Journal of Neurophysiology

Mechanisms That Initiate Spontaneous Network Activity in the Developing Chick Spinal Cord

Peter Wenner, Michael J. O'Donovan

Abstract

Many developing networks exhibit a transient period of spontaneous activity that is believed to be important developmentally. Here we investigate the initiation of spontaneous episodes of rhythmic activity in the embryonic chick spinal cord. These episodes recur regularly and are separated by quiescent intervals of many minutes. We examined the role of motoneurons and their intraspinal synaptic targets (R-interneurons) in the initiation of these episodes. During the latter part of the inter-episode interval, we recorded spontaneous, transient ventral root depolarizations that were accompanied by small, spatially diffuse fluorescent signals from interneurons retrogradely labeled with a calcium-sensitive dye. A transient often could be resolved at episode onset and was accompanied by an intense pre-episode (∼500 ms) motoneuronal discharge (particularly in adductor and sartorius) but not by interneuronal discharge monitored from the ventrolateral funiculus (VLF). An important role for this pre-episode motoneuron discharge was suggested by the finding that electrical stimulation of motor axons, sufficient to activate R-interneurons, could trigger episodes prematurely. This effect was mediated through activation of R-interneurons because it was prevented by pharmacological blockade of either the cholinergic motoneuronal inputs to R-interneurons or the GABAergic outputs from R-interneurons to other interneurons. Whole-cell recording from R-interneurons and imaging of calcium dye-labeled interneurons established that R-interneuron cell bodies were located dorsomedial to the lateral motor column (R-interneuron region). This region became active before other labeled interneurons when an episode was triggered by motor axon stimulation. At the beginning of a spontaneous episode, whole-cell recordings revealed that R-interneurons fired a high-frequency burst of spikes and optical recordings demonstrated that the R-interneuron region became active before other labeled interneurons. In the presence of cholinergic blockade, however, episode initiation slowed and the inter-episode interval lengthened. In addition, optical activity recorded from the R-interneuron region no longer led that of other labeled interneurons. Instead the initial activity occurred bilaterally in the region medial to the motor column and encompassing the central canal. These findings are consistent with the hypothesis that transient depolarizations and firing in motoneurons, originating from random fluctuations of interneuronal synaptic activity, activate R-interneurons, which then trigger the recruitment of the rest of the spinal interneuronal network. This unusual function for R-interneurons is likely to arise because the output of these interneurons is functionally excitatory during development.

INTRODUCTION

Spontaneous activity is a characteristic feature of developing circuits in virtually every part of the nervous system that has been examined to date (Ben-Ari et al. 1989; Christie et al. 1989; Fortin et al. 1995; Ho and Waite 1999; Itaya et al. 1995; Landmesser and O'Donovan 1984; Lippe 1995; Maffei and Galli-Resta 1990). This type of network-driven embryonic activity is remarkably similar in tissues as diverse as the hippocampus, retina, and spinal cord (see O'Donovan 1999 for review) and is manifest as recurrent depolarizing events, during which cells within the network are synchronously activated. During these events, intracellular calcium is elevated (Garaschuk et al. 1998; Kulik et al. 2000; Leinekugel et al. 1995; O'Donovan et al. 1994; Wong et al. 1995), suggesting a role in developmental or trophic processes. In the spinal cord, spontaneous activity has been implicated in the development of limb muscles, bones, and joints (Hall and Herring 1990; Persson 1983; Toutant et al. 1979), the projections of cutaneous afferents to the dorsal horn (Mendelson 1994), motoneuronal neurite outgrowth in culture (Metzger et al. 1998), and the maturation of motoneuron electrical properties in organotypic culture (Xie and Ziskind-Conhaim 1995).

Despite the presumed importance of this form of periodic activity, very little is known about the mechanisms that regulate its onset. In the developing hippocampus, it has been shown that spontaneously occurring giant depolarizing potentials (GDPs) are preceded by an increase in the frequency of spontaneously occurring synaptic events. Menendez de la Prida and Sanchez-Andres (1999) showed that GDPs occurred 100–300 ms after the frequency of excitatory postsynaptic potentials (EPSPs) in hippocampal neurons exceeded a specific threshold. Between GDPs, they also observed transient increases in EPSP frequency that were lower than those occurring before a GDP. These findings suggest that developing networks experience transient increases of synaptic activity, presumably arising from the coordinated firing of groups or clusters of interneurons and that these events can trigger synchronized network activity when they become large or frequent enough.

In the embryonic chick spinal cord, some progress has been made in characterizing the basic mechanisms underlying spontaneous activity (Chub and O'Donovan 1998; Fedirchuk et al. 1999; Milner and Landmesser 1999; Tabak et al. 2000). In this preparation, spontaneous bursting lasts for ∼1 min (referred to as an episode) and is followed by a period of little or no network activity that persists for many minutes (referred to as the inter-episode interval). Episodes are composed of many cycles of discharge whose period progressively lengthens throughout the episode (Landmesser and O'Donovan 1984) and are generated by combined action of GABAergic, glutamatergic, and cholinergic inputs (Chub and O'Donovan 1998;Sernagor et al. 1995).

The mechanisms responsible for initiating spontaneous episodes are unknown. Using optical recordings from the transversely cut face of the spinal cord, it has been shown that the earliest activity at the beginning of an episode occurred in and around the lateral motor column and then evolved as a dorsomedial wave (O'Donovan et al. 1994). Consistent with a role for motoneurons in episode initiation, Ritter et al. (1999) showed that motoneuron firing began hundreds of milliseconds prior to the start of an episode, before activity could be detected in interneurons. If motoneuron activity is involved in episode initiation, we hypothesized that the intraspinal neuronal targets of motoneuron recurrent collaterals might mediate this process (Wenner and O'Donovan 1999). These recently identified interneurons (R-interneurons), which appear to be the avian homologue of the mammalian Renshaw cell identified in the adult cat (Eccles et al. 1954; Renshaw 1946), receive excitatory cholinergic input from motoneuron collaterals and project depolarizing GABAergic synapses to motoneurons and to other spinal interneurons (Wenner and O'Donovan 1999).

In the present work, we have also investigated whether or not spinal neurons including motoneurons and interneurons experience transient increases of synaptic activity as has been reported in developing hippocampal networks (Menendez de la Prida and Sanchez-Andres 1999) and if such transients are associated with episode initiation. In addition, we have investigated the hypothesis that the pre-episode motoneuron discharge triggers an episode of spontaneous activity by exciting R-interneurons, which in turn excite the rest of the interneuronal network. To examine these questions, we have compared the timing of electrical activity in motoneurons, R-interneurons, and other spinal interneurons at episode onset. We have examined the ability of motor axon stimulation to trigger an episode and have used calcium imaging to visualize the recruitment of R-interneurons during such stimulation and compare it with the recruitment pattern occurring spontaneously. We have also investigated the effects of blocking the synaptic connections to and from R-interneurons on the recruitment patterns and timing of spontaneous activity. Some of this work has been published in a conference proceeding (Wenner et al. 1998).

