I. INTRODUCTION

Top Previous Next References

One of the main ways of neuronal communication in the nervous system is by chemical synaptic transmission, where the presynaptic nerve terminal releases transmitters that interact with the postsynaptic target to produce the synaptic response. This review deals with one specific aspect of the function of the secreting presynaptic nerve terminal: the role of the ion channels in its activity.

The primary function of the presynaptic nerve terminal is to release transmitter and thus activate the postsynaptic target cell. In almost every step leading to the release of transmitter, there is a substantial involvement of ion channels. A conservative estimate suggests that several tens of different ion channel types are present at the surface membrane and inside the presynaptic nerve terminal. The number of different ion channel molecules is probably well over a hundred, with obvious implications in control of transmitter release and synaptic efficacy. In this introduction, we want to review in brief the various steps leading to the release of transmitter. In doing so we hope to clarify the possible role of the large number of different ion channels involved in neurosecretion.

Conceptually, ion channels are very simple molecular machines. They exist in two main states: open and closed (named also shut state) (see Ref. 743). When they are in an open state, they pass ions passively according their electrochemical gradient (the combined electrical and concentration driving forces). The transition between the open and the shut state is termed gating. In this review, we see predominantly two types of gating of ion channels in the presynaptic nerve terminals: voltage gating and ligand gating. The voltage-gated channels change their probability of transition, from shut to open state, according to the membrane potential of the nerve terminal. The ligand-gated channels respond to the concentration of a specific ligand inside the nerve terminal, in the extracellular space or in the membrane domain.

When ion channels open and ions flow through an open channel according to their electrochemical gradient, three main changes occur: 1) there are changes in the concentration of the relevant ions inside the nerve terminal and in the immediate extracellular space, 2) there is a change in the membrane potential, and 3) there is a change in the membrane resistance: the opening of channels causes an increase in conductance and hence the membrane resistance decreases. This in turn affects the space constant and the time constant of the membrane, and other ionic currents have an altered efficacy.

We have seen in a previous section that the main "workhorse" in the process of quantal transmitter release is calcium. Therefore, we start the description of the ion channels in the presynaptic nerve terminal by discussing the properties of the calcium channels in some detail. Because most (but definitely not all) of the calcium channels in the nerve endings are voltage gated, we discuss thereafter the ion channels that control the membrane potential (and the membrane conductance) of the nerve terminal, namely, the potassium, the sodium, the nonselective, and the chloride channels (see Fig. 1). Finally, we briefly discuss the ligand-gated channels and the other channels present in the nerve terminal.

View larger version (31K): [in this window] [in a new window]   Fig. 1. Major categories of ion channels in presynaptic nerve terminals. Surface membrane contains calcium, potassium, sodium, chloride, as well as nonselective ligand-gated channels and probably stretch-regulated channels; the sodium/calcium exchanger at the surface membrane has channel-like properties. Ion channels were found also in intracellular organelles: calcium, nonselective, chloride, and anion channels.

	II. CALCIUM CHANNELS IN NERVE TERMINALS

Top Previous Next References

Calcium ions have many functions in neuronal cells. These include spike initiation (172, 226, 544, 582), rhythmic firing, neurite outgrowth, gene expression, and transmitter release (for review, see Ref. 161). The latter function is the central topic of this review article; hence, the other functions are mentioned only briefly.

Originally, it was assumed that there was only one type of calcium channel. However, the pioneering work of Hagiwara et al. (315) indicated that there may be more than one type of calcium current in the egg cell membrane of the starfish. Subsequently, it was found that such a distinction occurs in many other cells from different organisms (56, 58, 126, 595, 601, 702). The different currents have different voltage thresholds for activation; thus low voltage-activated channels (LVA) are those in which the activation is slightly above the resting potential, and high voltage-activated channels (HVA) are those in which the threshold for activation is substantially above the resting potential (towards 0 mV). In addition to the activation voltage, different calcium channels may be distinguished by their single-channel properties, activation kinetics, and inactivation kinetics. These properties of the channels in the nerve terminal are summarized in section IID.

2.  Pharmacological classification

A) L, N, AND T CHANNELS. Once it was realized that there are at least two different types of calcium currents and channels, a search started to find specific pharmacological agents. A family of chemical substances, the dihydropyridines (DHP), was found to affect HVA channels (336) but not LVA channels (56). Some members of the family, such as nitredipine, inhibit the HVA channel activity, whereas other compounds, such as BAY K 8644, activate the HVA calcium channels (56, 336).

View larger version (21K): [in this window] [in a new window]   Fig. 2. Calcium channels in the surface membrane of the presynaptic nerve terminal. Channels include voltage-gated calcium channels (N, L, P, and Q), the nonselective channel NS, the ligand-gated calcium-permeable channels LG, and the sodium/calcium exchanger. There are also intracellular calcium channels (not shown; see text).

B) MINDING THE P AND Q. The story does not end here. An additional type of HVA calcium channel, originally found in cerebellar Purkinje cells (156, 483, 910), is the P-type calcium channel, which is inhibited by funnel-web spider polyamine and peptide toxins FTX and -agatoxin IVA, respectively (551, 762). When the calcium channel sensitivity to agatoxin was tested on channels expressed in oocytes, it was found that these channels have a low sensitivity to agatoxin (~200 nM) (752), whereas the P-type sensitivity is much higher (~2 nM). This led to the proposal that there is yet another type of voltage-sensitive HVA calcium channel: the Q-type calcium channel (1003). The distinction between the P-type and Q-type calcium channels is not always obvious, and hence, they are frequently grouped together, as the P/Q type of calcium channels.

