Anatomical and Physiological Identification of Inhibitors of the Motor Giant and Segmental Giant Neurones In The Crayfish

The neuronal circuitry driving the escape tail-flip of Crustacea has been extensively studied at the level of single identified neurones, and it is one of the best understood ‘simple systems’ in the animal kingdom (see, for example, Wine and Krasne, 1982; Wine, 1984). The central feature of the circuit is the two pairs of giant fibre (GF) command neurones, the lateral giant (LG) and medial giant (MG). A spike in the former initiates a tail flexion causing an upward and forward movement of the whole animal, while a spike in the latter initiates a tail flexion causing a backwardly directed movement of the animal. In the abdomen, the GFs make major output to only two classes of neurone, the motor giant (MoG; Furshpan and Potter, 1959a) and segmental giant (SG; Kramer et al. 1981). In each case, the output is through rectifying electrical synapses. The MoG is a large abdominal flexor motor neurone, which provides the most important path for flexor motor activity. The SG acts as an interneurone transmitting excitation to the non-giant fast flexor (FF) motor neurones (Roberts et al. 1982). The SG is unusual in that, although it has an axon within the first root (R1), this axon is blind-ending and has no output (Heitler and Darrig, 1986).

The only known excitatory input to the MoG and SG in pre-terminal ganglia is from the GFs, but both the MoG and the SG neurones receive considerable inhibitory input. This inhibition exclusively takes the form of depolarising inhibitory postsynaptic potentials (dIPSPs; e.g. Furshpan and Potter, 1959b). One frequent source of inhibitory input is feedforward delayed inhibition from the GFs to the MoG (Wine, 1977). The function of this appears to be to prevent multiple spiking of the MoG in response to GF activation. Other inhibitory input is unrelated to GF activity (Fig. 1).

The only MoG inhibitor (MoG-I) that has been fully identified is a local interneurone in the terminal ganglion (G6), which inhibits the MoG in that ganglion (Kirk et al. 1986). This inhibitor, called the ‘C’ neurone because of its shape, is tightly dye-coupled to its contralateral homologue. It receives input from the SG in G6 and also from the corollary discharge interneurones I2 and I3, which originate in G2 and G3 respectively. In pre-terminal ganglia there are thought to be several MoG-I neurones, including some through-conducting interneurones, but they have not been anatomically identified (Wine, 1977). One source of the delayed inhibition of the MoG following a GF spike has been suggested to be the FF motor neurones (Wine, 1977), which are supposed to make central excitatory output to MoG-Is. The evidence for this is that extracellular stimulation of the third root (R3), which contains the axons of the FFs but no sensory neurones, can elicit dIPSPs in the MoG.

In this paper we describe the anatomy and physiology of several MoG-Is, including the pre-terminal homologues of the ‘C’ neurone, which have been physiologically and anatomically identified by microelectrode recording and staining. We show that one of the paths of MoG inhibition is indeed through the SG, emphasising the role of the latter as a ‘driver’ interneurone. We further show that there are interneurones that make inhibitory output to both the MoG and SG, and that it is unlikely that the path of excitation of the MoG-Is includes the FF motor neurones.

Experiments were performed on the central nervous system of the crayfish Pacifastacus leniusculus (Dana). Animals were obtained from Riversdale Farm (Stour Provost, Near Gillingham, Dorset, UK). The nerve cord was dissected as follows. An animal was anaesthetised by cooling on ice for 20min, and then decapitated. The legs and chelae were severed at the autotomy plane. The dorsal carapace and viscera were removed, and the nerve cord was dissected from the animal. The last two thoracic ganglia and the entire abdominal chain of ganglia were removed. The nervous tissue was then pinned dorsal surface upwards on a Sylgard platform and submerged in Van Harreveld’s crayfish saline.

Bipolar hook electrodes were used to record extracellularly and to stimulate the connectives. Pin electrodes were used to record extracellularly and to stimulate the roots, with a paired indifferent electrode for each pin placed adjacent to it in the preparation bath. Intracellular recordings were made with glass microelectrodes (resistance 15–40 MΩ) filled either with 5% Lucifer Yellow dissolved in 1mol l⁻¹ lithium chloride or with 2mol l⁻¹ potassium acetate. Penetrations were all made from the dorsal aspect of the ganglion into axonal or neuropile processes. The ganglionic sheath was removed prior to penetration.

