Calcium Sensitivity of Glutamate Release in a Calyx-Type Terminal

See allHide authors and affiliations

Science  11 Aug 2000:
Vol. 289, Issue 5481, pp. 953-957
DOI: 10.1126/science.289.5481.953


Synaptic efficacy critically depends on the presynaptic intracellular calcium concentration ([Ca2+]i). We measured the calcium sensitivity of glutamate release in a rat auditory brainstem synapse by laser photolysis of caged calcium. A rise in [Ca2+]i to 1 micromolar readily evoked release. An increase to >30 micromolar depleted the releasable vesicle pool in <0.5 millisecond. A comparison with action potential–evoked release suggested that a brief increase of [Ca2+]i to ∼10 micromolar would be sufficient to reproduce the physiological release pattern. Thus, the calcium sensitivity of release at this synapse is high, and the distinction between phasic and delayed release is less pronounced than previously thought.

In response to an action potential, the presynaptic release probability is strongly increased for a few milliseconds. This phasic release is thought to be triggered by a brief, localized increase in [Ca2+]i in the vicinity of open, presynaptic Ca2+ channels. The Ca2+ sensitivity of phasic release in mammalian central synapses is not yet known. On the basis of results obtained in other synapses, it has been assumed that a low-affinity Ca2+sensor, which is activated by local increases of [Ca2+]i to >100 μM, triggers phasic release in mammalian central synapses (1–4). In contrast, the more prolonged, delayed release period that, at most synapses, follows the phasic release may be controlled by a separate Ca2+ sensor with a much higher affinity for Ca2+ (5).

We measured the Ca2+ sensitivity of glutamate release at a giant synapse in the auditory brainstem, the axosomatic synapse formed by the calyx of Held with a principal cell in the medial nucleus of the trapezoid body. Using laser photolysis of caged Ca2+, we compared in the same terminals release evoked by a sustained, spatially uniform rise in presynaptic [Ca2+]i (6) with release triggered by action potentials, during which changes in [Ca2+]i are transient and highly localized (3). In 9-day-old rats, this synapse shows prominent synaptic depression during high-frequency signaling, which is most likely caused by rapid depletion of the releasable pool of vesicles (6–8). In order to relate the flash-evoked excitatory postsynaptic currents (EPSCs) to the size of the releasable pool in the same terminal, we first estimated the releasable pool size in the intact terminal. Simultaneous pre- and postsynaptic recordings were made from the calyx and a principal cell (9). With the presynaptic recording still in the cell-attached configuration, a train of action potentials was evoked by an extracellular electrode (Fig. 1A). A measure of release was obtained from the amplitudes of the glutamatergic EPSCs simultaneously recorded in the principal cell. During the train, the size of the EPSCs rapidly depressed, reaching a steady state within 100 ms. The cumulative amplitude of the EPSCs evoked by a train of afferent stimuli (200 ms, 200 Hz) was taken as a measure of the size of the releasable pool (7). This estimate was corrected for the steady-state component in the EPSCs (Fig. 1B). The cumulative EPSC was −9.7 ± 0.7 nA (n = 43, mean ± SEM) at a holding potential of −30 mV. The quantal EPSC amplitude was −32 ± 2 pA (n = 10 cells) at −80 mV. Assuming that the release of one vesicle gives an EPSC amplitude of −12 pA at −30 mV, this gave a releasable pool size of 810 ± 60 vesicles (6, 7). The amplitude of the first EPSC was 21 ± 2% (n = 43) of the amplitude of the cumulative EPSC. Taking the decay of the quantal EPSC into account, this means that about one-quarter of the releasable vesicle pool is released by a single action potential. In the presence of cyclothiazide, the 20 to 80% rise time of a single action potential–evoked EPSC was 424 ± 11 μs (n = 43). Its time course was not different at holding potentials of −80 and −30 mV (paired t test, P > 0.05; n = 7).

