Ca²⁺ influx–independent synaptic potentiation mediated by mitochondrial Na⁺-Ca²⁺ exchanger and protein kinase C

Activity-dependent modulation of synaptic transmission is a fundamental mechanism for the development and functions of the brain. Information processing, which is coupled tightly with firing of neuronal action potentials at different frequencies, often leads to changes in the efficacy of transmission lasting for a short or long period of time, a general phenomenon known as synaptic plasticity. Experimentally, this could be mimicked by repetitive stimulation of presynaptic neurons. Long-term forms of synaptic plasticity that last for at least 1 h, such as hippocampal long-term potentiation (LTP), are mediated mostly by postsynaptic mechanisms (Luscher et al., 2000). Short-term forms of synaptic plasticity lasting from seconds to minutes are often due to changes in presynaptic transmitter secretion (Zucker and Regehr, 2002). At the neuromuscular junction (NMJ), a brief, high-frequency stimulation results in an enhancement of transmitter release that lasts many minutes, called post-tetanic potentiation (PTP). Recently, tetanic stimulation has also been shown to induce a long-lasting enhancement of transmission similar to LTP at the neuromuscular synapses in Xenopus nerve-muscle cultures (Wan and Poo, 1999).

Although it is generally believed that presynaptic forms of plasticity are due to a prolonged elevation of intracellular concentrations of free Ca²⁺ ([Ca²⁺]_i) after the termination of tetanus, the precise molecular mechanisms for the enhancement of transmitter secretion remain unclear. At the resting nerve terminals, [Ca²⁺]_i is generally <100 nM. This is accomplished by the Ca²⁺ ATPase, which effectively pumps Ca²⁺ out of the terminals, and by the plasmalemmal Na⁺-Ca²⁺ exchanger, which allows entry of three Na⁺ in exchange for the efflux of one Ca²⁺ (Blaustein and Lederer, 1999; Garcia and Strehler, 1999). The key step in triggering transmitter secretion is an elevation of terminal [Ca²⁺]_i. This could be achieved by a number of mechanisms. First, a well-established mechanism is the action potential–driven membrane depolarization, leading to Ca²⁺ influx through voltage-gated Ca²⁺ channels. The second and more complex mechanism is the release of Ca²⁺ from intracellular organelles. One class of such organelles is the ER. Two types of ligand-gated Ca²⁺ channels are involved in Ca²⁺ release from the ER: the IP3 receptor, operated by inositol 1,4,5-trisphosphate (IP3), and the ryanodine receptor, gated by Ca²⁺ as well as cyclic ADP ribose (Berridge, 1998). Although still a fairly new concept, transmitter secretion triggered or modulated by Ca²⁺ release from the ER has been shown in a number of synapses (Smith and Cunnane, 1996; Cochilla and Alford, 1998; Mothet et al., 1998; Yang et al., 2001). The other class of organelles is mitochondria, which represents a transient storage mechanism for Ca²⁺. An accumulation of Ca²⁺ in the mitochondria induced by certain stimuli is released when the stimulus is terminated (Tang and Zucker, 1997; Melamed-Book and Rahamimoff, 1998). Under physiological conditions, mitochondrial Ca²⁺ release is achieved primarily by the Na⁺-Ca²⁺ exchanger on the mitochondrial membranes. It has been recently shown that the massive secretion of transmitters at the NMJ induced by α-latrotoxin is mediated by the mitochondrial Na⁺-Ca²⁺ exchanger (Tsang et al., 2000). Finally, when cells are overloaded with Na⁺ and extracellular Ca²⁺ is high, the plasmalemmal Na⁺-Ca²⁺ exchanger may operate in a “reverse mode” to allow Ca²⁺ entry into the cells (Zhong et al., 2001).

Many forms of activity-dependent synaptic plasticity require Ca²⁺ influx. Using a cultured neuromuscular synapse preparation in which Ca²⁺ influx has been completely prohibited, we report here a novel form of synaptic plasticity that would be difficult to reveal in normal circumstances. A train of tetanic stimulation induces a robust potentiation of neurotransmitter release, as well as an increase in [Ca²⁺]_i, at the developing NMJ in the absence of extracellular Ca²⁺. Detailed analyses using both pharmacological and molecular approaches indicate that this synaptic potentiation is independent of Ca²⁺ release from ER ryanodine or IP3 receptors, but requires Na⁺ influx. The increase in Na⁺ concentration in the nerve terminals triggers Ca²⁺ efflux through the mitochondrial Na⁺-Ca²⁺ exchanger, leading to the tetanus-induced synaptic potentiation (TISP). In addition, inhibition of PKC dramatically attenuated TISP as well as mitochondrial Ca²⁺ release. We also show that blockade of the mitochondrial Na⁺-Ca²⁺ exchanger inhibits the synaptic potentiation and [Ca²⁺]_i increase in normal extracellular Ca²⁺. Thus, this form of synaptic plasticity may occur during the bursting activity at the NMJ in vivo. Our studies may also help understand the contribution of mitochondria and PKC in transmitter release and provide a useful model to investigate molecular mechanisms for transmitter release without the interference of Ca²⁺ influx.

