THE JOURNAL OF EXPERIMENTAL ZOOLOGY 261~261-273 (1992)

Development of Spinal Motor Networks in the Chick Embryo MICHAEL O'DONOVAN, EVELYNE SERNAGOR, GERALD SHOLOMENKO, STEPHEN HO, MIKLOS ANTAL, AND WAYNE YEE

Section on Developmental Neurobiology, Laboratory of Neural Control, NINDS, NIH, Bethesda, Maryland 20892 ABSTRACT We have examined the cellular and synaptic mechanisms underlying the genesis of alternating motor activity in the developing spinal cord of the chick embryo. Experiments were performed on the isolated lumbosacral cord maintained in vitro. Intracellular and whole cell patch clamp recordings obtained from sartorius (primarily a hip flexor) and femorotibialis (a knee extensor) motoneurons showed that both classes of cell are depolarized simultaneously during each cycle of motor activity. Sartorius motoneurons generally fire two burstslcycle, whereas femorotibialis motoneurons discharge throughout their depolarization, with peak activity between the sartorius bursts. Voltage clamp recordings revealed that inhibitory and excitatory synaptic currents are responsible for the depolarization of sartorius motoneurons, whereas femorotibialis motoneurons are activated principally by excitatory currents. Early in development, the dominant synaptic currents in rhythmically active sartorius motoneurons appear to be inhibitory so that firing is restricted to a single, brief burst a t the beginning of each cycle. In E7-El3 embryos, lumbosacral motor activity could be evoked following stimulation in the brainstem, even when the brachial and cervical cord was bathed in a reduced calcium solution to block chemical synaptic transmission. These findings suggest that functional descending connections from the brainstem to the lumbar cord are present by E7, although activation of ascending axons or electrical synapses cannot be eliminated. Ablation, optical, and immunocytochemical experiments were performed to characterize the interneuronal network responsible for the synaptic activation of motoneurons. Ablation experiments were used to show that the essential interneuronal elements required for the rhythmic alternation are in the ventral part of the cord. This observation was supported by real-time Fura-2 imaging of the neuronal calcium transients accompanying motor activity, which revealed that a high proportion of rhythmically active cells are located in the ventrolateral part of the cord and that activity could begin in this region. The fluorescence transients in the majority of neurons, including motoneurons, occurred in phase with ventral root or muscle nerve activity, implying synchronized neuronal action in the rhythm generating network. Immunocytochemical experiments were performed in E14-El6 embryos to localize putative inhibitory interneurons that might be involved in the genesis or patterning of motor activity. The results revealed a pattern similar to that seen in other vertebrates with the dorsal horn containing neurons with y-aminobutyric acid tGABA)-like immunoreactivity and the ventral and intermediate regions containing neurons with glycine-like immunoreactivity.

Embryonic motor activity is a characteristic feature of vertebrate development that plays a n important role in the development of muscle (McLennan, '83; Renaud et al., '78; Toutant et al., '79; Srihari and Vrbova, '78) and motoneurons (Pittman and Oppenheim, '78, '79). In the chick, as in other species, several studies have shown that the motor output is generated by spinal circuits in the absence of descending or afferent activity (Hamburger et al., '66; Landmesser and O'Donovan, '84; O'Donovan and Landmesser, '87; O'Donovan, '89; Bekoff et al., '89), but little is known about the neuronal mech01992 WILEY-LISS,INC.

anisms generating this behavior. Over the last 20 years a number of detailed studies have generated a clear picture of how hindlimb activity develops in the chick embryo (Sharma et al., '70; Provine, '71, '72, '73; Ripley and Provine, '72; Bekoff et al., '75; Bekoff, '76; Watson and Bekoff, '90) and is expressed during locomotion in the hatched animal (Jacobson and Hollyday, '82a,b). Embryonic studies have shown that the alteration of flexor and extensor motoneurons emerges early in development (Bekoff et al., '75; O'Donovan and Landmesser, '87) and becomes progressively more refined there-

