Short-term potentiation of mEPSCs requires N-, P/Q- and L-type Ca²⁺ channels and mitochondria in the supraoptic nucleus

Abstract

The glutamatergic synapses of the supraoptic nucleus display a unique activity-dependent plasticity characterized by a barrage of tetrodotoxin-resistant miniature EPSCs (mEPSCs) persisting for 5–20 min, causing postsynaptic excitation. We investigated how this short-term synaptic potentiation (STP) induced by a brief high-frequency stimulation (HFS) of afferents was initiated and maintained without lingering presynaptic firing, using in vitro patch-clamp recording on rat brain slices. We found that following the immediate rise in mEPSC frequency, STP decayed with two-exponential functions indicative of two discrete phases. STP depends entirely on extracellular Ca²⁺ which enters the presynaptic terminals through voltage-gated Ca²⁺ channels but also, to a much lesser degree, through a pathway independent of these channels or reverse mode of the plasma membrane Na⁺–Ca²⁺ exchanger. Initiation of STP is largely mediated by any of the N-, P/Q- or L-type channels, and only a simultaneous application of specific blockers for all these channels attenuates STP. Furthermore, the second phase of STP is curtailed by the inhibition of mitochondrial Ca²⁺ uptake or mitochondrial Na⁺–Ca²⁺ exchanger. mEPSCs amplitude is also potentiated by HFS which requires extracellular Ca²⁺. In conclusion, induction of mEPSC-STP is redundantly mediated by presynaptic N-, P/Q- and L-type Ca²⁺ channels while the second phase depends on mitochondrial Ca²⁺ sequestration and release. Since glutamate influences unique firing patterns that optimize hormone release by supraoptic magnocellular neurons, a prolonged barrage of spontaneous excitatory transmission may aid in the induction of respective firing activities.

Magnocellular neurons of the supraoptic nucleus (SON) display an intermittent bursting or phasic activity to optimize the release of oxytocin (OXT) or vasopressin (VP) from their nerve terminals located in the neurohypophysis. These characteristic firing patterns are known to depend on the intrinsic regenerative properties of magnocellular neurons (Andrew & Dudek, 1983; Armstrong et al. 1994) and various extrinsic factors (Hu & Bourque, 1992; Jourdain et al. 1998; Bourque et al. 1998; Ludwig & Pittman, 2003; Li et al. 2007).

The excitatory neurotransmitter glutamate is one of the important components of the SON physiology and is known to be essential for the onset and maintenance of bursting activity of magnocellular neurons, both in vivo and in vitro (Wakerley & Noble, 1983; Hu & Bourque, 1992; Nissen et al. 1995; Jourdain et al. 1998; Brown et al. 2004). Increasing levels of frequency and fluctuation of basal firing between bursts are driven by excitatory synaptic inputs and promote bursting activity in OXT neurons (Israel et al. 2003; Moos et al. 2004), with clusters of EPSPs underlying bursts (Dudek & Gribkoff, 1987; Jourdain et al. 1998). Furthermore, synchronization of action potential firing between a pair of OXT neurons relies on excitatory transmission (Israel et al. 2003). Bursting of magnocellular neurons also depends on local neuropeptide release from the dendrites of these neurons (Lambert et al. 1993; Jourdain et al. 1998). Glutamate can initiate this mechanism by inducing dendritic neuropeptide release through activation of NMDA receptors (de Kock et al. 2004). Once dendritic release is triggered, neuropeptides may amplify themselves in a positive feedback manner stimulating further dendritic release (Moos et al. 1984; Ludwig et al. 2005). Furthermore, glutamate seems to be involved in OXT-induced morphological remodelling in the SON (Langle et al. 2003). Such structural plasticity is also known to occur in vivo during periods of high hormone demand such as lactation and dehydration (Hatton, 1997).

A train of stimulation to excitatory afferents innervating the SON induces a long barrage of spontaneous EPSPs/EPSCs leading to slow depolarization and afterdischarge of the postsynaptic magnocellular neurons that continue after the stimulation ended, for many seconds to many minutes depending on the frequency and pulse number of the stimulus (Hatton et al. 1983; Dudek & Gribkoff, 1987; Kombian et al. 2000). Recently it has been shown in the paraventricular nucleus, where VP and OXT neurons also exist, that asynchronous glutamate release following a single or a short burst of action potential firing of the afferents sufficiently prolongs the postsynaptic spike activity (Iremonger & Bains, 2007). In the SON, these spontaneous EPSCs induced by presynaptic stimulation are insensitive to tetrodotoxin (Kombian et al. 2000), indicating that they are equivalent to miniature EPSCs (mEPSCs) and that this form of synaptic plasticity is an extreme case of asynchronous, delayed release of glutamate maintained at presynaptic terminals rather than at the level of presynaptic soma.

It remains unknown how excitatory synapses in the SON can sustain the elevated rate of exocytosis for minutes without firing activity. Slow depolarization and afterdischarge that result from barrages of EPSCs are known to be diminished by low extracellular Ca²⁺, indicating that this may be a Ca²⁺-dependent process (Dudek & Gribkoff, 1987). Here, we examined the roles of various subtypes of voltage-gated Ca²⁺ channels (VGCCs) as well as intracellular Ca²⁺ stores in the initiation and maintenance of STP of mEPSC frequency (mEPSC-STP).

