Synaptic Plasticity and the Modulation of the Ca²⁺ Current

The study of the mechanisms of neuronal plasticity and the attempt to relate these mechanisms to actual instances of learning has accelerated in recent years as a result of the application of the techniques of biophysics and cell biology to central neurones and their interconnexions. Although support for it has been obtained only recently, the idea that learning might involve changes in the effectiveness of the connexions between neurones actually had its origins at the turn of the century. Following an earlier suggestion by E. Tanzi (1893), Ramon y Cajal postulated, in his Croonian Lecture to the Royal Society of London in 1894 (a lecture to which he was invited through the intervention of Charles Sherrington), that learning might involve changes in strength of connexions between neurones. An almost identical idea was put forth by Sigmund Freud also in 1894 in a fragmentary manuscript published only in 1950. A first requirement of this postulate is that some synapses have plastic properties, that they can change their efficacy following simple use or following more complex patterns of stimulation. This basic requirement has now been fully satisfied. A variety of experiments have shown that chemical synapses can undergo changes in effectiveness as a result of activity or inactivity in a given pathway (homosynaptic change) or as a result of activity in other pathways (heterosynaptic change). Some of the best evidence has come from studies of simple synaptic systems such as the synapses between vertebrate motor neurones and skeletal muscle (for reviews, see Katz, 1962; Eccles, 1964; Kandel & Spencer, 1968).

Soon after the initial discovery of the end-plate potential by Gopfert & Schaefer (1938), Schaefer, Scholmerich & Haass (1938) and subsequently Feng (1941) found that repeated (tetanic) stimulation increased the amplitude of the end-plate potential without producing any obvious changes in the action potential of the presynaptic axon. The increase in the end-plate potential was not restricted to the period of tetanic stimulation but persisted for several minutes after the tetanus. Feng called the synaptic enhancement that persisted after the tetanus post-tetanic potentiation. He also found that longer tetani produced greater potentiation than did shorter ones. In 1947, Larrabee and Bronk found post-tetanic potentiation at a peripheral neurone-to-neurone synapse between preganglionic and postganglionic cells of the stellate ganglion. In 1949, Lloyd described similar potentiation in the monosynaptic reflex of the spinal cord, thereby showing that these plastic changes also occur in central neurones. Lloyd found (as had Larrabee and Bronk) that post-tetanic potentiation produced by stimulating one afferent pathway did not increase the response of the postsynaptic cell to synaptic activation via another, unstimulated, afferent pathway. These experiments indicated that post-tetanic potentiation is homosynaptic: it is restricted to the stimulated pathway and results from a change in the synapse itself. Lloyd (1949) also described a second type of plastic change when he found that low frequencies of stimulation produced a decrease in synaptic effectiveness. This form of plastic change he called low-frequency or homosynaptic depresssion.

Analysis remained at this stage for a number of years because it was difficult to determine whether these changes were due to a presynaptic mechanism (a change in transmitter release) or to a postsynaptic mechanism (a change in receptor sensitivity). The solution to this problem was facilitated in 1954 when del Castillo & Katz (1954a) demonstrated that release of acetylcholine at the nerve muscle synapse is not graded but quantized. An action potential releases about 200 multimolecular packets of transmitter-called quanta -and each quantum contains several thousand molecules of ACh. Quantal transmission was soon found to be the general mode of transmitter release at chemical synapses (see Dudel & Kuffler, 1961a; Eccles, 1964). The discovery of quantal transmission not only established important new insights into the nature of transmitter release from the terminals but also provided a method for analysing the relative contribution to synaptic transmission of changes in presynaptic and postsynaptic mechanisms. Del Castillo & Katz (1954a, b), Liley (1956a, b) and subsequently others, analysed alterations in synaptic effectiveness in terms of quantal transmission and found that both homosynaptic depression and post-tetanic potentiation represented a presynaptic alteration in the number of transmitter quanta released per impulse. The sensitivity of the receptor seemed not to be affected.

This work was extended in a major direction in 1961 when, following the earlier suggestion of Frank & Fuortes (1957), Dudel and Kuffler described the first clear instance of a heterosynaptic interaction : presynaptic inhibition at the crayfish nervemuscle synapse. Dudel & Kuffler (1961a) found that in the crayfish the inhibitory axon to muscle has a double function: (1) it produces an inhibitory postsynaptic potential in the muscle, and (2) it depresses the excitatory postsynaptic potential produced by the excitatory axon. By applying a quantal analysis Dudel & Kuffler (19616) found that presynaptic inhibition reduces the number of transmitter quanta released by the excitatory axon without affecting the sensitivity of the receptor molecules. These experiments provided the first evidence that the membrane of the presynaptic terminals contains receptors to transmitter molecules, and that these receptors can control transmitter release. Subsequently, Kandel & Tauc (1964), Epstein & Tauc (1970), and Castellucci and his colleagues (1970) presented suggestive evidence for presynaptic facilitation of transmitter release in the synapses of Aplysia. Recently again based upon a quantal analysis, Castellucci & Kandel (1976) provided direct evidence for a presynaptic mechanism.

In 1967 Katz and Miledi advanced the study of synaptic plasticity still one important step further by finding that transmitter release is dependent on the influx of Ca²+ that occurs with each action potential. They proposed that Ca²⁺ allows the synaptic vesicles, the subcellular organelles that store transmitter (and which are thought to represent individual quanta), to bind to release sites. In addition, Katz & Miledi (1967 a, b) found that the presynaptic terminal contained a high density of voltagegated Ca²⁺ channels.

These findings suggested to Katz & Miledi (1968) and to Rahamimoff (1968) that changes in the intracellular level of free Ca²⁺ might be important for short-term synaptic plasticity. Rahamimoff (1968; Alnaes & Rahamimoff, 1975) proposed that the intracellular concentration of Ca²⁺ is controlled by intracellular organelles that buffer Ca²⁺-the mitochondria and endoplasmic reticulum -and that aspects of synaptic plasticity might depend upon this control. In support of this idea, Rahamimoff found that short-term homosynaptic facilitation seems to be due to residual Ca²⁺, the Ca²⁺ that remains in the terminal following a series of action potentials, and that is taken up slowly by the buffering organelles. A similar mechanism also seems to be operative in Aplysia (Kretz, Shapiro & Kandel, 1980).

