Analysis of Repolarization of Presynaptic Motor Terminals in Drosophila Larvae Using Potassium-Channel-Blocking Drugs and Mutations

The capability of combining genetic, electrophysiological and molecular techniques has made Drosophila melanogaster one of the best experimental systems for studying ion channels and membrane excitability. This multidisciplinary approach has permitted the electrophysiological and pharmacological characterization of particular ion currents, the identification of the genes that specify these currents, and molecular analysis of the encoded proteins (for reviews see Wu and Ganetzky, 1988; Papazian et al. 1988). These currents, which determine the electrical properties of excitable cells, have been studied in vitro and in vivo in muscle cells, photoreceptor cells and neuronal cell bodies (Salkoff, 1985; Gho and Mallart, 1986; Sole and Aldrich, 1988; Hardie, 1991). However, little is known about the currents present in other functional regions of neurons in Drosophila. For example, at the presynaptic terminal, the mechanisms underlying depolarization and repolarization are of fundamental importance in regulating transmitter release but little is known about the potassium currents that contribute to repolarization. Furthermore, little is known about the mechanisms underlying the synaptic plasticity observed at the presynaptic terminal in Drosophila larvae (Zhong and Wu, 1991a). It is known that modification of potassium activity, via several different second messenger systems, has important consequences for synaptic modulation and plasticity (Kandel and Schwartz, 1982; Alkon, 1984). Direct study of the presynaptic terminals in Drosophila has been difficult because of their small size. Here, as a first approach to investigating the repolarization mechanism of larval presynaptic terminals, we investigate the effects of drugs and mutations known to block potassium currents on the time course of transmitter release at the neuromuscular junction.

At least four distinct potassium currents have been described in Drosophila muscles and neurons (Salkoff, 1985; Wu and Haugland, 1985; Gho and Mallart, 1986; Wei and Salkoff, 1986; Sole and Aldrich, 1988; Saito and Wu, 1990). These include two fast, transient currents, I_A and I_CF. I_A is voltage-dependent and similar to the molluscan A current (Connors and Stevens, 1971; Neher, 1971), whereas I_CF is activated by calcium and is similar to other calcium-dependent currents previously described in muscles and neurons (Mounier and Vassort, 1975; Yamamoto and Washio, 1981; MacDermott and Weight, 1982). In addition, there are two slow, sustained potassium currents, I_K and I_cs. I_K corresponds to the classical non-inactivating, voltage-dependent delayed rectifier first described in the squid axon (Hodgkin and Huxley, 1952). I_cs is a calcium-dependent current similar to that originally described in molluscan neurons (Meech and Standen, 1975). These currents are selectively affected by several drugs and mutations in different genes. For instance, aminopyridines as well as Sh mutations block I_A in muscles and some neurons (Salkoff and Wyman, 1981; Salkoff, 1983; Sole et al. 1987; Baker and Salkoff, 1990; Saito and Wu, 1991). In other neurons, I_A is mediated by a distinct type of channel sensitive to aminopyridines but not affected by Sh mutations (Sole et al. 1987; Baker and Salkoff, 1990). I_CF is eliminated by slo mutations (Elkins et al. 1986; Singh and Wu, 1989; Komatsu et al. 1990) and quinidine selectively affects I_K (Singh and Wu, 1989; Hardie, 1991). In addition, calcium blockers or low calcium concentration in the external solution affect the calcium-activated currents Ic_F and I_cs.

Previous studies have inferred the presence of potassium currents in the presynaptic motor nerve terminal by the phenotypic effects of mutations on the postsynaptic potential, referred to as an excitatory junctional potential (EJP). For example, the presence of presynaptic potassium channels encoded by the Sh locus has been indicated by the enlarged prolonged EJPs observed in Sh mutants, consistent with a failure of normal repolarization of the presynaptic terminal (Jan et al. 1977). Similarly, a potassium current specified by the slo locus, a structural gene for calcium-activated potassium channels (Atkinson et al. 1991), may also be present at motor terminals because abnormally prolonged postsynaptic potentials are also seen in slo mutants (B. Ganetzky, unpublished results). However, because slo is important for proper repolarization of muscles (Elkins and Ganetzky, 1988; Singh and Wu, 1990), it is not clear whether the effect of slo on the postsynaptic potential is pre-or postsynaptic. Thus, alteration of the EJP by a particular mutation or drug does not always provide sufficient evidence that the potassium channels affected by these agents are found in the presynaptic membrane and steps must be taken to eliminate the contribution of postsynaptic conductances to the duration or time course of the synaptic current. We have circumvented this problem in these experiments by recording the excitatory junctional current (EJC) while keeping the muscle held at a constant membrane potential under voltage-clamp conditions.

