Electrical Interactions Between the Giant Axons of A Polychaete Worm (Sabella Penicillus L.)

Giant axons are conspicuous nervous structures which have evolved independently in a variety of invertebrate groups, notably in annelids, arthropods and cephalopod molluscs (Bullock & Horridge, 1965), for the rapid and reliable conduction of nervous impulses involved in certain critically important reflex responses.

In annelids, which exhibit considerable variation in the structural organization of giant axon systems, these axons appear to share a common involvement in the rapid withdrawal responses that apparently confer valuable selective advantages (Nicol, 1948a,Dorset, 1978). These systems are most highly developed in sabellid poly-chaetes, in which the giant axons can be exceptionally large, generally extend for the entire length of the nerve cord and occur in all species in the family (Nicol, 1948a). This is seen most dramatically in Myxicola infundibulum, in which the extremely rapid escape response is mediated by a single very large axon, of up to 1·0 mm in diameter (Nicol & Young, 1946; Nicol, 1948b), in which action potentials are propagated with a velocity of 20 m s⁻¹ (Nicol, 1948a). Other sabellids possess two giant axons, rapid and symmetrical body contractions being achieved by their synchronization via an anastomosis in the brain (Nicol, 1948 a) that enables the action potentials from one giant axon to be transmitted to the other (Bullock, 1953). Synchronization of the activities of these paired giant axons was also observed in posterior segments of the bodies of Protula and Spirographis, in which ‘cross-over’ of impulses occurs in a synaptic manner (Bullock, 1953). The existence of such transmission between the giant axons was subsequently confirmed by Hagiwara, Morita & Naka (1964) in Eudistyliapolymorpha, using intracellular recording techniques. These authors showed that an action potential induced a potential change in the other axon that resembled a synaptic potential and which could initiate an action potential. They concluded that electrotonic spread would be insufficient to explain this transmission and postulated that it is mediated by numerous chemical synapses, while recognizing that ‘no histological evidence has been obtained of such connexions’.

In this paper we describe results which provide electrophysiological and structural evidence for electrical transmission of impulses between the paired giant axons in the mid-body region of Sabella penicillus.

Specimens of Sabella penicillus L., trawled from the Tamar Estuary, were obtained from the Marine Biological Association, Plymouth, U.K. They were kept in marine aquaria at Cambridge in circulating sea water (at 8–10 °C) before use.

The electrophysiological experiments were carried out using isolated lengths of body wall, containing the paired nerve cords and giant axons, as previously described (Carlson, Pichón & Treherne, 1978; Treherne & Pichon, 1978).

One or both of the giant axons were stimulated, at a distance of at least 2-0 cm from the recording electrodes, by rectangular current pulses (50–300 μs in duration) from a Farnell pulse generator through an R.F. isolating unit and a pair of platinum wire electrodes. Resting and action potentials were recorded simultaneously with up to three intracellularly located KCl-filled glass microelectrodes, with resistances of from 5 to 20 MΩ, via separate high-impedance, negative capacitance, converters (W.P. Instruments Inc., Model M 701) a Tektronix 561 oscilloscope and a Medelec, four-channel, fibre optic oscilloscope. In some experiments, hyperpolarizing and depolarizing current pulses were delivered from the Farnell pulse generator, via one of the impedance converters, through an intracellularly located microelectrode.

The artificial sea water used in this investigation had the following composition Na⁺, 470 mM, K⁺ 10 mM, Ca²⁺ 11 mM, Mg²⁺ 55 mM, Cl^- 609·7 mM and HCO₃⁻ 2·3 mM. Variations in calcium and magnesium concentrations were accommodated by appropriate alteration in sodium concentration. Low chloride saline was produced by substituting NaCl for sodium methyl sulphate.

For ultrastructural studies the ventral nerve cord and giant axons were flooded with fixative in situ for 15–20 m. The ventral nervous system, together with a strip of ventral body wall, was then removed and left in fixative at room temperature from 2 to 15 h.