Spontaneous synaptic currents (SSCs) were recorded from innervated myocytes in 1-d-old Xenopus nerve-muscle cocultures (e.g., Fig. 1 D) under whole-cell, voltage-clamp conditions. Stimulation of the presynaptic motoneurons with a train of repetitive, high-frequency stimuli (or tetanus, 50 Hz, 10 s) elicited a striking potentiation of synaptic transmission. The frequency of SSCs increased more than 100 times immediately after the tetanus (Fig. 1 A). The average amplitudes of SSCs before and after the tetanus were not significantly different, suggesting that this form of synaptic plasticity is due primarily to an enhancement of presynaptic transmitter release (unpublished data). Surprisingly, the same tetanus induced a very similar synaptic potentiation in the complete absence of Ca²⁺ influx. Ca²⁺-free conditions were achieved by using medium containing 0 mM extracellular Ca²⁺ ([Ca²⁺]_o) plus 3 mM EGTA. Under these conditions, tetanus still elicited a marked enhancement of synaptic transmission (Fig. 1 B). Similar results were obtained in zero [Ca²⁺]_o plus 0.4 mM Cd²⁺ to block all voltage-gated Ca²⁺ channels (Fig. 1 C, right). The average frequency of SSCs increased by more than 60-fold. The onset of the potentiation was slightly slower (Fig. 1 B). It is important to note that whereas the tetanic stimulation induced evoked synaptic currents (ESCs) of various amplitudes in normal medium (Fig. 1 A, inset), the same tetanus elicited absolutely no ESCs in Ca²⁺-free medium (Fig. 1 B, inset).

Detailed analysis indicated that with the fixed stimulation duration (10 s), the magnitude of synaptic potentiation depended on the frequency of the stimulation. No potentiation was observed at 0.5 Hz, and an ∼10-fold increase in SSC frequency could be induced by stimulation at 2–10 Hz (Fig. 1 C, left). At 50 Hz, the synaptic potentiation reached a new level, and stimulation frequency >50 Hz did not induce further potentiation (unpublished data). In medium containing normal [Ca²⁺]_o, synaptic efficacy was increased almost immediately after the application of the tetanus (Fig. 1 A). In Ca²⁺-free medium, the SSC frequency began to rise ∼1–2 min after the tetanus, and the maximal enhancement was usually observed ∼5–10 min later (Fig. 1 C, middle). After the peak, the SSC frequency gradually decreased, but it never returned to baseline for as long as the recording can be held (up to 60 min).

Two effects are elicited by tetanic stimulation of presynaptic neurons in normal conditions: repetitive firing of high frequency action potentials and large influx of Ca²⁺. As the TISP was completely independent of Ca²⁺ influx, we tested the role of action potentials in this unusual form of plasticity. Firing of action potentials results from a rapid and large Na⁺ influx, followed by a delayed efflux of K⁺ ions. To determine whether the TIPS is mediated by Na⁺ influx, we reduced Na⁺ concentration in the extracellular medium ([Na⁺]_o) by half (from 115 mM to 57.5 mM), by replacing Na⁺ with N-methyl-d-glucamine (NMDG) (Simasko, 1994). Patch recordings were made on the nerve terminals distal to the synapses made between motor axons and muscle cells (Fig. 2 A). At this low [Na⁺]_o, electric stimulation could still reliably induce action potentials, with lower amplitudes (Fig. 2 C). Moreover, action potentials could be recorded at these terminals during the entire course of the tetanic stimulation, suggesting that action potentials fully invaded the presynaptic terminals (Fig. 2 B). Under the low [Na⁺]_o conditions, however, the effect of tetanus was dramatically reduced (Fig. 3 A). SSC frequency increased by 68-fold after tetanic stimulation in normal Na⁺ medium, but showed only fivefold increase when [Na⁺]_o was reduced to 57.5 mM (Fig. 3 B). To test whether Na⁺ influx into the nerve terminals is required for TISP, we inhibited Na⁺ channels at the terminals by rapid perfusion of tetrodotoxin (TTX, 0.5 μM) to a very restricted area around the nerve terminals. TTX at this concentration completely blocked Na⁺ channels, and therefore action potentials, in the spinal neurons (unpublished data). TTX was applied through a fine glass pipette positioned near the synapse under recording by gravity. A suction pipette was placed in the opposite side of the perfusion pipette to remove excess TTX. This method has been shown to restrict drug exposure to a very small area at terminals/axons (Stoop and Poo, 1995). As shown in Fig. 3 C, application of tetanus during TTX perfusion elicited virtually no potentiation, whereas subsequent application of the same tetanus induced a robust potentiation in the same neuron during perfusion of Ringer solution. Quantitative analysis indicated that local TTX perfusion virtually prevented the tetanus-induced potentiation (Fig. 3 D). A similar brief perfusion of Ringer solution to naïve synapses had no effect on TISP (Fig. 3 D). These results strongly suggest that Na⁺ entry into the nerve terminals is important for TISP.

