Inhibitory effects of L-glutamate on central processes of crustacean leg motoneurons

Profiling neurotransmitters in a crustacean neural circuit for locomotion

Locomotor systems are widely used to study rhythmically active neural networks. These networks have to be coordinated in order to produce meaningful behavior. The crayfish swimmeret system is well suited to investigate such coordination of distributed neural oscillators because the neurons and their connectivity for generating and especially for coordinating the motor output are identified. The system maintains a fixed phase lag between the segmental oscillators, independent of cycle period. To further the understanding of the system’s plasticity for keeping the phase lag fixed, we profiled the neurotransmitters used by the Coordinating Neurons, which are necessary and sufficient for coordination of the segmental oscillators. We used a combination of electrophysiological, immunohistochemical, and mass spectrometric methods. This arrangement of methods ensured that we could screen for several specific neurotransmitters, since a single method is often not suitable for all neurotransmitters of interest. In a first step, to preselect neurotransmitter candidates, we investigated the effect of substances known to be present in some swimmeret system neurons on the motor output and coordination. Subsequently, we demonstrated electrophysiologically that the identified synapse between the Coordinating Neurons and their target is mainly chemical, but neither glutamate antagonist nor γ-aminobutyric acid antagonist application affected this synapse. With immunohistochemical experiments, we provide strong evidence that the Coordinating Neurons are not serotonergic. Single-cell MALDI-TOF mass spectrometry with subsequent principal component analysis identified acetylcholine as the putative neurotransmitter for both types of Coordinating Neurons.

Invertebrate preparations and their contribution to neurobiology in the second half of the 20th century

This review summarized the contribution to neurobiology achieved through the use of invertebrate preparations in the second half of the 20th century. This fascinating period was preceded by pioneers who explored a wide variety of invertebrate phyla and developed various preparations appropriate for electrophysiological studies. Their work advanced general knowledge about neuronal properties (dendritic, somatic, and axonal excitability; pre- and postsynaptic mechanisms). The study of invertebrates made it possible to identify cell bodies in different ganglia, and monitor their operation in the course of behavior. In the 1970s, the details of central neural circuits in worms, molluscs, insects, and crustaceans were characterized for the first time and well before equivalent findings were made in vertebrate preparations. The concept and nature of a central pattern generator (CPG) have been studied in detail, and the stomatogastric nervous system (STNS) is a fine example, having led to many major developments since it was first examined. The final part of the review is a discussion of recent neuroethological studies that have addressed simple cognitive functions and confirmed the utility of invertebrate models. After presenting our invertebrate "mice," the worm Caenorhabditis elegans and the fruit fly Drosophila melanogaster, our conclusion, based on arguments very different from those used fifty years ago, is that invertebrate models are still essential for acquiring insight into the complexity of the brain.

GABAergic and glutamatergic inhibition of nonspiking local interneurons in the terminal abdominal ganglion of the crayfish

Nonspiking local interneurons in the terminal abdominal ganglion of the crayfish Procambarus clarkii receive inhibitory inputs from mainly glutamatergic spiking local interneurons and GABAergic nonspiking interneurons. In this study, the inhibitory responses of nonspiking interneurons to local application of glutamate and GABA into the neuropil were compared. Glutamate and GABA injection mediated the hyperpolarization of the nonspiking interneurons with an increase in membrane conductance. The glutamate-mediated membrane hyperpolarization was reversed by injection of 1 or 2 nA hyperpolarizing current. By contrast, more than 3 nA hyperpolarizing current was frequently necessary to reverse the GABA-mediated hyperpolarization. Bath application of a chloride channel blocker, 50 microM picrotoxin (PTX), reduced the glutamate-mediated hyperpolarization, but had no effect on the GABA-mediated hyperpolarization. The GABA-mediated hyperpolarization was not consistently affected by bath application of low chloride solution. These results suggest that the glutamate-mediated inhibition was related to the gating of a Cl(-) conductance, while the GABA-mediated inhibition was not. Electrical stimulation of sensory afferents innervating the exopodite elicited ipsps in uropod opener motor neurons. These sensory-evoked ipsps were also PTX-insensitive, suggesting GABAergic nonspiking interneurons could be the predominant premotor elements in organizing the uropod motor control system.

