Recurrent Inhibition in the Giant-Fibre System of the Crayfish and its Effect on the Excitability of the Escape Response

Twenty years ago Wiersma (1947) presented evidence to show that a single impulse in any one of the four central giant axons of the crayfish was sufficient to evoke a complete, complex escape movement involving enormous numbers of muscles and joints. The only other responses in which we know that a single impulse is sufficient to trigger a complete cycle of response are the escape reactions of fishes and cephalopods (Wilson, 1959 ; Young, 1938). In all three cases the escape responses are mediated by systems of giant neurones. The latter have attracted considerable attention from electrophysiologists interested in the types of synaptic contacts made by nerve cells (for example, Furshpan & Potter, 1959 a, b;Watanabe & Grundfest, 1961; Furukawa & Furshpan, 1963; Hagiwara & Tasaki, 1958). This biophysical work has laid the foundations for studies aimed at understanding central nervous mechanisms for the control of locomotion and behaviour.

The escape movements in fish, crayfish and squids are modified forms of more sustained swimming movements. For all these cyclic responses one would like to know what stimuli are effective in exciting the giant central command interneurones.. Work in this direction is in progress on fish and crayfish (Furukawa, 1966; Krasne, 1966). At the other, motor end of these complex reflex arcs one wants to know how single impulses in the command interneurones. generate the response (Young, 1938). Finally, one wants to know how the response is coordinated with the behaviour of the animal. In short, how can the animal, be it a fish squid, or lobster, ensure that it only does one thing at a time?

It is this last question which is approached in the present work. An animal just about to make an escape response is often subjected to a barrage of stimuli. It may be moving away or adopting some defence posture. However, at some point, if excitation reaches a critical level, a giant command interneurone will fire an impulse. This must evoke an escape response and also terminate all other actions or postures and ensure that no further, competing reactions occur during the escape response that would in any way interfere with it. The animal must wait for completion of its escape response before it attempts to escape again or to react with other responses using the same muscles.

The essential requirement is that no competing muscular activity must occur during the escape response. This may be achieved in fish by a period of inhibition which has been described in Mauthner cells by Furukawa & Furshpan (1963) and Furukawa (1966). When one of the two Mauthner cells fires an impulse it inhibits the other for a period of about 15 msec. At the same time the cell which fires also inhibits itself for a similar period. This recurrent inhibition must surely be involved in the suppression of too early a second response and consequently in the spacing of successive impulses and tail flips. The intemeuronal pathways responsible for this inhibition might also be involved in the inhibition of other competing responses for the same period of time. There are, in principle, other ways in which the required inhibition could be brought about. It could act anywhere in the reflex pathways mediating these escape responses. Its origin could be central or peripheral.

The crayfish giant-fibre system, which is analogous, allows one to test the generality of the role of recurrent inhibition suggested for the fish Mauthner cells.

The giant-fibre system of the crayfish consists of two pairs of central, longitudinal giant axons, the medial giants (MGs) and the lateral giants (LGs) and many pairs of peripherally directed motor giants (Motor Gs). The MGs originate in cell bodies in the brain where dendrites are probably excited by a variety of visual and other stimuli (Wiersma, 1947; Kao, 1960; Horiuch, Hayashi & Takahashi, 1966). The MG axons run caudally to the last abdominal ganglion. Like the MGs the LGs run the whole length of the animal. However, as their alternative name ‘segmental giants ‘suggests, the ‘fibres’ are composed of segmental units which join end to end with electrical junctions. These junctions allow the series of units to behave much like a single axon (Watanabe & Grundfest, 1961). The LGs send processes into the neuropile of all thoracic and abdominal ganglia of the ventral nerve cord and can be excited by stimulation to any segment posterior to the fourth thoracic, (Johnson, 1924; Kao, 1961). There are therefore many places in the giant-fibre system where afferent fibres can excite the central giant cells and evoke an impulse which will then generate an escape response. At least two of these sites for impulse initiation are rostrad in the MGs but most, perhaps twenty-two, are distributed along the LGs.

The presence of these numerous, separate command locations emphasizes the need for some coordinating mechanism to avoid confusions at the muscular level. This paper describes the search for a period of inhibition that might play a part in such coordination.

Male and female Procambarus clarkii (Girard) from California were held in tanks of fresh water at from 14 to 18°C. Only animals about 8 cm. from rostrum to telson tip were used, and before any operation they were cooled until the water around them began to freeze and they became much less reactive.

Two types of preparation were used. In the first the animal was intact and free to move in a small (bottom 12 by 23 cm.) water-filled aquarium. Recording and stimulation of these animals was via fine, pin electrodes. These were made from anodized insect pins, cut to various lengths and soldered to 40-gauge copper wire. The solder j unction was insulated with varnish and the electrodes were then sharpened at the tip, stuck through the exoskeleton of the crayfish and waxed in place. The copper wires were anchored mechanically to a loop of silver wire attached to the carapace at the top of the gill chamber and were than led from the top of the aquarium. Mounted in this way, electrodes often stayed in place for weeks and could be used for stimulation or recording. Ciné films of free-moving animals stimulated with these implanted electrodes were taken at 300 frames per second (frames/sec.) with an electrically operated Wollensak high-speed camera.

