The understanding of neurotransmitter release at vertebrate synapses has been hampered by the paucity of preparations in which presynaptic ionic currents and postsynaptic responses can be monitored directly. We used cultured embryonic Xenopusneuromuscular junctions and simultaneous pre- and postsynaptic patch-clamp current-recording procedures to identify the major presynaptic conductances underlying the initiation of neurotransmitter release. Step depolarizations and action potential waveforms elicited Na and K currents along with Ca and Ca-activated K (K Ca) currents. The onset ofK Ca current preceded the peak of the action potential. The predominantly ω-CgTX GVIA-sensitive Ca current occurred primarily during the falling phase, but there was also significant Ca2+ entry during the rising phase of the action potential. The postsynaptic current began a mean of 0.7 msec after the time of maximum rate of rise of the Ca current. ω-CgTX also blocked K Ca currents and transmitter release during an action potential, suggesting that Ca andK Ca channels are colocalized at presynaptic active zones. In double-ramp voltage-clamp experiments,K Ca channel activation is enhanced during the second ramp. The 1 msec time constant of decay of enhancement with increasing interpulse interval may reflect the time course of either the deactivation of K Ca channels or the diffusion/removal of Ca2+ from sites of neurotransmitter release after an action potential.
Neurotransmitter release from nerve terminals is triggered by the entry of Ca2+ through voltage-gated Ca channels (Katz, 1969; Augustine et al., 1987). Our understanding of the relationship between the presynaptic ionic currents and release is based largely on studies of the squid giant synapse (Katz and Miledi, 1967; Llinás et al., 1981; Charlton et al., 1982; Augustine et al., 1985a,b). Several critical questions remain, however, that one would like to address with equivalent biophysical rigor at a vertebrate synapse in which the presynaptic ionic currents and postsynaptic currents can be measured simultaneously and release can be resolved at the single quantum level. In this report we take advantage of a neuromuscular synapse preparation in which this can be done.
Among the important pending questions are the timing and delay between Ca2+ entry during an action potential and release, the roles of different Ca and Ca-activated K (K Ca) channels in the release process, and the quantitative relationship between Ca2+ influx and release. In squid, it has been shown that Ca2+ enters principally during the repolarization phase of the action potential (Llinás et al., 1982). Physiological studies of various excitable secretory cells have shown that different types of Ca channels play a dominant role in triggering release in different terminals, and often multiple Ca channel types are present in any given terminal (Kerr and Yoshikami, 1984; Pfrieger et al., 1992, Luebke et al., 1993; Artalejo et al., 1994; Dunlap et al., 1994; Fu and Huang, 1994; Regehr and Mintz, 1994;Yawo and Chuhma, 1994; Mintz et al., 1995; Sivaramakrishnan and Laurent, 1995; Wheeler et al., 1996). In many nerve terminals,K Ca channels are present, and in some cases they seem to be colocalized with Ca channels, potentially with important functional consequences (Augustine and Eckert, 1982; Lancaster and Nicoll, 1987; Lindgren and Moore, 1989; Roberts et al., 1990;Robitaille and Charlton, 1992; Robitaille et al., 1993; 1994; Blundon et al., 1995; Wheeler et al., 1996). Under different conditions and in different synapses, the relationship between external Ca2+concentration or measured Ca2+ influx and release is a power function with an exponent ranging from 1 to 5 (Dodge and Rahamimoff, 1967; Katz and Miledi, 1970; Llinas et al., 1981; Cohen and Van der Kloot, 1985; Augustine and Charlton, 1986; Stanley, 1986;Augustine, 1990; Borst and Sakmann, 1996; Takahashi et al., 1996).
Intense interest in these problems has stimulated attempts to obtain answers in a number of vertebrate preparations (Lim et al., 1990;Stanley and Goping, 1991; Stanley, 1993; Yawo and Momiyama, 1993;Artalejo et al., 1994; Heidelberger et al., 1994; Borst et al., 1995;Sivaramakrishnan and Laurent, 1995; Borst and Sakmann, 1996; Takahashi et al., 1996). In this paper, we describe results from a novel preparation that offers the accessibility and degree of control of pre- and postsynaptic currents that are needed to obtain answers to many of these questions.