METHODS

Physiology

Chick embryos were removed from the egg at E9–E11 (stage 35–37) and staged according to the criteria of Hamburger and Hamilton (1951). The great majority of the experiments were performed on E10 embryos; data in figures were obtained from E10 embryos unless stated otherwise. Embryos were decapitated and the spinal cords were isolated as described previously (O'Donovan 1989; O'Donovan and Landmesser 1987) in recirculating cold (15°C) Tyrode's solution [concentration (in mM): 139 NaCl, 3 KCl, 17 NaHCO3, 12 glucose, 3 CaCl2, and 1 MgCl2] in accordance with National Institutes of Health guidelines. The spinal cord was isolated together with certain muscle nerves (adductor = adductors and obturator; femorotibialis = external and medial head, femorotibialis internal head, sartorius). After the dissection, the solution was allowed to reach room temperature and left undisturbed for ≥2 h (overnight preparations were left ≤12 h at 17°C). The solution was then cooled to 17°C, dorsal pia was removed, and a horizontal cut was made using a vibrating razor blade at the midpoint of the dorsoventral axis, leaving equal dorsal and ventral halves from about the last thoracic segment (T) 7 to lumbosacral segment (LS) 5. The ventral piece, with intact ventral roots and muscle nerves, was then transferred to the recording chamber and the solution temperature was increased to 27°C for the remainder of the experiment. Muscle nerves (with cut dorsal roots) were drawn into suction electrodes for recording and/or stimulating. Whole cell electrodes [4–8 MΩ, with a K-gluconate solution concentration (in mM): 10 NaCl, 130 K-gluconate, 10 HEPES, 1.1 EGTA, 0.1 CaCl2, 1 MgCl2, and 1 Na2ATP] were driven ventrally through the dorsal aspect of the ventral piece of cord. The electrode was positioned directly over the R-interneuron region dorsal to the medial part of the motor column (see Fig. 1). Typically the electrode was driven 100–200 μm into the tissue before recordings were obtained. In two cells, recordings were obtained while driving the electrode through the transversely cut face as described previously (Wenner and O'Donovan 1999). All whole-cell recordings were obtained using an Axoclamp 2B amplifier and custom written data-acquisition software (Labview 4.0). Extracellular suction electrode recordings were obtained from muscle nerves or from a slip of the VLF (2–5 mm long) and were amplified 1,000 times and filtered at DC-1 kHz (low-pass) or 200–3 kHz (high pass). VLF recordings have been shown to reflect the population activity of spinal neurons (Ritter et al. 1999). Cells were only accepted for further study if their resting membrane potential was more negative than −40 mV. Single-pulses and stimulus trains (20–50 Hz for 0.5 ms) of 30 μA were delivered to muscle nerves to activate R-interneurons (Wenner and O'Donovan 1999). When testing whether motoneuronal inputs could trigger an episode, we stimulated the ventral root (to activate motoneurons antidromically). Although single shocks could sometimes evoke an episode of activity, stimulus trains were more effective and were used routinely.

R-interneurons were identified by the presence of short latency synaptic input following stimulation of muscle nerves. Cells falling into this category had latencies to the onset of the earliest synaptic potential of ≤5 ms (see Wenner and O'Donovan 1999). Our previous work has shown that the great majority of R-interneurons produce a depolarizing potential in motoneurons temporally coupled to the occurrence of a spike in the recorded interneuron. In the present work, we used spike-triggered averaging to identify the source of spiking in the adductor muscle nerve (see Fig. 2). For this purpose, we recorded spontaneous spiking in an adductor motoneuron and averaged traces acquired from the adductor muscle nerve time-locked to the motoneuron spikes.

Optical recordings

To visualize the interneurons activated by stimulation of motoneurons, we loaded a calcium dye (Ca-green1 dextran 10,000 MW; Molecular Probes) into ventrally located spinal interneurons. A section of the ventrolateral cord (LS4–LS5 border), often including a portion of the lateral motor column, was drawn into a suction electrode containing ∼20% wt/vol of the dye dissolved in distilled water containing 0.2% Triton X-100 detergent (O'Donovan et al. 1993). This configuration was left overnight to allow retrograde transport of the calcium-sensitive dye back to the interneuronal cell bodies. After this loading period, the pia was removed between adjacent roots and the cords were cut transversely (vibrating razor blade) rostral and caudal to a particular ventral root (LS2 or LS3) leaving a single segment slice of cord (∼1 mm thick). The cord was then positioned in a recording chamber on the stage of an inverted microscope (Nikon Diaphot). The rostral face of the slice was viewed with epifluorescence illumination. Images were continuously acquired to videotape using an intensified video camera (Stanford Photonics) while the preparation produced spontaneous episodic activity or while stimulus trains (20–50 Hz, 100–200 ms) were applied to the ventral root to define the R-interneuron region. The tissue was illuminated using a 75-W Xenon Arc lamp with an excitation filter of 450–490 nm, dichroic of 510 nm, and a barrier filter of 520 nm. Various ND filters were used to reduce photodynamic damage. 1–3 mM KCl was added to the circulating Tyrode's solution (increasing the K+ concentration to 4–6 mM) to increase the frequency of spontaneously occurring episodes.

Image analysis

During the experiment, video data (30 fps) were stored on S-VHS tape (Sony SVO-9500 MD). Images were digitized off-line, frame by frame, and processed on Metamorph software (Image Systems). To display regions activated by stimulation or spontaneous activity, we constructed difference images normalized to the background fluorescence (ΔF/F). These were generated by subtracting a 30-frame average obtained prior to the stimulus or spontaneous episode (background image) from consecutive frames during the activity. Resulting images were then divided by the background image. A 5 × 5 median filter was applied to the background and subtracted images to remove noise. These images were displayed in false color and were stretched to occupy the full 8-bit range (0–255) of the frame store. To determine the R-interneuron region, 10 consecutive frames were selected as those with the greatest intensity changes and averaged. Quantification of fluorescence changes was performed on specific regions of interest (ROI) as described previously (O'Donovan et al. 1994). Time series at the onset of an episode were generated by averaging several episodes time-locked by the frame that showed a 15% increase in fluorescence. Optical activity at the onset of an episode was defined as the area in the first frame that showed an increase of ≥2 SD over the background. Resolution of a slowly developing diffuse signal at episode onset in the presence of mecamylamine required a 3- to 10-frame average.

Definition of the R-interneuron region

We defined the R-interneuron region (in the cut transverse face of the spinal cord) as the area that showed an increase in fluorescence 6 SD above the mean fluorescence of the same region under control conditions when a stimulus train was applied to the ipsilateral ventral root or a muscle nerve (Wenner and O'Donovan 1999). We have argued that this optical signal is derived from the activity of interneurons monosynaptically activated by motoneuron recurrent collaterals. Because of the importance of this interpretation to the present experiments, we present here additional data supporting this conclusion. To address the possibility that activation of ventral root afferents (Jiang et al. 1991) might contribute to the optical signals, we bath-applied glutamatergic antagonists [50 μM AP5, 20 μM 6-cyano-7-nitroquinoxalene-2,3-dione (CNQX)] and found that neither the amplitude of the ventral root-evoked optical signal (peak transient 128.2 ± 56.5% of the pre-drug signaln = 3) nor its location dorsomedial to the lateral motor column were substantially changed (measured ≥15 min after application of the drugs). We then established if the location of the ventral root evoked optical signal changed when the outputs from R-interneurons were blocked with the addition of a GABAergic antagonist to the glutamatergic antagonists (Fig.1 C, 50 μM bicuculline and AP5/CNQX) or in a separate experiment, a combination of a GABAergic and a glycinergic antagonist (strychnine, 1 μM). Although the ventral-root-evoked ventral root response was abolished in the presence of the drugs (Fig. 1 D), we found that the optical response still occurred in the same region as under control conditions. In the presence of the drugs, the optical signal was more diffuse and less intense than under control conditions. We hypothesize that the reduction of the optical response occurred because bicuculline blocks the reciprocal GABAergic connections between R-interneurons themselves or between R-interneurons and motoneurons (Wenner and O'Donovan 1999). Alternatively, it may be that some component of the signal is derived the activation of non-R-interneurons by the GABAergic projections of R-interneurons that is lost in the presence of bicuculline. Whatever the reason for the decline of the signal, the unchanging location of the signal in the presence of GABAergic and glutamatergic blockade and its abolition following bath-application of the cholinergic antagonist mecamylamine strongly supports the idea that this region contains interneurons directly activated by cholinergic motoneuron collaterals. Further support for this idea was obtained from whole-cell recordings made from identified R-interneurons. In these experiments, we recorded from identified R-interneurons in a horizontal preparation of the ventral cord in which the dorsal half had been removed. We found that stimulation of the adductor muscle nerve (dorsal roots cut) not only produced a short latency synaptic response in R-interneurons but also an antidromic field potential field potential ∼2 ms before the onset of the evoked synaptic potential (data not shown). This observation suggested that R-interneuron cell bodies were close to those of the adductor motoneurons in the anterior segments (LS1–LS3) where most of these recordings were made. Indeed, we found that R-interneurons cell bodies were located ∼40 μm dorsal to the adductor motor nucleus. We reconstructed the location of both adductor motoneurons and R-interneurons in the rostrocaudal and mediolateral plane as shown in Fig. 1 E. Mediolateral position was defined using midline and lateral border of the spinal cord and expressed as a percentage of the distance from each boundary. Rostrocaudal position was obtained with respect to adjacent ventral roots and expressed as a percentage of the distance from one root to another. The mediolateral boundaries of the R-interneuron column mapped in this way could then be superimposed on the optically imaged cut transverse face of the cord, as shown in Fig. 1 B. It can be seen that this boundary coincides with that of the R-interneuron region estimated from the optical recordings. Collectively, these results strongly suggest that the region activated optically by ventral root stimulation in the transverse plane contains the cell bodies of neurons directly activated by motoneuron collaterals (R-interneurons).

