPUN2 in C) also received the same polysynaptic EPSP. B: EPSPs in the motor neuron followed spikes in the type IIIA neuron 1: 1. The motor neuron was hyperpolarized to prevent its firing. C: driving the type IRA neuron at 5 Hz caused strong firing in the one pedal motor neuron due to the monosynaptic EPSP and feedforward summation. A second motor neuron also fired strongly due to polysynaptic EPSPs. The stimulation failed to evoke PPCN bursting although it did trigger tonic firing by a unit in the PPCN. The large tonic unit in LP,, was inhibited, while other smaller units were excited by the type IRA neuron stimulation. There was no bursting characteristic of the PWMP.

ron, termed type IIIA, that made an excitatory monosynaptic connection with one motor neuron and a polysynaptic excitatory connection with a second motor neuron. Strong polysynaptic connections were distinguished by their variable latency and failure to follow the presynaptic neuron at higher firing frequencies. The first motor neuron in Fig. 14 received a ,polysynaptic excitatory postsynaptic potential (EPSP) from the type IIIA neuron in addition to the monosynaptic EPSP. The second motor neuron only re-

ceived the polysynaptic EPSP. When the type IIIA neuron was driven at 5 Hz, it caused strong (nearly 1: 1) firing in both motor neurons, particularly the first, due to feedforward summation. Other motor neurons with axons in the pedal nerves were inhibited. How- ever, ~-HZ stimulation failed to evoke any patterned activity resembling the PWMP.

Type IIIB neurons made inhibitory monosynaptic connections with pedal motor neurons (Fig. 15). Type IIIB neurons were unusual in that they did not decussate, having


FIG. 15. Type IIIB neurons made inhibitory monosynaptic connections with pedal motor neurons. A: intracellular stimulation of a type IIIB neuron (LCuIB) evoked constant-latency 1:l IPSPs in a pedal motor neuron (LPMVINI). The cerebral neuron also caused polysynaptic IPSPs in another pedal motor neuron (LPMNZ in C). B: IPSPs in the pedal motor neuron followed the driven spikes in the type IIIB neuron 1: 1. C: driving the type BIB neuron at 7 Hz failed to initiate the PWMP. The two intracellularly recorded pedal motor neurons were inhibited, LPMm monosynaptically and LP MNZ polysynaptically. The type IIIB neuron excited other motor neurons, as evidenced by the increased firing by units in the PPCN. The bursting characteristic of the PWMP was absent in nerves and neurons. Note: the Type BIB differed from the type IIIA neuron in that it was an LC neuron but projected to and made connections in the left pedal ganglion. Thus it did not decussate while the type IIIA neuron did.

their axons in the ipsilateral C-P connective. As with the type IIIA neuron, while one motor neuron received a monosynaptic IPSP, the IPSP in the other was polysynaptic. When the type IIIB neuron was driven at 7 Hz, both motor neurons were inhibited, as were units in Pg. Motor units in the PPCN were excited, however, indicating that type IIIB neurons make excitatory connections with some motor neurons. As with the type IIIA neurons, driving type IIIB neurons failed to initiate the PWMP. These results argue against models where command neurons initiate the PWMP by direct action on the pedal motor neurons themselves.

The type IIIC neuron differed from the others encountered in that while acting polysynaptically, it appeared to exert its effects by inhibiting the oscillator circuitry rather than the motor neurons themselves. Briefly depolarizing a type IIIC neuron altered the period of the spontaneous PWMP (Fig. 16). When this neuron fired at frequencies of OS- 1 Hz, the PWMP was inhibited. The type IIIC neuron was distinguished from the type IIIB neuron, in that it I) had an axon in the contralateral C-P connective and 2) did not excite units in the PPCN. As can be seen in Fig. 16, the inhibition produced by type IIIC neuron firing greatly outlasted the duration


PPCN

FIG. 16. Type IIIC neurons could inhibit the pedal wave motor program. The PWMP was spontaneously active, as evidenced by bursting in the PPCN. Briefly depolarizing (arrows) the type IIIC neuron (RCiiic) reset the PWMP. Not only was the next PPCN burst delayed, but the poststimulation program period was longer than the prestimulus program period. When the neuron was spontaneously active, the PWMP was completely inhibited. The resetting of the program period suggests that the type IIIC neuron acted at the PWMP oscillator.