Another factor that controls the free Ca²+ levels in the terminals is the Ca²⁺ influx. It seemed to us (as it had earlier to Zucker (1974) and to Stinnakre & Tauc (1973)) that influx might not be constant but might be modulated. In the past, however, this idea has proven difficult to examine in central neurones showing plastic changes, since the most direct test of the hypothesis would require recording of Ca²⁺ current in presynaptic terminals simultaneous with monitoring of transmitter release. This problem can be overcome, to a certain degree, in Aplysia neurones. First, the membrane of the cell body of Aplysia neurones contains Ca²⁺ channels whose properties seem to resemble those of the terminal membrane (Geduldig & Junge, 1968 ; Geduldig & Gruener, 1970; Stinnakre & Tauc, 1973; Llinas, Steinberg & Walton, 1976). Moreover, in certain cases changes in the calcium current of the cell body parallel the changes in transmitter release at the terminals (Klein & Kandel, 1978). In addition, in some of these neurones the presynaptic terminals controlling transmitter release appear to be sufficiently close to the cell body electrically so that it is possible to modify transmitter release from the terminals by injecting current into the cell body (Shimahara & Tauc, 1975; Shimahara & Peretz, 1978; Shapiro, Castellucci & Kandel, 1980a). The observations suggested to us that we might be able to examine the relationships between transmitter release and specific ionic currents of the presynaptic membrane. To this end we have combined analysis of ionic currents of the cell body of the presynaptic neurone, using voltage-clamp analysis and pharmacological blockade of specific ion channels, with assay of transmitter release from the presynaptic cell, using intracellular recordings of the synaptic potential in the postsynaptic cell (Fig. 1). These combined techniques provide a powerful method for studying changes in specific ionic conductances associated with various presynaptic mechanisms for synaptic plasticity.

We have applied these techniques to two identified synaptic connexions in the bdominal ganglion of Aplysia: (1) the connexions made between the multi-action cholinergic cell L10 and its identified follower cells, and (2) the connexions between the mechanoreceptor sensory neurones of the gill-withdrawal reflex and their follower cells.

The situation was particularly favourable for L10 and its synapses. When we blocked the Na+ channels with tetrodotoxin and axotomized L10 we found that graded depolarizing commands applied to the membrane of the presynaptic cell body under the voltage clamp caused graded transmitter release from the terminals, as determined by intracellular recording from postsynaptic cells (Figs. 2 A, 2 B, and 3 A). We obtained further evidence that release sites were being controlled by the clamp currents in the cell body by finding that transmitter release is also a graded function of the duration of the depolarizing step (Fig. 2B ₂). Voltage control can be further improved by adding pharmacological agents that block each of the three known outward K⁺ currents: tetraethylammonium ions (TEA+), which block the delayed K⁺ channel; 4-aminopyridine (4-AP), which blocks the early K⁺ channel; and substitution of Ba^24- ions for Ca^24-, which block the Ca²⁺-dependent K⁺ channel (see Adams, Smith & Thompson, 1980; Shapiro, Castellucci & Kandel, 1980a, b). Blocking these K⁺ channels also lengthens the effective space constant of the neurone. M

Using these K⁺ channel-blocking agents we could now examine the relationship between transmitter release and presynaptic Ca²⁺ current (Fig. 3 A). We found that in the range of depolarizing voltages in which the Ca²⁺ current was increasing, transmitter release increased in a linear fashion (Fig. 3B). These results suggest that in this voltage range (from about —40 to zero mV) we could control transmitter release from the terminals with voltage clamp of L10’s cell body.

Although in many cases (particularly in the studies of the connexions made by sensory neurones), we lacked ideal voltage control of the terminals, these procedures nonetheless allowed sufficient control to study transmitter release while at the same time examining ionic currents in the soma of the presynaptic neurone. Moreover, we repeated all experiments except those involving presynaptic inhibition (where the transmitter is not known) in mechanically isolated presynaptic cell bodies where optimal space clamp control can be achieved.

Using this approach, we have examined four types of synaptic plasticity, including two directly involved in simple nonassociative forms of learning: (1) the control of transmitter release in spike generating neurones by the membrane potential of the presynaptic neurones; (2) presynaptic inhibition; (3) homosynaptic depression (the mechanism underlying short-term habituation); and (4) presynaptic facilitation (the mechanism of behavioural sensitization). We have found that each of these involves modulation of the Ca²⁺ current, although the details of the mechanisms differ in each case.

In 1975 Shimahara and Tauc described a simple form of synaptic plasticity whereby the membrane potential of the presynaptic neurone exerts a powerful influence over the effectiveness of the connexions made by that neurone. Hyperpolarizing the presynaptic cell decreased the synaptic potential elicited by the presynaptic action potential whereas depolarizing the presynaptic cell enhanced the synaptic action. Although this effect was opposite to that described in the giant synapse of the squid (Hagiwara & Tasaki, 1958; Takeuchi & Takeuchi, 1962) and at first seemed paradoxical, similar potential dependent control of transmitter release by membrane potential has now been described in the leech (Nicholls & Wallace, 1978 a). A form of this plasticity occurs between the identified multiaction cholinergic cell L10 and its follower cells (Waziri, 1977; Shapiro, Castellucci & Kandel, 1980a).

