Presynaptic Mitochondrial Calcium Sequestration Influences Transmission at Mammalian Central Synapses

Abstract

Beyond their role in generating ATP, mitochondria have a high capacity to sequester calcium. The interdependence of these functions and limited access to presynaptic compartments makes it difficult to assess the role of sequestration in synaptic transmission. We addressed this important question using the calyx of Held as a model glutamatergic synapse by combining patch-clamp with a novel mitochondrial imaging method. Presynaptic calcium current, mitochondrial calcium concentration ([Ca²⁺]_mito, measured using rhod-2 or rhod-FF), cytoplasmic calcium concentration ([Ca²⁺]_cyto, measured using fura-FF), and the postsynaptic current were monitored during synaptic transmission. Presynaptic [Ca²⁺]_cytorose to 8.5 ± 1.1 μm and decayed rapidly with a time constant of 45 ± 3 msec; presynaptic [Ca²⁺]_mito also rose rapidly to >5 μm but decayed slowly with a half-time of 1.5 ± 0.4 sec. Mitochondrial depolarization with rotenone and carbonyl cyanidep-trifluoromethoxyphenylhydrazone abolished mitochondrial calcium rises and slowed the removal of [Ca²⁺]_cyto by 239 ± 22%. Using simultaneous presynaptic and postsynaptic patch clamp, combined with presynaptic mitochondrial and cytoplasmic imaging, we investigated the influence of mitochondrial calcium sequestration on transmitter release. Depletion of ATP to maintain mitochondrial membrane potential was blocked with oligomycin, and ATP was provided in the patch pipette. Mitochondrial depolarization raised [Ca²⁺]_cyto and reduced transmitter release after short EPSC trains (100 msec, 200 Hz); this effect was reversed by raising mobile calcium buffering with EGTA. Our results suggest a new role for presynaptic mitochondria in maintaining transmission by accelerating recovery from synaptic depression after periods of moderate activity. Without detectable thapsigargin-sensitive presynaptic calcium stores, we conclude that mitochondria are the major organelle regulating presynaptic calcium at central glutamatergic terminals.

Neurotransmitter release is triggered by Ca²⁺ influx through voltage-gated channels (Katz, 1969). Factors regulating the spatiotemporal profile of the presynaptic cytoplasmic calcium transient influence release probability, vesicle recycling, and information processing in the brain. Mitochondria are capable of having an important effect on these processes because they can sequester large quantities of calcium (Duchen, 1999; Nicholls and Budd, 2000), are present at high concentrations in presynaptic terminals, and can be tethered to vesicle release sites (Rowland et al., 2000). Presynaptic buffering of calcium by mitochondria occurs at the mouse (David and Barrett, 2000), lizard (David et al., 1998; David, 1999), and crustacean (Tang and Zucker, 1997) neuromuscular junctions, but is inconsistently observed at the goldfish retinal bipolar cell terminal (Kobayashi and Tachibana, 1995; Zenisek and Matthews, 2000). At the crustacean neuromuscular junction, mitochondrial calcium sequestration over 7–10 min of tetanic stimulation, and its subsequent release, causes post-tetanic potentiation (PTP) (Tang and Zucker, 1997). Although mitochondria are known to buffer calcium and influence secretion of peptides from sympathetic ganglia (Peng, 1998; Cao and Peng, 1999), pituitary gonadotropes (Kaftan et al., 2000), and chromaffin cells (Herrington et al., 1996; Park et al., 1996; Xu et al., 1997; Giovannucci et al., 1999; Sorimachi et al., 1999; Montero et al., 2000), relatively little is known about the role mitochondria play in influencing neurotransmission in the brain because conventional mitochondrial calcium imaging techniques are difficult to applyin situ. Additional complications arise because mitochondrial depolarization blocks both calcium accumulation and ATP synthesis, leading to indirect effects on synaptic transmission attributable to ATP depletion.