The fixative solution was modified from Baskin (1971) and included 2% glutaraldehyde plus 0·5 M sucrose in 0·1 M phosphate buffer, pH 7·4, plus 1 drop of 1 % CaCl₂ per 10 ml fixing solution, followed by washing in 0.1 mM buffer, pH 7·4, plus 0·64 M sucrose. Subsequent treatment included 1 h in 1 % OsO₄ in o-i M buffer, plus sucrose. After this, the tissues were stained for 30–60 min en bloc, in 2% aqueous uranyl acetate, dehydrated through an ascending series of ethanols, treated with propylene oxide and embedded in Araldite. Thick sections were cut both transversely and longitudinally through the ventral nerve cord and its associated giant axons and were stained with toluidine blue for examination under the light microscope. These sections were often cut serially through the region of the septum of the body wall in order to clarify the structural details of the relationship between the giant axons and the nerve cord. At intervals, thin sections were cut, mounted on grids, stained with uranyl acetate and lead citrate and examined in a Philips EM300.

Tissue was also fixed after 4% procion black was directly injected into the giant axons; this induced enhancement of the constriction of the giant axons at the septa. The fixing solution for this was similar to that outlined above. After embedding, both thick and thin sections were cut at the regions where the constriction had occurred.

Intracellular recordings were made with microelectrodes inserted into each of the two giant axons (ca. 200 μm in diameter) in a segment from the mid-body region. As shown in Fig. 1, the giant axons lie at the inner margins of the paired nerve cords which are linked by a pair of commissures in each segment of the body (see also Fig. 13).

When one of the two giant axons was stimulated in action potential was first recorded in that axon and, after a delay of at least 0·7 ms, an action potential was recorded in the other axon in the recording segment (Fig. 2). With increasing stimulus frequency, the delay between the action potentials increased and a pre-potential, reminiscent of a synaptic potential, became more obvious before the rising phase of the action potential recorded in the distal giant axon. As shown in Fig. 2, transmission between the giant axons eventually failed during prolonged stimulation at frequencies of between 3 and 10 Hz.

In the experiment illustrated in Fig. 3, action potentials were recorded in the two giant axons in the same, mid-body, segment following stimulation of, first, the right (Fig. 3B) and, subsequently, the left giant axon (Fig. 3C). Essentially similar responses were recorded in the unstimulated axons when either the left or right ones were stimulated. When both giant axons were stimulated simultaneously (by placing the stimulating electrodes across both giant axons) synchronous action potentials were recorded (Fig. 3 A). When the giant axon was cut (three segments before the recording segment) at a distance of approximately 20 mm from the stimulating electrodes, delayed action potentials with characteristic pre-potentials were recorded following stimulation of the other axon. This eliminates the possibility of any direct stimulation of this axon.

In a few preparations, deterioration occurred in transmission between the giant axons. This was seen in the progressive prolongation of the pre-potential and in the delay of the associated action potential in the indirectly stimulated axon (Fig. 4). This alteration cannot be attributed to effects on the giant axons themselves, for normal directly excited action potentials persisted even after impulse transmission between them had ceased. At a critical stage, when the delay between the directly and indirectly stimulated action potentials had increased to 5·8 ms, a second action potential was sometimes evoked in the first axon, shortly after the delayed action potential had occurred in the indirectly stimulated one (Fig. 4D). Thus if transmission of an action potential between the axons is delayed sufficiently, so as to fall outside the refractory period of the initially stimulated axon, then it is possible for it to re-excite the axon. These observations also show that delayed transmission in one direction can occur in the presence of more rapid transmission in the other.

The above observations demonstrate that transmission can occur between the giant axons in Sabella and that an action potential in one can initiate a pre-potential, resembling a synaptic potential, which precedes the action potential in the other giant axon. As concluded by Hagiwara et al. (1964), in studying Eudistylia, this indicates the possibility of some chemical synaptic connexion between the giant axons. Experiments were, therefore, performed to test the effects of alterations in the concentrations of external divalent ions which are known to affect chemically mediated synaptic mechanisms.