of the stimulation. Also when the PWMP did resume, the program period was nearly twice that of the prestimulus PWMP. This inhibition of the motor program followed by an increase in program period occurred each of 4 times that this neuron was stimulated. Thus type IIIC neurons were physiologically distinct from type IIIA and type IIIB neurons, apparently acting at the oscillator rather than the motor neuron level. Since very few type III neurons have been examined in detail, how they interact among themselves and with type I and II neurons has not yet been determined.

DISCUSSION

Since first reported by Wiersma and Ikeda (4 1) for interneurons in the crustacean nerve cord, neurons considered to serve a command function have been found in several systems. Putative command neurons have been examined in crustaceans (I), leech (40), mollusks (17), and fish (6, 7, 35, 43). The commanded behaviors have ranged from reflexes and relatively simple postural changes to complex motor patterns, such as locomotion and feeding. Several problems are encountered when these studies are compared. First, command neurons have tradi- tionally been only operationally defined. At- tempts to conform to a formal definition (3 1) have been made only recently. It is therefore misleading to compare a premotor command neuron, which monosynaptically excites motor neurons (30), with “higher order” command neurons, which release an entire complex behavior ( 17,40). More significantly, the

neuronal circuitry of some behaviors, such as crayfish tailflips (44, 45), the startle response of fish (7) and swimming in the leech (37), is reasonably well understood. This has permitted the rigorous identification of small numbers of command neurons (crayfish (30, 34), fish (35), leech (40)). For other systems, such as the lobster swimmerets (4, 5) and crustacean locomotion (32), the neuronal circuitry is (or was) unknown. In these studies, presumptive command neurons were only electrically stimulated. The position the stimulated neuron occupied in the neuronal circuits was not known. Individual command neurons could not be conclusively identified from animal to animal. Another complication is that complex behaviors such as crustacean locomotion can be divided into a series of limited fixed acts, which may be individually commanded (32). The problem then becomes one of relating the command elements to each other and the intact behavior. Further difficulties arise because some neurons do not fit into neat classifications (36). Swimming in Tritonia appears to lack command neurons, possibly because while rhythmic it is largely reflexive. The C2 neurons of Tritonia can be considered to have command neuron properties but are actually part of the swim oscillator (14, 15). Some neurons are not what they were initially thought to be. The “trigger” neurons of Tri- tonia were considered to trigger swimming (42). It has since been shown (14) that they are not directly involved. Similarly, the me- tacerebral cells of Pleurobranchaea were thought to command feeding (3) but now appear to have a modulatory rather than

1112 S.M.FREDMANANDB.JAHAN-PARWAR

command function. This latter system is further complicated by reports of several different command neurons mediating the same behavior (3, 16, 17). How these neurons re- late to each other and the feeding circuitry as a whole is not yet known. Finally, complex behaviors typically have several (or even many) neurons that interact, making clear- cut role definitions difficult. Given the range of behaviors and the different levels of organization that command neurons are either thought, or in the best-studied systems known, to act, coupled with the lack of general concensus as to what constitutes a command neuron, comparisons between other command neurons and type I neurons is both difficult and possibly misleading.

Type I neurons

As demonstrated in the RESULTS, type I neurons met the sufficiency criterion for being command neurons for initiating the pedal wave motor program. Low-frequency ( 1 S-2.0 Hz) intracellular stimulation of any one of the four identified type I neurons was sufficient to initiate and maintain neural correlates of the pedal locomotor waves. Driving type I neurons at a constant frequency (i.e., tonic stimulation) thus evoked stable patterned (bursting) activity in pedal nerves and neurons (Figs. 1, 4, 10, 12). Jahan-Parwar and Fredman (22) had shown that in semi- intact preparations, stretching the foot could evoke bursting in the PPCN and pedal waves in the foot. The use of an isolated CNS in the above experiments eliminated the possibility of initiating the PWMP by proprioceptive reflex. Similarly, since by nature the PWMP was patterned activity, the possibility that the type I neurons triggered some un- specified reflex can also be eliminated.

