Properties of Synchronous and Asynchronous Release During Pulse Train Depression in Cultured Hippocampal Neurons

Properties of Synchronous and Asynchronous Release During Pulse Train Depression in Cultured Hippocampal Neurons DONALD J. HAGLER, JR. AND YUKIKO GODA Division of Biology, University of California, San Diego, La Jolla, California 92093-0366 Received 17 October 2000; accepted in final form 13 February 2001 Hagler, Donald J., Jr. and Yukiko Goda. Properties of synchronous and asynchronous release during pulse train depression in cultured hippocampal neurons. J Neurophysiol 85: 2324 2334, 2001. Neurotransmitter release displays at least two kinetically distinct components in response to a single action potential. The majority of release occurs synchronously with action-potential-triggered Ca 2 influx; however, delayed release also called asynchronous release persists for tens of milliseconds following the peak Ca 2 transient. In response to trains of action potentials, synchronous release eventually declines, whereas asynchronous release often progressively increases, an effect that is primarily attributed to the buildup of intracellular Ca 2 during repetitive stimulation. The precise relationship between synchronous and asynchronous release remains unclear at central synapses. To gain better insight into the mechanisms that regulate neurotransmitter release, we systematically characterized the two components of release during repetitive stimulation at excitatory autaptic hippocampal synapses formed in culture. Manipulations that increase the Ca 2 influx triggered by an action potential elevation of extracellular Ca 2 or bath application of tetraethylammonium (TEA) accelerated the progressive decrease in synchronous release (peak excitatory postsynaptic current amplitude) and concomitantly increased asynchronous release. When intracellular Ca 2 was buffered by extracellular application of EGTA-AM, initial depression of synchronous release was equal to or greater than control; however, it quickly reached a plateau without further depression. In contrast, asynchronous release was largely abolished in EGTA-AM. The total charge transfer following each pulse accounting for both synchronous and asynchronous release reached a steady-state level that was similar between control and EGTA-AM. A portion of the decreased synchronous release in control conditions therefore was matched by a higher level of asynchronous release. We also examined the relative changes in synchronous and asynchronous release during repetitive stimulation under conditions that highly favor asynchronous release by substituting extracellular Ca 2 with Sr 2. Initially, asynchronous release was twofold greater in Sr 2. By the end of the train, the difference was 50%; consequently, the total release per pulse during the plateau phase was slightly larger in Sr 2 compared with Ca 2. We thus conclude that while asynchronous release like synchronous release is limited by vesicle availability, it may be able to access a slightly larger subset of the readily releasable pool. Our results are consistent with the view that during repetitive stimulation, the elevation of asynchronous release depletes the vesicles immediately available for release, resulting in depression of synchronous release. This implies that both forms of release share a small pool of immediately releasable vesicles, which is being constantly depleted and refilled during repetitive stimulation. INTRODUCTION A hallmark of neurotransmitter release is its exquisite temporal regulation. Arrival of an action potential triggers Ca 2 - dependent exocytosis of synaptic vesicles within less than a millisecond. Phasic (or synchronous) neurotransmitter release is central to proper brain function as evidenced by early postnatal death of mice deficient in synaptotagmin I whose phasic release is severely compromised (Geppert et al. 1994). When synaptic transmission is monitored from many synapses, synchronous release of neurotransmitters is followed by a delayed release that also exhibits Ca 2 dependence (Barrett and Stevens 1972; Goda and Stevens 1994; Meiri and Rahamimoff 1972; Miledi 1966). Such asynchronous release is thought to be sustained by residual Ca 2 in the presynaptic terminal. Consistently, asynchronous release is enhanced during conditions that elevate intracellular Ca 2 as observed in the course of repetitive stimulation. Asynchronous release is thus likely to significantly contribute to synaptic transmission in vivo where neuronal activity mostly occurs as spike trains (Hubel 1959; O Keefe and Dostrovsky 1971). The physiological relevance of asynchronous release, however, remains to be established. During repetitive stimulation, the increase in asynchronous release is accompanied by a progressive decline in the size of synchronous release. This decrease in synaptic transmission, which we refer to as pulse train depression (PTD), has been attributed to occur primarily by depleting a pool of releaseready synaptic vesicles (Dobrunz and Stevens 1997; Liley and North 1953; Rosenmund and Stevens 1996). Supporting the hypothesis that depletion of the readily releasable pool (RRP) causes PTD, conditions that promote synaptic vesicle fusion by increasing presynaptic Ca 2 influx increase the rate of PTD (Tsodyks and Markram 1997; Varela et al. 1997). Recent studies indicate that presynaptic mechanisms other than vesicle depletion may also contribute significantly to synaptic depression. Accordingly, in cultured hippocampal neurons, repetitive stimulation releases only 77% of the RRP, the size of which is determined by the application of hypertonic solution (Rosenmund and Stevens 1996). In addition, models of depression that take into account depletion only cannot completely match the properties of experimentally derived PTD data (Dittman and Regehr 1998; Matveev and Wang 2000). Potential presynaptic mechanisms include inhibition by presynaptic autoreceptors (Scanziani et al. 1997; Takahashi et al. 1996), Ca 2 - Address for reprint requests: Y. Goda, Div. of Biology, University of California, San Diego, 9500 Gilman Dr., La Jolla, CA 92093-0366 (E-mail: ygoda@biomail.ucsd.edu). The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 2324 0022-3077/01 $5.00 Copyright 2001 The American Physiological Society www.jn.org