Neurones were injected with Lucifer Yellow using 0.5s negative current pulses of 10–20nA delivered at 1Hz for up to 1h. Preparations were fixed in 5% formaldehyde, dehydrated in alcohol, cleared in methyl salicylate and visualized with a standard epifluorescent microscope.

We have anatomically and physiologically identified four classes of interneurones which inhibit the MoG specifically and one class which inhibits both the MoG and the SG.

We have identified a putative homologue of the G6 ‘C’ neurone (Kirk et al. 1986) in G2 (6 preparations) and G3 (14 preparations). We have not examined other pre-terminal abdominal ganglia. Because the name ‘interneurone C’ has been coined previously for an ascending sensory integrating interneurone of quite different function (Zucker et al. 1971), we term the neurone described below MoG-I1.

The anatomy of MoG-I1 is highly distinctive. When a neurone which was suspected to be MoG-I1 was injected with Lucifer Yellow, almost invariably two left–right homologous neurones were revealed (Figs 2, 3). These were so tightly dye-coupled that it was usually impossible to determine visually which neurone had actually been injected (although this could sometimes be determined by electrode position). This is the most complete dye-coupling that we have ever encountered in our studies on crayfish. In only one preparation out of more than 20 in which physiologically identified MoG-I1s have been encountered and stained has a neurone been found with an anatomy suggesting that it was a unilateral MoG-I1.

Each MoG-I1 has a cell body located ventrally just lateral to the midline. A neurite arises from the cell body, ascending towards the dorsal surface and moving slightly anteriorly, and then curves around to the contralateral side of the ganglion. Very extensive dendritic arborizations ramify within the central region of the ganglion, with one major branch extending anteriorly, while another extends posteriorly, giving the pair of coupled neurones an overall H shape. This shape is apparent in photographs taken in the plane of focus of the major dendritic branches (Fig. 2), but is less clear in camera lucida drawings, which superimpose all dendrites in the vertical axis (Fig. 3). The posterior branch gives rise to an axon that leaves the ganglion in the lateral margin of the posterior connective.

MoG-I2 (Fig. 4A) has a cell body located ventrally, just lateral to the midline, in approximately the anterior–posterior middle of the ganglion. A dorsal neurite ascends slightly anteriorly from the cell body and crosses the midline, where dendritic arborizations ramify within the central region of the ganglion, being restricted largely to the side contralateral to the cell body. Two axons arise from this dendritic region, one leaving the ganglion anteriorly, the other posteriorly, each in the approximate middle of the connective. MoG-I2 has been identified in G2 (two preparations) and G3 (four preparations).

MoG-I3 (Fig. 4B) has a ventral cell body located halfway between the midline and the lateral margin, towards the anterior end of the ganglion. A neurite ascends posteriorly and gives rise to a thick integrating region, which spans the midline aligned horizontally across the ganglion. An axon arises on the side ipsilateral to the cell body and leaves the ganglion posteriorly. MoG-I3 has been identified in two preparations (G3). In a third preparation (G3), a pair of bilaterally symmetrical homologous neurones was stained, the individual members of which had a structure and physiology similar to MoG-I3. Either this was another neurone entirely or MoG-I3 is sometimes dye-coupled to its contralateral homologue.

MoG-I4 (Fig. 4C) has a ventral cell body in approximately the middle of the hemiganglion. A dorsal neurite gives rise to dendritic branching, which is largely ipsilateral to the cell body but contains some contralateral elements. Two axons arise from these contralateral dendrites, one anterior and one posterior. MoG-I4 has been encountered in three preparation (G3).

