Centrally Generated Rhythmic and Non-Rhythmic Behavioural Responses in Rana Temporaria Embryos

By the time of hatching, most amphibian embryos have developed some form of rhythmic locomotion to enable them to survive the start of free-swimming larval life. The simple patterns of swimming shown by the late embryos of two amphibians, the urodele Triturus (T. helveticus and T. vulgaris) and the anuran Xenopus laevis, have been analysed in some detail (Blight, 1976,1979; Kahn et al. 1982; Soffe et al. 1983). The aim has been to seek, in the nervous systems of these relatively simple animals, basic principles of operation for vertebrate locomotor rhythm generation. Patterns of locomotion in X. laevis and T. vulgaris embryos show strong parallels and both are centrally generated by neural elements situated within the spinal cord (Soffe et al. 1983; Roberts et al. 1985; for a review, see Roberts et al. 1986). Swimming movements in X. laevis and T. vulgaris occur at relatively high frequency, generally 10–30 Hz, and motor activity in both is characterised by motoneurones firing a single spike per cycle.

As a precursor to more detailed studies of motor pattern generation and selection, the present investigation was carried out to describe the main behavioural responses of the frog R. temporaria L. close to hatching and to examine the extent to which these, too, are centrally programmed. Interest in this amphibian was stimulated because its embryos are rather larger and swim with a lower beat frequency than those of X. laevis and T. vulgaris. Initial studies (Sillar and Soffe, 1989) suggested that swimming in R. temporaria embryos is driven, at least in part, by a pattern of rhythmic motor bursts more typical of vertebrate locomotor drive than the rhythmic single spikes seen in X. laevis and T. vulgaris (see also Stehouwer and Farel, 1980, for results using older Rana larvae).

To understand the neural basis of a behaviour pattern, it is important to have a knowledge of the movements involved. This provides a background against which to assess the different contributions of centrally generated neural programmes, sensory modulation and mechanical considerations. In the case of larval anurans, detailed studies have been carried out for both Rana and Xenopus (Wassersug and von Seckendorf Hoff, 1985; von Seckendorf Hoff and Wassersug, 1986). However, relatively few studies have been carried out on younger amphibians or, indeed, on smaller swimming vertebrates generally. For this, video recording now offers an approach by which relatively large amounts of data can be obtained with ease, in situations where the frame speed gives sufficient resolution (e.g. Batty, 1984). This approach has been adopted here to examine the early movements of R. temporaria embryos, including both sustained rhythmic lashing and swimming movements and the simple flexions that precede them in development. The extent of central motor programming for these behaviour patterns is then examined by making electrical recordings of motor responses in paralysed animals. This makes it possible to investigate which properties the motor programmes of R. temporaria embryos share with those of other amphibian embryos and older larvae and also with the locomotor patterns of adult vertebrates.

Embryos of the frog Rana temporaria were maintained in the laboratory at temperatures between 10 and 20°C until required. Since there appear to be no detailed normal tables for R. temporaria, animals were staged using the criteria of Shumway (1942) for R. pipiens and the simplified scheme of Gosner (1960). Animals were examined at total lengths between 7 and 10 mm (developmental stages 19 and 20). Embryos of this age are available between approximately February and April each year and the work described here and in the following paper (Soffe and Sillar, 1991) is based on studies performed between 1986 and, 1989.

Body movements were recorded using a NAC 200 high-speed video recorder at 200 frames s⁻¹. Animals were placed in a Petri dish (diameter 90 mm) containing tapwater maintained at approximately 20°C. The temperature in the recording chamber was monitored continuously by means of a thermistor probe. Animals were kept at the recording temperature for at least 2h before recordings were made. They were viewed from directly above and were lit from directly below using reflected stroboscopic lighting synchronised to the recorder frame rate.

Movements were evoked by a light touch on one side using a fine etched tungsten needle or, more rarely, by a light pinch of the tail fin. Stimuli were applied as far as possible at a level just sufficient to evoke a response and responses should be considered threshold except where stated otherwise. During swimming, animals were recorded moving across the centre of the dish. On reaching the edge, they sometimes stopped but often continued, following the curve of the edge of the dish. Measurements were made only on animals swimming freely in a straight line across the dish.

