Serotonin Modulates the Central Pattern Generator for Locomotion in the Isolated Lamprey Spinal Cord

It is generally agreed that the sequences of motor commands for rhythmic stereotyped movements such as locomotion, ventilation and scratch are generated within the central nervous system by limited neuronal ensembles called central pattern generators (CPGs) (Delcomyn, 1980; Grillner, 1981; Stein, 1978). These CPGs are often located within nuclei of the brain stem or the spinal cord of vertebrates; decerebrate, deafferented animals can generate these simple stereotyped movements upon appropriate stimulation. However, even simple movements are rarely stereotyped in intact animals; instead, they exhibit remarkable plasticity, allowing the animal to adapt the movement according to need. Such adaptations are thought to arise from modulatory inputs which alter circuit interactions within the CPG, leading to changes in the motor pattern. For example, sensory feedback during locomotion can alter CPG output to guarantee that the movement remains adaptive (Forssberg, Grillner & Rossignol, 1975; Grillner & Wallén, 1982). Descending control systems from the brain have been shown to modify CPG output so as to alter the frequency or reset the rhythm of the movements (Shik, Severin & Orlovsky, 1966; Buchanan & Cohen, 1982; Drew & Rossignol, 1984), and to lengthen, shorten or change the amplitude of specific components of a movement (Boylls, 1978; Russell & Zajac, 1979; Udo, Matsukawa & Kamei, 1979; Orlovsky, 1972; Drew & Rossignol, 1984). The cellular mechanisms for this modulation of CPG activity are not well understood in vertebrates, due to the complexity of their nervous systems and the difficulty in identifying the components of the CPG. Some advances have been made in several simpler invertebrate preparations (Augustine, Fetterer & Watson, 1982; Cooke & Sullivan, 1982; Kristan & Weeks, 1984; Harris-Warrick & Kravitz, 1984; Beltz et al. 1984; Flamm & Harris-Warrick, 1984; Harris-Warrick & Flamm, 1984).

We have begun to study the modulation of the CPG for locomotion in the lamprey, Ichthyomyzon unicuspis. The nervous system of the lamprey, while clearly having a vertebrate organization, shares many experimental advantages with invertebrate preparations (Rovainen, 1979). For example, the spinal cord, when removed from the animal, survives in vitro for several days. It has relatively few neurones, many of which are visible under the dissecting microscope. Moreover, any group of four or more segments of cord can be induced to generate the motor pattern characteristic of swimming (called ‘fictive swimming’) by the addition of excitatory amino acids such as N-methyl-D-aspartic acid or D-glutamatic acid to the superfusing saline (Poon, 1980; Cohen & Wallén, 1980), demonstrating the presence of a distributed CPG network for locomotion in the spinal cord. Ayers, Carpenter, Currie & Kinch (1983) have proposed that this CPG should be thought of as a basal CPG for a family of related undulating body movements, including swimming forward and backward, burrowing in the mud and crawling on a solid surface. These movements have a similar basic motor pattern, differing only in quantitative parameters such as frequency and amplitude of motoneurone bursting. Presumably, descending pathways from the brain as well as sensory feedback modulate this basal CPG, either directly or indirectly, to produce this family of related behaviour patterns.

Our approach to the study of CPG regulation has been to identify endogenous compounds in the lamprey which are capable of modulating the CPG for locomotion in the isolated lamprey spinal cord. We have found that serotonin causes a dosedependent reduction of ventral root (VR) burst frequency, and enhances the duration and intensity of VR burst discharge. These effects appear to arise from actions of serotonin either directly or indirectly on the CPG itself. We have not detected any direct effects of serotonin on the motoneurones innervating the axial musculature. In related work (Filler, Simmons & Harris-Warrick, 1983; Harris-Warrick, McPhee & Filler, 1985), we have used immunocytochemical methods to show that serotonin is present in spinal neurones and processes of the lamprey spinal cord. Our combined anatomical and physiological results suggest that endogenous serotonin may play a functional role in altering the CPG rhythm to produce adaptive movements appropriate to the needs of the animal.