Intracellular stimulation of neurone L10 produces EPSPs in some follower cells (R15 and RB cells) and IPSPs in others (Li to L6, LB_VC, LD_HI, etc.; see Kandel et al. 1967; Wachtel & Kandel, 1967, Kehoe, 1972; Koester et al. 1974; Koester & Kandel, 1977). When the membrane potential of L10 is progressively increased, its ability to release transmitter at its various branches is reduced (Fig. 4). Even a slight increase in membrane potential of 5 mV produces significant depression of transmitter release. With larger hyperpolarizing changes of 20 – 40 mV, transmitter release can be completely blocked (Fig. 4 A). Conversely, depolarizing the cell’s membrane potential increases transmitter release. The changes in membrane potential affect the duration and height of the presynaptic action potential: spikes become narrower and shorter with hyperpolarization and broader and taller with depolarization (Fig. 4B). both of these potential induced alterations are likely to alter transmitter release because hey alter the Ca²⁺ current (Horn & Miller, 1977; Klein & Kandel, 1978). However.

However, as we shall discuss below, these alterations account for only part of the effects of Membrane voltage on release. Part of the effect of membrane voltage is independent of changes in spike height and duration.

The ability to alter transmitter release from the terminals by injecting current into the cell body implies that at least some release sites are electrically close to the soma. If the axon of the cell is cut close to the ganglion and the preparation is treated with TTX, graded depolarizing steps under voltage clamp lead to graded release of transmitter (Figs. 2, 3 and 5). The sigmoid function relating transmitter release to presynaptic depolarization is similar to that reported at the squid giant synapse (Katz & Miledi, 1969; Llinas et al. 1976). As is the case with the squid giant synapse the function relating transmitter release to presynaptic depolarization overlaps the voltage sensitivity of the presynaptic Ca²⁺ current (Fig. 3). However, in contrast with results in squid, hyperpolarizing the presynaptic cell decreases the size of the synaptic potential elicited by step depolarization to a given level (Fig. 5 B).

At depolarized holding potentials, currents elicited by step depolarizations under voltage clamp are less outward than at hyperpolarized holding potentials (Fig. 5 A and Connor & Stevens, 1971b; Adams et al. 1980). This difference in net current presumably accounts for the difference in the configuration of the action potential at the two holding potentials in unclamped cells (Fig. 4B). In addition, the finding that the synaptic potentials are graded with the duration of the command pulse (Fig. 2B ₂) provides further evidence that the K⁺ currents, which control spike duration and amplitude, could modulate transmitter release. However, the ability of the membrane potential to modulate transmitter release under voltage-clamp conditions, in which the duration and amplitude of the presynaptic command pulse is held constant, suggested that mechanisms other than modulation of the K⁺ current contribute to this form of plasticity. One possible additional mechanism is a change in transient Ca²⁺ current. To test this idea we blocked the several outward K⁺ currents which are responsible for alterations in the shape of the spike to see whether the membrane potential was still capable of modulating the transmitter release caused by a depolarizing command, and whether this modulation was due to a direct action on the transient Ca²⁺ current.

Three pharmacologically separable K⁺ conductances have been described in mol-luscan somata (Thompson, 1977; Adams et al. 1980). Two of these are voltagedependent; an early, rapidly inactivating K⁺ current sensitive to 4-AP (Connor & Stevens, 1971b; Byrne et al. 1979; Thompson, 1977), and a delayed K⁺ current sensitive to TEA+ (Connor & Stevens, 1971 a; Thompson, 1977; Byrne et al. 1979). A third K⁺ conductance is thought to be not dependent on voltage but on intracellular Ca²⁺ concentration, and can be blocked by agents which block the Ca²+ current (e.g. Ca²⁺, EGTA) and by substituting Ba^24- ions for Ca Ca²⁺ ions (Meech, 1972; Meech & standen, 1975; Eckert & Lux, 1976; Thompson, 1977; Adams et al. 1980; Shapiro et al. 1980a). Blocking the voltage-dependent K⁺ channels caused peak transies inward currents to appear relatively unchanged from different holding potentials (Fiji 6 A). With all K⁺ channels blocked pharmacologically (so that there is no voltagesensitive outward current with depolarizing command) and in a solution that was free of both Na+ and Ca²⁺ (265 mm-TEA+, 60 mm-Ba^24-, 10 mm-4-AP) the inward current through the Ca²⁺ channels, now carried by Ba²⁺ ions, can be observed directly. When this inward current was activated with a step depolarization it was not increased by changing the holding potential from —62 to —40 mV (Fig. 6C). At the same time, the synaptic potential elicited by depolarization from the two levels was still affected by membrane potential (Fig. 6B). When high concentrations of TEA+ and 4-AP were utilized together, the peak inward Ca²⁺ current even decreased slightly when evoked from more depolarized holding potentials (Fig. 6 A), perhaps as a result of steady-state inactivation, or because of an increase in the intracellular concentration of Ca²⁺ at depolarized holding potentials.

These results indicate that differences in the transient currents observed with steps from different holding potentials (Fig. 4A) are due to changes in K⁺ conductances and not to changes in Ca²⁺ current. The holding potential does not control transmitter release by regulating directly the transient activation of the Ca²⁺ channel (Fig. 6).

Although membrane potential regulation of transmitter release does not result from direct modulation of the transient Ca²⁺ current, changes in membrane potential could lead to changes in the steady-state activation of the Ca²⁺ channels (Fig. 6C). Steadystate activation-the contribution of Ca²+ current to the resting current-might be greater at depolarized levels. Such steady-state activation of Ca²⁺ channels has been described in molluscan somata by Eckert & Lux (1976) and by Akaike, Lee & Brown (1978).

A steady-state activation of Ca²⁺ could lead to an increased concentration of intracellular Ca²⁺ at more depolarized holding potentials, which would add to the Ca²⁺ influx during the action potential, resulting in a higher total intracellular concentration of Ca²⁺ available for transmitter release. In addition, increased intracellular Ca^24- could cause relative saturation of intracellular Ca²⁺ buffering systems that are thought to compete with the release process (Alnaes & Rahamimoff, 1975; Rahamimoff, 1968). A third possibility is that increased intracellular Ca²⁺ concentration causes changes in screening of surface membrane proteins by changing internal membrane surface charges, and transmitter release efficacy (Bass & Moore, 1966; Van der Kloot & Kita, 1973). In each of these possibilities invasion of the terminal by an action potential would provide a more release-effective Ca²⁺ concentration.

