Dorsal Spinal Interneurons Forming a Primitive, Cutaneous Sensory Pathway

Transcription

1 J Neurophysiol 92: , First published March 17, 2004; /jn Dorsal Spinal Interneurons Forming a Primitive, Cutaneous Sensory Pathway W.-C. Li, S. R. Soffe, and Alan Roberts School of Biological Sciences, University of Bristol, Bristol BS8 1UG, United Kingdom Submitted 8 January 2004; accepted in final form 10 March 2004 Li, W.-C., S. R. Soffe, and Alan Roberts. Dorsal spinal interneurons forming a primitive, cutaneous sensory pathway. J Neurophysiol 92: , First published March 17, 2004; / jn In mammals, sensory projection pathways are provided by just three classes of spinal interneuron that develop from the roof-plate. We asked whether similar sensory projection interneurons are present primitively in a developing lower vertebrate where function can be more readily studied. Using an immobilized Xenopus tadpole spinal cord preparation, we define the properties and connections of spinal sensory projection interneurons using whole cell patch recordings from single neurons or pairs, identified by dye filling. Dorsolateral interneurons lie in the tadpole equivalent of the spinal dorsal horn, and have dorsally located dendrites and an ipsilateral ascending axon that projects into the midbrain. These neurons receive direct, mainly AMPA receptor (AMPAR)-mediated, excitation from skin touch sensory neurons. They in turn produce AMPAR and N-methyl-D-aspartate receptor (NMDAR)-mediated excitation of spinal locomotor pattern generator neurons rostrally on the same side of the cord. During swimming, they receive glycinergic modulatory inhibition from ascending interneurons in the spinal locomotor central pattern generator. We conclude that spinal dorsolateral interneurons are a primitive class of excitatory sensory projection neurons activated by ipsilateral cutaneous afferents and carrying excitation ipsilaterally and rostrally as far as the midbrain to initiate or accelerate swimming. INTRODUCTION In vertebrates, sensory information from the body enters the dorsal spinal cord. What are the primitive sensory pathways that carry this information to the brain? The primary afferent axons can themselves carry this information for significant distances within the cord, but they also excite dorsal horn interneurons. Some of these are local, and others carry the information to distant parts of the spinal cord and to the brain. Recently, studies on the development of spinal cord neurons have raised hopes that the organization and different classes of spinal interneurons might soon become clearer (Goulding et al. 2002; Helms and Johnson 2003). In particular, three groups of interneurons whose development depends on the presence of the roof-plate seem to provide sensory projection pathways. Two of these groups project to the opposite side of the cord and then on to the brain, like the adult spino-thalamic tract, while one projects on the same side, like the adult spino-cervical tract (Brown 1981; Willis and Coggeshall 1991). Interneurons with similar projection patterns are present as soon as frog tadpoles emerge from the egg (Li et al. 2001). As a final step in the structural and physiological definition of spinal sensory pathways for initiation of swimming in the developing frog, we examined the properties, connections, and responses of dorsal spinal cord interneurons with ipsilateral ascending projections. Address for reprint requests and other correspondence: W.-C. Li, School of Biological Sciences, Univ. of Bristol, Woodland Rd., Bristol BS8 1UG, UK ( wenchang.li@bristol.ac.uk). The hatchling Xenopus tadpole (development stage 37/38) (Nieuwkoop and Faber 1956) can produce two main responses to trunk skin stimulation, even after spinalization. In the first, the tadpole bends its body away from the side where its tail skin is stroked and then swims forward (Boothby and Roberts 1995). In the second, the tadpole shows strong but slow rhythmic body flexions, called struggling, when it is held or the skin is stimulated repeatedly (Soffe 1991, 1993). The spinal cord has only eight anatomical classes of neuron (Fig. 1B), and the functions of six of these classes have been studied (Li et al. 2002; Roberts 2000). Functions for dorsolateral ascending interneurons (dlas) and Kolmer-Agdhur cells remain unknown. Two classes of possible sensory interneurons have been described (Fig. 1B; Clarke and Roberts 1984): dorsolateral commissural interneurons (dlcs) have dorsolateral somata and dorsal dendrites that receive inputs from skin sensory Rohon- Beard neurons (RB) axons and project axons to the contralateral side (Clarke et al. 1984; Roberts and Sillar 1990). They function as the relay neurons in the tadpole s disynaptic flexion reflex to contralateral skin stimulation (Li et al. 2003). Their activity can also increase swimming frequency when stimulation is applied during swimming (Sillar and Roberts 1988a). dlas are anatomically similar to dlcs, but their axons project rostrally on the same side, reaching the midbrain or forebrain (Clarke and Roberts 1984; Li et al. 2001). Are dlas also sensory projection interneurons? This study uses single and paired whole cell recordings to define the properties and responses of dla interneurons, to identify their input and output synaptic connections, and to determine their possible roles in the Xenopus tadpole s responses to skin stimulation. METHODS Details of the methods have been given recently (Li et al. 2002). Xenopus tadpoles at stage 37/38 (see Fig. 1A) were anesthetized with 0.1% MS-222 (3-aminobenzoic acid ester) and pinned in a small bath of saline (concentrations in mm: 115 NaCl, 3 KCl, 2 CaCl 2, 2.4 NaHCO 3, 1 MgCl 2, and 10 HEPES, adjusted with 5 M NaOH to ph 7.4). In many experiments, 1 mm MgCl 2 was replaced by 1 mm CaCl 2. The dorsal fin was cut, and the tadpole was transferred to 10 mm -bungarotoxin saline for immobilization (20 30 min). It was then re-pinned so skin and muscles over the right side of the spinal cord could be removed. A dorsal cut was made along the midline of the spinal cord to open the neurocoel and expose neuronal cell bodies. Some ependymal cells in the neurocoel were picked away to expose more ventral neurons. The animal was then re-pinned on a small rotatable Sylgard stage in a 700- l recording chamber that allowed brightfield illumination from below on an upright Nikon E600FN microscope. The animal was tilted to an angle that allowed exposed The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact /04 $5.00 Copyright 2004 The American Physiological Society 895

