Surprising multifunctionality of a Tritonia swim CPG neuron: C2 drives the early phase of postswim crawling despite being silent during the behavior

Abstract

In response to a suitably aversive skin stimulus, the marine mollusk Tritonia diomedea launches an escape swim followed by several minutes of high-speed crawling. The two escape behaviors are highly dissimilar: whereas the swim is a muscular behavior involving alternating ventral and dorsal whole body flexions, the crawl is a nonrhythmic gliding behavior mediated by the beating of foot cilia. The serotonergic dorsal swim interneurons (DSIs) are members of the swim central pattern generator (CPG) and also strongly drive crawling. Although the swim network is very well understood, the Tritonia crawling network to date comprises only three neurons: the DSIs and pedal neurons 5 and 21 (Pd5 and Pd21). Since Tritonia’s swim network has been suggested to have arisen from a preexisting crawling network, we examined the possible role that another swim CPG neuron, C2, may play in crawling. Because of its complete silence in the postswim crawling period, C2 had not previously been considered to play a role in driving crawling. However, semi-intact preparation experiments demonstrated that a brief C2 spike train surprisingly and strongly drives the foot cilia for ∼30 s, something that cannot be explained by its synaptic connections to Pd5 and Pd21. Voltage-sensitive dye (VSD) imaging in the pedal ganglion identified many candidate crawling motor neurons that fire at an elevated rate after the swim and also revealed several pedal neurons that are strongly excited by C2. It is intriguing that unlike the DSIs, which fire tonically after the swim to drive crawling, C2 does so despite its postswim silence.

NEW & NOTEWORTHY Tritonia swim central pattern generator (CPG) neuron C2 surprisingly and strongly drives the early phase of postswim crawling despite being silent during this period. In decades of research, C2 had not been suspected of driving crawling because of its complete silence after the swim. Voltage-sensitive dye imaging revealed that the Tritonia crawling motor network may be much larger than previously known and also revealed that many candidate crawling neurons are excited by C2.

INTRODUCTION

In response to an attack by a predatory sea star, or in response to suitably aversive skin stimuli, the marine mollusk Tritonia diomedea performs a stereotyped sequence of behaviors: an escape swim followed by several minutes of high-speed escape crawling (1) (Fig. 1A). The two behaviors are highly dissimilar: the escape swim is a muscular, rhythmic behavior in which whole body dorsal and ventral flexions propel the animal away from its predator (2), whereas escape crawling is a nonrhythmic gliding behavior mediated by the beating of foot cilia (1). The serotonergic dorsal swim interneurons (DSIs) of the swim central pattern generator (CPG) play central roles in both behaviors, firing rhythmically to generate the swim rhythm and then firing tonically postswim for tens of minutes to drive the foot cilia and crawling (3).

Figure 1.

A: illustration of the Tritonia escape swim and crawl behaviors. An escape swim is always followed by an escape crawl. B: diagram of the Tritonia swim and crawl circuitry. The swim circuit is well understood from the levels of the sensory neurons to swim command and central pattern generator (CPG) neurons. In addition to being members of the swim CPG, the dorsal swim interneurons (DSIs) are known to also drive crawling by exciting the only known crawling motor neurons, Pd5 and Pd21. Solid lines represent monosynaptic connections, and dashed lines represent polysynaptic connections. Bars represent excitatory synapses, and circles denote inhibitory synapses. DFN, dorsal flexion neuron; DRI, dorsal ramp interneuron; VFN, ventral flexion neuron; VSI, ventral swim interneuron. C: intracellular recording in the isolated Tritonia central nervous system (CNS) from C2 and 2 DSIs during the escape swim motor program and during the postswim period, showing that whereas the DSIs increase their firing rate after the swim, C2 is completely silent. D: intracellular recording from Pd5 during the swim motor program (indicated by the dashed gray line) and postswim period, showing that Pd5 is inhibited during the swim and then fires at an elevated rate in the postswim period. E: intracellular recording from C2 and Pd21 during the swim and postswim period, showing that Pd21 fires at an elevated rate postswim while C2 is completely silent. Fi: a C2 spike train (10 Hz, 2 s, gray bar) inhibits Pd5 for many seconds. Fii: a C2 spike train (10 Hz, 5 s, gray bar) weakly excites Pd21. The arrows in C–E denote when a stimulus (10 V, 10 Hz, 2 s, 5-ms pulses) was delivered to pedal nerve 3 (which contains sensory neuron axons) to elicit a swim motor program.

Whereas the swim circuit is well understood, especially in terms of the swim CPG (4) and the swim command neurons that drive it (5, 6) (Fig. 1B), the crawl network is less well characterized, with just the DSIs and two pedal neurons, Pd5 (7) and Pd21 (1, 8), identified so far. Pd5 is the largest neuron in the pedal ganglion (9) and is peptidergic (10). Popescu and Willows (7) found a strong correlation between Pd5’s spike rate and the animal’s crawling speed, and they found that depolarizing Pd5 to fire trains of spikes resulted in an increase in crawling speed. Pd21 is a large serotonergic neuron (11) located on the ventral side of the pedal ganglion, and its spiking strongly drives the beating of foot cilia (1, 8).