At the neuromuscular junction, transmitter release is triggered by an influx of calcium ions into the presynaptic terminal (Katz and Miledi, 1965, 1969; Dodge and Rahamimoff, 1967; Llinás and Nicholson, 1975) and is not initiated until the repolarizing phase at the end of the presynaptic action potential (Llinás et al. 1981, 1982). Prior to this, the presynaptic terminal is depolarized to a value near the calcium equilibrium potential and there is no net calcium entry. The delay before release of neurotransmitter and the resultant synaptic current in the postsynaptic cell are therefore primarily dependent upon the kinetics of repolarization under the control of potassium channels (Katz and Miledi, 1967a,b; Benoit and Mambrini, 1970; Datyner and Gage, 1980; Mallart et al. 1991). By measuring the EJC in larval muscle cells, we were able to obtain quantitative measurements of the delay before transmitter release and the onset of this response. These measurements provide information about the time before onset of the repolarizing phase of the presynaptic action potential, which in turn depends upon activation of potassium currents. Using this experimental paradigm, we examined the effects on evoked synaptic currents of several different potassium-channel-blocking mutations and drugs. Our data suggest that at least four potassium currents contribute to the repolarization of motor nerve terminals. These results provide a foundation for further genetic and molecular analyses of the molecular mechanisms involved in the regulation of transmitter release in Drosophila.

Experiments were performed on body-wall muscle 6 (for nomenclature see Johansen et al. 1989) in the second and third abdominal segments of D. metanogaster third-instar larvae. The larval preparation was similar to that previously described by Jan and Jan (1976). Previous measurements have shown that the larval muscles are essentially isopotential (Wu and Haugland, 1985). Wild-type flies were of the Canton-S strain. The mutations used were Sh^KSJ33, slo¹ and eag¹, which were all raised under standard laboratory conditions.

Synaptic currents were recorded using a two-electrode voltage-clamp. By use of the voltage-clamp to measure the synaptic currents, we avoided the problem of non-linear summation normally observed when the synaptic potential is used as a measure of synaptic activity (Martin, 1955; Augustine et al. 1985). Muscle fibers were impaled with microelectrodes of about 5–10 MΩ resistance filled with 3 mol l⁻¹ KC1. The current electrode was positioned at the center of the fiber. Unless otherwise indicated, the holding potential was –60 mV.

Synaptic current records were filtered at 3 kHz, digitized and recorded on a VCR system. Off-line analyses were performed using pClamp software (Axon Instruments). The principal parameters determined were (1) the onset slope, measured as the slope of the line approximating the onset phase of the synaptic current at its midpoint and (2) the synaptic delay, measured as the intersection between the baseline and a line representing the onset slope of the synaptic current drawn as described above.

Only those cells that gave reproducible responses in at least five successive trials were used for data collection.

The segmental nerves innervating the muscles were severed near the ventral ganglion. To obtain electrotonic depolarization of the terminal, 1μmoll⁻¹ tetrodotoxin (TTX) was added to the bathing solution to block action potentials and the nerve was stimulated close to the terminal. Unless otherwise stated, electrotonic depolarizing current pulses were 50 or 200 ms in duration and three times the threshold for evoking synaptic current (about 4–5 V). Under these conditions, the synaptic currents elicited always corresponded to the maximal response. Synaptic responses could not be evoked by current pulses of similar strength but inverted polarity.

The standard saline was (in mmol l⁻¹): NaCl, 128; KC1, 2; MgCl₂, 4; CaCl₂, 5.4; Hepes, 5; sucrose, 353; pH 7.1. Temperature was maintained at 6°C using a Peltier device (Cambion). The high concentration of sucrose in this saline prevented the muscle contractions that otherwise occurred at the external calcium concentrations used in these experiments. Hypertonic solutions have previously been used with insect muscle preparations for this purpose (Yamamoto and Washio, 1981; Ashcroft and Stanfield, 1982; Gho and Mallart, 1986). This procedure allowed us to maintain stable cell recordings for at least 10 min without any observed changes in response.