C) R-TYPE CALCIUM CHANNELS. If T, L, N, and P/Q comprised all the voltage-dependent channels that allow the entry of calcium into the nerve cell upon depolarization, then it would be possible to abolish all the calcium entry by a small and prolonged depolarization (to inactivate the T-type calcium channels) and by a "cocktail" containing an inhibitory DHP, -conotoxin GVIA, and -agatoxin IVA at high concentrations. This is not always the case, and some residual calcium channel activity remains. Thus it was proposed that this residual calcium entry activity be attributed to channels named R-type calcium channels. Their activation voltage is between that of HVA channels and LVA channels. This activity is blocked by a low concentration of nickel (1003).

3.  Molecular classification

A) THE SUBUNITS. Voltage-dependent calcium channels (VDCC), like many other channels, are composed of different subunits. There are five different subunits that are associated with the VDCC activity: ₁, ₂, , , and . The -subunit has been found in skeletal muscle but not in brain and is not discussed here (211). The ₂- and -subunits are connected by a disulfide bridge and are referred to as a unitary complex ₂. Hence, in the mammalian nervous system, we are dealing with three distinct subunits ₁, ₂, and  that together form a functional calcium channel (959, 1003).

B) ₁-SUBUNIT. Seven different genes were found to encode the ₁-subunit. They are named A, B, C, D, E, G, and S. The S gene encodes the ₁-subunit in the skeletal muscle; the products of the other six genes were found in the brain. Each of the genes can produce more than one gene product by alternative splicing. Altogether, at least 18 different ₁-gene products have been identified in the nervous system (79, 648, 806).

C) -SUBUNIT. Four different genes were found to encode the -subunit and are named 1, 2, 3, and 4. Each of the genes can produce more than one gene product by alternative splicing. At least eight different -gene products have been identified in the brain (79, 133, 374, 647).

D) ₂-SUBUNIT. The -subunit seems to have an identical structure to that of the COOH-terminal part of the ₂-subunit and hence may be the product of the same gene, altered by posttranslational modification. A splice variant was found in the rat brain (199). This subunit may be involved in the regulation of secretion (967).

There are three different classifications of the voltage-sensitive calcium channels. Clearly, it is of great interest to find which channel corresponds to what molecular structure; however, only a partial achievement of this aim is possible at present. The DHP-sensitive L-type calcium channels in the nervous system have the _1C- or _1D-subunit. The -conotoxin GVIA-sensitive N-type calcium channel has the _1B-subunit. The _1A-subunit is probably part of the P/Q-type calcium channels. The _1G-subunit is part of the T-type calcium channel (648). It is of interest to note that when various ₁-subunits are expressed without other subunits, the resulting calcium channels have properties similar to those of T-type channels (534). It was proposed that part of the calcium channels named R type are formed by the _1E-subunit (see Ref. 79). For clarity, a translation table follows (Table 1).

To appreciate the diversity of the calcium routes into the neuron and the nerve terminal, let us make a rough estimate of only one of the components of the calcium entry systems into the nerve: the voltage-gated calcium-selective ion channels. Such a calculation shows that the number of possible molecular species is really astounding. There are at least 18 different ₁-subunit genes including splice variants and at least 8 different -subunits, without taking into consideration posttranslational modifications. If there are no forbidden assemblies, at least 144 different types of calcium channels can be generated with these 2 subunits only. If one takes into account also the splice variants of the ₂-subunit, one reaches at least 288 different calcium channel molecules. Of course, not all the permutations are possible, since not all the genes are expressed in the same cell, but the possibility for diversity is extensive. The diversity is even larger envisaging the molecular interactions of calcium channels with other proteins (see, for example, Ref. 698), transmitters, and second messengers. If one also takes into consideration the other possible routes for calcium entry, then there is no doubt that the nerve cell possesses many hundreds (if not thousands) of different ways to alter the [Ca²⁺]_i. Even more surprising is that many of the various routes are present in the same nerve terminal and may be responsible for transmitter release. Why does the nerve terminal need such a wide variety of calcium entry routes? We speculate that such multitude of control of the same basic function allows a great plasticity in transmitter release and thus in synaptic transmission in the nervous system (see below). The presence of different calcium channels in the same nerve terminal may also have an importance in the partial protection against animal toxins. If a toxin is "administered" in the body, it will block only part of the calcium channels; this may have an obvious evolutionary advantage.

It is no wonder then that such molecular diversity leads to a large number of pharmacological actions. In section IIC8 and Table 2 we present the pharmacological modification of the presynaptic nerve terminal function, using agents acting on one or more molecular targets.

