Neuronal Interactions Mediated by Neurally Evoked Changes in the Extracellular Potassium Concentration

It is well established that the concentration of potassium ions can be transiently altered in the restricted extracellular spaces of the central nervous system (CNS), as a result of neuronal activity (Nicholson, 1980a,bSomjen, 1979; Orkand, Nicholls & Kuffler, 1966; Orkand, 1980; Sykova, Hnik & Vyklicky, 1981). The changes in the concentration of K⁺ ions are a direct consequence of two of the main design features of the nervous system: (a) rapid transmission of signals (i.e. action potentials) along axons is associated with the influx of Na⁺ (and in some cases Ca²⁺) into a neurone, and the efflux of K⁺ from it ; (b) the narrow and restricted extracellular spaces (which appear to be a consequence of the evolutionary trend to concentrate nerve cells into tightly packed groups) separating neurones, as well as non-excitable cells, enable the rather minute K⁺ efflux during the falling phase of the action potential to alter effectively the extracellular potassium concentration ([K⁺]_o). Despite the various mechanisms which control [K⁺]_o and prevent its increase, various degrees of K⁺-mediated interactions between neurones are inevitable. Moreover, despite mechanisms that prevent free K⁺ diffusion, the spread of these interactions may be quite significant, involving elements that are not nearest neighbours.

The interactions between active neurones and their inactive neighbours produced by the efflux and redistribution of K⁺ in the extracellular spaces may exhibit a wide range of efficiency. Thus in low efficacy interactions, the effects of increased K⁺ around the inactive neurones can be detected only after massive activity of a large population of neurones, or following the activity of a few neurones firing at very high rates. On the other hand, high efficacy interactions in which a single spike or a short train of action potentials in a single neurone is sufficient to produce a substantial depolarization in an inactive neurone have been demonstrated (Spitzer, 1976; Alkon & Grossman, 1978; Yarom & Spira, 1982; Spira & Yarom, 1983a,b). The efflux of K⁺ from an active axon may also alter the properties of the active axon itself (Franken-haeuser & Hodgkin, 1956) and thereby act as a self modulator. For example, it can block propagation along an axon with special geometry by reducing the safety factor (Parnas, Spira, Werman & Bergman, 1969; Spira, Yarom & Parnas, 1976), or it can block the propagation of impulses into one branch of a bifurcating axon and thereby channel impulses in one branch only (Grossman, Spira & Parnas, 1973; Grossman, Parnas & Spira, 1979a,b; Parnas, 1972, 1979).

The extent to which the activity of neurones affects inactive ones depends on a large number of factors which include: (a) the amount of K⁺ liberated from the active neurones; (b) the dimensions, shape and properties of the extracellular space; (c) the activity of neurones and glial cells in regulating the [K⁺]_o and, (d) the membrane properties of the affected neurones, especially their sensitivity to K⁺ concentration gradients across their membrane (Somjen, 1979; Nicholson, 1980a,b).

The highly ordered organization of neurones in brains and ganglia into defined aggregates and axonal tracts is associated with the formation of a highly structured network of extracellular spaces. As mentioned previously the three dimensional structure of these extracellular pathways, and their properties, define the efficiency of potassium-mediated interactions. Moreover they define the spatial specificity of the interactions. Accordingly, the spatial specificity of neuronal interactions mediated by K⁺ ions may exhibit some variety. Thus we find local interactions, for example, the interaction between A photoreceptors and hair cells of Hermissenda (Alkon & Grossman, 1978) or the interactions between adjacent giant interneurones of the cockroach (Yarom & Spira, 1982). Less restricted but spatially defined interactions between a given group of neurones can also be observed, as in the case of the primary afferent terminals in the spinal cord of various vertebrates (Barron & Mathews, 1938; Eccles, Magni & Willis, 1962; Kriz, Sykova, Ujec & Vyklicky, 1974; Krnjevic & Morris, 1972; Vyklicky, Sykova & Kriz, 1972; Vyklicky, Sykova, Kriz & Ujec, 1975; Rudomin, Stefani & Werman, 1979). The defined extracellular pathways that normally restrict communication between neurones may not function under various pathological conditions. For example, during epileptic episodes, K⁺ accumulation may affect a large population of neurones which are normally isolated from one another (Moody, Futamachi & Prince, 1974; Prince, 1974, 1978).