Distribution of glutamatergic immunoreactive neurons in the terminal abdominal ganglion of the crayfish

Using an antiserum directed against glutamate, we have analyzed the distribution of glutamate-like immunoreactive neurons in the terminal abdominal ganglion of the crayfish Procambarus clarkii. Approximately 160 central neurons (157 +/- 8; mean +/- SEM, n = 8) showed positive glutamate-like immunoreactivity, which represents approximately 25% of the total number of neurons in the terminal ganglion. Using a combination of intracellular staining with the marker Lucifer yellow and immunocytochemical staining has shown that most excitatory motor neurons are glutamatergic and that glutamate acts as an excitatory transmitter at peripheral neuromuscular junctions. Seven of 10 identified spiking local interneurons and only 2 of 19 identified ascending interneurons, showed positive immunoreactivity. Our observation that inhibitory spiking interneurons were immunopositive, whereas excitatory ascending interneurons were immunonegative, indicates that glutamate is likely to act as an inhibitory neurotransmitter within the central nervous system. Local pressure injection of L-glutamate into the neuropil of the ganglion caused a hyperpolarization of the membrane potentials of many interneurons. gamma-Aminobutyric acid (GABA)ergic posterolateral nonspiking interneurons and the bilateral nonspiking interneuron LDS showed no glutamate-like immunoreactivity, whereas non-GABAergic anterolateral III nonspiking interneurons showed glutamate-like immunoreactivity. Thus, not only GABA but also glutamate are used in parallel as inhibitory neurotransmitters at central synapses.

Inhibitory component of the resistance reflex in the locomotor network of the crayfish

The aim of this study was to investigate the inhibitory components of a resistance reflex in the walking system of the crayfish. This study was performed using an in vitro preparation of several thoracic ganglia including motor nerves and the proprioceptor that codes movements of the second joint (coxo-basipodite chordotonal organ-CBCO). Sinusoidal movements were imposed on the CBCO, and intracellular responses were recorded from levator (Lev) and depressor (Dep) motoneurons (MNs). We found that in MNs that oppose the imposed movements (e.g., the Lev MNs during the imposed downward movement), the response consists in a depolarization resulting from the summation of excitatory postsynaptic potentials (EPSPs). A movement in the opposite direction resulted in hyperpolarization during which inhibitory postsynaptic potentials (IPSPs) summated. The inhibitory pathway to each MN is oligosynaptic (i.e., composed of a small number of neurons in series) and involves spiking interneurons because it was blocked in the presence of a high-divalent cation solution. The IPSPs were mediated by a chloride conductance because their amplitude was sensitive to the chloride concentration of the bathing solution and because they were blocked by the chloride channel blocker, picrotoxin. Resistance reflex IPSPs related to single CBCO neurons could be identified. These unitary IPSPs were blocked in the presence of 3-mercapto-propionic acid, an inhibitor of gamma-amino-butyric acid (GABA) synthesis, indicating that they are mediated by GABA. In addition to this GABAergic pathway, electrical stimulation of the CBCO sensory nerve induced compound IPSPs that were blocked by glutamate pyruvate transaminase (GPT), indicating the presence of glutamatergic inhibitory pathways. These glutamatergic interneurons do not appear to be involved in the resistance reflex, however, as GPT did not block the unitary IPSPs. Functionally, the resistance reflex is mainly supported by movement-coding CBCO sensory neurons. We demonstrate that such movement-coding CBCO neurons produce both monosynaptic EPSPs in the MNs opposing imposed movements and oligosynaptic IPSPs in the antagonistic motoneurons. These results highlight the similarities between the inhibitory pathways in resistance reflex of the crayfish and in the stretch reflex of vertebrates mediated by Ia inhibitory interneurons.

Presynaptic modulation of sensory neurons in the segmental ganglia of arthropods

The afferent terminals of arthropod sensory neurones receive abundant input synapses, usually closely intermingled with the sites of synaptic output. The majority of the input synapses use the neurotransmitter GABA, but in some afferents there is a significant glutamatergic or histaminergic component. GABA and histamine shunt afferent action potentials by increasing chloride conductance. Though glutamate can also have this effect in the arthropod central nervous system, its action on afferent terminals appears to be mediated by increases in potassium conductance or by the action of metabotropic receptors. The action of the presynaptic synapses on the afferents are many and varied. Even on the same afferent, they may have several distinct roles that can involve both tonic and phasic patterns of primary afferent depolarisation. Despite the ubiquity and importance of their effects however, the populations of neurones from which the presynaptic synapses are made, remain largely unidentified.