Methods

All experiments in this study were carried out in accordance with the guidelines established by the Canadian Council on Animal Care and as approved by the Memorial University Internal Animal Care Committee. Effort was given to use only the necessary number of animals required to yield reliable results.

Slice preparation

Male Sprague–Dawley rats (60–100 g) were deeply anaesthetized using halothane until respiration almost stopped, prior to decapitation. The brain was rapidly removed and 250 μm thick coronal sections containing the SON were generated in a 0–2°C buffer solution composed of the following (mm): 87 NaCl, 2.5 KCl, 1.25 NaH₂PO₄, 7 MgCl₂, 0.5 CaCl₂, 25 NaHCO₃, 25 glucose, 30 sucrose, 3 pyruvic acid and 1 ascorbic acid, bubbled with 95% O₂–5% CO₂. Slices were incubated at 33–34°C for 45 min and then at room temperature until recording in bubbled artificial cerebrospinal fluid (aCSF) composed of the following (mm): 126 NaCl, 2.5 KCl, 1.2 NaH₂PO₄, 1.2 MgCl₂, 2 CaCl₂, 25 NaHCO₃, 10 glucose. For the experiments using CdSO₄, NaCl was used instead of NaH₂PO₄ to prevent precipitation. For Ca²⁺-free aCSF, CaCl₂ was replaced with equimolar MgCl₂ and 5 mm EGTA was added.

Electrophysiological recording

Slices were hemisected, placed into a recording chamber and perfused at 1.5–2.5 ml min⁻¹ with aCSF at 33–34°C. Whole-cell patch-clamp recordings were done in the SON with either an Axopatch 1D or MultiClamp 700B amplifier (Molecular Devices, Sunnyvale, CA, USA). Nystatin was used as a perforating agent to obtain access unless stated otherwise, where conventional whole-cell access was attained. For nystatin-perforated patch recording, the internal solution contained the following (mm): 120 potassium gluconate, 5 MgCl₂, 10 EGTA and 40 Hepes, pH 7.3. Nystatin was dissolved in dimethyl sulfoxide with Pluronic F127 and added to the internal solution to yield a final concentration of 450 μg ml⁻¹. Glass electrodes had a tip resistance of 3–7 MΩ when filled with the internal recording solution. Series/access resistance of 10–40 MΩ was normally attained within less than 20 min after formation of a gigaohm (1–8 GΩ) seal. For conventional whole-cell recording, the internal solution consisted of (mm): 123 potassium gluconate, 2 MgCl₂, 8 KCl, 0.2 EGTA, 10 Hepes, 4 Na₂-ATP and 0.3 Na-GTP, pH 7.3. EGTA was increased to 10 mm when necessary.

Magnocellular neurons were identified based on the characteristic delayed onset to action potential generation in response to a depolarizing current injection (Armstrong, 1995). Experiments were conducted in voltage-clamp mode at a holding potential of −80 mV in the presence of picrotoxin (50 μm) to isolate EPSCs. Input and access resistance of all cells were monitored every minute by applying a 20 mV hyperpolarizing pulse for 100 ms. Cells that showed a significant change in these parameters were not included in the data analysis.

Bipolar tungsten-stimulating electrodes were placed dorsal to the SON in order to evoke synaptic responses and to apply high-frequency stimulation (HFS; 50 Hz for 1 s or 100 Hz for 1 s, twice in a 10 s interval). All cells had a graded evoked synaptic response to increasing stimulus intensity. The stimulus intensity giving 50–70% of the maximum evoked EPSC was used to apply HFS.

Both evoked EPSCs and spontaneous EPSCs (sEPSC) were completely blocked by 6,7-dinitroquinoxaline-2–3-dione (DNQX; 10 μm), indicating that they were non-NMDA receptor mediated. However, while evoked EPSCs were abolished by tetrodotoxin (TTX; 1 μm), sEPSCs were TTX resistant as previously reported (Kabashima et al. 1997; Boudaba et al. 1997; Kombian et al. 2000). Therefore, sEPSCs recorded in our preparation were equivalent to miniature EPSCs (mEPSC). In this study, all recordings were done without TTX because STP induction is dependent on high-frequency action potential firing of the presynaptic fibres (Kombian et al. 2000). All data were collected using pCLAMP 7 and 9 software (Molecular Devices). Membrane currents were filtered at 1 kHz, acquired at 2–10 kHz sampling rate, and stored for offline analysis.

Data analysis

mEPSCs were detected using Mini Analysis 6.0 software (Synaptosoft, Decatur, GA, USA). For amplitude analysis, events were visually inspected on an expanded scale and only those with a clearly defined baseline and smooth rising phase were used. Statistical comparisons were performed by using appropriate tests, i.e. two-way repeated measures ANOVA for multi-factor paired group data, two-way ANOVA for multi-factor unpaired group comparisons, Bonferroni post test, paired and unpaired Student's t tests for group comparisons as appropriate. A Kolmogorov–Smirnov test was used to analyse individual cells. Values are indicated as mean ± s.e.m. and P < 0.05 was considered significant.