The most desirable methods to study the properties of the calcium channels in the presynaptic nerve terminals are direct electrophysiological methods and specific molecular probes for calcium channels. However, there is a great gap between what is desirable and what is available. The calcium channels in the nerve terminals can be studied directly, at the single-channel level using the patch-clamp technique, only in a very limited number of preparations, namely, the chick calyx synapse (820, 845), the rat calyx of Held (52, 95, 264, 372, 861), and the peptidergic nerve terminals of the rat neurohypophysis which release their content into the bloodstream (292, 417, 766, 945). (Calcium currents can be studied in a larger number of preparations.) Thus some extrapolation from these preparations, where direct methods are feasible, to other interesting structures is necessary. Most of our knowledge, regarding calcium channel properties in nerve terminals, comes from indirect methods. In section I, we outlined the various steps in presynaptic function. They may also serve as an indication of how far the method is from the actual activity of the calcium channels in the presynaptic nerve terminal. We briefly summarize here these indirect methods and their inherent difficulties, to assess the strength of various arguments regarding the role of calcium channels in the regulation of presynaptic function.

The primary function of the nerve terminal is to release quanta of transmitter (200, 247). Because the release of transmitter quanta is a function of [Ca²⁺]_i, and because [Ca²⁺]_i is a function of the activity of the calcium channels at the surface membrane, it was assumed that measurements of evoked quantal transmitter release represent mainly the activity of the calcium channels. The drawback of this method, however, is the nonlinearity between calcium concentration and transmitter release (209). Because this nonlinear relation has a sigmoidal shape (45, 209, 360) and because [Ca²⁺]_i can be affected by additional pathways and not only by the surface membrane calcium channels, the conclusions from such studies are indirect.

The release of transmitter can be measured not only by physiological means, but also by biochemical methods using pinched-off nerve terminals (591) (for review, see Ref. 962), as a function of calcium concentration (256, 323, 367, 384, 489, 623, 696, 705, 736, 895, 899, 900). The difficulty with this method is that the synaptosome is usually not amenable to physiological stimulation, and the depolarization needed for the opening of the calcium channels to evoke transmitter release is achieved by elevating the extracellular potassium concentration. Because changes in extracellular potassium concentration ([K⁺]_o) produce usually a much slower and more prolonged effect on the membrane potential, channel inactivation probably occurs. An additional difficulty is that the release of transmitter to the extracellular medium reflects not only evoked transmitter release from the presynaptic nerve terminal, but also the molecular leakage found to occur in a number of synapses (see Ref. 262, 408, 433).

Using radioactively labeled calcium, one can measure calcium influx into the nerve terminal (mainly synaptosomes but also in brain slices) (287, 486, 580, 703) and the resulting transmitter release. The main disadvantage of this technique for both influx and transmitter release is the time resolution, which is usually too slow for detailed characterization of the presynaptic processes involved.

By using fluorescent compounds, which change their emission in response to the calcium concentration in their vicinity, one can measure the changes in [Ca²⁺]_i (and intraterminal calcium concentration). This method is also widely used for studying calcium channels in presynaptic terminal membranes (105, 119, 189, 235, 332, 445, 461, 538, 550, 685, 805, 832, 833, 880, 977). In passing, we want to mention that fluorescence probes have also been used to monitor the membrane potential and the shape of the nerve terminals (539).

In most studies, the release of a quantum of transmitter was measured by its postsynaptic action. However, this information can also be obtained from the properties of the releasing cell, using variants of the patch-clamp technique (see Ref. 743). The whole cell patch-clamp technique enables one to monitor the changes in membrane surface area via changes in capacitance, and therefore to measure the exocytosis process directly. Combining this method with pharmacological studies reveals the contribution of the calcium channels to the transmitter release process (418, 469, 766, 767, 927).

In the calyx synapse of the chick, it was possible to record both "whole terminal" and single calcium channel activity (817, 824, 989, 990). This presynaptic terminal is exposed after enzymatic treatment. Almost all the methods described above were used in the terminals of the rat neurohypophysis. The role and contribution of calcium channels to secretion could be compared by different methods such as single-channel recordings (458), intraterminal recordings (99), whole terminal recordings (264, 943), calcium imaging, neurotransmitter release (833), and exocytosis measurements (766).

Single-channel recording has obvious advantages for monitoring the channel activity directly, enabling detailed electrophysiological and pharmacological characterization. The disadvantage is that the channel is usually studied in isolation, and the interpretation of its contribution to a physiological process is not direct. On the other hand, whole terminal recording and intraterminal recording supply the characteristics of the macroscopic current that compensates for this disadvantage in the single-channel recording. Single-channel recordings were made also in altered preparations of presynaptic terminal such as fused synaptosomes (535, 536) or presynaptic channels expressed in model systems (904).

Extracellular recording was used in many preparations to overcome the difficulty of recording calcium channels in the small presynaptic terminal. By blocking the potassium and sodium currents using pharmacological tools, many laboratories were able to record extracellular currents defined as carried by calcium ions either by calcium depletion or by agents known to block calcium currents. It was used mainly in neuromuscular junction preparations (see, for example, Refs. 23, 508). One of the main drawbacks of this method is that it is impossible to define in detail the biophysical characteristics of the current measured.

Usually the various methods are combined with specific agonists or antagonists that determine the existence and define the type of calcium channel. It should be mentioned that the basic pharmacology was usually investigated not in nerve terminal preparations; hence, some extrapolation was often necessary to reach the conclusions regarding the nerve terminal. Because the number of pharmacological studies of transmitter release is vast, most of them are omitted from this review. Some examples are given in Table 2.