Figure 1

Rapid depletion of the releasable vesicle pool by [Ca2+]i jumps. Data in (A) to (C) are from the same synapse. (A) A high-frequency train of afferent stimuli induced presynaptic action potentials (I pre, presynaptic cell-attached voltage-clamp recording) and EPSCs (I post, postsynaptic whole-cell voltage-clamp recording). Vertical scaling as in (C). Postsynaptic holding potential was −30 mV. Stimulus artifacts have been removed. (B) The peak-to-peak amplitudes of the individual EPSCs shown in (A) were summed (•) to estimate the releasable pool size in the intact terminal. The solid line is a linear regression of the steady-state component of the train. Back-extrapolation to the start of the train gave an estimate for the cumulative amplitude of the train in the absence of pool replenishment (6). (C) After presynaptic whole-cell dialysis, a UV laser pulse (arrow) evoked a rapid and sustained [Ca2+]i increase (top trace) to 26 μM. The increase in [Ca2+]i resulted in a rapid, large EPSC (bottom trace), whose amplitude approximated the (corrected) sum of the amplitudes of the EPSC train in (A). Its amplitude was larger and its rise time was faster than the action potential–evoked EPSC in the same terminal. Pre- and postsynaptic holding potentials were −80 and −30 mV, respectively. (D) Relative size of the EPSC evoked by the UV flash compared with the cumulative EPSC amplitude evoked by afferent trains in the same terminals, displayed on log-log coordinates. Data were pooled from 26 experiments.

After establishing the whole-cell configuration, the terminal was loaded via the patch pipette with a solution containing the ultraviolet (UV)-sensitive Ca2+ buffer DM-nitrophen (DM-n) (10) and a low-affinity Ca2+ indicator. This enabled us to evoke transmitter release by rapidly uncaging Ca2+ by laser photolysis of DM-n (11–13). The spatially uniform rise in [Ca2+]i to levels between 0.5 and 100 μM was monitored with a photodiode (Fig. 1C), starting ∼200 μs after the UV pulse. At this point, the rapid decay component of the transient [Ca2+]i spike (11, 14) had already subsided (15). The measured [Ca2+]i decayed by about 30% in 50 ms. The increase in [Ca2+]i induced a small, slow outward current in the presynaptic terminal that was not further investigated. Laser photolysis triggered EPSCs, whose amplitude depended on the [Ca2+]i levels that were reached. Increases in [Ca2+]i to concentrations of 7 μM and higher evoked an EPSC with an amplitude that was as large as the cumulative amplitude of the EPSCs evoked by the brief afferent stimulus train before the whole-cell configuration was established (Fig. 1D). At lower [Ca2+]i the laser-evoked EPSCs were smaller (Figs. 1D and 2A), probably because of the decay of the quantal EPSCs during the rising phase of these slower EPSCs. Therefore, our results suggest that the measured increases in [Ca2+]i after laser photolysis targeted the same, kinetically distinct, pool of vesicles as released by the transient, localized increase in [Ca2+]i after an action potential (6).

Figure 2

Relation between [Ca2+]i and the rate of exocytosis in the calyx of Held. (A) (Top) Photodiode traces of [Ca2+]i jumps to 26 (trace 1), 4.5 (trace 2), and 2.3 μM ( trace 3) in three different experiments. During the period indicated by the horizontal dashed lines, the excitation wavelength was briefly switched to the isosbestic wavelength. (Bottom) Corresponding postsynaptic currents. EPSCs were normalized to the size of the cumulative EPSC amplitude obtained in the same experiment. (B) A uniform increase of the [Ca2+]i in the terminal to <1.5 μM (top) resulted in a clear increase in the frequency of small EPSCs (bottom traces, V h = −80 mV). UV pulses were separated by ∼3 min. (Top) Overlaid [Ca2+]i of the first and the last of four sweeps. (Bottom) Asterisks mark putative quantal release events. (C) Summary of the relation between peak release rates and [Ca2+]i, displayed on log-log coordinates. Peak release rates of compound EPSCs (⧫) were corrected for the finite rise time of the average quantal EPSC. For [Ca2+]i jumps of <1.5 μM, the average quantal EPSC rate (▴) was analyzed during the 20 ms after the mean first latency in three to five sweeps per experiment. The solid line is derived from a kinetic model of the Ca2+ sensor (22). The measured relation between release rate and [Ca2+]i followed a power dependence of 4.4 for [Ca2+]i of <5 μM. Release rates are specified per vesicle. Data of 31 synapses were pooled. (D) [Ca2+]i dependence of the delays between the [Ca2+]i jump and the onset of release. The onset of compound release (⧫) was defined as the time when the EPSC intersected a threshold of −35 pA; the onset of quantal release (▴) was defined as the mean first latency of the quantal EPSCs. Solid line, the predicted mean delay between the [Ca2+]ijump and the release of the first transmitter quantum. A delay of 250 μs was added to the simulated delays to match the experimental data (48).