We developed a technique to image mitochondrial calcium concentrations ([Ca²⁺]_mito) in single cells and synaptic terminals by including the AM form of a rhod-based dye (rhod-2 or rhod-FF) in the patch pipette. These dyes are cationic (Minta et al., 1989) and are preferentially accumulated into the mitochondria when AM is loaded into cells. However, after AM loading of intact cells with rhod dyes, a substantial cytoplasmic component is still observed (typically 30%) (Kaftan et al., 2000), which must be reduced by up to 24 hr of washing (Simpson and Russell, 1996, 1998), permeabilizing cells with digitonin (Rohacs et al., 1997; Szabadkai et al., 2001), or manganese quenching (Mironov and Richter, 2001). By including the AM dye in the patch pipette we can avoid problems of cytoplasmic loading because cytoplasmic rhod fluorescence is removed by dialysis (back into the patch pipette), leaving fluorescence solely from the mitochondrial compartment. Extensive periods of washing or quenching of cytoplasmic fluorescence were not required, and measurement of cytoplasmic calcium concentrations ([Ca²⁺]_cyto) in the same terminal was possible by also including fura-FF potassium salt in the patch pipette.

In this study, we examined the role of mitochondria in a mammalian glutamatergic synapse known as the calyx of Held (Forsythe, 1994; Borst et al., 1995; von Gersdorff et al., 1997). Using simultaneous electrophysiological recording combined with intracellular calcium imaging of both [Ca²⁺]_cyto and [Ca²⁺]_mito, we examined the influence of mitochondria on neurotransmitter release on millisecond time scales.

MATERIALS AND METHODS

Electrophysiology. Transverse brainstem slices (∼150-μm-thick) were prepared from 10- to 12-d-old Lister Hooded rats, killed by decapitation. Slices were incubated for 1 hr at 37°C in artificial CSF (aCSF), and then stored for up to 8 hr at room temperature before transferring to the experimental chamber for recording at 37°C. Medial nucleus of the trapezoid body (MNTB) neurons and calyces of Held were visualized with infrared differential interference contrast (DIC) optics on a Nikon E600FN microscope with a 60×, numerical aperture 1.0, water-immersion, fluor lens. The aCSF contained (in mm): 125 NaCl, 2.5 KCl, 10 glucose, 1.25 NaH₂PO_4, 26 NaHCO_3, 1 MgCl_2, 2 CaCl_2, 3 myo-inositol, 0.5 ascorbic acid, 2 Na-pyruvate, 10 TEA-Cl, and 0.001 TTX, pH 7.4, when gassed with 95% O₂ and 5% CO₂. Whole-cell patch-clamp recordings were made from both the synaptic terminal and postsynaptic cell using thick-walled glass pipettes (GC150F-7.5; Clark Electromedical, Reading, UK) with Axopatch 200B amplifiers (Axon Instruments, Foster City, CA), filtered at 5 kHz (8-pole Bessel filter) and sampled at 10 kHz. Currents were recorded with pClamp software (Axon Instruments). Whole-cell access resistances were <10MΩ for the postsynaptic cell and <30MΩ for the presynaptic terminal, and were compensated >70% with a 10 μsec lag time. The intracellular solution for the postsynaptic cell contained (in mm): 110 CsCl, 40 HEPES, 10 TEA-Cl, 12 Na₂-phosphocreatine, and 1 EGTA, pH adjusted to 7.3 with CsOH. The intracellular solution for the presynaptic terminal contained (in mm): 110 CsCl, 40 HEPES, 10 TEA-Cl, 12 Na₂-phosphocreatine, 0.2 EGTA, 10 Na-glutamate, 2 Mg-ATP, and 0.5 Na-GTP, pH adjusted to 7.3 with CsOH. A free calcium concentration of 100 nm (measured with fura-2) was attained by the addition of 8 μm CaCl₂. A liquid junction potential of 3.4 mV was not corrected for. During paired presynaptic and postsynaptic recordings postsynaptic AMPA receptor responses were pharmacologically isolated, and desensitization and saturation minimized by the addition of 50 μmcyclothiazide and 3 mm kynurenate to the external medium. NMDA receptors were blocked with 50 μmdl-2-amino-5-phosphonopentanoic acid and 10 μm (+)-MK 801 maleate.