Fig. 1.

Evidence that R-interneuron cell bodies are located within the region optically activated by ventral root or muscle nerve stimulation.A: video-micrograph of the rostral transverse cut face at LS2–LS3 border showing interneurons labeled with calcium green dextran applied to the cut ventrolateral cord at LS4. The outlines of the spinal cord and the lateral motor column are indicated. B: difference image showing the optical activity induced by a train of stimuli applied to the ipsilateral LS3 ventral root. The red lines are derived from the data shown in E. (The arrow in Eshows the approximate location of the transverse plane shown inA–C.) The red outline defines the R-interneuron region as the area that shows an increase in the optical signal that is 6 SDs above the background. C: difference image of optical activity induced by a ventral root train in the presence of AP5, 6-cyano-7-nitroquinoxalene-2,3-dione (CNQX) to block glutamatergic transmission and bicuculline to block the GABAAoutput of R-interneurons. D: electrical recordings from the LS3 ventral root in response to a train of stimuli applied to the LS2 ventral root (before cutting transversely as described inA). E: comparison of the location of adductor motoneurons (blue dots) and interneurons with short-latency input from motoneurons (R-interneurons; red dots) recorded from a spinal cord in which the dorsal half had been removed (15 experiments have been combined to generate the map). The approximate distance from the lateral edge of the cord and the midline is indicated in percentage by the lines. These lines are shown in the transverse plane superimposed on the video-micrograph in B. Data in E show combined experiments from E9.5 to E11 embryos.

RESULTS

Inter-episode activity of motoneurons and interneurons

Motoneurons and spinal interneurons in the chick embryo spinal cord are spontaneously activated in recurrent episodes, which typically last from 30 to 90 s. The episodes are composed of depolarizing cycles that often become most evident toward the end of the episode as the cycle period lengthens. These spontaneously occurring episodes are separated by quiescent periods that vary from 10 to 20 min. During an episode there is a massive recruitment of spinal networks in which motoneurons and interneurons are activated synchronously (O'Donovan et al. 1994).

Fig. 2.

Spontaneous spike activity recorded from an adductor motoneuron (add mn, whole-cell recording), the adductor muscle nerve, and the ventrolateral funiculus (VLF) before, during, and after a spontaneous episode (A). The bar indicates the duration of spike-triggered averaging from the whole-cell recording to the adductor muscle nerve. This is illustrated in B and shows that the spike activity recorded in the nerve is derived from motoneuronal firing (n = average of 40 traces). C: recording from the LS2 ventral root showing that spike activity stops for several minutes after an episode and then resumes (↓) and progressively intensifies during the last 2/3 of the inter-episode interval.

To better understand the events leading to a spontaneous episode, we examined the activity occurring during the inter-episode interval and just before an episode. In the first 2–3 min after an episode, spinal neurons were inactive. After this time, certain classes of motoneuron (adductor and sartorius) began to discharge and their activity became progressively more intense until the next episode occurred. Other classes of motoneuron (e.g., femorotibialis) did not discharge during the inter-episode interval. Figure 2 A compares activity recorded from the adductor muscle nerve and intracellularly from an adductor motoneuron with the discharge recorded from interneurons projecting into the VLF. Discharge begins in the motoneurons and the muscle nerve record well before the occurrence of the episode, but little or no discharge could be recorded from the VLF during this period. After the episode, the motoneuron activity stopped. Spike-triggered averaging from the adductor motoneuron spike revealed a short-latency action potential in the muscle nerve (Fig.2 B), indicating that the muscle nerve discharge originates from the firing of motoneurons. Inter-episode motoneuron discharge was also recorded from the ventral roots and began ∼5 min after the end of an episode (Fig. 2 C, ↓).

During the inter-episode interval, we also recorded transient depolarizations from the ventral roots (Fig.3, A–C) that often appeared to be associated with the initiation of an episode (initial depolarization; Fig. 3 C). These transient depolarizations may be generated by interneuronal activity because optical recordings from interneurons in the cut transverse face of the cord revealed transient increases of fluorescence synchronized with the ventral root depolarizations (Fig. 3 B, asterisk). Alternatively, the interneuronal fluorescence transients might originate from spontaneous action-potential independent transmitter release (seediscussion). Although transients were evident two minutes before an episode (Fig. 3 A, a), they were not observed in the two minutes after an episode (Fig. 3 A, b). This finding is consistent with the progressive inter-episode increase in the amplitude of evoked and spontaneous synaptic potentials reported in earlier work (Chub and O'Donovan 2001;Fedirchuk et al. 1999; Tabak et al. 2000).

Fig. 3.

Spontaneous depolarizing events recorded from the ventral root increase in frequency and amplitude before an episode and are depressed after an episode. A: spontaneous episode recorded from a ventral root (LS4) showing the increase (a) of depolarizing events and their cessation (b) after the episode. B: spontaneous depolarizations recorded from a ventral root just before an episode are accompanied by fluorescence transients (asterisk) recorded from interneurons retrogradely labeled with calcium green dextran. The boxed region has been expanded in C and D. C: expansion of the ventral root record shows that the time course of the initial depolarization at episode onset is very similar to depolarizing transients recorded just before the episode onset. In this example, the gray transient has been scaled in amplitude only. D: expansion of the optical interneuronal signal shows that the time course of the signal at episode onset is very similar to the transients recorded before the episode onset. As in C, an earlier transient (B, gray panel) has been scaled in amplitude and superimposed over the optical signal at episode onset to show the similarity of their time course. E: recordings from the sartorius and femorotibialis muscle nerves show that the initiating transient (demarcated by box) is accompanied by a substantial increase in motor discharge in the sartorius muscle nerve.

About 500 ms before the onset of the episode, a depolarization was observed in virtually all ventral root and muscle nerve recordings (see also Ritter et al. 1999). This pre-episode or initial depolarization had a similar time course to the depolarizing transients. This is shown by scaling and superimposing a transient on the initial depolarization of the episode (Fig. 3 C). These initial depolarizations were larger in amplitude (19.7 ± 7.0% of the episode amplitude, 13 episodes measured in 5 experiments) than transient depolarizations averaged in the 2 min before an episode (9.4 ± 2.5% of the episode amplitude; 13 episodes measured in 5 experiments—P = 0.015, t-test). In some motoneurons (e.g., sartorius and adductor), the initial depolarization was accompanied by significant discharge (Fig. 3 E). This discharge was not observed in individual interneurons or in the VLF (Ritter et al. 1999).

Can motoneuron discharge trigger episodes?

The presence of intense pre-episode discharge in some motoneurons raised the possibility that it might be causally involved in triggering an episode. To test the ability of ventral root stimulation to trigger episodes, stimulus trains (50 Hz, 10 pulses) were delivered every 2 min during the inter-episode interval, while recording spontaneous episodes from another ventral root. The experiments were performed in two segment preparations (LS2–LS3 or LS3–LS4) in which spontaneous activity occurred at regular intervals (Fig.4). Because episodes occur spontaneously, it was important to ensure that the ventral root stimulus actually triggered an episode and was not simply coincident with a spontaneously occurring episode. We adopted two criteria for this purpose. The first was that the latency from the stimulus to episode onset should be <700 ms. We chose this comparatively long latency because the time it takes for a stimulus to trigger an episode can be several hundred ms depending on the stimulus intensity (Ritter et al. 1999) and when in the in the inter-episode interval the stimulus is presented. The second was that the stimulus should cause the episode to occur significantly prematurely. For this purpose, we first obtained control recordings of spontaneous episodes to establish the control inter-episode interval. Stimulus trains (10 pulses at 50 Hz) were then delivered to one of the ventral roots every 2 min while recording from an adjacent ventral root.

Fig. 4.

