OUP user menu

Evidence that the Central Pattern Generator for Swimming in Tritonia Arose from a Non-Rhythmic Neuromodulatory Arousal System: Implications for the Evolution of Specialized Behavior

Paul S. Katz, David J. Fickbohm, Christina P. Lynn-Bullock
DOI: http://dx.doi.org/10.1093/icb/41.4.962 962-975 First published online: 1 August 2001


Comparisons of the nervous systems of closely related invertebrate species show that identified neurons tend to be highly conserved even though the behaviors in which they participate vary. All opisthobranch molluscs examined have a similar set of serotonin-immunoreactive neurons located medially in the cerebral ganglion. In a small number of species, these neurons have been physiologically and morphologically identified. In the nudibranch, Tritonia diomedea, three of the neurons (the dorsal swim interneurons, DSIs) have been shown to be members of the central pattern generator (CPG) underlying dorsal/ventral swimming. The DSIs act as intrinsic neuromodulators, altering cellular and synaptic properties within the swim CPG circuit. Putative homologues of the DSIs have been identified in a number of other opisthobranchs. In the notaspid, Pleurobranchaea californica, the apparent DSI homologues (As1–3) play a similar role in the escape swim and they also have widespread actions on other systems such as feeding and ciliary locomotion. In the gymnosomatid, Clione limacina, the presumed homologous neurons (Cr-SP) are not part of the swimming pattern generator, which is located in the pedal ganglia, but act as extrinsic modulators, responding to noxious stimuli and increasing the frequency of the swim motor program. Putative homologous neurons are also present in non-swimming species such as the anaspid, Aplysia californica, where at least one of the cerebral serotonergic neurons, CC3 (CB-1), evokes neuromodulatory actions in response to noxious stimuli. Thus, the CPG circuit in Tritonia appears to have evolved from the interconnections of neurons that are common to other opisthobranchs where they participate in arousal to noxious stimuli but are not rhythmically active.


Organisms do not have the luxury of designing new components to best meet their current needs. Rather, during the course of evolution, innovations were built upon previously accumulated adaptations in a particular lineage. Thus, present day function is a reflection of an organism's ancestral history. This is particularly true of the nervous system, which tends to be highly conserved in closely related species (Kavanau, 1990; Katz, 1991; Paul, 1991; Arbas et al., 1991; Katz and Harris-Warrick, 1999). Yet, closely related species can exhibit important differences in their behavior even without exhibiting obvious differences in the gross structure of their nervous systems. Our understanding of the kinds of changes in the nervous system that can underlie species-specific behavior is still very limited (Katz and Harris-Warrick, 1999). Opisthobranchs provide a model for examining how different behaviors may have evolved from a common neuronal substrate.

Although opisthobranchs display many similarities in the organization of their nervous systems that derive from their common ancestry, there are distinct differences in the external form and lifestyle of these creatures. All opisthobranchs have paired cerebral, buccal, pedal, and pleural ganglia (see Fig. 4). Furthermore, putative homologous neurons can be identified in a broad variety of different species (Dorsett, 1974; Dickinson, 1979; Pentreath et al., 1982; Bulloch and Ridgway, 1995). These homologies are based on neurotransmitter content, axonal and dendritic projection pattern, physiology, and function (Croll, 1987b; Schmitt, 1995). Despite their apparent common origin and similarities in nervous system organization, the opisthobranchs display a wide variety of body forms and locomotory behaviors (Fig. 1). Some opisthobranchs swim via whole body movements, which are either dorsal/ventral or lateral. Others swim by flapping or waving specialized parapodia. Still others are incapable of swimming (Morton, 1964; Farmer, 1970).

Fig. 1. Members of the subclass Opisthobranchia exhibit a variety of swimming modes. Tritonia diomedea (Order: Nudibranchia) utilizes dorsal/ventral flexions for swimming. Some other nudibranch species, such as Phestilla sibogae have not been observed to swim. Other nudibranchs, such as Melibe leonina, swim using lateral body flexion. Pleurobranchaea californica (Order: Notaspidea) is a dorsal/ventral swimmer like Tritonia. Clione limacina (Order: Gymnosomata) swims constantly by flapping specialized wing-like parapodia. Some members of the genus Aplysia (Order: Anaspidea), such as A. brasiliana, swim by producing wave-like movements of their parapodia, while others such as A. californica do not swim at all. Tritonia image courtesy of W. N. Frost. Phestilla image provided courtesy of M. Hadfield. Melibe image courtesy of W. H. Watson III. Pleurobranchaea image was modified from Davis and Mpitsos, 1971. Clione image modified from Satterlie et al., 1997. Aplysia brasiliana image modified from Farmer, 1970. Aplysia californica image modified from Kandel, 1976. Phylogeny based on Bulloch and Ridgway, 1995