Some recordings were also made with a low-band U-matic video recorder at 50 frames s⁻¹. These records were used to obtain some data on cycle periods during whole episodes of swimming and also the form of slower movements. Analysis was carried out as above where appropriate.

For electrophysiological recording, animals were anaesthetised initially with MS222 and pinned on their sides onto a Sylgard block. An area of skin overlying the myotomes of the mid-trunk on the right-hand side was then removed using fine etched tungsten pins. Animals were kept immobilised by the addition of 70 μmol I⁻¹ 4-tubocurarine to the saline, the MS222 being washed off immediately after dissection. Extracellular recordings were made by means of glass suction pipettes (60–80μmi in diameter) applied to the intermyotome clefts either at, or near, the same level on the two sides or rostral and caudal on the same side. During experiments, animals were maintained in a bath of approximately 1.5 ml in volume and perfused at a rate of 5–7 ml min⁻¹ with saline. This contained the following (in mmoll⁻¹): NaCl, 105; KC1, 2.5; CaCl₂, 2 and NaHCO₃, 15, equilibrated to pH 7.2 by bubbling continuously with 5% CO₂, 95% O₂.

Statistical analysis was carried out using the Minitab software package.

Early movements in R. temporaria embryos occurred both spontaneously and in response to touch. Movements also occasionally occurred in response to dimming the lights, though with a longer latency. The first movements made by embryos during development were small twitches or larger flexions of the body to one side (Fig. 1). These apparently occurred spontaneously within the egg membranes but could also be evoked by lightly touching the skin of the trunk or head in animals released from their egg membranes. During the onset of each flexion, bending occurred maximally at the level of the trunk. In stronger flexions, this was often associated with an opposite bending of the tail and sometimes head, presumably due to recoil. Only during straightening from the flexion did a wave of bending pass along the tail. The stronger flexion responses involved a near-synchronous and maintained bending along one side (usually the opposite side in the case of a single-sided stimulus). This could involve just a small part of the trunk or, in the tightest flexions, the whole trunk and rostral part of the tail. Even in the strongest flexions, where the total angle of curve of the body could be as much as 325°, the caudal part of the tail showed little bending except as recoil to the initial trunk movement and during the last stage of straightening from the flexion.

When embryos were touched on one side, the first response was usually a bend to the opposite side (chi-square analysis, N=290, P<0·00;l) (Fig. 1). If the trunk skin was touched, this bend usually led directly to swimming (in those animals that had developed the ability to swim). In contrast, touching the head usually evoked a strong flexion that could be enough to turn the animal up to 180° away from the stimulus (Fig. 1C) before often leading to swimming.

It was not possible to determine the latency of a response to touch by direct observation or even from video analysis, in most cases, since the precise timing of each hand-held stimulus could not be clearly detected.

The first sustained rhythmic movements to appear were alternating tight flexions on each side of the body, with the body held maximally flexed for up to 1–2 s before the opposite flexion commenced. Once again, these movements could be seen to occur apparently spontaneously while the animals were still in their egg membranes, but could also be evoked in released animals. By the time of hatching, a clear head-to-tail progression could be seen and the extremes of flexion were held less long, leading to clearly alternating lashing movements.

At around the time of hatching, at total lengths of 8–9 mm, embryos became capable of rhythmic swimming.

Episodes of sustained rhythmic movements occurred both spontaneously and in response to a light touch on the skin of the trunk or head. These movements fell into a range of forms whose extremes were typified by slow alternating flexions at 1–6 Hz (cycle period 170–1000 ms) and higher-frequency movements of up to 10–12Hz (cycle period 80–100ms). At lotver frequencies, the amplitude of movements was greatest (see below) and there was no clear net forward progression. Movements at this extreme of the range are here termed ‘lashing’ (Fig. 2) to distinguish them from higher-frequency movements producing clear forward ‘swimming’ (Fig. 3). Other rhythmic movements fell between these extremes without obvious transitions from one form to another.