262

M. O’DONOVAN ET AL.

after (Bekoff, ’761, raising the possibility that the line) and in some cases fast green dye was added spinal circuits responsible for embryonic motility to the perfusate of one of the chambers to monitor are the precursors of locomotor networks in the the adequacy of its isolation. mature animal. Recording of muscle nerve activity Progress in understanding the organization and Recordings from muscle nerves or ventral roots properties of developing spinal networks is complicated by the small size of embryonic neurons, which were made with tight fitting suction electrodes. The are difficult to record from in ovo. For this reason electrical signals were amplified with a bandwidth we have investigated the development of motor net- of DC to 1-5 kHz, which allowed recording of both works using isolated preparations of the lumbosa- spike activity (AC) and slow (DC) electrotonically cral cord and brainstem that express rhythmic propagated population potentials (O’Donovan,’89; motor activity similar to the embryonic activity in Lee and O’Donovan, ’91). Neurograms were ampliovo (Velumian,’81;Landmesser and ODonovan,’84; fied between two and ten ( X 1,000)times and stored ODonovan and Landmesser, ’87; Barry and O’Dono- on magnetic tape for subsequent analysis. Pectovan, ’87; O’Donovan, ’89;Sholomenko and O’DO~O-ralis signals were usually filtered from 100to 10 kHz. van, ’90). In this paper we review recent progress Brainstem stimulation towards identifying the synaptic inputs to motoBrainstem stimulation was accomplished using neurons during motor activity, the likely location and properties of the interneurons involved, and ini- either a glass-coated platinumhridium or Tyrodetial studies of the descending projections to lum- filled glass microelectrodes (1.5-2 megohms).Sites that initiated motor activity were established by bosacral motor networks. advancing the stimulating electrode into brainstem MATERIALS AND METHODS and recording the evoked activity in the muscle nerves. The optimum stimulation sites were deterIsolated spinal cord and brainstem mined by advancing the electrode to the point where preparation White leghorn chicken embryos were removed lumbosacral motor activity was evoked at the lowfrom the egg and placed in a superfusion chamber, est current (Sholomenko and Steeves, ’87). where they were perfused with cooled (lO-lS°C) Intracellular recording and patch clamp reTyrode’s solution equilibrated with 95%02/5%C02. cording from motoneurons For brainstem experiments, the embryo was quickly Intracellular and whole cell patch clamp recorddecerebrated to the mesencephalic level, whereas, ings were made from antidromically identified lumfor experiments involving the lumbosacral cord, the bosacral motoneurons in the isolated cord during animal was rapidly decapitated. After evisceration, episodes of motor activity. Intracellular recordings a ventral or dorsal laminectomy was performed to were made using high-impedance microelectrodes expose the spinal cord (Landmesser and O’Donovan, ’89). Whole cell current and voltage (O’Donovan, ’84; Barry and O’Donovan, ’87).The dura was usuclamp recordings were obtained from motoneurons ally removed, and in many experiments the cord using the “blind patch” technique (Blanton et al., was hemisected. The sartorius (a hip flexor and ’89), the application of which to the embryonic chick knee extensor) and femorotibialis (a knee extencord has been discussed elsewhere (Sernagor and sor) muscle nerves were dissected free for recordO’Donovan, ’91). Signals were amplified using an ing and stimulation. The pectoralis (wingdepressor) Axoclamp-2A amplifier (Axon intruments) and nerves were also dissected free bilaterally in the recorded onto magnetic tape for further analysis. spinal cordhrainstem preparation. The cord was then transferred to a Sylgard lined Voltage clamp was performed using the continuous, recording chamber as described previously (ODono- single-electrode mode of the amplifier. van, ’89).The chamber was modified for the spinal Ablation experiments cordhrainstem preparations to accommodate two To isolate regions of the lumbosacral cord that movable partitions to isolate different portions of might be essential for the generation of motor activthe neuraxis into independently superfused chamity, we removed various parts of the lumbosacral bers. The brainstem was isolated into one chamber, the cervical and thoracic cord (from C2 to T5) cord and recorded the effects on the pattern of sarinto another, while the remainder of the cord (below torius and femorotibialis discharge. The ablations T5) was restricted to a third chamber. Seals between were performed with a vibrating needle (Hamthe baths were made with petroleum jelly (Vase- burger et al., ’66) mounted on a micromanipula-