Calcium imaging. Calcium-sensitive fluorescent dyes (200 μm fura-FF potassium salt and 10 μm of either rhod-2 AM or rhod-FF AM) were simultaneously added to the presynaptic intracellular solution to visualize the calcium transients in the cytosol and mitochondria of the terminal. After establishment of whole-cell recording, ∼10 min was required for rhod fluorescence to be detected in the mitochondria. Dyes were excited at 380 and 550 nm, respectively, with a Polychrome II monochromator (T.I.L.L. Photonics, Martinsried, Germany). Emitted light was separated by 400 or 575 nm dichroic mirrors and filtered with either a 420 or a 590 nm long-pass emission filter. Fluorescence images were acquired every 10 msec with a PentaMAX cooled CCD camera via a Gen IV image intensifier (Princeton Instruments, Trenton, NJ) and analyzed with MetaFluor software (Universal Imaging, West Chester, PA). Imaging of [Ca²⁺]_cyto and [Ca²⁺]_mito in the same cell was achieved by stimulating once while imaging at 380 nm, followed by once at 550 nm, repeating this alternate excitation and averaging the relevant signals. An interval between each stimulus of 2 min was required to allow re-establishment of baseline calcium levels; hence long duration stable recordings were required for data acquisition. Presynaptic calcium transients were imaged with fura-FF using only 380 nm excitation to enhance the rate of data acquisition and also because kynurenic acid absorbs much of the excitation light below 360nm. The fluorescence was converted to [Ca²⁺]_i using the following equation (Jaffe et al., 1992; Helmchen et al., 1997):

$$[{\text{Ca}}^{2+}{]}_{\text{i}}=\frac{[{\text{Ca}}^{2+}{]}_{\text{rest}}+K_{\text{d}}(\mathrm{\Delta }F/F)/(\mathrm{\Delta }F/F{)}_{max}}{(1-(\mathrm{\Delta }F/F)/(\mathrm{\Delta }F/F{)}_{max})}$$

A value of −0.76 ± 0.03 was established for (ΔF/F)_max at the end of 20 experiments by repeated electrical zapping of the terminal (−1.3 V pulses for 5 msec) to cause a saturating [Ca²⁺]_cyto rise. AK_d of 10 μmwas used, as previously determined for fura-FF in this preparation (Bollmann et al., 2000; Schneggenburger and Neher, 2000).

Fluorescent dyes were obtained from Molecular Probes (Eugene, OR), glutamate receptor antagonists from Tocris Cookson (Bristol, UK), TTX from Latoxan (Valence, France), and thapsigargin and Ru360 from Calbiochem (Nottingham, UK). All other chemicals were obtained from Sigma (Poole, UK). Ru360 was diluted from a 1 mm stock made daily in deoxygenated 1 mm Na-ascorbate. Frozen Ru360 stock solution lost activity overnight. Data are expressed as the mean ± SEM. Statistical significance (p < 0.05) was tested with paired or unpaired two-tailed t test as appropriate. All experiments were performed at physiological temperature (35–37°C).

RESULTS

Calcium influx into the calyx of Held (Fig.1A) was triggered by 2 msec step depolarizations under voltage-clamp conditions in the presence of sodium and potassium channel blockers (Helmchen et al., 1997; Borst and Sakmann, 1998). To maintain the balance between physiological calcium homeostatic mechanisms, experiments were conducted at a temperature of 35–37°C, and intracellular mobile calcium buffering was maintained with 200 μmEGTA (Borst and Sakmann, 1996). Inclusion in the patch pipette of the low-affinity calcium indicator fura-FF (Fig. 1B) allowed quantitative imaging of [Ca²⁺]_cyto(Helmchen et al., 1997; Bollmann et al., 2000; Schneggenburger and Neher, 2000) simultaneously with measurement of the presynaptic calcium current generated by P-type calcium channels (Forsythe et al., 1998;Iwasaki and Takahashi, 1998) (Fig. 1C). Presentation of four stimuli at 100 Hz caused a rapid rise of presynaptic [Ca²⁺]_cyto to 8.5 ± 1.1 μm (n = 5) (Fig. 1D) and decayed in a double exponential with time constants (τ) of 43 ± 2 and 311 ± 59 msec (τ_fast was 83 ± 2% peak amplitude).

Fig. 1.

Rapid time course of presynaptic [Ca²⁺] transients in the calyx of Held.A, DIC image of a calyx of Held (arrow) surrounding a postsynaptic MNTB neuron (asterisk). Scale bar, 15 μm. Patch pipette is to the right.B, Fluorescence image from the same calyx filled with fura-FF via diffusion from the patch pipette. C, Calcium currents (bottom trace) recorded in the calyx of Held in response to four voltage steps (−80 to 0 mV depolarization, 2 msec duration at 100 Hz; top trace). D,[Ca²⁺]_cyto time course during stimulation recorded simultaneously with the calcium current from the same terminal as in C using fura-FF imaging at a frame frequency of 100 Hz (stimulation at arrow as shown in C).