Here we describe experiments using simultaneous pre- and postsynaptic voltage clamp in Xenopus nerve–muscle cocultures and characterize the ionic currents of the presynaptic varicosities. Previous work on this preparation measured Na currents (Kidokoro and Sand, 1989) and Ca currents (Hulsizer et al., 1991; Meriney et al., 1991) in varicosities. In this report, we emphasize the currents carried by the N-type Ca channels and K Cachannels that are functionally coactivated in presynaptic varicosities and coupled to transmitter release.
MATERIALS AND METHODS
Cell culture. Nerve–muscle cocultures were prepared on the basis of methods described previously (Spitzer and Lamborghini, 1976; Tabti and Poo, 1991). In brief, stage 20–22 Xenopus laevis embryos (Niewkoop and Faber, 1967) were rinsed in sterile 10% normal frog Ringer’s solution (NFR) (116 mm NaCl, 1 mm NaHCO3, 2 mm KCl, 1.8 mm CaCl2, 1 mm MgCl2, 5 mm HEPES, 3 mm d-glucose, pH 7.3), and the spinal cord and associated myotomes were dissected away and allowed to dissociate in a Ca2+- and Mg2+-free Ringer’s solution (125 mm NaCl, 2 mm KCl, 1.2 mm EDTA, and 5 mm Na-HEPES, pH 7.4) for 30–60 min. Disaggregated cells were then plated onto glass coverslips or plastic tissue-culture dishes and incubated at room temperature (22–24°C) for 24–48 hr in a medium composed of 49% NFR and 51% L-15 (Life Technologies, Gaithersburg, MD) supplemented with 3 mg/ml glutamine, 0.1 mg/ml insulin, 0.7 mg/ml sodium selenite, 0.6 mg/ml transferrin, 1 mg/ml sodium pyruvate, and 35 ng/ml brain-derived neurotrophic factor [kindly provided by Amgen (Thousand Oaks, CA) and Regeneron (Tarrytown, NY)]. In as little as 12 hr, spinal neurons extended neurites, occasionally elaborating varicose regions on the substrate and in contact with spindle-shaped muscle cells (Kidokoro and Yeh, 1982; Takahashi et al., 1987; Evers et al., 1989; Tabti and Poo, 1994). In 1 and 2 d cultures, the presynaptic contact often takes the form of a varicosity sufficiently large to be accessed with patch electrodes, allowing one to correlate presynaptic electrophysiological events with transmitter release. These synapses exhibit properties–both physiological and morphological–that appear to parallel closely those of their developing counterparts in vivo (Kullberg et al., 1977). Within hours of contact, they resemble mature cholinergic synapses in many respects. They exhibit spontaneous release, high quantal content-evoked release, membrane thickenings, clouds of vesicles, and postsynaptic aggregation of ACh receptors (Anderson et al., 1977; Weldon and Cohen, 1979; Cohen and Weldon, 1980; Kidokoro et al., 1980; Kidokoro and Yeh, 1982; Brehm et al., 1984; Takahashi et al., 1987; Buchanan et al., 1989; Evers et al., 1989).
Electrophysiology. Traditional whole-cell recording techniques were used to record voltages and currents from neuronal somata and muscle cells. Patch electrodes of 2–4 MΩ were filled with a quasi-internal solution (internal solution A) of the following composition (in mm): 116 KCl, 1 NaCl, 1 MgCl2, 10 EGTA, 5 HEPES, pH 7.3. Because the Ca current at presynaptic varicosities is labile, we adopted a modification of the perforated patch recording configuration (Horn and Marty, 1988) for use in recording at the varicosity. For current clamp and in experiments designed to measure both K and Ca currents, the presynaptic pipette was filled with internal solution B (in mm): 52 K2SO4, 38 KCl, 1 EGTA, and 5 HEPES, pH 7.3, plus 900 μg/ml amphotericin B (Rae et al., 1991). Except where noted, for experiments designed to isolate Ca currents, the pipette was filled with internal solution C (in mm): 52 CsMeSO4, 38 CsCl, 1 EGTA, 1 3,4 diaminopyridine (3,4 DAP), 50d-glucose, and 5 HEPES, pH 7.3, plus 900 μg/ml amphotericin B. Similarly, the bath solution for all experiments was NFR, unless noted otherwise. Voltage-clamp depolarization waveforms (pCLAMP version 6.02, Axon Instruments, Foster City, CA) were followed by two identical hyperpolarizing waveforms of one-half or four of one-quarter amplitude for linear leak and capacitive current subtraction (P/−2 subtraction method was used for all voltage-clamp experiments except those of Fig. 5 D, in which P/−4 was used). We selected for use varicosities yielding records with fast tail currents, indicating good spatial control of voltage at sites of current flow, and accepted perforated patch recordings with series resistances under 20 MΩ. In all experiments, the holding potentials were −70 mV (varicosity) and −80 mV (muscle cell). Currents and voltages were recorded with patch-clamp amplifiers (Axopatch 1B, 200, 200A; Axon Instruments), filtered with a 4-pole Bessel filter at 5–10 kHz digitized at 111–167 kHz, and stored on a PC-based microcomputer for analysis. Action potentials were recorded in somas using the Axopatch 1B and in the varicosities using the Axopatch 200A (I-clamp, normal setting). Off-line digital Gaussian filtering was performed at 10–20 kHz for the creation of the figures. In some cases, up to 100 μsec of the capacitance current artifacts were blanked.