We will describe experiments which demonstrate various levels of interactions mediated by local changes in potassium ion concentration. The giant axons (Gax) and non-giant axons from the central nervous system of the cockroach Periplaneta americana were used. The types of interactions discussed are: (a) interactions among populations of neurones induced by superfusion of the CNS with the convulsant drug picrotoxin, (b) restricted and limited interactions which are the consequence of the combination of the special geometry of the Gaxs in the metathoracic ganglion (T₃) and the accumulation of K⁺ in the periaxonal space; and finally, (c) local interactions among the Gaxs in T₃ which are postulated to be mediated by K⁺ ions. The experiments on which most of the present paper is based have been published previously (Spira et al. 1976; Spira & Yarom, 1983a,b; Yarom & Spira, 1982, 1983; Yarom, Grossman, Gutnick & Spira, 1982).

Each connective from the ventral nerve cord of the cockroach Periplaneta americana contains eight giant axons (Gax), each with a relatively large diameter (20–50 μm in adults). Their cell bodies are in the last abdominal ganglion (A₆), and it is here that their dendrites receive synaptic inputs from afferents originating in the cerci (Pumphrey & Rawdon-Smith, 1937; Roeder, 1948; Callec, Guillet, Pichon & Beistel, 1971). The afferents are connected to filiform hairs that are extremely sensitive to changes in direction and velocity of air movements (Westin, Langberg & Camhi, 1977). The axon of each giant is continuous through the ventral nerve cord (Fig. 1) (Farley & Milburn, 1969; Spira, Parnas & Bergman, 1969a,b), the suboesophageal ganglion, and terminates in the supraoesophageal ganglion (D. Zeldes & M. E. Spira, unpublished observations). In the abdominal connectives and ganglia the Gaxs are organized into a dorsal and a ventral group, each containing four fibres. The axons can be individually identified, histologically, since each assumes a characteristic position within the group (Fig. 1B) (Harris & Smyth, 1971). In each ganglion the fibres send off several neurites which extend and branch into the neuropile (Fig. 1C). A particular Gax has a similar number of neurites in each ganglion with similar patterns of branches from cockroach to cockroach (Yarom & Spira, 1983). The axon of a giant interneurone is wrapped by a glial envelope, often several layers thick (Fig. 2A,B) (Castel, Spira, Parnas & Yarom, 1976; Meiri, Dormann & Spira, 1983; Lane, 1974, 1981). The base of the neurites which extend from the axon are also covered by glial processes, but the distal branches of the neurites are virtually devoid of glial covering (Fig. 3A,B) (Castel et al. 1976). Other unidentified axonal profiles which either run parallel to the longitudinal axis of the Gax or cross the ganglia are in close proximity with the Gax neurites (Fig. 3A). The gap between the Gax neurites and other axonal profiles within the neuropile typically ranges between 15–20 nm. However, regions in which the extracellular space between the Gax’s neurites and other axonal profiles is reduced to 8–10nm are also observed (Fig. 3B) (Castel et al. 1976; Yarom & Spira, 1982).

The isolated ventral nerve cord of late nymphal, or adult male cockroaches Periplaneta americana was fixed to the bottom of a Perspex chamber, ventral side up. The metathoracic ganglion (T₃) and the connectives at its base were mechanically desheathed. Intracellular recordings were made from one or two giant axons at the caudal or rostral edge of T3. Extracellular stimulation and recording were made by either suction or hook electrodes.

Recording from a Gax while stimulating either the ipsi-or contralateral connectives at intensities below the threshold of the recorded Gax itself or at low frequencies (1–10 Hz), reveals that it is functionally well insulated from other axons. The membrane of the impaled Gax is not depolarized despite the activation of a large number of axons and measurable build up of [K⁺]_o (Grossman & Gutnick, 1981). However, this insulation becomes insufficient under conditions which promote massive activation of the nerve cord, such as high frequency stimulation, or following superfusion with a convulsive agent.