To obtain evidence for steady-state Ca²⁺-channel activation, we examined the steady-state membrane conductance of L10 at different holding potentials in the range of membrane voltages where modulation of release was occurring. We found that the steady-state conductance showed a region of reduced positive slope that is due to the steady-state Ca²⁺ channel. The steady-state Ca²⁺ current in turn activates a Ca²⁺-dependent K⁺ conductance that is blocked by substituting Ba²⁺ ions for extract Ba²⁺ flows through Ca²⁺ channels without activating the K⁺ conductance (see Figs. 6B C and 7B and Adams et al. 1980). In the presence of Ba²⁺, a voltagedependent steady-state inward current flows through the Ca²⁺ channels at depolarized levels of holding potential (Fig. 7A,₂, 7B); this inward current is not present at hyperpolarized levels (Fig. 7B). To activate this K⁺ conductance the concentration of intracellular Ca²⁺ in the steady state must be significantly larger at more depolarized than at more hyperpolarized membrane potentials.

The role of steady-state Ca²⁺ current in the modulation of transmitter release by the membrane potential is supported by the observations of Nicholls & Wallace (19786) that spontaneous quantal release of transmitter in leech neurones is increased at depolarized potentials. Our results indicate that an additional mechanism also operates in cell L10 : a decrease of both voltage-dependent K⁺ currents with depolarization. We have not as yet attempted to evaluate quantitatively the relative contributions of the changes in the steady-state Ca²⁺ and those in the transient K⁺ current In an unclamped cell, the decrease in the K⁺ currents accounts for the increase in the weight and duration of the spike. In this way the transient Ca²⁺ current can also be Tnodulated, albeit indirectly. Our findings also suggest that despite a transient Ca²⁺ current that is not directly affected by changes in holding potential, powerful control over synaptic transmission-ranging from total block to enhanced effectiveness-can be achieved by variations in the steady-state Ca²⁺ current (Fig. 8).

These results support the idea first proposed by Shimahara & Tauc (1975) that EPSPs and IPSPs have a dual function. In addition to controlling the probability of firing an action potential, synaptic potentials also set the level for transmitter release. The results further suggest a possible mechanism for long-term regulation of synaptic output by metabolic or hormonal actions which affect resting potential (Thomas, 1972; Mayeri, Brownell & Branton, 1979; Mayeri et al. 1979).

Presynaptic inhibition has been described in several vertebrate and invertebrate synapses and has been attributed to a depolarization of the synaptic terminals (reviewed in Burke & Rudomin, 1977; Ryall, 1978). This idea seemed consistent with the release properties described at the squid giant synapse where depolarization reduces transmitter release (Miledi & Slater, 1966). However, as described above, in Aplysia and in the leech, depolarization enhances rather than depresses transmitter release (Shimahara & Tauc, 1975; Nicholls & Wallace, 1978 a; Shapiro et al. 1980a). At these synapses, a different mechanism must account for presynaptic inhibition.

In Aplysia the synapses of cell L10 that are modulated by membrane potential can also be modulated by presynaptic inhibition. Stimulation of one of the fibre pathways to the abdominal ganglion causes a depression of both the monosynaptic excitatory and the inhibitory connexions made by L10 (Fig. 9; Waziri, Kandel & Frazier, 1969; Shapiro et al. 1980b). Recently, Byrne (1980) has discovered a group of cells (the L32 cells) which, when stimulated, produce presynaptic inhibition without producing any conductance changes in the postsynaptic follower cells of L10.

Each of the ways of producing presynaptic inhibition tends to produce an IPSP in the presynaptic neurone, L10. Thus one change in the presynaptic neurone that clearly contributes to presynaptic inhibition is the hyperpolarization that occurs as a result of the IPSP. This hyperpolarization would reduce the output of transmitter by reducing the steady-state Ca²⁺ influx and by removing some of the inactivation of the K⁺ channels. However, presynaptic inhibition outlasts the hyperpolarization and can also be observed in cases in which the cell does not hyperpolarize. To determine whether additional mechanisms are operative we controlled for changes in membrane potential by examining the presynaptic neurone under voltage-clamp conditions.

In a solution containing TTX and TEA+, and with the cell held at a membrane potential which inactivates the early outward K⁺ current, the inward current produced by a depolarizing clamp step is carried by Ca²⁺ ions, and the residual outward current is mainly due to Ca²⁺-dependent K⁺ current and the leakage current. Under these circumstances presynaptic inhibition is correlated with a decrease in transient inward current evoked by the depolarizing command (Fig. 10).

The decrease in inward current in voltage-clamp records based on net currents can be the result of an increase in small residual outward K⁺ current, or a decrease in the inward Ca²⁺ current. The observation that the decrease in early inward current during presynaptic inhibition is larger than the increase in late outward and leakage currents (Fig. 10, Fig. 11 A) suggests that the Ca²⁺ current may be modulated directly. But to distinguish between these alternative possibilities directly we performed two types of experiments: pharmacological block of all the K⁺ channels, and an examination of the voltage dependence of the presynaptic action.

To test the involvement of Ca²+ channels in presynaptic inhibition we examined the divalent cation current in isolation after blocking the Na+ channel with TTX and the three known K⁺ channels with TEA and 4-AP, and with substitution of Ba²⁺ for Ca²⁺. under these circumstances we found that connective stimulation reduced the inward rent (Fig. 11B, 11C). This reduction paralleled the presynaptic inhibition.

In addition, changes in the steady-state inward current also occurred with connective stimulation (Fig. 11 C). These results imply that the decrease in inward current reflects a decrease in the transient and in the steady-state Ca²⁺ current during presynaptic inhibition.