The second type of preparation was the isolated nerve cord which was used for intracellular recording (cf. Furshpan & Potter, 1959a, b; Takeda & Kennedy, 1964). To dissect out the nerve cord, the abdomen was cut from the animal and pinned down in a dish of Ringer cooled to about 7° C. with an ice jacket. The six ganglia of the abdominal nerve cord were exposed by removing a strip of ventral exoskeleton along the midline. Once this was done all the nerve roots were cut, the nerve cord was lifted out in a spoonful of Ringer and placed in a Perspex chamber. Here it was fixed down, stretched fairly taut, by jamming roots or connectives into grooves on the bottom of the chamber with short lengths of silver wire. With oblique illumination from below the giant fibres could generally be seen very clearly. For intracellular recording, fine forceps were used to tear a small hole in the connective tissue sheath of the nerve cord about the same diameter as, and just above, the axon to be penetrated. While in the chamber, the cord was continuously perfused with Ringer cooled to 15±1° C. by passage over a Peltier-effect cooler before entering the chamber. In earlier experiments van Harreveld’s crustacean saline was used (van Harreveld, 1936), but later Dudel’s Trizma maleate modification was adopted as its pH is more stable (Dudel, 1961). No differences were noticed as a consequence of the change.

Gross-stimulation and recording were with tapered platinum wires, which were insulated to their tips with varnish and pressed against root or cord under the saline. Intracellular recording was with glass capillary micro-electrodes (5−40 MΩ) filled with 2M potassium citrate and a Medistor negative capacity electrometer amplifier. Extracellular signal amplification and stimulation used conventional equipment. For constant-current stimulation a 500 MΩ resistance was placed in series with the current electrode. All intracellular records were photographed with a Polaroid camera from images stored on a Tektronix 564 storage oscilloscope.

To relate intracellular recordings from the isolated nerve cord to events during the escape response of the intact crayfish it was necessary to establish the timing of tail flips and to confirm that a single impulse in any of the central giant fibres is sufficient to evoke a normal-looking tail flip as Wiersma (1947) had reported. Since the tail flips of animals held in any kind of clamp look very abnormal, the timing of flips was studied by taking high-speed motion pictures at 300 frames/sec. of animals in a small aquarium. These animals were stimulated in two ways : (a) natural stimulation such as prods to the rostrum or uropods, (b) electrical stimulation with brief (0·1 msec.) current pulses delivered via fine pin electrodes (see Methods). These were inserted through the exoskeleton so that their uninsulated tips lay near the circumoesophageal connectives. In this location they were well away from any muscles involved in the tail flips (Schmidt, 1915). At threshold, a single central giant axon was stimulated, as could be shown by recording from the nerve cord of the same animal when filming was completed.

Both types of stimulation evoked tail flips. There were differences in the responses to different forms of stimulation (cf. Bullock & Horridge, 1965), but it was clear that electrical stimulation, evoking a single impulse in a single giant axon, produced complete and normal escape responses. Such a response is shown in Fig. 1 where the full cycle of tail flexion and re-extension can be seen.

In an 8 cm. crayfish at about 16° C., responses to single electrical stimuli were measured from the film. These responses to electrical stimuli are definitely evoked by single central giant-fibre impulses. This may not be true for more sustained swimming movements. Flexion was complete in 28 msec, (range 24−30), extension took a further 62 msec, (range 51−66), so the whole movement was complete in about 90 msec. These times were relatively constant in animals of similar size and are compatible with those given by Eckert (1961).

In the preceding section it was established that the escape response is evoked by single giant-axon impulses and takes about 90 msec. Therefore, a search was started for inhibitory effects of similar duration following giant-axon impulses. The numerous sites where inhibition of the reflex arc could act were divided into two categories, on the afferent and efferent sides of the giant fibres. Correspondingly, two questions were posed:

• Is there inhibition of the effects of a giant-axon impulse? (In this case the inhibition would act on synaptic pathways between the central giants and the motoneurones or at the neuromuscular junctions.)

• Is there inhibition of the generation of giant-axon impulses? (In this case it would act at the receptors or on the synaptic pathways between afferent fibres and the central giants.)

These two questions will be considered in turn.

The first experiments were in whole animals stimulated via implanted electrodes as in the preceding section. This time, however, the crayfish were held rigidly at the carapace and movements of the first few abdominal segments were monitored with a photocell. Flexion increased the fight falling on the cell and extension decreased it (see top of Fig. 2A).

The aim was to see if the abdomen could respond to closely spaced impulses in a central giant axon. The responses recorded under these circumstances were not normal but they sufficed to show that the musculature is responsive to giant-axon impulses at all phases of the tail flip. (See Fig. 2A.) This confirms Wiersma’s (1947) finding. Further confirmation came from recordings of abdominal flexor muscle activity taken with implanted electrodes from free-moving animals. Under these conditions peripheral feedback during an escape response must be very nearly normal yet the muscles still responded to very closely spaced (5 msec.) giant-axon impulses. There is therefore no indication of a peripheral or central mechanism to block response of motoneurones or muscles and thus prevent confusion during the tail flip.