The methods described in the previous section were used to determine the functional properties of calcium channels in the nerve terminals (see Table 3). Valuable information is available from nerve terminals where direct patch-clamp recordings were made, and calcium channels were characterized on the single-channel or on the whole terminal level. This was achieved in a very limited number of nerve terminals, among them the peptidergic terminals of the rat neurohypophysis (943), the calyx synapse of the chick (818), and at the synapse of Held (52, 264, 372, 861).

Before starting the description of the various channels, we want to point out two generalizations. First, the voltage-activated calcium channels characterized until now at the nerve terminals are of the high voltage-activated (HVA channels) variety and, therefore, are activated by an action potential invading the nerve terminal. Second, most of the channels with all their characteristics do not fit into the conventional nomenclature of voltage-activated calcium channels. In most cases, we use the names proposed in the original articles.

An L-type calcium channel has been described in rat neurohypophysial terminals; it displays a single-channel conductance of 25 pS (with 110 mM Ba²⁺). The threshold for activation is around 20 mV, and it shows no voltage-dependent inactivation. The decay time kinetics of the ensemble currents are <500 ms at the single-channel level (943) and 1,250 ms for the whole terminal current (943). The mean open time during voltage pulses from a holding potential of 50 mV to test potential of +10 mV is 0.49 ms, and the closed time distribution has two components with means of 2.02 and 79.91 ms. The activity (NP_o) of the channel increases with voltage. There is no information about the activation and deactivation kinetics.

In goldfish retinal bipolar synaptic terminals, the only calcium current is of the L type. Its threshold for activation is 50 mV, and its maximal amplitude is at 15 mV (332). This current shows slow calcium-dependent inactivation with a minimal time constant of 4.6 s. The whole cell calcium current inactivation is due not to change in the voltage dependence of the channel, but to the closure of channels (928).

In neurohypophysial terminals, the N_t type has a single-channel conductance of 11 pS (with 110 mM Ba²⁺). The threshold for activation is around 10 mV, and it shows voltage-dependent inactivation (half-maximal voltage is around 68.5 mV). The decay time kinetics of the ensemble currents are <50 ms at the single-channel level (943) and 100 ms for the whole terminal current (943). The mean open time during voltage pulses from a holding potential of 90 mV to test potential of 10 mV is 0.34 ms, and the closed time has two components with means of 1.78 and 86.6 ms. The activity (NP_o) of the channel increases with voltage. In some of the patches, the channel did not inactivate and continued to be open throughout the test pulse.

In the presynaptic nerve terminals of the chick ciliary ganglion, one type of channel was characterized on a single-channel level, and two components were dissected on pharmacological and inactivation bases (990). The single-channel conductance ranges between 11 and 14 pS (with 110 mM Ba²⁺) (818) and is reduced to 10 pS with 6 mM Ca²⁺ and 6 mM Ba²⁺ (819). The threshold for activation is around 30 mV, and it has little voltage-dependent inactivation. The activation rate constant in whole terminal current is ~1.5 ms at positive potentials, and the deactivation rate at 80 mV is 0.77 ms (824). Although no inactivation was observed in the "whole cell" current when barium was the charge carrier, clear inactivation was detected when calcium ions replaced it, and the rate of inactivation increased with external calcium concentration (990). This calcium-dependent inactivation may be applicable to other VDCC in nerve terminals that were examined with barium.

In the squid giant presynaptic terminal, the calcium current is probably carried via P channels (150). The threshold for activation is around 40 mV, and it shows no voltage-dependent inactivation during 25-ms voltage pulse. The activation is half-maximal at 13 mV and maximal at 20 mV. Activation half times were reduced from 1.5 ms at 20 mV to 0.5 ms at 20 mV (46).

This type of channel was not characterized directly in a nerve terminal preparation but was suggested to be involved in the transmitter release process and therefore to be located at the presynaptic nerve terminal (960). Characterization of this channel by expressing class A ₁-subunit in Xenopus oocytes revealed that this subunit is the Q type (752). The threshold for activation is around 10 mV, and it shows steep voltage-dependent inactivation. Its single-channel conductance is 16 pS, and the time to peak current amplitude is ~5 ms, time for half decay (inactivation) is 116 ms, and the single-channel open time distribution has two components of 0.5 and 2.6 ms.

Recently, it was shown by Wu et al. (976) that R-type calcium channels may be involved in the regulation of transmitter release at the calyx-type synapse of the rat medial nucleus of the trapezoid body. After abolishing the L, N, and the P/Q type of calcium channels by pharmacological agents, they recorded the remaining calcium current that constituted 26% of the total calcium current. It fitted the classification of R-type calcium current. This current was large enough to generate release of transmitter sufficient for a suprathreshold postsynaptic response. There are three very interesting properties of this R-type calcium channel: the activity of this channel is inhibited by metabotropic glutamate receptors, the activity is inhibited by GABA_B receptors, and the calcium sensitivity of release induced by the activation of these channels is substantially lower than in other cases (see sect. IIF).

E.  "Demography" and "Geography" of Calcium Channels

1.  Calcium channels are localized at presynaptic terminals

The existence of calcium-conducting pathways into the nerve terminal has been shown by a variety of methods. Some of the methods are rather indirect and show that the [Ca²⁺]_i increases in the nerve terminal; other methods are more direct in nature, enabling the visualization of the channel protein.