MoG-I1 is easily and unambiguously identified because of its bilateral dye-coupling. Furthermore, it has totally different physiological characteristics from the other MoG-Is (see below). MoG-I2 has an anatomy somewhat similar to MoG-I4, since both have ascending and descending axons with cell bodies contralateral to the axons, but the two neurone types can be distinguished because the dendritic arborization of MoG-I2 is mainly ipsilateral to its axons and rather sparse, while the dendritic arborization of MoG-I4 is more extensive, and is largely contralateral to its axons. MoG-I3 has only a descending axon, and its cell body is ipsilatateral.

Each MoG-I has a posteriorly directed axon which passes at least to the next posterior ganglion, but we do not know its final termination point. As the axon passes the base of R3, fine branches extend from it and ramify across the surface of the ipsilateral MoG and, in most preparations, also across the surface of the contralateral MoG (Figs 3, 4). The exact morphology of these branches is highly variable, even within a particular class of MoG-I. In some preparations, branches also arise from the axon or dendritic branches at the posterior edge of the ganglion and propagate to the MoG quite separately from the main axon. Sometimes, fine branches arising at the base of R3 may course back into the ganglion and ramify within the ganglionic neuropile. We have not detected any consistent pattern to these detailed branch structures.

The MoG-I branches form a cobweb-like network across the MoG, with expanded bleb-like structures, which are presumably the actual sites of synaptic contact with the MoG (Fig. 5). These appear to be more or less uniformly distributed across the surface of the MoG, from the finger-like dendritic projections of the MoG, which form the electrical contact with the GFs, out into R3 and the axon proper of the MoG. In one preparation, approximately 120 blebs were counted on a small (0.1mm by 0.23 mm) area of the surface of the MoG. Since the MoG usually has a rather flattened surface after fixation, this suggests a (very approximate) synaptic density of 5000contactsmm⁻².

Input to MoG-I1. Extracellular stimulation of a GF in the connectives produces a large (35–50mV), rapidly rising potential in the MoG-I1, which characteristically has one or more ‘blips’ on the falling phase (Fig. 6A). The response is the same whether the lateral giant or the medial giant is stimulated, and so no distinction is made between these two classes of GF in this report. The GF stimulus also induces spikes in the ipsilateral and contralateral SGs. The SGs can be activated individually by antidromic stimulation of their axons in the first roots (R1) and, when either the left or the right SG is stimulated, MoG-I1 receives rapidly rising depolarising potentials, about 15–25mV in amplitude (Fig. 6B,C). When both SGs are simultaneously antidromically stimulated, the resulting potential in the MoG-I1 is similar in size to that resulting from GF stimulation (not shown). The connection between the SGs and MoG-I1 is via rectifying electrical synapses. Depolarising current injected into the SG spreads to MoG-I1 preferentially compared with hyperpolarizing current (Fig. 6D), while hyperpolarizing current injected into the MoG-I1 spreads preferentially to the SG compared with depolarizing current (Fig. 6E).

We attempted to determine whether there was any direct input from a GF to MoG-I1 by removing the intervening SGs through photoinactivation (see Fraser and Heitler, 1991). This confirmed the connection between the SGs and MoG-I1, since during photoinactivation the SG spike underwent massive broadening, leading to a great increase in the input to MoG-I1 (Fig. 7A,B). We were unable to kill the SGs completely (i.e. to abolish their membrane potential) in these experiments because dye-coupling to MoG-I1 invariably developed during the course of the experiment, leading to the start of inactivation of the latter. Nonetheless, we were able to photoinactivate both the SGs to the point where they failed to spike, without noticeably inactivating MoG-I1. When this was done, there still remained a significant input to MoG-I1 in response to GF activation (Fig. 7C). This strongly suggests that there is another input path to MoG-I1 besides that from the homoganglionic SGs. The latency of this input closely matches that of the discontinuity on the falling phase of the MoG-I1 spike in the intact preparation and is probably too long to be a direct monosynaptic input from the GF. In the MoG-I1 of G3, an EPSP was elicited by stimulating the ipsilateral R1 of G2. By analogy with the ‘C’ neurone of G6, this EPSP might have originated from the corollary discharge interneurone I2 (which would have been activated by the antidromic stimulation of the axon of the SG in G2), and thus the ‘blip’ on the MoG-I1 spike, and the residual input after SG inactivation, could be input from I2. However, we have not confirmed this directly.