Stimulation of a ventral root (LS3) can evoke an episode prematurely but only if the stimulus is delivered late in the inter-episode interval. A: control episodes recorded from the LS2 ventral root. The stimulus to the ventral root started at the time indicated by the bar over the ventral root recording. To the right of the recording, the stimuli indicated (* and ▴) are shown on an expanded time scale. The stimulus marked by * shows the ventral root potential evoked by the stimulus train just before the stimulus triggered an episode (▴). B: application of mecamylamine blocked the ability of the ventral root stimulus to trigger an episode. The minimum interval between the stimulus and the next episode is shown to the right of the record. C: in the same experiment, application of 50 μm bicuculline and 1 μm strychnine lengthened the inter-episode interval and blocked the ability of the ventral root stimulus to evoke an episode. The shortest interval between the stimulus and the occurrence of an episode is shown in the expanded panel to the right of the records. In the experiment mecamylamine was applied after the inhibitory antagonists.

An example of this type of experiment is illustrated in Fig. 4. In this experiment, the control inter-episode interval was 11 min 37 s ± 19 s measured from three inter-episode intervals. We found that ventral root stimuli were capable of evoking episodes prematurely at 8 min 20 s and 8 min after the previous episode (70.3% of the normal interval) and that the latency from the stimulus to episode onset was 433 ms. We also found that the stimulus was only effective when presented in the last 1/4 of the interval. In four experiments, eight episodes were evoked by such stimuli at 73.0 ± 7.8% (P < 0.01, paired t-test) of the normal interval of spontaneously occurring episodes and with an average latency of 427 ± 76 ms from the onset of the stimulus.

In the next set of experiments, we investigated the mechanism of the ventral root triggering of the episode. Previous work had shown that motoneurons make direct cholinergic connections with a class of interneuron located dorsomedial to the lateral motor column (R-interneurons, see Fig. 1) (see also Wenner and O'Donovan 1999). R-interneurons project depolarizing GABAergic (and possibly glycinergic) connections onto motoneurons and other interneurons. To establish if R-interneuron activation mediated the activation of episodes by ventral root stimulation, we examined the ability of ventral root simulation to trigger episodes in the presence of mecamylamine (to block the nicotinic cholinergic inputs to motoneurons) or in the presence of bicuculline and strychnine (to block the depolarizing outputs of R-interneurons to other spinal neurons).

In the presence 50 μM mecamylamine (applied for >30 min; Fig.4 B), we found that the ventral root-evoked ventral root response was significantly reduced but not abolished during a train of stimuli (10 pulses at 50 Hz; Fig. 4 B, right, ▴). In addition, the presence of the drug reduced the frequency of spontaneous episodes but did not block them (inter-episode interval in mecamylamine 20 min 4 s ± 250 s vs. 12 min 44 s ± 102 s in control, P < 0.05). Despite the persistence of a small synaptic response in the adjacent ventral root, we found that the ventral root stimulus was ineffective at triggering an episode. This was true even when the stimulus was presented in the last 10% of the inter-episode interval when network excitability (defined as the ability of an external stimulus to trigger an episode) was at its highest (average time between stimulus and episode = 62 s, range 9–107 s, determined from 11 spontaneous episodes in 4 experiments). The shortest time between the ventral root stimulus and the next episode was 9 s (Fig. 4 B, ▴,right). The failure of ventral root stimulation to trigger episodes in the presence of mecamylamine suggests that the effect requires a functional synaptic connection between motoneurons and R-interneurons and is not due to some other nonspecific effect of activating motoneurons (e.g., K+ release, electrical coupling). This result also makes it very unlikely that the ventral root activation of an episode under control conditions was due to the stimulation of afferents in the ventral root.

We also repeated the stimulus protocol in the presence of bath-applied bicuculline (50 μM alone) or together with the glycine antagonist strychnine (1 μM). We found that the drugs abolished the ventral root evoked responses in motoneurons, suggesting that the R-interneuronal output was effectively blocked (Fig. 4 C, ▴,right). In contrast to control conditions, episodes were never evoked by the ventral root stimulus at short latency (average = 67 s, range 28–112 s, 6 episodes, 2 experiments) even when the stimuli were presented in the last part of the inter-episode interval when network excitability was high (Fig.4 C). The shortest time between the stimulus and the next episode was 28 s (Fig. 4 C, ▴, right). In one other experiment, 10 μM bicuculline and 1 μM strychnine significantly reduced but did not abolish the ventral root evoked ventral root response. In this experiment, two of two episodes occurred independently of the stimulus trains.

These results show that motoneuron stimulation can trigger an episode during the last quarter of the inter-episode interval when the network is most excitable (Fedirchuk et al. 1999;Tabak-Sznajder et al. 2000) and that the effect is likely to be mediated by the synaptic excitation of R-interneurons.

The R-interneuron–motoneuron loop can be substantially activated while the rest of the spinal network is relatively inactive

To understand better the activation of the network by the ventral root stimulus and the role of R-interneurons in this process, we obtained whole cell recordings from 24 R-interneurons in multi-segment, spontaneously active cords while stimulating ventral roots or muscle nerves (dorsal roots cut). R-interneurons were identified by the presence of short-latency (4.6 ± 0.5 ms, range 3.9–5.3 ms,n = 21, in 3 additional cells the latency was <5 ms but the stimulus artifact prevented a precise measurement) monosynaptic potentials following stimulation of the ventral roots or a muscle nerve with the dorsal roots cut (Wenner et al. 1999).

Stimulus trains were delivered to one muscle nerve while recording intracellularly from an R-interneuron and extracellulary from another muscle nerve or from the VLF. When the stimulus train was delivered early in the inter-episode interval (Fig.5, left), the intracellular membrane potential of the R-interneuron and the slow potential recorded from the adductor muscle nerve both rose to a peak within 200–300 ms but then decayed back to baseline within a second. Despite the failure of the stimulus to evoke an episode, a substantial depolarization of R-interneurons and motoneurons occurred whose amplitude could approach that of an episode (compare Fig. 5 A, right andleft). Despite the positive-feedback nature of this connection and the substantial recruitment of motoneurons, the stimulus did not trigger an episode. These large amplitude depolarizations were not simply aborted episodes because recordings from the VLF, a monitor of interneuronal activity, revealed only weak depolarizations that were much smaller than those accompanying an episode (Fig. 5 B). Therefore the R-interneuron-motoneuronal circuit could be significantly activated while other interneurons were relatively inactive.

Fig. 5.

The effect of muscle nerve (dorsal roots cut) stimulation on intracellularly recorded R-interneuron activity and the slow potentials and discharge recorded from muscle nerves or the VLF when the stimulus was delivered early (left) or late (right) in the inter-episode interval. A: simultaneous recordings from an R-interneuron and the adductor muscle nerve during stimulation of the femorotibialis muscle nerve with a train of 4 stimuli at 20 Hz.Left: an episode of regenerative activity was not triggered even though a substantial depolarization was recorded in both the R-interneuron and from the adductor muscle nerve. When the stimulus was delivered later in the inter-episode interval (right) an episode was triggered. The recordings were obtained from two different R-interneurons (R-int 1 and R-int 2). Note that the depolarization generated in the adductor muscle nerve was almost of the same amplitude as the depolarization associated with the episode. B: comparison of VLF and femorotibialis muscle nerve recording in response to stimulation of the adductor muscle nerve. Left: an episode of regenerative activity was not triggered despite producing an episode-like depolarization in the muscle nerve record while only a small response was observed in the VLF (↓). When the stimulus was delivered later in the inter-episode interval, an episode was triggered and a large depolarizing response was detected in the VLF.C: averaged (±SE) records of the intracellular membrane potential (de-spiked; see methods) of R-interneurons and the rectified integrated discharge of adductor motoneurons during a stimulus train applied to the femorotibialis muscle nerve early (left) and late (right) in the inter-episode interval. Note that even when an episode is not triggered, the muscle nerve discharge begins to increase after the second stimulus (*) applied to the muscle nerve. ▴, stimulus and stimulus artifacts.

The large amplitude of the R-interneuron and motoneuron depolarizations probably occurs because of the positive-feedback excitatory interconnections between motoneurons and interneurons. To illustrate this, recordings from several experiments were averaged to compare the timing and form of the rectified integrated muscle nerve discharge and the membrane potential trajectory of R-interneurons (Fig.5 C). They show that shortly after the second stimulus (Fig.5 C, bottom, *) in the muscle nerve train, asynchronous motoneuron discharge is recorded from the muscle nerve, presumably derived from the activated R-interneurons. This additional evoked discharge will further excite R-interneurons and probably accounts for their substantial depolarization.