The initial proposal that calcium channels exist in the nerve terminals came from electrophysiological experimentation at the squid giant synapse (45, 46, 151, 406) and at the neuromuscular junction of a number of species (see Refs. 23, 106, 458, 508). In this section we present some recent articles showing an increase in [Ca²⁺]_i in the nerve terminal, or the existence of presumed specific channel proteins, in different vertebrate and invertebrate species.

A) SQUID. With the use of fura 2 measurements, it was found at the squid giant presynaptic terminals that calcium concentration was highest in the compartment closest to the postsynaptic neuron. The colocalization of calcium transients and active zones strongly suggests that neurons cluster calcium channels selectively at active zones, and this localization enhances the magnitude of calcium signals in the vicinity of the active zones (484, 805).

B) BARNACLE. Presynaptic terminal region, and individual photoreceptor terminals of the barnacle, Balanus nubilus, were studied using the calcium indicator dyes arsenazo III and fura 2. It was found that calcium entry occurs in a restricted region <50 µm in length, which corresponded closely to the region of synaptic contact with second-order cells (830), and that depolarizing pulses produced voltage-dependent calcium entry, that was confined to the tips of the arborization (119).

C) TORPEDO. With the use of a combination of colloidal gold labeling and freeze-fracture techniques, it was found in nerve terminals isolated from the electric organ of Torpedo marmorata, that antagonist specific for voltage-activated calcium channels binds to intramembrane particles in presynaptic membranes. Biotinylated derivative of -conotoxin exerts an inhibitory action on the high potassium-evoked release of ATP (243).

D) GOLDFISH. In goldfish cultured retinal ganglion cell growth cones, with the use of a circular vibrating microprobe, it was found that cell growth cones generate steady inward currents at their tips. The major part of this current is carried by calcium ions, which are suggested to flow through a population of voltage-sensitive calcium channels located on the filopodial tips (271).

E) FROG. The initial proposal of the localization of calcium ions at the active zones came from early morphological studies (339, 340). More recently, fluorescent probes were used; at the frog motor nerve terminals, tetramethylrhodamine-conjugated -conotoxin fluorescent stain consisted of a series of narrow bands (in face views) or dots (in side views) ~1 µm apart on the synaptic rather than the nonsynaptic side of the nerve terminal. The bands and dots of stain were in spatial register with the postsynaptic junctional folds as revealed by combined staining of ACh receptors (166). In frog hair cells from sacculus, the calcium indicator fluo 3 was imaged by fluorescence confocal microscopy; when a cell was depolarized, on its basolateral surface, several foci of transiently enhanced fluorescence due to local calcium influx occurred. After protracted recording, each cell displayed on average 18 brightly and permanently fluorescent spots at the same positions. Measurement of currents through membrane patches at fluorescently labeled active zones demonstrated the presence of both voltage-activated calcium channels and calcium-activated potassium channels (376).

F) LIZARD. In lizard axon terminals on twitch and tonic muscle fibers in intercostal muscles, the freeze-fracture technique was used. Differences in quantal output are related to the observed differences in the number of active zone particles flanking synaptic vesicles at the active zone (which are probably the calcium channels; Ref. 935). Evidence for the presence of calcium channels in these nerve terminals was obtained by electrophysiological and optical methods (23, 473, 474, 538, 568).

G) CHICK. In chick calyx-type nerve terminal of the ciliary ganglion, atomic force microscopy revealed low (1/µm²) and high (55/µm²) calcium channel density. Prominent interchannel spacing of 20 nm indicated an intermolecular linkage. Particles were observed in clusters and short linear or parallel linear arrays (328).

H) RAT AND MOUSE. In rat neuromuscular junctions, using electron cytochemical analysis, it was found that "A" sites are located at the openings of junctional folds; these triangular elements are identical to presynaptic protrusions of the active zone and probably comprise calcium channels of the presynaptic membrane (184).

I) GUINEA PIG. In guinea pig transverse slices of the hippocampus, histochemical methods showed that electrical stimulation combined with application of aminopyridine compounds led to electron-dense deposits of 60-400 nm diameter, mainly restricted to the activated input layers. Deposits were predominantly found at presynaptic sites (435) and may represent calcium channels.

In the rat retina (and in some endocrine cells) (89, 485), L-type channels control secretion (628). On the other hand, on motor nerve terminals innervating both skeletal and smooth muscle, it seems at the moment that only N-type calcium channels have been found to control neurotransmission (37, 272, 318, 576, 707). In the rat central nervous system (CNS), it seems that the picture is even more complex; in spinal cord, brain stem, neurohypophysis, cerebellum, midbrain, hippocampus, and cortex, it was suggested that more than one type of VDCC exists in the nerve terminals (see also sect. IIE4). Spinal cord sensory neurons possess mainly N-type calcium channels, but also L- and P-type calcium channels have been described. (543). However, in dorsal horn and superior cervical ganglion, there is P-type dominance with smaller N-type contribution (300, 862); the same picture emerges in brain stem interneurons (905). In the neurohypophysis, it was established that there are L, N or N-like, and P/Q channels (832, 944, 945), but it was also suggested that only N-type calcium channels contribute to secretion (930). In the cerebellum, again the picture is of P-type dominance with smaller N-type contribution and no L-type calcium channels (695, 862), although L-type was suggested to be involved in the modulation of calcium currents and glutamate release by GABA (367). In the midbrain, the picture is even more complex, with different types of neurons releasing different types of neurotransmitters (898). In GABA release, there is N-type dominance and small L-type contribution (421). In dopamine release, the contribution is either almost equal for N, L, and P/Q (131) or mainly P with slight N (898, 981) while only P in glutamate release (898). In the modulation of the calcium signal by adenosine and ATP (655), both N and L types are involved. In cultures of hippocampal neurons, it is mainly N type that mediates exocytosis, with small contribution of P/Q or both P/Q and L (700). In hippocampal slice preparations, P/Q-type channels dominate transmission, and N-type channels contribute much less (489, 615, 862); in other studies, N-type channels were also suggested as the main route for calcium entry (379, 487, 666, 736, 876). In the different areas of the rat cortex, there are mainly P type in cerebrocortical synaptosomes (872, 880) 970), in frontal cortex synaptosomes (609), and in neocortical mini-slices (287), whereas L and N were found (736) in other cortical synaptosomes. The data on the types of channels in rat nerve terminals are summarized in Table 4.