In addition to Na⁺ influx, propagation of action potential to the nerve terminals elicits K⁺ efflux, leading to extracellular K⁺ accumulation. Two experiments were performed to address this possibility. (1) To prevent extracellular K⁺ accumulation, we perfused Ca²⁺-free Ringer solution to the neuromuscular synapses. Tetanic stimulation induced the same magnitude of synaptic potentiation while the synapses were continuously perfused by Ca²⁺-free Ringer solution (e.g., Fig. 3 C). (2) To mimic the extracellular K⁺ accumulation, we perfused high K⁺ solution locally at the synapses. This manipulation did not cause an increase in synaptic efficacy in Ca²⁺-free medium (Fig. 4 B). Thus, extracellular K⁺ accumulation does not appear to be involved in TISP.

Direct membrane depolarization has recently been shown to induce transmitter release in the cell bodies of dorsal root ganglion (DRG) neurons (Zhang and Zhou, 2002). To test whether TISP is due to direct depolarization of motoneuron nerve terminals, we perfused high K⁺ solution locally to the terminals. After a period of control recording, a stream of high K⁺ solution was rapidly perfused to the synapse, leading to a local membrane depolarization of both presynaptic terminal and postsynaptic myocyte. In medium with regular extracellular Ca²⁺ ([Ca²⁺]_o = 2 mM), depolarization of the nerve terminal triggered robust vesicle fusion at the presynaptic terminals, as shown by a cluster of synaptic currents (Fig. 4 A, inset). It also elicited a steady inward current in the voltage-clamped myocyte and, consequently, a sustained downward shift of baseline (Fig. 4 A), which returned to control levels after the removal of the high K⁺ pipette. In marked contrast, perfusion of the nerve terminal with high K⁺ solution in Ca²⁺-free medium did not induce any transmitter release during depolarization (Fig. 4 B, inset). Under such sustained depolarization, rapid inactivation of voltage-dependent Na⁺ channels would result in very little Na⁺ influx, providing further support that Na⁺ entry into the nerve terminals is critical for TISP. As summarized in Fig. 4 C, local perfusion of high K⁺ always elicited a significant increase in SSC frequency during depolarization in normal medium, but never did so in Ca²⁺-free medium. As depolarization alone does not cause any increase in transmitter release in the absence of Ca²⁺ influx, it seems unlikely that TISP is mediated by direct depolarization of the nerve terminals.

We next tested whether a rise in intracellular Ca²⁺ concentration ([Ca²⁺]_i) is required for TISP. Membrane-permeable EGTA (EGTA/AM) was used to buffer [Ca²⁺]_i. The effectiveness of tetanus was markedly reduced when cells were pretreated with EGTA/AM for >30–60 min. The synaptic efficacy increased only threefold in 0.1 mM EGTA/AM, and the increase in SSC frequency was further attenuated when 1 mM EGTA/AM was used (Fig. 5 A). Similar effects were achieved when [Ca²⁺]_i was buffered by another Ca²⁺ chelator, the membrane-permeable BAPTA/AM, for >30 min (Fig. 5 B). Interestingly, robust synaptic potentiation was induced by tetanus when the membrane-impermeable BAPTA was loaded directly into the postsynaptic muscle cells through the patch pipette (pipette concentration, 1 mM; Fig. 5 B). These results suggest that although independent of Ca²⁺ influx, TISP still requires an increase in [Ca²⁺]_i in the presynaptic neurons but not in the postsynaptic muscle cells. To determine whether the increase in SSC frequency truly reflects an enhanced transmitter release, and whether TISP is completely independent of postsynaptic muscle cells, we examined transmitter release at free nerve terminals using FM dye destaining techniques (Betz et al., 1992; Ryan et al., 1993). FM 1-43 dye was loaded into synaptic vesicles in the presynaptic terminals by exposing neurons to high K⁺ loading solution (60 mM K⁺, 2 mM Ca²⁺) for 3 min in the presence of FM 1-43 (2 μM), followed by extensive washes in Ca²⁺-free medium. Fluorescent spots, which represent clusters of recycled vesicles labeled by FM dye, were quite stable in the Ca²⁺-free medium for a long period of time, suggesting that baseline SSC does not lead to significant FM dye destaining (Fig. 5 D). TISP was initiated by tetanic stimulation of the cell body of presynaptic neurons in the absence of extracellular Ca²⁺. A marked destaining of FM dye was observed at NMJ as well as free nerve terminals (Fig. 5 D). Thus, the induction of TISP is completely independent of postsynaptic muscle cells.