2 896 W.-C. LI, S. R. SOFFE, AND A. ROBERTS neuronal cell bodies on the left and right sides of the cord to be seen using a 40 water immersion lens. Saline in the chamber was kept circulating at about 2 ml/min and was not oxygenated. Drops of antagonists were added to a 100- l chamber upstream to the recording chamber (bath application) or applied close to the neuron being recorded by applying gentle pressure to solution in a pipette with tip diameter of m (microperfusion). NBQX, D-AP5, and bicuculline were from Tocris Cookson; strychnine and all other chemicals were from Sigma unless stated otherwise. Patch pipettes were filled with 0.1% neurobiotin in intracellular solution (concentrations in mm: 100 K-gluconate, 2 MgCl 2, 10 EGTA, 10 HEPES, 3 Na 2 ATP, and 0.5 NaGTP, adjusted to ph 7.3 with KOH) and had resistances around 10 M. In some experiments, 0.1% Alexa Fluor 488 (Molecular Probes) was also used in the patch pipette solution to identify dlas in live tadpoles. Stimulating suction electrodes were placed on head and tail skin to apply 1-ms duration current pulses to start fictive swimming or struggling activity. Another suction electrode was usually placed on the 11th intermyotome cleft to record ventral root activity. Patch pipettes were manipulated under visual control to contact exposed neuron somata (Fig. 1A). Light positive pressure was always applied to the pipette solution before trying to get a seal. Signals were recorded with an Axoclamp 2B in conventional bridge or continuous single electrode voltage-clamp mode. Data were acquired with Signal software through a CED 1401 Plus (CED, Cambridge, UK) with sampling rate of 10 khz. Stimuli to the skin were controlled using the CED 1401 Plus configured by Signal and given via an optically coupled isolator. After the recording, tadpoles were fixed overnight in 2% glutaraldehyde in 0.1 M phosphate buffer (ph 7.2 at 4 C) and washed with 0.1 M PBS (120 mm NaCl in 0.1 M phosphate buffer ph 7.2). The animals were 1) washed twice in 1% triton X-100 in PBS for 15 min; 2) incubated in a 1:300 dilution of extravidin peroxidase conjugate in PBS containing 0.5% Triton X-100 for 2 3 h; 3) washed again in PBS; 4) presoaked in 0.08% diaminobenzidine in PBS (DAB solution) for 5 min; 5) moved to 0.075% hydrogen peroxide in DAB solution for 5 min; and 6) washed in running tap water. The brain and spinal cord were dissected free with the notochord and some ventral muscles, dehydrated, cleared in methyl benzoate and xylene, and mounted whole, between two coverslips using Depex (BDH Laboratory Supplies). Neurons filled with neurobiotin staining were observed using a 100 oil immersion lens and traced using a drawing tube. To compensate for shrinkage during dehydration, all neurobiotin staining measurements in this paper have been corrected by multiplying by 1.28 (Li et al. 2001). Fluorescent images were acquired using a Penguin 150 CLM camera (Pixera) with B-2A filter. All means are given with their SD unless otherwise stated. Off-line analyses were made with Minitab and Excel. Photos were processed with Photoshop. RESULTS Properties of dlas Recordings from 320 interneurons showed the short latency responses to skin stimulation that would be expected in sensory pathway interneurons (typically 3 8 ms depending on the spacing between stimulating and recording electrodes). Of these, 48 had the anatomical features of dlas defined by horseradish peroxidase backfilling and intracellular neurobiotin staining (Clarke and Roberts 1984; Li et al. 2001): a dorsolaterally located, usually multipolar soma, dorsolateral dendrites, and an ipsilateral axon projecting toward the brain (Figs. 3 8). The anatomical features used to identify the neurons were seen either during the experiment using fluorescent imaging of the Alexa Fluor 488 or after the experiment following processing to visualize neurobiotin. The recorded FIG. 1. Diagram of the experimental set-up and the tadpole spinal sensory pathways. A: experimental set-up. Bottom: Xenopus tadpole at stage 37/38 with some muscle segments numbered (1, 5, 10, 15). Top: part of the trunk and some skin and muscles were removed to expose the spinal cord. Stimulating electrodes, microperfusion pipette, and whole cell patchclamp recording pipettes were arranged as shown. The ventral root recording electrode (VR) was usually placed between the 11th and 12th muscle segments. B: diagram of spinal neurons and sensory pathways shown in a segment of spinal cord cut-open dorsally. RB, Rohon-Beard neuron; dla, dorsolateral ascending interneuron; dlc, dorsolateral commissural interneuron; KA, Kolmer-Agduhr cell; mn, motoneuron; din, descending interneuron, cin, commissural interneuron; ain, ascending interneuron. The last 4 types of neurons are central pattern generator (CPG) neurons. The dlas and dlcs are sensory pathway interneurons. Neurons with dark shading are inhibitory and the others are excitatory. Triangles represent glutamatergic synapses. Circles represent glycinergic synapses. Hollow symbols (RB to dla, ain to dla, dla to CPG) are synaptic connections to be established in this study. Axons from mn, din, cin, and KA are omitted for simplicity.

3 DEFINING SPINAL PROJECTION INTERNEURONS 897 dlas had somata in the mid-trunk region at longitudinal positions between the 4th and 11th postotic muscle segments ( mm from the midbrain) with a peak in numbers at the 7th segment (1.7 mm from the midbrain). The average resting membrane potential for dlas was mv (n 44). Responses to current injection were analyzed in 13 dla neurons. These all showed rectification in their voltage responses to current steps below the neuronal firing threshold (Fig. 2, A and B). The inflection point in the I/V curves was slightly more positive than the resting membrane potential ( mv; P 0.05, 2-tailed paired t-test) and was estimated from the crossing point of regression lines fitted to the two parts of each curve. When the membrane potential was deploarized to a level close to firing thresholds, most dlas responses had an early sag (12/13, Fig. 2A) suggestive of the activation of a voltage gated potassium current. Average input resistances measured from the regression lines using positive and negative current steps were and 1, M, respectively. The action potential threshold was first estimated in 11 dlas at mv from the maximum voltage reached during the largest subthreshold current injection, and second, at mv in 11 dlas from the peak value of the biggest subthreshold excitatory postsynaptic potentials (EPSPs) FIG. 2. Responses of dlas to current injection. A: dlas voltage responses to a series of current pulses injected via the recording patch pipette. *Position of the early sag of the voltage response. B: I-V curves of 3 dlas below the neuronal firing threshold. Thicker line is for the neuron shown in A. Regression lines give estimates of cellular input resistances of 389 M to positive ( ) and 807 M to negative ( ) current pulses. The 2 regression lines cross at 58.9 mv (resting membrane potential is 60.3 mv). C: the same dlas firing to 3 different levels of current injection. D: the same dlas firing frequency changes vs. time to 7 rectangular positive current pulses ( pa; E, ). Note duration of firing increases with current intensity and firing frequency is always highest at the start of current injection. E: plots for 2 dlas of the highest firing frequency to current injection against current intensity. Diamonds are measurements from the same dla as in A, C, and D. For both logarithmic regression lines R Neuron s anatomy is shown in Fig. 5C. evoked by skin stimulation (see next section). When current was just above threshold, dlas only fired one or a few spikes at the beginning of current injection. When the current was increased, they fired a train of spikes with decreasing amplitude and frequency that could stop before the end of the 500-ms current pulse (Fig. 2, C and D). The relation between the highest instantaneous frequency (in most cases between the 1st 2 spikes in the train) and the current intensity could be fitted by a logarithmic function (Fig. 2E). dlas are excited by skin stimulation and inhibited during