When the stimuli were delivered during the last quarter of the inter-episode interval (Fig. 5, right), the intracellular membrane potential of the R-interneuron and the slow potential recorded from the adductor muscle nerve both rose to a peak within 200–300 ms (Fig. 5 A), followed immediately by a full episode (at ↓) accompanied by large depolarizing potentials in the VLF (Fig.5 B, right).

Imaging the recruitment of interneurons following ventral root stimuli

Based on the pharmacological and electrophysiological experiments described in the preceding text, we hypothesized that a motor nerve stimulus applied in the latter part of the inter-episode interval would first activate R-interneurons, which would in turn activate the rest of the interneuronal network. To test this idea directly, we used calcium imaging to compare the onset of optical activity in the R-interneuron region with that of other interneurons retrogradely labeled by the application of calcium-green dextran to the ventrolateral cord 1–2 segments caudal to the site that was imaged (O'Donovan et al. 1993; Wenner and O'Donovan 1999). These and subsequent imaging experiments were performed on a single isolated segment of the cord (typically LS3) to minimize the possibility that the initiating activity would occur deep in the tissue, remote from the cut face.

We first established the location of the R-interneuron region stimulating the ventral root early in the inter-episode interval. This activated a region dorsomedial to the lateral motor column that we have argued is the location of synaptically activated R-interneurons (see Fig. 1) (see also Wenner and O'Donovan 1999). When the ventral root was stimulated in the last quarter of the interval, the optical activity was first observed in the R-interneuron region but then expanded to include many of the labeled interneurons ipsi- and contralaterally, as an episode was triggered (Fig.6 A) and the rhythmic cycling activity was observed (Fig. 6 B). This finding suggested that R-interneurons were among the first spinal neurons to become active following the stimulus and that other interneurons were recruited subsequently. This pattern of recruitment was observed in 3/3 additional preparations. The earliest optical activity was observed in the R-interneuron region for ∼120 ms (4 frames; range 3–5) before the episode was triggered (Fig. 6). This result, together with the pharmacological evidence indicating that ventral-root evoked episodes are blocked in the presence of cholinergic and GABAergic antagonists, strongly supports the hypothesis that motoneuron activitycan activate the interneuronal network through the activation of R-interneurons.

Fig. 6.

Optical recordings of the activity of ventral interneurons labeled retrogradely with calcium green dextran applied to the LS5 ventrolateral cord in response to a train of stimuli (20 Hz, 200 ms) applied to the ventral roots late in the inter-episode interval.A: sequence of individual frames showing the cut transverse face of the spinal cord (between LS2 and LS3). The sequence of images represent individual frames (33 ms) subtracted from and divided by a control image obtained in the absence of stimulation (ΔF/F, see methods). Optical activity first appears in the R-interneuron region dorsomedial to the lateral motor column and then spreads to encompass the other labeled interneurons as an episode is evoked. The last frame shows the labeling pattern of the ventral interneurons. B: time series of the normalized fluorescence (ΔF/F) measured over the R-interneuron region following the ventral root stimulus (stim. VR).

R-interneuron activity during spontaneous episodes

The next set of experiments was designed to establish if motoneuron activity and subsequent R-interneuron firing initiatespontaneously occurring episodes. For this purpose, we first obtained whole cell recordings from identified R-interneurons during spontaneously occurring episodes to determine their patterns of activity and to establish if these patterns were consistent with a role in episode initiation. We then used calcium imaging to establish if the R-interneuron region is activated before other labeled interneurons at the onset of spontaneously occurring episodes. Finally, we report the effects of cholinergic blockade on the pattern of activity and the sequencing of interneuronal recruitment at episode onset.

Whole cell recordings of R-interneuron activity during spontaneous episodes

During spontaneous episodes, R-interneurons received a depolarizing synaptic drive that was similar in form to the population potentials recorded from muscle nerves (Fig.7 A). Consistent with a role in the initiation of the activity, they fired an initial burst of spikes (Fig. 7 A, *) at the onset of the episode; 18 of 24 cells fired a burst of action potentials [63 ± 27 (SD) Hz], while the remaining 6 cells fired a single spike at the onset of the episode. The pattern of R-interneuron discharge during the episode was most clear when cycling was distinct (Fig. 7 A, right). In all of the cells (24/24), firing also occurred at the beginning of each cycle. Figure 7 B compares the timing of the initial burst of spikes recorded from 11 R-interneurons (from 9 experiments) at episode onset with the slow potential recorded from the sartorius muscle nerve averaged from the same nine experiments. The onset of spiking in R-interneurons occurred in 7/11 cells before the peak of the averaged depolarization recorded from the sartorius muscle nerve at an appropriate time to initiate activity in other interneurons. For technical reasons, we did not record from VLF because we were interested in comparing the electrical activity recorded from the muscle nerves with that of R-interneurons. However, in a previous study, we noted that the peak discharge of the sartorius nerve occurred before the onset of firing in a small sample of non-R-interneurons (Ritter et al. 1999) consistent with our hypothesis that R-interneurons trigger activity in the rest of the network.

Fig. 7.

R-interneuron firing behavior at the onset of and during spontaneous episodes. Muscle nerve (sartorius and femorotibialis external–femoro) and whole-cell (R-interneuron) records compare activity during a spontaneous episode in motoneurons and R-interneurons. An initial burst of spikes (*) occurs at the onset of the episode in the R-interneuron.Left: 5 cycles of the activity (■) have been expanded on the right to show the timing of spiking in the R-interneuron for comparison with the discharge pattern of the muscle nerves. Spikes in the cell occur at the onset of each cycle coincident with the onset of the pause in sartorius discharge. B: comparison of the timing of the initial spiking and depolarization in 11 different R-interneurons (recorded from 9 embryos) with the slow potential recorded and averaged from the sartorius muscle nerve (±SE). ░ (75 ms in duration) is drawn to facilitate comparison of the spike timing and the nerve record. It terminates at the peak averaged depolarization in the sartorius muscle nerve.

Optical recordings of R-interneuron activity at the onset of spontaneous episodes

In the next set of experiments, we tested the hypothesis that R-interneurons become active before other spinal interneurons at episode onset, by comparing the activity of the R-interneuron region with that of other interneurons. The cells were labeled with calcium-sensitive dyes as described in the preceding text. Single-segment slices were used so that the optical activity would not simply reflect the spread of activity from a distant initiation site. Our strategy was to visualize many labeled interneurons simultaneously at the onset of spontaneously occurring episodes and establish which region became active first. We recognize the limitations of this experiment because not all interneurons will be labeled. However, if the R-interneuron region does not become active before other labeled interneurons, it will refute the hypothesis.

Figure 8 illustrates the optical signals originating from labeled interneurons visualized in the cut transverse face of a cord segment at the beginning of a spontaneous episode. In these experiments, we first defined the R-interneuron region by ventral root stimulation applied during a period of low network excitability as described previously (Fig. 8 A). This region was marked, and then spontaneous episodes were monitored. Figure 8 Cillustrates the initial video frames at the onset of a spontaneous episode subtracted and normalized to the background fluorescence (ΔF/F, averaged from 4 spontaneous episodes where activity began on the side ipsilateral to the application of the dye). It can be seen that the earliest activity begins in the R-interneuron region (Fig. 8 C, ↓) from where it spreads to the contralateral cell group near the central canal (Fig.8 C, ▾) to encompass the labeled cells on both sides of the cord as an episode is triggered (Figs. 8 and9 A). In some instances, the spread to the contralateral side occurred very early in the progression of the optical activity (Fig. 8 C). To establish the stability of the R-interneuron recruitment pattern, we compared several spontaneously occurring episodes in a single preparation. In each case the R-interneuron region was the first to become active (Fig.8 B). In 2/2 other preparations, four or more episodes per preparation were imaged and in each case the R-interneuron region was the first region activated at episode onset.

Fig. 8.