Patch recording configurations
Figure 1 A illustrates a representative preparation, consisting of a motoneuron soma (S) and its neurite, ending in a synaptic varicosity (V) on a muscle cell (M). Sometimes the varicosities were simply enlargements of a neurite that made contact with a muscle cell en passant. In the case illustrated in Figure 1, the neuron soma, its varicosity, and the muscle cell were simultaneously patch-clamped—the varicosity with the perforated patch method (Horn and Marty, 1988) to avoid washout of cytoplasmic contents. This triple-patch configuration was used to demonstrate that an action potential could be generated in either the soma or the varicosity, eliciting postsynaptic currents that were equivalent in magnitude and time course (Fig.1 B).
Severing the neurite between the varicosity and the soma did not prevent either the generation of an action potential in the varicosity or the release of transmitter, and it had no significant effect on the action potential waveform, although the undershoot was slightly reduced (data not shown). Because our primary interest was in the correlation between presynaptic ionic currents and transmitter release, in subsequent experiments we restricted our measurements to the varicosity and the muscle cell. Unless specified otherwise, we ensured that the varicosity was forming a functional synapse by simultaneously monitoring transmitter release from the postsynaptic muscle cell. This was true even for those experiments in which only the presynaptic currents are shown (e.g., Figs. 2, 3). EPSCs, recorded under voltage-clamp conditions, were used to assess neurotransmitter release because of their greater accuracy (Katz and Miledi, 1979), and to prevent muscle contraction.
Presynaptic currents and transmitter release with step depolarizations
To measure the electrophysiological properties of the presynaptic membranes, we voltage-clamped the varicosity and dissected ionic currents by applying specific channel blocking agents in turn (Figs. 2,3). In NFR, depolarizing voltage steps evoked large, transient inward currents followed by delayed outward currents (Fig.2 A, left) reminiscent of the Na and K currents classically described in squid axon (Hodgkin and Huxley, 1952). The inward current was blocked by tetrodotoxin (TTX) (Fig.2 A, right). It should be noted that this current was too large to be maintained under adequate voltage clamp, given the resistance in series with the electrode in the perforated-patch method (10–15 MΩ).
As is illustrated in Figure 2 B in an experiment with another synapse, in the presence of TTX much of the outward current was blocked by the subsequent application of 3,4 DAP. Like the TTX-sensitive currents (Fig. 2 A), these outward currents were often too large to be adequately voltage-clamped with the perforated-patch method.
The residual currents obtained after the addition of DAP and TTX (Fig.2 B, right) displayed two interesting features. First, an inward current preceded an outward current, and second, at large depolarizations the inward current was not visible and the outward current began to decline in magnitude. These features suggested the presence of an early inward Ca current activating a delayed outward K Ca current (Marty, 1981; Blatz and Magleby, 1984). Figure 3 A (first two sets of traces) demonstrates that a large fraction of the outward current was blocked by charybdotoxin (CTX), a blocker of the high conductanceK Ca channel (Miller et al., 1985). Similar effects were observed with iberiotoxin, another blocker of the largeK Ca channel (Candia et al., 1992) (data not shown). The observation that the block of the outward current by CTX was incomplete (Fig. 3 A, second set of traces) may be explained by the fact that the CTX block of K Cachannels is reversed by large depolarizations (MacKinnon and Miller, 1988) or that higher concentrations of CTX are needed. The bell shape of the I–V plot of the CTX-sensitive current (Fig. 3 A, right; obtained from theDifference records) supports the conclusion that this is aK Ca current, and it will hereafter be termedI K-Ca.