swimming and struggling The Xenopus tadpole at stage 37/38 responds to skin stimulation by producing two types of locomotion. When the skin is stimulated by a brief touch or single electric stimulus, neurons on both sides of the spinal cord are excited (Zhao et al. 1998), and the animal normally bends its body away from the touch and starts swimming forward with alternating contractions at Hz (Roberts 1990). During fictive swimming, ventral roots show brief, tightly synchronized bursts of spikes (Fig. 3Aa) that alternate on left and right sides. Sustained pressure on the skin or repetitive electrical stimuli usually led to struggling with slower but stronger rhythmic contractions at 5 10 Hz (Soffe 1991). During fictive struggling, ventral root bursts are much longer (Fig. 3Ba). We therefore stimulated the skin and looked at the responses of dlas and their input during these two locomotion patterns. When the tail skin was stimulated, all dlas tested received a strong compound EPSP (n 40, Figs. 3, 5, 7C, and 10A). For stimuli less than twice threshold, these EPSPs could lead to a single action potential in most dlas (n 36). The latencies measured for 10 EPSPs in each of 10 dlas were short and very consistent, (ranges, and ms). In a few cases (n 5), this initial excitation was followed by an inhibitory postsynaptic potential (IPSP; Fig. 3Ab). The shortest latencies for these IPSPs were ms after stimulation but before swimming started. During swimming, all 42 dlas tested were inhibited by IPSPs loosely in phase with the ventral root activity recorded on the same side (Figs. 3Aa, 6,C and D; we define this timing as early-cycle ). When struggling was elicited by skin stimulation, all 22 dlas tested were also inhibited by IPSPs in phase with ventral root bursts (Fig. 3B). These results suggest that dlas are only active briefly prior to the initiation of swimming or struggling. Having shown that dlas are excited following skin stimulation and receive inhibition during swimming and struggling, paired recording was used to trace the neuronal sources of these synaptic inputs. Sensory RB neurons directly excite dlas The short latency responses of dlas to skin stimulation implied that they were directly excited by sensory RB neurons. In 370 simultaneous recordings from pairs of neurons of a variety of types on the same side of the spinal cord, 16 were found to be between RB neurons and dla interneurons. The RB neurons were identified by their large dorsal soma without dendrites and ascending and descending axons in the dorsolat-

4 898 W.-C. LI, S. R. SOFFE, AND A. ROBERTS FIG. 3. dla responses to skin stimulation and inhibition during fictive swimming and struggling. A: a: electrically stimulating the tail skin (arrow) excites the dla and evokes an action potential (seen expanded in b). After the action potential, swimming activity starts (*) in the ventral root record (vr). During swimming, the dla is rhythmically inhibited by early-cycle IPSPs (arrowheads in a). b: in this dla, the action potential is followed by an inhibitory postsynaptic potential (IPSP; arrowhead) at a latency of 14.2 ms from the stimulus. In a and b, the membrane potential is depolarized from 58 to 38 mv to reverse otherwise depolarizing IPSPs. c: dla responds to repeated skin stimulation with short latency excitatory postsynaptic potentials (EPSPs) with action potentials on 3 of 10 overlapped traces. d: photograph from the right side of the spinal cord to show the dla in a c labeled by neurobiotin, with its ascending axon (arrowhead) and dendrites (e.g., at arrow). e: tracing of the CNS from the right side showing the location of the dla soma and its axon projecting through the spinal cord and hindbrain to the midbrain. B: a: inhibition of a dla during struggling seen as bursts of spikes in the ventral root (gray bar below vr) and started unusually by a single current pulse to the tail skin (arrow). Membrane potential is depolarized from 65 to 38 mv to show the timing of inhibition (arrowheads). b: fluorescence image of the dla in a (2 focus planes combined) showing the ascending axon (arrowhead), dendrites (e.g., at arrow), and whole cell recording pipette (*). c: tracing of the CNS to show the same dla in a and b, revealed by neurobiotin staining, with its soma and its axon projecting to the midbrain. Both dlas in A and B are in the right side spinal cord and viewed from the right. eral tract. In six pairs, action potentials resulting from current injection into the RB neuron produced EPSPs in the dla at short latencies ( ms; Fig. 4A), suggesting that the interaction was direct and monosynaptic. These EPSPs were large ( mv; maximum amplitude, mv), had fast rise-times (10 90% in ms; time to peak, ms), and were of short duration ( ms at 50% amplitude; although most had a long, low-amplitude tail when recorded in 0 mm Mg 2 ). The neurons anatomy, revealed by neurobiotin or fluorescent dye injection, confirmed the identity of the pre- and postsynaptic neurons and showed that RB neuron axons came into close contact with dla dendrites or somata (n 4, Fig. 4B). The pharmacology of the RB to dla synapses was examined using glutamate receptor antagonists. NBQX (n 6) and D-AP5 (n 4) were applied separately or together (n 3) to look at effects on EPSPs produced in dlas by tail skin stimulation. In each case, the EPSP consisted of an early AMPA receptor (AMPAR) component, blocked by NBQX, and a slower-rising, longer N-methyl-D-aspartate receptor (NMDAR) component, blocked by D-AP5. EPSPs were abolished during combined application (Fig. 5). In the only dla recorded in voltage-clamp mode, the APMAR component appeared to be predominant (Fig. 5Bc), but a detailed comparison between the sizes of different glutamatergic receptor current components was not made. This recording and one current-clamp recording showed that Mg 2 was effective in inactivating NMDARs at the resting membrane potential level (Fig. 5, A and Bb). The short latency excitation of dlas following skin stimulation and the paired recordings with RB neurons show that dla interneurons are directly excited through the activation of AMPARs and NMDARs following impulses in skin sensory RB neurons. dlas are therefore likely to be sensory pathway interneurons. dlas are inhibited by premotor ascending interneurons In six paired whole cell recordings of premotor ascending interneurons (ains) and dlas, three of the ains inhibited the postsynaptic dla (Figs. 6 and 7). In the 42 IPSPs from these recordings, the durations ( ms at 50% peak height) and the rise times ( ms from 10 90% of peak height) were similar to those in dlcs, another type of sensory interneuron, which are inhibited by ains when the tadpole swims (Li et al. 2002, 2003). The latencies ( ms from the peak of the presynaptic spike), like the values for RB dla interactions, FIG. 4. Sensory RB neurons directly excite dla interneurons. A: a: currentevoked spikes in a RB produce EPSPs at short and consistent latencies in a dla (6 overlapped EPSP traces and 1 failure; failure rate 22%.). b: same pair showing 3 EPSPs at a slower time scale. B: a: scale drawing of the RB and dla in A to show that the RB ascending axon could contact dla dendrites (near arrow). b: large scale drawing shows neuron positions on the right side of the spinal cord and the RB axon projection. The dla axon is not shown (*) to simplify illustration.