The time-effect plot of the frequency of mEPSCs obtained from individual cells as well as the group average data were fitted with one- or two-exponential curves using Prism 4 (GraphPad Software Inc.). The half-life of each exponential component in individual cells was also calculated using the software.

Due to the large variability in the magnitude and duration of frequency change among cells, each drug condition was compared to a non-drug condition in the same cell unless stated otherwise. Within a given cell, STP was reversible and reliably reproduced by repeated application of HFS as shown previously (Kombian et al. 2000).

Drugs

All drugs were bath-perfused (with the exception of EGTA) at final concentrations by dissolving aliquots of stock in aCSF. The solvents were diluted by at least 1000 times. Nicardipine stock was kept in the dark until use. All compounds were purchased from Sigma-Aldrich (St Louis, MO, USA), except KB-R7943 and CGP37157 from Tocris Bioscience (Ellisville, MO, USA), EGTA from OmniPur, and Pluronic F127 from BASF.

Results

Characteristics of short-term potentiation of mEPSCs

Whole-cell patch-clamp recordings were performed on 117 magnocellular neurons of the SON from 86 rats. Application of 50–100 Hz presynaptic HFS consistently induced STP of mEPSC frequency (referred to as mEPSC-STP) as well as an obvious increase in the amplitude of mEPSCs immediately following the stimulation (Fig. 1A). Furthermore, some cells showed characteristic electrophysiological properties of OXT neurons (sustained outward rectification and inward rectification when hyperpolarized to various membrane potentials from −40 mV) while others had properties of VP neurons (absence of rectification) (Stern & Armstrong, 1995; Hirasawa et al. 2003), confirming that mEPSC-STP was a characteristic of excitatory afferents for both types of neurons. Following the most pronounced increase immediately after HFS (up to 50-fold), mEPSC frequency displayed a two-phase exponential decay (Fig. 1B). In every cell tested, a two-exponential decay fits better than a one-exponential decay, thus data shown are two-exponential fit only (50 Hz: n = 10, 100 Hz: n = 10). The first phase was characterized by a large magnitude and fast decay which was followed by the second phase that decays much slower. The half-life of each phase was similar in both 50 and 100 Hz HFS conditions (P > 0.05, Fig. 1C and D). Although most cells showed a smooth transition from the first to second phase (Fig. 1E), some cells displayed a noticeable second peak in the frequency time-effect plot (Fig. 1F and G), supporting the idea of two distinct phases. For the rest of the study, experiments using 50 or 100 Hz HFS were pooled, with the same frequency and pulse number always used for paired statistical comparison.

Figure 1. Presynaptic high-frequency stimulation (HFS) results in a short-term potentiation of mEPSCs in magnocellular neurons.

A, voltage-clamp traces from a representative cell showing STP induced by 50 Hz × 1 s HFS (arrow). Lower panels show expanded traces at various time points as indicated. Both the frequency and amplitude of mEPSCs show significant increase. B, time-effect plot of mEPSC frequency following HFS, fitted with two-exponential decay curves. HFS was applied at time 0. C and D, half-life of 1st and 2nd phase exponential decays. E, a typical cell showing STP with smooth transition from the 1st to 2nd phase. F and G, examples of STP that display a clear second peak (arrowhead).

Extracellular Ca²⁺ and locus of mEPSC-STP

To test the role of extracellular Ca²⁺ in mediating STP, we delivered HFS in Ca²⁺-free aCSF, which was applied to the slice starting 10 min prior to stimulation and throughout the experiment to ensure that extracellular Ca²⁺ is minimal. In this condition, the evoked response was completely eliminated and HFS had no effect on the frequency of mEPSCs (n = 5; P > 0.05 versus baseline frequency, Fig. 2Aa–c and B). Furthermore, in normal aCSF (2 mm Ca²⁺), amplitude distribution showed a shift to the right, demonstrating the emergence of larger events (n = 5; P < 0.05 by Kolmogorov–Smirnov test in all cells tested, Fig. 2C, left panel). Without extracellular Ca²⁺, a change in amplitude was not seen (n = 5 of 5; P > 0.05, Fig. 2C, right panel). Thus, potentiation of both mEPSC frequency and amplitude is entirely dependent on Ca²⁺ influx.

Figure 2. STP is dependent on presynaptic Ca²⁺ entry.

Aa, time-effect plot of mEPSC frequency in response to HFS, applied in the absence (0 mm) or presence of 2 mm Ca²⁺ in the bath. Arrows indicate HFS. Note that this cell shows robust STP in the presence of extracellular Ca²⁺. Ab, evoked EPSCs recorded in the absence or presence of Ca²⁺. Ac, sample traces prior to (control) and immediately after HFS recorded with no extracellular Ca²⁺. B, group data showing that HFS induces no change in mEPSC frequency in the absence of Ca²⁺. C, cumulative plots of mEPSC amplitude from representative cells; before (continuous line: control) and after HFS (dashed line: post-HFS). In 2 mm Ca²⁺ (left panel), mEPSC amplitude distribution shifts to the right, indicating appearance of larger events. Without extracellular Ca²⁺ (0 mm Ca²⁺; right panel), there is no change in the amplitude. D, EGTA application into the postsynaptic cell via recording pipette does not alter STP. E, holding the postsynaptic cell at 0 mV during HFS has no effect on STP.