Exposure of preparations to calcium deficient saline (50 μM-Ca²⁺) had no significant effect on the transmission of action potentials between the giant axons. Elevated concentration of magnesium ions (110 mM) and addition of cobalt ions (30 mM) to the saline bathing the preparations also failed to block transmission between the giant axons (Fig. 5). The cobalt ions caused an increase in the delay between the action potential recorded in the two axons and a slight increase in the duration of the in- directly stimulated action potential (Fig. 5 B). However, this effect can be related to the effects on the directly stimulated axon in which there was a noticeable increase in the duration of the potential.

Reduction in the chloride concentration of physiological saline, by a variety of anion substitutes, is known to be effective in uncoupling the junctions which mediate electronic communication between nerve cells (cf. Asada & Bennett, 1971). In these experiments substitution of NaCl (by sodium methyl sulphate 470 mM) was found to have no significant effects on the transmission of action potentials between the giant axons (Fig. 6).

This metabolic inhibitor is known to produce rapid electrical uncoupling of the low resistance junctions, for example in the septate lateral giant axons of the crayfish (Peracchia & Dulhunty, 1976). The effects of dilute dinitrophenol (0·5 mM) was, therefore, tested on the transmission of action potentials between the giant axons. As shown in Fig. 7, the metabolic inhibitor had no effects on the transmission of action potentials such as were observed in the crayfish lateral giant axons. Here the effects on transmission were presumed to be a consequence of the accumulation of free intracellular calcium ions (Peracchia & Dulhunty, 1976) or, possibly, as suggested for embryonic cells (Turin & Warner, 1977), changes in intracellular pH.

In these experiments alternating hyperpolarizing and depolarizing current pulses (500 ms duration) were passed into one giant axon. The resultant displacement of membrane potential was simultaneously recorded in both giant axons in the same segment. The relationship between applied current and the resulting voltage changes is illustrated in Fig. 8. The slope of the 1/V curve obtained from the same axon is similar to that obtained in other neurones. The attenuation factor for electrotonic voltage spread between the axons is approximately 10 to 1 (Fig. 8). There is thus appreciable attenuation in the passive spread of voltage between the axons, which is however not as great as that observed in Eudistylia by Hagiwara et al. (1964) in which the attenuation factor was greater than 30.

Experiments in which the follower axon was hyperpolarized (by current delivered through a second intracellular microelectrode) lend support to the idea that transmission of excitation between the giant axons is electrical in character. As shown in Fig. 9, increasing hyperpolarization was associated with action potential blockage and a reduction in size of the prepotential recorded in the indirectly excited axon following initiation of impulses in the directly stimulated axon. These data are inconsistent with those expected from a conventional chemically mediated synapse, in which the post-synaptic current density would be increased during post-synaptic hyperpolarization. The continuous decline in amplitude of the pre-potential, as observed, suggests that the strength of excitatory current responsible for the pre-potential was reduced in proportion to the degree of hyperpolarization. One explanation for this effect could be a progressively more distant blocking of regenerative activity in pathways between the giant axons.

In these experiments a length of worm containing eight successive pairs of commissures was isolated by severing all other commissures anteriorly and posteriorly (Fig. 10). The effects of successive cutting of the remaining commissures on transmission between the giant axons was then recorded as indicated in Fig. 10.

Impairment of transmission between the axons did not occur until only four pairs of commissures remained intact (i.e. at stage 5 in Fig. 10). At this stage transmission became intermittent. With only three pairs of commissives intact the secondary spike response was abolished (stage 5 in Fig. 10). Further cutting of commissures resulted in the slight reduction of the depolarizing response which had preceded the action potential in the indirectly stimulated axon. These observations suggest that the nerve cord commissures are involved in transmission of action potentials between the giant axons, that a minimum of four adjacent pairs are necessary for post-axon excitation to occur and that current density from the directly excited axon is reduced in a graded manner by successive elimination of commissures.

More direct support for the possibility of impulse passage along pathways between the giant axons was provided by intracellular microelectrode recordings from within the commissures which link the nerve cords associated with the giant axons (see Fig. 1). In these experiments three microelectrodes were used to record simultaneously in the two giant axons (in the same segment) as well as in one of the pair of commissures.