Though a period of complete motor inhibition was excluded, there still remained the possibility that some of the motoneurones might be inhibited, perhaps the motor giant fibres as the work of Hagiwara (1958) and Furshpan & Potter (19596) might suggest. Intracellular recording from the motor giant axon near its junction with the central giants excluded this last possibility. Experiments by Takeda & Kennedy (1964) indicate that the remaining non-giant motoneurones can also follow closely spaced central giant impulses and this I have confirmed by intracellular recording from the axons and cell somas of these motoneurones.

The preceding experiments have shown that, following a giant-axon impulse, there is no long-lasting inhibition of the muscles or motoneurones that might prevent a second response. Therefore, if such inhibition exists, it must act at an earlier stage in the reflex arc for the escape response. A preliminary experiment was therefore performed (in collaboration with Dr Frank Krasne) to see if excitation to one of the giant fibres, the lateral giant (LG), was depressed for some period following an impulse in the same LG axon.

In the isolated abdomen the nerve cord was exposed dorsally. The lateral giant (LG) was penetrated in the third abdominal ganglion just anterior to the septal junction at the point where a large process descends into the neuropile (Johnson, 1924). At this recording site compound excitatory potentials can be seen when the ipsilateral second root of the same ganglion is stimulated with a brief (0·01 msec.) current pulse. When the voltage of root stimulation is large enough, this excitatory potential reaches threshold for impulse initiation in the LG axon. The response, even at higher stimulus strengths, is only a single impulse (see Fig. 3 j and Krasne, 1966). The LG could also be stimulated to fire an impulse directly by electrodes placed on the cord a few segments away. Fig. 3 shows the effect of such directly evoked impulses on later excitatory potentials resulting from root stimulation. If the directly evoked impulse preceded the excitatory potential (EPSP) by less than about 80 msec, there is inhibition and the EPSP no longer evokes a second impulse (see Fig. 3 i, j).

The duration of this inhibition coincides with the time a crayfish of this size needs to complete a tail-flip movement, although this may have been due to the particular stimulus strength used ; it also seems to coincide with the presence of a depolarizing potential following the LG impulse (e.g. see Fig. 3 i). These depolarizing potentials have been seen previously in the LGs by Kao (1960) and by Watanabe & Grundfest (1961). These workers regard them as at least partly postsynaptic potentials because under certain circumstances they can be obtained alone, without any LG impulse. Their origin was unclear. However, similar potentials, seen in the motor giant fibres, have been shown to be initiated by impulses in the central giant axons by Hagiwara (1958) and to be inhibitory potentials by Furshpan & Potter (19596). On the basis of this similarity and the experiment just described, I thought that the slow potentials seen in the LG axons might be IPSPs (inhibitory postsynaptic potentials) which could be serving the function suggested in the introduction. If this is the case they should occur in the LGs after an impulse in any of the central giant axons (since each of them can initiate a flip). Also they should not be present in the MGs in the abdomen where these fibres have no excitatory input (Kao, 1960).

These hypotheses were tested in the isolated nerve-cord preparation described in the methods section. Intracellular recordings were taken from the medial and lateral giant axons in the third ganglion while the same two axons were stimulated at a more caudal site with intracellular electrodes. (By this method stimulation of other axons was avoided, cf. Furshpan & Potter, 1959b.) The results are shown in Fig. 3 k−n. As predicted, slow potentials occurred in the LGs when an impulse occurred in any giant, but no comparable slow potentials were present in the MGs.

The slow potentials in the LGs seem to behave according to expectations derived from the inhibitory period hypothesis; therefore, experiments were carried out to decide whether they are IPSPs (inhibitory postsynaptic potentials).

In the next few sections MG-evoked LG slow potentials will be emphasized since they are not complicated by post-spike phenomena. The criteria used to determine whether or not the slow potentials are inhibitory postsynaptic potentials are presented below.

In trying to reveal different classes of PSPs one of the most useful techniques has been that of artificially polarizing the membrane of the postsynaptic cell to determine the PSP reversal potential (Tauc, 1957; del Castillo & Katz, 1954). For an inhibitory PSP generated at a presumed chemical synapse this reversal potential should be more negative than the spike threshold.

The LG slow potentials were observed at different membrane potentials by inserting two micro-electrodes at the usual site just anterior to the septum in the third abdominal ganglion. One was used to record potentials while polarizing current was injected through the other.

The results of a typical polarization experiment are shown in Fig. 4. Considering the responses to MG impulses ((A) open circles and (B), in Fig. 4) it is clear that the major part of the slow potential behaves like one resulting from a specific postsynaptic permeability increase and having an equilibrium potential about 7 mV. depolarized from the resting potential of 77 mV. However, in addition there is, in this preparation, a small, early component which behaves differently (Fig. 4B). This component is rather a mystery but will be considered again in the discussion. The slow potentials following LG impulses (Fig. 4C) have an equilibrium potential close to that of the MG-evoked slow potentials.