Since there were a minimum of four type I neurons in the cerebral ganglion, together they constitute a command system. Com- mand systems can be divided into two classes. In the first, while each element may be sufficient to initiate the behavior, they function primarily as part of a network. Stimulating one command neuron not only triggers the behavior but also excites the other elements in the command system. This makes it difficult to distinguish the actions of any one element. The command neurons for swimming in the leech are an example of such a

network. Each segmental ganglion contains one command neuron (cell 204). Driving any single cell 204 turns on the swim cycle, as well as other 204’s in the other ganglia (39, 40). In the second class of command system, the command elements are redundant. Each command neuron functions independently of the others. Stimulating one element turns on the behavior but not the other command neurons. Thus in terms of sufficiency, each neuron functions as if there were only a single command neuron in the system. The type I neurons fit this second class. Type I neurons were not electrotonically coupled nor were there excitatory chemical synaptic connections among them. Due to reciprocal recurrent inhibition, stimulating one command neuron inhibited the others in the system. The effect on the PWMP was therefore due to that neuron alone. Normally, the type I neurons fired synchronously due to common synaptic input from sensory pathways and possibly the pedal wave oscillator as well.

That there were a minimum of four individually sufficient command neurons pre- cluded conclusive testing of whether they were actually necessary for generating the PWMP. The advantage to redundancy is that no single command neuron is necessary, only the command neuron population as a whole. Thus it was only under unusual circum- stances that reasonably convincing evidence for necessity could be obtained. The results seen in Figs. 8 and 9 are consistent with type I neurons being necessary. Along with experiments where pairs of type I neurons were simultaneously recorded, it suggests that if all the type I neurons could be hyperpolarized, they would collectively meet the necessity criterion.

The reported physiological properties of command neurons are as varied as the behaviors they command. It is hardly surprising that type I neurons had properties in common with some previously described command neurons but were different from others. For example, there was a linear decrease in the latency of PWMP initiation with increasing frequency (Fig. 5). This has also been described in the leech (40) and crayfish (41). Increasing the frequency of ongoing stimulation caused transient decreases in PWMP burst period. Type I neuron stimulation was highly effective over the entire


range of physiological frequencies once threshold was reached. More importantly, qualitatively the program evoked by type I neuron stimulation was always the same, regardless of frequency. This is in contrast to the lobster swimmeret system (4) where some command neurons were effective only over a limited range of frequencies and where different command neurons evoked different motor programs. For the range of physiological frequencies (up to 8 Hz), type I neurons were all equally effective and motor program specific.

Unlike the Mauthner cells of teleost fish (6) and the lateral giant fibers of crayfish (30, 44, 45), type I neurons did not synapse directly on pedal motor neurons (Figs. 4, 10, 12). This may be due to type I neurons acting on a higher level or organization than the Mauthner neurons and. lateral giants, which mediate defensive reflexes rather than rhythmic behaviors. Nor did the type I neurons exhibit mutual excitation (either chemical or electrical), as seen for the cell 204 population in the leech (40) and lateral giant fibers (34). On the contrary, there was reciprocal recurrent inhibition. This has also been observed in Mauthner cells (8, 29). While of short latency, it was not monosynaptic. Thus it appeared to be like the chem- ically mediated recurrent inhibition seen in Mauthner cells (13, 29). No evidence was found for electrotonic inhibition, also seen in Mauthner neurons.