Various impalements in the commissures indicated the penetration of numerous nervous processes. In six cases excitable responses were recorded which were all associated with the activity of the giant axons. As shown in Fig. 11, the responses measured in the commissures often consisted of a fast rising prepotential (apparently due to activity in the directly stimulated giant axon) superimposed upon which was a more rapidly rising excitable component having an amplitude comparable to that of the impulses in the giant axons. These responses were obtained in both the anterior and the posterior of the commissure of the pairs, although not in the same preparation.

Although the commissures were impaled apparently midway between the giant axons the action potentials recorded there often occurred synchronously with those recorded in the indirectly stimulated giant axon (Fig. 11B). This appears to be an artifact of impalement, for extracellularly recorded action potentials or those measured during partial impalement (Fig. 12), occurred midway in time between the action potentials recorded in the directly stimulated and the follower axon.

As illustrated in Fig. 1, Sabella possesses two ventral nerve cords which are joined in each segment at the level of the septum by two commissures. Each nerve cord contains a single giant axon, of up to 250 μm in diameter (Fig. 13A), together with many small axons of variable diameter (Fig. 16).

The giant axons contain mitochondria, fibrils, and vesicles of smooth membrane but few neurotubules (Fig. 16, inset); many nerve terminals containing both small electron-lucent and larger electron-opaque granules are found directly apposed to their axolemma, especially in areas where branching occurs (Fig. 17).

At points on either side of the septum where two body segments adjoin, each giant axon can be seen to become flattened as it produces a branch (Figs. 13 B, 13 C and 18) which joins the commissures that connect the two nerve tracts (Figs. 13 C, 14, 15 and 21). These branches are of variable diameter (from 5 μm or less to 10 μm) (Figs. 14 and 15) but their precise geometry, ascertainable only by serial sectioning through a number of commissures, has not yet been fully analysed. Such branches can be identified in thin sections as areas of increasingly attenuated axoplasm. As they run into and become part of the commissures, they contain irregular smooth vesicles, especially near their periphery (Fig. 18) and in some cases these become aligned across the axoplasm. In the areas near the body septa, where the commissures (Figs. 19 and 20) fuse with the ventral nerve tract (Fig. 21 A, B and C) and where the giant axons contribute a process to the commissure (Figs. 13 c, 14, and 15) there is a great reduction in the diameter of the giant axons from 200 μm or more (Fig. 13 A) to about 70 μm (Fig. 13B, C). This appears to be a normal phenomenon, which is, however, considerably enhanced under conditions of stress, such as when foreign materials (Procion black, Fast green or cobalt and horseradish peroxidase) were injected into the axoplasm (Fig. 22). This axonal constriction could account for the difficulties in demonstrating transfer of injected dyes and tracer substances from one giant axon to the other via the commissure processes, as well as between adjacent segments of the same axon. In these experiments only faint staining or reaction product could be detected in the adjacent axon, or segments of the same axon, presumably as a result of the restricted access into the contricted region of the giant axon (cf. Fig. 22).

The other, smaller axons in the nerve cords contain mitochondria and neurotubules, but little smooth membrane (Fig. 16). There are glandular cells, probably containing mucopolysaccharide (Fig. 18) as well as muscle fibres (Fig. 18), characterized by very large paramyosin filaments lying around the ventral nerve tracts, in close association with the giant axons (Fig. 26). There is very little glial investment of the smaller axons although the giant axons may have a number of attenuated glial layers wrapped like a very loose mesaxon round them (Figs. 16, 17, 18 and 26). Such processes are often separated from one another by regions of extracellular matrix containing collagen-like fibres (Figs. 17 and 26); although less frequently two glial cell processes may lie very close to one another and be associated by gap junctions (Fig. 27). The glial cells send attenuated cytoplasmic processes out in several directions from their cell bodies which lie at random in the mass of axons (Fig. 16). The glial cytoplasm is characterized by clumps of fibrillar material (Fig. 27) of considerable electron opacity and often contains extensive Golgi complexes (Fig. 25) ; in some cases, dense bodies which may be gliosomes (lysosomes) (Fig. 17) also occur.