Adaptive motor control in crayfish

This article reviews the principles that rule the organization of motor commands that have been described over the past five decades in crayfish. The adaptation of motor behaviors requires the integration of sensory cues into the motor command. The respective roles of central neural networks and sensory feedback are presented in the order of increasing complexity. The simplest circuits described are those involved in the control of a single joint during posture (negative feedback-resistance reflex) and movement (modulation of sensory feedback and reversal of the reflex into an assistance reflex). More complex integration is required to solve problems of coordination of joint movements in a pluri-segmental appendage, and coordination of different limbs and different motor systems. In addition, beyond the question of mechanical fitting, the motor command must be appropriate to the behavioral context. Therefore, sensory information is used also to select adequate motor programs. A last aspect of adaptability concerns the possibility of neural networks to change their properties either temporarily (such on-line modulation exerted, for example, by presynaptic mechanisms) or more permanently (such as plastic changes that modify the synaptic efficacy). Finally, the question of how "automatic" local component networks are controlled by descending pathways, in order to achieve behaviors, is discussed.

Glutamate is a transmitter that mediates inhibition at the rectifying electrical motor giant synapse in the crayfish

Spike transmission at the electrical synapse between the giant fibres (GFs) and motor giant neurone (MoG) in the crayfish can be blocked by depolarising postsynaptic chemical inhibition, which has previously been shown to be mediated in part by gamma-aminobutyric acid (GABA). The authors show that glutamate applied to the synaptic region of the MoG mimics the depolarisation of the chemical input and can also block spike transmission from the GFs. The glutamate induces an inward current mediated by a conductance increase that is 30-40% of that induced by GABA and that is blocked substantially by picrotoxin. Glutamate has no effect on the presynaptic GF, and the effects in the MoG are maintained in the presence of cadmium, indicating that the glutamate is acting directly on the MoG. Both GABA and glutamate have similar effects on the cell body, where the response reverses 10-20 mV positive to resting potential, is dependent on chloride concentration, and is inhibited by picrotoxin. Joint application of glutamate and GABA induces a nonadditive current under voltage clamp, suggesting that the transmitters can activate the same postsynaptic receptors. Immunocytochemical staining shows that, whereas some synaptic profiles impinging on the MoG contain pleomorphic agranular vesicles and are immunoreactive to GABA and not glutamate (as previously reported), there are at least as many other profiles that contain round, agranular vesicles and that are immunoreactive to glutamate and not to GABA. Thus, the authors conclude that some of the interneurones mediating inhibition of the electrical synapse use glutamate as their neurotransmitter.

GABA and glutamate-like immunoreactivity at synapses on depressor motorneurones of the leg of the crayfish, Procambarus clarkii

To investigate their synaptic relationships, depressor motorneurones of the crayfish leg were impaled with microelectrodes, intracellularly injected with horseradish peroxidase, and prepared for electron microscopy. Post-embedding immunogold labelling with antibodies against gamma-aminobutyric acid (GABA) or glutamate was carried out either alone or together on the same section and allowed the identification of three classes of input synapses: 51% were immunoreactive for glutamate and contained round agranular vesicles, 31% were immunoreactive for GABA and contained pleomorphic agranular vesicles, and the remainder were immunoreactive for neither and also predominantly contained pleomorphic agranular vesicles. Output synapses were abundant in some of the motorneurones but were not seen in others, suggesting that members of the motor pool differ in their connectivity.

Central control components of a 'simple' stretch reflex

The monosynaptic stretch reflex is a fundamental feature of sensory-motor organization in most animal groups. In isolation, it serves largely as a negative feedback devoted to postural controls; however, when it is involved in diverse movements, it can be modified by central command circuits. In order to understand the implications of such modifications, a model system has been chosen that has been studied at many different levels: the crayfish walking system. Recent studies have revealed several levels of control and modulation (for example, at the levels of the sensory afferent and the output synapse from the sensory afferent, and via changes in the membrane properties of the postsynaptic neuron) that operate complex and highly adaptive sensory-motor processing. During a given motor task, such mechanisms reshape the sensory message completely, such that the stretch reflex becomes a part of the central motor command.