To ascertain the locus of Ca²⁺ influx required to induce STP, experiments were conducted using conventional whole-cell recordings with the Ca²⁺ chelator EGTA (10 mm) in the intracellular solution to inhibit cytosolic free Ca²⁺ elevation in the postsynaptic cell (n = 5). The result obtained from this set of experiments was compared to that from 0.2 mm EGTA in the intracellular solution (n = 7). We found that even in the presence of 10 mm EGTA, the frequency of mEPSCs increased following HFS, akin to control (P > 0.05; Fig. 2D), demonstrating that Ca²⁺ entry into the postsynaptic cell has a minimal role in the induction of STP.

Nevertheless, in a more physiological situation where the postsynaptic cell is not voltage clamped, the cell may depolarize significantly in response to afferent HFS and initiate retrograde signalling to modulate STP. Therefore, to further test for a role of the postsynaptic cell in mEPSC-STP, the postsynaptic cell was depolarized to 0 mV during HFS. Following stimulation, the holding potential returned to −80 mV and recording resumed. As shown in Fig. 2E, this protocol induced a STP equivalent to the one induced with the postsynaptic cell voltage clamped at −80 mV throughout the experiment (n = 4; P > 0.05). This result indicates that the postsynaptic cell does not play a role in STP.

Role of voltage-gated Ca²⁺ channels in mEPSC-STP

A major route of Ca²⁺ influx into a neuron is VGCCs. We found that these channels are critical for STP, because Cd²⁺ (200 μm), a non-selective blocker of voltage-gated Ca²⁺ channels, largely inhibited STP (n = 4; P < 0.05, Fig. 3A and B). To determine whether low- or high-VGCCs are involved, HFS was given in the presence of Ni²⁺ (50 μm) or Cd²⁺ (50 μm). At these concentrations, Ni²⁺ and Cd²⁺ specifically block low- and high-VGCCs, respectively (Fisher & Bourque, 1995). Cd²⁺ at 50 μm inhibited STP to a similar degree as 200 μm (n = 5; P < 0.05 versus control STP, P > 0.05 versus 200 μm Cd²⁺, Fig. 3B and C), suggesting that high-VGCCs predominantly mediate the effect. Lack of Ni²⁺ effect further supported this result (n = 5; P > 0.05, Fig. 3D).

Figure 3. STP is dependent on high-voltage-gated Ca²⁺ channels.

A, representative time-effect plot showing the effect of Cd²⁺ on STP. In the presence of Cd²⁺, STP is largely abolished. Arrows indicate HFS. Right panel, evoked EPSC is completely blocked by Cd²⁺. B, group data showing the significant attenuation of STP by 200 μm Cd²⁺. C, low concentration of Cd²⁺ (50 μm) also significantly reduces STP. There is no significant difference between the effect of 50 and 200 μm Cd²⁺. D, Ni²⁺ has no effect on STP. E, sample traces showing mEPSCs prior to HFS in the presence of Cd²⁺ (top), immediately (middle) and 40 s following HFS (bottom). Note the delayed increase in mEPSCs observed in the bottom trace. Result of ANOVA, where significant, is indicated in the graphs. ***P < 0.001, Bonferroni post test.

In the presence of Cd²⁺, there was a small but significant increase in mEPSC frequency (n = 9; P < 0.05 versus basal frequency) that became more pronounced with time with a latency of 30 s to 3 min post-HFS (Fig. 3E). This result prompted us to question whether any Ca²⁺ influx occurred during HFS in Cd²⁺. Examination of synaptic response during HFS revealed that in 0 mm Ca²⁺ aCSF, there is no evoked transmission (n = 5; Fig. 4A). In contrast, 200 μm Cd²⁺ did not block the occasional synaptic response during the later portion of the HFS in 3 of 4 cells examined, although response to a single stimulus is completely blocked (Fig. 4A). One cell showed no evoked response to HFS. Therefore, during a train of synaptic activation, there seems to be some Ca²⁺ build-up in the presynaptic terminal, despite the presence of Cd²⁺, which triggers evoked as well as spontaneous glutamate release. A potential source of such Ca²⁺ influx is reverse operation of the Na⁺–Ca²⁺ exchanger (NCX) which may occur in periods of intense neuronal activity (Zhong et al. 2001; Minami et al. 2007). To ask whether the small STP seen in Cd²⁺ is explained by NCX, the reverse-mode specific blocker KB-R7943 (50–100 μm) was applied in combination with Cd²⁺. We found that even in the presence of these blockers, there was a delayed increase in the frequency of mEPSCs (n = 4; P < 0.05 compared to baseline frequency, Fig. 4B and C) similar to Cd²⁺ alone (P > 0.05, Fig. 4B), indicating that NCX is not involved in STP.

Figure 4. Low level of Ca²⁺ influx occurs during HFS.