Two species of lampreys were used in these experiments: adult silver lampreys (Jchthyomyzon unicuspis) from the Mississippi River, and small adult sea lampreys (Petromyzon marinus) from Cayuga Lake, New York. No significant differences were observed between these species. The animals were kept in freshwater circulating aquaria at 2–4 °C without feeding.

The in vitro preparation of the spinal cord was made essentially as described by Buchanan & Cohen (1982). Briefly, after decapitation, a 3-to 9-cm length of spinal cord (20-60 spinal segments) with notochord was removed from the animal. The dorsal meninx primitiva were removed to facilitate electrode penetration for intracellular recording. Such preparations were used during the first 3 days following dissection and were stored overnight at 4 °C in oxygenated saline.

Experiments were performed at 10-13 °C in a modified frog Ringer fluid (Buchanan & Cohen, 1982) containing (in mmol I⁻¹) NaCl, 115; KC1, 2.1; CaCl₂, 2.6; MgCl₂, 2.0; NaHCO₂, 3 and glucose, 3. In some experiments, an alternative saline (Wickelgren, 1977) was used (in mmol I⁻¹): NaCl, 91; KC1, 2.1; CaCl₂, 2.6; MgCl₂, 1.8; NaHCO₃, 20 and glucose, 4, buffered to pH 7.4 by bubbling with 95% O₂ and 5% CO₂. No differences were detected in experiments comparing the effects of these salines. In addition, the temperature of the saline was systematically varied from 6 to 16°C; the effects of serotonin were identical at all temperatures, although the rate of ventral root discharge varied as a function of temperature. Serotonin creatinine sulphate (Sigma Chemical Co., St Louis, MO) was dissolved in saline immediately before use.

The motor programme for swimming was induced by the addition of 0.25–1 mmol I⁻¹ D-glutamate to the saline, which was superfused over the preparation at 3-6 ml min⁻¹. Motor activity was monitored with bipolar suction electrodes whose tips (60-100µm) were placed over the ventral roots between the spinal cord and the wall of the spinal canal. Activity was monitored conventionally and recorded on tape. For analysis of the burst structure, the data were played back onto a chart recorder, digitized using a bitpad (Summagraphics, Inc.) and analysed on a MINC-11 microcomputer. In particular, we measured the frequency of ventral root burst discharge and the phase relationships between onset of bursts in different ventral roots. Frequency was defined as the inverse of the time between successive burst onsets. Intersegmental phase lags were computed by measuring the delay from the onset of the more rostral burst to the onset of the more caudal burst as a fraction of the rostral burst period.

To make intracellular recordings from motoneurones, we used glass micro-electrodes with resistances of 60–80 MΩ when filled with 4 mol I⁻¹ potassium acetate.

Current was injected through the recording electrodes via a bridge circuit. We identified motoneurones by a 1:1 correlation between action potentials recorded intracellularly from the soma and extracellularly from an adjacent ipsilateral ventral root. The resting potential was monitored conventionally. If the membrane potential decreased with time due to injury, the cell was rejected. The apparent threshold for action potential generation was monitored by delivering 100-ms depolarizing current steps through the recording electrode and slowly increasing the level of depolarization until an action potential was generated. An average of four to six determinations was made at each time point. This value reflects the spike threshold in the soma, which may be different from that at the spike initiation zone; however, it is useful for comparative purposes in the presence and absence of serotonin.

The central pattern generator for swimming was activated in the isolated lamprey spinal cord by addition of D-glutamate (0.25-1 mmol I⁻¹) to the superfusion saline (Poon, 1980; Cohen & Wallén, 1980). The motor programme was initially monitored with extracellular recordings of motoneurone activity in ventral roots along the spinal cord (Fig. 1A). This motor programme was characterized by three features:(1) rhythmic bursts of motoneurone action potentials from a ventral root, at a frequency of about 0-5-1 burst s⁻¹; (2) alternation in bursting between left and right ventral roots of a single segment; (3) a relatively constant phase lag separating the onset of bursts from ipsilateral ventral roots in different spinal segments, with rostral bursts leading caudal bursts by about 1% of the period per spinal segment separating the roots. This motor activity reflects that observed in intact animals (Wallén & Williams, 1982, 1984) and generates the undulating sinusoidal movement of the body that propels the lamprey through the water (Grillner & Kashin, 1976).