Mitochondrial calcium sequestration was first examined by imaging the fluorescent calcium indicator rhod-2 (Minta et al., 1989), which preferentially accumulates in mitochondria. The AM form of rhod-2 is cationic and membrane-permeant but possesses no detectable calcium-dependent fluorescence. When added via the patch solution, it accumulated in mitochondria because of the large transmembrane potential (Δψ_m) and was activated by mitochondrial esterases. Once de-esterified, rhod-2 is retained in the mitochondrial matrix, regardless of Δψ_m. Rhod-2 fluorescence was absent from the cytoplasm and axon (Fig.2C) because of dialysis of cytoplasmic de-esterified rhod-2 (and possibly the esterases themselves) into the patch pipette. The different fluorescent absorbance spectra of rhod-2 and fura-FF permitted measurement of both [Ca²⁺]_cyto and [Ca²⁺]_mito in the same presynaptic terminal. Depolarization-evoked calcium entry into the presynaptic terminal resulted in calcium accumulation in the mitochondria, as indicated by an increase in rhod-2 fluorescence (Fig.2D).

Fig. 2.

Slow time course of presynaptic mitochondrial [Ca²⁺] transients. A, DIC image of the presynaptic terminal. Scale bar, 15 μm. B,Projected stack of fura-FF images focused through the entire terminal.C, Projected stack of rhod-2 images showing that fluorescence was concentrated in the terminal but was absent from the pipette and axon regions. D,[Ca²⁺]_mito rose rapidly on stimulation (as in Fig. 1C, black trace) but decayed slowly. Four groups of four stimuli (the same protocol as in Fig. 1Crepeated 4 times 150 msec apart, gray trace) had a similar time course and amplitude, indicating saturation of the dye. Mitochondrial depolarization (perfusion of 25 μm rotenone and 1 μm FCCP in the presence of 5 μg/ml oligomycin) completely blocked calcium accumulation by the mitochondria after one group of four stimuli (bottom black trace).E, Rhod-FF fluorescence signals were not saturated by four groups of four stimuli (gray trace) compared with one group of stimuli (black trace). Thetraces in D and E are shown normalized to the change in fluorescence after one group of stimuli. F, Summary data show that mitochondrial calcium uptake was inhibited by mitochondrial depolarization and 10 μm TPP⁺ but not 1 μmthapsigargin. Mitochondrial calcium sequestration saturated rhod-2 (K_d, 570 nm) but not rhod-FF (K_d, 19 μm).Asterisks indicate statistical significance (p < 0.05).

Saturation of rhod-2 by calcium (K_d, 570 nm) severely limited the ability to resolve changes in [Ca²⁺]_mito. Repetition of the stimulus produced no additional increase in fluorescence signal (n = 4; p = 0.27) (Fig. 2D, gray trace), suggesting that the rhod-2 was indeed saturated in these experiments. In contrast, repetitive stimulation of presynaptic terminals after mitochondrial loading with the lower affinity dye rhod-FF (K_d, 19 μm), resulted in progressive summation and greater fluorescent responses to repetitive stimulation (n = 4; p < 0.01) (Fig. 2E, gray trace). Because rhod-FF was not saturated under these conditions, the time course of [Ca²⁺]_mito could be determined. The half rise-time of [Ca²⁺]_mito after four stimuli at 100 Hz was 47 ± 3 msec, and the half decay time 1.5 ± 0.4 sec (n = 3). Although fura-FF and rhod-FF have similar affinities for calcium, the rhod-FF signal was notably slower than the kinetics of [Ca²⁺]_cytorecorded with fura-FF (Fig. 1D), consistent with the notion that these two dyes are signaling calcium changes in different subcellular compartments. The peak [Ca²⁺]_mito could not be estimated from the rhod-FF data because we could not measure a calcium-saturated fluorescent signal from the mitochondria. However, the rhod-2 data (Fig. 2D) represents an increase in [Ca²⁺]_mito of at least 5 μm, taking into account itsK_d, saturation, and assuming a resting [Ca²⁺]_mito of 100–200 nm (Babcock et al., 1997; Zhou et al., 1998). The effects of mitochondrial inhibitors on the fluorescence change were used to confirm the mitochondrial origin of the rhod fluorescence signal.