To have an inhibitory effect a process producing depolarizing potentials must decrease postsynaptic membrane resistance. This decrease would tend to attenuate depolarizations that might initiate impulses. Therefore the LG resistance during the slow potentials was measured to see if there were changes that could account for the observed inhibition. The electrode arrangement was the same as in the previous polarization experiment except for the addition of a pair of extracellular stimulating electrodes on the ipsilateral second root of the third ganglion. The intracellular current electrode was used to pass constant current, polarizing pulses. The potential change produced by these pulses was measured at steady state and used as a relative measure of membrane resistance (Dudel & Kuffler, 1960). Such measurements are justified for the peak and falling phases of the slow potentials since, for this period, the time constant of the resting membrane, about 1 msec., is much shorter than the time course of the resistance or potential changes. The aim was to compare the effects of the MG-evoked and LG-evoked slow potentials on the current pulses and second root EPSPs.

The results and potentials measured are shown in Fig. 5 where the graphs show a similarity in the time course of the resistance decrease and the inhibition of the test EPSPs. The peak reduction in membrane resistance during MG-evoked slow potentials was about 30−35 %. (This reduction can also be determined from the slope of the curve for MG-evoked slow potentials in the polarization experiments of Fig. 4, which was from the same preparation. A change of 10 mV. in base potential produced a 3·5 mV. change in slow potential amplitude.) In other preparations the range was from 25 to 40%. The peak reduction in the first component of the second root EPSP was in this case about 50−60% but ranged up to as much as 90% being different in some cases for different components of the compound EPSP.

At all but electrical junctions there is a definite delay of 0·5 msec, or more between the start of depolarizing potentials in the presynaptic terminals and the consequent PSPs in the postsynaptic cell (Bullock & Hagiwara, 1957). The latency of the MG-evoked LG slow potentials is 1·5−5 msec, after the rising limb of the MG impulse. This measure is not very precise as in some cases the potentials develop rather gradually. The delay is long enough to exclude any direct electrical coupling close to the recording site. These results will be useful in later discussion of the synaptic pathways for inhibition.

Since the resistance change during the slow potentials relates so closely to the inhibition of the excitatory potentials it is concluded that these slow potentials are inhibitory. They are IPSPs. Is the position of the equilibrium potential for the slow potentials consistent with this interpretation?

The equilibrium potential for the slow potentials was commonly near 7 mV. (range 6-9 mV.). (This measurement, like all others in this section, is in mV. of depolarization from the resting potential.) If the inhibitory synapses are close to the recording electrode then this value will be accurate, but for more distant synapses the equilibrium potential will be overestimated to the extent of the potential drop between the recording electrode and the synapses. Therefore, the true equilibrium potential is probably equal to, or less than, 7 mV. This is well below the threshold of 20−30 mV. for spike initiation in response to direct stimulation (personal observation ; Watanabe & Grundfest, 1961). However, it is rather close to the thresholds of about 8 mV-(maximum 14 mV.) for spike generation by second-root EPSPs (Krasne, 1966; and see Fig. 3). These EPSPs are probably generated at a distance from the recording site, in the LG dendrites (for anatomy see Kendig, 1967). Were this so, an EPSP reaching threshold could appear much smaller in the main axon than it is nearer its origin or at the point where it triggers an impulse.This would lead to an underestimate of spike threshold which is probably therefore more than 8 mV. depolarized from the resting potential.

It is therefore likely that the equilibrium level of the slow potential is more hyperpolarized than the spike threshold. The process producing the slow potentials will consequently tend to clamp the membrane potential below threshold and therefore have an inhibitory effect.

It is interesting to note one of the paradoxical features of the depolarizing IPSP. This is that small EPSPs summate with the IPSP so that the total depolarization is greater than the EPSP alone. On the other hand, large EPSPs summate to produce less depolarization than they would alone (compare EPSP components in Fig. 5d−f). The critical feature is that those EPSPs which are approaching threshold and might therefore evoke an impulse are decreased and prevented from doing this by the IPSPs.

Picrotoxin seems to block all forms of neural inhibition investigated in crustaceans while having no effect on excitation in the absence of other drugs (Grundfest, Reuben & Rickies, 1959). In crayfish its injection causes convulsions, or, in smaller amounts, sensitization to stimulation (Florey, 1951). Further it has been shown to block inhibition at the neuromuscular junction (Robbins & van der Kloot, 1958), cardiac ganglion (Florey, 1957), and stretch receptors (Edwards, 1960). These are all peripheral synapses, but it might have a similar action at central inhibitory synapses. The effect of picrotoxin on the LG PSPs, if present, should therefore provide an independent check on the conclusion, reached in the previous section, that these PSPs are IPSPs.

Recordings were made from the LGs in the usual position. External electrodes were arranged so that the LGs, an MG and ipsilateral second root of the third ganglion could be stimulated. The second-root EPSP was used as a control to compare with the presumed IPSP evoked by MG stimulation. When all the electrodes were in place the preparation was perfused as follows : (a ) normal Ringer to establish base levels for the various potentials; (b) 10⁻⁴M picrotoxin in Ringer for about an hour; (c) normal Ringer to check that changes were reversible.