The R-interneuron region is the first to become active at the onset of spontaneous episodes. Spinal interneurons were retrogradely labeled with calcium green dextran, and the cord was transversely cut leaving a single segment piece that was then imaged at the onset of spontaneously occurring episodes. A: a 10-frame average of the percentage change in fluorescence following a motoneuron stimulus train demonstrates the R-interneuron region (red).B: schematic outline demonstrates where the first optical activity was detected at the onset of 4 different spontaneous episodes (4 different colors) in the spinal segment shown inA. To accurately define this region as the first to become active, we defined it as the area that rose 2 SDs above the background. In each case, the activity began in the R-interneuron region (red). C: individual frames of the normalized fluorescence changes at the onset of the 4 episodes shown inB were averaged (see text) and are displayed in sequence showing the onset of spontaneous activity beginning in the R-interneuron region. Notice the similarity between Aand the top right image of C where (↓) the first appearance of activity is shown clearly. D–H: schematic outlines show individual frames of the normalized fluorescence changes at the beginning of a spontaneous episode in 5 different preparations. Each color marks sequential frames showing the onset and progression of the fluorescence changes (defined as 2 SD above the background noise) at the beginning of a spontaneous episode. The R-interneuron region (white outline) was defined after stimulus trains were delivered to muscle nerves with dorsal roots cut (seemethods). In 1 experiment, the activity began in the R-interneuron region (H1), but then 1 min later (H2) an episode occurred prematurely and the optical activity began outside the R-interneuron region.

Fig. 9.

Comparison of the recruitment pattern of interneurons labeled with calcium green dextran in a single segment of the spinal cord (LS3) in the presence and absence of cholinergic blockade (mecamylamine 50 μM; atropine 2 μM). A, top left: 10-frame average of the change in normalized fluorescence following a motoneuron stimulus train, thereby defining the R-interneuron region. Next panel shows raw fluorescence image of interneurons labeled with calcium green dextran in the transversely cut spinal segment. Comparison of the recruitment pattern of labeled interneurons is shown in the following normalized fluorescence images in control conditions (average of 4 episodes) and in the presence of cholinergic blockade (B, average of 3 episodes). In B, the 1st frame shows the fluorescence in the absence of activity. In the presence of cholinergic blockade, the activity appears throughout the ventromedial region and increases more slowly than in the control.C: the initial optically active region at episode onset is compared under control conditions and in the presence of cholinergic blockade (mecamylamine 50 μM, atropine 2 μM). The control image was obtained by averaging 3 successive frames, and the image under cholinergic blockade was obtained by averaging 15 successive frames before the activity grew to encompass the entire ventral cord. The outlines of the motor column (LMC) and the R-interneuron regions are shown.

Figure 8, D–H, shows the pattern of recruitment observed at episode onset across 5 different preparations. As can be seen from the figure, in each of these experiments (Fig. 8, D–H1) the optical activity was initiated in the R-interneuron region at episode onset. In one unusual experiment, optical activity occurred within the R-interneuron region as in the other experiments (Fig. 8 H1). However, when a second episode occurred prematurely 1 min later, the earliest activity began outside this region (Fig. 8 H2). This was the only occasion that we observed this pattern in a single segment preparation. Collectively, therefore, these data demonstrate the consistency of the onset pattern both within individual preparations and between different preparations.

These optical results are consistent with the whole-cell recordings from individual R-interneurons and provide additional evidence that R-interneurons are among the first interneurons recruited at the onset of spontaneous episodes.

Effects of cholinergic blockade on interneuronal recruitment at episode onset

The results described in the preceding text are consistent with a role for R-interneurons in episode initiation. To investigate this hypothesis further, we applied cholinergic antagonists (3 experiments) and compared the timing of electrical activity recorded from ventral roots and optical activity recorded from labeled interneurons. In two experiments, mecamylamine and atropine were bath applied together. In another experiment, mecamylamine was first added to the bath solution alone and then atropine was added later. The results were similar for both conditions. In two of these experiments, interneurons were labeled bilaterally to allow comparison of the recruitment patterns on each side of the cord.

Following bath application of nicotinic and muscarinic cholinergic antagonists (50 μM mecamylamine, 2 μM atropine) to block the recurrent input from motoneurons to R-interneurons, the pattern of interneuronal recruitment changed. We found that interneurons located medial to the motor column on both sides of the cord, sometimes including the R-interneuron region, became active synchronously at episode onset. In addition, the rise time of the optical and electrical signals slowed in the presence of the drugs (Fig. 9 B).

These differences in the spatial pattern of recruitment can be seen more clearly in the averaged records shown in Fig. 9 C. In this experiment, three successive frames were averaged under control conditions to show the location of the initial optical activity at episode onset. In Fig. 9 C, right, a similar average was performed when the activity began in the presence of cholinergic blockade (mecamylamine and atropine). In this case, 15 frames could be averaged because episode initiation occurred more slowly and emphasizes the differences in the spatial distribution of activity under the two conditions. This finding raises the possibility that the region around the central canal is important in the initiation of activity in the absence of a functional motoneuron to R-interneuron connection. This medial region has also been implicated in rhythmogenesis in the neonatal rat spinal cord (Kjaerulff and Kiehn 1996;Kjaerulff et al. 1994).

In addition to these changes in recruitment pattern at episode onset, we also found that network excitability was decreased under cholinergic blockade. This was manifest in three ways. First, as we have mentioned, episode initiation was substantially slower in the presence of the drugs. Under control conditions, the average rise time (10–90%) of the optical signals in the R-interneuron region of the initiating side (measured in 10 episodes, 3 experiments) was 179.7 ± 50.3 ms, and this slowed to 298.1 ± 119.8 ms (measured in 8 episodes, 3 experiments). Second, the interval between episodes in single segment preparations lengthened by 82.5 ± 38.6% (n = 4 experiments, 13 control intervals, 8 drug intervals) in the presence of the drugs. This lengthening of the interval may also explain why episodes are longer in the presence of cholinergic antagonists (Ritter et al. 1999) because previous work has shown that longer intervals are accompanied by longer episodes (Tabak et al. 2000). As illustrated in Fig. 10 A, the inter-episode interval lengthened after application of the cholinergic antagonists as described previously for glutamatergic blockade (Barry and O'Donovan 1987; Chub and O'Donovan 1998; Tabak et al. 2000). Finally, the amplitude of the spontaneous transients recorded from the ventral roots and optically from labeled interneurons increased over the control values (Fig. 10 B). Under cholinergic blockade, the initial depolarizations were 166.5 ± 23.3% of their control value (n = 5 experiments, 8 episodes). The increased amplitude of these initial depolarizations is consistent with the idea that the threshold for episode initiation has increased in the presence of the drugs. Indeed, we also found that the amplitude of the transient ventral root depolarizations occurring in a 2-min pre-episode measuring period, was significantly larger in the presence of the drugs than in control (153 ± 37% of the control level, P < 0.01, paired t-test, 76 transients in drugs; 44 in control measured in 5 experiments, Fig. 10 B).

Fig. 10.

Cholinergic blockade depresses network excitability. A: plot of the inter-episode interval against time under control conditions and in the presence of cholinergic blockade is shown. Note that the inter-episode interval lengthened in the presence of the drugs. - - -, the time of application of the drugs. B: the depolarizing transients recorded from the ventral roots and optical transients recorded from interneurons retrogradely labeled with calcium green dextran are compared in control and cholinergic blockade conditions. Insets: expanded versions of ventral root recordings where the transient depolarization (*) is superimposed on the initial depolarization.

The increased amplitude of the transients may be related to the lengthening of the inter-episode interval that accompanied cholinergic blockade. We showed earlier (Fig. 3) these transients are depressed after an episode, suggesting that their appearance later in the inter-episode interval reflects the progressive increase in the amplitude of evoked and spontaneous synaptic potentials during the inter-episode interval. Because the inter-episode intervals are longer under cholinergic blockade, the recovery time is also longer, which may account for the larger amplitude of the events. Although we do not have causal evidence that the initial depolarizations actually initiate an episode, the increased amplitude of these initial depolarizations under cholinergic blockade is consistent with the idea that larger transients are required to initiate an episode in the absence of a functional motoneuron to R-interneuron connection.