The following drugs were used to block various ion channels: 3,4-diaminopyridine (DAP, Sigma); quinidine hydrochloride monohydrate (Sigma), tetraethyl-ammonium chloride (TEA⁺, Sigma) and tetrodotoxin (TTX, Sigma).

Previous studies of Drosophila have shown that prolonged synaptic responses are evoked in the presence of agents affecting potassium currents. These changes could result from effects on (1) the action potentials arriving at the terminal, (2) the postsynaptic membrane or (3) the presynaptic terminal itself. We used experimental conditions that enabled us to discriminate effects on the presynaptic terminal from the other possibilities. To eliminate the possibility that effects on synaptic response resulted from abnormal axonal action potentials arriving at the synaptic terminal, we examined synaptic responses when action potentials were blocked by the addition of 1μmoll⁻¹ TTX. Under these conditions, transmitter release was evoked by direct electrotonic depolarization of the presynatic terminal by application of a sufficiently large current pulse via the suction electrode (Katz and Miledi, 1969; Jan et al. 1977). Electrotonic depolarization of the terminal was achieved using stimuli of long duration (50–200 ms). The electrotonic nature of the depolarization under these conditions has been shown by the occurrence of graded responses that increase progressively to the maximal response as the stimulus intensity is gradually increased (Wu et al. 1978). Such graded responses are never observed under conditions of active nerve conduction.

To prevent any possible contribution of the postsynaptic cell to the evoked synaptic responses, the muscle cell was held at a constant voltage. Thus, any voltage-dependent conductances in the muscle activated during the synaptic response would be eliminated. In addition there is the possibility that a muscle current could be activated by the influx of some ion carried by the synaptic current. In this case, changes in the synaptic response could result from the blockage of these ion-activated postsynaptic currents by mutations or drugs. If this were occurring, the synaptic current should also change in the absence of blockers when the holding potential was equal to the equilibrium potential of the ion carried by this putative outward current. However, we observed that the shape of the synaptic current was invariant across the entire range of holding potentials from – 100 to –30mV (not shown). This eliminates the possibility that activation of non-voltage-dependent conductances in the muscle contributes to the synaptic responses recorded.

Because the onset of transmitter release occurs during the repolarizing phase of the nerve terminal, measurements of synaptic latency can be used to compare the effects of different drugs and mutations on the repolarization of the nerve terminal following electrotonic stimulation. Because current pulses of long duration (50–200 ms) were used to depolarize the terminal electrotonically, if the terminal were to repolarize with a normal time course, the resting potential should be restored prior to the end of the current pulse and synaptic current should be initiated before the end of the stimulus. However, if repolarizing currents were reduced, the onset of the synaptic current should be delayed.

In these experiments, the intensity of the stimulus pulse was always about three times higher than the threshold required to elicit a synaptic response. This stimulus strength evoked the maximal response; further increases in stimulus strength had no additional effect on the synaptic current.

The synaptic current evoked by a current pulse of long duration in wild-type muscle began 9 ms after the stimulus had been applied, peaked after about 20 ms and was followed by a slow decay. The rising phase of the synaptic current (onset slope) was about lOnAms¹ (Fig. 1). Neither the slo mutation, which eliminates I_CF in muscle and neuronal cell bodies (Elkins et al. 1986; Singh and Wu, 1989; Komatsu et al. 1990; Saito and Wu, 1991), nor the application to wild-type muscle of DAP, which selectively eliminates I_A in muscles and neuronal cell bodies (Salkoff and Wyman, 1981; Sole and Aldrich, 1988; Saito and Wu, 1991), had any significant effect on the synaptic delay or onset slope compared with wild-type muscle (Figs 1A, 3). However, when DAP (20μmoll⁻¹) was applied to slo larvae, the synaptic delay increased to 17 ms and the slope of the rising phase of the synaptic current (onset slope) was reduced to 3 nA ms⁻¹. Similar results were obtained using a Sh mutation, which eliminates I_A, at least in muscles (Salkoff and Wyman, 1981; Wu and Haugland, 1985), instead of DAP (Figs 1B, 3). No effects on synaptic delay were observed after application of DAP to Sh mutants (not shown), suggesting that in these experiments DAP is acting on the same set of channels altered by Sh mutations.