A, synaptic responses during HFS, recorded in control condition (normal aCSF), Ca²⁺ free (0 mm Ca²⁺), in Cd²⁺ or KB-R7943 + Cd²⁺, as indicated. Blocking VGCCs and NCX does not completely eliminate synaptic transmission. Stimulus artifacts are blanked for the purpose of clear presentation. B, frequency of mEPSCs after HFS in the presence of Cd²⁺ or KB-R7943 + Cd²⁺. HFS was applied at time 0. Note the Y-axis is linear, not in logarithm. C, representative traces showing mEPSCs in KB-R7943 + Cd²⁺, immediately or 1 min after HFS, as indicated. mEPSC increase was minimal immediately following the stimulation, but potentiation appeared with a delay.

Multiple subtypes of high voltage-gated Ca²⁺ channels mediate STP

Our result suggests that a large portion of mEPSC-STP is mediated by presynaptic high-VGCCs. To characterize the subtypes of Ca²⁺ channels involved, the effect of specific blockers for N-, P/Q- or L-type Ca²⁺ channels was tested. ω-Conotoxin GVIA (ω-CTx: 0.1–1 μm), the N-type channel blocker, partially inhibited evoked EPSCs as we previously reported (Hirasawa et al. 2001). However, when HFS was applied after the effect of ω-CTx on evoked EPSC maximized, it had no effect on the magnitude or time course of mEPSC-STP (n = 8; P > 0.05, Fig. 5A). ω-Agatoxin TK (ω-Aga: 0.2–1 μm), the P/Q-type channel blocker, also caused partial inhibition of evoked EPSCs (Hirasawa et al. 2001) but again, mEPSC-STP was comparable to control conditions in the presence of 0.2 μm ω-Aga (n = 10; P > 0.05, Fig. 5B). Nicardipine (10 μm), the L-type channel blocker, was also without any significant effect on mEPSC-STP (n = 9; P > 0.05, Fig. 5C). This result prompted us to speculate that there was a functional overlap between multiple subtypes of Ca²⁺ channels so that blockade of one subtype did not alter the outcome. To test if this was the case, we examined the effects of combinations of two or all three Ca²⁺ channel blockers. Every two-blocker combination tested failed to block STP, i.e. ω-CTx + ω-Aga (n = 4; P > 0.05, Fig. 5D), ω-Aga + nicardipine (n = 4; P > 0.05, Fig. 5E), and ω-CTx + nicardipine (n = 4; P > 0.05, Fig. 5F). In contrast, when a cocktail of all three Ca²⁺ channel blockers was added, the overall magnitude of STP was significantly attenuated (n = 6; P < 0.001, Fig. 6A and B) without statistical change in the half-life of the first or second phase of STP in each condition (P > 0.05, Fig. 6C and D). In summary, these results suggest that initiation of mEPSC-STP depends on Ca²⁺ entry, but there is no specificity in the subtype of VGCCs mediating its entry.

Figure 5. Specific Ca²⁺ channel blockers fail to block STP.

ω-Conotoxin GVIA (ω-CTx) (A), ω-agatoxin TK (ω-Aga) (B) or nicardipine (C) applied alone has no effect on STP. Pairs of specific Ca²⁺ channel blockers also do not inhibit STP. Two blockers are simultaneously applied: D, ω-CTx and ω-Aga; E, ω-Aga and nicardipine; F, ω-CTx and nicardipine. Blockers are administered prior to (until their effect on evoked EPSC, if any, reached a plateau) and during HFS (time 0).

Figure 6. Ca²⁺ channels redundantly mediate the induction of STP.

A, HFS induces a robust STP in a representative cell, but its effect is attenuated in the presence of specific blockers for N-, P/Q- and L-type channels. Arrows indicate HFS. Inset, evoked EPSC is completely abolished by the cocktail of 3 blockers. B, time-effect plot of control STP and in the presence of ω-CTx, ω-Aga and nicardipine (3 blockers). Result of ANOVA is indicated in the graph. ***P < 0.001, Bonferroni post test. C and D, half-life of exponential decays (1st and 2nd phase) in control condition and in the presence of 3 Ca²⁺ channel blockers.

Maintenance of STP requires mitochondrial calcium sequestration

Our data suggest that extracellular Ca²⁺ is required for the initiation of STP. However, once established, STP seems to continue independently of significant neuronal firing for minutes, since TTX applied after HFS does not block it (Kombian et al. 2000). Therefore, we hypothesized that following HFS, another source of Ca²⁺ such as intracellular stores became involved in the maintenance of mEPSC-STP.

The endoplasmic reticulum (ER) Ca²⁺ store is filled during physiological neuronal activity and serves as a Ca²⁺ reservoir. Ca²⁺ release from the ER is known to contribute to presynaptic Ca²⁺ transients and enhance spontaneous neurotransmitter release in some synapses (Llano et al. 2000; Emptage et al. 2001). To test whether the ER was a Ca²⁺ source for the maintenance of mEPSC-STP, slices were treated with thapsigargin, a specific inhibitor of SERCA pumps. As shown in Fig. 7, 1 μm thapsigargin, a concentration that effectively blocks all SERCA pumps (Thastrup et al. 1990), had no effect on the baseline frequency or STP (n = 6; P > 0.05). Thus, the ER store is not responsible for STP.