TRANSMITTER RELEASE DURING PULSE TRAIN DEPRESSION 2325 channel inactivation (Forsythe et al. 1998; Patil et al. 1998), action potential conduction failures (Brody and Yue 2000; Hatt and Smith 1976; Streit et al. 1992), and modulation of Ca 2 - sensitive exocytosis machinery (Bellingham and Walmsley 1999; Hsu et al. 1996; Wu and Borst 1999). The purpose of this study is to investigate the relation between synchronous and asynchronous release during PTD and how PTD affects the number of readily releasable vesicles in autaptic excitatory synapses formed in cultured hippocampal neurons. We demonstrate that during repetitive stimulation the decline of synchronous release is matched by increased asynchronous release. In addition, our findings are consistent with recent models of depletion and refilling that are based on a small subset of the RRP comprised of on average one or fewer vesicles which we term the immediately releasable pool (IRP). Our results suggest that this IRP is shared by synchronous and asynchronous release; thus depletion by elevated asynchronous release can reduce future synchronous release. METHODS Unless otherwise noted, all chemicals are from Sigma (St. Louis, MO). Hippocampal neuronal cultures Cultured hippocampal neurons were prepared from P0 to P2 rats as previously described with a few modifications (Bekkers and Stevens 1991; Segal and Furshpan 1990). Dissection solution consisted of Hank s balanced salt solution (Life Technologies, Gaithersburg, MD) supplemented with 25 mm HEPES, ph 7.35. The dentate gyrus was removed, and the remaining hippocampal tissue was proteolyzed in dissection solution plus 20 U/ml papain (Worthington, Freehold, NJ), 0.5 mm EDTA, 1.5 mm CaCl 2,1mM L-cysteine, and 0.1 g/ml DNAase for 30 min at 37. Tissue pieces were triturated in culture media (CM) Basal Media Eagle (Life Technologies) plus 1 mm HEPES ph 7.35, 1 mm Na-pyruvate, 50 U/ml penicillin, 50 g/ml streptomycin, 6 mg/ml glucose, 10% fetal bovine serum (Hyclone, Logan, UT or Omega Scientific, Tarzana, CA), mito serum extender (Fisher Scientific, Pittsburgh, PA), and 2% B27 (Life Technologies) and 0.5 ml of cell suspension was plated at 1 2 10 4 cells/ml onto 12-mm coverslips that had been preplated with astrocytes. One day after plating, 4 M cytosine -D-arabinofuranoside (arac) was added to prevent astrocyte proliferation. Every 3 4 days thereafter, cells were fed with 0.1 ml of CM. Astrocytes were prepared from extra cells obtained from each dissection. After a week, the confluent glial cells were shaken at 260 rpm overnight to enrich for type I astrocytes. Usually, the cells were then passaged and grown for an additional week to select for adherent, quickly spreading astrocytes. Astrocytes were plated at 6 10 3 /ml onto coverslips sprayed with microdots of collagen and poly-d-lysine. After 2 4 days, 4 M arac was added to limit the size of the astrocyte islands. Neurons were added 3 6 days after astrocyte plating, after aspirating most of the media. Cultures were used for experiments 8 14 days after plating. Recording solutions Medium Ca 2 extracellular bath solution (MC-EBS) contained (in mm) 135 NaCl, 5 KCl, 3 CaCl 2, 2 MgCl 2,10D-glucose, 5 HEPES- NaOH (ph 7.3), and 0.1 picrotoxin plus 1 M glycine. Osmolarity was adjusted with sorbitol to 315 325 mosm. 2-Amino-5-phosphonopentanoic acid [( )APV, 50 M] was included to prevent longterm plasticity. High Ca 2 EBS (HC) was modified MC-EBS with 10 mm CaCl 2 and 0.5 mm MgCl 2. Low Ca 2 EBS (LC) contained 1 mm CaCl 2 and 4 mm MgCl 2. The recording chamber was perfused with a constant flow rate of 0.5 1 ml/min. Pipette solution contained (in mm) 127.7 K-gluconate, 16.4 KCl, 8.4 NaCl, 2.9 MgCl 2, 9.4 HEPES- KOH (ph 7.2), 0.2 EGTA, 2 ATP, 0.5 GTP, and 10 creatine phosphate and 25 U/ml creatine phosphokinase. Drugs were prepared as stock solutions at the following concentrations in the indicated solvents: EGTA-AM, 5-nitro-2-(3-phenyl-propylamino) benzoic acid (NPPB), and cyclothiazide (CTZ) at 100 mm in DMSO, (S)- methyl-4-carboxyphenylglycine (MCPG) at 50 mm in 0.1 N NaOH, apamin at 1 mm in 50 mm acetic acid, and TEA (tetraethylammonium) at 1MinH 2 O. Final dilutions in EBS are stated in the text. Data acquisition Standard whole cell procedures were followed. Recordings were obtained with Axopatch 200B (Axon Instruments, Foster City, CA), filtered at 2 khz and digitized at 2 5 khz. Traces were stored and analyzed with programs written by DH using Visual BASIC 5.0 (Microsoft, Redmond, WA) and ComponentWorks 1.1 (National Instruments, Austin, TX). Neurons were held at 70 mv; series resistance was compensated to 80%. Leak currents were not subtracted and recordings with a leak 0.2 na were excluded. There was at least a 45-s delay between trains of 20 stimuli, which consisted of 1-ms step depolarizations to 30 mv. All experiments were performed at room temperature (22 25 C). Analysis of synchronous release Excitatory postsynaptic current (EPSC) amplitudes were measured by subtracting an average baseline value 1 10 ms before each pulse from the peak following the stimulus artifact. For cells in HC-EBS, the mean initial EPSC amplitude was 6.7 1.2 (SE) na, ranging from 0.5 to 31.5 na, with a standard deviation of 7.6 and a median of 3.4 na (n 39). No relationship between the rate of PTD and the initial peak EPSC amplitude was found (not shown). If short-term potentiation was observed in a given cell, responses were included only after the initial peak EPSC amplitude stabilized (typically after a single pulse train). Recovery of synchronous release following PTD was assayed by applying a train of 20 pulses followed by a test pulse at varying time intervals. Recovery was calculated by normalizing the amplitude of the test EPSC to the amplitude of the first EPSC of the depressing train. To adjust for incomplete depression (average depression to 9 1% of initial EPSC amplitude, n 7), the amplitude of the last EPSC of the train was subtracted from both values before normalization. Analysis of asynchronous and total release Asynchronous and total release were estimated by integrating the current in the appropriate time window following each pulse 40 50 ms for asynchronous and 5 ms to immediately before the next pulse for total release. A baseline current measured before the first pulse of the train was subtracted from current traces before integration. Total release was normalized to the first response, similarly to synchronous release. Asynchronous release was normalized to the last response as asynchronous release on the first pulse was small in most cases. The asynchronous fraction of release the relative contribution of asynchronous release to total release for responses to 20-Hz stimulation, which is shown in Fig. 5F, was approximated by dividing the average current 40 50 ms after each pulse by the average current 5 50 ms after each pulse. It should be noted that our method of estimating the amount of asynchronous release by measuring charge transfer has some limitations. First, a portion of the measured currents could be due to presynaptic currents, which are expected to be insensitive to EGTA- AM. EGTA-AM, however, greatly reduces the measured currents (see Figs. 6 and 10), suggesting that the presynaptic currents do not