Picrotoxin had rather dramatic effects (see Fig. 6). As the MG-evoked IPSP shrank away, the EPSP resulting from second-root stimulation increased. Within 5 min. the EPSP was generally increased enough to evoke impulses and at full effect (about 15 min. after the change to picrotoxin Ringer) the single root stimulus triggered a long (40 msec.), high-frequency (250 per sec.) burst of LG impulses (Fig. 6, B2 and B3). The MG-evoked IPSP was very much reduced in amplitude after about 30 min. in the picrotoxin Ringer. Reduction of the potentials following LG impulses was clear but less marked than the effect on the MG-evoked IPSPs.

The reduction in the LG PSPs when treated with picrotoxin further supports their identification as IPSPs. The repeated discharge of the LGs in picrotoxin is also in marked contrast to the normal response to second-root stimulation, a single impulse. This suggests another possible function for inhibition, namely the termination of the LG discharge after a single impulse. The repeated discharge also shows that there is no special property of the LG axon to prevent repeated firing.

A crayfish about 8 cm. long swam at an average of eight tail flips per second in a number of separate swimming sequences (range 5·5−23 per sec.). It was therefore interesting to see what happened to the inhibition at similar frequencies. Only the IPSPs evoked by MG impulses are considered. Responses are from the whole inhibi-tory pathway so limitations may be imposed by any excitatory or inhibitory synapses in this pathway.

When evoked at 8 per sec. the LG IPSPs are about half their full amplitude but the amount of inhibition at this frequency per unit time is still not at its peak plateau value (see Fig. 7B, C). (At frequencies higher than about 15 per sec., the IPSPs begin to overlap so that potential is no longer useful as a direct measure of inhibition.) Even at lower frequencies (1/sec) there is a considerable reduction in the amplitude of IPSPs after only a few responses (see Fig. 7 A). Nearly full response is only maintained at very low frequencies (0·1−0·5/sec.). Pairs of stimuli to an MG show that the MG-evoked IPSPs are smaller if they follow a similar IPSP. This effect persists for up to 15 sec. and the reduction can be as much as 30% even when the IPSPs are separated by more than their own duration. At shorter intervals, when the IPSPs are summing, there is no sign of any facilitation of the second response.

After bursts of PSPs at a variety of synapses there are often ‘after effects ‘such as post-inhibitory rebound and post-tetanic potentiation (Chalazonitis & Arvanitaki, 1961 ; Bullock & Horridge, 1965). No such effects have been observed for the LG IPSPs though a thorough search has been made. The response decrement described above is the only result of repeated activity that was noticed. Compared with nearly all vertebrate excitatory synapses the repeated response capability of this inhibitory pathway is very weak (Eccles, 1964).

Somewhere in the abdominal nerve cord of the crayfish there must be some interneurone or set of interneurones. which are responsible for the generation of the LG IPSPs. It is the purpose of this section to inquire about these pathways with certain questions in mind. Are the IPSPs unitary PSPs? What is the relation between the similar IPSPs evoked by impulses in the different central giant axons? Where in the LGs do the IPSPs originate? Are the interneurones generating the IPSPs confined to single segments or do they extend between segments?

There are at least three independent pathways or presynaptic fibres by which inhibition is evoked in the abdominal LGs. These are the LGs themselves, which always fire together (Watanabe & Grundfest, 1961), the ipsilateral MG and the contralateral MG. The latter pair often produce indistinguishable IPSPs in the LGs. The LG-evoked slow potentials have some features in common with these MG-evoked IPSPs: (1) The duration and shape of the two types of potential are broadly similar especially in more rostral ganglia (see Fig. 10 A2). Differences between animals tend to be similar for the two types also. For example, when the MG-evoked IPSP is flat-topped the LG slow potential is also (see Fig. 3,k, I), (ii) The inhibition of EPSPs during the two types of potential is similar in amount (Fig. 5). (iii) The reversal potentials of the two are close, near 7 mV. depolarized from the resting potential (Fig. 4). However, against this background of similarities there are clear differences such as the invariably larger amplitude of the LG-evoked slow potentials. This may be due to the IPSP being superimposed on a true LG spike afterpotential.

In view of the similarities between the effects produced by the three presynaptic pathways, experiments were performed to see if evidence could be obtained about the existence of some final common pathway involved in the production of the LG IPSPs. One of the classical methods for the demonstration of shared pathways is to study summation and look for occlusion between the effects evoked by different presynaptic pathways (Creed et al. 1932). Occlusion can be shown between all three pathways evoking the LG IPSPs. Figure 8 B shows responses of the left LG in the third ganglion where the IPSP evoked by synchronously arriving impulses in both MGs is no larger than that evoked by an impulse in the right MG alone. Similarly, in (A) of the same figure, the slow potentials after LG impulses are no larger when one, or both MGs fire impulses at the same time as the LGs.