Glial hyperpolarization upon nerve root stimulation in the leech Hirudo medicinalis

Hyperpolarizing responses in neuropil glial cells evoked by nerve root stimulation were studied in the central nervous system of the leech Hirudo medicinalis using intracellular recording and extracellular stimulation techniques. From a mean resting potential of -60.5 +/- 1.0, the glial membrane was hyperpolarized by -8.6 +/- 0.8 mV, via stimulation of the dorsal posterior nerve root in an isolated ganglion. Nerve root stimulation evoked biphasic or depolarizing responses in glial cells with resting potentials around -70 mV (Rose CR, Deitmer JW. J. Neurophysiol. 73:125-131, 1995). The hyperpolarizing response was reduced by the ionotropic glutamate receptor antagonist CNQX (50 microM) to 58% of its initial amplitude. In 15 mM Ca2+/15 mM Mg(2+)-saline the hyperpolarization was reduced by 44%. The hyperpolarization that persisted in high-divalent cation saline was not affected by CNQX. Bath-applied glutamate (500 microM) and kainate (2 microM) elicited glial hyperpolarizations that were sensitive to CNQX and 10 mM Mg2+/1 mM Ca(2+)-saline. The 5-HT-antagonist methysergide did not affect the hyperpolarizations evoked by nerve root stimulation. The results show that in the leech glial membrane responses to neuronal activity include not only depolarizations, as shown previously, but also hyperpolarizations, which are mediated by direct and indirect neuron-glial communication pathways. In the indirect pathway, glutamate is a transmitter between neurons.

Distribution of input and output synapses on the central branches of bushcricket and cricket auditory afferent neurones: immunocytochemical evidence for GABA and glutamate in different populations of presynaptic boutons

In order to investigate the synapses on the terminals of primary auditory afferents in the bushcricket and cricket, these were impaled with microelectrodes and after physiological characterisation, injected intracellularly with horseradish peroxidase. The tissue was prepared for electron microscopy, and immunocytochemistry for gamma-aminobutyric acid (GABA) and glutamate was carried out on ultrathin sections by using a post-embedding immunogold technique. The afferent terminals received many input synapses. Between 60-65% of these were made by processes immunoreactive for GABA and approximately 25% from processes immunoreactive for glutamate. The relative distribution of the different classes of input were analysed from serial section reconstruction of terminal afferent branches. Inputs from GABA and glutamate-immunoreactive processes appeared to be scattered at random over the terminal arborisation of the afferents both with respect to each other and to the architecture of the terminals. They were, however, always found close to the output synapses. The possible roles of presynaptic inhibition in the auditory afferents is discussed in the context of the auditory responses of the animals.

Neuromodulation of reciprocal glutamatergic inhibition between antagonistic motoneurons by 5-hydroxytryptamine (5-HT) in crayfish walking system

In an in vitro preparation of the crayfish thoracic locomotor system, paired intracellular recordings were performed from antagonistic depressor (Dep) and levator (Lev) motoneurons (MNs) that control the second joint of walking legs. Connections between these two groups of MNs consist mainly of inhibitory connections and weak electrotonic synapses. Injection of depolarizing current into a Lev MN results in a hyperpolarization in a Dep MN, and vice versa. This reciprocal glutamatergic inhibition, is not changed in the presence of the sodium channel blocker tetrodotoxin (TTX) and therefore is likely supported by a direct connection between MNs. By contrast, reciprocal inhibition is largely reduced in the presence of 5-hydroxytryptamine (5-HT; 10 microM). Direct micro-application of glutamate pressure-ejected close to an intracellularly recorded MN, evoked an inhibitory response in that MN, accompanied by a decrease of input resistance. These two effects were dramatically reduced in the presence of 5-HT. Thus 5-HT could be involved in mechanisms of dynamic reconfigurations of the neural network controlling leg movements in crayfish.