We quantified changes in the timing of optical activity at episode onset by measuring the fluorescence changes within the R-interneuron and another region medial to the lateral motor column (non-R-interneuron region, see Fig. 11,inset) in each video frame. These were then compared with the simultaneously recorded ventral root potentials (Fig. 11). The rise of the fluorescence in the R-interneuron region on the initiating side started before that in the non-R-interneuron region on the same side in every episode we examined (12 episodes in 3 experiments; Fig.11 A). The average delay of the optical signal measured from the R- to non-R-interneuron region on the initiating side of the cord was 52 ± 10 ms (range 14–116 ms). In these and in subsequent experiments, we measured the delay when the fluorescence change reached 5% because at higher levels the optical signal in both regions was contaminated by the fluorescence increases of cells outside these regions. In 4/12 episodes, the next region to become active was the non-R-interneuron region on the contralateral side of the cord (rather than the ipsilateral non-R-interneuron region). The delay to the activity of this contralateral region was only 17 ± 3 ms (measured at 5% fluorescence) and may suggest the existence of a contralateral connection to this medial group of cells.

Fig. 11.

The effects of cholinergic blockade on the timing of electrical motoneuron activity and optical activity recorded from retrogradely labeled interneurons. Time course of ventral root activity and optical activity recorded from the R- and non-R-interneuron (black and gray traces respectively) regions under control conditions (A) and under cholinergic blockade (B).Right: the optical traces on an expanded time scale. In cholinergic blockade, activity in the R-interneuron region no longer precedes that in the non-R-interneuron region. VR int. discharge, ventral root integrated rectified discharge.

If motoneurons activate R-interneurons that then recruit the rest of the interneuronal network into an episode, then blocking the recurrent excitation of the R-interneuron population should change the timing of recruitment in the R- and non-R-interneuron regions. We therefore investigated the effects of cholinergic blockade on the timing of optical and ventral root activity. In the presence of the nicotinic cholinergic antagonists mecamylamine or a combination of mecamylamine and the muscarinic antagonist atropine, the sequence of recruitment changed (Fig. 11 B). We found that the earliest optical activity began in the non-R-interneuron region in 8/10 episodes (3 experiments). The average delay between activity in the non-R-interneuron region and the next active region was 124 ± 31 ms (range 6–266 ms, 8 episodes/3 experiments).

Collectively, the results show that cholinergic blockade changes the way the network is recruited in single segment preparations and are consistent with our hypothesis that under normal conditions the pre-episode motoneuron discharge activates R-interneurons. Furthermore, the increase of the inter-episode interval in cholinergic blockade and the larger initial depolarizations suggest that, in the absence of the facilitating action of motoneuron discharge on R-interneurons, it takes longer for the interneuronal network excitability to achieve a level that can sustain an episode.

DISCUSSION

In this paper, we have investigated the mechanisms involved in triggering spontaneous episodes in developing spinal networks of the chick embryo. We have found that motoneurons and interneurons experience spontaneous transient depolarizations that can be observed 2 min before an episode and appear to be responsible for the pre-episode discharge of motoneurons. Such transients are not detectable in the 2 min immediately after an episode. These motoneuronal depolarizations and discharge can trigger network activity through activation of the intra-spinal target of motoneurons (R-interneurons). This result suggests a special importance for transient fluctuations of spontaneous synaptic activity and the output elements (motoneurons) of spinal networks in the initiation of spontaneous activity during development.

What causes the transient depolarizations and firing in motoneurons?

Transient depolarizations were recorded in the ventral roots and in the VLF 2 min before, but not after, spontaneous episodes. These transient synaptic events may be responsible for the motoneuronal firing and depolarization that were recorded immediately before an episode. The firing was particularly prominent in the adductor and sartorius motor nerves although it was seen to various degrees in all of the muscle nerves we examined. The prominence of the discharge in sartorius and adductor probably occurred because some of these motoneurons were already firing just before the episode. Both classes of motoneuron began to fire at ∼1/3 of the inter-episode interval and continued to fire until an episode occurred. The initial depolarizations in motoneurons were accompanied by spatially diffuse calcium transients in interneurons labeled with calcium green but, surprisingly, not by discharge recorded from the VLF. What, then, is the origin of these inter-episode depolarizing transients that appear to eventually trigger an episode?

We propose two possible sources for the transients that are not mutually exclusive. The first possibility is that the motoneuronal depolarizations originate from the spiking of premotor interneurons. In support of this idea is the observation that the transient ventral root depolarizations were accompanied by similar transients in the VLF, particularly after cholinergic blockade when the amplitude of the transients was maximal. In addition, calcium transients synchronized with the motoneuronal depolarizations were recorded from interneurons retrogradely labeled with calcium green. The coincidence of the transients in motoneurons and interneurons suggests that both may originate from the firing of premotor interneurons. Such interneurons are likely to be few in number and are probably scattered throughout the gray matter because the calcium transients accompanying the depolarizations were of low intensity and spatially diffuse. However, we have been unable to resolve any significant interneuronal spiking during the inter-episode transient depolarizations or accompanying the initial depolarizations when motoneurons fire briskly. Of course, it is possible that we have not observed such activity because the relevant interneurons are few in number.

An alternative possibility is that action potential-independent quantal release of transmitter might be responsible for the transient depolarizations of motoneurons and interneurons when, by chance, release from several sources is briefly synchronized. Recordings from voltage-clamped spinal neurons have shown that there is a progressive increase in the amplitude and possibly the frequency of spontaneously occurring synaptic currents during the inter-episode interval and that these events are activity-independent (Chub and O'Donovan 2001). It is possible that the spatially diffuse nature of interneuronal optical signals accompanying the motoneuronal depolarizations reflects pre- or postsynaptic calcium transients associated with stochastic increases of transmitter release. According to this idea, motoneurons fire during the transients because they are either already firing or closer to spike threshold than interneurons. In future experiments, it should be possible to establish if the spatially diffuse interneuronal calcium transients require action potentials by determining if they occur in the presence of TTX.

Definition of the R-interneuron region

During the initial depolarization, motoneurons began to fire intensely ∼500 ms before the onset of an episode. We propose that this firing activates R-interneurons, which then excite the rest of the network. Some of the evidence for this hypothesis is derived from the optical recordings that show that the earliest interneuronal activity following ventral root stimulation, or at the onset of a spontaneous episode, occurs in a region dorsomedial to the lateral motor column that we have defined as the R-interneuron region (Wenner and O'Donovan 1999). For this reason, we now consider additional evidence that this region contains interneurons monosynaptically activated by motoneuron recurrent collaterals. First, reconstruction of the mediolateral positions of physiologically identified R-interneurons in the LS1 and LS2 segments demonstrated that they were located dorsal to, and overlapped mediolaterally with, the adductor motor nucleus. This location, which was established electrophysiologically, coincided with the R-interneuron region (Fig. 1). Second, ventral root stimulation activated the same region in the presence of the glutamatergic antagonists AP5 and CNQX, the GABA antagonist bicuculline, and the glycinergic antagonist strychnine. When coupled with the observation that ventral root activation of the R-interneuron region is abolished by cholinergic antagonists (Wenner et al. 1999), these findings make it very unlikely that the region is activated exclusively through polysynaptic projections from motoneuron collaterals or by the activation of glutamatergic, ventral root afferents. Although glutamatergic ventral root afferents have been identified in the rat spinal cord (Jiang et al. 1991), they have not been described in the chick embryo. Chu-Wang and Oppenheim (1978) have argued that the number of axons in the ventral root corresponds to the number of cell bodies in the lateral and medial motor columns, suggesting the presence of few, if any, ventral root afferents at this stage of development. Finally, in horizontally cut ventral-half preparations of the cord (seemethods), we have observed an optical signal in the R-interneuron region with a single shock applied to the ventral root, again making it unlikely that the region was activated polysynaptically (unpublished observations). Thus while it is possible that the optical signal following ventral root stimulation (in the absence of drugs) may contain a contribution from non-R-interneurons, our evidence strongly indicates that neurons monosynaptically activated by motoneuron collaterals are located in this region.