5 DEFINING SPINAL PROJECTION INTERNEURONS 899 ventral root bursts in each normalized swimming cycle (668 spikes from 16 ains and 691 IPSPs from 9 dlas). The expected longitudinal delays between the positions of recorded neurons and the ventral root electrode were compensated using a rostro-caudal delay of 3.5 ms/mm (Li et al. 2002). The correlation between the two distributions is significant (P 0.001, Pearson). Since most of the other neurons active during swimming fire in a tightly synchronized burst and earlier in the swimming cycle than ains (Li et al. 2002), this suggests that Downloaded from FIG. 5. Effects of glutamate receptor antagonists and Mg 2 on dla interneuron responses to skin stimulation. A: a single tail skin stimulation (arrow) usually leads to a dla spike riding on an EPSP (spikes go off scale). Effects of bath-applied 5 M NBQX, 50 M D-AP5, 5 M NBQX plus 50 M D-AP5, and 1 mm Mg 2 are shown using 3 superimposed traces for each case. B: a: another dlas responses to tail skin stimulation (arrow) in current-clamp mode. b: in voltage clamp, averages of 5 EPSCs at a holding potential of 65 mv in response to skin stimulation in Mg 2 -free saline (control, wash). Bath-application of 0.2 mm Mg 2 abolishes the slow NMDAR component. c: same control record as in b at a different scale to show the large size of the fast AMPAR component (arrowhead) relative to the slow NMDAR component (arrow). Dotted line indicates control holding current level. C: drawing of the anatomy of the dla in A at high and low powers. The dla axon appears to be broken (at arrowhead). Viewed from the right and the dla is in the right side spinal cord. all lay within the range found in previous cases where the connection was considered to be direct (see Li et al. 2002, 2003). In addition, the anatomy in all three pairs revealed possible contact sites between presynaptic ain axons and dla dendrites or soma (Figs. 6B and 7B). During swimming, some ain spikes in paired recordings seemed to correlate closely with dla IPSPs (Fig. 6C; n 3 pairs). We therefore compared the timing of ain action potentials and dla IPSPs during swimming (Fig. 6D). For ain spikes or dla IPSPs, phases were calculated by their timings relative to FIG. 6. Ascending interneurons (ains) produce inhibition of dlas. A: a: 10 superimposed traces show the short and consistent latencies of IPSPs in a dla produced by current evoked spikes in an ain. b: ain spikes evoked by repeated positive current pulses produce reliable IPSPs in the dla (3 traces superimposed). Voltage scale in a also applies in b. B: anatomy of the ain and dla in A. Both neurons are in the left side of the spinal cord. The ain was identified by fluorescent imaging but its soma was broken while withdrawing the electrode. a: the descending ain axon makes close contact with dla dendrites (arrowheads). The dla axon ascends toward the hindbrain. Arrows with dotted lines indicate axon projection directions. b: large scale drawing of the CNS viewed from the left side to show the dla soma and the position of the ain soma (dotted circle) and its descending axons, 1 of which passes the dla. The dla axon was omitted for simplicity (*). C: ain spikes and dla IPSPs (*) during swimming (vr). These are the same neurons as in A and B. Latency between the ain spike and the dla IPSP at the dotted line is 3.0 ms. This falls into the 95% confidence limits of latencies calculated from interactions in A ( ms). In A and C, the dla was deploarized to make the IPSPs clear. D: to compare the timing of ain spikes and dla IPSPs their occurrence is plotted as a function of swimming cycle phase as defined by the arrow on the vr record in C. by on November 30, 2017

6 900 W.-C. LI, S. R. SOFFE, AND A. ROBERTS bicuculline had no effect (n 1). We conclude that the transmission from ain to dla is probably glycinergic and that ains produce glycinergic inhibition of dlas during swimming. dlas excite rostral central pattern generator neurons FIG. 7. Inhibition of dlas by ains is blocked by strychnine, but not bicuculline. A: In a paired recording, IPSPs produced in a dla by ain spikes (dotted line) were blocked by bath application of 5 M strychnine but little affected by 10 M bicuculline. B: a: high power drawing of the ain filled by neurobiotin and a reconstructed fluorescence photo of the dla (from 4 focus planes, the soma is out of focus); *recording electrode for dla; arrowheads, the descending ain axon passing dla dendrites. Arrows with short dotted lines indicate the direction each axon projects. b: large scale drawing of the CNS and the neurons in A. The dla soma (dotted circle) broke while withdrawing the electrode and its axon is omitted for simplicity. Neurons in the left side of the spinal cord and viewed from the left. C: bath application of 5 M strychnine blocks IPSPs (arrowheads) in another dla (same neuron as Fig. 3A) during swimming (vr) started by tail skin stimulation (arrow). Initial dla spikes are truncated. All antagonists were bath-applied and wash times were 4 min. most of the IPSPs that dlas receive during swimming come from ains. During struggling, ains are rhythmically active in phase with ventral root activity, again at the time when dlas receive inhibition. They may be the source of this inhibition too, but detailed correlation between ain spikes and dla IPSPs was not attempted during struggling because it is not possible to compare the phases of these events with sufficient precision. Bath application of the glycine antagonist strychnine blocked the inhibition from ains onto dlas in two paired recordings, and when one of these was tested further, the GABA A antagonist bicuculline did not have any obvious effect (Fig. 7A). During swimming, the rhythmic IPSPs in dlas were also blocked by strychnine (Fig. 7C, n 6), and again PAIRED RECORDINGS. Having found the main sources of synaptic inputs to dlas, excitation from sensory RBs and inhibition from ains, we investigated their output connections. dla cell bodies are mainly located in the mid-trunk region of the spinal cord between 1.25 and 2 mm (segments 5 9) from the midbrain (Li et al. 2001). Their axons project rostrally in the dorsal half of the spinal cord, but are often quite wavy, so they could contact dendrites in a wide range of dorso-ventral positions. In paired recordings from dlas and 17 more rostral ipsilateral CPG neurons (9 cins, 6 ains, 2 dins), current-evoked dla spikes excited two ains and one cin. The two ains also made reciprocal inhibitory synapses back onto the recorded dlas. An example of a paired recording with reliable interaction between a dla and an ain is shown in Fig. 8. The mean latencies from the peak of the dla spike to the start of the CPG neuron EPSPs were between 1.4 and 1.8 ms (3 neurons, 12 EPSPs). This indicates direct, monosynaptic connections from dlas to these CPG neurons (Li et al. 2003). Because the total number of dlas is small (Li et al. 2001), it has been difficult to get paired recordings with interaction between dlas and ipsilateral CPG neurons. To identify the neuronal types that may receive dla excitation and investigate the pharmacology of this excitation, we analyzed CPG neuron responses to ipsilateral tail skin stimulation below swimming threshold. We looked for evidence that CPG neurons are excited di-synaptically via dlas (sensory RB neurons 3 dlas 3 CPG neurons). FIG. 8. A dla interneuron excites an ipsilateral rostral ain interneuron. A: a: dla spikes evoked by 100-pA current injection produced EPSPs in an ain. Four traces are overlapped: 2 show EPSPs, in 1 the EPSP fails, and in 1 the EPSP sums with 30-pA injected current to produce a spike. b: 7 more overlapped traces at smaller time scale to show that the latencies of EPSPs are short and consistent. B: scale drawings of the same dla and ain after neurobiotin staining. The dla axon (arrow) appears swollen where it leaves the soma. It ascends to pass the ain with a possible contact site onto an ain dendrite (arrowhead). ain axon is not drawn (*) to simplify illustration. Neurons in the right side spinal cord and viewed from right.