Since the Ca²⁺ current is voltage dependent, one would expect that the synaptic transmitter that mediates presynaptic inhibition would have no effect on ionic currents at hyperpolarized levels of membrane potential where the Ca²+ current is not activated. We therefore examined the effects of presynaptic inhibition on currents in the voltage range where the Ca²+ channel was activated and in the range where it was not activated. When cell L10 was held in the voltage range in which the Ca²⁺ channel was activated, connective stimulation caused decreases in inward current (Fig. 12A). When the cell was held in the voltage range in which the Ca²⁺ channel was not activated, no conductance change was observed after connective stimulation (Fig. 12) These data support the idea that presynaptic inhibition is due to a direct action on the Ca²⁺ channel (Fig. 13). A similar mechanism for presynaptic inhibition has been discovered in dissociated dorsal root ganglion cells by Dunlap & Fischbach (1978) and by Mudge, Leeman & Fischbach (1979).

In vertebrate heart muscle Giles & Noble (1976) have found that ACh, acting on muscarinic receptors in the heart, decreases the Ca²+ current in the heart muscle. This action of ACh is similar to the mechanism of presynaptic inhibition in Aplysia and in dissociated dorsal root ganglion cells. Reuter (1979; Reuter & Scholz, 1977) has suggested that cGMP, or the ratio of cAMP/cGMP in heart muscle, may control the maximum calcium conductance of the muscle. An alternative possibility (Fig. 27) is that the modulatory transmitter closes a Ca²⁺ channel-receptor complex directly without the mediation of an intracellular second messenger.

In mammals, presynaptic inhibition has been correlated with presynaptic depolarization and with increased excitability in afferent fibres. The observed conductance change at some of these synapses follows the chloride Nernst potential (reviewed in Burke & Rudomin, 1977). However, as indicated above, results similar to those reported here have been obtained by Fischbach and his colleagues (Dunlap & Fischbach, 1978; Mudge et al. 1979) on chick dorsal root ganglia. Whether the presynaptic depolarization and the increased Cl⁻ conductance in vertebrates are epiphenomena, that are not directly related to the mechanism of presynaptic inhibition, and are due only to the artificial modes of activation, or whether there are several distinct mechanisms for presynaptic inhibition, needs now to be determined (for a critical review At the studies of presynaptic inhibition in the mammalian brain, see Ryall, 1978).

Homosynaptic depression -a self-induced depression in excitatory transmission at the synapses made by the sensory neurones on their central target cells-is the mechanism for short-term habituation of the siphon and gill-withdrawal reflexes in Aplysia (Fig. 14; Castellucci et al. 1970). This depression, which is very profound and rapid at these sensory neurone synapses, is due to a progressive decrease in the amount of transmitter released by each action potential (Castellucci & Kandel, 1974).

As a first step in examining homosynaptic depression, we analysed the Ca²⁺ contribution to the action potential in the sensory neurone. Under normal circumstances, the action potential is rapidly terminated by the repolarizing action of large outward potassium currents, masking the small contribution of the calcium current to the action potential. We therefore reduced the voltage-dependent potassium current substantially by bathing the ganglion in TEA⁺. This caused the peak amplitude of the action potential to increase and a plateau to develop on the descending limb, thereby increasing the width of the action potential at half-maximal amplitude by about 50fold. During this plateau the inward current, which is carried by Ca²⁺, is unmasked sufficiently to produce a self-sustaining depolarization of the membrane. That the plateau is due to calcium current is supported by the finding that it occurs in the absence of sodium, and that the amplitude of the plateau follows the external calcium in concentration as predicted by the Nernst equation. Furthermore, the plateau can be abolished with cobalt or nickel ions (Klein & Kandel, 1978), which block the calcium channels of other neurones (Adams et al. 1980).

The duration of the action potential in TEA is a sensitive measure of the calcium current because it is determined by the balance between the inward calcium current and the outward potassium and leakage currents. As long as the calcium current is at least equal to the outward currents (the leakage and the K⁺ currents) the plateau is maintained; when the outward currents outweigh the inward current the membrane repolarizes. The total calcium influx during the plateau phase can therefore be regulated both by intrinsic changes in the calcium channels as well as by changes in potassium or leakage currents. Hence, changes in the duration of the TEA spike are a good indication of changes in total Ca²⁺ influx, but they do not indicate whether the changes result from a direct action on the calcium channel or whether the influx is affected indirectly by a potassium conductance not blocked by TEA. In addition, some part of the plateau also represents firing of other parts of the neurone outside the cell body. Calcium spikes in the axon and terminal regions can contribute to the duration of the TEA action potential recorded in the cell body, except, of course, when the cell body is isolated from its processes by tying it off, for example.

We next examined transmitter release from the terminals of the sensory neurones, as measured by the size of the synaptic potential in postsynaptic neurones, and the simultaneous change in calcium current, as measured by changes in the duration of the action potential (Fig. 15). Repeated stimulation of the sensory neurone at rates that produce habituation led to a progressive decrease in the PSP (with kinetics similar to what is observed in normal sea water) together with a reduction in the duration of the TEA action potential, indicating a progressive decrease in the calcium current of the action potential. Spontaneous recovery of synaptic transmission with rest was accompanied by an increase in the calcium current.

To determine whether the decrease in Ca²⁺ current was due to a direct action on the Ca²⁺ current or an action on a K⁺ current, we voltage clamped the cell body of the sensory neurone. In normal sea water, depolarizing commands elicit EPSPs in the follower cells and an inward current in the sensory neurone which is due largely to Na⁺ and a lesser degree to Ca²⁺, followed by an outward current (Fig. 16 A). With repeated depolarizing commands repeated at every 10 sec the EPSPs decrease in amplitude but there is no change in the presynaptic current. This is consistent with the lack of a change in the action potential in normal sea water in the absence of a⁺ or K⁺ channel-blocking agents.