MoG-I1 induces dIPSPs in the MoG. MoG-I1 can be induced to spike by injecting it with depolarizing current. Simultaneous recordings from the ipsilateral MoG reveal depolarising synaptic potentials phase-locked to the MoG-I1 spikes (Fig. 8A). Similar phase-locked potentials can also be recorded in the contralateral MoG (Fig. 8B), but we do not know whether this is because each MoG-I1 makes output to both MoGs, or whether the output is strictly ipsilateral, but strong electrical coupling between the bilateral MoG-I1s (as suggested by the dye-coupling) phase-locks their spikes. A spike in the MoG-I1 of G2 has not been observed to produce synaptic potentials in the MoG of G3 (despite the posteriorly directed axon of MoG-I1), although this has only been tested in two preparations. The depolarizing potentials induced in the MoG by MoG-I1 can be inverted by injecting depolarizing current into the MoG through a second microelectrode, showing that they are dIPSPs rather than EPSPs (Fig. 8C).

In a normal preparation, the dIPSPs produced by MoG-I1 are rather small compared with some dIPSPs produced by other MoG-Is (see below). We have never observed the dIPSPs induced in MoG by depolarizing current injected into MoG-I1 to be powerful enough to prevent the MoG from spiking in response to GF activation (e.g. Fig. 9A). However, during the process of photoinactivation, MoG-I1 undergoes massive spike broadening and increases its output considerably (Fraser and Heitler, 1991). Under these circumstances, the dIPSPs summated into a depolarized plateau that almost totally abolished any coincident MoG spikes (Fig. 9B).

The physiological characteristics of MoG-I2, -I3 and -I4 are very similar. The data presented (Fig. 10) are specifically for MoG-I2 but, unless stated to the contrary, apply equally to MoG-I3 and MoG-I4. In each case, spikes can be initiated in the MoG-I by injecting depolarising current and they cause large, sometimes very large (up to 25mV), dIPSPs in the MoG (Fig. 10A). The dIPSPs follow the spikes 1:1, with a fixed latency of approximately 2ms from the peak of the ganglionic spike in the MoG-I to the start of the dIPSP in the MoG at the base of R3 (Fig. 10B). Sometimes, small depolarising potentials can be detected in the LG synchronously with the spikes in MoG-I and the dIPSPs in the MoG (Fig. 10A). These may indicate direct inhibitory input to the LG from MoG-I, since the LG–MoG electrical synapse is strongly rectifying, and it is unlikely that depolarising potentials could spread antidromically back across the synapse. However, the functional significance of this has not been established. When MoG-I2, -I3 or -I4 is induced to spike, small extracellular potentials can usually be recorded in R3 synchronously with the start of the large dIPSPs in the MoG (Fig. 10B). These small extracellular potentials probably originate from spikes in the fine branches of the MoG-Is which propagate into R3 on the surface of the MoG.

The ability of the dIPSPs induced by each of MoG-I2, -I3 and -I4 to inhibit GF–MoG transmission was tested by bracketing extracellular stimulation of the GFs with intracellular stimulation of the MoG-I. The GF stimulation alone induced a spike in the MoG recorded intracellularly, and extracellularly in R3 (Fig. 10C). When the GF stimulation was coincident with dIPSPs caused by the MoG-I, the MoG spike potential recorded intracellularly was massively attenuated, and no MoG spike was recorded in R3 (Fig. 10D). In contrast, extracellular recordings from R1 show that the SG axon spike is maintained under conditions where the MoG axon spike is abolished, suggesting that MoG-Is do not make effective inhibitory input to the SG (Fig. 10D).