Figure 7. Endoplasmic reticulum Ca²⁺ store is not involved in STP.

A, a representative cell showing that application of thapsigargin for 30 min has no effect on the magnitude of STP. Arrows indicate HFS. B, group data indicating no change in STP in the presence of thapsigargin.

In addition to the ER, mitochondria also act as an intracellular Ca²⁺ sink to regulate cytosolic Ca²⁺ levels and neurotransmitter release, in particular as the Ca²⁺ load rises to higher levels (David et al. 1998; Rusakov, 2006). In order to determine if sequestration of Ca²⁺ into mitochondria is necessary for the maintenance of STP, carbonyl cyanide 3-chlorophenylhydrazone (CCCP), a protonophore, was used to dissipate the mitochondrial inner membrane potential which would prevent Ca²⁺ uptake and induce release of already existing Ca²⁺ in the mitochondria (Werth & Thayer, 1994; Herrington et al. 1996). Disruption of mitochondrial membrane potential, however, would also cause depletion of cellular ATP through reverse operation of ATP synthase as the cell attempts to restore mitochondrial membrane potential. Depletion of ATP may in turn cause non-specific attenuation of various cellular processes. Thus, to specifically address the role of Ca²⁺ uptake into mitochondria, an inhibitor of ATP synthase oligomycin was used to avoid ATP depletion. Occasionally, a cocktail of CCCP and oligomycin induced a transient increase in the basal frequency of mEPSCs which may be due to the release of already stored Ca²⁺ (data not shown). Conversely, CCCP + oligomycin dramatically shortened the overall duration of STP as compared to control (n = 6; P < 0.0001; Fig. 8A and C), preferentially attenuating the second phase of STP (P < 0.05, Fig. 8B, bottom panel) without significantly altering the first phase (P > 0.05, Fig. 5B, top panel). These inhibitory effects were due to CCCP, because oligomycin alone had no effect (n = 5; P > 0.05, Fig. 8D). Lack of oligomycin effect also indicates that transient ATP production by mitochondria is not necessary for STP.

Figure 8. Mitochondrial Ca²⁺ sequestration underlies the second phase of STP.

A, a representative cell that showed a robust STP. Co-application of CCCP and oligomycin caused a faster return of mEPSC frequency to baseline following HFS (arrows). Note that the initial increase is intact. Following prolonged washout of the drugs, the 2nd phase shows a sign of recovery (last arrow). B, half-life of 1st and 2nd phase exponential decays. *P < 0.05. C, group result of the time course of STP in control and in the presence of CCCP and oligomycin. Compared to control, the initial potentiation is similar but the overall duration is curtailed. D, oligomycin alone has no effect on STP. E, CGP37157 also shortens the duration of STP. F, SPBN does not alter STP. Result of ANOVA, where significant, is indicated in the graphs.

One of the pathways by which Ca²⁺ can be released from mitochondria is through the mitochondrial Na⁺–Ca²⁺ exchanger (mitoNCX), which has been shown to mediate synaptic potentiation (Garcia-Chacon et al. 2006). Indeed, application of the mitoNCX inhibitor CGP37157 showed an effect similar to CCCP: inhibition of the second phase of STP without affecting the initial potentiation immediately following HFS (n = 6; P < 0.0001 versus control STP, P > 0.05 versus CCCP + oligomycin, Fig. 8E). Ca²⁺ uptake into mitochondria is also known to stimulate production of reactive oxygen species (ROS) such as superoxide (Bindokas et al. 1996; Hongpaisan et al. 2004). To test whether STP requires release of ROS from the mitochondria, the ROS scavenger, N-tert-butyl-a-(2-sulfophenyl)nitrone sodium (SPBN: 100 μm) was bath applied prior to stimulation at a concentration previously shown to inhibit superoxide-mediated effects (Hongpaisan et al. 2004). It was found that SPBN application had no effect on the magnitude or duration of STP (n = 5; P > 0.05, Fig. 8F), suggesting that ROS is not involved.

We also investigated the role of various sources of Ca²⁺ in the potentiation of mEPSC amplitude (Figs 1A and 2C). While the amplitude change depends on extracellular Ca²⁺ (Fig. 2C), none of the Ca²⁺ sources tested seem to be crucial, namely VGCCs, NCX, ER and mitochondria, since their respective blockers failed to prevent the rightward shift in amplitude distribution in most cells tested (data not shown). This result suggests that a very small amount of Ca²⁺ influx through a pathway independent of VGCCs or NCX may be sufficient to induce a significant amplitude increase.

Discussion

The present study demonstrates that HFS of excitatory afferent inputs to the SON induces a robust STP of mEPSC frequency that displays a two-phase exponential decay, indicating that the mechanisms underlying initiation and maintenance of this form of synaptic potentiation may be different. The first phase is immediately visible, large and fast decaying whereas the second phase becomes apparent successive to the first phase and decays much slower. The half-life of each component induced by stimulation frequencies of 50 or 100 Hz is not significantly different, suggesting that both mechanisms are activated by this frequency range. We demonstrate that this plasticity is exclusively dependent on extracellular Ca²⁺. Its induction involves activation of VGCCs, while the second phase of STP is supported by Ca²⁺ sequestration and release by the mitochondria. Furthermore, the activity state or Ca²⁺ rise in the postsynaptic cell has no apparent influence on these processes. Thus, mEPSC-STP is a form of presynaptic plasticity.