MOTOR DEVELOPMENT IN CHICK

263

chrome alum-gelatin. Sections were reacted to visualize y-aminobutyric acid (GABA)and glycine immunoreactivity using the method described by Somogyi and Hodgson ('85). After etching the resin with ethanolic sodium hydroxide and removing the Os04 Optical detection of neural activity with sodium metaperiodate, the sections were incuThe cellular calcium transients associated with bated with rabbit anti-GABA (Pel-Freez; diluted motor activity were imaged in E10-El3 embryos 1:1,000) or rabbit antiglycine (Chemicon, diluted using the calcium sensitive fluorescent indicator 1:400) antisera. Subsequently, biotinylated goat Fura-2 (Grynkiewicz et al., '85).The cut rostra1 face antirabbit IgG (Vector Labs; diluted 1:200) was layof the spinal cord (LS31LS4)was loaded for 1h with ered over the slides, followed by a n avidin-biotinya 10-20 p.M solution of the Fura-2AM (Molecular lated peroxidase complex (ABC; Vector Labs; diluted Probes), and then washed for a further 1h r before 1:lOO). All of the antibodies were diluted in Tris recordings were taken (Grynkiewicz et al., '85).The (10 mM) phosphate (10 mM)-bufferedisotonic saline dye-loaded cord was mounted in a perfusion cham- (TPBS; pH 7.41, to which 1% normal goat serum ber fixed to the stage of an inverted microscope. The was added. Immunoreactions were visualized with cord face was pressed onto a coverslip forming the diaminobenzidine chromogen and the reaction end chamber's base, and illuminated at 380 nm or 340 product was intensified with Os04. To test the specificity of the immunostaining nm using a 75 W Xenon lamp. The fluorescence changes accompanying motor activity were recorded method, some sections were incubated in normal at 510 nm using a n intensified CCD camera (Video- rabbit serum (Vector Labs; diluted 1:lOO) instead scope KS1381, Pulnix TM-845).Video signals were of the primary antiserum. Some of the sections, stored on tape and digitized with a frame grabber processed using the full sequence of the immuno(Data Translation DT-2861 and DT-2858) for fur- cytochemical procedure, were incubated in the peroxidase medium without H202,to control for the ther analysis. During neural activity, cellular fluorescence binding of diaminobenzidine to the tissue. No spedecreased when the cord was illuminated at 380 cific staining was observed in these sections. nm and increased when it was excited at 340 nm RESULTS corresponding to the excitation spectrum for the calcium-Fura complex (Grynkiewicz et al., '85).The Motor behavior of the isolated spinal cord in vitro signal to noise ratio was largest using 380 nm illumination, so this wavelength was routinely used At early stages of development (E6, E7) the patin the experiments. tern of motor activity produced by the isolated spiTo visualize the decreased neuronal fluorescence nal cord is very simple and consists of one or during motor activity, we inverted the video image two recurring cycles. Recordings from muscle and subtracted each video frame from an inverted, nerves reveal that activity in sartorius (flexor) averaged control image taken before the activity motoneurons is restricted to a single brief burst began. This procedure resulted in a positive difference image in which brightness was proportional at the beginning of each cycle, whereas each cycle of extensor discharge (femorotibialis and to the decrease in fluorescence. caudilioflexorius) lasts for about 1sec (Landmesser Immunocytochemical characterization of and O'Donovan, '87; see Fig. 1A). With further spinal interneurons development, recurring bouts of activity (referred Chick embryos (E14-El6) were perfused through to as episodes) contain several cycles, and, by E l 0 the heart with oxygenated (95% 02,5% CO2) or E l l (stage 36-37), a n episode can comprise five Tyrode's solution and then by a fixative contain- to 20 cycles. While there is little developmental ing 2.5% glutaraldehyde and 0.5% paraformalde- change in the pattern of extensor firing, a second hyde in 0.1 M phosphate buffer (PB; pH 7.4). The burst appears in each cycle of sartorius activity. The lumbosacral cord was removed and fixed in the second burst follows the initial discharge after a same fixative for 2-3 hr, then 1-2-mm-thick slices pause during which firing is depressed. Initially were cut from it. Following several washes in 0.1 M (E8,stage 32-33) thissecondburst is weak (O'DonoPB, tissue blocks were treated with Os04, dehy- van and Landmesser, '871, but it increases in amplidrated, and embedded in araldite. Semithin sections tude and duration so that, by stage 36, a pattern of were cut at 0.5 pm and dried onto slides coated with alternation is established with the extensor firing tor and could be made in any plane of the cord. The ablations were made in the recording chamber, which allowed motor activity to be monitored before and immediately after the lesion.

M. O'DONOVAN ET AL.

264

A.

E7

I

I

I

I

5 sec

I

5 sec Fig. 1. Pattern of motor activity recorded from hindlimb muscles nerve changes between E7 (A) and E l l (B). Recordings were made from the nerves innervating the sartorius (pri-

marily a hip flexor) and the femorotibialis (a knee extensor) muscles.

obtain intracellular recordings from identified motoneurons. Initially this was achieved using high-impedance microelectrodes to penetrate the cells (O'Donovan,'891, but more recently it has been accomplished using whole cell patch recordings from motoneurons (Sernagor and O'Donovan, '91). The results are similar using both techniques. Intracellular and whole cell patch recordings from femorotibialis and sartorius motoneurons reveal that both classes of cell are depolarized simultaneously during each cycle of activity, as predicted from the muscle nerve recordings. This is illustrated in Figure 2, which shows current and voltage clamp recordings from femorotibialis and sartorius motoneurons during an episode of motor activity. Femorotibialis motoneurons discharge durIntracellular and whole cell patch clamp ing the peak of their depolarization, indicating the recording from identified motoneurons excitatory nature of their synaptic drive (Fig. 2A). To investigate the cellular mechanisms underly- This depolarization is produced by rhythmic inward ing the muscle nerve activity, it was necessary t o currents that can be detected under voltage clamp

(see Fig. 1B;Landmesser and O'Donovan, '84; Barry and O'Donovan, '87; O'Donovan, '89). In addition to the propagated discharge, electrotonically decremented synaptic potentials can be recorded from the muscle nerve (O'Donovan, '89; Lee and O'Donovan, '91). During motor activity, the time course of these potentials is very similar to that of the membrane potential trajectory of individual motoneurons and is similar in flexor and extensor muscle nerves (O'Donovan, '89). This finding suggests that both classes of motoneuron are depolarized simultaneously during motor activity, a suggestion that has been supported by intracellular and whole cell patch clamp recordings from motoneurons.

MOTOR DEVELOPMENT IN CHICK

A

265

C

................................................................................................................. -5BrnV

43rnV

-

j.. ...............................................................................................................................

5 sec

B ,-BDrnV

..................................................................................................................................... ~ ;........................................................................................................................................ 60mV

it

5 sec

5 sec

Fig. 2. Whole cell patch clamp recordings from two identified femorotibialis (A,B) and two identified sartorius motoneurons (C,D) during episodes of motor activity. The upper traces in A and C illustrate whole cell recordings obtained in current clamp with the resting potential indicated to the left of the traces. The upper traces in B and D illustrate the synaptic currents recorded using continuous single electrode voltage clamp, with the holding potential shown to the left of the traces. Two current traces are shown for the sartorius motoneuron in D, each of which was recorded during a separate episode of motor

activity. The record obtained at a holding potential of - 60 mV reveals a rapid inward current (asterisks), which coincides with the pause in firing recorded in the muscle nerve (displayed in the lowest trace). At a holding potential of - 35 mV, the rapid inward current can no longer be detected, leaving a smaller inward current that extends throughout the cycle. The lower traces in each of the records show the neurograms simultaneously obtained from the respective nerves (A and B, femorotibialis; C and D, sartorius). Data in A and B from E l l embryos, in C from an E l 2 embryo, and in D from an E l 0 embryo.