The neuronal basis for the swimming behavior in the nudibranch, Tritonia diomedea, has served as a model for understanding the organization of central pattern generator (CPG) circuits (Getting and Dekin, 1985b; Getting, 1989a). This animal swims to escape predators by repeatedly flexing its body dorsally and ventrally (Fig. 1). Identified neurons that comprise the CPG circuit for this rhythmic behavior are located in the cerebral portion of the fused cerebropleural ganglia (except for one cell type, VSI-B, which is located in the pleural portion). The apparent homologues of some of these neurons have been identified in other opisthobranch species. In particular, a set of serotonergic neurons, including neurons intrinsic to the Tritonia swim CPG, can be found in all opisthobranchs, regardless of whether that species is capable of swimming. In Tritonia, the serotonergic neurons fulfill an intrinsic neuromodulatory function (Katz and Frost, 1996; Katz, 1998), whereas in species that utilize other modes of swimming, the homologous neurons may be extrinsic modulators of the respective CPG circuits, but are not part of those CPGs. Comparisons of different opisthobranchs suggest that the presence of the neurons in the ancestral species predated their participation in a swim CPG circuit and that intrinsic neuromodulation may have arisen by co-opting an extrinsic neuromodulatory arousal system.


The neural basis for the swimming behavior has been well studied in Tritonia diomedea (Willows and Hoyle, 1969; Dorsett et al., 1973; Getting, 1989a). Swimming consists of a series of undirected dorsal/ventral whole body flexions (Willows et al., 1973). This rhythmic swimming behavior is produced in response to contact with a predator and results in thrashing movements that cause the animal to rise off the bottom and be carried away by currents. The CPG for this behavior is bilaterally symmetric with electrical coupling between the equivalent neurons in the left and right cerebral ganglia (Getting, 1981).

The motor pattern for Tritonia swimming (Fig. 2A) is generated as a result of the cellular properties of CPG neurons and their synaptic interconnections (Getting, 1989a, b). Sensory neurons, whose cells bodies are located in the pleural portion of the fused cerebropleural ganglion (Getting, 1976), synapse directly and indirectly on a command neuron, the dorsal ramp interneuron (DRI) (Frost and Katz, 1996; see also Frost et al., 2001). DRI monosynaptically excites a set of three neurons, the dorsal swim interneurons (DSI-A, B, and C) in each side of the bilaterally symmetric cerebropleural ganglion. The DSIs synapse onto two other CPG interneuron types: cerebral neuron 2 (C2) and the ventral swim interneurons (VSIs) (Getting et al., 1980; Getting, 1981; Getting and Dekin, 1985a).

Fig. 2. Intracellular recordings of the activity of serotonergic cerebral cells. A. In Tritonia, simultaneous recordings from a Dorsal Swim Interneuron (DSI) (a serotonergic neuron) and Cerebral interneuron 2 (C2), show a pattern of activity typical of the swim motor program. Upon electrical stimulation of a pedal nerve that innervates the body wall, both neurons fire repeated bursts of action potentials that underlie the swim motor program. Note the increased frequency of firing of the DSIs after the swim motor program. (Original recordings.) B. In Pleurobranchaea, the apparent homologues of the DSIs are the serotonergic As1–3 cells. The As2/3 cells are equivalent to DSI-B and C. A1 is the putative homologue of C2. As in Tritonia, nerve stimulation elicits rhythmic bursts in an As2/3 and A1. In contrast to Tritonia however, the burst period is about half as long. The As2/3 cell fires at a higher rate after the swim motor program, as is the case for the DSIs in Tritonia. However, unlike C2 in Tritonia, A1 continues to fire after the cessation of the swim motor program. Modified from Jing and Gillette, 1999. C. In Clione, the serotonergic Cr-SP neurons, apparent homologues of the DSIs and As1–3, are not active in phase with the swim interneurons. However, activation of a Cr-SP by depolarization can cause an acceleration of the swim motor pattern. Modified from Satterlie and Norekian, 1995. D. In Aplysia californica, electrical stimulation of the tail elicits a massive increase in spiking of CC3 (CB-1). CC3 does not display rhythmic activity. Modified from Mackey et al., 1989. E. The response of the DSIs in Tritonia when the stimulus is subthreshold for eliciting a rhythmic swim motor program is very similar to that of CC3 in Aplysia. The DSIs increase their tonic firing rate for many minutes after a nerve stimulus. (Original recordings.)