In Ca²⁺-free medium, ER Ca²⁺ stores are the major sources of intracellular Ca²⁺. Two approaches were employed to investigate the role of Ca²⁺ from ER Ca²⁺ stores in TISP. First, we examined whether inhibition of Ca²⁺ release from the ER stores (the ryanodine receptor and the IP3 receptor) would prevent TISP. A high concentration of ryanodine (Rya, 100 μM) may be used as a ryanodine receptor antagonist. If the TISP is mediated by opening of ryanodine receptor channels on the ER, pretreatment with ryanodine should prevent the effect of tetanus. However, TISP still occurred when the cultures were pretreated with ryanodine (Fig. 6 B, a). IP3 is generated from phosphatidylinositol 4,5-bisphosphate (PI[4,5]P2) by PLC-γ cleavage. Pretreatment of the nerve-muscle cocultures with the PLC-γ inhibitor U73122 (5 μM) did not block TISP (Fig. 6 B, a). Application of the IP3 receptor inhibitor Xestospongin C (XeC, 1 μM) (Gafni et al., 1997) also had no effect (Fig. 6 B, a). We also used heparin, a more specific and reliable inhibitor of the IP3 receptor. As heparin is a hydrophilic molecule incapable of entering cells through bath application, we loaded it into the presynaptic neurons through embryo injection techniques (Yang et al., 2001). The presence of heparin in presynaptic neurons was indicated by the fluorescence of coinjected rhodamine dextran. Tetanic stimulation induced very similar synaptic potentiation in both heparin-positive and -negative synapses (Fig. 6 B, a). In the second series of experiments, we determined whether TISP could still occur after ER Ca²⁺ stores are completely exhausted. Thapsigargin is known to inhibit ER Ca²⁺-ATPase activity and therefore has frequently been used to deplete ER Ca²⁺ stores (Thastrup et al., 1990). With a few minutes of delay, the application of thapsigargin in Ca²⁺-free conditions induced a marked increase in SSC frequency due to the release of Ca²⁺ from the ER Ca²⁺ stores. Within 60–80 min or so, the SSC frequency returned to control levels as a consequence of the depletion of ER Ca²⁺. Tetanic stimulation at this point could still induce the synaptic potentiation (Fig. 6 A). Taken together, these results strongly indicate that the TISP was not mediated by the release of Ca²⁺ from the ryanodine or IP3 receptor stores.

We further investigated the role of a number of key signaling molecules and extracellular factors implicated in the control of transmitter release. Ca²⁺/calmodulin-dependent kinase II (CaMKII) is well known for its ability to regulate transmitter release from nerve terminals (Greengard et al., 1993). Loading of a specific peptide inhibitor for CaMKII (Ishida et al., 1995) (estimated concentration, 10 μM) into the presynaptic neurons by embryo injection had no effect on TISP (Fig. 6 B, b). PI3K has recently been implicated in the modulation of transmitter release (Yang et al., 2001). However, expression of a dominant-negative form of PI3K (PI3K#) in presynaptic neurons by injecting PI3K# mRNA into the Xenopus embryos did not interfere with synaptic potentiation induced by tetanus (Fig. 6 B, b). A number of secretory factors, such as neurotrophins, glutamate, and substance P, have been shown to potentiate synaptic transmission at NMJ (Lohof et al., 1993; Fu et al., 1995), and the release of some of these factors could be independent of Ca²⁺ influx (Canossa et al., 2001). We tested whether tetanic stimulation induces the release of these factors, leading to the synaptic potentiation. Pretreatment of the cultures with TrkB-IgG (4 μg/ml) or TrkC-IgG (10 μg/ml), scavenger molecules specific for the neurotrophins BDNF and NT3, respectively, had no effect on the TISP (Fig. 6 B, c). Pretreatment of the cultures with a cocktail of glutamate receptor antagonists (CNQX, 5 μM, + Apv, 50 μM) also failed to block the effect of tetanus (Fig. 6 B, c). Finally, inhibition of the substance P receptors NK1 and NK2 by L-668,169 (6 μM) and L-659,877 (6 μM) did not prevent the synaptic potentiation (Fig. 6 B, c). Thus, it seems unlikely that tetanus-induced release of the secretory factors from presynaptic neurons could explain the synaptic potentiation.

Tetanic stimulation rapidly increases the concentration of intracellular Na⁺, which triggers the exchange of cytoplasmic Na⁺ with Ca²⁺ in the mitochondria (Friel, 2000; Gunter et al., 2000). The requirement of intracellular Ca²⁺ as well as Na⁺ influx, and the independence of Ca²⁺ influx and ER Ca²⁺ stores, suggests that the TISP is mediated by mitochondrial Ca²⁺. To test this hypothesis, we first uncoupled mitochondrial membrane potential by the protonophore carbonyl cyanide 4-(trifluoromethoxy) phenyl hydrazone (FCCP, 1 μM), which depolarizes and depletes Ca²⁺ from mitochondria. Application of FCCP in Ca²⁺-free conditions induced a huge increase in SSC frequency, which returned to control levels in 20–50 min (Fig. 7, A and B). FCCP also caused a transient and slow drift of membrane potential, which was recovered within minutes (Fig. 7 A). Compared with the effect of thapsigargin on ER Ca²⁺ stores, the depletion of mitochondria Ca²⁺ by FCCP had a faster onset and higher magnitude, but lasted for a relatively shorter period of time. Importantly, tetanic stimulation 40 min after FCCP application could no longer induce synaptic potentiation (Fig. 7 A). Quantitative analysis indicated that FCCP treatment elicited a transient, 67-fold increase in SSC frequency, and TISP failed to occur when the tetanic stimulation was applied after the effect of FCCP subsided (Fig. 7 B). These data suggest that the release of Ca²⁺ from mitochondria is critical for TISP.