After several minutes of superfusion of the isolated ventral nerve cord with the convulsant agent picrotoxin (10⁻⁵ mol l⁻¹), a single extracellular stimulus (which under normal conditions produces a short burst of impulses) produces a very long (1 s) burst of activity in many fibres (Fig. 4A). This burst is associated with a prolonged depolarization (PD) of the giant axon. The amplitude of the PD can reach 10 mV and last for 10 s (Yarom et al. 1982) and is associated with a large increase in membrane conductance of the Gax (Fig. 4B,C).

The observed conductance increase can be explained by either rectification of the Gax’s membrane (i.e. to a voltage-dependent decrease in membrane resistance; Spira et al. 1976), or to prolonged synaptic activity in T₃, (i.e. to a voltage-independent conductance). To distinguish between these possibilities, the effective membrane resistance of the Gax was measured at different steady state membrane potentials induced by current injection. The plot in Fig. 4C demonstrates that the conductance change during the PD can be entirely explained by membrane rectification.

Since the PD is associated with intense burst activity in the connectives and since the conductance change during the PD can be entirely accounted for by a voltage-dependent conductance increase, we speculated that the PD is a consequence of Accumulation of K⁺ in the periaxonal space. This possibility was confirmed by direct measurements of [K⁺]_o using ion-sensitive electrodes placed as close as possible to the Gax (Fig. 5D,E). An additional estimate of the changes in [K⁺]_o in the immediate vicinity of a Gax membrane was made by examination of the amplitude and absolute peak value of the Gax spike after-hyperpolarization (AHP) before, during and after the PD (Fig. 5). The AHP peaks at a membrane potential (V_m) close to the reversal potential of potassium ions (E_K) (Frankenhaeuser & Hodgkin, 1956; Baylor & Nicholls, 1969; Brodwick & Junge, 1972). In the experiment illustrated by Fig. 5A, action potentials were triggered during the course of a PD. The peaks of the AHPs were more depolarized, indicating that EK is changed during the PD. These changes, which closely followed the PD time course, are not due to changes in the membrane potential (Fig. 5B). Subtraction of the PD from the changes in E_K (as measured from the shift of the AHP) reveals that the PD cannot be fully accounted for by the change in E_K (Fig. 5C). The mechanism underlying this discrepancy is not known. However, several independent mechanisms can account for it: (a) contribution of chemical synaptic potentials to the PD amplitude; (b) a direct action of increased [K⁺]_o on other ionic conductances (Adelman & Palti, 1969); (c) the depolarization of the membrane by raised [K⁺]_o itself increases conductance to other ions, such as sodium or calcium and thereby increases the membrane potential change; (d) assuming that the resting potential of a Gax is to some extent (a few mV) influenced by an electrogenic pump, then the large conductance increase during the PD would shunt the hyperpolarizing current generated by the pump. It is unlikely that the discrepancy between the PD amplitude and the change in EK is due to activation of chemical synapses, since a similar discrepancy is seen when potassium ions are iontophoresed in the vicinity of a Gax in the connective, where chemical synapses are absent and the nearest ones are more than 1 mm away in the ganglion. Thus the most likely possibilities to explain the added depolarization involve direct or indirect activation of cationic conductances in addition to the changes in EK produced by K⁺ accumulation.

The effects of picrotoxin in producing the PD were far more prominent in nymphal than in adult stages. This difference may be mainly attributed to the fact that the insulation to K⁺ diffusion between Gaxs and other axons is better in the adult then in nymphal stages (Grossman & Gutnick, 1981).

These experiments illustrate that, as a consequence of treatment which alters the normal pattern of neuronal activity and produces massive synchronized activity of large populations of neurones, the glial envelope that normally provides insulation between neurones is no longer capable of doing so. Under such conditions, interactions among groups of neurones that are normally isolated take place, and phenomena such as epileptogenic foci may be seen (Moody et al. 1974; Prince, 1974, 1978).