2326 D. J. HAGLER AND Y. GODA contribute significantly to the estimate of asynchronous release. A second concern is that accumulation of extracellular glutamate and desensitization of glutamate receptors may affect the size of the measured currents. The contribution of these phenomena to the measurement of asynchronous release is difficult to estimate, and so we compared our method with a straightforward count of individual asynchronous release events. In a cell with an initial EPSC amplitude of 1 na, in medium Ca 2 i.e., under conditions in which individual miniature EPSCs (mepscs) can be resolved we find that the charge transfer and number of mepscs increases with matching time courses during the pulse train and that the charge transfer per release event closely matches the average charge transfer measured for well-isolated individual releases (data not shown). Measurement of hypertonic-evoked response Hypertonic solution was applied as previously described (Stevens and Tsujimoto 1995). A 3-s puff of MC-EBS plus 0.5 M sucrose was applied at the end of a train of 100 pulses at 20 Hz, and the peak of the hypertonic response (HR) arrived 1 s after the end of the train. Cells in which the peak EPSC amplitude at the end of the train did not depress to 5% of the initial peak EPSC size were excluded (1 of 5 cells tested). HR currents were integrated using 200-ms bins; the value from the bin with the largest charge transfer the peak of the integrated response was used as a measure of the RRP (Stevens and Tsujimoto 1995). Control measurements of the HR were obtained 60 s before and after each test HR measurement ( test HR refers to the HR 1 s after the pulse train). To correct for possible rundown in these measurements, the test HR measurement was normalized to the average of the preceding and subsequent control HR measurement. To correct for the baseline shift due to residual asynchronous release following the pulse train, the integrated charge of a control pulse train was subtracted from that of a pulse train plus hypertonic solution application. On average, the uncorrected test HRs were 76 6% of the control HRs. Following each hypertonic solution application or control pulse train, there was at least a 60-s delay before subsequent stimulation to allow for recovery of the RRP. RESULTS Extracellular Ca 2 and frequency dependence of PTD We first examined the basic properties of the decline of synchronous release (peak EPSC amplitude) during pulse trains by varying extracellular Ca 2 and stimulation frequency. The effects of changing extracellular Ca 2 and thus release probability on PTD caused by repetitive stimulus trains in low (1 mm), medium (3 mm), or high (10 mm) extracellular Ca 2 are shown in Fig. 1A. Relative to high Ca 2, synchronous release to the first pulse was decreased to 84 6% (n 8) in medium Ca 2 and 34 4% (n 6) in low Ca 2. While the rate of PTD was fast in high Ca 2, PTD was slowed in lower Ca 2. Moreover, facilitation of the second EPSC of the train relative to the first was observed in low Ca 2. When stimulus frequency was varied 5, 10, and 20 Hz higher stimulus frequency resulted in a faster rate of depression (Fig. 1, B D), in agreement with previous studies (Dittman and Regehr 1998; Galarreta and Hestrin 1998; Janz et al. 1999; Tsodyks and Markram 1997; Varela et al. 1999). At 20 Hz in high Ca 2, the EPSC amplitude usually decayed to zero within 10 20 pulses (Figs. 1A and 2B). RRP recovers quickly following complete depression PTD has been generally interpreted to result from depleting the RRP of vesicles (Liley and North 1953) (see INTRODUCTION). FIG. 1. Dependence of pulse train depression (PTD) on extracellular Ca 2 and stimulation frequency. A: excitatory postsynaptic current (EPSC) amplitudes in response to 20 pulses at 20 Hz in low ( ), medium ( ), and high (Œ) Ca 2 were normalized to the first response of the train in high Ca 2 (n 14). Recordings from a given cell were made in high Ca 2 and in low (n 6) or medium (n 8) Ca 2. B D: normalized EPSC amplitudes to 5 ( ), 10 ( ), and 20 (Œ) Hz train are shown in low (B), medium (C), and high (D) Ca 2. For B, responses from 9 cells are shown. In C, responses are from 17, 38, and 35 cells for 5, 10, and 20 Hz, respectively. In D, responses are from 27 cells for 5 Hz and 39 cells for 10 and 20 Hz. Error bars in this and other figures represent SE. LC, MC, and HC, low, medium, and high Ca 2, respectively. The observed increase in the rate of PTD caused by a higher stimulus frequency and conditions that increase release probability is consistent with this model. The depletion model would then predict that the lack of synaptic response observed for PTD at higher stimulus frequencies reflects the emptied RRP (Rosenmund and Stevens 1996). To address whether the RRP is indeed emptied following PTD, we first applied a prolonged pulse train (100 pulses) at 20 Hz (in medium Ca 2 ) to achieve a fully depressed state, where the synchronous release is negligible or nonexistent. Immediately after this depression, the size of the RRP was measured by applying extracellular solution made hyperosmolar with 0.5 M sucrose (Rosenmund and Stevens 1996). Surprisingly, 62 6% (n 4) of the RRP was available for release just 1 s after the end of the train, the shortest time at which the RRP can be reliably assayed by hypertonic solution application (Fig. 2D). The

TRANSMITTER RELEASE DURING PULSE TRAIN DEPRESSION 2327 FIG. 2. Readily releasable pool (RRP) size following complete depression of synchronous release. A: example EPSC trace from a cell recorded in high Ca 2, stimulated at 20 Hz; stimulus artifacts were removed. In B, the EPSC trace to each pulse from A are overlaid with artifact present. C: example overlaid EPSC traces during a pulse train from a cell recorded in medium Ca 2 prior to application of hypertonic solution. The 1st, 5th, and 10th responses and every 10th response thereafter of the 100-pulse train are shown. Inset: a plot of average normalized EPSC amplitudes from 4 such cells. D: example of hypertonic-response (HR). Left trace: HR at rest; middle trace: decay of asynchronous release following a pulse train; and right trace: HR following a pulse train. E: recovery of RRP (, n 4) is plotted against the time delay after a train of 100 pulses. Recovery of EPSC amplitude (, n 7) is plotted as a function of the time delay after a 20-pulse train. Recordings were made in medium Ca 2. - - -, the refilling of the RRP, assuming exponential recovery with a time constant of 3 s (Stevens and Wesseling 1998). recovery of RRP was also matched by the recovery of the EPSC amplitude tested at various times following PTD (Fig. 2D). The refilling of the RRP has been well characterized as a single exponential recovery with a time constant ranging from 3 to 11 s (Dobrunz and Stevens 1997; Pyle et al. 2000; Rosenmund and Stevens 1996; Stevens and Tsujimoto 1995; Stevens and Wesseling 1998). The lower limit of 3 s is achieved by a Ca 2 -dependent acceleration of refilling that occurs following repetitive stimulation; however, this effect saturates at 10 pulses at 10 Hz, such that more pulses or higher stimulus frequencies produce approximately the same rate of refilling (Stevens and Wesseling 1998). Figure 2D, ---, shows the predicted recovery from an emptied RRP that would be observed under accelerated conditions. Note that the observed recovery of the RRP attained within 1 s following PTD is much faster than the prediction. While this could indicate an exceptionally accelerated refilling, it is plausible that the fast recovery results from incomplete depletion of the RRP. In support of this latter proposal, at the end of a pulse train substantial asynchronous release occurs even though synchronous release is completely depressed (see following text). Thus factors other than depletion