To determine the specific route through which Ca²⁺ came out of mitochondria, we used CGP37157 (CGP), a specific inhibitor for the mitochondrial Na⁺-Ca²⁺ exchanger (Cox et al., 1993). Pretreatment of cells with CGP markedly reduced the effectiveness of tetanus to induce synaptic potentiation. Dose–response experiments showed that CGP was effective at as low as 10 μM (Fig. 7 C). Quantitative analysis indicated that the SSC frequency increased 68-fold after tetanus in control conditions, but only 3.5-fold at the NMJ treated with 30 μM of CGP (Fig. 7 C). To ensure the specificity of CGP, we performed a number of control experiments. In either normal or Ca²⁺-free conditions, CGP (30 μM) did not affect the frequency or any other parameters of SSCs (Fig. 7 D and not depicted), suggesting that it does not alter the release apparatus. Treatment of the neuromuscular synapses with CGP at the same concentration had no effect on the properties of ESCs induced in normal Ca²⁺ conditions (Fig. 7 E). Identical action potentials were recorded in spinal neurons treated with or without CGP (not depicted). Thus, TISP appears to require Ca²⁺ release from the mitochondrial Na⁺-Ca²⁺ exchanger.

We next examined whether tetanic stimulation elicits an increase in [Ca²⁺]_i via the mitochondrial Na⁺-Ca²⁺ exchanger. Ca²⁺ imaging was performed by confocal microscopy using the cell-permeable Ca²⁺ indicator fluo-4. Cells in the nerve-muscle cultures were loaded with this indicator, and tetanic stimulation was applied to neuronal soma through a loose-patch pipette (touched, but not sealed) to induce action potentials. We determined the fluo-4 fluorescence in Ca²⁺-free medium (Fig. 8 A, inset) and normalized it to averaged basal fluorescence before stimulation. The change in fluorescence (ΔF/F₀) was plotted against time. Tetanic stimulation evoked a gradual increase of [Ca²⁺]_i at nerve terminals (Fig. 8 B), with the time course very similar to that of TISP. This increase was significantly attenuated in low [Na⁺]_o (57.5 mM, Fig. 8 B). Moreover, pretreatment with CGP markedly attenuates the tetanus-induced mitochondrial Ca²⁺ release (Fig. 8 B). The basal [Ca²⁺]_i level in the terminals, however, was not affected by CGP (see Fig. 10 C). In the presence of the IP3 receptor inhibitor XeC, tetanus still induced the same magnitude of increase in [Ca²⁺]_i (Fig. 8 B). Thus, in zero [Ca²⁺]_o conditions, tetanus-induced increase in intracellular Ca²⁺ is mediated by the mitochondrial Na⁺-Ca²⁺ exchanger. These results, together with the findings that TISP could be attenuated by FCCP and CGP, strongly support the model that tetanic stimulation induces Ca²⁺ release through the mitochondrial Na⁺-Ca²⁺ exchanger at the nerve terminals, leading to a marked potentiation of transmitter release.

While many intracellular signaling pathways and secretory factors we tested had no effect on TISP (Fig. 6), inhibition of PKC had a profound effect on TISP as well as tetanus-induced [Ca²⁺]_i increase. Loading of a specific, peptide inhibitor for PKC (PKC pep; House and Kemp, 1987) into the presynaptic neurons by embryo injection dramatically reduced the potentiation induced by tetanic stimulation of the same neurons (N+), while TISP was observed in the nearby synapses made by neurons not loaded with the PKC pep (N−) in the same culture dishes (Fig. 9 A). Again, loading of the PKC pep into the postsynaptic muscle cells had no effect on TISP (unpublished data). In addition, TISP was greatly attenuated by pretreatment with the specific PKC inhibitor chelerythrine (CheT, 5 μM) or GF 1092303X (GF, 0.8 μM), respectively (Fig. 9 A). On the other hand, application of the PKC activator PMA (5 μM) alone was capable of potentiating transmitter release in the absence of Ca²⁺ influx, although the magnitude of the potentiation was smaller than that of TISP, and pretreatment with CGP significantly reduced the PMA-induced potentiation (Fig. 9 B). We then investigated the role of PKC in intracellular Ca²⁺. Inhibition of endogenous PKC by CheT or GF significantly reduced the tetanus-induced increase in [Ca²⁺]_i (Fig. 9 C). On the other hand, activation of PKC by PMA rapidly and significantly increased the [Ca²⁺]_i in Ca²⁺-free medium (Fig. 9 D). The time course of the [Ca²⁺]_i rise was similar to that induced by tetanus, but the level of [Ca²⁺]_i gradually declined even in the presence of PMA (Fig. 9 E). Further, the PMA-induced increase in [Ca²⁺]_i was inhibited by CGP (Fig. 9 F). Thus, while PKC is required for TISP and the tetanus-induced increase in [Ca²⁺]_i, activation of PKC is not sufficient to induce the same magnitude of synaptic potentiation as TISP, or the same patterns of [Ca²⁺]_i change as that induced by tetanus. These results suggest that PKC serves as a modulator, rather than a mediator, for tetanus-induced [Ca²⁺]_i release through the mitochondrial Na⁺-Ca²⁺ exchanger.