An off-response, recognized as a second surge of synaptic current at the end of the pulse, was clearly observed when the slo mutation was combined with the application of DAP or the presence of the Sh mutation (Fig. 1). This off-response was even more pronounced when additional potassium channels were blocked (cf. Figs 2 and 5). The fact that significant changes in synaptic latency and onset slope of the synaptic current were observed only after the combined action of DAP (or Sh) and slo indicates that the repolarizing current of the motor nerve terminal contains at least two components: one is blocked by DAP or Sh and the other is blocked by slo. Apparently, under the experimental conditions used, either one of these currents alone is sufficient to repolarize the terminal.

The existence at the presynaptic terminal of additional repolarizing components other than those affected by DAP or slo was tested by examining whether application of quinidine and TEA⁺ had any additional effect on synaptic delay beyond those caused by application of DAP to slo (Fig. 2). When quinidine (0.1 mmol l⁻¹), which selectively blocks I_K in muscle and photoreceptors cells (Singh and Wu, 1989; Hardie, 1991), was applied together with DAP to slo, the delay before transmitter release was further increased to 39 ms (Figs 2, 3A) and the slope of the onset phase was reduced further to 0.5 nA ms⁻¹ (Figs 2, 3B). These results indicate that quinidine blocks an additional repolarizing component in the terminal that is different from those affected by DAP and slo.

In the experiments with quinidine, there was considerable variability between larvae and also between segments within a single larva in the synaptic latency and the onset slope of the synaptic current. In some cells, the effect of quinidine on synaptic currents resembled the more extreme results obtained only after the further addition of TEA⁺ (see below). This variability was not the result of a slow action of quinidine because the synaptic delay and the onset slope remained fairly constant during long (approximately 20 min) recordings from the same cell (not shown). Instead, the variability may indicate that some class of TEA⁺-sensitive potassium channels is absent or nonfunctional in a few terminals; depending on the presence or absence of this class of channels in a particular terminal, the addition of TEA⁺ would or would not be required to block repolarization fully (see below).

When TEA⁺, a more general blocker of potassium currents, was applied together with quinidine and DAP to slo mutants, the synaptic current was never initiated until the end of the current pulse. Thus, the synaptic current was always observed as an off-response (Fig. 2); the synaptic delay was the same as the duration of the stimulus and the onset slope was zero. This additional effect of TEA⁺ on the synaptic current suggests that at least one other repolarizing potassium current not affected by quinidine, DAP or slo is present at the terminal. The fact that the synaptic current was always observed at the end of the stimulus pulse in the presence of a generalized potassium channel blocker, such as TEA⁺, suggests that the intrinsic repolarization mechanisms of these terminals were virtually eliminated under these conditions. Thus, it is unlikely that any other non-potassium current, such as a chloride current, participates significantly in repolarization of the presynaptic motor terminal.

Because the eag mutation has previously been shown to reduce the sustained potassium currents in muscle (Wu et al. 1983; Zhong and Wu, 1991b), we tested whether this mutation also had effects on the synaptic terminal. Synaptic delay was not further increased by the combination of eag with slo plus DAP (not shown). Furthermore, synaptic delay was increased by the application of quinidine or TEA⁺ to DAP-treated eag;slo larvae (not shown), suggesting that eag did not substantially diminish the quinidine-or TEA⁺-sensitive currents of these terminals. Because the effect of eag on muscle is more extreme at 20°C than at 6°C (C.-F. Wu, personal communication), we repeated these experiments at a higher temperature (21°C), but we still failed to observe an effect of eag on the synaptic current.

As shown in Fig. 4A, when presynaptic terminals were stimulated by short electrotonic stimuli, varying the stimulus strength over a narrow range could produce substantial differences in the synaptic current. The relationship between synaptic currents and stimulus intensity is plotted in Fig. 4B for slo larvae treated with DAP. A sudden jump in the amplitude of the synaptic current was observed when the stimulus was increased above a certain threshold. A similar result was observed in wild-type larvae (not shown). Stimuli lower than this threshold could also evoke these explosive synaptic responses if the pulse duration was increased. Subthreshold graded release, sometimes followed by a full-blown response (Fig. 4A), was observed only in the Sh;slo double mutant or after application of DAP to slo larvae, suggesting that, in the wild type, potassium currents prevent the presynaptic terminal from depolarizing gradually.

This explosive synaptic current is probably triggered by an all-or-nothing regenerative potential at the presynaptic terminal. Because 1TX was present in the bath, this presynaptic action potential was probably sustained by the activation of an inward calcium current. The observation that graded responses could be evoked by lower-strength stimulation indicates that the depolarization allows the internal calcium concentration to increase sufficiently to trigger synaptic release, even before the threshold for triggering the presynaptic action potential is reached.