likely contribute to the depression of the synchronous EPSC. Factors that could potentially contribute to PTD We next addressed mechanisms other than depletion that might potentially contribute to PTD in cultured hippocampal neurons. To determine whether desensitization of AMPA-type glutamate receptors plays a role in PTD, repetitive stimulation was applied in the presence of CTZ to block AMPA-receptor desensitization. CTZ (100 M) had no effect on the rate of PTD despite a threefold increase in the initial EPSC amplitude (Fig. 3A). A small increase in the normalized amplitude of the second response was observed in CTZ at 20 Hz; this difference was not observed for 5 or 10 Hz and was not statistically significant (n 4, P 0.23, 2-tailed, paired t-test). We also examined the potential contribution of metabotropic glutamate receptors in modifying presynaptic glutamate release during repetitive stimulation (Scanziani et al. 1997; Takahashi et al. 1996). Bath application of 250 M MCPG, a metabotropic glutamate receptor antagonist, was without effect on PTD (Fig. 3B). AMPA receptor desensitization and activation of metabotropic receptors therefore do not contribute significantly to PTD in our recording conditions. It has been suggested that PTD could arise from action potential propagation failures at axonal branch points during repetitive stimulation (Brody and Yue 2000). For example, Ca 2 that builds up during repetitive stimulation could activate Ca 2 -dependent K or Cl channels. Ensuing hyperpolarization could then impede action potential propagation or simply reduce the extent of depolarization at the synapse (Lüscher et al. 1996). Such modulation of action potential propagation could result in a progressive reduction of EPSC amplitudes. To test this possibility, PTD was measured in the presence of blockers of Ca 2 -dependent K or Cl channels (Fig. 3, C E). Apamin (1 M), which blocks the small conductance K channels, had no effect on the rate of PTD. Nevertheless, 1 mm TEA, which blocks the big conductance channels, greatly

2328 D. J. HAGLER AND Y. GODA of Ca 2 -dependent K or Cl channels contributing to PTD by promoting action potential propagation failures. FIG. 3. Effects of various drugs on PTD. Normalized EPSC amplitudes (20 pulses, 20 Hz) in control conditions ( ) are compared with responses in the presence of the following drugs (E): 100 M cyclothiazide (A, n 4), 250 M (S)- -methyl-4-carboxyphenylglycine (MCPG, B, n 4), 1 M apamin (C, n 4), 1 mm TEA (D, n 5), and 10 M 5-nitro-2-(3-phenyl-propylamino) benzoic acid (NPPB, E, n 5). Recordings were made in high Ca 2 for A and B and in medium Ca 2 for C E. increased the rate of PTD. The TEA-dependent increase in PTD rate most likely resulted from higher release probability caused by enhanced Ca 2 influx, similar to the effects of elevating extracellular Ca 2 (Katz and Miledi 1966; Kusano et al. 1967) (see following text). NPPB (10 M), a chloride channel blocker, had no effect on PTD; when applied at a higher concentration (100 M), NPPB slightly increased the rate of PTD (n 3, data not shown). A reduction in the rate of PTD was not observed by blocking Ca 2 -dependent K or Cl channels. Our results, therefore are inconsistent with activation Depression of EPSC amplitude depends on accumulation of intracellular Ca 2 The dependence of PTD on the levels of extracellular Ca 2 and stimulus frequency both of which determine the intracellular Ca 2 concentration attained during repetitive stimulation (Tank et al. 1995) suggests that the accumulation of intracellular Ca 2 contributes to the depression of synaptic strength. To test this hypothesis, PTD was measured following bath application of 100 M EGTA-AM to chelate intracellular Ca 2. EGTA-AM caused a slight but significant decrease in the initial EPSC size (78 5% of control; P 0.002, 2-tailed, paired t-test; Fig. 10C), presumably because of the slight reduction in peak Ca 2 concentration brought about this slow chelator (Atluri and Regehr 1996). In contrast to responses recorded in control condition, which gradually decreased toward zero, EGTA-AM greatly attenuated the extent of PTD, most noticeably later in the train when responses reached a plateau (Fig. 4E). Furthermore, the stimulus frequency and extracellular Ca 2 dependence of PTD were greatly reduced in the presence of EGTA-AM (compare Figs. 1 and 4, A D). A buildup of intracellular Ca 2 during repetitive stimulation is thus likely to play a role in the mechanism of PTD. Decrease in synchronous release is matched by increase in asynchronous release We next characterized the properties of asynchronous release during repetitive stimulation in relation to the depression of synchronous release. Asynchronous release was examined by measuring late release that which follows the decay of synchronous release by integrating the current between 40 and 50 ms after each pulse (see METHODS). As expected from its Ca 2 dependence, more asynchronous release was initially observed in high extracellular Ca 2 (Fig. 5, A C). Elevated Ca 2, as well as higher stimulus frequency, also promoted a faster rise in asynchronous release during the pulse train. Once a maximal level was reached, asynchronous release remained constant or even declined slightly. This plateau level was higher with increased stimulus frequency or Ca 2 concentration (Fig. 5, A C). We then examined the relative changes in synchronous and asynchronous release with increasing stimulus number. The fractional contribution of asynchronous release to the total release sum of synchronous and asynchronous release (see METHODS) was initially minimal at each extracellular Ca 2 concentration tested. In low Ca 2 the contribution of asynchronous release increased gradually during the pulse train (Fig. 5F). In high Ca 2, however, there was a complete or nearly complete shift to asynchronous release, especially at higher stimulus frequency (Fig. 5F). Despite the differences in the relative contributions of synchronous and asynchronous release at different extracellular Ca 2, at higher stimulus frequency the total release per pulse tended to eventually reach the same value, regardless of extracellular Ca 2 concentration (Fig. 5E). The sensitivity of asynchronous release to intracellular Ca 2 is illustrated in Fig. 6. Bath application of EGTA-AM largely blocked the rise in asynchronous release that normally devel-