Inhibition of mitochondrial calcium sequestration

Accumulation of calcium by mitochondria was inhibited by three separate methods. First, agents were applied that disrupt Δψ_m because mitochondrial calcium accumulation is driven by Δψ_m (Gunter and Pfeiffer, 1990; Gunter and Gunter, 1994; Duchen, 1999; Nicholls and Budd, 2000). Calcium uptake into mitochondria was inhibited by >93% (n = 5; p < 0.05) (Fig.2D) with perfusion of rotenone (a blocker of complex I of the respiratory chain) combined with carbonyl cyanidep-trifluoromethoxyphenylhydrazone (FCCP) (a protonophore) and oligomycin (a blocker of F₀F₁-ATPase, preventing use of ATP to maintain proton gradients). Second, tetraphenylphosphonium (TPP⁺), which blocks mitochondrial calcium efflux without effecting ATP production (Aiuchi et al., 1985) (but see Nicholls and Budd, 2000) and mitochondrial calcium uptake (Tang and Zucker, 1997) also blocked the [Ca²⁺]_mito rise after stimulation of the calyx of Held (n = 4;p < 0.05) (Fig. 2F). Third, the inclusion in the patch pipette of Ru360 (a specific inhibitor of the mitochondrial calcium uniporter) (Matlib et al., 1998) prevented the stimulus-induced rise in [Ca²⁺]_mito(n = 3; p < 0.01) (Fig.3F). By contrast, thapsigargin, which blocks calcium pumps into intracellular stores, did not prevent the rise of rhod-2 fluorescence (n = 3;p = 0.19) (Fig. 2F). The antagonist profile outlined in Figure 2 and the prolonged time course of the [Ca²⁺]_mito rise clearly demonstrate that the rhod-AM dyes, loaded by this method, report Δ[Ca²⁺] from the mitochondrial compartment alone.

Fig. 3.

Mitochondrial calcium sequestration buffers [Ca²⁺]_cyto. A, The [Ca²⁺]_cyto transient (using the same stimulation protocol as Fig. 1) was significantly slowed by mitochondrial depolarization, but the peak rise was not significantly altered (97 ± 6% of control; n = 5;p = 0.66). B, The τ_fast [Ca²⁺]_cyto decay was slowed by mitochondrial depolarization and TPP⁺but not by thapsigargin (129 ± 9% of control;n = 3; p = 0.08). C, D, Trains of presynaptic depolarization (−80 to 0 mV, 1 msec stimulation repeated 20 times at 200 Hz) produced rises in [Ca²⁺]_cyto that were also significantly slowed by mitochondrial depolarization or 1 μm Ru360. E, F, [Ca²⁺]_mito recorded with rhod-FF increased during stimulation (at arrow as inC). This increase was blocked by mitochondrial depolarization or Ru360. Asterisks indicate statistical significance (p < 0.05).

Mitochondria rapidly buffer presynaptic calcium

Dissipation of Δψ_m by rotenone and FCCP in the presence of oligomycin significantly slowed the [Ca²⁺]_cytotransient after stimulation of the presynaptic terminal (Fig.3A). The τ_fast was slowed by 45 ± 6% (n = 5; p < 0.01) (Fig.3B). This slowing was not attributable to ATP depletion from the terminal, because ATP was continually available by dialysis from the patch pipette, and application of oligomycin alone had no significant effect on τ_fast (110 ± 7% of control; n = 3; p = 0.30). This excludes the possibility that mitochondrial inhibition slows the [Ca²⁺]_cytotransient by reducing cytoplasmic [ATP], which could inhibit the plasma membrane calcium ATPase and, indirectly, Na⁺-Ca²⁺exchange. In addition, TPP⁺ and Ru360, which do not affect mitochondrial ATP production, had similar effects on τ_fast, slowing it to 203 ± 9% and 412 ± 88%, respectively (Fig.3B,D, note the different stimulus protocols). Thapsigargin had no significant effect on [Ca²⁺]_cyto(n = 3; p = 0.08) (Fig. 3B), consistent with a lack of substantial calcium sequestration into the endoplasmic reticulum. Longer, faster trains of stimuli (1 msec step depolarizations, repeated 20 times at 200 Hz) were also used, which mimic the action potential firing patterns observed at this synapse (Brownell, 1975). As with the briefer stimuli, mitochondrial depolarization significantly slowed the decay of the [Ca²⁺]_cytotransient (Fig. 3C,D), τ_fast being slowed to 436 ± 32% (from 45 ± 3 to 239 ± 22 msec; n = 6;p < 0.01). Concurrent [Ca²⁺]_mitorecordings with rhod-FF demonstrated that mitochondrial depolarization or inclusion of Ru360 inside the patch pipette prevented mitochondrial calcium sequestration during the longer trains of stimulation (Fig.3E,F).