In response to a light touch, most episodes of rhythmic movement consisted purely of swimming; that is, movements in the higher frequency range and with a clear forward motion of the body. Some episodes, however, started with one or more cycles of lashing, which appeared to progress smoothly into swimming. Occasionally, a response consisted entirely of lashing. During sustained swimming, the cycle period usually remained fairly constant. Long episodes of unhindered swimming could not be examined because of the restrictions imposed by the dish (see Materials and methods). However, a few measurements made on animals that continued to swim around the edge of the dish showed that the cycle period remained fairly constant throughout longer episodes.

Analysis of sequences of sustained rhythmic movements showed a remarkable homogeneity of form over the whole frequency range, the movements differing largely in magnitude. Rhythmic movements, from slow lashing to fast swimming, all involved a travelling wave of bending that passed along the body from head to tail. By fitting sine functions to plots of angles of bending at intervals along the body (Fig. 4) several features were determined for sequences of movement analysed in detail. As with any curve fitting, results must be interpreted with caution; however, use of this method appears justified by the closeness of the fit of the theoretical curves to the data.

The patterns of phase delay along the body, obtained from parameter c, were similar at different cycle periods (Fig. 5A). Phase delay between points 0.1 L and 0.9 L for cycle periods between 88 and 197 ms was 0.54. This would correspond to a total phase delay of 0.68. However, phase delay computed from bending angles was not uniform along the body (Fig. 5A). Delay was largest between points along the trunk and rostral tail and lowest towards the head and caudal part of the tail. Extrapolating from phase delays between 0.4 L and 0.6 L gives a total phase delay of 0.96. During swimming it took approximately one cycle for a wave crest to pass along the body from snout to tail tip (Fig. 3), i.e. a phase delay of 1.0. The body therefore formed approximately one wavelength at all times, with the snout and tail tip, for example, showing equivalent degrees of lateral excursion at the same point in a cycle (see Fig. 5D).

During rhythmic movements, the maximum bending angle for each point was lowest rostrally and highest at about mid-animal, decreasing again towards the tip of the tail (Fig. 5B). During lashing, the maximum bending angle was generally higher than during swimming, though for the caudal part of the tail angles were similar.

Throughout the whole frequency range of rhythmic alternating body movements, lateral displacement of the body was smallest at the rostral end of the trunk, 0.2 body lengths from the snout (Fig. 5C). It increased both rostral and caudal to this level, reaching a maximum at the tip of the tail. Once again, the only clear difference between swimming and lashing was the much greater degree of lateral displacement during lashing, reaching nearly 1 body length at the tip of the tail.

For most points along the body, motion was forward throughout each cycle, despite lateral movement. The track followed by more-caudal points along the body, however, frequently had a component of backward movement (Fig. 5D). This was particularly noticeable at the tip of the tail. Specific forward velocity during swimming, calculated for the point of minimum lateral displacement 0.2 L from the snout, was up to 4.7Ls⁻¹. At the other extreme, during lashing there seemed to be no net movement in any one direction and forward velocity was, therefore, effectively zero.

Motor root discharge could be recorded from the intermyotome clefts in response to the same stimuli that evoked the behavioural responses outlined in the previous section. Mechanical stimulation was effective, and similar responses were also obtained using brief electrical stimulation of the skin. Dimming the illumination was again a less effective stimulus, being more unreliable and having a much longer latency to response.

The simplest motor root responses made by R. temporaria embryos paralysed with curare to a light touch or brief electrical stimulation of the trunk skin were single spikes. Responses were graded, with progressively stronger stimulation in the same region producing longer bursts of motor root discharge of up to about 2 s in duration (Fig. 6A). Usually, bursts started at the more rostral electrode before appearing caudally. Activity was never seen synchronously on the two sides at the same level in the spinal cord. Bursts of motor root discharge could also be evoked by a light touch or electrical stimulation of the head skin. As with behavioural responses to stimulation in this region, the first response in the curarised preparation was usually contralateral. Once more, the smallest responses could be restricted to a small part of one side, recorded at a single site. Stronger responses were more widespread. In some cases the initial contralateral response was followed by an ipsilateral burst or by briefly alternating activity (Fig. 6B). The long burst duration and alternating nature of such responses would make them appropriate to drive the alternate tight flexions seen behaviourally.

Episodes of rhythmic motor root discharge were recorded in response to a light touch to the skin of the trunk, tail or head or to electrical stimulation of any of these regions (Fig. 7). On a very few occasions, they were recorded in response to dimming the lights. These episodes could last for approximately 4 s (about 30 cycles), but more often lasted about 2s.