Distribution of GABAergic premotor nonspiking local interneurones in the terminal abdominal ganglion of the crayfish

The inhibitory neurotransmitter of premotor nonspiking local interneurones in the crayfish terminal abdominal ganglion was investigated physiologically and immunocytochemically. Depolarization of a nonspiking interneurone evoked a hyperpolarization in a uropod motor neurone. The amplitude of hyperpolarization in the motor neurone was gradually decreased under low-calcium/high-magnesium saline. Local pressure injection of gamma-aminobutyric acid (GABA) into the neuropil caused a similar hyperpolarization of the motor neurone. These physiological studies suggested a GABAergic inhibitory interaction between nonspiking interneurones and the motor neurones. Premotor nonspiking interneurones are classified into two subgroups ofposterolateral (PL) and anterolateral (AL) interneurones, and AL interneurones are further divided into three subtypes. A combination of intracellular staining from nonspiking local interneurones with Lucifer yellow and immunocytochemical staining with an antiserum directed against GABA revealed that all the PL interneurones sampled in this study showed GABA-like immunoreactivity. A population of cell bodies (n = 6-11) with a small diameter (15-30 microm) packed together forming a cluster showed GABA-like immunoreactivity, and the cell bodies of most PL interneurones were found in this cluster. To compare the number and the pattern of main branches of PL interneurones, cells were classified into three identifiable sets of interneurones, called PL-1, PL-2, and PL-3. By contrast, about one-half of AL interneurones, especially the third subtype of AL interneurones, which have cell bodies located ventrolaterally in the ganglion, did not show GABA-like immunoreactivity. Furthermore, the position of cell bodies of GABA-immunoreactive AL interneurones was scattered compared to that of PL interneurones.

Two types of identified ascending interneurons with distinct GABA receptors in the crayfish terminal abdominal ganglion

More than half of the identified ascending interneurons originating in the terminal abdominal ganglion of the crayfish received inhibitory sensory inputs from hair afferents innervating the tailfan on the side contralateral to their main branches. Biochemical aspects of this transverse lateral inhibition of ascending interneurons were examined by the use of neurophysiological and pharmacological techniques. Local application of gamma-aminobutyric acid (GABA) and its agonist muscimol into the neuropil induced membrane hyperpolarization of identified ascending interneurons with an increase in membrane conductance. Because the reversal potential of inhibitory postsynaptic potential (IPSPs) in ascending interneurons elicited by the sensory stimulation and GABA injection was similar, and the sensory-stimulated IPSPs of the interneurons were blocked by GABA and muscimol application, this study strongly suggests a GABAergic nature for transverse lateral inhibition of ascending interneurons. According to the response to the GABAA antagonists bicuculline and picrotoxin, ascending interneurons were classified into two types, picrotoxin-sensitive and picrotoxin-insensitive interneurons. Identified ascending interneurons VE-1 and RO-4 showed a pharmacological profile similar to that of the classical GABAA receptor of the vertebrates. Bath application of both bicuculline and picrotoxin reversibly reduced the amplitudes of IPSPs. The other identified ascending interneurons CA-1, RO-1, and RO-2 were not affected significantly by the bath application of GABAA and GABAB antagonists, although bath application of low-chloride saline reversed the sensory-stimulated IPSPs. IPSPs of the picrotoxin-sensitive interneurons had a rather faster time course and shorter duration in comparison with those of the picrotoxin-insensitive interneurons.

Inhibitory glutamate receptor channels

Inhibitory glutamate receptors (IGluRs) are a family of ion channel proteins closely related to ionotropic glycine and gamma-aminobutyric acid (GABA) receptors; They are gated directly by glutamate; the open channel is permeable to chloride and sometimes potassium. Physiologically and pharmacologically, IGluRs most closely resemble GABA receptors; they are picrotoxin-sensitive and sometimes crossdesensitized by GABA. However, the amino acid sequences of cloned IGluRs are most similar to those of glycine receptors. Ibotenic acid, a conformationally restricted glutamate analog closely related to muscimol, activates all IGluRs. Quisqualate is not an IGluR agonist except among pulmonate molluscs and for a unique multiagonist receptor in the crayfish Austropotamobius torrentium. Other excitatory amino acid agonists are generally ineffective. Avermectins have several effects on IGluRs, depending on concentration: potentiation, direct gating, and blockade, both reversible and irreversible. Since IGluRs have only been clearly described in protostomes and pseudocoelomates, these effects may mediate the powerful antihelminthic and insecticidal action of avermectins, while explaining their low toxicity to mammals. IGluRs mediate synaptic inhibition in neurons and are expressed extrajunctionally in striated muscles. The presence of IGluRs in a neuron or muscle is independent of the presence or absence of excitatory glutamate receptors or GABA receptors in the cell. Generally, extrajunctional IGluRs in muscle have a higher sensitivity to glutamate than do neuronal synaptic receptors. Some extrajunctional receptors are sensitive in the range of circulating plasma glutamate levels, suggesting a role for IGluRs in regulating muscle excitability The divergence of the IGlu/GABA/Gly/ACh receptor superfamily in protostomes could become a powerful model system for adaptive molecular evolution. Physiologically and pharmacologically, protostome receptors are considerably more diverse than their vertebrate counterparts. Antagonist profiles are only loosely correlated with agonist profiles (e.g., curare-sensitive GABA receptors, bicuculline-sensitive AChRs), and pharmacologically identical receptors may be either excitatory or inhibitory, and permeable to different ions. The assumption that agonist sensitivity reliably connotes discrete, homologous receptor families is contraindicated. Protostome ionotropic receptors are highly diverse and straightforward to assay; they provide an excellent system in which to study and integrate fundamental questions in molecular evolution and adaptation.