Responses to paired stimuli are also pertinent to the question of shared pathways. When a pair of stimuli, 500 msec, apart, are given to the right MG the IPSP evoked by the second impulse is smaller than that evoked by the first (see previous section). This response decrement could be due to a variety of processes in the pathway be-tween the MG and the LG. If the other giants excited the same pathway then, for example, stimulation of the left MG or the LGs should cause a similar response decrement in a succeeding IPSP evoked by the right MG. Such crossed response decrement was found between all three pathways evoking the LG IPSPs. It could be induced in any ‘direction’ as is illustrated in Fig. 9.

Each picture shows two responses ; the larger is a control response after a 30 sec. rest, the other is a test response evoked 500 msec, after a conditioning stimulus given to the same or a different giant axon. In all cases there is a similar decrement in the test IPSP or slow potential. For example, the reduction in amplitude of a right MG-evoked IPSP in the left LG is the same if, 500 msec, previously, there is an impulse in the LGs, the left MG or the right MG itself.

To this point all recordings have been from one place in the lateral giants. To complete the picture and to try to find where the LG IPSPs originate, recordings were made from a variety of positions along the LGs from the second to the fifth ganglion of the abdominal nerve cord.

Figure 10 shows records from two such preparations in which the LG of one side was penetrated at a number of different locations. In general these observations show that very similar potentials are present throughout the abdominal nerve cord indicating that a single, localized source for the IPSPs is out of the question. Further, the lack of appreciable amplitude differences in potentials recorded on either side of the LG septa (even in isolated ganglia), suggests that the IPSPs may not be generated exclusively on one or the other side of these septa (Fig. 10A). Efforts at precise localization are frustrated by the large space constant of the LGs (Watanabe & Grundfest, 1961), and by the fact that the IPSPs most probably arise at many sites along the LGs. Were these sites confined to the ganglia of the nerve cord this would explain the consistently smaller amplitude (about 70% of ganglionic PSPs) of the IPSPs recorded between ganglia (see, for example, Fig. 10 B 2). The slightly smaller amplitude of IPSPs recorded on the rostral sides of the septum could be due to the shunting effect of the large LG branch dipping into the neuropile at this point (cf. Watanabe & Grundfest, 1961).

The last topic to be considered is the extent and organization of the interneurones. in the pathways excited by the central giants and generating the LG IPSPs. By cutting through the connectives on either side of single abdominal ganglia it was established that very normal slow potentials could occur in an isolated ganglion. This shows that there is a sufficient mechanism in one ganglion to respond to giant-fibre impulses and produce inhibition of the LGs. The inhibition is thus not dependent on intersegmental spread or conduction.

However, the experiment of Fig. 10 C shows that such intersegmental connexions do exist. The right MG between the third and fourth ganglia was very carefully desheathed as if for penetration (see Methods). The dorsal wall of the axon was then grasped with forceps in two places and the whole right MG torn in two. This produced a conduction block in the MG axon. However, right MG-evoked IPSPs were not completely abolished caudal to the cut. Despite the absence of a right MG impulse in ganglia four and five, there were still slow potentials (see Fig. 10C). However, in these ganglia, the responses to the right MG impulse were smaller than in more rostral ganglia. Further, rostral to the cut, the responses to the right MG were 92% of those to the left MG, whereas caudal to the cut this percentage was 58%. What are these reduced slow potentials? The changes in amplitude distal to the cut were not due to local damage since full IPSPs resulted when the stimulating electrode was moved to the proximal side of the cut. If the IPSPs originated in the third ganglion and were seen electrotonically at the fourth, then they should be twice as large on the rostral as on the caudal side of the LG septum in the fourth since Watanabe & Grundfest (1961) have shown that the septum attenuates to about one half. The IPSPs are nearly identical at these two sites so the possibility of passive spread is eliminated.

I therefore conclude that a right MG impulse in the third ganglion can evoke reduced slow potentials, IPSPs, in the fourth and fifth ganglion even though the MG impulse does not reach these ganglia. This conclusion, and the increased latency of the right MG-evoked IPSP in the fifth ganglion (Fig. 10C3), suggest that there must be some inter-ganglionic pathway responding to MG impulses in one segment and causing inhibition in others. The fact that IPSPs evoked by this pathway are smaller than normal implies that there must be at least two sites for impulse initiation arranged in parallel in the normal pathway for IPSP generation.

In the introduction the possibility of a period of inhibition following an impulse in any central giant axon was suggested. Such a period was found. After an LG impulse there is a long depolarization and, during this slow potential, LG excitability is depressed. It was argued that if these slow potentials are the result of an inhibitory process, such as the one proposed, they should also occur after impulses in the MGs. They do, and this confirmed prediction is a strong argument in favour of the view that the two classes of LG slow potential, evoked by LG and MG impulses, are similar in function and are inhibitory. Subsequent experiments on the slow potential properties showed that : (i) the equilibrium potential is close to the normal resting potential, (ii) coinciding with the potentials there is a considerable resistance decrease and finally, (iii) the potentials virtually disappear when picrotoxin is present. There is thus good evidence for the presence of a postsynaptic inhibitory process and the slow potentials should be interpreted as IPSPs.