The voltage-dependent inactivation of the presynaptic inward current could be examined when the normal repolarization mechanism of the terminal was completely blocked by application of DAP, quinidine and TEA⁺ to slo larvae. Terminals were depolarized with strong electrotonic stimuli (three times above threshold) for 0.2–3 s. Under these conditions, the synaptic response was always observed at the end of the pulse as an off-response (15 of 15 muscle fibers studied). Superimposed traces of off-responses obtained with 200 ms and 3 s stimulus pulses are shown in Fig. 5. In this particular example, the off-response after a 3 s stimulus was not diminished in amplitude compared with the off-response after a 200 ms stimulus. In all the other cases examined, the synaptic currents evoked by the 3 s stimulus were never reduced by more than 30% compared with those elicited by the 200 ms stimulus.

The results presented here indicate that mutations and drugs, which have known effects on potassium currents in muscles and neuronal cell bodies of Drosophila, also affect repolarization of presynaptic motor terminals. The failure of the terminals to repolarize properly was observed as an increase in the delay and a decrease in the onset slope of the evoked synaptic current recorded from the muscle.

We believe that under our experimental conditions the effects of the drugs and mutations on synaptic release result from an action on the terminal itself for the following reasons. (1) Any possible contribution of voltage-dependent muscle currents to the synaptic response recorded was eliminated by holding the membrane potential of the muscle constant. (2) The effects were observed even when the muscle was clamped at different holding potentials, suggesting that contributions to the synaptic response from any non-voltage-dependent post-synaptic current are also unlikely. (3) The presynaptic terminal was depolarized directly by electrotonic stimulation using suprathreshold stimulation in the presence of TTX to block axonal conduction. Because, under these conditions, the synaptic currents evoked represent the maximal response, the possibility that changes in the passive properties of the axonal membrane contribute to the modification of the synaptic responses is unlikely. (4) To our knowledge, there have been no reports demonstrating any direct actions of these drugs and mutations on the mechanism of neurotransmitter release itself or on the properties of postsynaptic receptors (Augustine, 1990). Thus, the most likely explanation of our results is a selective effect on potassium currents in the presynaptic terminal.

Among the useful parameters that could be reliably quantified in these experiments were the synaptic latency and the slope of the onset phase of the synaptic current. Each of these varied in stepwise fashion when the drugs and mutations were added progressively (Fig. 3). The synaptic latency depends directly upon the time course of repolarization of the presynaptic terminal (Katz and Miledi, 1967a,b; Benoit and Mambrini, 1970; Augustine, 1990; Mallart et al. 1991). Thus, it is expected that the synaptic latency will increase progressively as different components of the repolarizing current at the terminal are blocked or eliminated.

The decreasing slope of the onset phase of synaptic current observed as the various potassium currents were progressively blocked is more difficult to interpret. One possibility is that the slope is correlated with the rate of the falling phase of the presynaptic action potential. When the terminal is electrotonically depolarized by long-duration current injection, repolarization of the presynaptic terminal will be slower when some of the repolarizing currents are blocked. Since, in our conditions, calcium channels show only little, if any, voltage inactivation during this time (see below), the entry of calcium into the terminal probably follows the driving force determined by the membrane potential. Partial blockage of potassium currents at the terminal could thus slow the falling phase of the action potential, leading in turn to a low rate of calcium entry and slow release of neurotransmitter.

In muscle fibers of Drosophila larvae, morphological studies have shown two classes of synaptic boutons (Johansen et al. 1989) and electrophysiological studies have shown two components of the evoked EJP (Jan and Jan, 1976) suggesting that the fibers are innervated by at least two different axons. Similar polyinnervation of larval muscle has been found in other dipterans (Hardie, 1976). Thus, one could envisage that the various repolarizing currents described in the present study could be differentially distributed among different terminals. Focal studies of synaptic terminals will be necessary to examine this possibility.