We have not identified specific paths activating any of MoG-I2, -I3 or -I4. Extracellular stimulation of either the anterior connective between the thorax and G1, or the posterior connective between G5 and G6, initiates EPSPs on each MoG-I, and these can cause spikes, but the spikes do not appear to be antidromic. Extracellular stimulation of R1 initiates smaller EPSPs. Extracellular stimulation of homoganglionic R3 usually initiates a large unitary depolarising potential in each of MoG-I2, -I3 and -I4, with the same stimulus activation threshold as a dIPSP in the MoG. This may indicate that the MoG-Is receive input from fast flexor motor neurones activated antidromically (Wine, 1977), but direct activation of individual FF motor neurones by injecting depolarising current (not shown) has never been seen to elicit EPSPs in any MoG-I. In our opinion, the depolarising potentials observed in the MoG-Is resulting from extracellular stimulation of R3 are more likely to result from direct stimulation of the terminals of the MoG-Is in R3 as they spread across the MoG surface. A spike in these terminals could propagate back to the ganglion as an EPSP-like depolarization, due to spike failure at the large discontinuity in diameter where the fine branches join the main axon at the base of R3.

Simultaneous intracellular recordings from the SG and MoG show that both neurones frequently receive a barrage of depolarising potentials, and that some are common to the two neurones (Fig. 11A). In seven preparations (five G2, two G3) we have encountered interneurones that activated phase-locked depolarizing synaptic potentials on both the homoganglionic MoG and SG when induced to spike by injecting depolarizing current (e.g. Fig. 11B). In three of these (G2), the interneurone was stained and revealed a through-conducting axon with restricted dendritic arborizations within the ganglion in the region of the SG and at the base of R3. The cell body was not located in the stained ganglion, and hence its position was not determined.

An important question is whether the depolarizing potentials are dIPSPs or EPSPs. In two preparations we were able to show that, by injecting sufficient depolarizing current into the interneurone to induce multiple spiking, the depolarizing synaptic potentials in the MoG could inhibit transmission at the GF–MoG synapse, indicating that they are indeed functional dIPSPs. In these preparations, transmission at the GF–SG synapse was not functionally inhibited at this dIPSP frequency, suggesting that the GF–SG electrical connection is more resistant to modulation than the GF–MoG connection. In a further two preparations, we were unable to demonstrate functional inhibition, but we were able to reverse the polarity of the depolarizing potentials by injecting subthreshold depolarizing current into the MoG, again indicating that the potentials were dIPSPs. In only one preparation were we able to demonstrate functional inhibition of both GF–MoG and GF–SG transmission (Fig. 11C–E). In this preparation, the MoG was not spiking, but its electrical EPSP was massively attenuated by the dIPSPs. The SG spike could also be abolished by superimposing GF stimulation onto dIPSPs induced by depolarising current injected into the MoG/SG-I, but more current had to be injected into the inhibitor to block the SG spike centrally than was required to attenuate the MoG EPSP. Interestingly, the axon spike of the SG recorded extracellularly in R1 was abolished at a lower intensity of inhibition than that required to abolish the central SG spike (Fig. 11D,E)

There is considerable evidence from earlier work that some through-conducting interneurones mediate multisegmental inhibition onto the MoGs (Wine, 1977). We have not concentrated upon the identification of these interneurones, but have encountered at least one interneurone, which was anatomically similar to one of the MoG–SG inhibitors described above, which induced dIPSPs on both the MoGs of G2 and G3 (Fig. 12).

In this paper we describe the final links in a complete circuit mediating feedforward inhibition from the GFs to the MoG (Fig. 13). The GFs have been shown previously to drive the SGs through rectifying electrical synapses (Kramer et al. 1981; Roberts et al. 1982; Heitler and Darrig, 1986); here we show that the SGs in turn drive the MoG-I1 through rectifying electrical synapses, and that MoG-I1 inhibits the MoG. This is almost certainly not the only pathway which mediates feedforward inhibition (since the dIPSPs that occur on the MoG following GF activation can sometimes be considerably larger than those mediated by MoG-I1), but it is probably one of the most reliable. The GF to SG synapse is extremely stable, and simultaneous spikes in the two bilateral SGs (as would normally occur in response to a GF spike) normally elicit spikes in the MoG-I1. Thus, the activation of the MoG-I1 via a disynaptic pathway of powerful electrical synapses ensures that the inhibition of the MoG mediated by the MoG-I1 will reliably arrive at a constant short latency after the GF-activated MoG spike. In our experiments, we found that the dIPSP mediated by MoG-I1 is not strong enough by itself to prevent the MoG from spiking, but the timing of its occurrence, which is tightly controlled by the electrical synapses, ensures that it will augment intrinsic inhibitory cellular events within the MoG (delayed rectification, sodium inactivation, reversed biasing of the rectifying electrical input synapse; Edwards, 1990). It is thus likely to make a significant contribution to the reduced excitability of the MoG following a GF-activated spike.