TRANSMITTER RELEASE DURING PULSE TRAIN DEPRESSION 2329 FIG. 4. Effect of EGTA-AM on PTD. A: EPSC amplitudes after EGTA-AM (abbreviated as E) treatment in response to 20 pulses at 20 Hz in low ( ), medium ( ), and high (Œ) Ca 2. Recordings from a given cell were made in high Ca 2 as well as in low (n 3) or medium (n 7) Ca 2, and responses were normalized to the 1st response of the train in high Ca 2 (n 10). B D: normalized EPSC amplitudes in response to 5 ( ), 10 ( ), and 20 (Œ) Hz after treatment with EGTA-AM in low (B, n 4), medium (C, n 11), and high (D, n 11 for 5 Hz, n 17 for 10 and 20 Hz) Ca 2. E: responses before ( ) and 10 min after (E) bath application of 100 M EGTA-AM (in high Ca 2 ) normalized to the 1st response of the train without EGTA-AM (n 11). ops during repetitive stimulation (Figs. 6A and 10C). In contrast, synchronous release reached a higher plateau in EGTA-AM relative to control (Fig. 4E). To determine whether the sustained synchronous release observed in EGTA-AM can be accounted for by the reduction in asynchronous release, we compared the total release per pulse before and after EGTA-AM treatment. In high extracellular Ca 2, by the end of the train total release is approximately equal between that measured with or without EGTA-AM despite greatly reduced synchronous release in control conditions (Figs. 4E and 6C). This demonstrates that part of the decrease in synchronous release during a pulse train is accompanied by a matching increase in asynchronous release. Effect on PTD of increasing asynchronous release with extracellular Sr 2 The preceding results suggest that depression of synchronous release involves a shift to asynchronous release. To examine how tightly coupled these two forms of release are, we substituted extracellular Ca 2 with Sr 2, a manipulation that increases asynchronous release while decreasing synchronous release. Sr 2 substitution of high extracellular Ca 2 decreased the initial EPSC amplitude by half while increasing the initial asynchronous release by about twofold (Figs. 7, A and B, and 10A). At 20 Hz, the rate of depression of the EPSC amplitude was similar in Ca 2 and Sr 2. At 5 and 10 Hz, however, the initial depression on the second pulse (pairedpulse depression: PPD) was significantly greater in Sr 2 than in Ca 2 (extent of depression was 71 11 and 21 5% greater for 5 and 10 Hz, respectively, n 6, P 0.01 for both, 2-tailed, paired t-test). As in the presence of Ca 2, asynchronous release reached a plateau for each of these frequencies tested; however, the plateau levels reached in Sr 2 were somewhat higher than in Ca 2 (for example, in Fig. 7B, 20-Hz plateau was 45 15% greater in Sr 2, n 6, P 0.03, 2-tailed, paired t-test). As a result, by the end of the train, the total release evoked per pulse was greater in Sr 2 than in Ca 2 by 31 5% at 20 Hz (Fig. 7C: P 0.01, 2-tailed, paired t-test) and 48 17% at 5 Hz (Fig. 7F: P 0.04, 2-tailed, paired t-test). We also examined the effects of Sr 2 substitution of medium Ca 2. The initial EPSC amplitude was decreased to one-fifth of the original size, and the initial asynchronous release was slightly elevated. Although PPD was observed at 10 and 20 Hz in Ca 2, we found facilitation at those frequencies in Sr 2 (Fig. 8, A and D), consistent with a reduction in release probability in Sr 2 (Xu-Friedman and Regehr 2000). Subsequent PTD was largely unaffected (compare Figs. 1C and 8D). In contrast to Sr 2 replacement of high Ca 2, the increase in asynchronous release with Sr 2 replacement of medium Ca 2 was not sufficient to cause a greater total release per pulse (Fig. 8, C and F). At 20-Hz stimulation, the differences in asynchronous and total release between Ca 2 and Sr 2 became very small by the end of the train. Effect on PTD of prolonged Ca 2 influx While testing for a possible contribution of Ca 2 -dependent K channels, we found that TEA caused a large increase in the rate of PTD. This result is readily explained by the broadening of the action potential caused by blockade of K channels and the subsequent increase in Ca 2 influx (Katz and Miledi 1966; Kusano et al. 1967). TEA increased both the initial synchronous and asynchronous release (Figs. 9, A and B, and 10B). Total release per pulse was in general greater in TEA; however, at 20 Hz this difference quickly diminished (Fig. 9, C and F). As discussed in the preceding text, the plateau level of

2330 D. J. HAGLER AND Y. GODA FIG. 5. Asynchronous and total release during pulse trains with varying extracellular Ca 2 and stimulation frequency. In A and B, the asynchronous release the late charge transfer measured 40 50 ms following each pulse is shown for 5 ( ), 10 ( ), and 20 (Œ) Hz stimulus trains in medium (A, n 14 for 5 Hz and n 25 for 10 and 20 Hz) and high (B, n 17 for 5 Hz and n 27 for 10 and 20 Hz) Ca 2. For A and B, measurements are normalized to the late charge transfer for the 20th response of the train at 20 Hz. C: late charge transfer for 20-Hz stimulation in low ( ), medium ( ), and high (Œ)Ca 2 ; recordings from a given cell were made in high Ca 2 as well as in low (n 3) or medium (n 8) Ca 2, and responses were normalized to the last response of the train in high Ca 2 with 20-Hz stimulation. High Ca 2 responses in C are the pooled data from the 11 cells. D and E: the total release the charge transfer measured from 5 ms after each pulse to just before the next pulse in low ( ), medium ( ), and high (Œ) Ca 2,with5(D)- and 20 (E)-Hz stimulation. F: fractional contribution of asynchronous release to the total release (see METHODS for normalization procedures) for each pulse at 20 Hz in low ( ; n 5), medium ( ; n 25), and high (Œ; n 27) Ca 2. asynchronous release later in the pulse train increases with higher stimulus frequency (Fig. 5A). In the presence of TEA, the frequency dependence of the plateau reached by asynchronous release was greatly reduced (Fig. 9E). In addition, TEA caused an increase in the maximal asynchronous release (Fig. 9B). During 20-Hz stimulation, asynchronous release was 49 17% greater in TEA on the 10th pulse (n 5, P 0.05, 2-tailed, paired t-test) but only 23 17% greater on the 20th pulse (P 0.25, 2-tailed, paired t-test). DISCUSSION We studied the properties of synchronous and asynchronous release during repetitive stimulation to gain insight into the underlying mechanisms of PTD and the pool of vesicles that supplies the two components of release. In view of its extracellular Ca 2 and frequency dependence, the progressive decline in synchronous release is consistent with depletion of available vesicles. Nevertheless, RRP recovery following complete depression of synchronous release was faster than the expected refilling of an empty pool, suggesting that the RRP is not fully depleted during PTD. Thus it is plausible that depletion of the RRP is not the only source of depression of synchronous release. We have also shown that synchronous and asynchronous releases are coordinately regulated such that manipulations that increase the rate of decline of synchronous release elevate asynchronous release. The sum of synchronous and asynchronous release was similar between control and EGTA-AM, despite the differential effects of EGTA-AM on the two components of release. We use these data to advance a model describing an immediately releasable vesicle which supplies and is depleted by both forms of release. FIG. 6. Asynchronous and total release following EGTA-AM treatment. A: asynchronous release for 20-Hz stimulation before ( and E) and after ( and ) EGTA-AM treatment in medium (E and ; n 5) and high ( and ; n 11) Ca 2, normalized to the last response of the train with 20-Hz stimulation before EGTA-AM treatment. B and C: the total release before ( ) and after (E) EGTA-AM treatment in medium (B, n 5) and high (C, n 11) Ca 2 with 20-Hz stimulation, normalized to the 1st response of the train before EGTA-AM treatment.