Excitatory feedback from the pattern generator was another property that type I neurons had in common with other command neurons (17, 40). Type I neurons tended to fire in phase with PWMP bursts (Fig. 12). Synaptic connections from the recorded motor neurons to the command neurons were never observed (Figs. 4, 10, 12). Because only a small fraction of the pedal motor neuron population was recorded, the possibility that other motor neurons mediated this feedback cannot be completely discounted. Since most pedal neurons do not have axons in the C-P connectives (unpublished observations), it seems more likely that the weak excitatory feedback comes from the PWMP oscillator circuitry. In Pleurobranchaea, cerebral command neurons receive excitatory feedback from the feeding motor network via “corol- larv discharge” interneurons (17). As noted

above, cell 204, one of the command neurons for swimming in the leech, also receives excitatory feedback from the swim oscillator (40). On the other hand, type I neurons did not exhibit spike broadening, which contrib- utes to the command ability of the buccal ganglia ventral white cells of Pleurobran- chaea (16). It must be noted that spike broadening has not been reported for cerebral neurons also thought to command feeding in Pleurobranchaea ( 17). In summary, given the diversity of putative command neurons, comparisons with type I neurons must be made in a functional context.

Type I neurons appeared to be coordinated by common synaptic input (Figs. 12, 13). All the type I neurons were excited by sensory inputs, which are known to evoke locomotion in freely behaving animals. There was also excitatory feedback during the PWMP (Fig. 12). By firing in synchrony, the command system might be able to operate at lower firing frequencies than would otherwise be possible. This assumes that there is summation of type I neuron inputs onto their followers in the PWMP generating circuitry. Limiting the firing frequency of the command neurons may be significant since high-frequency stimulation in particular appeared to increase the rate of decrement that occurred with repeated stimulation (Fig. 3). At this time the nature of the decremental process is not known. It may be habituation, frequency depression, or some combination of the two. The excitatory feedback to the command neurons may serve two functions. The first is that it provides a mechanism for and probably helps sustain spontaneous locomotor activity. Second, for sensory-evoked locomotion, it may help compensate for de- clines in sensory input due to sensory adaptation and habituation in the sensory pathways. This too would help sustain long bouts of locomotion. The function of the recurrent inhibition (Figs. 1, 12, 13) is less clear. The most obvious function is to limit the command neuron fi.ring frequency. Since l-Hz firing (or even less, judging from spontaneous activity) was adequate to maintain the PWMP (Fig. 4) there probably is no need for higher frequency type I neuron firing except when initiating the motor program. The recurrent inhibition may be significant for another reason. Due to the positive feedback from the


oscillator, there would be a tendency toward Finally, type II neurons might be less effec- accelerating activity by the command neu- tive in initiating the PWMP because their rons. Recurrent inhibition would counteract synaptic connections are made with a less this by slowing the firing frequency of the crucial part of the oscillator circuitry. command neurons, as every increase in firing Type III neurons also cannot be consid- due to oscillator feedback would also be ac- ered command neurons. While intracellular companied by increased inhibition. This stimulation of type III neurons had signifi- would prevent runaway activity within the cant effects on the pedal motor neurons, they locomotion control system. Supporting this consistently failed to initiate the PWMP view is the observation that type II neurons, (Figs. 14, 15). These neurons have properties which weakly excite the PWMP oscillator that are consistent with a modulatory rather (Fig. lo), also caused polysynaptic inhibition than a command function. By being able to in the command neurons. As with the re- increase or decrease the firing of pedal motor current inhibition among the type I neurons, neurons, type III neurons could play a sig- this would counterbalance the resulting ex- nificant role in determining the expression citatory feedback from the oscillator to the of the PWMP. There appears to be a broad command neurons. spectrum of type III neurons. At one extreme

Type II and type III neurons Type II neurons, while sufficient to initiate

the PWMP, cannot be considered true command neurons. First, they did not appear to be absolutely necessary for PWMP generation. Second, their overall efficacy was con- siderably less than that of the type I neurons (Figs. 10, 11). Type II neurons had to be driven at higher frequencies than type I neurons to evoke the PWMP. When the PWMP was initiated by type II neuron stimulation, it was weaker, running at much longer program periods than for type I neurons. Re- peated stimulation of type II neurons produced a much more rapid decline in the number of PWMP cycles evoked than did repeated stimulation of type I neurons. Given this, one cannot account for the effectiveness of the type I neurons by their ability to excite type II neurons polysynaptically. While type II neurons can initiate the PWMP, that probably is not their primary function. There are several alternatives that can account for the effects of the type II neurons. One is that, like type IIIA and IIIB neurons, type II neurons may have a modulatory function, but act primarily to increase the excitability of the oscillator. Another possibility is that type II neurons may trigger some other part of the locomotor sequence. Initiating the PWMP may be secondary to this. The type II neurons might be involved in the control of one of the phases of the locomotor sequence that precedes pedal wave generation, such as head extension (24). This would be consistent with their apparently indirect effect on the PWMP.