(Fig. 2B). These currents reverse near 0 mV consistent with the action of excitatory synapses (Sernagor and O'Donovan, '91). Direct pharmacoological demonstration of the transmitter at these synapses has been difficult because bath applied amino acids and their antagonists disrupt the genesis of motor activity, presumably by a n action on premotor interneurons (Barry and O'Donovan, '87). Moreover, the excitatory interneurons mediating the synaptic drive to motoneurons may be located close to the motoneuron pool (Ho and O'Donovan,

'90; and see below), complicating selective pharmacological manipulation of synapses on motoneurons. While excitatory synapses appear to be the dominant input to femorotibialis motoneurons it is also possible to record IPSPs in these cells, particularly at the beginning of an episode (O'Donovan, '89). In addition, Fura-2 imaging (see below) has detected the presence of rhythmic calcium transients in these motoneurons during motor activity, that may be generated by calcium currents flowing through voltagedependent calcium channels (McCobb et al., '90).

266

M. O'DONOVAN ET AL.

Sartorius motoneurons behaved very differently from femorotibilias motoneurons, even though they were depolarized in a similar way. The cells fired on the rising part of their depolarization at the beginning of each cycle, stopped firing, and then resumed as the membrane potential returned t o rest. In some cells, such as the one illustrated in Figure 2C, firing was depressed during the initial part of an episode when cycles recur rapidly. Injection of depolarizing current into the cells during activity can reverse the trajectory of the membrane potential from depolarizing to hyperpolarizing during the pause in sartorius discharge, leaving the hyperpolarizing components superimposed on a small, sustained depolarizingpotential (O'Donovan, '89). These observations suggest that the depolarization of sartorius motoneurons is generated by more than one type of synaptic input. Voltage clamp experiments have confirmed this idea and revealed the presence of two types of synaptic current in sartorius motoneurons. One of these currents is inward at holding potentials below - 50 mV, has an apparentreversalpotential between - 50and - 30mV (at Ell-E13, but it tends t o be lower in younger embryos), and coincides with the pause in firing recorded from the nerve. The other is a more prolonged inward current that coincides with the period of firing recorded from the nerve (Fig. 2D). The phasic current reversing between - 50 and - 30 mV is presumably inhibitory, because it coincides with cessation of firing whereas the smaller current is probably excitatory, although we have not yet determined its reversal potential. Intracellular recordings using high-impedance electrodes and ventral root recordings have provided evidence that the equilibrium potential for the IPSP is actually above rest potential (discussedby ODonovan, '89). Although the ionic nature of the inhibitory current is unknown, it could be chloride mediated because developing spinal and hippocampal neurons receive chloride dependent IPSPs with a reversal potential above rest (Takahashi, '84; Ben-Ari et al., '89). This idea could be tested using chloride injectionwith high-impedance electrodesor by intracellular perfusion with a chloride-filled patch electrode. One difficulty with this hypothesis is that the reversal potential of chloride mediated IPSPs should be substantially lower than the - 50 mV in our whole cell patch recordings, assuming the cell contents are completely replaced with the patch solutionwhich has a low chloride concentration (1-2 mM). This discrepancy could arise if the IPSP was not chloride mediated or, alternatively, for techni-

cal reasons, including leaky seals, inadequate space clamp, or incomplete dialysis of the cells. The inhibition of firing in sartorius motoneurons is probably the result of shunting due t o the very large decrease in motoneuron membrane impedance that accompanies the IPSP (O'Donovan, '89). The inhibition is probably not due t o the action of voltage-dependent conductances because depolarizing current injection produces steady firing (O'Donovan, '89). Of course, the active properties of motoneurons might be modified by neuromodulators released during motor activity, so that current injection at rest would not reveal them, but we have no evidence in support of this idea. We have also obtained preliminary recordings from sartorius motoneurons in younger embryos (E7, E8) when the discharge recorded from the nerve is restricted to the beginning of each cycle. At this age, the dominant synaptic current in rhythmically active sartorius motoneurons has a n apparent reversal potential near - 50 mV and is presumably inhibitory. It is not yet clear if the initial discharge of sartorius motoneurons is due to the presence of an additional excitatory current or to the depolarization at the onset of the IPSP. The appearance of a second burst in sartorius motoneurons later in development could be due to a decrease in the efficacy of this inhibitory current, to an increase in excitatory drive, or to both. Resolution of this issue must await identification of the relevant excitatory and inhibitory interneurons and determination of the developmental changes in their output to motoneurons.