Because of the small size of the terminals in Drosophila, direct study of ionic currents responsible for repolarization of the presynaptic terminal has not yet been possible. However, our results provide some insight about the number and properties of the potassium currents present at the terminal. Repolarization of the presynaptic terminal and release of neurotransmitter are known to depend on potassium currents (Katz and Miledi, 1967a,b; Benoit and Mambrini, 1970; Datyner and Gage, 1980; Augustine, 1990; Jackson et al. 1991; Mallart et al. 1991). Therefore, if the drugs or mutation used alter the time course of synaptic release, we can infer the presence in the presynaptic terminal of potassium currents affected by these agents. The observation that the synaptic response changed in a stepwise fashion when the different drugs or mutations were introduced one at a time suggests that at least four distinct components contribute to the repolarization of presynaptic motor terminals of Drosophila larvae. One component is affected by Sh or DAP. The observation that DAP and Sh produce identical effects on synaptic release indicates that DAP is not affecting any class of potassium channels at the presynaptic terminal not affected by Sh and vice versa. A second component is affected by mutations of slo, which is a structural gene for a subunit of calcium-activated potassium channels present in muscle and neuronal cell bodies (Atkinson et al. 1991), and a third is affected by quinidine. The additional effects of TEA⁺ beyond those caused by these agents suggest the presence of at least one other TEA⁺-sensitive component. An alternative possibility is that the effects of TEA⁺ are caused by a more complete blockage of the same channels that were only partially blocked by DAP or quinidine. However, we think this possibility is less likely because in other Drosophila neurons I_A and I_K are blocked completely by the concentrations of DAP and quinidine used here. Although direct recording from synaptic terminals will be necessary to match with certainty a particular current with each of these components and to characterize their properties in detail, it is worth noting that the genetic and pharmacological sensitivities of the repolarizing currents in the terminal parallel those of the potassium currents identified by voltage-clamp analysis of larval and adult muscle. Thus, the effects of Sh and DAP on repolarization of the terminal suggest the presence in the terminal of a current that may resemble I_A; the effects of slo suggest the presence of an inactivating calcium-dependent potassium current (I_CF-like). Since quinidine selectively blocks I_K in both muscle and photoreceptor cells at the concentration used here (Singh and Wu, 1989; Hardie, 1991), the effect of quinidine on synaptic release suggests the presence of a delayed rectifier current (I_K-like). We cannot rule out the possibility that quinidine has some additional effect on the presynaptic terminal that alters synaptic release by a mechanism not involving potassium channels. Nonetheless, this possibility is unlikely to be a significant factor in our experiments because the application of quinidine alone in the absence of other potassium channel blockers did not have any observed effect on synaptic release (Wu and Ganetzky, 1988; M. Gho and B. Ganetzky, unpublished observations) and at the concentrations used here no effects of quinidine in Drosophila other than the selective blockage of IK have been reported (Singh and Wu, 1989; Hardie, 1991). The additional action of TEA⁺ indicates the presence of at least one other potassium current, perhaps the non-inactivating calcium-dependent current (Ics-like).

Despite the possible similarity between potassium currents in muscles and presynaptic terminals suggested above, it is still necessary to interpret our results with caution because our analysis is indirect. For example, drugs such as DAP could have different pharmacological effects in muscles and motor terminals. Similarly, although mutations such as Sh and slo may affect potassium channels in both synaptic terminals and muscle, the channels in each location could represent alternative gene products encoded by the same gene with distinct functional properties. In addition, a differential distribution of some subunits encoded by other genes could lead to the assembly of potassium channels with different properties in the two regions. Thus, the biophysical properties of the potassium channels and their molecular composition may not be identical in muscle and presynaptic terminals. In view of such considerations and the variety of potassium channel genes identified in Drosophila (Butler et al. 1989), it is somewhat surprising that the genetic and pharmacological effects on potassium currents in muscles and presynaptic terminals are as comparable as they appear from this analysis. Ultimately, more detailed biophysical and molecular studies of all the potassium channel subunits present in muscle and presynaptic terminals will be necessary for a definitive comparison to be made.