Molecular biology and electrophysiology of glutamate-gated chloride channels of invertebrates

In this chapter we summarize the available data on a novel class of ligand-gated anion channels that are gated by the neurotransmitter glutamate. Glutamate is classically thought to be a stimulatory neurotransmitter, however, studies in invertebrates have proven that glutamate also functions as an inhibitory ligand. The bulk of studies conducted in vivo have been on insects and crustaceans, where glutamate was first postulated to act on H-receptors resulting in a hyperpolarizing response to glutamate. Recently, glutamate-gated chloride channels have been cloned from several nematodes and Drosophila. The pharmacology and electrophysiological properties of these channels have been studied by expression in Xenopus oocytes. Studies on the cloned channels demonstrate that the invertebrate glutamate-gated chloride channels are the H-receptors and represent important targets for the antiparasitic avermectins.

Cholinergic control of the walking network in the crayfish Procambarus clarkii

The output of a neuronal network results generally from both the properties of the component neurons and their synaptic relationships. This article aims at synthesizing various results obtained on the neural network generating locomotion in vitro. In the preparation used, consisting of the last three thoracic ganglia (3-5) along with motor nerves from the 5th leg ganglion to the promotor, remotor, levator and depressor muscles, motor nerve recordings generally revealed only tonic activity in several different motoneurons (MNs). However, rhythmic activity can be obtained by the use of cholinergic agents such as the oxotremorine (Oxo) superfused in the bath (5 x 10(-5) M). If Oxo is pressure-ejected locally in the ganglion, it is possible, depending upon the locus where the drug is applied, to elicit a rhythmic activity restricted to a group of antagonistic MNs. To analyze how cholinergic agents are able to induce such rhythmic activity, very small volumes of drug (50-200 pl), were applied close to the recording electrode. Two types of depolarizing response occurred: a fast large amplitude depolarization (5-20 mV) and a long lasting (10s to several minutes) low amplitude depolarization (1-3 mV). These responses persisted in the presence of TTX and Co(2)+. The transient initial depolarization is a mixed nicotinic and muscarinic voltage-independent response during which the input resistance decreases by 20 to 40%. In contrast, the long lasting component is voltage-dependent, exclusively muscarinic and associated to a 5-10% increase of input resistance due to the closing of a K+ conductance that is active at the resting Vm, and totally suppressed at holding potentials below -70 mV. More generally, K+ currents activated at resting potential are responsible for membrane potential stability. The injection of TEA, a blocker of the K+ currents, through the recording electrode is able to unmask plateaus above a threshold depolarization. These plateaus are TTX-sensitive but persist in the presence of Ca(2)+ channel blockers. Moreover, in 10% of TEA-filled MNs a spontaneous pacemaker activity was revealed. The organization of the locomotor network is also based upon connections between MNs and INs. Within a MN pool, connections are only loosely established, appearing to consist mainly of electrical coupling. Inhibitory synaptic connections between MNs of opposite pools are mediated by chloride channels. However, the neurotransmitter involved could be either GABA or glutamate. Therefore, at the level of a given joint, a basic rhythm occurs due to both motoneuronal membrane properties and motoneuronal connectivity. However, the coordination of all MNs of an entire leg during fictive walking activity requires the involvement of INs. Based upon these data, we propose a two-stage model of the locomotor network organization: a joint motoneuronal level and a whole leg interneuronal level.