7 DEFINING SPINAL PROJECTION INTERNEURONS 901 INDIRECT EVIDENCE FOR DLA OUTPUT CONNECTIONS. Following skin stimulation, sensory pathway interneurons receive reliable, large, short-latency, monosynaptic EPSPs from RB neurons that usually lead to an action potential (Clarke and Roberts 1984; Roberts and Sillar 1990; Sillar and Roberts 1988a; Figs. 3 5). These EPSPs are glutamatergic (Sillar and Roberts 1988a), with a strong AMPAR-mediated component (Li et al. 2003) (Fig. 5). As a result, when 5 M NBQX was applied in the bathing saline onto 12 dlcs and 5 dlas, EPSPs were strongly attenuated so that firing in response to skin stimulation was abolished in 14 of them (88%, Fig. 5A). The spikes produced by NMDAR-mediated EPSPs in the other two dlcs and one dla were unreliable and had long and variable latencies. In NBQX, therefore, neither dlcs nor dlas would be able to produce reliable, short latency synaptic excitation of CPG neurons. By applying NBQX globally in the saline, we therefore predicted that disynaptic excitation of CPG neurons in response to ipsilateral skin stimulation would disappear as a result of the failure in spiking of dla sensory interneurons (Fig. 9B). We tested the effect of bath-applied 5 M NBQX on EPSPs produced in 25 CPG neurons by ipsilateral tail skin stimulation below swimming threshold. In 14 neurons, slow-rising (presumed NMDAR) EPSPs remained during NBQX application. The latencies of these EPSPs were short (5 8 ms) and consistent, like those seen in dlas following skin stimulation. They were therefore probably produced by direct connections from sensory RB neurons and are not considered further here. In the remaining five ains and six cins, EPSPs could be totally blocked by NBQX. These EPSPs were therefore disynaptic and presumed to be from ipsilateral dla sensory interneurons. To confirm that the disynaptic EPSPs following skin stimulation in the 11 CPG neurons resulted from dla spikes, we checked that the latencies of these two events occurred in a FIG. 9. Pharmacology of EPSPs in a CPG neuron (cin) following ipsilateral, subthreshold skin stimulation (at arrow). A: a: diagram shows that with microperfusion of antagonists, excitation and firing of dlas remains, so we expect to see residual EPSPs in the cin. b: microperfused 5 M NBQX and 50 M D-AP5 partially block EPSPs, but 5 M NBQX 50 M D-AP5 abolishes the EPSPs. B: a: diagram shows that with bath application of antagonists, disynaptic EPSPs in the cin are expected to disappear due to failure in spiking in the relaying dla neurons. b: in the same cin as in A, bath-applied 5 M NBQX totally blocks cin EPSPs to skin stimulation. FIG. 10. Comparing the latencies of dla spikes and CPG neuron EPSPs evoked by ipsilateral tail skin stimulation. A: top traces are EPSPs in a cin following skin stimulation (at arrow, 10 traces overlapped). Spikes are off scale. Bottom traces are dla action potentials to skin stimulation (10 traces overlapped) recorded in another animal. B: a combined histogram summarizes and compares the distribution of latencies for spikes in 11 dlas and EPSPs in 11 CPG INs. C: distribution of EPSP latencies from individual CPG neurons. Numbers of EPSPs are 77 for the cin, 50 for the ain, 90 for the din, and 70 for the mn. The illustrated din neuron also receives reliable EPSPs at shorter latencies that are probably from RB neurons (early hatched bar truncated). The gray stripe marks the latency range from 7 to 13 ms in which EPSPs from dlas are most likely to occur. strict sequence. We compared the distribution of latencies of EPSPs (n 140 EPSPs, 10 stimuli in each of the 11 neurons) and dla spikes (n 126 spikes, 10 stimuli in each of 11 dlas tested, Fig. 10A) produced by ipsilateral tail skin stimulation. The combined distribution histograms for both dla spikes and CPG EPSPs (Fig. 10B) had very similar skewed shapes, and as expected, CPG EPSPs were slightly delayed relative to dla spikes (median value 9.5 vs. 6.8 ms). After correcting for this delay by subtracting the difference in median values between the two distributions from the EPSP latency values, the two latency distributions were highly significantly correlated (coefficient 0.796, P 0.001). The difference in median latencies between the distributions (2.7 ms) was also very similar to the latencies measured between dla spikes and CPG neuron EPSPs in paired recordings: a combination of synaptic delay and a conduction delay resulting from the more caudal recording positions of dlas compared with CPG neurons. We produced EPSP latency distribution histograms for 14 individual CPG neurons where a sufficient number of skin stimuli were delivered. If the EPSPs were from dlas, their latency distribution should be very similar to the combined distribution described above, produced by using EPSPs from many neurons (hatched bars in Fig. 10B). This was the case for 1/2 mns, 2/2 dins, 1/3 ains, and 4/7 cins (3cINs were from the above 11 CPG neurons; Fig. 10C). Together, these results give us confidence that the EPSPs in CPG neurons having a latency of around 10 ms in response to skin stimulation come from ipsilateral dlas and that dlas can excite all types of CPG neurons found in the tadpole spinal cord swim circuitry. We studied the pharmacology of the 11 cases of disynaptic EPSPs, presumed to be from dlas, using very local application of glutamatergic antagonists. Microperfusion was used to min-