Location of essential elements of rhythmogenic and patterngenerating networks The goal of the ablation experiments was to establish if particular regions of the lumbar cord are specialized for rhythmogenesis. Our motivation derived from previous work in the chick and other vertebrates, suggesting the existence of some degree of regionalization in the networks responsible for rhythmic motor activity. Previous studies in the chick suggested that the rhythm generating networks were limited to the ventral part of the lumbar cord, because surgical removal of the dorsal lumbosacral cord at E2 does not alter embryonic motility until quite late in embryonic development (Hamburger et al., '66). Moreover, in other vertebrates, some of the interneurons influencing rhythmic motor activity appear t o be located rostra1 to the motoneurons involved (Arshavsky et al., '86; Currie and Stein, '90; Schefchyk et al., '90).

MOTOR DEVELOPMENT IN CHICK

267

oped an optical method to identify the behavior and location of cells activated during episodes of motor activity (O'Donoran et al., '90). As described in the methods we used the calcium indicator Fura-2AM (Grynkiewicz et al., '85) to load the cut transverse face (LS3/LS4 boundary) of the spinal cord. Dyeloaded cells were widely distributed throughout the cut face but tended to be restricted to its most superficial layers. During activity, a number of loaded cells, including motoneurons, exhibited oscillations of fluorescence in phase with the electrical activity recorded from ventral roots or from the femorotibialis nerve. The fluorescence change in motoneurons depended on the frequency of their antidromic stimulation and, in addition, was temporally correlated with firing during motor activity, implying that the fluorescence signal in motoneurons is primarily dependent on impulse activity. One of the most striking features of the activation pattern of neurons in the cut face was their near synchrony during each cycle of activity. This was true for cells in all regions of the cord, although the fluorescenceoscillations of dorsally located cells tended to decay more rapidly than those in ventral cells. The synchronous activation of interneurons was anticipated because flexor and extensor motoneurons are depolarized simultaneously, implying that they are driven by activity in a common set of excitatory interneurons (O'Donovan, '89; Sernagor and O'Donovan, '91; and see above). In addition, the inhibitory interneurons responsible for the sartorius inhibition are likely to be coactive with excitatory neurons at the beginning of the cycle (O'Donovan, '89). Moreover, our results have confirmed earlier studies using extracellular recording from the chick cord in ovo, which showed that activity in different regions of the ventral cord is synchronized during the burst discharges underlying embryonic motility (Provine, '71, '73). The active cells outside the motor nucleus were distributed throughout the transverse face of the cord, with high numbers in the ventral and intermediate regions of the grey matter (Fig. 4).This distribution could arise if neurons in some regions of the cord lack calcium channels or fail to express fluorescence signals during motor activity. While this is unlikely to account for our results (most dyeloaded cells responded with large fluorescence changes when they were depolarized with high-K+ solutions), we cannot exclude the possibility that Optical imaging of motor activity some interneurons are not loaded with the calcium in the spinal cord indicator or do not generate adequate fluorescent To obtain more detailed information about the signals to be detected by our present optical system. The active cells outside the lateral motor column cells that participate in motor activity, we devel-

To investigate these issues we acutely ablated sections of the lumbar cord in the recording chamber while monitoring activity from the hindlimb muscle nerves (see Materials and Methods). Initially we were interested in establishing if the rhythmogenic capacity was uniformly distributed along the rostrocaudal axis of the lumbosacral cord or whether there was any evidence for regionalization. For this purpose, the lumbosacral cord of E10-El1 embryos was divided into a rostral (T7-LS3) and a caudal section (LS4-LS8). The results showed that both parts were capable of generating cyclic activity, indicating that the rhythm-generating capacity was distributed along the rostrocaudal axis of the cord. However, we found that the rostral part of the cord was capable of generating more cycles/ episode than the caudal part of the cord (6.4/episode vs 1.8/episode;n = 4). These findings raise the possibility of a difference in the rhythm-generating mechanisms in the rostral and caudal lumbosacral cord, although other factors such as network excitability or the extent of deafferentation might also play a role. To investigate the possibility that the premotor networks might be concentrated in the ventral part of the cord, as suggested by earlier work (Hamburger et al., ,661, we removed the dorsal part of the cord over several segments. In such preparations we found that motoneurons could still be activated rhythmically, although the number of cycles in each episode was reduced (Fig. 3). Remarkably, however, the pattern of discharge recorded from the sartorius and femorotibialis nerves showed little change (Fig. 3B). When the medial part of the cord was removed, the pattern of activity remained unaltered, providing that the lesion did not encroach on the lateral motor column. These findings suggest that the ventrolateral part of the cord contains the necessary neural elements for rhythmogenesis and for the appropriate phasing of activity in femorotibialis and sartorius motoneurons. We cannot exclude the possibility that dorsal or medial neurons contribute to the activity in the unlesioned cord, although our results suggest that they are not essential for it. Moreover, rhythmic dorsal root potentials, which are present in the intact cord, disappear followingseparation of the dorsal and ventral parts of the lumbar cord, suggesting that the dorsal cells are synaptically driven by ventral neurons.