Multiple subtypes of high-voltage-gated calcium channels mediate STP

Application of the non-selective VGCC blocker Cd²⁺ (200 μm) largely abolished mEPSC-STP, suggesting that Ca²⁺ entry through these channels is critical. High-VGCCs rather than low-VGCCs seem to be particularly crucial, evident from the lack of effect of Ni²⁺ at a concentration (50 μm) that specifically blocks low-VGCCs, in contrast to the strong inhibition by 50 μm Cd²⁺ that is specific for high-VGCCs (Fisher & Bourque, 1995). Ni²⁺ at 50 μm is also known to block R-type Ca²⁺ channels (Randall & Tsien, 1995; Tottene et al. 2000), therefore this result also excludes the contribution of these channels. Millimolar concentrations of Cd²⁺ also inhibit P2X receptor-mediated Ca²⁺ influx (Capiod, 1998) which can induce glutamate release if located at the presynaptic terminal (Shigetomi & Kato, 2004). However, the concentrations of Cd²⁺ used in this study (50 and 200 μm) are known to selectively block VGCCs without significantly affecting P2X receptors (Shigetomi & Kato, 2004; Song et al. 2007).

In order to mimic the effect of Cd²⁺, it required co-application of specific blockers for N-, P/Q- and L-type channels, namely ω-CTx, ω-Aga and nicardipine, respectively. This indicates a functional redundancy for VGCCs with respect to mEPSC-STP induction, i.e. any one of N-, P/Q- and L-type channels can initiate it. In other words, if one of these subtypes is functional, it is capable of fully inducing STP regardless of the status of other channel subtypes. This may be due to overlapping Ca²⁺ microdomains caused by N-, P/Q- and L-type channels at every release site, allowing each subtype of channel to mediate maximum release. However, this is improbable because N- and P/Q-type channels but not L-type channels mediate isolated evoked EPSCs at these synapses, with N-type being more efficient than P/Q-type channels (Hirasawa et al. 2001). Rather, different subtypes of VGCCs probably have distinct localizations in the terminal relative to the active zone. L-type channels may not be expressed at the active zone and only come into play when a train of action potentials causes enough build up of Ca²⁺ through these channels (Fisher & Bourque, 2001). Reduction in buffering capacity of the terminal due to cytosolic Ca²⁺ accumulation may further augment free cytosolic Ca²⁺ transients generated by additional Ca²⁺ entry (Rusakov, 2006). Furthermore, strong depolarization induced by HFS may potentiate Ca²⁺ channels such as L-, N- and P/Q-type channels (Cloues et al. 1997; Brody & Yue, 2000; Currie & Fox, 2002). These alterations in Ca²⁺ dynamics during HFS may underlie the difference in VGCC subtype dependency of synaptic activation by single action potential versus a train of action potentials. The possibility remains that at lower stimulation intensities (low frequency or low pulse number), each VGCC subtype contributes differently to influence asynchronous transmitter release.

The fact that Cd²⁺, NCX blocker or the combination of VGCC blockers does not completely abolish synaptic response during HFS or mEPSC-STP is intriguing. Together with the observation that removal of extracellular Ca²⁺ eliminates the synaptic effects of HFS, our result suggests that there is some Ca²⁺ influx, even in a condition where VGCCs and NCX are inhibited. This may also explain the potentiation of mEPSC amplitude in the presence of VGCC and NCX blockers. While the increase in mEPSC amplitude is also exclusively dependent on extracellular Ca²⁺, as HFS had no effect in the Ca²⁺-free condition, none of the Ca²⁺ sources examined in this study is necessary, i.e. VGCCs, NCX, ER and mitochondria. Therefore, it seems that a very small amount of Ca²⁺ entry from the extracellular space is sufficient to induce the potentiation of mEPSC amplitude.

Although 200 μm Cd²⁺ is a commonly used concentration to block VGCCs, and although we have shown previously that 100 nm ω-CTx and 200 nm ω-Aga induce the maximal inhibitory effect on evoked EPSCs at excitatory synapses in the SON (Hirasawa et al. 2001), incomplete block of VGCCs cannot be ruled out, which may allow some Ca²⁺ influx during intense synaptic stimulation to induce a minor transmission. Nonetheless, the difference in the time course of mEPSC-STP seen in the absence or presence of Cd²⁺, i.e. immediate versus delayed increase in mEPSC, may signify an involvement of yet another mechanism of Ca²⁺ influx which remains to be determined. The Cd²⁺-insensitive delayed response may at least partially account for the second phase of STP.