are the type III neurons that made monosynaptic connections with the motor neurons. Others acted polysynaptically and thus were less influential. At the other extreme were neurons like the type IIIC, which appeared to act at a level several synapses removed from the motor neurons. Although type IIIA and IIIB neurons had opposite effects on some of the motor neurons, they might still function in a coordinated fashion. Types IIIA and IIIB neurons might fire out of phase with each other due to appropriate feedback from the oscillator. In this case, the firing of type III neurons would serve to sharpen and reinforce the oscillator input to the motor neurons. Another possibility is that types IIIA and IIIB neurons mediate reflexes that are only indirectly related to locomotion. These alternatives may be distinguished when more is known about the synaptic and timing relationships of the type III neurons.

Type IIIC neurons were distinctly different from the others in that they appeared to act primarily on the PWMP oscillator rather than the motor neurons. Type IIIC neurons could inhibit the PWMP and increase the program period (Fig. 16). Type IIIC neurons provided an important element in the locomotor control system, since they effectively shut off the PWMP. Such inhibition is necessary for the execution of competing behaviors, such as feeding and copulation.

Control of locomotion From the results presented above and from

the data obtained in previous studies, it is


possible to make a realistic model for the control of locomotion in Aplysia. While presented in general terms, many of the elements of this model have been examined in detail. The control of locomotion in Aplysia has features in common with other motor systems: 1) command neurons that activate and 2) a central pattern generator or oscillator. This in turn drives 3) pedal wave motor neurons. There are also 4) modulatory inputs to adjust the output of the system. Figure 17 shows how these elements are interconnected in Aplysia. Several other rhythmic motor systems (2, 18, 36) have similar organizations. As demonstrated above, the command neurons for pedal waves in Aplysia are of the tonic (or gating) type, with the PWMP running essentially only as long as the type I neurons fire. The command neurons receive several excitatory inputs. The first is polysynaptic input from the tentacles. The second- and fourth-order neurons in the me- chano- and food sensory pathways have been previously described (11, 27). The latency of the type I neurons was longer than the sec-

ond-order B-neurons and roughly the same as the fourth-order A-neurons (unpublished observations). This suggests that they are approximately fourth- or fifth-order along the food sensory pathway. The type I neurons did not receive direct synaptic input from either the A- or B-neurons. Thus, their exact position in the sensory-to-motor pathway is not yet certain. There is probably divergence of the pathway at the level of the third-order neurons ( 11). The command neurons receive two other sources of excitatory input. One is from the foot via pleural neurons. Neurons in the pleural ganglia respond to proprioceptive, nociceptive, and mechanosensory input from the foot (25). These neurons make monosynaptic connections with neurons in the cerebral ganglion (9) as well as with pedal motor neurons (38). A final source of command neuron input is the feedback from the oscillator described above. The command neurons in turn drive the pedal wave oscillator. Although very little is known about the oscillator, it is almost certainly of the inter- neuronal network type (10). The oscillator

r ________ ---------- -t --------------------_ 1

A I , ,

Cerebral Command I I

Modulatory p-------- I I

Neurons I I

( Type II, Ill 1 (Type 1) I I I I ,

I

I I I I

I

A 7 ’

I I

A Pedal Pleural Pleural Wave [2- -- nl--- 1 ---

Oscillator I I l L I I

Modulatory Modulatory

I I

I I

I I

I I

I r-- ---

r 1_

I r - -

- - -

LOCOMOTION

FIG. 17. Control of locomotion in Aplysia. This diagram summarizes interconnections among elements involved in the control of locomotion. Open triangles, excitatory synapses; filled triangles, inhibitory synapses; solid lines, presumed monosynaptic pathways; dashed lines, presumed polysynaptic pathways. For simplicity, the control of only one-half of the foot (by one pedal ganglion) has been represented. Interneurons mediating bilateral coordination have thus been omitted. See text for details.