M. O'DONOVAN ET AL.

268

A

I

SART I

I

1

I

1

FE M I'

B l

I

'J

I I

Fig. 3. Ablation procedures demonstrate that the essential elements of the network generating motor activity lie close to the lateral motor column. A: Evoked motor activity recorded from an isolated section of the rostra1 part (T7-LS4) of the lumbosacral cord of an E l 0 embryo. B: The alternation between sartorius (SART) and femorotibialis (FEM) remains after re-

II

FF M

moving the dorsal part of the cord over the whole section. C: Photomicrograph of the ventral section of the cord (at LS3) that the recordings in B were made from (dorsal is up, lateral to the left). The contralateral part of the cord is shown in D to illustrate the extent of the dorsal ablation.

presumably include premotor interneurons, al- ent, however, they must be in phase with the rhyththough we cannot eliminate the possibility that mic fluorescence changes in motoneurons that are glial cells might also generate calcium signals dur- closely correlated with firing. In some experiments we detected another type ing neuronal activity (Cornell-Bell et al., '90; Glaum et al., '90). If rhythmic glial signals are pres- of signal that was spatially diffuse and was not

H

DEVELOPMENT IN CHICK

1" ' Fig. 4. Fura-2 imaging of motor network activity in the lumbosacral spinal cord. The location of the cells activated during motor activity is shown in the video micrograph of the transverse face of the spinal cord at the level of LS3. The boundaries of the grey matter and lateral motor column have been outlined (lateral to the left, dorsal up). The cord had previously been loaded with the calcium indicator dye Fura-2AM as described in Materials and Methods. The map was constructed by subtracting an image averaged (30 frames) dur-

269

I 2 sec

ing motor activity from a control image acquired just before the burst began. The timing of the frames acquired during the burst is indicated by the bar (1sec duration) over the femorotibialis neurogram, which is displayed in the box under the photomicrograph. The difference image is positive, so that brightness is proportional to the decrease in fluorescence. The range of pixel values was linearly expanded to occupy the full range (0-255) of the eight-bit digital frame store.

270

M. O'DONOVAN ET AL.

GABA

Fig. 5. Camera lucida drawings and micrographs of semithin sections immunostained for GABA and glycine. Drawings show the distribution of GABA (A)- and glycine (B)-immunoreactive perikarya, which appeared in a single semithin transverse section through one side of the lumbosacral spinal cord. C: Micrograph of GABA-immunoreactiveperikarya in the dorsal horn. D: Micrograph of glycine-immunoreactive perikarya

in the intermediate gray matter. Areas shown on C and D are framed on A and B. Borders of the gray matter and the lateral motor column are drawn with dashed lines. Arabic numerals indicate Rexed laminae of the spinal cord. The spinal laminae for the chicken spinal cord are as in Martin ('79). Bars = 100 km (A,B) and 50 Fm (C,D).

localized to individual cells. The signal occurred at the start of each cycle and took the form of a wave that spread dorsally or medially. The origin of the signal is unclear, but it could represent fluorescence changes in dye-loadedpresynaptic terminals belonging to the axons of premotor interneurons, or alternatively it could originate from the dendrites of motoneurons or interneurons. The apparent ventro-dorsal or ventromedial movement of the fluorescence change suggests that activity is initiated in the ventrolateral part of the cord, in or around the lateral motor column, and that more dorsal and medial regions are activated later. Such a view is consistent with the ablation studies indicating that the essential components of the motor circuitry are located in the ventrolateral part of the

cord. One possible interpretation of these findings is that the synaptic activation of motoneurons and interneurons in the cord face is derived from premotor axons travelling in the ventrolateral white matter tracts. We have recently determined that the cell bodies of lateral axons are located in the lateral part of the grey matter, above the motor nucleus.

Descending control of locomotor activity Experiments were performed in E6-El3 embryos to establish if descending pathways are capable of initiating rhythmic motor activity and if such effects are mediated by long descending axons. For this purpose we developed a spinal-cord brainstem

MOTOR DEVELOPMENT IN CHICK

preparation that exhibits spontaneous motor activity indistinguishable from that produced by the isolated spinal cord (see Materials and Methods). As early as E6-E7, microstimulation in the brainstem elicited episodes of brachial and lumbar motor activity that appear similar to spontaneously occurring episodes. Brainstem stimulation evoked lumbosacral motor activity when chemical synaptic transmission through the cervical and brachial cord was blocked by bath application of zero calcium solutions, which abolished pectoralis (brachial) nerve activity. These results suggest that functional connections from the brainstem to the spinal cord are established by E6-E8. Such connections could be mediated by descending axons, which are known to reach the lumbosacral cord by E5.5 (Okado and Oppenheim, '85), although it is also possible that the brainstem-evoked lumbosacral activity could result from the antidromic activation of spinal axons ascending to the brainstem or could be mediated through electrical synapses not blocked by reduced calcium solutions.

Immunocytochemical localization o fputative inhibitory spinal in terneurons In this section we present immunocytochemical results on the distribution of neurons exhibiting GABA- and glycine-immunoreactivity in E14-E 16 spinal cords. At this stage, it has not been possible to correlate the immunocytochemical findings directly with the electrophysiology or imaging. However, in future experiments we plan to combine immunocytochemistry with imaging to characterize further the rhythmically active interneurons in the developing cord. We have, therefore, included our immunocytochemical observations as a preliminary effort in this direction.GABA-immunoreactive perikarya were detected almost exclusively in the dorsal horn. Most of the somata expressing positive immunoreaction for GABA were found in laminae 11,111, and IV (Fig. 5A,C). Stained cell bodies were also revealed in a significant number in lamina V but occasionally only appeared in deeper layers (Fig. 5A). The distribution of perikarya displaying immunoreactivity for glycine was strikingly different from that of GABA-immunoreactive ones. Cell bodies immunoreactive for glycine were primarily found in laminae V, VI, and VII (Fig. 5B,D). A few labelled somata were also seen at the medial border of the lateral motor column and in laminae 111 and IV (Fig. 5B), but the rest of the ventral and dorsal horns were devoid of labelling. Further experiments are in progress to establish if this distribu-