8 902 W.-C. LI, S. R. SOFFE, AND A. ROBERTS imize the risk of a direct effect of the drugs on RB to dla transmission and therefore dla firing (Fig. 9Aa). In each case, microperfusion of either NBQX or D-AP5 could only partially block the EPSPs. When they were applied together, however, the EPSPs were totally blocked (Fig. 9Ab). We conclude that after being excited to fire by RB sensory neurons, dla interneurons excite ipsilateral CPG neurons located more rostrally. This excitation is mediated by both NMDARs and AMPARs. Role for ipsilateral sensory ascending excitation in swimming When the tail skin is stimulated with a brief current pulse ( 1 ms), a few RB sensory neuron peripheral neurites are excited (Clarke et al. 1984; Soffe 1997). The action potentials propagate via the soma to the central, longitudinal RB axons that excite dlc and dla interneurons to fire action potentials. Despite their ipsilateral excitatory connections, dlas do not appear to mediate ipsilateral reflex responses, and previous work suggests that this is partly the result of early ipsilateral inhibition following skin stimulation (Roberts et al. 1984; Zhao et al. 1998). After some period of time, depending on the stimulation intensity ( ms, 12 tadpoles, data not shown), swimming starts (Figs. 3A and 7C). Direct evidence suggests that dlas play a role in the initiation of swimming activity. In one case, current induced multiple firing of a single dla was capable of starting fictive swimming (4 observations, Fig. 11Aa). This phenomenon was seen more often in artificial 0 Mg 2 saline (7 observations in 4 dlas, data not shown) where magnesium s blocking effect on NMDARs was absent. A feature of tadpole swimming is that frequency increases if skin stimulation is given at certain phases of the swimming cycle (Sillar and Roberts 1988a, 1993). Recruitment of silent contralateral CPG neurons by dlcs was proposed to be the mechanism for this increase in frequency (Sillar and Roberts 1992b, 1993). Like dlcs, dlas receive early-cycle IPSPs from ains during swimming (Figs. 3A and 6, C and D), and can also be excited to fire spikes by skin stimulation during swimming (Fig. 11B). Excitation from dlas could be important in recruiting silent ipsilateral CPG neurons when skin stimulation is delivered during swimming. For example, in one dla recorded in 1 mm Mg 2 saline, currentinduced multiple firing could result in faster swimming (Fig. 11Ab). We therefore looked at the responses of eight cins and two ains that fired unreliably during swimming to see if they could be recruited by ipsilateral skin stimulation during swimming. Nine of these neurons were reliably recruited by a skin stimulus given during swimming. Analysis of the latencies of EPSPs in these neurons following skin stimulation suggested that EPSPs with shorter and consistent latencies ( 5 7 ms) were produced directly by RB neurons, while those with longer and more variable latencies (7 13 ms) were most likely produced by dlas (Figs. 10C and 11, C and D). We conclude that dlas are a component of the swim initiation pathway and can also contribute to speeding up swimming when stimuli are given during swimming. FIG. 11. Possible roles for dlas in initiating swimming and recruiting silent neurons during swimming after skin stimulation. A: a: positive current injection produces multiple firing of a left side dla (l.dla) and initiates fictive swimming recorded from a right side ventral root (r.vr). b: current-induced multiple firing of the same dla leads to an immediate increase in swimming frequency. Gray arrows indicate the positions where ventral root bursts would be expected if the swimming frequency was maintained at its prestimulus level. B: a right side dla (r.dla) is excited to fire a spike by ipsilateral skin stimulation applied during swimming (arrow), which also leads to faster swimming activity (r.vr). C: a: ain EPSPs in response to ipsilateral tail skin stimulation (arrow) below swimming threshold (7 traces overlapped), 2 with later IPSPs (*). Gray bar here and in D shows the expected timing of direct input from dla interneurons (Fig. 10C, 7 13 ms). Membrane potential was depolarized to discriminate IPSPs (the latency distribution histogram for EPSPs in this ain is given in Fig. 10C). b: ain is recruited to fire 6 spikes on 5 swimming cycles by ipsilateral tail skin stimulation (arrow) at the same current level as that used in a during swimming. D: a: EPSPs in a cin after subthreshold, ipsilateral skin stimulation (arrow;11 traces overlapped, 2 with truncated spikes). EPSPs with longer latencies (arrowhead) are likely to be from dlas and other earlier ones from RB neurons. Membrane potential is depolarized. b: cin is recruited to fire spikes after ipsilateral skin stimulation during swimming (same level as in a, arrow). DISCUSSION dla interneurons form an ipsilateral cutaneous sensory pathway We provide the first evidence on the responses and physiology of Xenopus tadpole dorsolateral ascending interneurons (dlas). Both the anatomical features of dye filled neurons and their location, in a longitudinal region of the spinal cord mm from the midbrain (muscle segments 4 10; cf. Fig. 9B in Li et al. 2001), matched those of dlas defined previously (Li et al. 2001; Roberts and Clarke 1982). This evidence gives us confidence that our recordings were from a single class of spinal interneuron. dla interneurons are excited by skin stimulation and form an ipsilateral cutaneous sensory projection pathway. They show strong parallels with dorsolateral commissural interneurons (dlcs), which we previously identified as the major crossed sensory projection pathway in the developing Xenopus spinal cord (Clarke and Roberts 1984; Li et al. 2002; Roberts and Sillar 1990; Sillar and Roberts 1988b, 1992a). Both types of interneurons 1) receive strong, direct, mainly AMPAR-mediated glutamatergic excitation that can lead to action potentials, from cutaneous sensory RB neurons; 2) receive glycinergic

9 DEFINING SPINAL PROJECTION INTERNEURONS 903 inhibition early on each swimming cycle from premotor ains and do not fire during swimming; 3) are inhibited and do not fire during struggling; and 4) make excitatory glutamatergic synapses onto spinal neurons that are components of the central pattern generator for swimming. Surprisingly, dlas have a significantly higher input resistance than dlcs (dla, M ; dlc, M ; P 0.04; W.