TRANSMITTER RELEASE DURING PULSE TRAIN DEPRESSION 2331 FIG. 7. Synchronous, asynchronous, and total release in high Sr 2 vs. high Ca 2. A C: the EPSC amplitude, asynchronous release, and total release are compared for high Sr 2 (E) and high Ca 2 ( ) with 20-Hz stimulation. In A and C, each response is normalized to the 1st response of the train in high Ca 2.InB, each response is normalized to the last response of the train in high Ca 2. D and E: EPSC amplitudes and asynchronous release are shown, respectively, for 5 ( )-, 10 ( )-, and 20 (Œ)-Hz stimulation in high Sr 2. D: responses are normalized to the 1st response of the train. E: responses are normalized to the last response of the train in high Ca 2 with 20-Hz stimulation. F compares the total release in high Sr 2 (E) and high Ca 2 ( ) with 5-Hz stimulation. Responses are normalized to the 1st response of the train in high Ca 2 (n 6). Possible factors governing depression of synchronous release Since PTD likely results from the combination of several processes, we have sought to rule out several potential mechanisms to simplify the analysis of neurotransmitter release properties during PTD. We have ruled out desensitization of AMPA receptors because CTZ has no significant effect on PTD under the present experimental conditions as reported for several other preparations (Bellingham and Walmsley 1999; Dittman and Regehr 1998; Dobrunz and Stevens 1997; Galarreta and Hestrin 1998; Varela et al. 1997). Metabotropic glutamate receptors do not contribute to PTD as an antagonist for these receptors had no effect, similar to a previous demonstration in hippocampal slices (Dobrunz and Stevens 1997). Furthermore, earlier work showed that PPD was not affected by MCPG in hippocampal cultures (Maki et al. 1995). We were unable to test the role of Ca 2 channel inactivation in PTD; however, such a role is expected to be minor since the fractional reduction in Ca 2 currents is relatively small compared with the extent of PTD (Forsythe et al. 1998). Furthermore, closed-state Ca 2 channel inactivation is not sensitive to buffering of Ca 2 (Patil et al. 1998) unlike the PTD observed here. Another potential mechanism contributing to PTD that de- FIG. 8. Synchronous, asynchronous, and total release in medium Sr 2 vs. medium Ca 2. A C: the EPSC amplitude, asynchronous release, and total release are compared for medium Sr 2 (E) and medium Ca 2 ( ) with 20-Hz stimulation. A and C: each response is normalized to the 1st response of the train in medium Ca 2. B: each response is normalized to the last response of the train in medium Ca 2. D and E: EPSC amplitudes and asynchronous release are shown, respectively, for 5 ( )-, 10 ( )-, and 20 (Œ)-Hz stimulation in medium Sr 2. D: responses are normalized to the 1st response of the train. E: responses are normalized to the last response of the train in medium Ca 2 with 20-Hz stimulation. F compares the total release in medium Sr 2 (E) and medium Ca 2 ( ) with 5-Hz stimulation. Responses are normalized to the 1st response of the train in medium Ca 2 (n 3).

2332 D. J. HAGLER AND Y. GODA FIG. 9. Synchronous, asynchronous, and total release with and without TEA. A C: the EPSC amplitude, asynchronous release, and total release are compared for medium Ca 2 ( ) and medium Ca 2 plus TEA (E) with 20-Hz stimulation. A and C: each response is normalized to the 1st response of the train in medium Ca 2. B: each response is normalized to the last response of the train in medium Ca 2. D and E: EPSC amplitudes and asynchronous release are shown, respectively, for 5 ( ; n 3)-, 10 ( ; n 5)-, and 20 (closed triangles; n 5)-Hz stimulation in medium Ca 2 plus TEA. D: responses are normalized to the 1st response of the train. E: responses are normalized to the last response of the train in medium Ca 2 with 20-Hz stimulation. F compares the total release in medium Ca 2 ( ) and medium Ca 2 plus TEA (E) with 5-Hz stimulation. Responses are normalized to the 1st response of the train in medium Ca 2. serves attention is alteration of action potential (AP) propagation; this could include a reduction in AP amplitude or actual AP conduction failure at axonal branch points (Brody and Yue 2000; Hatt and Smith 1976; Streit et al. 1992). If AP amplitude is progressively reduced, or if the frequency of AP conduction failures increases with repetitive stimulation, it is expected that the EPSC amplitude would decrease during a pulse train. Although AP conduction in hippocampal axons is highly reliable, even for pairs of pulses (Mackenzie and Murphy 1998; Stevens and Wang 1995), a reduction in AP amplitude in the second of a pair of stimuli has been reported with a recovery time constant of 8 ms (Brody and Yue 2000). Such a fast recovery rate should preclude significant effects on PTD at the frequencies tested in the current study. Nonetheless, a potential mechanism for reducing AP amplitude is activation of Ca 2 - dependent K or Cl channels (Lüscher et al. 1996). A blockade of these channels might be expected to decrease the rate of PTD. PTD, however, was either unaffected or increased, suggesting that this type of mechanism is unlikely to be involved in the EGTA-sensitive component of PTD. Presently, we cannot completely exclude the possible contribution to PTD by channels that are insensitive to the drugs tested. Finally, it has also been proposed that high extracellular Ca 2 promotes conduction failure by reduced excitability resulting from charge screening effects (Brody and Yue 2000). That the Ca 2 dependence of PTD is sensitive to buffering intracellular Ca 2 with EGTA demonstrates that changes in membrane excitability by charge screening effects is not a significant issue. Coordinate regulation of synchronous and asynchronous release FIG. 10. Examples of pulse train responses in single neurons. A: comparison of responses in high Ca 2 (gray line) and high Sr 2 (black line). B: comparison of responses in medium Ca 2 with (black line) and without (gray line) TEA. C: comparison of responses in high Ca 2 with (black line) and without (gray line) EGTA-AM. In each panel, the 1st 4 pulses of a 20-pulse train are shown; stimulus artifacts have been removed. We find that manipulations that increase Ca 2 influx accelerate both the depression of synchronous release and the increase in asynchronous release with subsequent stimulus pulses. We also observe that the sum of synchronous and asynchronous release at the end of the train was similar between control and EGTA-AM. These results indicate an interesting scheme whereby elevating asynchronous release