27 1

tion is similar throughout the lumbosacral cord and whether it changes during development. CONCLUSIONS This paper has reviewed recent work on the cellular and synaptic basis of rhythmic motor activity in the developing spinal cord of the chick embryo. Intracellular recordings from motoneurons indicate that the rhythmic activity of motoneurons is a synaptic process, generated by activity in a network of premotor interneurons. Although this network appears to be distributed throughout the rostrocaudal axis of the lumbosacral cord, the imaging and ablation studies suggest that it is concentrated in the ventral cord and that its organization or mode of operation may differ in the rostra1 and caudal parts of the lumbar cord. Although we know very little about the organization of this network, our results suggest that pattern generation (the alternation of flexors and extensors) and rhythm generation may be separate mechanisms. The alternation of sartorius and femorotibialis motoneurons is brought about by a selective inhibition of sartorius motoneurons. By contrast, excitatory inputs appear to be more symmetrically distributed to both classes of motorneuron. The mechanism responsible for rhythmogenesis is less clear, although the synchronized action of both motoneurons and interneurons during rhythmic activity suggests the operation of recurrent excitation within the premotor network. Further progress in understanding the genesis of embryonic rhythmicity, and the relative importance of network and intrinsic membrane properties, will require direct identification and characterization of the interneurons involved. This is now technically possible using the patch clamp and imaging techniques described in this paper. ACKNOWLEDGMENTS We thank Drs. Ken Spring, Bill Marks, and Robert Burke for their comments and advice and Mr. George Dold for his excellent technical assistance and the construction of some of the equipment used in the experiments. LITERATURE CITED Arshavsky, Y.I., G.N. Orlovsky.,G.A. Pavlova, and L.B. Popova (1986)Activity of C3-C4 propriospinal neurons during ficticious forelimb locomotion in the cat. Brain. Res., 3633:354-357. Barry, M., and M.J. ODonovan (1987) The effects of excitatory amino acids and their antagonists on the generation of motor activity in the isolated chick cord. Dev. Brain.Res., 36:271-276. Bekoff, A. (1976) Ontogeny of leg motor output in the chick embryo:A neural analysis. Brain Res., 106:271-291.

272

M. O’DONOVAN ET AL.

Bekoff, A., J.A. Kauer, A. Fulstone, and T.R. Summers (1989) Neural control of limb coordination. 11. Hatching and walking motor output patterns in the absence of input from the brain. Exp. Brain Res., 74:609-617. Bekoff, A., P.S.G. Stein, and V. Hamburger (1975) Coordinated motor output in the hindlimb of the 7-day-old chick embryo. Proc. Natl. Acad. Sci. USA, 721245-1248. Ben-Ari, Y., E. Cherubini, R. Corradetti, and J.-L. Gaiarsa (1989) Giant synaptic potentials in immature rat CA3 hippocampal neurons, J. Physiol., 416:303-325. Blanton, M.G., J . J . Lo Turco, and A.R. Kriegstein (1989) Whole cell recording from neurons in slices of reptilian and mammalian cerebral cortex. J. Neurosci. Methods, 30:203210. Cornell-Bell, A.H., S.M. Finkbeiner, M.S. Cooper, and S.J. Smith (1990) Glutamate induces calcium waves in cultured astrocytes: Long- range glial signaling. Science, 247:470473. Currie, S.N., and P.S.G. Stein (1990) Cutaneous stimulation evokes long-lasting excitation of spinal interneurons in the turtle. J. Neurophysiol, 64:1134-1148. Glaum, S.R., J.A. Holzwarth, and R.J. Miller (1990) Glutamate receptors activate Ca2+ mobilization and Ca2+ influx into astrocytes. Proc. Natl. Acad. Sci USA, 87:34543458. Grynkiewicz, G., M. Poenie, and R. Tsein (1985) A new generation of Ca2+ indicators with greatly improved fluorescence properties. J. Biol. Chem. 260:3440-3450. Hamburger, V., E. Wenger, and R. Oppenheim (1966) Motility in the chick embryo in the absence of sensory input. J . Exp. Zool., 162.133-155. Ho, S., and M.J. O’Donovan (1990) The effects of selective ablations on the rhythmic motor output of the isolated chick spinal cord. SOC.Neurosci. Abstr., 16:114. Jacobson, R.D., and M. Hollyday (1982a) A behavioral and electromyographic study of walking in the chick. J . Neurophysiol., 48:238-256. Jacobson, R.D., and M. Hollyday (1982b) Electrically evoked walking and fictive locomotion in the chick. J . Neurophysiol., 48:257-270. Landmesser, L.T., and M.J. O’Donovan (1984) Activation patterns of embryonic chick hindlimb muscles recorded in ovo and in an isolated spinal cord preparation. J. Physiol., 34 7: 189-204. Lee, M.T., M.J. Koebbe, and M.J. ODonovan (1988) The development of sensorimotor synaptic connections in the lumbosacral spinal cord of the chick embryo. J. Neurosci., 8: 2530-2543. Lee, M.T., and M.J. O’Donovan (1991) Organization of hindlimb muscle afferent projections to lumbosacral motoneurons in the chick embryo. J. Neurosci., in press. Martin, A.H. (1979) A cytoarchitectonic scheme for the spinal cord of the domestic fowl, Gallus gallus domesticus: Lumbar region. Acta Morphol. Neerl. Scand. 17.105-117. McCobb, D.P., P.M. Best, and K.G. Beam (1989) Development alters the expression of calcium currents in chick limb motoneurons. Neuron, 2.1633-1643. McLennan, I.S. (1983) Neural dependence and independence of myotube production in chicken hindlimb muscles. Dev. Biol., 98:287-294. O’Donovan, M.J. (1989) Motor activity in the isolated spinal cord of the chick embryo: Synaptic drive and firing pattern of single motoneurons. J. Neurosci., 9:943-958. O’Donovan, M., W. Yee, and M. Antal(1990) Real time optical