The DSIs are immunoreactive for the neurotransmitter serotonin (5-hydroxytryptamine, 5-HT) (Fig. 3A) (Katz et al., 1994; McClellan et al., 1994). Physiological evidence strongly suggests that the DSIs release 5-HT and that it is used both as a classical neurotransmitter, evoking fast and slow synaptic actions, and as a neuromodulator, altering the cellular and synaptic properties of other neurons (Katz and Frost, 1995a; Katz, 1998; Fickbohm and Katz, 2000; Clemens and Katz, 2001). The neuromodulatory effects include an enhancement of synaptic strength and an increase in the excitability of CPG neuron C2 (Katz et al., 1994; Katz and Frost, 1995b, 1997). Thus, when the DSIs are active, such as during the production of a rhythmic swim motor pattern, they dynamically modulate the cellular and synaptic properties of another CPG neuron. The ability of a neuron within a circuit to evoke neuromodulatory actions on other members of the circuit has been termed “intrinsic neuromodulation” (Katz and Frost, 1996). This is in contrast to “extrinsic neuromodulation,” in which neurons independent of a circuit evoke neuromodulatory actions upon cells in that circuit.

Fig. 3. Serotonin immunoreactivity in five opisthobranch species. In each example, the left giant serotonergic neuron is labeled with an asterisk (*). The medial serotonergic neurons highlighted in this paper are circled on the left side. All cells are bilaterally represented. The names of identified neurons are indicated on the right side. A. Tritonia diomedea. There are 5 serotonergic neurons located bilaterally near the midline. The group of three neurons located most laterally are the dorsal swim interneurons (DSI-A, B, and C). The other two neurons have not been identified in Tritonia. (Original photo.) B. The non-swimming nudibranch, Phestilla sibogae has a set of 5 unidentified serotonin-immunoreactive neurons in the same position as those in Tritonia. Photo courtesy of R. P. Croll. C. Pleurobranchaea californica. All 5 serotonergic neurons have been identified. The apparent DSI homologues are known as As1,2,3. As4 is more posterior and has more widespread axonal projections than As1–3. As-rh has an axon in the rhinophore nerve (Jing and Gillette, 1999). Photo modified from Sudlow et al., 1998. D. Clione limacina. There are two groups of medial serotonergic neurons. The anterior group (Cr-SA) contains 7 or 8 neurons. The posterior group (Cr-SP) contains 3 or 4 neurons (Satterlie and Norekian, 1995). Photo modified from Satterlie et al., 1995. The giant serotonin neuron is located more laterally than in other opisthobranchs (*). E. Aplysia californica. There are five neurons of the posterior cell cluster (PCC) in a very similar position to those in Tritonia and Pleurobranchaea. Only the most posterior of these cells has been named (CC3/CB-1). (Original photo.)

The intrinsic neuromodulatory actions of the DSIs appear to be necessary for the production of the swimming behavior in Tritonia: blocking the serotonergic receptors that mediate the neuromodulatory actions also prevents the production of the swim motor program (McClellan et al., 1994; Katz and Frost, 1995a). Moreover, the DSIs are the only cells in the CPG known to receive direct input from the command neuron, DRI (Frost and Katz, 1996), further suggesting a critical role in the initiation of the behavior. The command neuron can be bypassed by directly driving a DSI to fire action potentials. This is sufficient to initiate and maintain rhythmic activity in the CPG circuit (Fickbohm and Katz, 2000), illustrating that activation of these serotonergic neurons is also sufficient to initiate the swimming behavior.

The current hypothesis about how the swim is initiated and maintained is as follows. Sensory input excites the DSIs via DRI. The DSIs synaptically excite C2 while simultaneously increasing C2 excitability and synaptic strength. The enhanced synaptic strength allows C2 to recruit DRI polysynaptically. This sets up a positive-feedback loop (DRI to DSI to C2 and back to DRI). C2 also excites the VSIs, which provide negative feedback to DSI and C2, momentarily interrupting the positive feedback and initiating a ventral flexion. Thus, the rhythmic motor pattern arises from the properties of the network, including the neuromodulatory actions of the serotonergic DSIs.


The primary mode of locomotion for most members of the molluscan class Gastropoda, is crawling by muscular contractions of the foot or, in the case of Tritonia and other related species, gliding by means of cilia that line the foot. However, as with Tritonia, many other members of the subclass Opisthobranchia also exhibit swimming behavior (Morton, 1964; Farmer, 1970). Swimming appears to have evolved at least three times in this subclass because it involves at least three different means of propulsion (Fig. 1). Furthermore, the location of the CPG and the patterns of activity suggest that these different modes of swimming derive from independently evolved CPG circuits.