The serotonergic DSIs of the swim CPG fire at an elevated rate after swimming (Fig. 1C), and intact-animal treadmill experiments have shown that a single DSI can drive crawling (3). Further, the DSIs excite Pd5 and Pd21 (3), which also fire at an elevated rate after swimming (Fig. 1, D and E; Ref. 3). Popescu and Frost (3) also showed that the ventral swim interneurons (VSIs) of the swim CPG inhibit Pd5 and Pd21 and thus may inhibit crawling. Tritonia has been suggested to have evolved from a nonswimming ancestor that used crawling as its main mode of locomotion (12). Thus, since the escape swim network may have arisen from a preexisting crawling network, we wondered whether another member of the swim CPG, neuron C2, also plays a role in crawling. Because of its silence during postswim crawling in isolated brain preparations (Fig. 1C), C2 has not heretofore been implicated in driving crawling. In fact, based on its strong inhibition of Pd5 and weak excitation of Pd21 (Fig. 1, Fi and Fii; Refs. 8, 13), it was hypothesized that skin stimuli might instead activate C2 to inhibit crawling (3).

Given the above, we were surprised to find that in semi-intact preparations a 5-s C2 spike train strongly drives the beating of foot cilia that begins after the termination of the C2 spike train and continues for ∼30 s. Intracellular recordings from C2 in intact-animal swimming preparations confirmed that C2 is indeed silent before and after swimming. Since C2’s known connections to Pd5 and Pd21 cannot explain its ability to strongly drive crawling, we performed voltage-sensitive dye (VSD) imaging experiments on both the dorsal and ventral sides of the pedal ganglion and identified many neurons that fire at an elevated tonic rate after the swim motor program. Imaging also revealed several neurons in the dorsal pedal ganglion that are excited for many seconds after a brief C2 spike train. These neurons could represent previously unidentified crawling motor neurons that mediate C2’s ability to drive crawling. Our data demonstrate that C2, in addition to being a member of the escape swim CPG, also strongly drives the early phase of escape crawling, which interestingly occurs while the cell is silent. Furthermore, our findings demonstrate the existence of many candidate pedal crawling neurons that could be explored further in the future.

METHODS

Preparation

Tritonia diomedea, which are simultaneous hermaphrodites, were obtained off the coast of British Columbia, Canada by Living Elements. Animals were maintained in chilled (11°C) artificial recirculating seawater (Instant Ocean) systems before experiments. Experiments (except for the intact-animal preparations) were performed in filtered artificial seawater (fASW).

Intracellular Recording

Intracellular recordings were obtained with 15- to 30-MΩ electrodes filled with 3 M KCl connected to Dagan IX2-700 dual intracellular amplifiers. The resulting signals were digitized at 2 kHz with a BioPac MP 150 data acquisition system and recorded in the associated AcqKnowledge software, version 3.91. Swim motor programs were elicited by stimulation (10 V, 10 Hz, 2 s, 5-ms pulses) of pedal nerve 3, which was drawn into a suction electrode (hand pulled over an alcohol lamp from PE 100 tubing) by gentle negative pressure.

Semi-Intact Preparation

Tritonia were anesthetized by placing them in 1.4 L of fASW containing 1 mL of 1-phenoxy 2-propanol for 4 h (14). Once the animal was anesthetized, most of the internal organs were removed, and the animal was placed in a chamber ventral side up. The brain (bilateral cerebropleural and pedal ganglia) was inverted to render it dorsal side up (to allow for intracellular recording/stimulation of C2) and was pinned to the bottom of a small Sylgard-lined chamber at the top of a post, with slits in its wall to accommodate pedal nerves 2 and 3, which innervate the foot of the animal, on both sides of the brain. Furthermore, Pd5 sends axons out of the ipsilateral pedal nerves 2 and 3 (9, 15, 16), whereas Pd21 sends its axon out the ipsilateral pedal nerve 3 (1). To enable flipping of the brain to be dorsal side up, the cerebral and pleural nerves were severed. The preparation was placed in a refrigerator overnight, and the experiments were performed the following day, allowing more time for the effects of the anesthetic to wear off. Chilled fASW was fed into the chamber containing the brain to keep it between 10 and 11°C. C2 was impaled either through the sheath or after the sheath had been surgically removed. Since C2’s axon projects to the contralateral pedal ganglion, carbon particles were sprinkled on the side of the foot contralateral to C2 and high-definition video was taken to monitor the particles’ head-to-tail transit, which is an indicator of crawling speed. Five-millisecond current pulses were injected into C2 to depolarize it to action potential threshold; activity was observed on an oscilloscope to ensure that an action potential was elicited by each current pulse. C2 was typically driven at 10 Hz for 5 s, but in some experiments it was driven at 20 Hz for 5 s or at 10 Hz for 10 s. An output of the intracellular amplifier was connected to an audio monitor to mark precisely on the audio track of the video when C2 fired action potentials.