The types of VDCC differ in different species and anatomical localization. In motor nerve terminals (innervating skeletal muscle), all the types of nerve terminal calcium channels (L, N, P, and Q) were found in different species. Usually a combination of more than one type of channel in each terminal was suggested (see sect. IIE4). In insect (grasshopper and housefly) motor nerve terminals, it is P and/or Q (78). In lobster, it is the N-type channel that underlies transmitter release (310). In the lizard, it is L type (473); in the frog and Xenopus, it is L and N (30, 254, 273, 324, 414, 718, 720); and in electromotor neurons of electric fish, it is N (244) and Q (705). In mammals, L type was found in the mouse (19, 356) together with N (357, 579), P (669, 901), and P/Q (844), whereas in rat only N type (318) and in humans only P type (668) were found.

It should be noted that similar pharmacology does not constitute a proof of channel identity. In this context, we want to quote the experiments of Fisher and Bourque (259) on the somata and nerve terminals of rat magnocellular neurons of the supraoptic nucleus (see also Ref. 796). Both structures possess calcium currents sensitive to -agatoxin IVA. However, the -agatoxin IVA-sensitive currents at the nerve terminals have a rapid inactivation, whereas those of the cell body have a much slower inactivation. It will be of interest to see in the future whether this biophysical difference between the -agatoxin IVA-sensitive currents is due to subunit composition or posttranslational modification and how the targeting is achieved. For these types of questions, one needs "dissecting" toxins and antibodies for a relevant answer.

The contribution of L-type calcium channels to transmitter release or presynaptic calcium currents is small (see Table 5). Single L-type channels were recorded in the rat neurohypophysis (943), and their contribution to the whole terminal calcium current was estimated to be ~20% (938, 943). The L-type DHP channel blockers partly inhibited the release or the presynaptic calcium current in Aplysia neuronal synapse in buccal ganglion (266), in lizard motor nerve ending (473), and in mouse neuromuscular junction (355, 356). In some other terminals, the L-type contribution was estimated as 11-20% of the total: rat central terminals of visceral sensory neurons (543), rat striatal midbrain synaptosomes (131, 421), and guinea pig hippocampus (134). In rat cultured hippocampal neurons, the L-type contribution was found to occur in only a percentage of the boutons and was estimated to be 23% (700). On the other hand, there was no contribution for L type: in the rat tail arteries (272), in the rat brain stem inhibitory interneurons (905), in the rat cerebellum (550, 695, 862), in the rat spinal slice (862), in the rat (215, 237, 615, 862, 961), and in the rat frontal cortex slice preparation (897). In conclusion, it seems that only a small fraction of the terminal calcium channels is of the L type, and their contribution to transmission is usually smaller than that of other channels.

Most of the studies on the colocalization of different types of calcium channels in nerve terminals indicate that the majority of channels are either N- or P/Q-type channels. The fraction of these types varies in different preparations. In rat tail artery, 70-80% of the calcium currents are of the N type (272); there is no evidence yet for the identity of the other types responsible for the unaccounted fraction. In rat central terminals of visceral sensory neurons, 57% of the calcium current is carried by N type and only 12% by P/Q type (543). In rat striatal synaptosomes, GABA release is dominated by N type (421) (48%), with no indication of P/Q while dopamine and glutamate release are dominated by P/Q type (898) (50-70%). In the rat brain stem inhibitory interneuron terminals, 50% of the calcium channels are of the P type and ~26% are of the N type (905). In three types of synapses, both excitatory and inhibitory in rat cerebellum, there is a clear dominance of P/Q type (550, 695, 862). In rat hippocampal slice preparations, most of the channels are of the P/Q type (215, 237, 489, 862, 961). On the other hand, it should be noted that in cultured hippocampal neurons, the picture is quite the opposite, with the majority of the channels of the N type (615, 700). In guinea pig hippocampal slices, most of the channels are of the P/Q type, colocalized with the N type (134, 983); the same pattern is seen in rat frontal cortex synaptosomes (897). The N-like channel (N_t) described at the rat neurohypophysis (943) was thought to be responsible for the transient component of the calcium current. Later, it was found that a large part of this transient current is sensitive to FTX and -agatoxin IVA (939). The most prominent channels, therefore, appear to be the Q type and the N type, with the L type next in prevalence.