When serotonin was added with D-glutamate to the superfusing saline, the swimming motor programme was altered in several ways (Fig. 1B). The most obvious effect was a reduction in the frequency of ventral root burst discharge. This reduction was dependent upon the serotonin concentration (Fig. 2A). A detectable reduction was observed at 5x10⁻⁹moll⁻¹ serotonin, and this effect became stronger with increasing serotonin concentrations. A 50% reduction in burst frequency was seen with 10⁻⁷ mol I⁻¹ serotonin, and an apparent saturation of the effect was seen between 10⁻⁶ and 10⁻⁵moll⁻¹ (Fig. 2C). At saturation, the frequency of ventral root burst discharge was reduced about 20-fold.

It has been shown previously (Cohen & Wallén, 1980) that the duration of the ventral root burst is a constant proportion of the cycle over a wide range of frequencies. Therefore, we expected to see a prolongation of burst duration as serotonin reduced the burst frequency, and this was observed (Figs 1B, 2A). In Fig. 2A for example, each ventral root burst lasted nearly 4 s during superfusion with 10⁻⁶ mol I⁻¹ serotonin, long enough for five to six complete cycles in the pre-serotonin controls. In addition, the number and frequency of action potentials within a burst increased with higher serotonin concentrations, as can be seen at higher sweep speed (Fig. 2B). The change in burst intensity was fairly modest at serotonin concentrations below 10⁻⁷ mol I⁻¹ (see also Fig. 6) but became pronounced at higher concentrations (10⁻⁶ and 10⁻⁵ mol I⁻¹; see also Fig. 7). This enhancement appeared to result in part from the recruitment of new motoneurones that were previously inactive; this is seen most clearly in intracellular records from such motoneurones (Figs 6, 7). The activity of previously spiking motoneurones was also increased but it was difficult to quantify this from the extracellular records due to the overlap of action potentials during serotonin superfusion.

Serotonin had only slight effects on intrasegmental coordination (Fig. 1). The normal alternation of motoneurone bursting between left (L10) and right (RIO) ventral roots of a single spinal segment was preserved during serotonin superfusion. In six experiments, the average left-right phase lag changed from 0.53 ± 0.04 (S.D.) to 0.57 ± 0.05 during 10⁻⁶ or 10⁻⁵ mol I⁻¹ serotonin superfusion. Fig. 3 shows a typical result.

In contrast, the intersegmental phase lag separating the onset of ventral root burst discharge in different spinal segments could be markedly lengthened by serotonin superfusion. In the experiment shown in Fig. 4, for example, intersegmental phase lag between R10 and R30 was 0.17±0.07(s.D.)in control measurements, or 0.85% per spinal segment; this is similar to the values of 1% per segment reported by other workers (Ayers et al. 1983; Wallén&Williams, 1982, 1984). Addition of 10⁻⁷ mol l⁻¹ serotonin slowed and prolonged the VR burst discharges, and the R10-R30 intersegmental phase lag was significantly prolonged to 0.54 ± 0.05 (s.D.) (2.7% per segment). Raising the serotonin concentration to 10⁻⁶moll⁻¹ resulted in a further prolongation of the intersegmental phase lag to 0.61 ±0.07 (S.D.), or 3.1% per segment. If serotonin prolonged the intersegmental phase lag uniformly down the body during locomotion in vivo, it would cause a significantly increased degree of body curvature. When combined with serotonin’ s effects on burst frequency, this would result in multiple slow waves of contraction moving down the body. A similar prolongation of the intersegmental phase lag was observed in 15 out of 20 experiments. However, in five experiments there was a smaller (10%) shift in intersegmental phase lag. We do not know why animals varied in the strength of this response. The extent of change in intersegmental phase lag with serotonin did not correlate with age, sex, portion of spinal cord used, initial intersegmental phase lag or degree of change in burst frequency with serotonin.