A more localized interaction due to change in [K⁺]_o is observed around a Gax at T₃ when the connectives are electrically stimulated at high frequencies.

In the isolated adult ventral nerve cord, the giant axons can transmit single action potentials in both ascending and descending directions. When the stimulus frequency is increased above 30–40 Hz, however, intermittent conduction failures are observed at the level of the thoracic ganglia but not at the abdominal ones. Complete blockage of impulse propagation at T ₃ can be produced after stimulation at frequencies around MO Hz, or after long periods of stimulation at 30 Hz.

The events leading to, and associated with the conduction block are illustrated in Figs 6 and 7. In general, during stimulation at high frequency, the following changes are observed: (a) consecutive reduction in the amplitude of either ascending or descending action potentials approaching ganglion T₃ (Figs 6, 7); (b) a gradual decrease in the amplitude of the after-hyperpolarization associated with membrane depolarization of 10–20 mV (Figs 6, 8); (c) a gradual appearance of a pre-potential and a delay in the initiation of the action potential across the ganglion (Fig. 6) (under these conditions the delayed action potential spreads back across the ganglion and is seen as a reflection on the falling phase of the pre-ganglionic action potential) (Fig. 7) ; (d) intermittent failure of the action potential to propagate across the ganglion (Figs 6, 7) (under these conditions only a decremental potential, resulting from electrotonic spread can be seen on the far side) ; (e) after cessation of stimulation and a period of rest, the conduction of action potentials across T₃, and the resting potentials recover as normal.

The development of conduction block at restricted and defined regions along a Gax at the thoracic ganglia is a consequence of the combination of two factors; the inhomogeneous geometry of the giant fibres in T₃ and the increase in Gax membrane conductance at these regions. While passing through the metathoracic ganglion the giant axon diameter is reduced to about half of its value in the connectives. This narrowing provides the morphological basis for the formation of a low safety region for impulse propagation (Parnas, Hochstein & Parnas, 1976;,Yarom & Spira, 1983). During high frequency stimulation of the connectives, the Gax membrane is depolarized and the membrane conductance is increased. These changes further decrease the safety factor for impulse propagation leading eventually to conduction block (Spira et al. 1976). The increased conductance at T₃ can be attributed to two independent factors: activation of chemical synapses which terminate on Gax’s neurites (Spira et al. 1976; Yarom & Spira, 1983), and Gax membrane rectification following depolarization induced by accumulation of K⁺ ions.

If high frequency conduction block is mainly due to voltage-dependent membrane rectification and not to voltage-independent changes (i.e. chemical synaptic inputs), it would be expected that membrane hyperpolarization to resting level should restore the membrane resistance and restore conduction. The experiment illustrated in Fig. 9 confirms this assumption: the conduction blockade produced by descending impulses at 50 Hz was released by membrane hyperpolarization. Further support for the contention that the block is not produced by voltage-independent processes is provided by the observation that conduction block can be produced under conditions in which chemical synapses are unlikely to be activated, namely at low Ca²⁺ (1 mmoll⁻¹), high Mg²⁺ (9 mmol l⁻¹) concentrations.

Finally, direct recording of potassium activity with ion-sensitive electrodes in the vicinity of a Gax in T₃ revealed that the high frequency conduction block is associated with an increase in extracellular K⁺ concentration (Grossman & Gutnick, 1981).

The phenomenon of conduction block in the thoracic ganglia is attributed to a combination of local increase in Gax membrane conductance and the changes in the shape of the axons in T₃ (Parnas et al. 1969, 1976; Spira et al. 1976). Evidence for the importance of the special structure of the Gax in producing the low safety region r impulse propagation was recently provided by comparison of the properties of Gaxs from adult and nymphal cockroaches (Spira & Yarom, 1983a; Yarom & Spira, 1983). The Gaxs of nymphal cockroaches do not narrow extensively while traversing T₃. In spite of the fact that the rate of K⁺ accumulation around the Gax in nymphal stages is faster, and the extent of K⁺ accumulation is larger, conduction block cannot be produced in early nymphal stages. The safety factor for spike propagation is reduced gradually as the nymphs develop into the adult form (Spira & Yarom, 1983a). Thus the phenomenon of conduction block at T₃ is a localized one in which the efflux of K⁺ ions from many fibres in the ganglia affects the capacity of a single Gax to conduct impulses at high frequencies. It is not known yet whether this interaction plays any physiological role. However, mean frequencies of 300 spikes s⁻¹ and instantaneous frequencies of up to 900 s⁻¹ have been recorded from Gaxs in response to physiological stimuli (Westin et al. 1977). These frequencies are far beyond those required to produce conduction block in the isolated ventral nerve cord.