The ultrastructure of the commissures (Figs. 19 and 20) appears to be similar to that of the ventral nerve tract in that they possess many small axons, and a few glial cells (Fig. 23). They have some larger axons in their periphery, as can be seen in cross-sections (Fig. 23), and one of these must represent the branch that arises from the giant axon. The commissures lack a very definite outer perineurial layer, which appears to be formed of loose attenuated cell processes separated by connective tissue (Fig. 23) as occurs in the ventral nerve tract. Many neurosecretory-bearing fibres are present as is also the case in the ventral nerve tract and these are particularly striking in longitudinal section (Fig. 24). Glia once more exhibit the characteristic dense fibrillar cytoplasmic inclusions (Fig. 24) and as in the nerve cord, some muscle fibres lie in the connective tissue sheath around the commissures.

The electrophysiological observations described above fully confirm those of Hagiwara et al. (1964) on Eudistylia in showing a transfer of impulses between the giant axons in each body segment in the mid-body region of Sabella. However, our observations do not support the hypothesis that this transfer is mediated by chemical synapses. This is indicated by the spread of depolarizing and hyperpolarizing voltage changes between the giant axons, and the lack of effects of increased concentrations of magnesium and of cobalt ions in inter-axonal transmission. Furthermore, the effect of hyperpolarization of the indirectly stimulated axon in reducing the amplitude of the depolarizing prepotential is inconsistent with the expected behaviour of a conventional chemical synapse.

The histological evidence shows that the commissures which link the two halves of the central nervous system contain processes from both of the giant axons. These axonal processes are approximately 5–10 μm in diameter and are larger than the majority of the axons linking the nerve cords in the mid-body region. The presence of the axonal processes between the giant axons confirm the prediction of Hagiwara et al. (1964), that ‘there must be some kind of fibre-like connexion between ‘the giant axons, and they provide a structural path for the action potentials recorded in the commissures during transmission between the giant axons in our study.

We lack detailed structural evidence as to the precise nature of the connexion represented by the processes between the giant axons. There are two possibilities. There could be direct cytoplasmic continuity, as occurs in the first order giant axons of the squid (Young, 1939), or there could be electrotonic communication across gap junctions such as occurs, for example, at septate gap junctions between the segmented giant axons of the crayfish (e.g. Perrachia, 1973a, b). The electrophysiological evidence does not support the latter possibility, since the agents which are known to uncouple electronic junctions in other nerve cells (Ca-free saline, dilute dinitrophenol and reduced external chloride concentration) were without effect on impulse transmission between the giant axons of Sdbbella. It is difficult to attribute the lack of effect of these uncoupling agents to restricted intercellular access to the axonal surfaces, for previous electrophysiological observations indicate that there is no appreciable restriction to the intercellular diffusion of water-soluble ions and molecules to the axon surfaces in Sabella in isosmotic conditions (Treherne & Pichon, 1978). The present ultrastructural observations reveal no morphological evidence for restriction since the axons show only loose glial coverage, intercalated with connective tissue, with only occasional gap, not tight, junctions occurring between adjacent glial processes. The available evidence thus tends to favour the possibility that the processes within the commissures are in cytoplasmic communication with both of the giant axons.

The observation of impulse traffic within the commissures indicates that transmissions between the giant axons are not achieved solely, if at all, by passive spread of potential. Furthermore, the ten-to-one attenuation of electronic potentials between the axons also argues against the possibility of an exclusively passive spread of potential along the axonal processes linking the giant axons across the commissures. Thus depolarizing currents passed across an axonal membrane were ineffective in producing action potentials even when they generated voltage drops of 20 mV. The measured action potentials of no mV would therefore be insufficient to excite action potentials in the follower axon with the ten-to-one alteration of a passive potential spread between the axons.