The residual outward current that was insensitive to block by either DAP or CTX (Fig. 3 A, middle set of traces; 3B, left set of traces) was blocked by tetraethylammonium chloride (TEA), revealing a sustained inward current (Fig. 3 B, second sets of traces). This inward current was blocked completely by 100 μm CdCl2 (data not shown) and will hereafter be termed I Ca. This I Cais characterized by the I–V plot at the right of Figure 3 B (filled symbols). It should be noted that the peak of this I–V (at approximately +10 mV) is coincident with the peak of theI K-Ca I–V (Fig.3 A, right), further supporting the assertion that the CTX-sensitive current (Fig. 3 A, Differencerecords) is Ca2+-activated. Although this combination of pharmacological agents was relatively effective at isolating Ca currents, the inward current often was still contaminated by a small residual outward current if the internal presynaptic electrode contained K+. Moreover, because extracellular TEA blocks postsynaptic ACh receptors (Katz and Miledi, 1979), its use was contraindicated in studies designed to investigate the relationship between presynaptic Ca currents and transmitter release. Both problems could be circumvented by replacing the K+ in the recording pipette with Cs+. With this electrode solution, and the use of 10 mm Ca2+ in combination with TTX and DAP in the bathing solution, almost all of the outward current could be suppressed, allowing near-complete isolation of the presynaptic Ca current (Fig. 3 B, third panel, labeledCs+ internal). Close inspection of the traces of Figure 3 B (especially the third panel) reveals that at the onset of the voltage steps there were rapid, transient outward currents that could not be subtracted with the P/−2 pulse protocol (see Materials and Methods). These were seen even for small depolarizations and thus were not artifacts caused by saturation of the amplifier by the capacitive transients. These may represent gating currents of the various voltage-gated channels in the varicosity.
The isolation of the presynaptic I Ca by using internal Cs+ and external TTX and DAP allowed us to correlate the Ca2+ entry into varicosities with transmitter release. Figure 4 A shows a family of presynaptic currents and corresponding EPSCs obtained over a range of presynaptic voltage steps in a synapse bathed in normal extracellular Ca2+ (1.8 mm). As expected, with depolarizations to membrane potentials approaching the Ca2+reversal potential, the Ca current and the EPSC during the voltage step were suppressed. With these larger depolarizations, an increasing percentage of the EPSC arose from the Ca tail current at the end of voltage step (for example, see +60 and +90 mV). Peak presynapticI Ca ranged from 95 to 934 pA, whereas the peak EPSC ranged from 1 to 15 nA. Figure 4 B illustrates the mean current-to-voltage relationships of both the presynaptic varicosity (top panel) and the postsynaptic muscle cell (bottom panel) when measured simultaneously. In this correlation of the presynaptic Ca current with neurotransmitter release, we chose only those synapses in which the peak postsynaptic current was <5 nA. This was done to minimize the voltage error resulting from resistance in series with our pipettes. Note that the low-voltage arm of the Ca current I–Vrelationship is shallower than the corresponding arm of the EPSCI–V, consistent with a power-law relationship. Curiously, the peak of the EPSC I–V occurs at lower voltages than does the Ca current I–V, which is inconsistent with a simple power law and may imply saturation of the synaptic transfer function. The fact that the presynaptic current reversed polarity before the postsynaptic response decrements to zero is a reflection of contamination of the Ca current with an outward current. Finally, the asymmetry in the EPSCI–V implicates a possible asymmetry in the release process.
In Figure 4 C we illustrate the quantitative relationship between the presynaptic Ca current and peak postsynaptic response for the same junctions from which we plotted the I–Vvalues of Figure 4 B. To determine the maximum power-law relationship, we chose the first few voltages (−35 mV to −5 mV) in which Ca2+ entry produced a postsynaptic response. At larger depolarizations outward current contamination made measurements of the Ca current unreliable. The slopes of the linear regression fits of these data yielded a mean value of 1.76 ± 0.36 (SD, n = 7), with a range of 1.30 to 2.29.