Adjacent Gaxs interact with each other in T ₃ by a reciprocal synapse (Gin-to-Gin synaptic potential; GGSP). These interactions appear to be due to the efflux of potassium ions during the falling phase of an action potential of one Gax into a restricted and defined extracellular space between it and an adjacent Gax (Yarom & Spira, 1982). The reciprocal interaction is illustrated in Fig. 10. In this experiment two adjacent Gaxs were impaled by microelectrodes at the base of T ₃; each electrode could serve for either current injection or voltage recording. Generation of an action potential in one Gax (Fig. 10A) evokes a synaptic response in the other, and vice versa (Fig. 10B). The GGSPs differ in several features from chemical synapses recorded in the same region (T₃) and from the chemical synapses between the cereal nerve and the Gax in the last abdominal ganglion (A₆). The GGSP has a shorter latency than potentials produced by monosynaptic chemical junctions (0·1–0·2ms, and 0·5–1 ms respectively at 23 °C, Bennett 1972, 1974) (Fig. 10D). Furthermore the GGSP has a faster rise time (1–1·5 ms) than chemical PSPs (2·8–3 ms) recorded in the same region and its decay time is slower (60–100 ms) than the decay of potentials produced by chemical synapses (about 5–10ms) (Fig. 10).

The GGSP differs from conventional chemical synapses in three additional major aspects. First, it is insensitive to cobalt ions. A compound chemical PSP generated by cereal nerve stimulation is blocked after 10 to 15 min of superfusion with 5 mmol I⁻¹ cobalt ions, the GGSP is almost unaffected (Fig. 11), though in some experiments the GGSP decay time becomes brief and its amplitude is slightly reduced. The mechanisms underlying these changes are still not clear. They can be attributed to an effect of cobalt ions on a delayed release of transmitter, or to delayed Ca²⁺-dependent potassium currents in the neurites (Meech & Strumwasser, 1970; Meech, 1978; Yarom & Spira, 1982; Hounsgaard & Nicholson, 1983). Second, the amplitudes of depolarizing PSPs generated by stimulation of the contralateral thoracic connective are largely increased by hyperpolarization of the Gax membrane, the GGSP is almost unaffected (Fig. 12). The extrapolated reversal potential of the GGSP is between 0 and −10 mV, and that of other synapses is between −57 and −65 mV. An apparent reversal potential of 0 to −10 mV is expected if the GGSP results from an increase in the extracellular K⁺ concentration (Rudomin et al. 1979). Third, it is insensitive to change in temperature. By contrast the synaptic delay of a chemical synapse is extremely sensitive to temperature, with a Q₁₀ as high as 3–4 (Katz & Miledi, 1965; Datyner & Gage, 1980). The experiment of Fig. 13 illustrates the effect of lowering the bath temperature from 21 to 12 °C on a GGSP. Unlike the chemical responses (see Fig. 11), no clear delay between the peak inward current of the ‘presynaptic’ Gax spike and the GGSP can be recorded. Lowering of the temperature caused: (1) a decrease in the conduction velocity of the action potential along the Gax, (2) an increase in the peak amplitude of the action potential, (3) an increase in duration of the Gax action potential and (4) an increase in the peak value of the AHP. These changes are also reflected in the extracellular recording of the ‘presynaptic’ action potential by the intracellular electrode in the second Gax. The alterations in the shape of the action potential can be attributed to a decrease in the rate of activation of the potassium conductance. Unlike conventional chemical synapses, lowering the temperature did not produce changes that reveal a clearly discernible synaptic delay. The increased duration between the peak of the inward current of the ‘presynaptic axon’ and the GGSP can be entirely attributed to the effect on the rate of K⁺ efflux during the repolarizing phase of the action potential. The decreased rate of K⁺ efflux from the ‘presynaptic’ Gax may also be the cause of the slowing of the GGSP rate of rise.