Episodes showed a range of forms (Figs 7, 8): cycle period could be relatively short and relatively constant throughout most of an episode; it could start relatively long and then progressively shorten, reaching a fairly constant level; it could progessively shorten without reaching a constant level by the time the episode finished; or it could remain relatively long throughout the episode. Very occasionally, progressive lengthening of cycle period was seen in the last two or three cycles of an episode (though see Soffe and Sillar, 1991). In all cases, discharge on the two sides alternated (Figs 7, 8). The phase, measured with respect to the start of each motor root burst, was around 0.5, but varied somewhat between cycles. In no case was activity seen synchronously on the two sides.

The motor root activity itself displayed a range of forms. Most commonly recorded were bursts of discharge that occupied up to 40–50% of the cycle period. This is consistent with motoneurones each firing a burst of impulses on each cycle. The duration of the motor bursts was correlated with cycle period for much of the range of periods (Fig. 9A). Towards the end of episodes, where cycle periods tended to be shorter, motoneurones were probably each able to fire only a single spike per cycle (see also Soffe and Sillar, 1991).

Motor root discharge recorded at more caudally placed electrodes always started later than that recorded at more rostral sites (Figs 7, 9). In two preparations where it was measured in detail, the longitudinal delay was correlated overall with cycle period (Fig. 9B), though it was rather variable between cycles. Values for longitudinal delay extrapolated to the length of the whole animal corresponded to mean total phase delays of 0.62 and 0.63 for these two embryos.

Stimulation of the skin on one side of the head usually evoked a large burst of discharge on the opposite side lasting 0.5-2.Os. In some cases this response could occur alone. Often, though, the initial response was followed by an episode of rhythmic discharge (Fig. 10). Such motor root responses would be appropriate to drive the contralateral flexions shown by freely moving animals (described above) that turn them away from stimuli to the head and that can also be followed by swimming.

In this study, video analysis and electrical recordings of motor root discharge have been used to examine the movements made by R. temporaria embryos at around the time of hatching, and their central nervous origin. These are the movements that allow the animal to survive the period immediately after hatching (and possibly assist with hatching itself). Movements start as simple flexions within the egg membranes and develop in complexity so that, by the time of hatching, embryos can make a variety of responses to stimuli. These responses include flexions and also sustained lashing and swimming movements suitable for escape and locomotion.

In animals paralysed with curare, it is possible to evoke patterns of motor root discharge that correspond to all the different behavioural responses. This means that not only simple rhythmic and non-rhythmic patterns, like swimming and flexing, but also more elaborate responses, like the turning and swimming that occur in response to touching the head, are largely centrally programmed. In view of the ubiquity of central patterning of locomotor behaviour, it would have been surprising if this had not been true, to some extent at least, for swimming in R. temporaria embryos. However, since rhythmic motor patterns in R. temporaria differ somewhat from those previously described for X. laevis and T. vulgaris (Kahn and Roberts, 1982; Kahn et al. 1982; Soffe et al. 1983), it was important to establish that they too were indeed centrally programmed. The degree to which each pattern of movement is determined by central mechanisms is, of course, harder to establish.

As far as major components of the rhythmic patterns are concerned, the important features appear to be centrally programmed. Cyclic rhythmicity, alternation between the sides and a longitudinal delay are all present in the paralysed animal and clearly, therefore, do not critically require sensory inflow. It has not yet proved possible to make electromyographic recordings in freely moving embryos so the temporal relationships between motor root discharge, muscle activation and movement of the body cannot yet be determined. Therefore, in the case of the longitudinal delay, for example, it is not yet possible to eliminate completely some role of sensory inflow or to determine the influence of mechanical factors. In older larval stages of R. catesbeiana and, presumably, at equivalent later stages of development in R. temporaria also, longitudinal delay requires sensory inflow through dorsal roots (Stehouwer and Farel, 1980). However, it is not known just how early in development this dependence becomes established.