We therefore next blocked the Na⁺ channels with TTX and the delayed and early K⁺ channel with a very high concentration of TEA (Fig. 16B). With the larger Na+ current blocked, this inward current is due to Ca²⁺. Depolarizing commands now still produced EPSPs in the motor cell and these declined with repetition. Fig. 17 shows averaged data on the depression of the EPSP and of the inward current from three sensory neurones where 15 EPSPs were elicited at 10 sec intervals in two consecutive training sessions separated by 5 min of rest. The changes in the EPSP parallel quite closely the changes in the peak inward current. However, in these experiments some portion of the current results from uncontrolled active responses in other parts of the neurone, including the fine terminal processes of the neurone where transmitter release occurs. As a result, the interpretation of the decline in inward current is not unambiguous. It is possible, for example, that the threshold for firing calcium spikes in fine processes increases during habituation, resulting in transmitter release from fewer terminals. Alternatively (or in addition), the decline in calcium current could occur in a graded manner throughout all parts of the neurone, although perhaps not equally everywhere, so that progressively less transmitter is released from each ter-mnal during habituation. The experiments do show, however, that drastic reduction of the K⁺ currents by TEA does not affect the ability of repeated activity to depress the EPSP. The parallel decline of the PSP and of the inward current therefore appears to result from an intrinsic change in the calcium current.

To examine the inward current in a situation where the cell body is optimally spaceclamped so that uncontrolled responses are eliminated we isolated the cell body of the sensory neurone by tying it off. We next blocked all the K⁺ currents by substituting Ba²⁺ for Ca²⁺ and using 4-AP in addition to high concentrations of TEA. Under these circumstances we still found a progressive decrease in the inward currents with repeated depolarizing commands (Fig. 16C).

We do not know whether this graded decline in inward current is quantitatively the same as that which occurs in the intact neurone, and whether the depression of the inward current with repeated stimulation is the only factor that accounts for habituation. However, habituation clearly does not involve an increase in K⁺ current. In fact, in normal sea water with long depolarizing commands the K⁺ current actual decreases with repeated depolarization, whereas the depression of the PSP is unaffected (see, for example, Fig. 22 B, left half of figure).

The term ‘inactivation’ has been applied to a decrease in an ionic current elicited at a constant membrane voltage that is not due to an increase in an opposing current or to a change in driving force. We have provided evidence that the decrease in inward current is not the result of the activation of a residual K⁺ current (Fig. 16 C). It is also unlikely that the reduction in inward current is due to a change in driving force (resulting from an accumulation of divalent cations in the cell or near the inner surface of the membrane) since we have found that while the transient inward current decreases with repeated depolarizing command pulses of 60 to 70 mV (designed to simulate the action potential), the steady-state current through the Ca²+ channel (elicited with small voltage-clamp steps) does not decrease. A change in driving force should affect both transient and steady-state components of the current. Thus, the homosynaptic depression that underlies habituation is correlated with a prolonged inactivation in the transient inward current resulting from its repeated activation (Fig. 18).

Tillotson & Horn (1978) have described the tendency of the transient Ca²⁺ current to undergo prolonged inactivation in cells R2 and R15 of Aplysia. This inactivation differs from Na⁺ inactivation in that the inactivation of the Ca²⁺ channel is not voltage dependent, but may depend on the intracellular calcium concentration (see Brehm & Eckert, 1978; Tillotson, 1979). The inactivation of the Ca²⁺ current underlying habituation in the sensory neurones, however, seems to differ in certain respects from that studied by Tillotson in cells R2 and R15 (Tillotson, 1979; Tillotson & Horn, 1978). First, in R2 and R15 the Ca²⁺ channels do not inactivate when Ba²⁺ carries the inward current, whereas in the sensory neurones the channels do inactivate with Ba²⁺ as the charge carrier. In addition, the time course of recovery from inactivation is miuch slower in the sensory neurones than in cells R2 and R15. It would therefore be of interest to explore the specific mechanisms that underlie the prolonged Ca²⁺ current inactivation in the sensory neurones.

As is often the case with behavioural habituation, reflexes that illustrate this form of learning also show an opposite learning process: sensitization. Presynaptic facilitation was first encountered while trying to develop acellular analogue for reflex sensitization, a simple form of learning which involves the enhancement of a reflex response by a strong stimulus (Kandel & Tauc, 1965 a, b). This process was subsequently shown to be actually utilized in behavioural sensitization of the gill-withdrawal reflex (Kupfermann et al. 1970).

As we have seen, repeated sensory stimulation at rates that produce habituation in the intact animal produces a depression in the monosynaptic excitatory connexions between sensory neurones and their followers due to a decrease in the number of transmitter quanta released per impulse (Castellucci & Kandel, 1974). On the other hand, stimulation of a different pathway for a few seconds causes an increase in transmitter release from the sensory neurones (Castellucci & Kandel, 1976and Fig. 19). This presynaptic facilitation is simulated by cyclic AMP and by serotonin, but not by a number of other candidate transmitter substances that were examined. A sensitizing stimulus produces two actions on the sensory neurones: (1) a slow EPSP thatlastsupto half an hour (Klein & Kandel, 1978), and (2) a similarly long facilitation of synaptic transmission at the sensory-to-motor synapses (Fig. 19; Castellucci et al. 1970). To determine what was responsible for the increase in transmitter release, we examined the slow EPSP and other changes in the membrane properties of the sensory neurone cell body that accompany presynaptic facilitation.

We first examined the changes in the membrane conductance produced in the presynaptic neurone by the slow EPSP. Using electrotonic potentials produced by constant current pulses in the sensory neurones, we found that the slow EPSP was accompanied by an apparent decrease in the conductance of the membrane and that this decrease was simulated by serotonin, the putative facilitating transmitter, by cyclic AMP, and by phosphodiesterase inhibitors (Klein & Kandel, 1978). This was confirmed with voltage-clamp experiments. Stimulating the facilitatory pathway or incubating the ganglion or the isolated cell body with serotonin under voltage clamp caused an inward shift in the holding current and a decrease in the ‘leakage’ currents elicited by small step hyperpolarizations. Moverover, the slow EPSP and the inward shift in the holding current produced by connective stimultion or by serotonin were blocked by K⁺ channel-blocking agents (compare Fig. 22 C₁ and C₂), and by replacing intracellular K⁺ in the sensory neurones with Cs⁺, an impermeant cation. These results imply that this EPSP is due to a decrease in K⁺ conductance. The EPSP is voltage dependent. Current-voltage curves are highly non-linear. At levels of membrane potentials more depolarized than — 50 mV the non-linearity is reduced by connective stimulation and serotonin but at more hyperpolarized potentials than — 60 mV synaptic stimulation or serotonin gave rise to little or no synaptic current. By contrast, when the external K⁺ concentration was doubled, the current reversed at about — 50 mV. The association of a decrease in a voltage-sensitive K⁺ conductance with presynaptic facilitation suggested that the facilitation might occur through the reduction of a presynaptic K⁺ current that is activated by the action potential, thereby leading to an increase in the duration of the action potential and increased transmitter release. We therefore examined the presynaptic action potential and the currents activated by them under a variety of conditions to see whether or not a reduction in the K⁺ current of the presynaptic action potential could account for presynaptic facilitation.