Mitochondrial depolarization had no effect on the magnitude of the presynaptic calcium current (I_Ca). A statistically insignificant time-dependent rundown ofI_Ca was observed at physiological temperature (I_Ca declined by 15 ± 4% over ∼30 min; n = 5; p = 0.06). I_Ca was stable in experiments performed at room temperature, whereas mitochondrial depolarization still slowed [Ca²⁺]_cytotransients at these lower temperatures (data not shown). TPP⁺, however, reducedI_Ca by 44 ± 8% (n = 3; p < 0.05), and the peak [Ca²⁺]_cyto by 36 ± 9%, suggesting that TPP⁺ has additional nonspecific effects on presynaptic voltage-gated calcium channels.

Presynaptic mitochondria influence neurotransmitter release

To assess whether mitochondrial calcium buffering affects neurotransmitter release, simultaneous recordings were made from the presynaptic terminal and postsynaptic neuron (Fig.4A) (Borst et al., 1995; Takahashi et al., 1996). This was done in the presence of cyclothiazide and kynurenic acid to minimize receptor desensitization and saturation (Neher and Sakaba, 2001). A conditioning train of presynaptic step depolarizations (as in Fig. 3C) was followed 500 msec later by one test depolarization to assess the degree of recovery from synaptic depression. EPSCs recorded from the postsynaptic MNTB neuron during the conditioning train showed substantial synaptic depression (Fig. 4B), which recovered within 500 msec. TheI_Ca elicited by the test depolarization was reduced to 90.4 ± 1.6% (n = 3; p = 0.03) compared with the firstI_Ca of the conditioning train (Fig.4C,D) because of the build-up of inactivation of the presynaptic calcium channels (Forsythe et al., 1998). However, the magnitude of this reduction was unchanged by mitochondrial depolarization or inhibiting mitochondrial calcium sequestration with Ru360 (n = 3; p = 0.17 and 0.27, respectively) (Fig. 4D), indicating that the action of these drugs on neurotransmitter release (see below) was not mediated by a change in I_Ca. The size of the test EPSC was measured under control conditions and after mitochondrial inhibitors in the same terminals. A difference in the test EPSC amplitude should indicate whether mitochondrial calcium sequestration modulates neurotransmitter release. Under control conditions, the test EPSC recovered to 81 ± 4% (n = 3) of the initial train amplitude after 500 msec. After mitochondrial depolarization (reducing Ca²⁺ sequestration but maintaining presynaptic [ATP] as for Fig. 3) the rate of recovery of the EPSC train was slowed, such that the magnitude of the test EPSC recovered to only 40 ± 5% of control after 500 msec (n = 3; p < 0.01) (Fig.4E,F). Similarly, Ru360 slowed the recovery rate of the EPSC train, resulting in the magnitude of the test EPSC recovering to 51 ± 6% of the initial amplitude (n = 3; p = 0.01). To demonstrate that the synaptic depression induced by mitochondrial depolarization is mediated by changes in the dynamics of the presynaptic calcium transient, the concentration of the calcium buffer (EGTA) in the presynaptic terminal was increased from 200 μmto 1 mm (Borst and Sakmann, 1996; Dittman and Regehr, 1998; Wang and Kaczmarek, 1998; Wu and Borst, 1999). With this additional presynaptic calcium buffering the τ_fast of the [Ca²⁺]_cytotransient was 35 ± 3 msec under control conditions and only slowed to 73 ± 10 msec during mitochondrial depolarization (compare Figs. 3C, 4G). The effects of mitochondrial depolarization on the rate of recovery of the EPSC from synaptic depression were completely abolished by the additional EGTA (p = 0.38; n = 3) (Fig.4H), strongly implicating [Ca²⁺]_cyto in mediating this effect. Thus, our present data using two independent methods of blocking mitochondrial calcium sequestration, indicate that mitochondria can modulate synaptic transmission by accelerating the recovery from short-term presynaptic depression.

Fig. 4.