Some general discrepancies suggest that the centrally programmed component of each behavioural response may not explain the whole behaviour. First, in the paralysed preparation, episodes of rhythmic activity are generally briefer than the behaviour itself, although long episodes can be evoked. Second, cycle periods may be longer in paralysed animals than in freely moving animals. Third, cycle periods are often more variable on a cycle-by-cycle basis in paralysed animals than during rhythmic behaviour. Although mechanical or other factors could be involved, these discrepancies could also point to an involvement of movement-related feedback by some as yet undiscovered route.

Perhaps the most striking difference between rhythmic motor patterns in R. temporaria embryos and those previously described for X. laevis embryos is the lack of any clear discontinuity between slower and faster responses. In X. laevis, there is an obvious difference between the slow rhythmic bursts of struggling (Kahn and Roberts, 1982) and the faster rhythmic single spikes of swimming. Changes between the two are exaggerated by a reversal in the longitudinal delay, which is from tail to head during struggling. In R. temporaria, there is a smooth gradation between lashing and swimming and the same head-to-tail direction of the delay is maintained throughout. Behaviourally, lashing grades into swimming as the strength of bending (angle and lateral displacement) declines and a clear forward movement of the body is established. Underlying this, relatively slow rhythmic bursts accelerate into shorter, faster rhythmic bursts or single spikes.

The mechanisms underlying generation of rhythmic motor discharge raise a number of interesting points. First, the pattern of rhythmic bursts underlying the slower end of the range of rhythmic movements is clearly self-sustaining. Although rhythmic bursts of discharge can be generated by a wide range of other vertebrate motor systems, there are few examples where sustained rhythm generation is not driven by either maintained sensory stimulation or some other maintained excitation, such as the use of pharmacological excitants or stimulation of descending pathways. McClellan (1984) has shown a few cycles of self-sustained rhythm in the lamprey and the in vitro central nervous system of R. catesbeiana larvae can also produce spontaneous and apparently self-sustaining episodes of rhythmic locomotor bursts (Stehouwer and Farel, 1980). Second, from the perspective of the behaviour of the animal, pre-programming of activity apparently extends beyond simple responses. For example, although relatively mild stimulation will evoke swimming, stronger stimulation evokes lashing that grades into swimming. This could allow an animal to break free from restraint (perhaps a predator) and then to swim off. Similarly, a touch to the head evokes an appropriate turn away from the stimulus and only then leads to swimming. In both cases, the whole response can be executed without further sensory input. A remarkably similar response to head skin stimulation has been descibed for T. vulgaris embryos (Soffe et al. 1983).

Motor responses in R. temporaria are evoked by a variety of sensory stimuli. No attempt was made here to classify the sensory systems involved beyond broad distinctions such as dimming the illumination or a light touch, either single-sided or bilateral. Responses to light dimming, which were much less reliable and of longer latency for R. temporaria than for X. laevis (Roberts, 1978), probably involve the pineal eye, as in X. laevis. The lateral eyes in R. temporaria are still developing and at this stage are covered by pigmented skin. Trunk stimulation in X. laevis acts through either Rohon-Beard neurones in the spinal cord (Clarke et al. 1984) or propagated skin impulses. Rohon-Beard neurones are present in the Rana spinal cord (Hughes, 1957) and the skin supports skin impulses (Roberts, 1971), so either of these possibilities also applies to R. temporaria. Although not clearly established, the pronounced sidedness of responses argues in favour of Rohon-Beard cell stimulation being the important route in mediating the responses to light touch described here. This is supported by the results of intracellular recording from spinal cord neurones (Soffe and Sillar, 1991). The head skin of R. temporaria embryos contains both rapid transient and movement type detectors (Roberts, 1980). In X. laevis, stimulation of the rapid transient detectors, like stimulation of the Rohon-Beard neurones of the trunk, turns on behaviour. Stimulation of the movement detectors, particularly those around the cement gland, appears to turn off behaviour. It is not yet possible to say whether this situation also holds true for R. temporaria.

This work was supported by an MRC grant. I should particularly like to thank Dr Nicholas Dale for writing the software used in the movement analysis. I should also like to acknowledge the SERC for loan of the NAC200 recorder and Dr P. Goodyer for instruction in its use. I thank Paul Court, Linda Teagle, Fiona Wyatt and Mark Shannon for expert technical assistance and Drs Nicholas Dale, Alan Roberts and Keith Sillar for advice on the manuscript.