Intact-Animal Preparation

In these preparations, done at the University of Washington’s Friday Harbor Laboratories, an incision was made in the dorsal skin overlying the brain, and a set of hooks was placed around the edges of the opening. The hooks were connected by threads to rotatable posts at the tops of the chamber walls, allowing the opening to be maintained and the animal to be relatively immobilized but still free to execute swim movements. The brain was next exposed and fixed with stainless steel minutien pins placed through its surrounding connective tissue to a wax-covered, manipulator-mounted platform positioned within the animal. The sheath overlying one of the cerebropleural ganglia was then surgically removed, allowing intracellular recording with sharp electrodes from specific neurons during the swim behavior. The entire chamber was perfused with natural seawater, and the area around the brain was separately perfused with 12°C filtered seawater. Animals were induced to swim by squirting 4 M NaCl solution onto the skin, and the neurons were recorded while simultaneously filming the swim behavior. Data were stored on a Vetter model 402 PCM/FM recording adapter, which stored the video plus two channels of DC-coupled data on the audio tracks of the videotape.

Optical Imaging

In experiments aimed at identifying candidate escape crawling neurons, we used optical recording with a voltage-sensitive dye (VSD) to seek neurons in the pedal ganglia that increase their tonic firing in the minutes after the end of the swim motor program. The loose overlying connective tissue was removed, but the ganglion sheath was not removed and pedal ganglion neurons were stained by periodically (approximately every 5 min) applying pressure to a PE tube filled with 0.2 mg/mL RH-155 (Toronto Research Chemicals) pressed for 1 h against the surface of the ganglion. In experiments where we drove C2 and imaged in the contralateral dorsal pedal ganglion, the sheath over the cerebropleural ganglia was removed on the same side of the brain as the C2 to be impaled, and then the entire central nervous system (CNS) was bath perfused with 0.2 mg/mL RH-155 (AnaSpec) for 5–10 min; for the remainder of the optical imaging experiment, the ganglia were perfused with artificial seawater containing 0.02 mg/mL RH-155. During the experiment the temperature was maintained between 10.0 and 11.0°C by passing fASW through a feedback-controlled, in-line Peltier cooling system (model SC-20, Warner Instruments) with a peristaltic pump (model 720, Instech Laboratories). Temperature was monitored with a Bat-12 thermometer fitted with an IT-18 microprobe (Physiotemp Instruments).

Optical imaging experiments were performed with a RedShirtImaging PDA-III photodiode array. The PDA-III contains 464 photodiodes, with the full array sampled at 1.6 kHz, along with up to eight channels of DC-coupled electrophysiology. Data were acquired with RedShirtImaging’s NeuroPlex software. Optical data were AC coupled and then amplified (100×) by the PDA-III. In some experiments (e.g., driving C2 while imaging) a tungsten-halogen lamp house (Optiquip) powered by an Olympus TH4 DC power supply (2.5–12.5 V, 150 W) was used for illuminating the preparation. Light from the lamp house passed through the adjustable base diaphragm of the Olympus BX51WI upright microscope and was then filtered by a 725/50-nm band pass filter (Chroma Technology). Light was then passed through an electronic shutter (model VS35, Vincent Associates) and through a Nikon 0.9 numerical aperture (NA) substage condenser before reaching the preparation. In other imaging experiments we used a 735-nm LED (Thorlabs) to illuminate the preparation. Light passing through the preparation was collected by either a ×20/0.95 NA or a ×10/0.6 NA water immersion objective lens (Olympus) and passed through a phototube to reach either the photodiode array or an Optronics MicroFire 1.0 camera. Swim motor programs were elicited by stimulation of pedal nerve 3 contralateral to the imaged ganglion (10 V, 10 Hz, 2 s, 5-ms pulses).

Imaging Combined with Intracellular Stimulation

In some experiments C2 was impaled under the compound microscope and then the microscope was moved on its motorized translation stage (EXFO-Burleigh) to enable VSD imaging of the contralateral dorsal pedal ganglion. The contralateral pedal ganglion was imaged because C2’s axon projects contralaterally. Five-millisecond current pulses were injected into C2 to elicit the firing of action potentials, and C2’s activity was monitored on an oscilloscope to ensure that it fired spikes on all the current pulses. Physiological data from C2 were acquired simultaneously with the imaging data in NeuroPlex.

Data Analysis

Optical data were filtered in NeuroPlex (band pass Butterworth with 5 Hz and 100 Hz cutoffs) and saved as text files. Independent component analysis (ICA) was run on the optical data in MATLAB; customized MATLAB code can be found in Ref. 17. After ICA, components were thresholded to generate binary spike trains. Neuronal mapping was accomplished by fitting two-dimensional Gaussians to the weights of the diode signals contributing to each independent component, as in Ref. 18.