Finally, we determined whether the mitochondrial Na⁺-Ca²⁺ exchanger also contributes to the synaptic plasticity in normal extracellular Ca²⁺ ([Ca²⁺]_o = 2 mM). The synaptic potentiation induced by the same tetanus (50 Hz, 10 s) in normal medium was even greater than that in Ca²⁺-free medium (Fig. 10 A). Pretreatment with CGP greatly attenuated the effect of the tetanus (Fig. 10 A). The SSC frequency increased 143-fold after tetanus, but only 38-fold at the NMJ treated with 30 μM CGP. To detect high-amplitude, rapid changes of nerve terminal [Ca²⁺]_i in normal [Ca²⁺]_o conditions, we used a low affinity Ca²⁺ indicator Oregon green 488 BAPTA-5N, AM (OG488, 5 μM). Consistent with a previous report (David et al., 1998), a train of tetanic stimulation induced a rapid increase in [Ca²⁺]_i due to Ca²⁺ influx into the cytosol, followed by a slow but sustained increase in [Ca²⁺]_i, most likely due to Ca²⁺release from mitochondria (Fig. 10 B). Pretreatment with CGP blocked the late onset [Ca²⁺]_i increase (Fig. 10 B), but not the increase of [Ca²⁺]_i immediately after the tetanus (Fig. 10 B, inset). Application of CGP did not affect the basal level of [Ca²⁺]_i in either normal or Ca²⁺-free medium (Fig. 10 C). Taken together, these results suggest that the mitochondrial Na⁺-Ca²⁺ exchanger contributes to TISP under normal physiological conditions.

The mechanism underlying modulation of transmitter release is an important but difficult problem to study, because many extracellular and intracellular molecules that regulate transmitter release also affect Ca²⁺ influx. Thus, it is difficult to establish whether an effect on synaptic transmission is due to modulation of Ca²⁺ influx or to direct regulation of transmitter release. The present study was done in the complete absence of Ca²⁺ influx, avoiding the potential problem of channel modulation. Using this approach, we have discovered a novel form of synaptic plasticity in which transmitter release is dramatically potentiated in response to a presynaptic tetanic stimulation. This form of synaptic plasticity would not have been observed if the experiments were performed in normal extracellular Ca²⁺. There are a number of interesting findings: (1) Na⁺ influx appears to be the trigger for TISP; (2) it requires an increase in [Ca²⁺]_i; (3) the increase in [Ca²⁺]_i results from Ca²⁺ efflux through the mitochondrial Na⁺-Ca²⁺ exchanger; and (4) PKC is involved in TISP as well as mitochondrial Ca²⁺ efflux. The enhancement of transmitter release via Ca²⁺ efflux from mitochondria also occurs in normal [Ca²⁺]_o, and may contribute to synaptic modulation in heightened activity under physiological conditions such as physical exercise, and/or in pathological conditions such as neuromuscular fatigue. These results demonstrate the role of Na⁺ influx and the mitochondrial Na⁺-Ca²⁺ exchanger in synaptic modulation and reveal a new form of synaptic plasticity. Further, our study may help understand how mitochondria and PKC contribute to the regulation of transmitter release and provide a useful model to investigate molecular mechanisms for transmitter release without the interference of Ca²⁺ influx.

Although our study demonstrated that transmission and synaptic plasticity could occur without Ca²⁺ influx, it seems that a rise of presynaptic [Ca²⁺]_i is still necessary. This is quite different from the Ca²⁺-independent vesicle fusion in the cell body of DRG neurons reported recently (Zhang and Zhou, 2002). Vesicle fusion, as measured by membrane capacitance recording, was induced by membrane depolarization in Ca²⁺-free medium. Fura-2 measurements could not detect any change in [Ca²⁺]_i under these conditions. Even intracellular dialysis with BAPTA for 10 min could not block this secretion. It was therefore concluded that a Ca²⁺-independent but voltage-dependent vesicular secretion may exist in the cell body of the DRG neurons (Zhang and Zhou, 2002). As the authors pointed out, the depolarization-induced vesicle fusion was not observed in a number of other cells tested. In our system, depolarization of the nerve terminals in Ca²⁺-free conditions by rapid perfusion of high K⁺ solution could not induce transmitter release at the neuromuscular synapses. Although the present study also found an increase in vesicle fusion in the absence of Ca²⁺ influx, we attribute this phenomenon to Na⁺ influx driven by action potentials and Ca²⁺ release through mitochondrial Na⁺-Ca²⁺ exchanger. It will be interesting to determine whether mitochondrial Ca²⁺ release also contributes to the depolarization-induced vesicle fusion seen in DRG cell bodies.