Ca²⁺ uptake into mitochondria is necessary for the maintenance of STP

The time course of mEPSC-STP lasting up to tens of minutes suggested that it did not simply rely directly on the Ca²⁺ influx through presynaptic Ca²⁺ channels during HFS. Indeed, our results indicate that mitochondrial Ca²⁺ stores prolong asynchronous release into the second phase of STP whereas ER Ca²⁺ stores appear to be not involved. The mitochondrial uncoupler CCCP induced a strong inhibition only on the second phase of the STP without affecting the initial potentiation (first phase). Mitochondria are commonly found in glutamatergic terminals impinging on magnocellular neurons in the SON (Meeker et al. 1993), thus strategically positioned to sense the Ca²⁺ microdomains generated by the opening of Ca²⁺ channels and to influence transmitter release.

Ca²⁺ sequestration into mitochondria may have several consequences that can impact transmitter release. Following Ca²⁺ load during HFS, Ca²⁺ may be released from mitochondria slowly over minutes, sustaining cytosolic Ca²⁺ above resting levels which may in turn trigger exocytosis (Tang & Zucker, 1997; Medler & Gleason, 2002). We have found that preventing Ca²⁺ efflux from mitochondria through mitoNCX significantly attenuates STP, supporting this idea. Conversely, Ca²⁺ uptake by mitochondria has been shown to help keep the low level of cytosolic Ca²⁺, preventing desensitization of VGCCs (Medler & Gleason, 2002) and synaptic depression (Billups & Forsythe, 2002). This is an unlikely mechanism, however, because once STP is initiated, it does not require significant depolarization (TTX insensitive) and hence VGCC activation.

ATP synthesis and ROS production are also stimulated by Ca²⁺ uptake into mitochondria (Carriedo et al. 2000) which can in turn affect transmitter release. Mitochondrial ATP production has been shown to facilitate mobilization of reserve pool vesicles (Verstreken et al. 2005). However, lack of an oligomycin effect, a mitochondrial ATP synthase inhibitor, suggests that this is not the case. This result also indicates that ATP production through glycolysis is sufficient to support STP, although a long-term effect of oxidative phosphorylation blockade is possible. ROS such as superoxide has been shown to act as small messenger molecules in physiological conditions, for example by activating protein kinases including PKC, PKA and CaMKII (Knapp & Klann, 2002; Hongpaisan et al. 2004), which may subsequently facilitate transmitter release (Hirasawa & Pittman, 2003). Nonetheless, our data suggest that STP is independent of ROS, since the ROS scavenger SPBN shows no inhibitory effect.

Synaptic plasticity of mEPSC versus evoked EPSCs

Panatier et al. reported a long-term potentiation (LTP) of evoked EPSCs in the SON by using an intense HFS somewhat similar to our current study, i.e. 100 Hz, 1 s, 4 times applied in current-clamp mode (Panatier et al. 2006). Evoked EPSC-LTP is dependent on NMDA receptor activation and postsynaptic Ca²⁺ (Panatier et al. 2006). In our study, a holding potential of −80 mV was used for most of the experiments which would prevent the activation of NMDA receptors on the postsynaptic cell. To eliminate the possibility that NMDA receptors on neighbouring cells that are activated during HFS can somehow signal to modulate the excitatory synapses under investigation, we also tested bath application of the NMDA receptor antagonist d-AP5 and found that this had no effect on mEPSC-STP (data not shown). Furthermore, we were able to induce mEPSC-STP in the presence of EGTA in the recording pipette, a condition that has been shown to block evoked EPSC-LTP. Finally, potentiation of mEPSCs can be induced with a very modest stimulation, for example stimulation frequencies as low as 1 Hz or the number of stimulations as few as 10 pulses at 100 Hz (M Hirasawa, unpublished observation). Collectively, it is clear that mEPSC-STP is not a by-product of evoked EPSC-LTP but is synaptic plasticity in its own right with a distinct mechanism of induction.

Physiological implication

STP may occur in physiological conditions when presynaptic glutamatergic neurons fire trains of action potentials and may correspond to EPSP clusters that underlie the bursting activities of magnocellular neurons (Jourdain et al. 1998) or higher level of interburst basal firing (Israel et al. 2003; Moos et al. 2004). Although mEPSCs are exclusively mediated by AMPA receptors in our preparation, the magnitude of mEPSC-STP may also influence NMDA receptor-dependent events such as LTP of evoked EPSCs (Panatier et al. 2006) and dendritic neuropeptide release (de Kock et al. 2004) by providing glutamate in the synaptic cleft and prolonged background excitation.

Unlike fast neurotransmitters such as glutamate, release of neuropeptides from dense core vesicles requires certain firing patterns and a long latency. To realize adequate amounts of hormone release in response to peripheral needs, it is conceivable that the excitatory neurons projecting to the SON employ a mechanism in a form of mEPSC-STP to maintain the activity of magnocellular neurons even after the termination of their own firing. STP does not sensitize or desensitize, meaning repeated HFS reliably induces STP of the same magnitude each time. This would allow the presynaptic neurons to conserve energy, as only intermittent, brief trains of firing are necessary to maintain the excitatory tone to control the postsynaptic cell. Redundant involvement of multiple subtypes of Ca²⁺ channels in this process could represent a fail-proof mechanism, indicating STP may be important for the physiological function of the SON.