Within the order, nudibranchia, there is a variability in the propensity to swim and the mode of swimming. Many nudibranchs, such as the aeolid nudibranch Phestilla sibogae (R. P. Croll, personal communication), have not been observed to swim at all (Fig. 1). Even within the genus Tritonia, some species show more of a tendency to swim than others. For example, T. hombergi, produces less robust swims than T. diomedea (Willows and Dorsett, 1975) and T. festiva has not been observed to swim much at all (A. O. D. Willows, personal communication).

Some nudibranch species, such as Melibe leonina, swim with lateral bending movements of the body rather than dorsal/ventral flexions (Fig. 1) (Farmer, 1970). Melibe swims even more readily than Tritonia diomedea, doing so whenever disturbed and not just in response to a predator. Although this swimming behavior is superficially similar to swimming in Tritonia because it involves whole body flexions, a fundamental difference is that the muscles on the left and right sides are alternately activated. These lateral flexions are generated by alternation of activity in the left and right sides of the central nervous system (see Watson et al., 2001), in contrast to the bilaterally synchronous swim motor pattern in Tritonia. The CPG circuit seems to have components that are located in the fused cerebropleural ganglia and the pedal ganglia (Watson et al., 2001). These neurons do not appear to be homologous to the swim interneurons in the Tritonia swim CPG (W. H. Watson III, personal communication). Unlike in Tritonia, where the CPG members on the left and right sides are synchronized, in Melibe, the CPG neurons in the left and right ganglia mutually inhibit one another and are active in alternation.

The notaspid, Pleurobranchaea californica, has a dorsal/ventral swimming behavior that resembles that of Tritonia (Davis and Mpitsos, 1971; Jing and Gillette, 1995) (Fig. 1). Furthermore, the underlying neuronal circuit has many similarities. Many neurons in the CPG circuit seem to be homologous to those in Tritonia, based on location, synaptic connectivity, activity pattern during a swim motor program (Fig. 2B), and, for the presumed DSI homologues, neurotransmitter content (Fig. 3C) and axon projections (Fig. 4B) (Jing and Gillette, 1995, 1999). As in Tritonia, identified neurons located in the cerebral ganglia are electrically coupled to their contralateral homologues (Jing and Gillette, 1999), yielding a bilaterally symmetrical swim. Given the phylogenetic relationship of pleurobranchids to ancestral nudibranchs, Jing and Gillette (Jing and Gillette, 1995) hypothesized that the swimming behavior in Tritonia is derived from that of a pleurobranchid ancestor. However, it should be noted that as with Nudibranchia, there are members of Notaspidea that do not swim and others with different modes of swimming locomotion, such as parapodial waves (not shown) (Farmer, 1970). Thus, as will be discussed below, it is possible that swimming in these two groups may have evolved in parallel from a similar ancestral state where the serotonergic neurons provided neuromodulatory input to other circuits.

Fig. 4. Projection patterns of the putative cerebral serotonergic neuron homologues. A. In Tritonia, each DSI has a neuropil region in the cerebral portion (Ce) of the fused cerebropleural ganglion, ipsilateral to its cell body. They project to the contralateral cerebral hemiganglion where they have an additional arborization. An axon continues to the contralateral pedal ganglion (Ped). Electrophysiological evidence suggests that the axon projects further, to the ipsilateral pedal ganglion through the pedal-pedal connective. Drawing based on data from Getting et al. (1980). B. In Pleurobranchaea, the As1–3 neurons have very similar axonal arborizations to that of the DSIs in Tritonia. The As4 cells of Pleurobranchaea are probably homologous to the large serotonergic neuron located medial to the DSIs in Tritonia (Fig. 3A). The projections of As4, although more widespread than those of As1–3 are nonetheless exclusively contralateral. The nervous system of Pleurobranchaea resembles that of Tritonia except that the cerebral-pedal connectives are much longer. Modified from Jing and Gillette (1999). C. In Clione, the Cr-SP neurons appear to be homologous to the DSIs because their projection pattern is so similar, despite the differences in gross nervous system structure. The Cr-SP neurons have a strong projection to the pedal ganglion contralateral to the cell body. In Clione, as in Aplysia, but unlike Tritonia and Pleurobranchaea, the pleural ganglia are separate from the cerebral and pedal ganglia. The Cr-SV neuron may be homologous to the As4 cell of Pleurobranchaea except that it projects bilaterally and, thus, more strongly resembles the CC3 (CB-1) cell of Aplysia (see below). The Cr-SA neurons are probably not homologous to the DSIs because they project purely ipsilaterally and do not have neuropil in the cerebral ganglia. They may be homologous to anteriorly located serotonergic neurons in Tritonia and Pleurobranchaea that are located near the giant serotonergic neuron. Drawings of Cr-SA and Cr-SP modified from Satterlie and Norekian (1995). The drawing of Cr-SV is based on data presented in Norekian and Satterlie (1996). D. In Aplysia californica, the projection patterns of the presumed DSI homologues are still preliminary (McPherson and Katz, 2001) and not shown here (see text). The projection pattern of CC3 (CB-1) resembles that of Cr-SV in Clione. Modified from Wright et al. (1995). Abbreviations for ganglia: Ce, Cerebral ganglion; Pl, Pleural Ganglion, Ped, Pedal Ganglion