Presynaptic mitochondrial calcium sequestration influences synaptic transmission. A, The postsynaptic MNTB neuron and presynaptic calyx were simultaneously voltage clamped. Scale bar, 10 μm. B, Trains of stimuli (as in Fig. 3C) produced a markedly depressing EPSC train. At 500 msec after the end of the train, the EPSC had recovered to 81% of control amplitude. C, D,Presynaptic calcium currents recorded in response to the train were unaffected by mitochondrial depolarization. E,Postsynaptic currents normalized to the magnitude of the first EPSC. The absolute magnitude of the EPSC was −4.6 ± 0.4 nA before mitochondrial depolarization and −4.3 ± 0.4 nA after. This 6 ± 2% reduction was a time-dependent run-down and was not statistically significant (n = 3;p = 0.07). Mitochondrial depolarization induced by rotenone and FCCP in the presence of oligomycin had no effect on the initial train amplitude and time course, but it significantly reduced the relative magnitude of the test EPSC. F, Identical results were observed after mitochondrial depolarization or by blocking the calcium uniporter with intracellular Ru360 (1 μm).G, Inclusion of 1 mm EGTA in the presynaptic patch pipette accelerated the decay and reduced the effect of mitochondrial depolarization on the presynaptic calcium transient.H, Presynaptic EGTA at 1 mm abolished the effects of mitochondrial depolarization on the rate of recovery from synaptic depression. Asterisks indicate statistical significance (p < 0.05).

DISCUSSION

We have described a technique for imaging both cytoplasmic and mitochondrial calcium in the same presynaptic terminal at a mammalian central synapse. Using this method we showed that mitochondria rapidly sequester substantial quantities of cytoplasmic calcium, significantly buffering [Ca²⁺]_cyto and influencing neurotransmitter release on millisecond time scales. The selective loading of fluorescent dyes relies on the ability to include the AM ester of the rhod dye in the patch pipette, resulting in accumulation of de-esterified dye in the mitochondria. Loading is thus limited to the mitochondria within the patch-clamped cell, and cytoplasmic fluorescence is removed by dialysis into the pipette. Loading occurs rapidly (within a few minutes), and extensive periods of washing to remove cytoplasmic rhod are not required. Manganese quenching is also not required, which enables simultaneous use of a cytoplasmic dye and avoids problems with manganese block of calcium channels and neurotransmission (Hagiwara and Nakajima, 1966; Llinas et al., 1981; Augustine et al., 1985).

We are confident that the fluorescence labeling seen with the rhod dyes is mitochondrial for three reasons: First, contrasting localization of rhod and fura dyes (Fig. 2B,C). Second, different time course of the mitochondrial calcium signal compared with [Ca²⁺]_cyto(compare Figs. 1D, 2E). And third, the sensitivity of the fluorescence signal to mitochondrial depolarization, TPP⁺ and RU360. The lack of effect of thapsigargin on the rhod fluorescence signal demonstrates that the signal does not originate from the endoplasmic reticulum and implies that mitochondria are the predominant presynaptic calcium store.

Simultaneous mitochondrial and cytoplasmic imaging

The long excitation wavelength of rhod-2 and rhod-FF (500–600 nm) enables simultaneous measurement of [Ca²⁺]_mito and [Ca²⁺]_cyto by combined loading with shorter wavelength dyes such as fluo-3 (Simpson and Russell, 1998; Boitier et al., 1999; Gonzalez et al., 2000;Trollinger et al., 2000; Collins et al., 2001; Rakhit et al., 2001), fura-2 (Simpson and Russell, 1996; Drummond and Tuft, 1999; Drummond et al., 2000), calcium green (Babcock et al., 1997; Peng and Greenamyre, 1998), and Oregon green BAPTA 5N (David et al., 1998). In our studies we combined rhod AM loading with fura-FF salt in the patch pipette. The salt of the fura dye is not membrane-permeable and will remain in the cytoplasm, ensuring that fura and rhod report [Ca²⁺] from separate subcellular compartments (Fig. 2B,C). Fura-FF has a relatively low affinity for calcium (Bollmann et al., 2000;Schneggenburger and Neher, 2000) and is ideal for imaging the large presynaptic rises in [Ca²⁺]_cyto with minimal spectral overlap with rhod dyes.