A major input to MoG-I1 is undoubtedly from the SG neurones. However, non-specific stimulation of the ganglionic roots and connectives shows that there are other unidentified sources of input. It is thus possible that MoG-I1 may be activated by pathways other than those from the SGs and in circumstances other than GF-mediated escape tail-flips.

We strongly suspect that the MoG-I1 that we describe in this paper is a pre-terminal homologue of the neurone named ‘C’, which occurs in the terminal abdominal ganglion (Kirk et al. 1986). The evidence for this is as follows. (1) The neurones have similar connectivity (input from the SGs, inhibitory output to the MoG). (2) The neurones both occur as a tightly dye-coupled bilateral pair. (3) The cell bodies of the neurones occur in the same relative location. (4) The neurones have dendritic arborizations of broadly similar shape (within the constraints imposed by the terminal location of the C neurone).

The dIPSPs mediated by the identified interneurones MoG-I2, -I3 and -I4 can be extremely large (up to 25mV) and are highly effective in preventing GF activation of the MoG. This effectiveness is consistent with the synaptic morphology revealed by staining the MoG-Is. It is extremely unusual to be able to identify sites of synaptic contact at the light microscope level, but the rather simple anatomy of the MoG, which lacks any extensive dendritic arborizations, combined with the location of the MoG-I to MoG synapses at the base of R3, away from the major dendritic neuropile of the ganglion, enables us to visualise the sites of contact in this case (Fig. 5). Dual staining of the MoG-I2 and the MoG shows that the surface of the MoG is covered with a cobweb-like network of synaptic blebs. These extend from the central hemiconnective region, where the GF makes electrical synaptic contact with the MoG, right out along the axon of the MoG into R3. In some cases branches of the inhibitor have been traced as far as the second branch point of R3 as it courses towards the muscle. This extensive area of high-density synaptic contact is the anatomical corollary of the large and effective dIPSPs observed in the MoG in response to activation of the presynaptic inhibitory neurone.

It has been suggested previously (Wine, 1977) that one pathway mediating feedforward excitation of the MoG-Is from the GFs is via the fast flexor (FF) motor neurones. The evidence for this was that stimulating R3, which contains the axons of the FFs, induced dIPSPs in the MoG. Consistent with this, we have found that stimulating R3 extracellularly can induce depolarising potentials in MoG-Is. However, we have never observed dIPSPs arising in the MoG as a result of specific intracellular stimulation of any FF motor neurone, nor have we observed EPSPs in any MoG-I arising from this stimulation (K. Fraser and W. J. Heitler, unpublished data). Furthermore, in this paper we show that small potentials can be recorded extracellularly from R3; these are phase-locked to spikes in the MoG-Is when the latter are induced by intracellular current injection. It is perhaps possible that these might be field potentials arising from the large dIPSPs occurring on the MoG, and propagating by cable conduction some distance along R3, but a more likely explanation is that they are spike potentials arising from branches of the MoG-Is which themselves extend out into R3. The reason we favour this explanation is that Lucifer Yellow stainings of MoG-Is show that such branches do indeed frequently occur. When R3 is stimulated extracellularly with pulses of increasing amplitude, the initial response in a MoG-I is usually a depolarising potential, which then shows stepped increments in amplitude as the stimulus is increased, eventually giving rise to a spike. We interpret this as being due to progressive recruitment of spikes in the fine branches of the MoG-I, until the summed branch-spike current is sufficient to propagate a spike antidromically past the expansion in axon diameter that occurs at the base of R3.

This work was supported by a grant to W.J.H. from the Science and Engineering Research Council of the UK, and from the Hasselblad Foundation.