Pharmacology of Ca currents and transmitter release
Pharmacological tests showed that I Ca in the varicosities was carried through more than one type of channel, only one of which seemed capable of mediating evoked transmitter release. The N-type Ca channel blocker ω-CgTX (1 μm) (Kerr and Yoshikami, 1984) reduced the varicosityI Ca to ∼8.6 ± 12% (n = 6) of the original value and blocked all neurotransmitter release (n = 20) (Fig. 5 A). TheI Ca block was partially reversible. On average, washout of the toxin restored I Ca within minutes to an average 32 ± 35% of control values (n = 6) (Fig. 5 B). There was never full recovery ofI Ca, however, with little or no recovery of transmitter release. This suggests that there may be at least two types of ω-CgTX-sensitive Ca channels in these preparations: one that is blocked irreversibly and the other that is blocked reversibly (Jones and Marks, 1989; Plummer et al., 1989).
In a parallel series of experiments, the Ca current was completely isolated when a combination of internal Cs+ and extracellular TEA with 10 mm CaCl2 was used. For these experiments we did not simultaneously monitor neurotransmitter release, but endeavored to choose junctions where we saw muscles twitching (spontaneously or after stimulation). In this preparation, uninnervated muscle cells fail to contract. In these experiments, the varicosity Ca current was reduced to 14.3 ± 2.5% by ω-CgTX (n = 8) and to 82.5 ± 5.1% by 2 μm dihydropyridine nimodipine (n = 3), but was insensitive to 500 nm ω-agatoxin IVA (n = 4), which blocks P-type channels (Mintz et al., 1992) (Fig. 5 C). The block by ω-CgTX and nimodipine applied together was essentially the same as the block by ω-CgTX alone (Fig. 5 C). This suggests that both drugs are working on the same Ca channels. Figure 5 D shows representative experimental results from those pooled in Figure 5 C. The nature of the remaining 15–17% of I Ca not blockable by ω-CgTX or nimodipine is not known, but it is probably not attributable to T-type channels, which are activated with small depolarizations and exist in the soma (Barish, 1991); we saw no low voltage-activated component to I Ca, and the ω-CgTX-resistant I Ca was insensitive to a decrease in the holding potential to −40 mV.
Presynaptic currents and transmitter release with action potential waveforms
One of the advantages of this coculture preparation is that it permits study of the activation kinetics of the presynaptic ionic currents underlying neurotransmitter release under physiological conditions of activation. We recorded action potentials elicited in varicosities (Fig. 6 A,left) and then used the action potential waveforms as command voltages for varicosities in voltage-clamp experiments (Llinás et al., 1982; Yazejian et al., 1995; Borst and Sakmann, 1996). The EPSC occurred during the falling phase of the action potentials (Fig. 6 A, left) and during the downward stroke of the voltage-clamped action potential (VCAP) waveforms (Fig. 6 A, right). The magnitude and delay of the evoked EPSCs were similar in both current-clamp and voltage-clamp conditions.
To study further the specific ionic currents underlying release, we blocked the major Na and K conductances in other varicosities with TTX and DAP. Under these conditions, the VCAP waveforms evoked biphasic ionic currents (linear currents were subtracted by the P/−2 method; see Materials and Methods) consisting of an early outward component correlated with the depolarizing phase of the voltage, followed by an inward current associated with the repolarizing phase (Fig.6 B, left). Each of these current components was much smaller than the Na current observed before block with TTX. Interestingly, CTX eliminated most of the early outward current, suggesting that this current is carried throughK Ca channels, and unmasked a prominent Ca current that preceded transmitter release (Fig. 6 B,right).
After K+ was replaced with Cs+ in the presynaptic pipette, the time course of the isolated Ca current during a VCAP stimulation then could be compared with the release of transmitter and with the action potential (Fig. 6 C). Interestingly, most of the Ca current occurred during the downstroke of the action potential; however, the inflection in the Ca current, which coincides with the peak of the action potential (Fig. 6 C), reveals that significant Ca2+ entry can also occur during the upstroke of the action potential. This early Ca current may enable the activation of the prominent I K-Ca seen in Figure 6 B.
This preparation and the use of the VCAP waveform has allowed direct comparison of the timing of Ca2+ entry during an action potential with transmitter release. For comparison with studies in other preparations, we have also measured the characteristic delays from the time of maximum rate of rise of the action potential and from the termination of a depolarizing step to the onset of transmitter release (Table 1).