TRANSMITTER RELEASE DURING PULSE TRAIN DEPRESSION 2333 depresses synchronous release by depleting the vesicles immediately available for synchronous release. Such a proposal implies that the two forms of release share a common pool of releasable vesicles. Our observations are similar to those reported recently for inhibitory synapses in nucleus magnocellularis where the enhanced contribution of asynchronous release during high-frequency synaptic transmission may play a significant role in auditory processing (Lu and Trussell 2000). The fact that interdependent regulation of synchronous and asynchronous release also occurs at hippocampal excitatory synapses suggests that it might be a general property expressed by all central synapses. It is of interest to note that our results differ from a previous study in cultured hippocampal neurons in which synchronous release monitored in the presence of EGTA-AM did not reach a significantly higher plateau at the end of the stimulus train compared with control (Cummings et al. 1996). Likely reasons for the apparent disagreement are the lower range of extracellular Ca 2 levels and shorter stimulus trains used in the earlier study. Model for a common pool supplying synchronous and asynchronous release Based on the coordinate regulation of the two components of release we propose that synchronous and asynchronous release share a common pool of releasable vesicles. We describe a model that has been refined within the framework of our observations and those made by others. RELEASE OCCURS FROM AN IRP CONSISTING OF A SINGLE QUANTUM. Studies of transmitter release properties at single central synapses indicate that neurotransmitter release is limited to at most a single quantum (Arancio et al. 1994; Edwards et al. 1976; Stevens and Wang 1995). A plausible explanation for the release of just a single vesicle is that the IRP, which supplies synchronous release, represents a small subset of the RRP consisting of a single, most primed vesicle (Matveev and Wang 2000). In our model, synchronous release occurs exclusively from the IRP, and asynchronous release can also gain access to the IRP (see Fig. 11). This model accounts for features of short-term plasticity of synchronous release. An IRP that is limited to one vesicle will produce prominent PPD under conditions of high release probability as found in this and other studies (Fig. 1) (Brenowitz et al. 1998; Dittman and Regehr 1998; Thomson 1997; Varela et al. 1997). Paired-pulse facilitation occurs with low initial release probability if, at rest, the fraction of synapses with an occupied IRP is less than one and refilling accelerates upon stimulation. Our model accounts for the increased PPD of synchronous release in particular for 5- and 10-Hz stimulation in high extracellular Sr 2. Although it would be expected that greatly reduced initial release probability should result in less depression, we observe increased PPD (compare Figs. 1D and 7D). Assuming asynchronous release can also gain access to the IRP, elevated levels of asynchronous release would deplete the IRP, thereby preventing that synapse from later releasing synchronously with the action potential (Fesce 1999). The PPD that we observed in EGTA-AM, however, cannot be fully explained by our model without an additional assumption. We postulate that a Ca 2 -dependent process accelerates refilling, or priming, of the IRP. Slower IRP refilling in the FIG. 11. Schematic representation of multiple pools of synaptic vesicles. The immediately releasable pool (IRP) is shown as a single vesicle (E) that can undergo Ca 2 -dependent fusion. The IRP is replenished by priming from the readily releasable pool (RRP; ). This priming step may not require Ca 2 per se but is accelerated by Ca 2 (Dittman and Regehr 1998). The RRP in turn is replenished by the docking of reserve pool (RP; gray circles) vesicles to the plasma membrane. This docking step presumably also involves an early priming step that enables hypertonic-evoked release. Together, these steps are likely involved in the refilling of the RRP following depletion by hypertonic solution. Similar to the filling of the IRP, filling of the RRP does not strictly require Ca 2 but is accelerated by Ca 2 (Stevens and Wesseling 1998). presence of EGTA-AM, therefore reduces the likelihood of occupancy of the IRP; thus PPD is increased despite the slight reduction in release probability. Consistent with this model, a previous study has also shown that EGTA-AM slows the recovery from PPD (Dittman and Regehr 1998). Moreover, although depression of synchronous release in control conditions quickly overtakes that observed in EGTA-AM in particular with 20-Hz stimulation in high Ca 2 total release is greater in control conditions throughout most of the pulse train. A gradual decline of total release in the control condition may reflect the slow depletion of the RRP, which is likely to affect the rate of refilling of the IRP. MECHANISMS UNDERLYING ADDITIONAL INCREASES IN ASYNCHRO- NOUS RELEASE. We observed that while asynchronous release increases to match a portion of the depression of synchronous release, asynchronous release could be increased further in the presence of TEA or by extracellular substitution of Ca 2 by Sr 2. Both TEA and Sr 2 are expected to greatly increase the residual divalent concentrations; TEA because of prolonged Ca 2 influx (Katz and Miledi 1966; Kusano et al. 1967), and Sr 2 because of slower clearance (Xu-Friedman and Regehr 1999). This suggests that while synchronous release may be limited to a single, most primed vesicle (IRP), asynchronous release can, under special conditions such as in the presence of TEA or high Sr 2 access a slightly larger pool. We suggest two alternative mechanisms to account for this greater access. In the first scenario, asynchronous release can access vesicles other than the IRP in the RRP. Such non-irp vesicles normally have a much lower affinity for Ca 2 or Sr 2 and require much greater concentrations to undergo fusion. Asynchronous release, therefore can access these vesicles under the extreme conditions achieved with TEA or Sr 2. In the second case, very high Ca 2 or Sr 2 further accelerate the rate of

2334 D. J. HAGLER AND Y. GODA priming, thereby increasing the occupancy of the IRP. In this mechanism, synchronous and asynchronous release share the identical pool of releasable vesicles. In either case, the implication of our model is that under normal conditions the IRP is the primary supplier of vesicles for both synchronous and asynchronous release. We thank Dr. Silvio Varon for generously providing equipment support, G. Ramirez for excellent technical assistance, and Dr. Marla Feller and members of the Goda lab for helpful comments on the manuscript. This work was supported by grants from the National Institute of Mental Health (MH-57710) and the Whitehall Foundation to Y. Goda. D. J. Hagler was supported by a training grant from the National Institutes of Health. REFERENCES ARANCIO O, KORN H, GULYAS A, FREUND T, AND MILES R. Excitatory synaptic connections onto rat hippocampal inhibitory cells may involve a single transmitter release site. J Physiol (Lond) 481: 395 405, 1994. ATLURI PP AND REGEHR WG. Determinants of the time course of facilitation at the granule cell to Purkinje cell synapse. J Neurosci 16: 5661 5671, 1996. BARRETT EF AND STEVENS CF. The kinetics of transmitter release at the frog neuromuscular junction. J Physiol (Lond) 227: 691 708, 1972. BEKKERS JM AND STEVENS CF. Excitatory and inhibitory autaptic currents in isolated hippocampal neurons maintained in cell culture. Proc Natl Acad