We estimated the release rate per single vesicle after a [Ca2+]i jump. Larger increases in [Ca2+]i evoked EPSCs with a smaller delay and a shorter rise time (Fig. 2A). Apparently, the time needed to deplete the releasable pool depended on [Ca2+]i, suggesting that a sustained increase in [Ca2+]i of >4 μM was sufficient to deplete the releasable pool of vesicles on a millisecond time scale. The rising phase of the compound EPSCs could therefore be used to calculate release rates (16). The peak release rate during a laser-evoked EPSC was divided by the estimated number of releasable vesicles for the same synapse, thus correcting for pool size variability between synapses. In contrast to results obtained in other preparations (17, 18), there was no clear threshold for transmitter release. [Ca2+]iincreases to ∼1 μM, which is close to the [Ca2+]i during the delayed release phase in the calyx of Held (19), triggered a sequence of individually resolvable quantal EPSCs (Fig. 2B). Their frequency provided a direct measure of the evoked change in release rate. The calculated release rate varied more than 10,000-fold as [Ca2+]i varied from 0.5 to 100 μM (Fig. 2C) (20, 21). The fastest rise times of the laser-evoked EPSCs measured were 220 ± 12 μs (n = 4), corresponding to a maximal release rate of ∼6 ms−1 per vesicle. The delay from the UV pulse to the start of the EPSC was less than 0.3 ms at a [Ca2+]i of >30 μM, whereas at a [Ca2+]i of ∼1 μM, the delay to the first quantal EPSC was still on average <10 ms (Fig. 2D). This indicates that the Ca2+ sensor binds Ca2+ rapidly before it triggers the final steps of transmitter release.

We fitted the relation between peak release rate and [Ca2+]i using a kinetic model of the Ca2+ sensor and its interaction with the releasable vesicles (Fig. 2C). The model features five identical Ca2+-binding steps, followed by a final, reversible, Ca2+-independent isomerization step that promoted vesicle fusion (22). A satisfactory prediction of the [Ca2+]i dependence of both the release rates and the delays was obtained with the parameters given in (22). Although this parameter set was not unique, several conclusions could be drawn from the fitting procedure. To reproduce the fast depletion of the pool at high [Ca2+]i, a large isomerization rate constant and fusion rate constant were needed. To reproduce the apparent saturation of release rates at [Ca2+]i of >30 μM, a dissociation constant (K d) of ∼10 μM for the individual binding steps was needed, not very different from the estimated affinities of the Ca2+ sensor that triggers the release of large dense-core vesicles (23–25), but clearly lower than previously estimated for the release of clear vesicles from bipolar cells of the goldfish retina (26).

The laser photolysis experiments can be used to calculate the typical [Ca2+]i transient observed by a Ca2+ sensor during action potentials (27). The rise times of the action potential–evoked EPSCs indicated that peak release rates were 0.42 ± 0.04 ms−1 per vesicle (n = 43). A sustained increase of [Ca2+]i to 5 μM gave release rates similar to the ones observed during action potentials (Fig. 3A). This concentration is therefore a lower estimate, because the peak [Ca2+]ireached during an action potential will be reached only very briefly and will not trigger release as efficiently as a steady increase to the same level.

Figure 3

Comparison of release rates after action potentials and [Ca2+]i jumps. (A) A prolonged [Ca2+]i increase to 4.5 μM (top) evoked an EPSC (EPSCexp, 4.5 μM, bottom) that rises almost as fast as an action potential–evoked EPSC (EPSCexp, AP). The action potential–evoked EPSC was aligned such that the putative peak of the presynaptic Ca2+ current coincided with the [Ca2+]i jump (arrow). Scaling as in (B). (B) Simulation of EPSCs evoked by a brief increase in [Ca2+]i. The time course of the [Ca2+]i transient was assumed to be the same as that of the measured Ca2+ current during an action potential (top) (22). It was scaled and used to drive the kinetic release model to simulate a release rate (Rate) that produced an EPSC (EPSCsim, AP) of the same amplitude as observed during action potential–evoked release (EPSCexp, AP.). The simulated release rate and EPSC were shifted to the right by 250 μs (48).