Evidence for coactivation of presynaptic Ca and KCachannels during an action potential
In mature neuromuscular junctions there is evidence for structural colocalization of N-type Ca channels and K Cachannels (Robitaille and Charlton, 1992; Robitaille et al., 1993). A tight functional colocalization of dihydropyridine-sensitive (L-type) Ca channels and K Ca channels has been demonstrated in amphibian hair cells (Roberts et al., 1990). To determine whether in our preparation there is this type of coactivation of N-type Ca channels and K Ca channels during an action potential leading to transmitter release, we explored the effects of ω-CgTX on the activation of K Cachannels. We used VCAP waveforms to elicit Ca andK Ca currents mimicking those occurring during a physiological action potential. ω-CgTX blocked significantly both the inward Ca current and the outward K Ca current, and eliminated transmitter release (Fig.7 Ab). Hence, K Cachannels and neurotransmitter release are coactivated during an action potential by Ca current through N-type Ca channels. Interestingly, after washout of ω-CgTX, the K Ca current component recovered within 20 sec, whereas the EPSC and the net inward current showed slower, incomplete recovery (Fig. 7 A,c and d). The rapid recovery of theK Ca current before full recovery ofI Ca and transmitter release suggests that theK Ca channels display a higher affinity for Ca2+ ions, a closer proximity to the Ca channels, or different cooperativity among Ca2+ ions for their activation than does the release mechanism. Nevertheless, these data demonstrate that Ca2+ entry under physiological conditions during an action potential can coactivate K Cachannels and the release of neurotransmitter.
Because K Ca channels respond to Ca2+during an action potential, we used their activation as an assay of the decay of intracellular [Ca2+] after an action potential. We used a double-pulse paradigm with identical ramp waveforms and varied the interpulse interval. Figure 7 B (left) shows that the early K Ca current elicited during the rising phase of the first voltage ramp was greatly potentiated in the second ramp when the latter was delivered with an interpulse interval of under 4 msec. Increasing the interpulse interval led to decreased potentiation. We have shown previously with VCAP waveforms (Fig. 6 C) that Ca2+ entry occurred primarily during the repolarization phase of the action potential. If one assumes that a similar time course of Ca2+ entry occurred with the ramp waveform, then it is conceivable that the potentiation of theK Ca current in the second pulse of Figure7 B reflects the sensitivity of theK Ca channel as a Ca2+ concentration sensor. The magnitude of the K Ca current in response to the second ramp pulse is plotted against the interval between pulses in the right panel of Figure 7 B. The decay time constant (τ) of the potentiation was ∼1 msec and may represent the time course of deactivation of K Ca channels or reflect the persistence of Ca2+ in the vicinity ofK Ca channels.
We describe here a novel use of a culturedXenopus neuromuscular junction preparation. Presynaptic varicosities and postsynaptic muscle cells were patch-clamped to detect simultaneously the magnitude and time course of their respective ionic currents during synaptic transmission. After the major conductances responsible for the excitable properties of the presynaptic varicosity (Na and K V) are blocked, the smaller residual conductances (Ca and K Ca), which are potentially more interesting because of their involvement in neurotransmitter release, can be studied precisely under voltage control of the presynaptic varicosity.
Coupling of neurotransmitter release to ICaunder steady-state voltage conditions
The quantitative correlation of presynaptic calcium entry with the amount of neurotransmitter release requires ideal conditions and possibly various methods of synaptic stimulation and analysis. In our preparation we were confident of these comparisons only for Ca currents generated by low-voltage stimulations. We are aware that analysis of data obtained with other protocols, for example using Ca tail currents or action potential waveforms, may yield additional information. Nevertheless, our mean power-law relationship (1.76) is similar to that of Takahashi et al. (1996), who also used step waveforms and compared peak calcium tails with peak postsynaptic currents. On the other hand, values of 1–4 have been obtained at the squid giant synapse when presynaptic currents were compared with postsynaptic responses during step depolarizations at normal, constant extracellular Ca2+concentrations (Llinás et al., 1981; Augustine and Charlton, 1986). In the classic paper on mature neuromuscular junctions, Dodge and Rahamimoff (1967) reported a value of 4 obtained with changes in external Ca2+ concentration. In a recent report of experiments on the calyx of Held, Borst and Sakmann (1996), using action potential waveform depolarizations, found a fourth order relationship between peak Ca current and peak postsynaptic response. It is possible that these different numerical correlations betweenI Ca and release can be explained by the different methods used. For example, longer depolarizations during step pulses allow more Ca2+ entry, which may partially saturate the Ca-acceptor molecule(s), reducing the cooperativity, as has been suggested by Stanley (1986).