The pteropod, Clione limacina (Order: Gymnosomata) differs from the nudibranchs and notaspids discussed so far in that it swims continuously using paired parapodial appendages that resemble wings (Fig. 1) (Arshavsky et al., 1993, 1998). Swimming movements are used to pursue prey and escape from predators (Arshavsky et al., 1986, 1993). The CPG circuit for this motor pattern is localized to the pedal ganglia rather than the cerebral ganglia (Arshavsky et al., 1998). Furthermore, the swimming motor pattern has a period of less than one second, much faster than that seen in Tritonia or Pleurobranchaea (Fig. 2C). Thus, this behavior is not homologous, or even analogous, to swimming in Tritonia and Pleurobranchaea.

The order Gymnosomata is thought to have arisen from an ancestor in the order Anaspidea (Bulloch and Ridgway, 1995). Among the Anaspidea, numerous species of the genus Aplysia, such as A. brasiliana, are capable of swimming by undulating their parapodia (Fig. 1) (Johnson and Willows, 1999; Medina and Walsh, 2000). Although the commands for initiating swimming seem to arise from the cerebral ganglia (Gamkrelidze et al., 1995), the motor pattern is most likely generated in the pedal ganglia (Parsons and Pinsker, 1988) as is true in Clione. The similarities in action and gross anatomical origin suggest that swimming in Clione is directly descended from swimming in an Anaspid ancestor and therefore the behaviors may be homologous.

Other members of the Aplysia genus, such as A. californica, are incapable of producing the swimming behavior (Fig. 1) although their nervous systems appear outwardly identical (Johnson and Willows, 1999; Medina and Walsh, 2000). It is thought that swimming in this group may have originated as pumping movements involved in respiration (Johnson and Willows, 1999).

Thus, the opisthobranchs exhibit a great diversity of swimming behaviors (Fig. 1). Behaviors that resemble one another in form, such as the dorsal/ventral swimming of Tritonia and Pleurobranchaea, are generated by similar neuronal circuits, involving homologous neurons. However, behaviors that are outwardly different, such as the lateral bending of Melibe or the parapodial flapping and undulations of Clione and Aplysia brasiliana differ in their neuronal substrate.


Despite the fact that the neuronal circuits underlying different forms of swimming are distinct in their neuronal composition, neurons that appear to be homologous to the Tritonia swim interneurons can be identified in other species, based on neurotransmitter content, location, and axonal branching pattern. The neuron C2, a prominent white peptidergic neuron located in the cerebral ganglion of Tritonia (Getting, 1977; Taghert and Willows, 1978; Snow, 1982) appears to be homologous to A1 in Pleurobranchaea, and has a similar pattern of activity during the swim motor program (Jing and Gillette, 1995, 1999) (Fig. 2A, B). In Melibe, a putative C2 homologue is present in the cerebral ganglia, but it is not active during the production of the lateral swimming motor pattern (W. H. Watson III, personal communication).

Apparent homologues of Tritonia swim interneurons may even be present in non-swimming species. Although the peptide content of C2 is not known, it displays immunoreactivity to some FMRFamide and SCPB antisera. Antibody staining reveals a peptidergic neuron, similar in location to the C2 neuron, in a non-swimming species (Aeolidia) (Longley and Longley, 1987). Thus, certain neurons may be more highly conserved than the behaviors they serve.


If homologues of swim interneurons are present in species that do not use them for swimming, what is their function? An answer to this question is suggested by the serotonergic neurons that are physiologically and morphologically similar to the DSIs and found in a variety of species (Figs. 3, 4). In the cases presented here, we now know something about the function of these neurons.

The DSIs of Tritonia are located midway between the anterior and posterior edges of the fused cerebropleural ganglion, near the central commissure (Figs. 3A, 4A). Their axons project through the commissure to the contralateral cerebropleural ganglion and the contralateral pedal ganglion (Fig. 4A) (Getting et al., 1980). Action potentials of the DSIs can be recorded from the pedal-pedal connective, suggesting that the DSIs innervate both pedal ganglia. In addition to the three DSIs, there are two other unidentified serotonin-immunoreactive neurons located more medially in the same region (Fig. 3A).