S. M. FREDMAN AND B. JAHAN-PARWAR

network excites the pedal motor neurons in their proper phase and sequence. There is evidence that the oscillator receives feedback from the pedal motor neurons ( 10). Sensory input from the foot also excites the oscillator (22, 26) probably via pleural neurons. Since there are proprioceptive reflexes in the foot (22, 23), there is mono- and polysynaptic sensory input to the pedal motor neurons themselves. Although modulation of locomotion is typically considered in terms of sensory feedback to the oscillator and/or motor neurons (36), there is also higher order modulatory input. This is well known in vertebrates (18). This can (and in Aplysia does) act at both the oscillator and motor neuronal levels. Type II neurons can be considered as modulatory since they appear to excite the oscillator polysynaptically. Type IIIC neurons, as noted above, inhibit the oscillator. Types IIIA and IIIB neurons also appear to be modulatory, acting both mono- and polysynaptically on the pedal motor neurons. The locomotor control system of Aplysia thus exhibits many of the organizational features of other invertebrate and vertebrate locomotor systems. Given this similarity, elu- cidating the principles of organization for locomotion in Aplysia should have general significance.

REFERENCES

1. BOWERMAN, R.F. ANDLARIMER, J.L.Command neurons in crustaceans. Comp. Biochem. Physiol. A 54: l-5, 1976.

2. DAVIS, W. J. Neuronal organization and ontogeny in the lobster swimmeret system. In: Control of Pos- ture and Locomotion, edited by R. B. Stein, K. G. Pearson, R. S. Smith, and J. B. Redford. New York: Plenum, 1973, p. 437-455.

3. DAVIS, W. J. The command neuron. In: Identz$ed Neurons and Behavior of Arthropods, edited by G. Hoyle. New York: Plenum, 1977, p. 293-305.

4. DAVIS, W. J. AND KENNEDY, D. Command interneurons controlling swimmeret movements in the lobster. I. Types of effects on motoneurons. J. Neu- rophysiol. 35: I-12, 1972.

5. DAVIS, W. J. AND KENNEDY, D. Command interneurons controlling swimmeret movements in the lobster. II. Interaction of effects on motoneurons. J. Neurophysiol. 35: 13- 19, 1972.

6. EATON, R. C. AND BOMBARDIERI, R. Behavioral functions of the Mauthner neuron. In: Neurobiology of the Mauthner Cells, edited by D. S. Faber and H. Kom. New York: Raven, 1978, p. 221-244.

7. EATON, R. ~.,.BOMBARDIERI, R. A., AND MEYER, D. The Mauthner-initiated startle response in teleost fish. J. Exp. Biol. 66: 65-8 1. 1977.

In summary, very few command systems for complex (nonreflexive) behaviors have been described in detail. In most, such as the crayfish swimmeret system (4, 5) feeding in Pleurobranchaea ( 17), and leech swimming (40), the sufficiency criterion alone has been adequately demonstrated. Only in those systems where the number of elements involved was low, such as the crayfish lateral giant fibers (34) and the anuran tadpole Mauthner cells (39, was it possible to demonstrate both sufficiency and necessity criteria. The command system for locomotion in Aplysia is unusual in that the system is redundant with at least four individually sufficient and seem- ingly necessary command neurons. The identification of these neurons offers the possibility of investigating many important questions about how locomotion is controlled.

ACKNOWLEDGMENT

This work was supported by Public Health Service Grant NS 12483 to B. Jahan-Parwar.

Present address of S. M. Fredman: Dept. of Psychia- try, University of Iowa College of Medicine, Iowa City, IA 52242.

Received 24 May 1982; accepted in final form 8 De- cember 1982.

8. FARBER, D. S. AND KORN, H. Electrophysiology of the Mauthner cells: basic properties; synaptic mechanisms, and associated networks. In: Neurobiology of the Mauthner Cell, edited by D. S. Faber and H. Korn. New York: Raven, 1978, p. 47- 13 1.