The timing of Ca2+ entry during an action potential and coupling of neurotransmitter release to ICa in the action potential
The varicosity–muscle cell preparation has advantages over many other preparations in allowing isolation and direct measurement of each ionic current, including direct measurements of Ca2+ entry, during an action potential waveform. Currents can be correlated with release, synaptic delays can be measured under (relatively) physiological conditions, and release can be resolved at the quantal level.
We show that Ca2+ entry occurred primarily during the falling phase of the action potential (Fig. 6 C). Similar results were obtained at squid synapses (Llinás et al., 1982), and comparable conclusions have been reached in studies of other, nonsynaptic preparations (McCobb and Beam, 1991; Scroggs and Fox, 1992;Wheeler et al., 1996) and recently in a CNS synapse (Borst and Sakmann, 1996). In addition, we detect a significant Ca2+ entry during the rising phase of the action potential (Fig. 6 C), as has been reported recently at a mammalian synapse (Sabatini and Regehr, 1996). Although it is difficult to determine whether this early Ca2+ current is itself sufficient to evoke transmitter release, it probably participates in the activation ofK Ca channels.
The average synaptic delay of 630 μsec between the end of a step depolarization and the onset of transmitter release (Table1 D) identifies the cultured Xenopusneuromuscular junction as a fast synapse. Interestingly, this value was approximately the same as the “physiological” synaptic delay we measured between the time of maximal rate of rise of theI Ca during the repolarization phase of the action potential and the onset of the EPSC (700 μsec) (Table1 C). The shortest delay we measured was 350 μsec, slightly longer than the 180 μsec minimum delay seen at the squid synapse (Llinás et al., 1981).
We observed similar synaptic delays between the maximum rate of rise of the action potential and the onset of the EPSC and between the maximum rate of rise of the voltage in a VCAP waveform and the EPSC (∼1.3 and 1.5 msec) (see Table 1, columns A and B). This finding verifies the validity of using the VCAP waveform in experiments measuring presynaptic currents coupled to transmitter release. Moreover, these values were only slightly longer than the value (∼1.1–1.2 msec) found at the mature neuromuscular junction by Katz and Miledi (1965). This similarity to the mature synaptic delay adds to the evidence that the varicosity synapses have well developed release machinery. Why both varicosity and mature neuromuscular junctions exhibit longer delays than does the squid synapse (minimum delay of 200 μsec) is not obvious. Ca2+ triggering and vesicle exocytosis are complicated processes, undoubtedly capable of adaptation to an increase or decrease in the speed of coupling, and the squid giant fiber system is a critical link in what has evolved to become a fast escape response.
Types of Ca channels and their coupling to transmitter release
In view of our finding that 1–2 μm ω-CgTX blocks all transmitter release in parallel with ∼85% of the presynaptic Ca current, we conclude that transmitter release is dependent on Ca2+ entry through N-type Ca channels. This is consistent with findings at mature frog neuromuscular junctions, at which it has been shown that ω-CgTX blocks all evoked transmitter release (Kerr and Yoshikami, 1984; Koyano et al., 1987; Katz et al., 1995) and some spontaneous release (Grinnell and Pawson, 1989). In the present study, nimodipine blocked ∼17% of I Ca, suggesting the presence of L-type Ca channels as well. Although we did not test the effect of nimodipine on release, application of nimodipine and ω-CgTX had no greater blocking effect than ω-CgTX alone, suggesting that both agents may work on the same channels, for which there is some precedent (Jones and Jacobs, 1990; Wang et al., 1992; Reeve et al., 1994). The remaining ω-CgTX-insensitive current has not been identified but was not blocked by ω-AGA-IVA; therefore, it was apparently not P-type (Mintz et al., 1992). Finally, the residual current was incapable of evoking release and not merely insufficient in magnitude, because similarly small ω-CgTX-sensitive currents were capable of evoking release. Because our primary interest was in the Ca channels important for transmitter release, we have not yet made a more systematic study of the residual Ca current.