In cultured rat hippocampus, different types of channels are localized and colocalized in different boutons of the same axon; in 45% of the boutons, exocytosis is completely N type dependent, whereas in 55% of the boutons, L and P/Q type also contribute to the release process (700).

Evidence thus points to the P/Q-type channel as the most prominent in many different nerve terminals, with N type also playing a very significant role while the L type contributes little to the transmitter release process at many synapses. It should be noted, however, that this is not the case in all terminals. For example, in goldfish bipolar retinal neurons (332), L type was the only calcium current found and in the calyx-type synapse of the chick ciliary ganglion, and the N-like channel was the only type found both in the whole terminal and in single-channel recordings (328, 817-819, 821).

It has been suggested that the different types of channels are required for kinetically distinct phases of release (55, 470, 581, 646), depending on the electrophysiological characteristics of the type of channel involved. The density and distribution of calcium channels are probably developmentally regulated at some synapses (see Ref. 757). There may be important functional implications in the variations of the distribution of the different channels among terminals. This may reflect the differences in the strength of the synapse and may be responsible for the speed and duration of transmission and the ability of hormones and transmitters to modulate transmission (223).

Data on the fractions of the different channel types in the same terminal are summarized in Table 5.

F.  Cooperativity of Action of Calcium Channels in Transmitter Release

1.  Cooperativity in transmitter release

In previous sections we have shown that different types of calcium channels coexist on the same nerve terminals. Inhibition of the activity of each of calcium channel type in such cohabitation produces a substantial decline in transmitter release from the presynaptic nerve terminals. How is it possible that the sum of all the fractional inhibitions is larger than unity? This phenomenon can be attributed to the cooperative action of calcium ions in the process of transmitter release.

The cooperative relation between calcium and transmitter release was first demonstrated at the frog neuromuscular junction (167, 209). In this preparation, changes in the extracellular calcium concentration ([Ca²⁺]_o) produced a significant change in the number of transmitter quanta liberated by the nerve impulse. In certain concentration ranges of [Ca²⁺]_o, doubling the calcium concentration produced an almost 16-fold increase in quantal release. This observation was interpreted as a cooperative relation between calcium and quantal liberation of transmitter. Subsequently, it was found that a similar relation exists in many other synapses (220, 360, 407, 439, 481). The question remained whether the cooperativity exists between [Ca²⁺]_o and calcium influx into the terminal, as suggested by some investigators (481), or between [Ca²⁺]_i and transmitter release. A voltage-clamp experiment of the nerve terminals at the squid giant synapse examined the relation between transmitter release and calcium influx. In the squid preparation, it is difficult to detect individual transmitter quanta; hence, the relation between the calcium influx and the postsynaptic response was investigated. These studies showed a highly nonlinear dependence of postsynaptic response on calcium entry through the presynaptic membrane, which strengthens the notion of cooperative action of intracellular calcium ions in the process of transmitter release (42, 45) (see Ref. 462, 820).

More recently, this problem was approached by using photolabile calcium chelators that release their calcium upon light (440). Here the entry stage was bypassed, and hence it was possible to examine directly the relation between the [Ca²⁺]_i and transmitter release, measured as the amplitude of the postsynaptic response. These experiments (440) showed that the cooperativity occurs between [Ca²⁺]_i and the process of transmitter release.

The cooperativity of calcium ions in transmitter release is also highly relevant for understanding the interaction of different types of calcium channels in the release process. To illustrate the calcium channel interaction, we will cite the results of Mintz et al. (550). They measured the effects of two different calcium channel blockers on calcium currents (measured as calcium transients by imaging furaptra) and on the amplitude of the excitatory postsynaptic currents (EPSC) in granular cells of rat cerebellar slices. Because the calcium channel blockers had almost no postsynaptic effects, one can take the amplitude of the EPSC as a measure of transmitter release. Calcium current is reduced by 27% with -conotoxin GIVA (that blocks specifically N-type calcium channels) and by 50% with -agatoxin IVA (at concentrations of 50-400 nM that probably block both P- and Q-type calcium channels). These inhibitory effects account for 77% of the calcium currents. Addition of cadmium ions inhibited the remaining calcium currents. The summation of the different channel blockers was linear on the calcium signal. A completely different picture emerged when the actions of the same pharmacological agents were examined at the level of the EPSC. The EPSC were reduced by 50 and 93%, respectively, with the same toxin concentrations. The sum of the fractional inhibitions was greater than unity.

Two important conclusions can be drawn from these experiments. First, the cooperativity between calcium and release is at the level of [Ca²⁺]_i and not at the level of calcium entry. Second, different types of calcium channels act on the same pool of transmitter quanta and are probably intermingled. This conclusion supports the involvement of many calcium channels types in the release of a single quantum of transmitter, discussed in section IIG.

We will illustrate this more than linear summation of the effect of two different toxins with the aid of Figure 3.

View larger version (19K): [in this window] [in a new window]   Fig. 3. Calcium channel cooperativity. A schematic representation of the cooperative action of calcium channels in transmitter release. It is assumed that different channels are inhibited by different inhibitors. Abscissa: normalized calcium concentration after stimulation, without action of any inhibitor. Ordinate: normalized transmitter release. We assume that each one of the inhibitors reduces calcium entry by the same amount. The sum of the effects of the 2 inhibitors is much more than linear. For details, see text.