Sci USA 88: 7834 7838, 1991. BELLINGHAM MC AND WALMSLEY B. A novel presynaptic inhibitory mechanism underlies paired pulse depression at a fast central synapse. Neuron 23: 159 170, 1999. BRENOWITZ S, DAVID J, AND TRUSSELL L. Enhancement of synaptic efficacy by presynaptic GABA(B) receptors. Neuron 20: 135 141, 1998. BRODY DL AND YUE DT. Release-independent short-term synaptic depression in cultured hippocampal neurons. J Neurosci 20: 2480 2494, 2000. CUMMINGS DD, WILCOX KS, AND DICHTER MA. Calcium-dependent pairedpulse facilitation of miniature EPSC frequency accompanies depression of EPSCs at hippocampal synapses in culture. J Neurosci 16: 5312 5323, 1996. DITTMAN JS AND REGEHR WG. Calcium dependence and recovery kinetics of presynaptic depression at the climbing fiber to Purkinje cell synapse. J Neurosci 18: 6147 6162, 1998. DOBRUNZ LE AND STEVENS CF. Heterogeneity of release probability, facilitation, and depletion at central synapses. Neuron 18: 995 1008, 1997. EDWARDS FR, REDMAN SJ, AND WALMSLEY B. Statistical fluctuations in charge transfer at Ia synapses on spinal motoneurones. J Physiol (Lond) 259: 665 688, 1976. FESCE R. The kinetics of nerve-evoked quantal secretion. Philos Trans R Soc Lond B Biol Sci 354: 319 329, 1999. FORSYTHE ID, TSUJIMOTO T, BARNES-DAVIES M, CUTTLE MF, AND TAKAHASHI T. Inactivation of presynaptic calcium current contributes to synaptic depression at a fast central synapse. Neuron 20: 797 807, 1998. GALARRETA M AND HESTRIN S. Frequency-dependent synaptic depression and the balance of excitation and inhibition in the neocortex. Nat Neurosci 1: 587 594, 1998. GEPPERT M, GODA Y, HAMMER RE, LI C, ROSAHL TW, STEVENS CF, AND SÜDHOF TC. Synaptotagmin I: a major Ca 2 sensor for transmitter release at a central synapse. Cell 79: 717 727, 1994. GODA Y AND STEVENS CF. Two components of transmitter release at a central synapse. Proc Natl Acad Sci USA 91: 12942 12946, 1994. HATT H AND SMITH DO. Synaptic depression related to presynaptic axon conduction block. J Physiol (Lond) 259: 367 393, 1976. HEMPEL CM, HARTMAN KH, WANG X-J, TURRIGIANO GG, AND NELSON SB. Multiple forms of short-term plasticity at excitatory synapses in rat medial prefrontal cortex. J Neurophysiol 83: 3031 3041, 2000. HSU SF, AUGUSTINE GJ, AND JACKSON MB. Adaptation of Ca(2 )-triggered exocytosis in presynaptic terminals. Neuron 17: 501 512, 1996. HUBEL DH. Single unit activity in striate cortex of unrestrained cats. J Physiol (Lond) 147: 226 238, 1959. JANZ R, GODA Y, GEPPERT M, MISSLER M, AND SÜDHOF TC. SV2A and SV2B function as redundant Ca 2 regulators in neurotransmitter release. Neuron 24: 1003 1016, 1999. KATZ B AND MILEDI R. Input-output relation of a single synapse. Nature 212: 1242 1245, 1966. KUSANO K, LIVENGOOD DR, AND WERMAN R. Tetraethylammonium ions: effect of presynaptic injection on synaptic transmission. Science 155: 1257 1259, 1967. LILEY AW AND NORTH KAK. An electrical investigation of effects of repetitive stimulation on mammalian neuromuscular junction. J Neurophysiol 16: 509 527, 1953. LU T AND TRUSSELL LO. Inhibitory transmission mediated by asynchronous transmitter release. Neuron 26: 683 694, 2000. LÜSCHER C, LIPP P, LÜSCHER HR, AND NIGGLI E. Control of action potential propagation by intracellular Ca 2 in cultured rat dorsal root ganglion cells. J Physiol (Lond) 490: 319 324, 1996. MACKENZIE PJ AND MURPHY TH. High safety factor for action potential conduction along axons but not dendrites of cultured hippocampal and cortical neurons. J Neurophysiol 80: 2089 2101, 1998. MAKI R, CUMMINGS DD, AND DICHTER MA. Frequency-dependent depression of excitatory synaptic transmission is independent of activation of MCPGsensitive presynaptic metabotropic glutamate receptors in cultured hippocampal neurons. J Neurophysiol 74: 1671 1674, 1995. MATVEEV V AND WANG XJ. Implications of all-or-none synaptic transmission and short-term depression beyond vesicle depletion: a computational study. J Neurosci 20: 1575 1588, 2000. MEIRI U AND RAHAMIMOFF R. Neuromuscular transmission: inhibition by manganese ions. Science 176: 308 309, 1972. MILEDI R. Strontium as a substitute for calcium in the process of transmitter release at the neuromuscular junction. Nature 212: 1233 1234, 1966. O KEEFE J AND DOSTROVSKY J. The hippocampus as a spatial map. Preliminary evidence from unit activity in the freely-moving rat. Brain Res 34: 171 175, 1971. PATIL PG, BRODY DL, AND YUE DT. Preferential closed-state inactivation of neuronal calcium channels. Neuron 20: 1027 1038, 1998. PYLE JL, KAVALALI ET, PIEDRAS-RENTERIA ES, AND TSIEN RW. Rapid reuse of readily releasable pool vesicles at hippocampal synapses. Neuron 28: 221 231, 2000. ROSENMUND C AND STEVENS CF. Definition of the readily releasable pool of vesicles at hippocampal synapses. Neuron 16: 1197 1207, 1996. SCANZIANI M, SALIN PA, VOGT KE, MALENKA RC, AND NICOLL RA. Usedependent increases in glutamate concentration activate presynaptic metabotropic glutamate receptors. Nature 385: 630 634, 1997. SEGAL MM AND FURSHPAN EJ. Epileptiform activity in microcultures containing small numbers of hippocampal neurons. J Neurophysiol 64: 1390 1399, 1990. STEVENS CF AND TSUJIMOTO T. Estimates for the pool size of releasable quanta at a single central synapse and for the time required to refill the pool. Proc Natl Acad Sci USA 92: 846 849, 1995. STEVENS CF AND WANG Y. Facilitation and depression at single central synapses. Neuron 14: 795 802, 1995. STEVENS CF AND WESSELING JF. Activity-dependent modulation of the rate at which synaptic vesicles become available to undergo exocytosis. Neuron 21: 415 424, 1998. STREIT J, LÜSCHER C, AND LÜSCHER HR. Depression of postsynaptic potentials by high-frequency stimulation in embryonic motoneurons grown in spinal cord slice cultures. J Neurophysiol 68: 1793 1803, 1992. TAKAHASHI T, FORSYTHE ID, TSUJIMOTO T, BARNES-DAVIES M, AND ONODERA K. Presynaptic calcium current modulation by a metabotropic glutamate receptor. Science 274: 594 597, 1996. TANK DW, REGEHR WG, AND DELANEY KR. A quantitative analysis of presynaptic calcium dynamics that contribute to short-term enhancement. J Neurosci 15: 7940 7952, 1995. THOMSON AM. Activity-dependent properties of synaptic transmission at two classes of connections made by rat neocortical pyramidal axons in vitro. J Physiol (Lond) 502: 131 147, 1997. TSODYKS MV AND MARKRAM H. The neural code between neocortical pyramidal neurons depends on neurotransmitter release probability. Proc Natl Acad Sci USA 94: 719 723, 1997. VARELA JA, SEN K, GIBSON J, FOST J, ABBOTT LF, AND NELSON SB. A quantitative description of short-term plasticity at excitatory synapses in layer 2/3 of rat primary visual cortex. J Neurosci 17: 7926 7940, 1997. WU LG AND BORST JG. The reduced release probability of releasable vesicles during recovery from short-term synaptic depression. Neuron 23: 821 832, 1999. XU-FRIEDMAN MA AND REGEHR WG. Presynaptic strontium dynamics and synaptic transmission. Biophys J 76: 2029 2042, 1999. XU-FRIEDMAN MA AND REGEHR WG. Probing fundamental aspects of synaptic transmission with strontium. J Neurosci 20: 4414 4422, 2000.