Mathematical models of coupled, non-linear oscillators predicted that an increased intersegmental phase lag could result from differential effects of serotonin on the inherent frequency of bursting in different segments of the spinal cord. That is, if serotonin slowed the segmental oscillators for swimming in the caudal part of the cord to a greater extent than the rostral cord, a prolonged intersegmental phase lag would result (P. Holmes & R. Rand, personal communication; Cohen, Holmes & Rand, 1982; see Discussion). We have tested this by comparing the effects of serotonin on burst frequency in isolated rostral, medial and caudal segments of the spinal cord. In the experiment shown in Fig. 5, for example, a 60-segment length of spinal cord was used, and motor activity was monitored from spinal segments 10, 30 and 50. Serotonin reduced the frequency of bursting and lengthened the intersegmental phase lag from 0.72 ± 0.04% (S.E.) per segment to 1.9 ± 0.23% (s.E.) per segment (Fig. 5A). The spinal cord was then cut at segments 20 and 40 to separate the three recording sites within segments of equal length. Before the addition of serotonin, segments within the three pieces of cord had similar burst frequencies (Fig. 5B). However, after addition of serotonin (5x 10⁻⁸moll⁻¹) to the bathing fluid, the burst frequencies of the three segments became quite different. The burst frequency of the rostral segment decreased to 51% of controls, while the medial segment decreased to 39% and the caudal segment was completely inhibited and became silent. Similar results were seen in five other experiments, with serotonin concentrations from 10⁻⁸ to 10⁻⁷ mol I⁻¹. There was some variability in the extent of serotonin’ s effect on rostral and caudal segments, but in every experiment the caudal segments were affected to a greater extent than the rostral segments. This would increase the difference in inherent burst frequency between rostral and caudal segments, thereby causing an increased phase lag between the segments (see Discussion).

We have not pharmacologically characterized the receptors that serotonin interacts with to modulate fictive swimming. However, other amines we have tested did not have the same effect as serotonin. Norepinephrine (1–5×10 ⁴moll ^J) and octopamine (10⁻⁴moll⁻¹) had no detectable effect on fictive swimming. Dopamine (5 × 10⁶–10⁻⁴moll⁻¹) disrupted the motor pattern in a complex way. Spontaneous motor activity at first increased, then declined within 10 min. The rate of ventral root discharge tended to increase initially, but with time (5-10 min) the swimming motor pattern was disrupted completely. Although the effects of these amines have not been completely analysed, it is clear that they do not resemble serotonin in their effects on fictive swimming.

The increased intensity of each ventral root burst discharge observed with higher concentrations of serotonin (Fig. 2) could result from two contributing factors: (1) serotonin could enhance the CPG output onto the motoneurones, and/or (2) it could directly excite the motoneurones. The experiments described below suggest that the major effect of serotonin is an enhancement of the output from the CPG, and that the amine’ s direct actions on motoneurones are minimal.

Intracellular recordings from motoneurones during D-glutamate-induced fictive swimming show regular oscillations in resting potential in phase with ipsilateral VR burst discharge (Buchanan & Cohen, 1982; Russell & Wallén, 1980). This membrane potential oscillation arises from an alternation of active synaptic excitation and active synaptic inhibition from the CPG (Russell & Wallén, 1980; Kahn, 1982). In the example in Fig. 6A, the membrane potential of the myotomal motoneurone was oscillating, but it was not firing action potentials. Addition of 10⁻⁷moll⁻¹ serotonin caused a slight increase in the frequency of large excitatory postsynaptic potentials (EPSPs) during the depolarizing phase of the oscillation, generating occasional action potentials (Fig. 6B).

Higher concentrations of serotonin had more dramatic effects on motoneurone activity during fictive swimming. In the experiment shown in Fig. 7A,B, a different motoneurone was monitored. Addition of 3x 10⁻⁵ mol I⁻¹ serotonin caused a marked reduction in burst frequency and enhancement of burst intensity, as seen in the ventral root records. Simultaneously, the motoneurone began to receive a greatly enhanced synaptic input that was time-locked with the activity of the swimming CPG. This was manifested by a large increase in the frequency of EPSPs during the depolarizing phase of the membrane potential oscillation cycle (seen more easily at higher sweep speed in Fig. 7Bii). This caused an increased depolarization and the production of multiple action potentials per burst. In addition, enhanced inhibitory input during the hyper polarized phase of the cycle was also observed.