When we bathed the abdominal ganglion in TEA and examined action potentials in the sensory neurones, we found that they were prolonged by stimulating the connective, indicating an increase in the Ca²⁺ current. As with homosynaptic depression, the changes in the duration of the action potential as a result of presynaptic facilitation paralleled the changes in transmitter release (Fig. 20). A similar prolongation of the action potential was also produced by serotonin, by cAMP, and by phosphodiesterase inhibitors (Klein & Kandel, 1978).

The increase in Ca²⁺ current could be produced by a direct action on the Ca²⁺ channel or by an indirect action by means of a reduction in an opposing K⁺ current. To distinguish between direct and indirect actions on the Ca²⁺ current, we voltage clamped the cell body of the sensory neurone and analysed the changes in ionic currents produced by connective stimulation and by serotonin, the presumed facilitating transmitter.

We again introduced depolarizing commands in the sensory neurones under voltage clamp and recorded both the presynaptic currents and PSPs in follower cells. In normal sea water stimulation of the connective produced facilitation of the EPSP and a decrease in the outward current (Fig. 21). This experiment confirms in the absenced of drugs the finding made in TEA solution that connective stimulation can alter the configuration of the currents contributing to the action potential. To analyse this change in the presynaptic current, we isolated the cell body to maximize voltage-clamp control, using serotonin to simulate the presynaptic facilitation. In sea water containing no drugs, serotonin caused a decrease in transient outward currents elicited with depolarizing steps, as well as the decrease in resting conductance (Fig. 22 B, CJ. However, when potassium currents were blocked with TEA and 4-AP, serotonin had no effect on any membrane conductance (Fig. 22 A, C₂). The effects of serotonin were blocked with either Ca²⁺ or Ba²⁺ as the divalent cation that was carrying current through the Ca²⁺ channels.

To rule out the possibility that the K⁺ channel-blocking agents were interfering with the action of serotonin on the sensory neurone, we also eliminated the K⁺ current in a different way. We substituted the non-permeant cation Cs+ for the intracellular K⁺ using the antibiotic nystatin, which makes the membrane permeable to monovalent cations (Russell, Eaton & Brodwick, 1977). We then washed out the nystatin and recorded from the cell in K⁺-free solution with electrodes filled with CsCl. Under these circumstances, serotonin produced no modulation of the current. By contrast, when a small amount of K⁺ was reintroduced into the cell with a microelectrode filled with KC1, the net current became outward, and now was again affected by serotonin (Fig. 23). These experiments imply that serotonin, and presumably presynaptic facilitation, act on K⁺ currents, and have no significant direct action on the calcium channel.

The finding that presynaptic facilitation works by means of the depression of K⁺ currents implies that the increase in Ca²⁺ current is secondary to a change in the configuration of the action potential. This in turn leads to three predictions: (1) The action potential of the sensory neurones in normal solutions should be prolonged by connective stimulation and by serotonin; (2) Prolonging a depolarizing step under voltage clamp should produce similar enhancement of synaptic transmission; and (3) Facilitation should cause a characteristic change in the shape of the synaptic potential. We have tested each of these predictions experimentally and have found good agreement with the experimental results.

Connective stimulation (or serotonin) produces a broadening of the action potential. In the cell body the broadening is modest (Fig. 24A). Since the density of the Ca²⁺ channels relative to the Na⁺ channels may be higher in the terminals than in the cell body, we increased the extracellular Ca²⁺ concentration and lowered the Na+ concentration, to simulate the action potential that may be present in the terminals. Under these circumstances, the increase in duration is greater and there is also an increase in the peak amplitude of the action potential (Fig. 24 B).

To strengthen the conclusion that increased transmitter release during sensitization is caused by broadening of the action potential, we next increased the duration of the presynaptic depolarization under voltage clamp in an attempt to simulate the changes that occur in the action potential. We clamped a sensory neurone in the presence of Na⁺ and K⁺ channel-blocking agents and elicited an EPSP in a follower cell with a depolarizing command step. When we increased the duration of a 20 msec depolarizing step by 25%, the EPSP doubled in amplitude (Fig. 25). This occurs because the Ca²⁺ current has a very long rise time (approximately 10–30 msec) compared to the duration of the step, and, in this cell, is still increasing at the end of the 20 msec command. Lengthening the command by 5 msec increases appreciably the calcium influx (measured by the area under the current). If the depolarization were as short as an action potential (approximately 2 msec at half amplitude), an even smaller percentage increase in duration of the spike might increase the EPSP significantly. The observed changes in action potential duration could therefore cause the increase in transmitter release that underlies sensitization.

If increased release were a result of broadening of the presynaptic action potential, the Ca²⁺ current would be activated for a longer time, and we might expect to see a corresponding change in the configuration of the EPSP. When we superimposed the two EPSPs of Fig. 25 produced by command pulses of different duration, adjusting for the amplitude difference, we found, not surprisingly, that the PSP elicited with the 25 msec depolarization reached its peak later than the PSP elicited with a 20 msec step (Fig. 25 C). Similarly, the PSPs produced by action potentials after a sensitizing stimulus show longer rise times compared to the PSPs elicited prior to sensitization. These changes in shape are qualitatively similar to those produced by increasing the duration of a voltage-clamp command (Fig. 26).