Mitochondria are observed in virtually all nerve terminals. It has long been thought that the function of the terminal mitochondria is to provide energy necessary for synaptic vesicle cycling. However, a number of studies have demonstrated that high frequency stimulation could induce mitochondrial Ca²⁺ release at the motoneuron terminals, and this could go on over a period of ∼10 min after the cessation of the stimulation (David et al., 1998; David, 1999). The release of mitochondrial Ca²⁺ has been shown to enhance transmitter release induced by repetitive stimulation or drug application (Tang and Zucker, 1997; Scotti et al., 1999; Tsang et al., 2000). As these experiments were done in normal extracellular Ca²⁺, loading of Ca²⁺ into the mitochondria during stimulation and unloading after stimulation is thought to be the mechanism for the transient increase in [Ca²⁺]_i (Tang and Zucker, 1997; Melamed-Book and Rahamimoff, 1998). Mitochondria are therefore proposed to serve as a Ca²⁺ buffering system that controls the extent of cytosolic Ca²⁺ rise in the nerve terminals (Stuenkel, 1994; Tang and Zucker, 1997; David et al., 1998; David, 1999). We found, however, that FCCP could induce mitochondrial Ca²⁺ release when Ca²⁺ influx was completely prohibited. Thus, resting mitochondria contain substantial amounts of Ca²⁺, and the release of this Ca²⁺ is sufficient to enhance transmitter release at NMJ. Another advance by the present study is to demonstrate synaptic modulation by Ca²⁺ release through the mitochondrial Na⁺-Ca²⁺ exchanger. Several previous studies have suggested that the plasmalemmal Na⁺-Ca²⁺ exchanger, but not the one on mitochondria, is important for presynaptic Ca²⁺ regulation (Scotti et al., 1999; Zhong et al., 2001). We show that in Ca²⁺-free medium, inhibition of the mitochondrial Na⁺-Ca²⁺ exchanger prevented the tetanus-induced increase in [Ca²⁺]_i and greatly attenuated TISP. Thus, our study reveals a previously unexpected role of the mitochondrial Na⁺-Ca²⁺ exchanger in modulation of transmitter release.

A form of plasticity most relevant to the present study is PTP. Using the crayfish NMJ as a model system, Tang and Zucker (1997) have investigated the mechanism for PTP in detail. Inhibition of mitochondrial Ca²⁺ uptake or release by tetraphenylphosphonium (TPP⁺), carbonyl cyanide m-chlorophenylhydrazone (CCCP), or ruthenium red all block PTP and the “residual Ca²⁺” in the nerve terminals after tetanic stimulation. They proposed that mitochondria accumulate Ca²⁺ during tetanic stimulation (as a consequence of Ca²⁺ influx) and release Ca²⁺ back to the cytoplasm after the tetanus. In a later study, they found that the specific plasmalemmal Na⁺-Ca²⁺ exchanger inhibitor KB R7943 significantly reduced PTP as well as Ca²⁺ accumulation caused by Na⁺ influx (Zhong et al., 2001). In contrast, the specific mitochondrial Na⁺-Ca²⁺ exchanger inhibitor CGP had no effect on either PTP or Ca²⁺ accumulation. It was therefore concluded that the plasmalemmal Na⁺-Ca²⁺ exchanger acting in reverse mode is the key mediator for PTP, and the mitochondrial Na⁺-Ca²⁺ exchanger is not involved in this form of plasticity. The TISP described in the present study is very different from PTP. First, TISP does not involve residual Ca²⁺ resulting from tetanus-induced Ca²⁺ influx. Rather, it can be induced in the complete absence of Ca²⁺ influx. Second, the kinetics of TISP is quite different from that of PTP. PTP occurs right after the termination of tetanus and gradually decreases over time. In Ca²⁺-free medium, TISP occurs several minutes after the termination of the tetanus, peaks around 10 min, and lasts for a much longer period of time (>1 h). Third, TISP and PTP are mediated by very different mechanisms. We demonstrated that TISP requires the mitochondrial Na⁺-Ca²⁺ exchanger, while Zhong et al. (2001) showed that PTP is completely independent of the molecule. Taken together, the present study has identified a new form of synaptic plasticity and elucidated its underlying mechanisms.

Tetanic stimulation has also recently been shown to induce LTP at the NMJ (Wan and Poo, 1999). Injection of BAPTA, EGTA, or a peptide inhibitor for the Ca²⁺-dependent phosphatase calcineurin into postsynaptic muscle cells blocks this LTP. Thus, the induction of LTP requires Ca²⁺ influx and perhaps calcineurin activity in postsynaptic muscle cells. We found that TISP in the absence of Ca²⁺ influx could also last for a long period of time (for >1 h, or as long as the recording could be held, SSC frequency never returned to the baseline levels before tetanus). However, introduction of BAPTA or PKC inhibitor into postsynaptic muscle cells could not prevent TISP. FM dye destaining experiments also indicate that tetanic stimulation of neuronal soma elicited a robust transmitter release in free nerve terminals. Thus, TISP described in this study is independent of postsynaptic muscle cells.