9. FREDMAN, S. M. AND JAHAN-PARWAR, B. Intra- and interganglionic synaptic connections in the CNS of Aplysia. Brain Res. Bull. 4: 393-406, 1979.

10. FREDMAN,~. M. ANDJAHAN-PARWAR, B. Role of pedal ganglia motor neurons in pedal wave generation in Aplysia. Brain Res. Bull. 5: 179-193, 1980.

11. FREDMAN, S. M. AND JAHAN-PARWAR, B. Pro- cessing of chemosensory and mechanosensory information in identifiable Aplysia neurons. Comp. Biochem. Physiol. A 66: 25-34, 1980.

12. FREDMAN,~. M. ANDJAHAN-PARWAR, B.Cerebral neurons constituting a command system for locomotion in Aplysia. Sot. Neurosci. Abstr. 7: 642, 1981.

13. FURUKAWA, T. AND FURSHPAN, E. J. Two inhibitory mechanisms in the Mauthner neurons of goldfish. J. Neurophysiol. 26: 140- 176, 1963.

14. GETTING, P. A. Neuronal organization of escape swimming in Tritonia. J. Comp. Physiol. 12 1: 325- 342, 1977.

15. GETTING, P. A., LENNARD, P. R., AND HUME,

PEDAL WAVE COMMAND NEURONS

16.

17.

18.

19.

20.

21.

22.

23.

24.

25.

26.

27.

28.

29.

R. I. Central pattern generation mediating swimming in Tritonia. I. Identification and synaptic interactions. J. Neurophysiol. 44: 15 1- 164, 1980. GILLETT, R., GILLETT, M. U., AND DAVIS, W. J. Action-potential broadening and endogenously sus- tained bursting are substrates of command ability in a feeding neuron of Pleurobranchaea. J. Neuro- physiol. 43: 669-685, 1980. GILLETT, R., KOVAC, M. P., AND DAVIS, W. J. Command neurons in Pleurobranchaea receive synaptic feedback from the motor network they excite. Science 199: 798-801, 1978. GRILLNER, S. Some aspects on the descending control of the spinal circuits generating locomotor movements. In: Neural Control of Locomotion, edited by R. M. Herman, S. Grillner, P. S. G. Stein, and D. G. Stuart. New York: Plenum, 1977, p. 35 l- 375. HENING, W. A., WALTERS, E. T., CAREW, T. J., AND KANDEL, E. R. Motor neuronal control of locomotion in Aplysia. Brain Res. 197: 23 l-253, 1979. JAHAN-PARWAR, B. Behavioral and electrophysiological studies on chemoreception in Aplysia. Am. Zool. 12: 525-537, 1972. JAHAN-PARWAR, B. ANDFREDMAN, S.M.Cerebral ganglion of Aplysia: Cellular organization and origin of nerves. Comp. Biochem. Physiol. A 54: 347-357, 1976. JAHAN-PARWAR, B. ANDFREDMAN, S. M.Control of pedal and parapodial movements in Aplysia. I. Proprioceptive and tactile reflexes. J. Neurophysiol. 41: 600-608, 1978. JAHAN-PARWAR, B. ANDFREDMAN, S. M. Control of pedal and parapodial movements in Aplysia. II. Cerebral ganglion neurons. J. Neurophysiol. 4 1: 609-620, 1978. JAHAN-PARWAR, B. AND FREDMAN, S. M. Neural control of locomotion in Aplysia: role of the central ganglia. Behav. Neural Biol. 27: 39-58, 1979. JAHAN-PARWAR, B. AND FREDMAN,~. M. Role of interganglionic synaptic connections in the control of pedal and parapodial movements in Aplysia. Brain Res. Bull. 4: 407-420, 1979. JAHAN-PARWAR, B. AND FREDMAN, S. M. Motor program for pedal waves during Aplysia locomotion is generated in the pedal ganglia. Brain Res. Bull. 5: 169-177, 1980. JAHAN-PARWAR, B. ANDFREDMAN,S.M.I~~~~~~- tion of chemosensory and mechanosensory information in the nervous system of Aplysia. In: Ad- vances in Physiological Science. Neurobiology ofIn- vertebrates, edited by J. Salanki. New York: Pergamon, 198 1, vol. 23, 97-l 13.