Coactivation of KCa channels with Ca channels during an action potential
Our results show that K Ca channels are activated by Ca2+ entry during the rising phase of an action potential that initiates transmitter release (Fig.7 A). A close physical and functional association amongK Ca channels, Ca channels, and presynaptic release sites has been demonstrated at adult neuromuscular junctions (Cohen et al., 1991; Robitaille and Charlton, 1992; Robitaille et al., 1993) and in hair cells (Roberts et al., 1990). BecauseK Ca channel activation and transmitter release occur in a small window of time (the action potential) in parallel with the Ca2+ entry, it is probable that this is true in theXenopus varicosities as well.
It is desirable to know the channel densities and distributions in the varicosities. Assuming a Ca single channel conductance of 1.2 pS in normal extracellular Ca2+ (1.8 mm) (Church and Stanley, 1995), it can be estimated that the mean open Ca channel density obtained with step pulses was 2232 ± 1313 channels/varicosity, or 3.3 ± 2.2 channels/μm2(data presented in Fig. 3). For K Ca channels, the mean number of open channels/varicosity was estimated to be 109 ± 53, or 0.10 ± 0.04/μm2, assuming a single-channel conductance of 77 pS (Roberts et al., 1990). If one assumes, as has been demonstrated in other preparations (Roberts et al., 1990; Robitaille et al., 1993), that most of theK Ca channels are located at the synaptic contact area where the Ca channels are concentrated, this would result in a calculated ratio of Ca channels to K Ca channels of ∼15–30. Furthermore, it is estimated on the basis of the data from Figures 6 C and 7 A that the opening of ∼700 Ca channels and 15 K Ca channels occurs in the presynaptic varicosity during the generation of EPSCs by action potentials.
KCa channels as Ca2+ sensors
In our preparation, K Ca channels respond very quickly to Ca2+ entry during an action potential. Because of this behavior, the K Ca channel activation can be used to estimate the persistence of Ca2+in the submembranous space after an action potential. As is shown in Figure 7 B, the second of a pair of depolarizing ramps elicits a K Ca current that is enhanced compared with the first. The decay time constant of enhancement was 1 msec (Fig.7 B, right). This might represent the closing time of K Ca channels opened by Ca2+ entry in the first pulse. Alternatively, because the return ofK Ca current precedes the return of transmitter release after removal of block by ω-CgTX (Fig. 7 A),K Ca channels have an apparently higher effective sensitivity to Ca2+ than does the release mechanism. Given this higher sensitivity, the time constant of decay of potentiation of the K Ca current in Figure 7 B may represent a sensitive measure of the persistence of free Ca2+ near K Ca channels. This decay may give an upper limit to the time within which the presynaptic Ca2+ concentrations fall below threshold for neurotransmitter release after an action potential.
The cultured Xenopus varicosity–muscle preparation allows the simultaneous study of pre- and postsynaptic currents at a synapse where release can be resolved at the single quantum level. We have shown that ω-CgTX GVIA-sensitive Ca currents (N-type) predominate and regulate neurotransmitter release at these varicosities but that other minority Ca currents also exist. Calcium entry occurs primarily during the falling phase of the presynaptic action potential, and the delay between Ca2+ entry and release in this preparation is 600–700 μsec. In addition, we have demonstrated that calcium-activated potassium channels are expressed in these transmitter-releasing varicosities and have provided biophysical evidence that they are coactivated with calcium channels at release sites. This preparation should prove useful for many future physiological, biophysical, biochemical, cellular, and molecular studies of synaptic function.
This work was supported by grants from the National Science Foundation (BNS 8919481 to A.D.G.) and National Institutes of Health (NS 30673 to A.D.G., AR25201 to J.L.V., and NS32345 to S.D.M.). D.A.D. and R.E.P were partially supported by National Institutes of Health Fellowships GM 08042, NS10197, and MH 18273, respectively. We thank Phuong Hoang for preparation of the cell cultures and Jonathan Monck for comments on this manuscript.
Correspondence should be addressed to Bruce Yazejian, Department of Physiology, Jerry Lewis Neuromuscular Research Center, UCLA School of Medicine, Los Angeles, CA 90095-1751.