Determination of which neurons fired at higher frequencies postswim was accomplished with a custom MATLAB script that counted the number of spikes in time periods of equal length before and after the swim (ranging from 1 to 3 min postswim versus preswim). We determined which neurons fired >150% the number of spikes after the swim versus before the swim, and these were manually classified in relation to their activity during the swim: dorsal phase bursters, ventral phase bursters, double bursters (see Ref. 19 for a description of Tritonia flexion neurons), tonically firing, or inhibited. These neurons were then mapped onto the ganglion image in each preparation.

The same MATLAB script was used to determine which neurons were excited by C2: the spike rate of each neuron before and after the C2 spike train was calculated, and neurons that fired at a rate >150% of the pre-C2 spike train rate were considered to be excited by C2.

Statistical Analyses

Repeated-measures ANOVAs were performed in SigmaPlot 11.0.

RESULTS

C2 Stimulation Potently Activates the Foot Cilia

Since C2 is known to strongly inhibit Pd5 and to only weakly excite Pd21 (Fig. 1, Fi and Fii; Refs. 8, 13), we had previously hypothesized that C2 would inhibit crawling (3). To test this directly, we impaled C2 in semi-intact preparations while filming the movement of carbon particles on the foot (i.e., crawling) (Fig. 2A). To our surprise, we found that a 10-Hz, 5-s C2 spike train strongly drove particle movement on the foot, which began immediately upon the termination of the C2 spike train and lasted for ∼30 s [Fig. 2B and Supplemental Video S1 (see https://figshare.com/articles/media/Hill_et_al_Supplementary_video_mp4/24923211); 1-way repeated-measures ANOVA F(8,8) = 10.53, P < 0.001, n = 9 (9 trials in 4 C2 neurons in 4 animals), with multiple comparisons of postswim particle movement vs. baseline particle movement, Holm–Sidak method, P ≤ 0.002]. A striking feature was that this 30-s period of C2-activated particle movement began after C2 stopped firing and thus occurred when C2 was silent.

Figure 2.

Experiments in semi-intact preparations demonstrated that C2 strongly drives carbon particle movement on the foot, i.e., ciliary beating. A: schematic of the experimental setup. B: movement of carbon particles on the foot was significantly increased after the termination of a 10-Hz, 5-s C2 spike train (gray bar), with the effect lasting ∼30 s.

Although in deafferented isolated brain preparations C2 is silent after the swim motor program, it remained possible that in intact animals it fires in response to sensory feedback during postswim crawling. To test this possibility, we employed intact-animal electrophysiology preparations in which the brain was exposed and pinned to a rigid platform positioned within animals that were tethered but still able to perform the ventral and dorsal flexions of the swim (see methods). In all five preparations, although C2 fired vigorously as a CPG member during the swim (mean 3.4 ± 0.7 cycles per swim), it was completely silent (0 action potentials) in the minutes recorded before (5.2 ± 1.3 min) and afterward (13.7 ± 2.7 min). These recordings show that with feedback from the periphery fully intact, C2 remains silent throughout the postswim period, as in isolated brain preparations. This finding is in accordance with a previous study that also found that C2 is silent after the swim in intact animal preparations (20). Thus despite its silence after the swim, C2 potently drives the foot cilia for tens of seconds, a delayed effect of its firing during the swim.

VSD Imaging Identifies Pedal Neurons with Elevated Postswim Tonic Firing Appropriate for Crawling Motor Neurons

Since C2 inhibits Pd5 and only weakly excites Pd21, we reasoned that there must be other neurons in the pedal ganglion that are capable of driving cilia movement, and C2 must excite at least some of them to strongly drive carbon particle movement on the foot. Pedal motor neurons that drive crawling should fire at an elevated rate during the postswim period, as do the DSIs, Pd5, and Pd21 (Fig. 1, C–E). We therefore used VSD imaging to survey for such neurons on the dorsal and ventral sides of the pedal ganglion. We indeed found that many neurons in the dorsal pedal ganglion fired at an elevated rate postswim. Figure 3A shows the activity of neurons that in the postswim period (1 min) fired >150% the number of spikes they fired in the 1-min preswim period (33 out of a total of 41 neurons whose activity was imaged in this experiment). After a stimulus to the contralateral pedal nerve 3 (arrow in Fig. 3A), many neurons fired rhythmic bursts during the swim motor program, which lasted five cycles. A schematic of the experimental setup is shown in Fig. 3B, and the ganglion locations of the neurons shown in Fig. 3A are shown in Fig. 3C, color-coded by their activity during the swim: dorsal phase, ventral phase, double burster, tonic, and inhibited (key on right of Fig. 3D). A composite map of the ganglion locations of neurons that fired more frequently postswim from seven experiments is shown in Fig. 3D [the black oval represents the outline of the right pedal ganglion (note that in some experiments the left pedal ganglion was imaged; in these cases the maps were flipped horizontally)]. The ganglion locations of the different types of neurons are shown in Fig. 3E, color-coded by experiment (key at bottom of Fig. 3E). The five different subtypes of neurons that fired more postswim do not appear to be spatially segregated in the dorsal pedal ganglion.