In contrast to the effects of Sh and slo, which alter repolarization of both muscle and presynaptic terminals, eag did not appear to have any pronounced effect on repolarization of the motor terminal even though muscle potassium currents are altered in eag mutants (Wu et al. 1983; Zhong and Wu, 1991b) and previous evidence suggests that eag affects the motor terminal (Ganetzky and Wu, 1982, 1983). Furthermore, synaptic transmission in eag;Sh double mutants is greatly enhanced beyond that of Sh mutants alone. The synergistic interaction between these two mutations has been interpreted as the result of a combined deficit in repolarization of the synaptic terminal affecting potassium currents other than I_A (Ganetzky and Wu, 1982). Consistent with this interpretation is the observation of very large prolonged EJPs, resembling those in eag;Sh double mutants, after application of quinidine to Sh larvae or larvae treated with 4-aminopyridine (Wu et al. 1989). These data suggest that, as in muscles, a current that is affected either by eag or by quinidine is present in the presynaptic motor terminal. We are as yet unable to reconcile this conclusion with our failure to observe effects of eag on repolarization of the presynaptic terminal in these experiments. However, all of the earlier EJP analysis was performed in a low-calcium external solution, where the efficacy of calcium-dependent potassium currents in repolarizing the presynaptic terminal is minimal. The lack of effect of eag on synaptic release under our conditions could be because a partial reduction in one or more potassium currents is not sufficient to change the rate of repolarization of the synaptic terminal if some other potassium current similar to l_cs remains fully functional. It will be necessary to await procedures to block selectively any remaining potassium currents to test this possibility.

In addition to the inferences about potassium currents, these experiments enabled us to derive some new insights about the presynaptic inward current. The explosive synaptic responses observed are probably due to the activation of a calcium current at the presynaptic terminal. The surge of synaptic current at the end of the stimulus (off-response) is probably the result of a presynaptic calcium tail current that occurs at the end of the stimulus (Katz and Miledi, 1967b). When all of the presynaptic potassium currents were blocked with DAP, quinidine, TEA⁺ and slo, the synaptic current was always observed as an off-response at the end of the electrotonically applied current pulse. Under these conditions, the electrotonic stimulus is apparently sufficient to depolarize the presynaptic terminal to near the calcium equilibrium potential such that, during the pulse, there is no calcium entry and transmitter release is prevented. However, at the end of the stimulus pulse, the calcium driving force again favors calcium entry and the synaptic current is immediately restored (Katz and Miledi, 1971). This off-response was observed even when the terminal was fully depolarized for as long as 3 s by stimulus pulses of long duration. This result suggests that voltage-dependent inactivation of the presynaptic calcium current during prolonged depolarization of the motor terminal is only slight, if it occurs at all. Barium currents recorded from cultured embryonic neurons of Drosophila have indicated the presence in this cell population of inactivating as well as non-inactivating calcium channels (Leung and Byerly, 1991), although barium itself can exert pronounced effects on the inactivation of calcium channels. Calcium currents not displaying voltage-dependent inactivation have been described in other excitable cells (Katz and Miledi, 1971; Keynes et al. 1973; Llinás et al. 1976) including insect muscle fibers (Ashcroft and Stanfield, 1982). We cannot rule out the possibility that the off-response results from calcium channels in larval motor terminals undergoing an obligatory transition through an open state during the recovery process after inactivation, as has been described for calcium channels in mouse cerebellar neurons (Slesinger and Lansman, 1991) and I_A channels (Demo and Yellen, 1991). Nevertheless, the observation that the probability of calcium channels reopening during the transition from inactivation to rest states increases with long periods of inactivation (Slesinger and Lansman, 1991) argues against this explanation of our results, because we never observed an increase (and in some cases even observed a decrease) in the off-response when the stimulus duration was increased from 10 ms to 3 s.

The presence of several potassium currents in motor terminals of Drosophila is reminiscent of observations on synaptic terminals in other organisms. For example, it has been shown that transient and sustained voltage-dependent or calcium-dependent potassium currents are present in motor terminals of frog (Mallart, 1984), lizard (Benoit et al. 1989; Lindgren and Moore, 1989; Morita and Barrett, 1990) and mouse (Tabti et al. 1989). The diversity of potassium currents has been interpreted as a general feature of presynaptic terminals. This diversity attests to the existence of an elaborate mechanism for repolarization which, in turn, controls the amount and time course of transmitter release. In Drosophila, regulatory mechanisms that modulate the strength of one or more of these currents could underlie the plasticity that has been observed at this synapse (Stern and Ganetzky, 1989; Zhong and Wu, 1991a). The results of this study provide the basis for further elucidation of these regulatory mechanisms as well as for the analysis of other mutations whose effects on synaptic transmission are not yet understood.

We are grateful to G. Robertson, K. A. Schlimgen, M. Stern and C.-F. Wu for helpful comments on the manuscript. This research was supported by grants from the National Institutes of Health (NS 15390) and the Markey Charitable Trust and fellowships from the Klingenstein and McKnight Foundations. M.G. was also partially supported by the Fondation pour la Recherche Médicale. This is paper number 3189 from the Laboratory of Genetics.