-C. Li, unpublished whole cell recordings from dlcs). Despite this, both types of sensory interneurons share similar firing properties, giving adapting trains of spikes to depolarizing current injection (see also Aiken et al. 2003). The major difference between dlas and dlcs is that their postsynaptic target neurons are on opposite sides of the spinal cord. We propose that dlas and dlcs form two primitive ascending cutaneous sensory pathways in the tadpole central swimming circuitry. Since dlas do not produce a significant short latency ipsilateral reflex response, we now need to consider their role in the initiation of locomotion. Role of dlas in initiation of swimming Current injection that evokes spikes in individual sensory RB neurons can often start fictive swimming activity in immobilized tadpoles (Soffe 1997). Current evoked multiple firing of single dla or dlc sensory interneurons (Roberts and Sillar 1990) can also start fictive swimming. Following skin stimulation, dlcs and dlas are the only two groups of interneurons that fire reliably before swimming starts. Our paired recordings have confirmed the monosynaptic excitation of dlcs and dlas by skin sensory RB neurons (Li et al. 2003) (Fig. 4). All these data support the proposal that the dlas and dlcs are the first order sensory interneurons in the swimming-initiation pathway. When the skin is stimulated locally, a few sensory RB neurons will fire (Clarke et al. 1984). RB neurons have central axons about 2 mm long that all pass the mid-trunk region (segments 4 10; Hartenstein 1993) wherever their somata lie. RB neurons innervating all parts of the body surface can therefore make synapses to excite dlas located in the mid-trunk. The first role of dlas is therefore to distribute this excitation rostrally through the spinal cord, hindbrain, and to the midbrain, independent of the site of stimulation. Because the axons of more caudal sensory RB neurons only reach the mid-trunk region (Hartenstein 1993), dlas are necessary to ensure that excitation from these RB neurons reaches more rostral parts of the CNS. The second role of dlas, like dlcs (Clarke and Roberts 1984), is to amplify the skin stimulation signal. A few RB neuron spikes can lead to the excitation of many dlas, which would normally be active together to initiate swimming. We have examined the output connections of dlas within the spinal cord, although we assume that their axons in the brain make further en passant synapses here too. In the spinal cord, we have direct evidence from paired recordings and less direct evidence from responses to skin stimulation that dlas activate AMPAR and NMDAR to excite all classes of central pattern generator neurons (motoneurons, reciprocal inhibitory cins, recurrent inhibitory ains, and excitatory dins; see Fig. 1B). Again, these connections are similar to those made by dlcs (Li et al. 2003). When the tadpole is at rest and receives a skin stimulus, swimming normally starts after a 30- to 100-ms delay. The dla-derived EPSPs in CPG neurons appear after a 10-ms delay and are long-lasting (Figs. 8 and 9). It is highly likely that these EPSPs contribute to ipsilateral CPG neuron firing when swimming starts. Together with some RB-derived EPSPs (Fig. 11D), dla EPSPs may partly account for the more reliable firing of many CPG neurons at the beginning of swimming. Another similarity to the dlc interneurons in the crossed sensory pathway is that dlas receive gating inhibition during swimming (Sillar and Roberts 1988a, 1992a). We have shown that in both neuron classes, this inhibition comes from glycinergic ascending interneurons (ains; Li et al. 2002; Figs. 6 and 7). The effect of the inhibition is strong when swimming starts and may effectively close down the sensory pathway but later, as the inhibition weakens, stimulation at certain phases can escape inhibition and lead to higher frequency swimming (Sillar and Roberts 1993). Because the timing of modulatory inhibition is similar in both types of sensory interneuron, sensory stimuli that escape inhibition will excite CPG neurons on both sides via dlcs and dlas. Such phasic inhibitory gating during locomotion is a general feature of sensory pathways in most animals (Pearson and Collins 1993). Relation of dlas to neurons in other vertebrates dla interneurons in Xenopus have very long ascending axons so seem suited to a role as sensory projection neurons. At early stages in their development, the spinal cords of the newt Triturus vulgaris and zebrafish have interneurons that may be homologues of Xenopus tadpole dlas, with ascending axons and similar morphology (Roberts 2000). If they are homologous, we would expect them all to have similar sensory relay functions. Dorsal horn interneurons excited by cutaneous stimulation and with long distance ipsilateral ascending projections are a basic feature of the vertebrate spinal cord (Brown 1981; Willis and Coggeshall 1991). In adult mammals, these interneurons fall into two groups, having axons in the dorso-lateral funiculus (spinocervical tract neurons, Morin 1955) or in the dorsal columns (postsynaptic dorsal column pathway, Brown and Fyffe 1981). Interneurons that are possibly homologous to those forming the mammal spinocervical tract have also been described in adult Xenopus (Munoz et al. 1996). Anatomically, dlcs and dlas in Xenopus tadpole spinal cord may be the counterparts of classes of interneurons recently described in the developing mammal dorsal spinal cord (Goulding et al. 2002; Helms and Johnson 2003; Lee and Pfaff 2001; Sharma and Peng 2001). Classes of possible somatosensory relay interneurons are defined first by the dependence of their progenitors on the presence of the roof plate and second by a lack of expression of the transcription factor Lbx1. Individual populations are defined by the combinatorial expression of other transcription factors, the position of the soma, and their axonal projections (Gowan et al. 2001; Gross et al. 2002; Lee and Pfaff 2001; Muller et al. 2002). Particularly intriguing is the possible relationship between Xenopus dlas and the dl3 class. Neurons in this class lie deep in the dorsal horn and are thought to have ipsilateral ascending axon projections and to be excited by sensory input. CONCLUSIONS We have defined the anatomy, properties, connections, and some of the physiological roles of a class of spinal sensory