Role of R-interneurons in the initiation of spontaneous episodes

Several lines of evidence implicate R-interneurons in episode initiation. First, stimulation of the ventral roots or muscle nerves was capable of triggering an episode when the stimulus was delivered in the last quarter of the inter-episode interval when network excitability was highest. This effect was mediated by R-interneuron activation because it persisted in the presence of glutamatergic blockade and was prevented by blockade of either the cholinergic motoneuronal inputs to R-interneurons or the GABAergic output from R-interneurons. Second, whole-cell recordings from physiologically identified R-interneurons revealed that their initial spiking at episode onset occurred ∼50 ms before the peak depolarization recorded from the sartorius muscle nerve. Although we did not compare the timing of R-interneurons and non-R-interneurons, we can estimate their relative timing indirectly from an earlier study. Ritter et al. (1999) showed that the peak discharge of the sartorius nerve occurred before the onset of firing in a small sample of non-R-interneurons. Future experiments will be necessary to more directly compare the timing of initial activity in R-interneurons versus other interneurons. Third, optical recordings from spinal cord segments, containing interneurons labeled with calcium green dextran, revealed that the earliest interneuronal fluorescence change at episode onset began within the R-interneuron region and then spread to other interneurons. Finally, we found that blocking the recurrent cholinergic connection from motoneurons decreased the occurrence of spontaneous episodes and altered the sequence in which interneurons were recruited. Instead of the R-interneuron region leading the activity, we observed an early signal experienced bilaterally in the medial part of the cord. Further, we found that in the presence of cholinergic antagonists, the inter-episode interval lengthened, and the recruitment of motoneurons and interneurons was substantially slowed at episode onset. In addition, the initial depolarizations and corresponding optical transients observed at episode onset were larger than under control conditions, consistent with an increase in the threshold for episode initiation.

We propose that the reduction of network excitability in the presence of cholinergic antagonists occurs, in part, because the facilitating influence of the initiating motoneuron discharge on R-interneurons is removed. As a result, the remainder of the interneuronal network has to achieve a higher level of excitability to sustain episodes. We believe in this condition, other interneurons, possibly in the region around the central canal, are involved in the recruitment of the rest of the spinal network. It is conceivable, however, that cholinergic antagonists could depress network excitability independently of their effects on the motoneuron to R-interneuron connection. Milner et al. (1999) showed that cholinergic blockade reduced episode frequency in chick embryos at E4–5, when the motoneuron/R-interneuron pathway is unlikely to be functional (unpublished observations). While it is reasonable to believe that the effects of cholinergic blockade differ between the two ages for developmental reasons, such a result indicates that caution is required in attributing all of the effects of cholinergic blockade on network excitability to an interruption of the motoneuron/R-interneuron circuit.

Comparison with previous studies

In the neonatal hippocampus, spontaneous activity is expressed as network-driven GDPs, which are similar to the spontaneous episodes generated by the spinal cord. Menendez de la Prida and Sanchez-Andres (1999) used intracellular recording from hippocampal pyramidal cells to investigate the mechanisms leading to the initiation of a GDP. They found that some pyramidal cells in the CA3 region fired during the interval between GDPs and then stopped after a GDP. During the inter-GDP interval, the firing and the underlying EPSPs exhibited transient increases that reached a maximum 100–300 ms before a GDP. The occurrence of a GDP could be predicted when the underlying EPSP frequency exceeded a critical threshold, suggesting that, like the spinal cord, the transients were involved in triggering a GDP.

The spontaneous activity of hippocampal pyramidal cells described by Menendez de la Prida and Sanchez-Andres shares many similarities to that of the motoneurons we have described in the present work. This raises the possibility that pyramidal cells—the output neurons of the hippocampus—might also play an important role in GDP initiation. However, a recent study using paired intracellular recordings from different cell types in the neonatal hippocampus has argued that GDP initiation does not depend on a specific neuronal population but involves the co-operation between several cell classes (Menendez de la Prida and Sanchez-Andres 2000). Nevertheless, these authors did not directly attempt to record the initial activity within the slice or if such activity was reproducibly initiated by a particular cell class. It is generally assumed in both the hippocampus and in other developing networks that activity is initiated within a small group of cells and then rapidly propagates to invade the rest of the network. How the rest of the network is invaded will depend on the details and connectivity of the particular network. Unfortunately, activity does not originate from a fixed site so that capturing the cellular details of this process is very difficult. High-resolution optical imaging is one approach to this problem, but it is often difficult to visualize a large field sufficient to capture the initiation site and achieve cellular resolution. Calcium imaging has been used to visualize activity within the developing hippocampus (Canepari et al. 2000; Leinekugel et al. 1995), but no studies have focused on the recruitment of different cell types at the initiation site. However, it might be possible to approach this question in the hippocampus by imaging small pieces of tissue or islands that continue to generate spontaneous GDPs (Kazipov et al. 1997).

Instead of attempting to visualize the whole slice to capture the initiation site, the alternative is to record the activity of cellular populations. In the spinal cord, this is possible because the axons of motoneurons and some interneurons are accessible for recordings in the ventral roots and the VLF, respectively. These recordings allow us to monitor the earliest active members of the population at episode onset. Unfortunately, in a structure like the hippocampus this type of population recording is not possible.

Optical imaging to visualize initiation might prove more profitable in the retina than in the hippocampus. This is because the retina is essentially a planar structure that generates spontaneous waves that propagate across the retinal surface (Feller et al. 1996; Meister et al. 1991). It has been postulated that retinal waves are initiated in the amacrine layer, which is presynaptic to the output ganglion cells. According to a recent model, waves are initiated when a critical number of amacrine cells become coactive (Feller et al. 1997). This model predicts that amacrine activity should be more common than ganglion cell activity, although some recent experimental evidence argues against this conclusion (Zhou 1998).

Functional significance of episode initiation by motoneurons and R-interneurons

We have argued that spontaneous episodes in the chick cord are triggered by the motoneuron/R-interneuron circuit but have also shown that spontaneous activity still occurs when this circuit is interrupted pharmacologically. What then is the functional significance of this particular mode of triggering, if network activity can still occur in its absence? In previous work, we have proposed that the recovery after pharmacological blockade occurs because of a progressive increase in the strength of the remaining network connections, in particular, GABAergic connections (Chub and O'Donovan 1998; Tabak-Sznajder et al. 2000). Once spontaneous activity has recovered in the presence of cholinergic blockade, we propose that the distribution of activity and synaptic strength across the population of active neurons has changed compared with control conditions. We believe that some neurons (e.g., GABAergic neurons) now contribute comparatively more to network activity than under control conditions. For example, preliminary results have suggested that GABAergic synaptic potentials increase in amplitude during recovery from excitatory blockade and remain elevated when the inter-episode interval stabilizes (Tabak et al. 2000). The precise mechanisms of this increase are not understood but may involve a change in intracellular chloride homeostasis (Chub and O'Donovan 2000). At present, the developmental consequences of these changes in the distribution of activity and synaptic efficacy within developing networks are unknown. However, it seems reasonable to suppose that alterations of activity that occur by blocking the recurrent connection and therefore changing the normal recruitment pattern (i.e., lengthening inter-episode interval) may have important developmental effects on neurons and their synaptic targets.

A second functional question to arise from our results is their significance in terms of the adult function of R-interneurons. In adult animals, the homologue of the R-interneuron is the Renshaw cell. Studies in the adult cat have shown the Renshaw cell receives motoneuron input but inhibits its synaptic targets (Eccles et al. 1954). Although the precise function of Renshaw cells is unknown, it is important to stress that the inhibitory output of adult Renshaw cells ensures that their function will be very different from that during development (see Noga et al. 1987) when the output of the R-interneurons can be functionally excitatory. It is highly unlikely therefore that Renshaw cells would initiate rhythmic activity in the adult spinal cord.

However, after blockade of the motoneuron/R-interneuron pathway, the initial activity at episode onset occurred bilaterally in the medial part of the cord in the vicinity of the central canal. This region has been implicated in rhythmogenesis in the rat spinal cord (Kjaerulff and Kiehn 1996; Kjaerulff et al. 1994). The results of the recurrent blockade suggest that this medial region may have a similar importance when the output of R-interneurons becomes functionally inhibitory in the developing chick.

Acknowledgments

We are grateful to R. Burke, N. Chub, C. McBain, and J. Tabak for comments on the manuscript.

This study was supported by the National Institute of Neurological Disorders and Stroke Intramural Research Program.

Footnotes

  • Address for reprint requests: P. Wenner, NIH, NINDS, Lab. of Neural Control, Bldg. 49, Rm. 3A50, 49 Convent Dr., Bethesda, MD 20892-4455.

REFERENCES

View Abstract