JAHAN-PARWAR, B. ANDFREDMAN, S.M.Control of extrinsic feeding muscles in Aplysia. J. Neuro- physiol. 49: 1481-1503, 1983. KORN, H., TRILLER, A., AND FABER, D. S. Strut- tural correlates of recurrent collateral interneurons producing both electrical and chemical inhibitions

30.

31.

32.

33.

34.

35.

36.

37.

38.

39.

41.

42.

43.

44.

45.

of the Mauthner cell. Proc. R. Sot. London Ser. B 202: 533-538, 1978. KRAMER, A. P., KRASNE, F. B., AND BELLMAN, K. L. Different command neurons select different outputs from a shared premotor interneuron of crayfish tail-flip circuitry. Science 2 14: 8 1 O-8 12, 1981. KUPFERMANN, I. AND WEISS, K.R.Thecommand neuron concept. Behav. Brain Sci. 1: 3-39, 1978. LARIMER, J. L. Command interneurons and locomotor behavior in crustaceans. In: Neural Control oflocomotion, edited by R. M. Herman, S. Grillner, P. S. G. Stein, and D. G. Stuart. New York: Plenum, 1976, p. 293-326. MCCAMEN, R.E. AND MCKENNA, D.G.Neurons of the cerebral ganglion of Aplysia with extensive monosynaptic connections projecting to neurons in other ganglia. Sot. Neurosci. Abstr. 4: 201, 1978. OLSON,G. C. AND KRASNE, F. B. The crayfishlat- era1 giants as command neurons for escape behavior. Brain Res. 2 14: 89- 100, 198 1. ROCK, M. K., HACKETT, J.T., ANDBROWN, D.L. Does the Mauthner cell conform to the criteria of the command neuron concept? Brain Res. 204: 2 l- 27, 1981. STEIN, P. S. G. Motor systems with specific reference to the control of locomotion. Annu. Rev. Neurosci. 1: 61-81, 1978. STEN'T, G. S., KRISTAN, W. B., JR., FRIESEN, W. O., ORT, C. A., POON, M., AND CALABRESE, R. L. Neuronal generation of the leech swimming movement. Science 200: 1348-l 357, 1978. WALTERS, E. T., CAREW, T. J., AND KANDEL, E. R. Identification of sensory neurons involved in two forms of classical conditioning. Sot. Neurosci. Abstr. 7: 353, 1981. WEEKS, J. C. Neuronal basis of leech swimming: separation of swim initiation, pattern generation, and intersegmental coordination by selective le- sions. J. Neurophysiol. 45: 698-723, 198 1. WEEKS, J. C. AND KRISTAN, W. B. Initiation, maintenance and modulation of swimming in the medicinal leech by the activity of a single neuron. J. Exp. Biol. 77: 7 l-88, 1978. WIERSMA, C. A. G. AND IKEDA, K. Interneurons commanding swimmeret movements in the crayfish, Procambarus clarkii (Girard). Comp. Biochem. Physiol. 12: 509-525, 1969. WILLOWS, A.O.D., DORSETT, D.A., ANDHOYLE, G. The neuronal basis of behavior in Tritonia. III. Neuronal mechanism of a fixed action pattern. J. Neurobiol. 4: 255-285, 1973. ZOTTOLI, S. J. Correlation of startle reflex and Mauthner cell auditory responses in unrestrained goldfish. J. Exp. Biol. 66: 243-254, 1977. ZUCKER, R. S. Crayfish escape behavior and central synapses. I. Neural circuit exciting lateral giant fiber. J. Neurophysiol. 35: 559-621, 1972. ZUCKER, R. S., KENNEDY, D., AND SELVERSTON, A. I. Neuronal circuit mediating escape responses in crayfish. Science 173: 645-650, 197 1.

Command Neurons for Locomotion in Aplysia

Documents

Transcript of Command Neurons for Locomotion in Aplysia