As recognized by Hagiwara et al. (1964) (as a less favoured alternative to the hypothesis of chemically mediated transfer of impulses) the transmission delay could be due to conduction time of action potentials along the connecting processes between the giant axons. In the Sabella preparation the conduction time along the axonal processes (ca. 10 μm in diameter and 0·32 mm in length) can be roughly estimated by comparison with the measured conduction velocity (2·7 ms⁻¹) and diameter ca. 200 μm) of the giant axons. This yields an estimated conduction time for transmission along the axonal processes within the commissures of 0·73 m s⁻¹. This falls short of a minimal measured transmission delay of approximately 0·7 m s⁻¹. This discrepancy can be accounted for by the marginal nature of the transfer of excitation to the follower axon due to the very large size discrepancy between the very much smaller axonal processes and the giant axons themselves. Action potentials occurring along the axonal process within the commissures will be severely loaded as they approach the expanded membrane area of the follower giant axon. Given that the space constant of the giant axons is around 7 mm and that the conduction velocity is 2·7 m s⁻¹, then it follows that at least five pairs of commissural axonal processes will be excited simultaneously within the transmission time of the action potential. We assume, therefore, that the overall density of injected current at the multiple sites on the follower axon are sufficient to overcome the reduction in safety factor imposed by the geometry of the system. Our findings that at least four pairs of intact commissures are necessary for inter-axonal transmission supports this interpretation. However, the marginal nature of the transmission between the commissural processes and the giant axons may also account for additional conduction delay between the measured action potentials in the directly stimulated and follower axons. Thus, the depolarizing prepotential which precedes the follower action potential could be an electrotonic reflexion of the action potential within the processes which increases appreciably the transmission delay between the peaks of the action potentials in the directly stimulated and in the follower axon. This interpretation is supported by our observation that hyperpolarizing potential changes in the follower axon reduce the amplitude of the depolarizing pre-potentials while decreasing the latency of the maximal change of pre-potential (see Fig. 9). The above results would be expected if the imposed hyperpolarization blocks inter-axonal impulses at progressively greater distances from the recording site along the follower giant axon.

Bullock (1953) describes the transmission between the giant axons of sabellid worms as occurring ‘in a synaptic manner, delaying and labile but unpolarized’. However, Hagiwara et al. (1964) showed, in Eudistylia, that repetitive stimulation of one giant axon caused fatigue in transmission but did not affect conduction in the opposite direction, an effect which was attributed to the properties of chemically mediated synapses. Our observations in Sabella, which indicate electrical transmission, also show that apparent polarization of impulse transmission can occur in certain conditions (see Fig. 4). These observations could be explained in terms of geometrical asymmetries in the system which are known to operate along equivalent electrically conducting pathways in other nervous preparations (cf. Tauc & Hughes, 1963; Mellon & Kaars, 1974; Spira, Yarom & Parnas, 1976).

An important aspect of inter-axonal transmission in Sabella is the appreciable constriction of the giant axons, in the septate regions, where the axonal processes originate (Figs. 13 C and 22). The extreme narrowing of the diameter of the giant axons would be expected to increase current density from action potentials within the processes and, consequently, the efficacy of post-synaptic currents generated by other central neurones in this region. It is probably significant that numerous nerve terminals are present, in the constricted region of the giant axon, which contain dense and hollow core vesicles (Fig. 17). It is conceivable, therefore, that transmission between the giant axons could be influenced by synaptic inputs and could, thus, provide an explanation for the otherwise inexplicable observations of Bullock (1953) who described the interactions between the giant axons as being mediated by ‘quasiartificial synapses’, in which transmission occurs ‘at a localized spot, but this spot may shift smoothly along, millimetres in seconds’.

The available evidence shows that the co-ordinated activities of the two giant axons in sabellids is achieved in the brain by a single anastomosis (Nicol, 1948a; Bullock, 1953) and by electrical transmission, apparently, in each segment of the body. As in the earthworm (Wilson, 1961), the presence of the segmental transmission between the giant axons ensures effective synchronization of impulse traffic initiated in any region of the body and, thus, co-ordination of muscular contractions during rapid withdrawal responses of the worm. Furthermore, this segmental transmission preserves the co-ordinated body contractions even in portions of truncated worms, a mechanism which is likely to be of selective advantage in animals capable of body regeneration.

This research was supported in part by a grant from the U.S. European Research Office.