The same organization of five serotonin-immunoreactive neurons in the cerebropleural ganglion has been observed in other nudibranchs, including the non-swimmer Phestilla sibogae (Croll et al., 2001) (Fig. 3B) and the lateral flexion-swimmer Melibe leonina (W. H. Watson III, personal communication). Hermissenda crassicornis, which has been reported to swim (T. Crow, personal communication), also has 5 serotonin-immunoreactive neurons in this position (Land and Crow, 1985; Croll, 1987a; Auerbach et al., 1989). Thus, the presence of potential DSI homologues is common for all nudibranch species that have been examined.


Apparent homologues of the DSIs can be recognized in notaspid, Pleurobranchaea, along with the two more medial serotonergic neurons (Fig. 3C) (Sudlow et al., 1998). In that species, the lateral cells, named As1–3, have the same projection pattern as the DSIs and appear to be DSI homologues (Fig. 4B, Left). The most posterior and medial of the five serotonergic neurons has been named As4 (Jing and Gillette, 1999). Although its axonal projection pattern differs from As1–3 (Fig. 4B, Right) it is also a member of the swim CPG. It will be interesting to determine whether the homologue of this cell in Tritonia has a similar function. The fifth serotonergic neuron (As-rh) (Fig. 3C) projects to the rhinophores and is not involved in swimming (Jing and Gillette, 1999).

In addition to their location and transmitter content, the As1–3 neurons resemble the DSIs in a number of other ways. Their synaptic actions on post-synaptic targets in the swim circuit are remarkably similar to those of the DSIs and their pattern of activity during and after the swim motor program resembles that of the DSIs (Fig. 2B) (Jing and Gillette, 1999). Furthermore, As1–4 excite other serotonergic neurons, including the giant cerebral neuron that enhances feeding (Jing and Gillette, 2000). We have seen a similar synaptic connection of the DSIs to C1 in Tritonia (unpublished observation). As1–4 also excite serotonergic neurons in the pedal ganglia that enhance ciliary locomotion (Jing and Gillette, 2000), as do the DSIs in Tritonia (Popescu and Frost, 2000). Since the As1–4 neurons are activated by noxious stimuli, and project to many types of neurons involved in the generation of feeding and locomotion behaviors, they seem to serve as a general arousal system (Jing and Gillette, 2000). The situation is likely similar in Tritonia: stimulation of swimming in intact animals, which activates the DSIs for a prolonged period of time, is followed by a period of increased creeping locomotion and feeding (unpublished observations).


Putative homologues of the DSIs also can be recognized in the Gymnosomatid, Clione, despite gross differences in the organization of the nervous system. There are two groups of medial serotonergic neurons on the dorsal side of the Clione cerebral ganglia (Fig. 3D). The cerebral serotonergic anterior group (Cr-SA, also known as CPB2 [Arshavsky et al., 1998]) does not appear to be homologous to the DSIs because neurons in that group project ipsilaterally to the pedal ganglia instead of contralaterally (Fig. 4C, Right) (Satterlie and Norekian, 1995). However the cerebral serotonergic posterior cells (Cr-SP, also known as CPB1 [Arshavsky et al., 1998]), a cluster of 3 or 4 neurons on the dorsal surface near the cerebral commissure, have axonal projections that more closely resemble those of the DSIs (Fig. 4C, Left). The Cr-SP neurons have branches in both cerebral hemiganglia, project contralaterally to the pedal ganglion, and continue through the pedal-pedal connective to the pedal ganglion ipsilateral to the cell body (Satterlie and Norekian, 1995). An additional serotonergic neuron (Cr-SV) is located on the ventral surface and projects bilaterally to both pedal ganglia (Fig. 4C, Center) (Norekian and Satterlie, 1996) Cr-SV in Clione may be homologous to As4 in Pleurobranchaea, although As4 does not project bilaterally.

Unlike the DSIs of Tritonia and As1–4 of Pleurobranchaea, none of the cerebral serotonergic neurons of Clione are active in phase with the swim motor pattern (Fig. 2C) (Satterlie and Norekian, 1995; Norekian and Satterlie, 1996; Arshavsky et al., 1998). Instead, these neurons appear to act as extrinsic modulators of the swim motor program in Clione. Increased spiking in any one of these neurons accelerates the swim rhythm (Arshavsky et al., 1992; Satterlie and Norekian, 1995, 1996; Norekian and Satterlie, 1996). As with the As1–4 neurons in Pleurobranchaea, the serotonergic cerebral cells in Clione also have widespread effects on the activity of other neurons (Satterlie and Norekian, 1995, 1996; Norekian and Satterlie, 1996). These serotonergic neurons are activated by touch of the tail (Arshavsky et al., 1992) and may serve as a general arousal system that accelerates swimming.