Figure 3.

Voltage-sensitive dye (VSD) imaging in the dorsal pedal ganglion revealed that many neurons increased their spiking rate postswim. A: neurons that fired >150% of the spikes postswim that they fired preswim (33 neurons out of a total of 41 neurons recorded) are shown. A 10-V, 10-Hz, 2-s stimulus was given to the contralateral pedal nerve 3 (PdN3) at the arrow to elicit a swim motor program. B: schematic of the Tritonia central nervous system (CNS) showing the location of the photodiode array (PDA) (hexagon) over the right dorsal pedal ganglion and placement of the stimulating extracellular electrode. Ce, cerebral ganglion; Pd, pedal ganglion; Pl, pleural ganglion. C: map of the ganglion locations of the neurons shown in A, color-coded by their activity during the swim (key on right of D). The hexagon corresponds to the PDA’s field of view. D: composite map showing neurons from 7 preparations that fired >150% the number of spikes postswim compared to preswim. The black oval represents the outline of the right pedal ganglion. E: neurons from the 7 preparations that fired >150% the number of spikes after swim vs. before swim are shown by type (dorsal phase, ventral phase, double burster, tonic, inhibited), color-coded by experiment (exp). All of the types of neurons appear to be distributed throughout the ganglion without any clear spatial segregation.

As mentioned above, Pd21, one of the two known Tritonia crawling motor neurons, is located on the ventral side of the pedal ganglion. Therefore we also performed VSD imaging on the ventral side of the pedal ganglion to see whether, as on the dorsal side, many neurons fired more spikes postswim compared to preswim. Figure 4A shows the activity of neurons that in the 1-min postswim period fired >150% the number of spikes they fired in the 1-min preswim period (28 out of a total of 37 neurons whose activity was imaged in this experiment). After a stimulus to the contralateral pedal nerve 3 (arrow in Fig. 4A), many neurons fired rhythmic bursts during the swim motor program, which lasted five cycles. Most neurons on the ventral side of the pedal ganglion fired bursts on the dorsal phase of the swim. The experimental setup is shown in Fig. 4B. The ganglion locations of the neurons that fired more spikes postswim versus preswim are shown in Fig. 4C, color-coded by their activity during the swim (key on right of Fig. 4D). A composite map of the ganglion locations of neurons that fired more postswim from six experiments is shown in Fig. 4D [the black oval represents the outline of the pedal ganglion (note that in some experiments the left pedal ganglion was imaged; in these cases the maps were flipped horizontally)]. The ganglion locations of the different types of neurons are shown in Fig. 4E, color-coded by experiment (key at bottom of Fig. 4E). This revealed that dorsal phase neurons that fire more spikes after swim were arranged in a “C” shape on the ventral side of the pedal ganglion. Furthermore, based on their location, size, and activity during and after the swim motor program, some of the dorsal phase neurons that fired more spikes after swim are likely to be Pd21 (medium-sized cells in the upper lateral portion of the pedal ganglion, recorded in five preparations, outlined by the dashed rectangle in Fig. 4E). The neuron shown in the top trace in Fig. 4A, based on its location, size, and activity during and after the swim motor program, is likely to be Pd21, and it is noteworthy that the majority of neurons imaged in this experiment fire in a fashion very similar to Pd21 before, during, and after the swim, lending support to the possibility that these could be unidentified crawling motor neurons.

Figure 4.

Voltage-sensitive dye (VSD) imaging revealed that many neurons in the ventral pedal ganglion also increased their spiking rate after the swim. A: neurons that fired >150% of the spikes after the swim that they fired before swim (28 neurons out of a total of 37 neurons recorded) are shown. A 10-V, 10-Hz, 2-s stimulus to the contralateral pedal nerve 3 (PdN3) was given at the arrow. B: schematic of the experimental setup showing the location of the photodiode array (PDA) (hexagon) over the right ventral pedal ganglion. Ce, cerebral ganglion; Pd, pedal ganglion; Pl, pleural ganglion. C: map of the ganglion locations of the neurons shown in A, color-coded by their activity during the swim (key on right of D). D: composite map showing neurons from 6 preparations that fired >150% the number of spikes postswim compared to preswim. The black oval represents the outline of the right pedal ganglion. E: neurons from the 6 preparations that fired >150% the number of spikes postswim vs. preswim shown by type (dorsal phase, ventral phase, tonic, inhibited), color-coded by experiment (exp). Whereas the dorsal-phase neurons are arranged in a “C” shape along the periphery of the ganglion, there is no clear spatial segregation of the other types of neurons. The large dorsal-phase neurons in the top right quadrant (outlined by a dashed rectangle) are likely to be Pd21 (based on their location, size, and activity during and after the swim) recorded in 5 separate experiments.