This change in shape is not simply a consequence of the difference in size before and after facilitation. The changes in PSP amplitude produced by habituation and spontaneous recovery from habituation do not affect the shape of the synaptic potential (Fig. 26). Similarly, changes in PSP amplitude produced by altering the calcium and magnesium concentrations do not produce shape changes (Castellucci & Kandel, 1974). On the other hand, when the PSP is increased by adding TEA to the bathing solution, blocking K⁺ current, the increase in PSP amplitude is accompanied by a prolongation of its rise time, similar to that seen in presynaptic facilitation (Castellucci & Kandel, 1974). These findings strengthen our earlier conclusion that habituation is caused directly by an intrinsic decline in calcium current rather than increased potassium current, and that habituation and sensitization reflect different modes of calcium current modulation.

These several findings indicate that the increase in Ca²+ current that accompanies presynaptic facilitation is due to an indirect action on the Ca²⁺ current mediated by a depression of the opposing K⁺ current. Turning off the K⁺ current causes an increase in the duration of the action potential which increases the Ca²⁺ influx (Fig. 27). Thus, one of the interesting consequences of the slow modulatory synaptic action that produces sensitization is that it does not simply change the membrane potential. Instead, the synaptic action has a novel effect: it acts on voltage-dependent K⁺ channels to alter the ionic makeup and the configuration of the action potential, thereby altering transmitter release.

Our studies so far have indicated that presynaptic facilitation is due to a decrease in the K⁺ currents of the presynaptic neurone, but we have not delineated which of the K⁺ currents is critically involved. Using the various blocking agents we can now dissect out these currents and determine how much each current contributes to the modulation of the total outward current. Having done that, it might also be possible to specify what aspect of the current is altered (number of channels activated, channel selectivity, or rate constant of activation or inactivation).

The ability to analyse simultaneously the Ca²⁺ current of the presynaptic cell body and transmitter release from the terminals has allowed us to explore the alterations in Ca²⁺ current underlying a variety of plastic changes in transmitter release: the control of transmitter release by membrane potential, presynaptic inhibition, homosynaptic depression, and presynaptic facilitation. In this type of analysis most of the total voltage-clamp current is recorded across the soma membrane, which parallels the transmitter-releasing membrane area but probably does not directly contribute to release. Nevertheless, it is possible to demonstrate that these currents correlate with transmitter release. Moreover, the cell body currents examined in isolated somata where the clamp control is optimal have properties similar to the currents recorded in intact cells.

It has been known for more than a decade that the influx of Ca²+ into the terminals with each action potential is an essential step for transmitter release (Katz & Miledi, 1967a). Whether changes in Ca^24- influx are involved in the well-known capability of chemical synapses to undergo plastic changes in synaptic effectiveness was not known, however. The experiments presented in this paper provide direct evidence that the calcium influx into the terminals can be regulated in both directions, that the regulation can be direct or indirect, and that this regulation can contribute to control of synaptic efficacy lasting minutes to hours (Table 1). The indirect modulation of the Ca²⁺ current is achieved through the regulation of K⁺ currents that control the height and duration of the action potential and thereby determine the time during which the Ca²⁺ channels are activated. Direct modulation of the Ca²⁺ current is achieved, on the other hand, perhaps through actions on the Ca²⁺ channel protein itself (Fig. 28).

Our data suggest that the ability to modulate the Ca²⁺ current directly derives from an interesting characteristic of the Ca²⁺ channel protein: that it is endowed with several regulatory sites. These regulatory sites are illustrated schematically in Fig. 28 B₁. Based on our data we can identify sites that are sensitive to at least two different regulatory signals: (1) membrane voltage and (2) chemical transmitters. Data of others studying properties of the Ca²⁺ channel in the vertebrate heart, in paramecium, and in molluscan nerve cell bodies suggest the presence of two other sites on the Ca²⁺ channel: those that respond to (3) cyclic nucleotide action (Reuter & Scholtz, 1977) and those that are sensitive to (4) intracellular Ca²⁺ concentration (Brehm & Eckert, 1978; Tillotson, 1979). The ability to modulate the Ca²⁺ current indirectly seems to derive from the existence of a similar set of regulatory bites on the K⁺ channel (Fig. 28 B₂). In particular, both the K⁺ and the Ca²⁺ channel are capable of being affected by membrane voltage on the one hand and by chemical transmitters (and perhaps by one or more intracellular regulatory signals) on the other.

That certain ionic channels are sensitive to both transmitters and membrane voltage has now been confirmed in a number of systems. At certain synapses the voltage sensitivity of a chemically activated channel affects primarily the decay of the synaptic current; the voltage sensitivity does not alter the rate of rise or the peak current (Kordas, 1969; Anderson & Stevens, 1973). At other synapses the voltage dependence is more dramatic and can alter significantly the amplitude and the configuration of the PSP at various membrane potentials. (Pellmar & Wilson, 1977; Pellmar & Carpenter, 1979; Hartzell et al. 1977; Weiss et al. 1978). Our data suggest that voltage-gated molecules that are transmitter sensitive occur in or near the terminal region of the neurone. The sensitivity to transmitter actions of these voltage-dependent channels can lead to changes in the ionic makeup of the action potential that can exert a powerful control over transmitter release.

Thus, the existence on both the Ca²+ and K⁺ channels of regulatory sites for chemical transmitters on the one hand and for membrane voltage on the other allows a mechanistic explanation for several different forms of long-term synaptic plasticity: presynaptic inhibition, presynaptic facilitation, homosynaptic depression, and the dependence of transmitter release on membrane potential. Although these forms of plasticity vary in their detailed mechanisms, they all share a common mode of expression-the modulation of the Ca²⁺ current. The finding of multiple controls of the Ca²⁺ current also makes the modulation of this current an attractive candidate mechanism for long-term synaptic control of the sort that might be involved in the development of synaptic connexions on the one hand and long-term learning on the other.