An upper estimate can be obtained for the [Ca2+]i transient peak value for the hypothetical situation that all release sites faced the same [Ca2+]i transient. We assumed that the time course of the [Ca2+]i transient at the Ca2+ sensor is not faster than the Ca2+ current during an action potential (Fig. 3B), which was measured previously (22, 28). With this time course, the amount of release evoked by the simulated [Ca2+]itransient matched the release evoked by real action potentials if the peak [Ca2+]i was ∼9 μM. This estimate was largely model-independent. After adjustment of the parameters of other kinetic models (24, 26,29) to satisfy the relation between [Ca2+]i and release rates, a similar estimate was obtained (30). Assuming a linear relation between Ca2+ influx and the peak of the [Ca2+]i transient, the simulated action potential–evoked release shared several features with the experimentally characterized release. Delays and rise times of EPSCs were largely independent of the amount of Ca2+ influx during the action potential (28), although for very high Ca2+ influx, a decrease in the synaptic delay and the rise time was observed. The time course of the release probability matched the experimentally observed time course (28). The model predicted a fourth-power dependence of EPSC amplitudes on external [Ca2+], somewhat higher than previously measured (6). Our results do not indicate that Ca2+ sensors never experience [Ca2+]i of >10 μM during action potentials. However, they suggest that, in contrast to earlier suggestions (1–4), most Ca2+sensors in the calyx of Held do not experience [Ca2+]i of hundreds of μM, because even if they were exposed for a brief period, release during action potentials would be faster and larger than experimentally observed.

We conclude that transmitter release from the calyx of Held exhibited a high Ca2+ sensitivity compared with previous estimates for the release of clear vesicles from other synapses (18, 26). Our characterization of the Ca2+ sensitivity of synaptic transmitter release may be of use in identifying possible Ca2+ sensors and in elucidating the molecular mechanisms of transmitter release. For example, synaptotagmin I and II are prominent candidates for the Ca2+ sensor that triggers phasic release (2, 4). Our results suggest that its binding to syntaxin is unlikely to be involved in the final steps before fusion at the calyx of Held because it requires very high [Ca2+]i (2). Synaptic terminals contain a plethora of other Ca2+-binding proteins with a higher affinity for Ca2+, which may be considered as alternative candidates for the Ca2+ sensor (4).

We recorded from calyces of young rats, and the Ca2+ sensitivity of release may change during development. However, the observed high sensitivity agrees with other observations. First, [Ca2+]i of <10 μM evoke substantial release in the squid giant synapse (31, 32) and in the crayfish neuromuscular junction (33). Second, the slow Ca2+ buffer EGTA not only abolishes delayed release (28, 34, 35), but also affects phasic release at many synapses (28, 34,36). [Ca2+]i of hundreds of μM are reached only in the immediate vicinity of open Ca2+channels, where EGTA would be ineffective. Third, at the frog neuromuscular junction, EGTA inhibits release, but not the Ca2+-dependent potassium channels. This suggests that the Ca2+ sensor for release is farther away from the Ca2+ channels than the Ca2+-dependent potassium channels (37). At the cultured neuromuscular junction, simulations suggest that [Ca2+]i is <10 μM in most regions of a presynaptic Ca2+ entry site (38). Combined with our result that a low-affinity Ca2+ sensor is not a prerequisite for phasic transmitter release, these results suggest that the high Ca2+sensitivity of phasic release at the calyx of Held may be a property of many synapses.

The phasic-release Ca2+ sensor equilibrated rapidly to changes in [Ca2+]i and triggered release with a high maximal speed, much faster than for dense-core vesicles (39). However, the Ca2+ sensitivity observed in the calyx was not very different from the sensitivity of the release of large dense-core vesicles in, for example, melanotrophs (23). Similar Ca2+-binding mechanisms may therefore be at work. Finally, because the Ca2+ sensor that triggers phasic release has relatively high Ca2+sensitivity, it may be predicted that delayed release is to a large extent a consequence of a delayed triggering of the same sensor, rather than a result of the triggering of a different sensor with much higher affinity.

  • * To whom correspondence should be addressed. E-mail: jbollman{at}


Stay Connected to Science

Navigate This Article