While these data suggest that serotonin enhances synaptic output from the swimming CPG onto the motoneurones, they do not preclude an alternative hypothesis: serotonin may directly enhance the motoneuronal responses to an unchanged synaptic input (for example, by a decreased conductance mechanism). To test this hypothesis, we monitored the effects of serotonin on several parameters of motoneuronal excitability.

The resting potential and threshold for action potential initiation were monitored in quiescent motoneurones, in the absence of fictive swimming. When D-glutamate was removed from the saline, fictive swimming rapidly ceased and oscillations in motoneuronal resting potential slowly faded, completely disappearing within a few minutes. Under these conditions, the motoneurones received occasional excitatory and inhibitory synaptic input (see Fig. 8, for example). Addition of serotonin alone (10⁻⁷–3x 10⁻⁵ moll⁻¹) did not elicit fictive swimming and the motoneurones did not fire action potentials. With low concentrations of serotonin (10⁻⁷ mol I⁻¹), neither the resting potential nor the action potential threshold were significantly affected by the amine; however, a slight increase in the frequency of inhibitory synaptic potentials was observed. In two experiments with high concentrations of serotonin (3x 10⁻⁵ mol I⁻¹), a slow oscillation in the membrane potential was observed; a portion of one cycle is shown in Fig. 8A. This oscillation had a period of almost 1 min, which was considerably longer than the oscillations in motoneurone membrane potential seen during serotonin-modulated fictive swimming. Recordings made at higher sweep speeds during the depolarized and hyperpolarized phases of the cycle are shown in Fig. 8B. During the depolarized phase of the oscillation, the motoneurone was indistinguishable from pre-serotonin controls: the resting potential and threshold for action potential generation were unaffected, and occasional excitatory and inhibitory synaptic potentials were observed. The hyperpolarized phase of the oscillation resulted from a long dense cluster of IPSPs (Fig. 8B, lower trace). The inhibitory input hyperpolarized the membrane potential by 2-4 mV and increased the apparent threshold for action potential generation, probably by a conventional shunting action. We know neither the source of this periodic synaptic inhibition seen during superfusion of high concentrations of serotonin, nor whether it has any functional significance.

Serotonin also had no detectable effect on the motoneuronal input resistance measured near the normal resting potential. This was tested in two ways (Fig. 9). During D-glutamate-induced fictive swimming, short (30-ms) hyperpolarizing current pulses were delivered to the neurone throughout the membrane potential oscillation (Fig. 9A). Serotonin (10⁻⁷moll⁻¹) reduced the ventral root burst frequency by more than 50%, and the motoneurone began to fire occasional action potentials. However, the input resistance, estimated from the amplitude of the voltage response to the hyperpolarizing current pulses, was not affected by serotonin. We also monitored the input resistance in quiescent motoneurones (without D-glutamate in the bath), so that we could generate more complete l/V curves with long hyperpolarizing and depolarizing current steps (Fig. 9B). Again, serotonin did not change the input resistance. In conclusion, serotonin had no detectable effects directly on motoneurones, as determined from measurements of their resting potential, threshold for action potential generation, and input resistance.

The data presented in this paper show that bath-applied serotonin can modulate the motor programme for fictive swimming in the isolated spinal cord of the lamprey. Serotonin had three major effects on the swimming motor programme: (1) a dosedependent reduction in the frequency of ventral root bursts; (2) an increase in the intensity of each ventral root burst, especially at higher serotonin concentrations; (3) prolongation of the intersegmental phase lag between different spinal segments. In principle, all of these effects can be envisaged as resulting from a single primary action of serotonin. However, since there is no evidence yet for a common origin, we will treat them as separate effects of the amine.