The reciprocity, relatively short latency, fast rise time, insensitivity to cobalt ions and insensitivity to temperature and transmembrane voltage are inconsistent with the known properties of conventional chemical synapses. These properties resemble to some extent the properties of electrotonic synapses (Bennett, 1977). Three lines of evidence, however, argue against this possibility. First, hyperpolarizing or depolarizing current pulses sufficient to produce −40 to +20 mV shifts in membrane potential of one Gax did not spread to a second Gax (Yarom & Spira, 1982; Spira & Yarom, 1983b). Second, intracellular injections of the fluorescent marker Lucifer Yellow, known to traverse gap junctions in invertebrates, do not reveal evidence of spread from one Gax to another. Third, TEA, which can cross invertebrate low resistance junctions (Deschenes & Bennett, 1974), does not cross between Gaxs that produce GGSPs (Yarom & Spira, 1982).

Through elimination of the possibilities that the GGSP is mediated either by a conventional chemical or an electrotonic synapse, it has been suggested that the GGSP is the result of efflux of potassium ions during the falling phase of a spike in one Gax into a narrow gap between two Gaxs (Yarom & Spira, 1982). This possibility was supported by the observation that intracellular injection of TEA, which decreases the outflow of K⁺ during an action potential (Armstrong, 1969; Hille, 1967), decreases the rate of rise and amplitude of the GGSP (Yarom & Spira, 1982).

In previous sections we have emphasized that the glial envelopes around the Gaxs in the connectives and the ganglia prevent cross-talk. However, distal regions of the neurites which emerge from two Gaxs do come into close proximity. The space between the neurites in these regions is only 8–10nm (Fig. 3B). Several such junctions have been observed among neurites of adjacent Gaxs. These contact regions may serve as the morphological substrate for the K⁺-mediated reciprocal interactions among the Gaxs.

We have described three levels of neuronal interactions mediated by changes in extracellular potassium concentration. The first level is non-specific interactions which are produced as a consequence of abnormal and massive activation of neuronal networks in the CNS. Under these conditions the amount of potassium liberated from a large number of neurones overcomes the regulating capacity of the highly organized insulating structures of the glial envelopes (Lane, 1974, 1981) as well as the capacity of the extracellular matrix to regulate the concentration of ions in extracellular spaces (Treherne & Schofield, 1981).

The second level is interactions that take place in a more restricted fashion and were demonstrated by the blockade of impulse propagation following high frequency stimulation of the Gaxs. Whereas impulse propagation along the abdominal ganglia is insensitive to stimuli at high frequency, in T₃ it is blocked as a consequence of two factors: (a) the non-homogeneous geometry of the Gaxs in T₃ and (b) the local conductance increase of a Gax membrane due to potassium accumulation. Although a high frequency of impulses in the abdominal connectives was recorded in response to physiological stimuli, it is not clear that the conduction block appears, or plays any role in the intact animal. It is possible that the non-specific electrical stimulation of the connectives activates pathways which are normally not activated at such frequencies or in such patterns in the intact animal. In any event, conduction block is a reflection of the structural relations between the various neurones, glial envelopes and extracellular spaces in the thoracic ganglia only. It is of importance to note however, that conduction block and information channelling in bifurcating axons as a consequence of potassium ion accumulation in the periaxonal space of peripheral neurones occur in isolated preparations and in semi-intact and even whole animals (Grossman et al. 1973, 1979a,b; Parnas, 1972, 1979; Parnas & Segev, 1979).