10 904 W.-C. LI, S. R. SOFFE, AND A. ROBERTS interneuron (dlas) forming the ipsilateral ascending cutaneous sensory pathway in a simple developing vertebrate. There appear to be just two classes of spinal cutaneous sensory projection interneurons that are excited by brief skin stimulation in the frog tadpole. One projects ipsilaterally (dlas) and the other projects contralaterally (dlcs). Both classes are excitatory, have similar properties, play a role in the initiation of swimming, and project rostrally to higher brain regions. Their definition emphasizes the fact that the central pattern generator for swimming requires excitation rostrally and on both sides of the body (Roberts and Tunstall 1990). Since both classes are inhibited when pressure stimulation to the skin leads to violent struggling movements, they cannot be responsible for the initiation of this motor output (Soffe 1993). We believe that the ability to define anatomical and physiological classes of neurons remains a key issue in understanding how the vertebrate spinal cord works in controlling behavior. Since we are close to understanding the function of each class of neuron in the Xenopus tadpole spinal cord, this primitive vertebrate system has great potential for establishing links between genetically marked neuronal types and function (Sharma and Peng 2001). ACKNOWLEDGMENTS We thank T. Colborn, D. Dunn, J. Maxwell, and B. Porter for technical assistance. GRANTS We thank the Wellcome Trust for generous support. REFERENCES Aiken SP, Kuenzi FM, and Dale N. Xenopus embryonic spinal neurons recorded in situ with patch-clamp electrodes conditional oscillators after all? Eur J Neurosci 18: , Boothby KM and Roberts A. Effects of site and strength of tactile stimulation on the swimming responses of Xenopus laevis embryos. J Zool (Lond) 235: , Brown AG. Organization in the Spinal Cord: The Anatomy and Physiology of Identified Neurones. Berlin: Springer-Verlag, Brown AG and Fyffe REW. Form and function of dorsal horn neurones with axons ascending the dorsal columns in cat. J Physiol 321: 31 47, Clarke JDW, Hayes BP, Hunt SP, and Roberts A. Sensory physiology, anatomy and immunohistochemistry of Rohon-Beard neurones in embryos of Xenopus laevis. J Physiol 348: , Clarke JDW and Roberts A. Interneurones in the Xenopus embryo spinal cord: sensory excitation and activity during swimming. J Physiol 354: , Goulding M, Lanuza G, Sapir T, and Narayan S. The formation of sensorimotor circuits. Curr Opin Neurobiol 12: , Gowan K, Helms AW, Hunsaker TL, Collisson T, Ebert PJ, Odom R, and Johnson JE. Crossinhibitory activities of Ngn1 and Math1 allow specification of distinct dorsal interneurons. Neuron 31: , Gross MK, Dottori M, and Goulding M. Lbx1 specifies somatosensory association interneurons in the dorsal spinal cord. Neuron 34: , Hartenstein V. Early pattern of neuronal differentiation in the Xenopus embryonic brainstem and spinal cord. J Comp Neurol 328: , Helms AW and Johnson JE. Specification of dorsal spinal cord interneurons. Curr Opin Neurobiol 13: 42 49, Lee SK and Pfaff SL. Transcriptional networks regulating neuronal identity in the developing spinal cord. Nat Neurosci 4(Suppl): , Li WC, Perrins R, Soffe SR, Yoshida M, Walford A, and Roberts A. Defining classes of spinal interneuron and their axonal projections in hatchling Xenopus laevis tadpoles. J Comp Neurol 441: , Li WC, Soffe SR, and Roberts A. Spinal inhibitory neurons that modulate cutaneous sensory pathways during locomotion in a simple vertebrate. J Neurosci 22: , Li WC, Soffe SR, and Roberts A. The spinal interneurons and properties of glutamatergic synapses in a primitive vertebrate cutaneous flexion reflex. J Neurosci 23: , Morin F. A new spinal pathway for cutaneous impulses. Am J Physiol 183: , Muller T, Brohmann H, Pierani A, Heppenstall PA, Lewin GR, Jessell TM, and Birchmeier C. The homeodomain factor lbx1 distinguishes two major programs of neuronal differentiation in the dorsal spinal cord. Neuron 34: , Munoz A, Munoz M, Gonzales A, and ten Donkelaar HJ. Evidence for an Anuran homologue of the mammalian spinocervicothalamic system: an in vitro tract-tracing study in Xenopus laevis. Eur J Neurosci 8: , Nieuwkoop PD and Faber J. Normal Tables of Xenopus laevis (Daudin). Amsterdam: North Holland, Pearson KG and Collins DF. Reversal of the influence of group Ib afferents from plantaris on activity in medial gastrocnemius muscle during locomotor activity. J Neurophysiol 70: , Roberts A. How does a nervous system produce behaviour? A case study in neurobiology. Sci Prog 74: 31 51, Roberts A. Early functional organization of spinal neurons in developing lower vertebrates. Brain Res Bull 53: , Roberts A and Clarke JDW. The neuroanatomy of an amphibian embryo spinal cord. Phil Trans Roy Soc 296: , Roberts A, Dale N, and Soffe SR. Sustained responses to brief stimuli: swimming in Xenopus embryos. J Exp Biol 112: , Roberts A and Sillar KT. Characterization and function of spinal excitatory interneurons with commissural projections in Xenopus laevis embryos. Eur J Neurosci 2: , Roberts A and Tunstall MJ. Mutual re-excitation with post-inhibitory rebound: a simulation study on the mechanisms for locomotor rhythm generation in the spinal cord of Xenopus embryos. Eur J Neurosci 2: 11 23, Sharma K and Peng CY. Spinal motor circuits: merging development and function. Neuron 29: , Sillar KT and Roberts A. Control of frequency during swimming in Xenopus embryos: a study on interneuronal recruitment in a spinal rhythm generator. J Physiol 472: , Sillar KT and Roberts A. A neuronal mechanism for sensory gating during locomotion in a vertebrate. Nature 331: , 1988a. Sillar KT and Roberts A. Unmyelinated cutaneous afferent neurons activate two types of excitatory amino acid receptor in the spinal cord of Xenopus laevis embryos. J Neurosci 8: , 1988b. Sillar KT and Roberts A. Phase-dependent modulation of a cutaneous sensory pathway by glycinergic inhibition from the locomotor rhythm generator in Xenopus embryos. Eur J Neurosci 4: , 1992a. Sillar KT and Roberts A. The role of premotor interneurons in phasedependent modulation of a cutaneous reflex during swimming in Xenopus laevis embryos. J Neurosci 12: , 1992b. Soffe SR. Triggering and gating of motor responses by sensory stimulation: behavioural selection in Xenopus embryos. Proc R Soc Lond B Biol Sci 246: , Soffe SR. Two distinct rhythmic motor patterns are driven by common premotor and motor neurons in a simple vertebrate spinal cord. J Neurosci 13: , Soffe SR. The pattern of sensory discharge can determine the motor response in young Xenopus tadpoles. J Comp Physiol 180: , Willis WD and Coggeshall RE. Sensory Mechanisms of the Spinal Cord. New York: Plenum Press, Zhao FY, Burton BG, Wolf E, and Roberts A. Asymmetries in sensory pathways from skin to motoneurons on each side of the body determine the direction of an avoidance response in hatchling Xenopus tadpoles. J Physiol 506: , 1998.