Even in the absence of any other inhibitory process, this postsynaptic inhibition can, in some cases, account for 70% of the observed reduction in EPSPs. However, further postsynaptic inhibition may occur far from the recording electrodes in branches of the dendrites of the LGs where EPSPs may originate (see Kendig (1966) for anatomy). Such inhibition could have only an insignificant effect on the resistance as measured in the main axon where the electrodes are inserted. Similar ‘remote dendritic inhibition’ was first proposed by Frank (1959) and has been invoked more recently by Granit, Kellerth & Williams (1964) and Green & Kellerth (1966) for motoneurones in the cat and by Furukawa (1966) and Diamond (1968) for goldfish Mauthner cells. In these two cases, as in the present one, presynaptic inhibition of terminals of the excitatory inputs could also play some role and is very hard to exclude (Eccles, 1964).

Does the depolarizing polarity of the LG IPSPs have any special significance? In the crayfish LGs and probably also in the Mauthner cells of fish the equilibrium potential for recurrent inhibition is close to the normal resting potential. As a consequence there is little current flow or change in potential during the IPSPs. This may be important in preventing complex after effects of the IPSPs which might be enhanced by strongly hyperpolarizing IPSPs (e.g. Chalazonitis & Arvanitaki, 1961 ; see also Florey, 1951).

The depolarizing polarity probably results from a relatively high internal chloride concentration. Wallin (1966) has measured the chloride concentration in the MG axon of Procambarus(crayfish) and found that the chloride equilibrium potential, like that of the IPSPs in the present work, is on the depolarized side of the resting potential. This coincidence supports Watanabe & Grundfest’s (1961) suggestion that these potentials in the LGs are the result of an increase in permeability to chloride.

Higher internal chloride concentration, which would lead to depolarizing IPSPs, may be of significance because it reduces the internal axon resistance and therefore increases the speed of impulse propagation (Keynes, 1963). Depolarizing IPSPs may be the secondary result of such an adaptation for increased velocity.

It is certain from the stimulation and filming experiments (Fig. 1) that a single impulse in any one of the central giant fibres is sufficient to evoke a complete escape response (cf. Wiersma, 1947). Recordings from intact animals have so far provided no evidence for repeated bursts of impulses in the crayfish giant axons like those in the lateral giants of earthworms (Bullock, 1945). Though it is likely that a single impulse is the normal command to escape, bursts could conceivably occur under some circumstances.

The possibility of some excitatory coupling between the MGs and LGs in the abdomen is of interest when considering the normal firing patterns of the giant axons. In some preparations an MG impulse evoked an early, depolarizing potential in the LG (see Figs. 4B and 10A 1, 2). These potentials ranged from 0·5 to 3 mV. In preparations where the early component was large it could trigger an LG impulse if the LG was depolarized by applied current. This may allow some sort of synchronization or summation so that the LGs can be excited to fire by the MGs if the former are already partially excited (depolarized).

The enormous potentiation of the LG response to second-root stimulation by picrotoxin seems best explained in terms of decreased inhibition. This suggests that, normally, pathways and synapses exciting the lateral giants are inhibited to a considerable extent and that there is a large reserve of excitability which can be released by a reduction in inhibition. How this excitability is normally controlled is an extremely interesting question.

The fact that picrotoxin blocks the MG-evoked IPSP is useful confirmation of the other criteria used to identify these potentials. It is also interesting that picrotoxin has similar effects at central and peripheral inhibitory synapses. It should therefore be a useful experimental tool for the study of crustacean central nervous processes.

Features of the synaptic pathways for LG inhibition must account for the following observations:

• Three presynaptic fibres evoke LG IPSPs, the LGs and the ipsi-and contralateral MGs.

• IPSPs evoked by these presynaptic fibres and by some other interganglionic pathway (see Figs. 9 A 2, 3 and 10C) have slightly different forms,

• There is complete occlusion between these same presynaptic fibres (Fig. 9).

If there are three separate pathways excited by the three presynaptic fibres, how can the crossed decrement results be explained? Some active but weak inhibition, or some form of crossed transmitter de-sensitization between nearby inhibitory terminals, which lasts for about 10 sec., could explain the crossed decrement results. The occlusion is more difficult. It is hard to see how this complete lack of summation could occur if there are three separate sets of inhibitory synapses on the LGs at one ganglion. Mutual inhibition between the three presynaptic fibres could not account for the lack of summation when the impulses are synchronous. An asymmetrical inhibition whereby the LGs always blocked response to the MGs and one MG always blocked response to the other seems the only form of inhibition that could account for the occlusion.

Both occlusion and crossed decrement results are, on the other hand, readily explained in terms of a common inhibitory pathway utilized by the three presynaptic fibres. There are two principal possibilities. The first is that each presynaptic fibre excites a population of inhibitory interneurones.. When excited by any one of the presynaptic fibres, each of these interneurones. could then respond maximally so that extra excitation, from one or more further presynaptic fibres, could not change its response. Sharing of these interneurones. would cause the occlusion. Utilization of different numbers of interneurones. would lead to IPSPs of different forms. This is the classical explanation (Creed et al. 1932).