We assume that the relation between [Ca²⁺]_i and transmitter release is sigmoidal in nature (for short pulses) and that in the physiological range there is an almost linear relation between the activity of the calcium channels and the [Ca²⁺]_i. The abscissa in Figure 3 is the normalized [Ca²⁺]_i, and the ordinate is the normalized release. For simplicity, we take that each one of the toxins reduces the [Ca²⁺]_i following the impulse by the same amount, 40%. Because of the shape of the curve, each one of the toxins (inhibitors in Fig. 3) reduces the release by ~21% when given alone. When the two inhibitors are given together, they reduce the release not by 42%, as expected from linear summation, but by >70%. This more-than-linear effect has many important ramifications regarding the modulation of calcium channel activity in nerve terminals and thus of regulation of the strength of synaptic transmission by different modulators.

This "more than linear inhibitory effect" was observed in a number of other preparations (134, 215, 489, 695, 862, 905, 961, 978).

The major role of calcium channels in nerve terminals is to induce the release of transmitter quanta (see Ref. 223). Therefore, it is of particular interest to determine how many calcium channels have to open within a short time frame to cause the liberation of a quantum of transmitter. Is one calcium channel in the open state sufficient to evoke release of a quantum, or do multiple calcium channels have to be open, almost simultaneously, to generate an exocytotic event?

Let us first examine some of the evidence for the "one channel" suggestion provided by Stanley (819). In Stanley's experiments, single calcium channel openings were recorded at the ciliary ganglion presynaptic nerve terminal of the chick. After dissociation of these nerve terminals, the exposed presynaptic surface can be approached by a patch pipette and high-resistance seals formed. The cell-attached (on-cell) configuration of the patch-clamp technique was used to record single calcium channel openings. It is known that at this synapse, there is a quantal liberation of ACh (514), whereas a previous study has shown that the calcium channels are confined to the release face of the presynaptic calyx. The same patch pipette was used to monitor ACh release by a photon emitting reaction (375). It was found that the photon emission was frequently preceded by a calcium channel opening. This finding by itself is not yet complete evidence that a single calcium channel opening caused the release, since although there may be a single channel opening in the membrane facing the patch pipette, it cannot be excluded that there are many channel openings at the presynaptic membrane in the immediate vicinity outside the patch pipette. However, the association between calcium channel openings and photon emission was found in cases where the rate of channel openings was very low, and the probability of adjacent opening was negligibly low. (It is noteworthy that the current trace showed a clear peak after the photon emission, indicating that an additional channel opens after the fusion. Is this the vesicle channel discussed elsewhere in this review?) The experimental results were compared to a single domain model (69). (The name of this model indicates that the quantal release machinery is the domain of influence of a single calcium channel.) These experiments provide a strong support that a single calcium channel opening can in principle cause the release of a quantum, but is this the situation in normal synaptic transmission?

There is evidence, however, for an alternative model: the model of overlapping domains, where the quantal release machinery is under the influence of many calcium channels. In two recent articles, Borst and co-workers (95, 96) provide evidence that under normal physiological conditions, many more channels are necessary, on the average, to release a quantum of transmitter at the synapse of Held. This synapse is at the rat medial nucleus of the trapezoid body. It is large enough to permit accurate measurements of both the presynaptic and the postsynaptic currents and to measure individual transmitter quanta. They found that the calcium influx generated by the action potential occurs during the repolarizing phase of the action potential when the calcium conductance is large and the driving force increases (tail currents). Thereafter, they compared the number of calcium channels that open by the action potential, with the number of transmitter quanta that were liberated, and reach the conclusion that ~60 calcium channels have to open to cause the release of a quantum of transmitter. In principle, it can be argued that some of the calcium channels open in the presynaptic terminal at locations not associated with transmitter release. They address this argument by using "slow" and "fast" calcium buffers. The result is that even slow calcium buffers like EGTA are able to suppress quantal transmitter release, indicating that calcium ions have to travel a substantial distance before they trigger transmitter release. This distance is much larger than envisaged for a single, high calcium concentration domain (6, 484, 791, 842, 1012, 1013). Hence it appears that although a single calcium channel is able occasionally to trigger transmitter release, under normal circumstances at the synapse of Held, many calcium channels are involved in the liberation of a single transmitter quantum. This number of channels-to-number of quanta ratio is also very important in understanding the modulation and plasticity of synaptic transmission. If this ratio is always one, it is much more difficult to envisage the calcium channel cooperativity.

H.  Other Routes for Calcium Entry Into the Nerve Terminal

1.  Channels

A) NONSELECTIVE CHANNEL. One of the channels identified in the preparation of fused presynaptic nerve terminals of Torpedo electromotor nerve is a large, calcium-permeable and highly voltage-dependent ion channel (536). Single-channel conductance is on the average 846 pS. At voltages below 0 mV (inside negative), the probability of the channel to open is negligible but increases dramatically, within a very narrow voltage range, to >50% at 8 mV. In short pulse experiments from holding potential of 70 mV, it was found that the channel is activated shortly after depolarization and deactivated shortly after the membrane voltage returns to the resting potential. Therefore, it may be activated and deactivated by a typical action potential invading the nerve terminal.