VSD Imaging Identifies Pedal Neurons Excited by C2

The above data showed that many neurons on both sides of the pedal ganglion fire more spikes postswim and thus could be previously unidentified neurons that generate the crawling behavior. Since a C2 spike train strongly drives ciliary movement, and since C2 inhibits Pd5 and only weakly excites Pd21, we reasoned that C2 must excite additional, as-yet-unknown crawling motor neurons in the pedal ganglion. As a first step toward finding such neurons, we performed simultaneous intracellular stimulation of C2 and VSD imaging in the contralateral pedal ganglion, the ganglion to which C2 sends its axon. After a C2 spike train (10 Hz, 2 s), many neurons in the contralateral dorsal pedal ganglion were excited and fired spikes at an elevated rate (>150% of their spike rate before the C2 spike train; 19 out of 43 neurons imaged) for many seconds (Fig. 5A). A schematic of the experimental setup is shown in Fig. 5B. The ganglion locations of the neurons excited by the C2 spike train are shown in Fig. 5C and are color-coded by their activity during the swim (dorsal phase bursters shown in blue, tonically active neurons shown in black). In two additional preparations, a C2 spike train similarly excited many neurons in the contralateral dorsal pedal ganglion to fire at an elevated rate for many seconds (24 out of 48 neurons imaged and 23 out of 51 neurons imaged), but the activity of these neurons in the swim motor program was not determined. Analysis of the firing rate of these neurons before and in three 5-s bins (i.e., for the duration of the recording) following the C2 spike train of the experiment shown in Fig. 5A reveals a significant increase in spike rate following the C2 spike train (Fig. 5D). Analysis of the other two experiments also revealed a significant increase in spike rate in the three 5-s bins following the C2 spike train (Fig. 5, E and F). For all three experiments the data were analyzed with a Friedman repeated-measures analysis of variance on ranks, which revealed a highly significant effect (P < 0.001), followed by multiple pairwise comparisons (Tukey test) that revealed significant differences (P < 0.05) between the pre-C2 train spike rate and the spike rate in all three 5-s bins following the C2 train for all three experiments. Finally, in all three experiments, some neurons in the contralateral dorsal pedal ganglion were inhibited (<50% of the spike rate before the C2 train; 5 out of 43, 2 out of 48, and 13 out of 51 neurons in the 3 experiments) or were unaffected (spike rate between 50% and 150% of the spike rate before the C2 train; 19 out of 43, 24 out of 48, and 15 out of 51 neurons in the 3 experiments.) The present study thus demonstrated the presence of several candidate pedal ganglion cilia-driving neurons that could be examined in future studies.

Figure 5.

Voltage-sensitive dye (VSD) imaging combined with intracellular stimulation shows that C2 excites many neurons in the contralateral pedal ganglion. A: many neurons in the contralateral dorsal pedal ganglion were excited by a 10-Hz, 2-s C2 spike train (gray bar) to fire at >150% the spike rate they fired at before the C2 spike train. Other neurons in the contralateral pedal ganglion were either inhibited or not affected by the C2 spike train (data not shown). B: schematic of the experimental setup. The hexagon shows the location of the photodiode array. Ce, cerebral ganglion; Pd, pedal ganglion; Pl, pleural ganglion. C: map of the ganglion locations of the neurons excited by the C2 spike train color-coded according to their activity during the swim: blue, neurons that fired during the dorsal phase; black, neurons that were tonically active during the swim. D: bar graph showing the significant effect of the C2 spike train on the spike rate of the neurons shown in A for three 5-s bins (i.e., for the duration of the recording) following the C2 spike train. E and F: bar graphs showing similar significant effects of the C2 spike train on neurons in the contralateral dorsal pedal ganglion in 2 other experiments. In D–F * represent significant differences (P < 0.05).

DISCUSSION

The main finding of this study is that, after the termination of its spiking during the swim motor program, the Tritonia swim CPG neuron C2 surprisingly and strongly drives the beating of foot cilia for ∼30 s, and thus drives the early phase of the postswim escape crawling behavior when it is completely silent. Taken together with prior findings, swim CPG neuron C2 and the DSIs both drive postswim crawling but by very different mechanisms. The DSIs do so via elevated tonic firing lasting for tens of minutes, whereas C2 is likely to do so via slow synaptic connections that long outlast its spikes, to be confirmed in future studies.