detection of rhythmic motoneuron and interneuron activity in the isolated spinal cord using digital imaging of calcium transients. SOC.Neurosci. Abstr., 16: 1132. O’Donovan, M.J., and L.T. Landmesser (1987) The development of hindlimb motor activity studied in an isolated preparation of the chick spinal cord. J. Neurosci., 7:32563264. Okado, N., and R.W. Oppenheim (1985) The onset and development of descending pathways to the spinal cord in the chick embryo. J. Comp. Neurol., 232:143-161. Pittman, R., and R.W. Oppenheim (1978)Neuromuscular blockage increases motoneuron survival during normal cell death in the chick embryo. Nature. 271:364-161. Pittman, R., and R.N. Oppenheim (1979) Cell death of motoneurons in the chick embryo spinal cord. IV. Evidence that a functional neuromuscular interaction is involved in the regulation of naturally occurring cell death and the stabilization by synapses. J . Comp. Neurol. 187:425-446. Provine, R.R. (1971)Embryonic spinal cord: Synchronized spatial distribution of polyneuronal burst discharges. Brain. Res., 29:155-158. Provine, R.R. (1972) Ontogeny of bioelectric activity in the spinal cord of the chick embryo and its behavioral implications. Brain Res, 41 :365-378. Provine, R.R. (1973) Neurophysiological aspects of behavior development in the chick embryo. In Behavioral Embryology, Vol l.,G. Gottlieb, ed. Academic Press, New York, pp. 77-102. Renaud, D., G.H. LeDourain, and A. Khaskiye (1978) Spinal cord stimulation in chick embryo: Effects on development of the posterior latissimus dorsi muscle and neuromuscular junctions. Exp. Neurol., 60:189-200. Ripley, K.L., and R.R. Provine (1972) Neural correlates of embryonic motility in the chick. Brain Res., 45:127-134. Sernagor, E., and M.J. ODonovan (1991)Whole cell patch clamp of rhythmically active motoneurons in the isolated spinal cord of the chick embryo. Neurosci. Lett., 128:211-216. Sharma, S.C., R.R. Provine, V. Hamburger, and T. Sandel (1970) Unit activity in the isolated spinal cord of the chick embryo in situ. Proc. Natl. Acad. Sci. USA, 66:40-47. Shefchyk, S., D. McCrea, D. Kriellaars, P. Fortier, and L. Jordan (1990) Activity of interneurons within the L4 spinal segment of the cat during brainstem-evoked fictive locomotion. Exp. Brain Res., 80:290-295. Sholomenko, G., and M.J. O’Donovan (1990) Hindlimb motor activity elicited by brainstem stimulation in an isolated brainstemlspinal cord preparation of the chick embryo. SOC. Neurosci. Abstr., 16:118. Sholomenko, G.N., and J.D. Steeves (1987) Effects of selective spinal cord lesions on hind limb locomotion in birds. Exp. Neurol., 95r403-418. Somogyi, P., and A.J. Hodgson (1985) Antisera to y-aminobutyric acid. 111. Demonstration of GABA in Golgi-impregnated neurons and in conventional electron microscopic sections of cat striate cortex. J . Histochem. Cytochem., 33: 249-257. Srihari, T., and G. Vrbova (1978) The role of muscle activity in the differentiation of neuromuscular junctions in slow and fast chick muscles. J. Neurocytol., 7~529-540. Takahashi, T. (1984) Inhibitory miniature synaptic potentials in rat motoneurons. Proc. R. SOC.London Biol., 221:103109.

MOTOR DEVELOPMENT IN CHICK Toutant, J.P., M.N. Toutant, D. Renaud, and G.H. Le Douarin (1979) Enzymatic differentiation of muscle fiber types in embryonic latissimus dorsi ofthe chick: Effects of spinal cord stimulation. Cell. Differ., 8:375-382. Velumian, A.A. (1981)Intracellular recordings from chick mob-

273

neurons in the isolated perfused spinal cord. Brain Res. 229502-506. Watson, S.J., and A. Bekoff (1990) A kinematic analysis of hindlimb motility in 9- and 10-day-old chick embryos. J. Neurobiol., 22:651-660.