We do not yet know the cellular targets of serotonin action that cause these changes in CPG-elicited motor output. The dramatic dose-dependent modulation of the frequency of ventral root bursts suggests that serotonin may directly modulate the oscillator components of the CPG that determine the rhythmicity of the bursting. Alternatively, serotonin could affect other inputs to the oscillators of the CPG that constrain them to burst at frequencies different from their natural or inherent frequencies (Wilson, 1966); modulation of the activity of these inputs could lead to a change in CPG burst frequency without any direct effect of serotonin on the CPG itself. Resolution of these possibilities will have to await the identification of the components of the CPG.

Serotonin’ s potentiation of motoneurone activity during fictive swimming could result either from an enhanced synaptic output from the swimming CPG, or from a direct excitatory effect of serotonin on the motoneurones themselves, as has been described in other species (Barasi & Roberts, 1974; Parry & Roberts, 1980; Roberts & Wright, 1981; White & Neuman, 1983). Evidence for the first hypothesis is quite strong. Serotonin did increase appropriately phased excitatory and inhibitory synaptic input to the motoneurones during D-glutamate-activated fictive swimming. While these results suggest that the motoneurones receive enhanced synaptic input from the CPG, it is theoretically possible that serotonin instead amplifies the motoneurone’ s response to previously existing synaptic input. However, we have not detected any direct effects of serotonin on the membrane potential, threshold for action potential generation, and soma input resistance of myotomal motoneurones either during fictive swimming or in its absence. We cannot exclude the possibility that serotonin affects the membrane properties of distal dendrites of the motoneurones in a manner that was not detected from the cell body. However, it appears that the major mechanism by which serotonin potentiates motoneurone activity during fictive swimming is by causing a greater synaptic output from the swim CPG.

The effect of serotonin on intersegmental coordination could follow simply from its effects on the CPG burst frequency. Cohen et al. (1982) have presented a mathematical model of intersegmental coordination in the lamprey spinal cord. The circuit generating swimming was modelled as a double chain of coupled non-linear oscillators where each oscillator controls the activity of a single VR. Using this model, they have predicted that the intersegmental phase lag between oscillators is dependent upon two variables: (1) the difference in the inherent burst frequency of the two oscillators; if this difference increases, the phase lag will increase; (2) the strength of the intersegmental coupling between the oscillators; if this decreases, the phase lag will increase. We have observed a gradient in serotonin’ s effects on the burst frequencies in isolated spinal segments with the rostral oscillators affected least and the caudal oscillators affected most (Fig. 5). In the intact cord, this would produce an increased difference in inherent frequency between oscillators in the different spinal segments, and thereby lead to an increase in the phase lag between rostral and caudal segments, as predicted by the mathematical model. More detailed, quantitative comparisons are difficult, as we have no estimates of the intersegmental coupling strength in either the absence or the presence of serotonin. However, qualitatively, the lengthening of the intersegmental phase lag can be explained at least in part by differential effects of serotonin on the inherent frequency of CPG oscillation in different parts of the spinal cord.

Our dose-response studies (Fig. 2) indicate that the threshold for a serotonin effect is very low (5x 10⁻⁹moll⁻¹). These physiological results suggest that receptors sensitive to nanomolar concentrations of serotonin exist in the spinal cord. We have not yet done serotonin binding studies or pharmacological analyses of serotonin receptors in the lamprey. In other species, at least three pharmacologically distinct binding sites for serotonin have been detected, with apparent dissociation constants of 10⁻⁹–10⁻⁶molI⁻¹ (Peroutka & Snyder, 1979; Pedigo, Yamamura & Nelson, 1981; Monroe & Smith, 1983). These receptors may be associated with different physiological actions on target neurones (Peroutka, Lebovitz & Snyder, 1981; Peroutka & Snyder, 1982). Further work will be needed to correlate receptor type with modulation of swimming in the lamprey. In addition to possible classical transmitter roles for serotonin, the ability to modulate fictive swimming at very low concentrations suggests the possibility that serotonin may act in a diffuse, non-synaptic manner in the lamprey spinal cord. Similar proposals for a local hormonal action of serotonin as well as norepinephrine have been made in the mammalian brain (Beaudet & Descarries, 1976, 1978; Leger & Descarries, 1978; Moore, 1981; however, see Bloom, 1981), as well as the rat spinal cord (Maxwell, Leranth & Verhofstad, 1983). Strong evidence for a local hormonal action of a released transmitter has been presented for the peptide LH-RH in frog sympathetic ganglia (Jan & Jan, 1982).