Aplysia also has serotonergic neurons in the same general position as the DSIs and associated cerebral serotonergic neurons of Tritonia (Fig. 3E). Serotonergic neurons are located in this area of the cerebral ganglion in a non-swimming species, Aplysia californica (Goldstein et al., 1984; Longley and Longley, 1986; Jahan-Parwar et al., 1987; Hawkins, 1989; Kemenes et al., 1989; Wright et al., 1995; Marois and Carew, 1997) and a swimming species, Aplysia brasiliana (McPherson and Blankenship, 1991). The most posterior of the five serotonergic neurons, CC3 (formerly called CB-1), has been physiologically and morphologically identified in Aplysia (Hawkins, 1989; Wright et al., 1995; Xin et al., 2001). The projection pattern of CC3 (CB-1), determined in both Aplysia brasiliana (McPherson and Blankenship, 1991) and Aplysia californica (Wright et al., 1995) (Fig. 4D) is very similar to that of Cr-SV in Clione (Fig. 4C, Center), suggesting that these neurons may be homologues.

The actions of CC3 on swimming in Aplysia brasiliana are unknown. However, CC3 is not likely to be an intrinsic component of the pattern generator for swimming because it is in the cerebral ganglion whereas the CPG has been localized to the pedal ganglia (Parsons and Pinsker, 1988). As with the serotonergic cerebral cells in the other species, CC3 has widespread actions and appears to act as an arousal neuron in the non-swimming Aplysia californica. CC3 excites the giant metacerebral cell (the homologue of C1 in Tritonia) and the heart excitor cell in the abdominal ganglion (Xin et al., 2001). It has neuromodulatory effects on neurons involved in the gill and siphon withdrawal response that may help mediate behavioral sensitization of this reflex response. The modulatory actions include presynaptic facilitation of sensory neuron synapses and enhancement of cellular excitability (Mackey et al., 1989), two effects that are very similar to the actions of the DSIs in Tritonia (Katz, 1998). In addition, CC3 fires tonically, at a high rate, for many minutes in response to tail shock (Mackey et al., 1989) (Fig. 2D). A similar long lasting tonic activation occurs in the DSIs of Tritonia following a swim response or if the stimulus falls below the threshold for initiating a rhythmic swim response (Fig. 2E). Oscillatory activity has not been reported for CC3 in Aplysia.

Preliminary evidence suggests that at least two of the other serotonergic neurons in this cluster are homologous to the DSIs (McPherson and Katz, 2001). These neurons project to the pedal ganglion through the cerebral-pedal commissure and have synaptic actions that resemble the DSI-B,C in Tritonia and As2/3 in Pleurobranchaea. These neurons also resemble previously described “command neurons” for swimming in Aplysia brasiliana (Gamkrelidze et al., 1995). Thus, the putative homologues of the DSIs in Aplysia may be involved in activating locomotion.


The neuronal circuit underlying the swimming response of Tritonia is a model system for studying rhythmic motor pattern generation. Understanding its evolutionary history can help explain its design and perhaps clarify observations that have been made regarding its operation.

Swimming is not unique to Tritonia. Indeed, it appears to have evolved multiple times during the evolution of opisthobranchs. The diversity of mechanisms underlying swimming and the different neurons involved in production of the swim attest to the independent origin of this behavior in the various lineages. Even within a lineage, there is variability in the expression of the behavior (Fig. 1). Thus, although most nudibranch species do not swim, some such as Melibe, swim via lateral body flexions, whereas members of the genus Tritonia swim dorsal-ventrally. Even within a genus, there may be differences in the ability to swim as is seen in Aplysia.

Despite the behavioral variability, the underlying neurons, in particular the putative homologues of the DSIs and C2 seem to be highly conserved. Conservation of neuronal structure seems to be a general rule in the evolution of behavior (Katz, 1991; Katz and Harris-Warrick, 1999). Since putative DSI homologues have been found in all opisthobranchs examined (over 9 different species), regardless of whether they exhibit a swimming behavior, it is likely that they represent a primitive characteristic of opisthobranchs. This viewpoint is supported by an outgroup comparison; gastropod molluscs outside the opisthobranchs, such as the prosthobranch, Littorina littorea (the periwinkle) (Croll and Lo, 1986) and the pulmonate snails, Lymnaea stagnalis (Croll and Chiasson, 1989) and Helisoma trivolvis (Diefenbach et al., 1998), also have serotonergic neurons in this position, although the number of cells varies.