Snow (21) provided early evidence that C2 releases a peptide, and later work showed that it in fact uses two as its neurotransmitters: FMRFamide and small cardioactive peptide (SCP) (22). Peptides are known to produce long-lasting effects. For example, they have persistent effects that can last for minutes in the Aplysia feeding network and have been shown to play important roles in network modulation (23). Next, peptidergic neurons known as bag cells control egg laying in Aplysia, which involves a series of behaviors that last for hours (24). Furthermore, many peptidergic neurons provide input to the crustacean stomatogastric ganglion (STG), and this peptidergic modulation acts to reconfigure the networks of the STG to produce a variety of different outputs (25). We hypothesize that C2’s spiking causes it to release FMRFamide and SCP, which then cause cilia-activating neurons in the contralateral pedal ganglion to fire at an elevated rate for ∼30 s. C2 is known to produce complex, long-lasting effects on neurons in the contralateral pedal ganglion. The C2-to-Pd5 synapse has four components: fast inhibition followed by fast excitation, slow inhibition, and then even slower excitation lasting tens of seconds (13). Snow also showed that Pd7 (the third-largest neuron in the pedal ganglion) receives slow excitatory input from C2 that lasts ∼30 s. Such slow, long-lasting excitatory actions from C2 to other as yet unidentified crawling motor neurons seem likely to mediate C2’s strong effects on cilia movement.

The fact that carbon particles on the foot only start moving after the end of the C2 spike train (Fig. 2A and Supplemental Video S1) leads us to speculate that while C2 is spiking, it may exert fast inhibitory effects on cilia motor neurons that keep them from firing. In this scenario, after the cessation of C2 spiking the fast inhibition would end and the slow excitatory effects of C2, which had been building in the background, would be unopposed to produce rapid-onset excitation and thus crawling. C2 is known to have precisely this type of effect on Pd10, a neuron that fires rhythmically in the ventral phase of the swim. Pd10 receives fast inhibitory input from C2 followed by slow excitatory input that lasts about a minute (19). We have shown here that many neurons that fire during the ventral phase of the swim then fire at an elevated rate postswim (Fig. 3 and Fig. 4). However, driving C2 and imaging in the contralateral dorsal pedal ganglion detected few neurons that were inhibited and then excited by C2. Although such neurons may be located in deeper layers of the dorsal pedal ganglion, or on its ventral side, another possibility is that C2 may excite neurons that transiently inhibit the movement of foot cilia.

Whereas crawling does not have to be preceded by a swim, the escape swim is always followed by an escape crawl [which is faster than run-of-the-mill crawling (1)]; thus, rapid crawling is an important second part of the escape behavior. Although the animal does crawl for many reasons other than escape, there is a distinct sequence to Tritonia’s escape behaviors: swimming followed by rapid crawling. Recent work has elucidated the neural underpinnings of sequential behaviors in a variety of species. For example, a neural-chain activation mechanism is thought to underlie song production in zebra finches (26). Next, a parallel activation scheme with hierarchical inhibition has been shown to underlie sequential grooming of Drosophila (27), and a ramp-to-threshold mechanism involving the spiking activity of a single descending neuron has been shown to drive sequential stages of Drosophila copulation (28). We propose a very different sequencing scenario in Tritonia: the escape swim and crawl behaviors are performed in the proper sequence because of the bursting of C2 and the DSIs during the swim, followed by C2’s slow, long-lasting excitatory effects on crawling motor neurons and by the postswim elevated spiking of the DSIs, which take over to drive sustained crawling for many minutes. Driving one DSI to fire at 20 Hz for 1 min causes other DSIs to fire at an elevated rate for many minutes (29), demonstrating that it is actually the DSIs’ own firing during the swim that drives their elevated postswim firing. Thus the sequencing mechanism in Tritonia is essentially “built in,” with the firing of the DSIs and C2 during the swim driving the postswim escape crawling behavior.

It should furthermore be noted that, in addition to driving crawling, the elevated postswim firing of the DSIs holds the memory for sensitization so that if the animal is attacked again it begins the swim with a shorter onset latency (30). This is a highly efficient neural mechanism, in which the CNS of Tritonia “kills two birds with one stone”: while the animal is performing its escape crawling behavior it is simultaneously prepared for another bout of swimming should it be preyed upon again. Sensitization of the escape swim of Tritonia is conceptually similar to serotonergic behavioral arousal that has been observed in other species of sea slugs (31, 32).

On the basis of our results, C2 now joins the DSIs as being multifunctional neurons playing central roles in generating both the swim and crawl behaviors. Based on their inhibition of Pd5 and Pd21, the VSI neurons of the swim CPG may also play dual roles; however, experiments in semi-intact preparations are needed to confirm that these interneurons do indeed inhibit crawling. Similar multifunctionality has been described in a variety of species. For example, the As neurons in Pleurobranchaea, another sea slug, are involved in swimming, turning, arousal, and crawling (33). Next, in the medicinal leech 93% of the neurons that oscillate with swimming also oscillate with crawling (34), though both are muscle-driven behaviors and are therefore more similar to one another than Tritonia’s swim and crawl behaviors. In addition, a multifunctional network in the crustacean STG produces two distinct outputs, the pyloric and gastric rhythms (35). Furthermore, many Aplysia pedal ganglion neurons burst rhythmically during both the gallop and crawl escape motor programs (18). Additionally, normal respiration and sighing in mammals is generated by a multifunctional network (36). Interestingly, peptidergic pathways comprise the core of the neural circuit that controls sighing (37). Finally, a multifunctional network in the brainstem underlies both respiration and nonrespiratory motor behaviors such as coughing, swallowing, and vomiting (38). Thus multifunctional neurons/networks are increasingly appreciated to produce many diverse behaviors in species ranging from invertebrates to mammals.