The behavioural relevance of serotonin’ s effects on the CPG for fictive swimming is at present unclear. Although our data are suggestive, they certainly do not prove that serotonin normally modulates locomotion or other related undulatory movements in the intact animal. Our method of bath application introduces serotonin simultaneously to all sites of the spinal cord, thus eliminating any specificity of spatial or temporal release of the amine in vivo. Thus, the effects observed in our experiments may not mimic the effects of endogenously released serotonin in intact animals. However, our experiments do demonstrate that the CPG for fictive swimming is sensitive to serotonin at low concentrations. Moreover, we have recently localized serotonin in the lamprey spinal cord; using immunohistochemical techniques, we have observed serotonin-like immunofluorescence in ventromedial cell bodies and processes located diffusely throughout most of the fibre tracts in the cord (Filler et al. 1983; Harris-Warrick et al. 1985). Thus, serotonin is present and appropriately located to be a candidate neuromodulator of the CPG for swimming. Serotonin also probably affects many other aspects of normal spinal cord function that we have not yet monitored with our current experiments.

Ayers et al. (1983) have used cinematographic analysis to study the parameters of a family of related undulatory movements in the lamprey, including slow and fast swimming, burrowing, crawling and reverse crawling. These basically similar movements vary only in quantitative details of the period length and strength of motor output. Based upon these results, Ayers et al. (1983) have proposed that the CPG for fictive swimming may be in fact a general CPG for undulatory movements, and that modulation of the CPG by descending processes from the brain could generate all of the undulatory movements seen behaviourally. Aswe have described above, serotonin modulates both the frequency of VR burst discharge and the duration and strength of motoneurone firing during the burst. Such modulation of the CPG could transform the motor output from swimming to burrowing or other slow movements. However, serotonin also usually lengthened the intersegmental phase lag, which was remarkably constant for all the movements studied by Ayers et al. (1983). This discrepancy may arise from the non-physiological method of bath application of the amine over the whole cord in our in vitro experiments, as we have discussed above. A local synaptic release of serotonin could have a different effect on the animal’ s behaviour from that we have observed in our in vitro preparation. Alternatively, serotonin may be involved in modulation of the CPG to produce a rare undulatory movement that was not analysed by Ayers et al. In addition, the isolated spinal cord is not subject to the constraints of the intact animal: stretch-receptive neurones have been reported in the lateral margins of the spinal cord (Grillner, Williams & Lagerback, 1984), and these cells, as well as sensory receptors sensitive to deformation of the muscle and skin, are known to provide sensory feedback that influences the CPG (A. H. Cohen, personal observation). Such feedback may limit the possible variability of the intersegmental phase lag in vivo. We also do not know what constraints are placed on the intersegmental phase lag in situ by the elastic properties of the intact body wall, which is removed in our dissection.

In conclusion, we have shown that bath-applied serotonin can modulate the motor pattern generated by the central pattern generator for fictive swimming in the isolated lamprey spinal cord. In other work, we have shown that serotonin is present in neurones and processes in the spinal cord. These results suggest that serotonin, released at the appropriate sites and with appropriate timing within the spinal cord, might be one of the neuromodulators normally used to fine-tune the motor programme generated from this CPG to produce undulatory movement appropriate to the needs of the animal.

This study was supported by a Muscular Dystrophy Association Basic Research Grant and NIH Grant NS 17323 (to RH-W) and NIH Grant NS 16803 (to AC). We thank Philip Holmes and Richard Rand for important discussions, Margaret Baker for technical assistance, Ronald Hoy and Robert Flamm for their careful reading of the manuscript, M. Nelson for figure preparation and Barbara Seely for preparation of the manuscript.