The third level of interactions are specific and local ones, like those provided by conventional chemical and electrotonic synapses. Local interactions of this type mediated by potassium ions have been demonstrated in several animals; between type A photoreceptors and hair cells in Hermissenda (Alkon & Grossman, 1978) ; between the post-and presynaptic elements of the squid giant synapse (Erulkar & Weight, 1977) ; between the giant axons of the cockroach (Yarom & Spira, 1982), and, under certain conditions, between the presynaptic nerve terminal and muscle fibres of the frog (Katz & Miledi, 1981, 1982).

Based on these observations and on the ultrastructural studies of various neuronal tissues (for example see Boyne & Tarrant, 1979, 1982; DiFiglia, Pasik & Pasik, 1979) it seems possible that K⁺-mediated interactions are prevalent. As pointed out by Katz & Miledi (1982) ‘One may wonder whether some of the brief synaptic potentials that have been attributed to electric coupling between neurons were not due to potassium release’. It is possible that K⁺-mediated interactions were not identified as such because phenomenologically the K⁺ PSPs are similar to electrotonic and chemical PSPs (Spira & Yarom, 1983b).

Some of the basic characteristics of a hypothetical synapse in which potassium ions serve as the mediator may be very similar to those of chemical and electrotonic junctions. Therefore the conventional manipulations that are used to differentiate between chemical and electrotonic junctions may be insufficient to identify the existence of a potassium-mediated interaction. We will describe briefly some of the features that may be common to all these modes of transmission.

Unidirectional conduction is a basic property of a chemical synapse and may be possible in electrotonic junctions (rectifying electrotonic synapses, for review see Bennett, 1972, 1974, 1977). Basically, a potassium-mediated synapse should be bidirectional (for example the GGSP). Unidirectional transmission, however, could be produced if the density of K⁺ channels is larger on one side of the junction than on the other, or if more K⁺ is liberated from one neurone than from the other. That the release of K⁺ during an action potential may be strongly dependent on local membrane properties was recently demonstrated by Hounsgaard & Nicholson (1983). Another condition which may produce unidirectional propagation in a K⁺-mediated synapse is when the two interacting neurones are asymmetrical in terms of their input resistances (Bennett, 1972, 1974, 1977). Thus directionality of transmission is not a diagnostic feature of any type of synapse.

A very short synaptic delay usually indicates the presence of an electrotonic junction. However, it would be difficult to differentiate between an electrotonic synapse and a K⁺-mediated one on this basis (Katz & Miledi, 1982 and see the latency of the GGSP).

Low Ca²⁺, high Mg²⁺ solutions block chemical synapses but have little effect on electrotonic ones (Bennett, 1977). If calcium ions are not involved in the generation of the action potential of a K⁺-mediated interaction, then lowering the Ca²⁺ concentration will not affect the transmission. On the other hand if Ca²⁺ ions are involved in the pike generation, reduction in [Ca²⁺]_o may lower the efflux of K⁺ from the neurone and thereby reduce the amplitude of a K⁺-mediated PSP (Meech & Strumwasser, 1970; Meech, 1978; also see Hounsgaard & Nicholson, 1983). Thus again this conventional tool for identification of the mode of synaptic transmission may not be very useful in recognizing a K⁺-mediated synapse.

The inversion of a chemical PSP by shifting the postsynaptic membrane potential is a most powerful diagnostic tool. This is because at most chemical synapses the generation of synaptic conductance is almost entirely independent of transmembrane voltage. However, in non-isopotential neurones where the synapses are located far from the point of current injection, it may be impossible to invert the PSP. In most electrotonic junctions the changes in PSP amplitude are restricted by the extent to which the presynaptic membrane potential is influenced by postsynaptic current injection (Bennett, 1972, 1974, 1977). K⁺-mediated potentials are only slightly dependent on transmembrane voltage (Rudomin et al. 1979 and see for example the GGSP) and therefore behave in a similar manner to either an electrotonic synapse or a remote chemical synapse.

To summarize, this partial list demonstrates some of the expected experimental difficulties in identification of localized potassium interactions. We expect, however, that additional studies will provide sufficient information and experimental criteria that will enable this mode of transmission to be identified.

We would like to thank R. Werman for helpful comments on this manuscript. This work was supported by a grant from The United States-Israel Binational Science Foundation No. 015/5639 to the laboratory of MES.