Given that components of the swimming circuit in Tritonia appear to have been present in the common ancestors of all opisthobranchs, it is likely that the swim circuit arose from a previously existing network of neurons that was not involved with swimming. A likely scenario is that escape swimming in Tritonia arose out of the interconnections of neurons involved in responses to noxious stimuli. A common trait of the cerebral serotonergic neurons in the different opisthobranchs is that they respond strongly to noxious sensory stimuli and that they activate a variety of different motor systems. Even in Pleurobranchaea, where the putative DSI homologues, As1–3, are part of the swim pattern generator, they nonetheless have widespread actions on the feeding system and on the ciliary locomotion of the foot, consistent with their probable origin as an arousal system (Jing and Gillette, 2000).

The evolution of the Tritonia swim circuit from a general arousal circuit can help explain certain peculiar aspects of its operation. For example, in Clione, where the serotonergic neurons are extrinsic to the swim CPG circuit, treatment with the serotonin precursor, 5-HTP, enhances the swim program in the same fashion as serotonin application (Sakharov, 1990). However, in Tritonia, where the DSIs are components of the CPG circuit (Getting et al., 1980; Getting, 1981), 5-HTP treatment enhances the potency of the DSIs at evoking serotonergic actions, but does not produce a more vigorous swim pattern. Instead, the rhythmic oscillations become disabled (Fickbohm and Katz, 2000). A conclusion that can be drawn based on the function of serotonin in the two systems is that enhancing the output of serotonergic neurons extrinsic to a circuit may be less deleterious than altering the synaptic balance within a network oscillator by changing the serotonin levels of intrinsic components.

Understanding the evolution of the swim circuit and its components in Tritonia is also important for making generalizations about how similar circuits function in other species. For example, the many similarities between the swim circuit in Pleurobranchaea and that in Tritonia led to the conclusion that the mechanisms for producing oscillations are similar (Jing and Gillette, 1999). However, given the paucity of other pleurobranchids and nudibranchs that swim, it seems possible that the common ancestor of Pleurobranchaea and Tritonia did not swim. This may explain why there are neurons involved in the generation of the swim motor program in one species that have not been found in the other. If the two circuits evolved in parallel from a common substrate, they might utilize different subsets of neurons from pre-existing arousal networks to form an oscillator. For example, As4 plays a role similar to that of As1–3 (the putative DSI homologues). Although immunohistochemical staining for serotonin shows that the apparent homologue of As4 is present in Tritonia (see Fig. 3A), it has not been physiologically identified as a member of the swim circuit despite more than 20 years of recording from this region of the ganglion. It should be noted that the three DSIs were physiologically identified prior to the discovery that they are serotonin-immunoreactive (Getting, 1977; Katz et al., 1994; McClellan et al., 1994). Therefore, if there were a fourth DSI-like neuron associated with the swim, it probably would have been identified already. This suggests that the As4 homologue in Tritonia may not be a swim interneuron.

Opisthobranch swimming provides a lesson about how novel behaviors evolve. Selection pressure for a behavior must act on the existing circuitry. Simple changes in cellular properties or the distribution of receptors can result in large changes in the output of a circuit (Katz and Harris-Warrick, 1999). Neuronal circuits are often multifunctional, performing different behaviors at different times (Getting and Dekin, 1985b). Given the complexity of neuronal circuits, natural selection may act by restricting existing circuit multifunctionality or building upon it, rather than designing entirely new circuits to produce new behaviors. Thus, circuits underlying one behavior may have been co-opted from a different behavior or set of behaviors.


We would like to thank the participants of the symposium on Opisthobranch swimming for their useful discussions in formulating the ideas underlying this paper. We especially thank Yuri Arshavsky who first suggested to us that the Tritonia swim circuit might have arisen from a command network. We thank Win Watson for kindly providing the drawing of Melibe, Michael Hadfield for the image of Phestilla, and Roger Croll for the photomicrograph of Phestilla serotonin immunoreactivity. We thank Jiang Jing for comments on the manuscript and Vincent Rehder, Rhanor Gillette, and two anonymous reviewers for helpful suggestions. Our research is supported by a grant from the NIH, a Research Program Enhancement grant from GSU, and the Center for Behavioral Neuroscience. The symposium was supported by National Science Foundation grant, BN 990 5990.


  • 1 From the Symposium Swimming in Opisthobranch Mollusks: Contributions to Control of Motor Behavior presented at the Annual Meeting of the Society for Integrative and Comparative Biology, 4–8 January 2000, at Atlanta, Georgia.

  • 2 E-mail: pkatz{at}gsu.edu


View Abstract