At a variety of synapses, activation of PKC potentiates transmitter release primarily in steps downstream of Ca²⁺ influx (Majewski and Iannazzo, 1998). Given the same magnitude of [Ca²⁺]_i rise, two factors control the fusion of synaptic vesicles and the release of transmitters: the number of readily releasable vesicles (also called the readily releasable pool) and the release probability of these vesicles. Using chromaffin cells and cultured hippocampal neurons as model systems, PKC has been implicated in the potentiation of synaptic transmission by increasing the size of the readily releasable pool (Gillis et al., 1996; Stevens and Sullivan, 1998). At calyx-type synapses in the brain stem, PKC is thought to facilitate vesicle exocytosis by increasing the probability of release (Yawo, 1999; Wu and Wu, 2001). These conclusions, however, were based on experiments using phorbol ester as a PKC activator. A recent study indicated that phorbol ester–induced potentiation of transmitter release is mediated not by PKC but by Munc-13, a protein localized in the presynaptic active zone and involved in priming vesicles to fusion competence (Rhee et al., 2002). PKC may also modulate voltage-gated ion channels, leading to an increase in presynaptic Ca²⁺ influx (Fu and Huang, 1994; Byrne and Kandel, 1996). By eliminating Ca²⁺ influx into the nerve terminals, we were able to address the role of PKC in transmitter release induced by Ca²⁺ released from mitochondria. Inhibition of presynaptic PKC by several specific PKC inhibitors markedly attenuated the tetanus-induced Ca²⁺ release from mitochondria and virtually prevented TISP, while activation of PKC rapidly increased the [Ca²⁺]_i and potentiated transmitter release in Ca²⁺-free medium. Our results therefore support the notion that PKC is required for optimal function of the mitochondrial Na⁺-Ca²⁺ exchanger at nerve terminals.

As previously described (Yang et al., 2001), dominant-negative PI3 kinase mRNA, IP3 receptor inhibitor heparin (Sigma-Aldrich), PKC peptide inhibitor, or CaMKII peptide inhibitor (both from Calbiochem) was mixed with GFP mRNA and injected (total 10 ng) into one of the cells of two-cell stage Xenopus embryos. At stage 20–22, the normal or injected embryos were dissociated in Ca²⁺-Mg²⁺–free medium, plated on glass coverslips, and grown for 1 d at room temperature (20–22°C). Synaptic currents were recorded from innervated muscle cells by whole-cell recording in either normal or Ca²⁺-free medium. In the low [Na⁺]_o experiments, the extracellular Na⁺ was reduced by half, plus 57.5 mM NMDG to maintain the ionic strength (Sigma-Aldrich). A time course of SSC frequency was first constructed on a minute-to-minute basis for each synapse. The SSC frequencies recorded 10 min before tetanic stimulation and those recorded during a 10-min period with the maximal response after tetanic stimulation were compared. Tetanic stimulation was applied to neuronal soma under loose seal conditions through a patch electrode filled with Ringer solution. For local perfusion of TTX to the NMJ, whole culture dishes were continuously perfused with Ca²⁺-free or Ca²⁺-containing Ringer solution. A TTX-containing pipette was positioned near synapses. After a period of control recording of SSCs, TTX (0.5 μM in Ringer solution) was rapidly perfused to the synapse using a multi-barrel perfusion system (Warner Instruments). For high K⁺–mediated depolarization experiments, the same system was used to perfuse high K⁺ solution (Ringer solution with 60 mM KCl) to the synapses. The duration and the extent of depolarization were reflected by the sustained downward shift of baseline. The clusters of synaptic currents were quite reproducible among different depolarization episodes.

The FM dye experiments were performed essentially as previously described (Wang et al., 2001). FM 1-43 (Molecular Probes) was loaded into the spinal neurons by incubating 1-d-old cultures with high K⁺ loading solution containing KCl (60 mM), NaCl (57.6 mM), CaCl₂ (2 mM), Hepes (10 mM, pH 7.6), and FM 1-43 (2 μM) for 3 min. Cells were then rinsed with Ca²⁺-free medium and imaged with a Noran Odyssey II confocal unit, using a laser with a band pass excitation filter around 488 nm and a 515-nm long pass emission filter. Four 640 × 480 pixel frames were averaged (123 ms), and images were acquired at one image/10 s. The fluorescence images were stable with minimum bleach for >10 min in Ca²⁺-free medium. After acquiring 12 images (2 min) as baseline, FM 1-43 destaining was initiated by tetanus, and acquisition was continued for another 8 min. Fluorescence intensity was measured using a region-of-interest tool outlining the varicosities, corrected for photo bleaching, and post hoc normalized to initial fluorescence (ΔF/F₀).

Cells from 1-d-old cultures were loaded with Ca²⁺ indicators for 30–60 min. Fluo-4 (Molecular Probes; final concentration, 5 μM) was used for terminal [Ca²⁺]_i changes in Ca²⁺-free medium, whereas Oregon green 488 BAPTA-5N, AM (Molecular Probes; final concentration, 5 μM) was used for Ca²⁺ imaging in normal medium. Cells were extensively washed after loading the Ca²⁺ indicators in respective media, and then excited at 488 nm. The fluorescence images were acquired at the same rate using the same procedures as described in the previous paragraph.

We thank Regeneron Pharmaceutical for TrkB-IgG and TrkC-IgG.

This work is supported by the National Institute of Child Health and Human Development Intramural Research Program.