In the case of Tritonia, the multifunctionality of the DSIs and C2 may simply be a by-product of evolution. Evidence suggests that Tritonia evolved from an ancestor that did not swim (12) and used crawling as its only mode of locomotion and escape. Thus the Tritonia escape swimming network likely arose from preexisting elements of a crawling neural network. From this perspective, it may be expected that all the members of the Tritonia swim CPG play some role in crawling. The DSIs drive crawling by exciting Pd5 and Pd21, and the VSIs, based on their inhibition of Pd5 and Pd21, may serve as stimulus-induced “brakes” for the crawling network. Our present data demonstrate that C2 potently drives the early phase of crawling, helping ensure that it gets off to a fast start after the swim.

We have shown here that Tritonia swim CPG neuron C2 strongly, albeit rather briefly, drives the postswim crawling behavior. Our data suggest that the postswim crawl of Tritonia is driven by C2 and the DSIs in the early part of the behavior, and then sustained escape crawling, which can last for tens of minutes, is driven by continued elevated firing of the DSIs (Fig. 6A, Ref. 3). Despite several decades of research, this is the first report of C2’s ability to drive crawling. It is likely that because C2 is silent during crawling it was not even considered as a neuron that could drive this behavior. We also provide evidence that the Tritonia crawling network may have many more motor elements than have been previously identified. Many neurons on both the dorsal and ventral sides of the pedal ganglion fire at an elevated rate postswim, as would be expected of crawling motor neurons. In fact, the ventral pedal ganglion neurons that burst on the dorsal phase of the swim motor program and fire at an elevated rate postswim (Fig. 4E) may be the “crescent cells,” some of which Audesirk (8) found could sometimes cause an increase in foot cilia beating. Moreover, the location of the neurons that fire more postswim in the ventral pedal ganglion bears a striking resemblance to that of ventral pedal ganglion neurons that have been shown to be serotonin immunoreactive (39), lending further support to the notion that these could be previously unidentified cilia activating neurons. Furthermore, VSD imaging in the present study also revealed that a C2 spike train causes many neurons in the contralateral dorsal pedal ganglion to be excited for many seconds, prompting the hypothesis that many are previously unidentified crawling motor neurons that mediate C2’s ability to strongly drive the foot cilia (Fig. 6B). In summary, C2 and the DSIs are both multifunctional: they first drive Tritonia’s escape swim as members of the swim CPG, after which they drive the postswim escape crawling behavior by different mechanisms, C2 by excitatory connections whose effects long outlast its firing and the DSIs by elevated tonic firing that lasts for tens of minutes. For investigators in the field, the finding that C2 drives the early phase of crawling while silent broadens the group of neurons that must be evaluated as participants when mapping out circuits underlying behaviors.

Figure 6.

The data suggest that C2 and the dorsal swim interneurons (DSIs) work together to drive the initial phase of the escape crawl, after which extended crawling, which can last for many tens of minutes, is driven solely by the DSIs. A: C2 and the DSIs first generate the swim behavior as central pattern generator (CPG) members and then both drive the early phase of the escape crawl by different mechanisms: C2 by slow, long-lasting excitation and the DSIs by elevated tonic firing. Prolonged crawling is then driven solely by the DSIs. B: the Tritonia crawling network may be much bigger than was previously thought, with many hypothesized unidentified crawling motor neurons receiving fast inhibitory followed by slow excitatory input from C2. Solid lines represent monosynaptic connections, and dashed lines represent polysynaptic connections. Bars represent excitatory synapses, and circles denote inhibitory synapses. The red line between C2 and the hypothesized unidentified efferent crawl neurons signifies that although we speculate that the connection is monosynaptic, it has not yet been shown to be. DFN, dorsal flexion neuron; DRI, dorsal ramp interneuron; VFN, ventral flexion neuron; VSI, ventral swim interneuron.

DATA AVAILABILITY

Data will be made available upon reasonable request.

SUPPLEMENTAL MATERIAL

Supplemental Video S1: https://figshare.com/articles/media/Hill_et_al_Supplementary_video_mp4/24923211.

GRANTS

This work was supported by NIH 1R01NS121220 to W.N.F.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

E.S.H. and W.N.F. conceived and designed research; E.S.H., J.W., J.W.B., and V.K.M. performed experiments; E.S.H. analyzed data; E.S.H. and W.N.F. interpreted results of experiments; E.S.H. prepared figures; E.S.H. drafted manuscript; E.S.H., J.W.B., V.K.M., and W.N.F. edited and revised manuscript; E.S.H., J.W., J.W.B., V.K.M., and W.N.F. approved final version of manuscript.