Alternatively, the IPSPs could be produced by activity in a single inhibitory interneurone excited on different branches by the presynaptic fibres. In each of these branches impulses could be generated which would then travel to all parts of the neurone. As a consequence, the presynaptic fibre causing the greatest response (highest impulse frequency) would dominate since impulses originating in other branches of the same neurone would be abolished by impulse collison (cf. Hughes & Wiersma, 1960; Larimer & Kennedy, 1966; Bullock & Terzuolo, 1957). This second explanation is more attractive since arthropods tend to accomplish much with single neurones (e.g. central giant fibres, highly branched motor axons and abdominal stretch receptors).

Though there are still unexplained differences between the LG-evoked slow potentials and the MG-evoked IPSPs, the experiments discussed in the previous paragraphs provide further evidence that these two types of potentials share some common inhibitory component. Also the LG IPSPs cannot be unitary (generated by a single impulse in a single inhibitory interneurone), since a rested unitary PSP could not be graded in amplitude.

It seems very likely that the central giant fibres excite some common inhibitory pathway which produces inhibition of the lateral giants. The latency of the IPSPs would allow two or even more 0·5 msec, synaptic delays. We can now ask about the extent and design of the neurones in this pathway, remembering that there may be one or many such neurones for each segment of the LGs. Figure 11 shows schematic diagrams of some simpler possible interneurone arrangements. An extensive inter neurone excited in one ganglion as in Fig. 11 (1) is ruled out by the demonstration that inhibition can be evoked in single abdominal ganglia. Interneurones. confined to single ganglia (2) cannot alone explain the intersegmental conduction when the MGs are cut. This intersegmental effect cannot be explained by linking up units as in (3) because then there would not be the observed decrease in the amplitude of ISPSs in segments beyond a cut in the MGs. To achieve this, more sites for excitation (4) or more synapses must be added (5) and (6) (cf. Hughes & Wiersma, 1960). Either of these last schemes, if duplicated in each abdominal ganglion, can account for all present observations. The intersegmental connexions would seem to ensure that all ganglia receive some inhibition even if the impulse fails to propagate completely along a giant axon. Whether this double assurance effect is their true function is not clear.

Two functions are suggested for the recurrent inhibition in the crayfish giant-fibre system. The inhibition starts very shortly after an LG impulse and is therefore likely to prevent the LG from giving more than one impulse in response to a single, large EPSP. This role is suggested by the fact that when inhibition is removed (by picrotoxin) the LG responds to an EPSP with a burst of impulses, something that never happens normally. Alternatively, the LG response could be limited by some feature of the pathways exciting the LGs. This would be complicated for a system with as many input sites as the LGs (Johnson, 1924; Kao, 1960). The desirability of some control over the number of impulses that the LGs give is emphasized by Fig. 2 which shows that even a closely spaced pair of impulses in the LGs result in a marked prolongation of the tail flip.

The first role suggested for the recurrent inhibition is the limitation of the response of the LGs to a single impulse per excitatory volley. The inhibition thus helps in the formation of the command to escape. The second role is the limitation of the repetition rate of these commands.

Though in some cases it seems a little brief (e.g. 40 msec.), the duration of the LG inhibition is appropriate to suppress impulses for the period of about 70 msec, (see Fig. 1) that it takes the crayfish to flex its abdomen and become reasonably well extended again. The duration of the LG inhibition will therefore limit the repetition rate of impulses in the LGs and ensure that these commands are separated sufficiently so that successive flips do not interfere with one another. It is interesting that this apparatus for determining the cycle rate exists in an isolated part of the central nervous system. Possibly in the intact animal the inhibition can be modified by proprioceptive feedback but this would not appear to be essential (Eckert, 1961).

The similarities between the recurrent inhibition in goldfish Mauthner cells (Furukawa, 1966) and that in the crayfish lateral giants is striking. Though there are differences, mainly as a result of the paired symmetrical arrangement in fish, the inhibition in the two animals must surely serve very similar functions. Those suggested here would be in addition to the reciprocal role of the Mauthner cell inhibition which has been emphasized by previous workers (e.g. Diamond & Yasargil, 1966). It would be interesting to relate the duration of the ‘late collateral inhibition’ in Mauthner cells (Furukawa & Furshpan, 1963) to the time course of a goldfish tail flip.

To conclude, these relationships between the inhibitory mechanisms in fish and crayfish suggest that the phenomena studied may be fairly general features of command systems evoking cyclic responses of predictable duration.

I would like to thank Dr T. H. Bullock for guidance, and support through grants from the National Science Foundation and the U.S. Public Health Service. Drs T. H. Bullock, B. M. H. Bush, J. T. Enright, A. D. Grinnell, S. Hagiwara, F. B. Krasne, D. Lange and J. P. Segundo have offered criticisms of drafts of the manuscript and helped in many other ways.