The magnitude of [Ca²⁺]_mito rises

A brief stimulus train (100 msec at 200 Hz) resulted in a rapid rise in [Ca²⁺]_mitothat saturated rhod-2 (Fig. 2D). Rhod-FF is a new low-affinity indicator that overcomes this saturation problem and enables the large rises in [Ca²⁺]_mito to be resolved (Fig. 2E). During stimulation, a minimum [Ca²⁺]_mito rise of ∼5 μm was calculated based on the resting calcium level and saturation of rhod-2. This rise is consistent with studies using mitochondrially targeted aequorin proteins that have shown [Ca²⁺]_mitorises of 5–20 μm in transfected cell lines on agonist application (Rizzuto et al., 1992; Rutter et al., 1996) and ∼300 μm in chronically depolarized bovine chromaffin cells (Montero et al., 2000). Our findings contrast with studies at the lizard neuromuscular junction where [Ca²⁺]_mito rises after stimulation are delayed and do not saturate rhod-2 (David et al., 1998). It seems likely that the proximal location of presynaptic mitochondria to release sites (Rowland et al., 2000) has a major impact on the rate and functional significance of mitochondrial calcium sequestration at the mammalian synapse. The distal location of mitochondria at goldfish bipolar cell synapses may explain their minor contribution to sequestration at that site (Zenisek and Matthews, 2000).

Prevention of ATP depletion

Blocking mitochondrial calcium sequestration had a significant effect on the release of neurotransmitter after a train of stimulation (Fig. 4E). There are four reasons why this effect was not mediated by depletion of ATP after mitochondrial poisoning: first, ATP was continually applied via the patch pipette. Second, oligomycin alone had no effect on EPSC amplitude. Third, TPP⁺ and RU360 prevented mitochondrial calcium uptake without depolarizing Δψ_m and depleting ATP (Aiuchi et al., 1985;Matlib et al., 1998). And forth, the effects of mitochondrial inhibitors were reversed by the inclusion of 1 mmEGTA in the intracellular solution, indicating mediation by an intracellular calcium buffering mechanism.

Recovery from synaptic depression

Previous reports have shown that increased presynaptic calcium buffering slows the recovery from synaptic depression (Dittman and Regehr, 1998; Stevens and Wesseling, 1998; Wang and Kaczmarek, 1998). This is contrary to our results where presynaptic terminals with lower calcium buffering (caused by removal of the mitochondrial calcium sequestration) show slower recovery from synaptic depression. The residual calcium hypothesis (Katz, 1969) suggests that neurotransmitter release would be enhanced rather that depressed by this additional free calcium. There is good evidence that [Ca²⁺]_cyto plays multiple roles in regulating the vesicle cycle. For instance, elevated [Ca²⁺]_cyto can inhibit endocytosis at the goldfish bipolar cell synapse (von Gersdorff and Matthews, 1994) and interact with regulators of dynamin to inhibit endocytosis in mammalian nerve terminals (Cousin and Robinson, 2000). Furthermore, the Rab GTPases are activated by Ca²⁺–calmodulin, causing inhibition of vesicle cycling (Coppola et al., 1999), although Ca²⁺–calmodulin has also been shown to promote refilling of the rapidly releasing pool of vesicles after high-frequency stimulation (Sakaba and Neher, 2001). The overall effect of [Ca²⁺]_cyto on transmitter release will therefore depend on the balance between these processes. In addition, the subcellular location of the [Ca²⁺]_cyto rise is critical in determining its action. The exogenous mobile calcium buffers applied in previous experiments (Dittman and Regehr, 1998;Stevens and Wesseling, 1998; Wang and Kaczmarek, 1998) will freely diffuse throughout the cytoplasm and influence [Ca²⁺]_cyto close to the site of calcium influx. Because mitochondria are located 100–500 nm from the release sites (Lysakowski et al., 1999; Rowland et al., 2000) and constitute 22% of the presynaptic volume (data from the endbulb of Held; M. Nicol and B. Walmsley, personal communication) (Nichol and Walmsley, 2002), they are ideally situated to sequester calcium from the bulk cytoplasm. Hence, they differentially influence the vesicle cycle compared with mobile exogenous buffers, and they could provide a crucial link between presynaptic metabolic activity and neurotransmitter release.

Functional relevance

Reduced calcium sequestration by mitochondria is associated with aging (Leslie et al., 1985) and neurodegenerative conditions such as amyotrophic lateral sclerosis, Huntington's disease, and Alzheimer's disease (Beal, 2000). In addition, symptoms similar to Parkinson's disease can result from inhibition of mitochondrial respiration by rotenone (Betarbet et al., 2000). Hence, the influence of mitochondrial calcium sequestration on synaptic transmission has important implications for cognitive function. Although there are some studies suggesting that calcium release from endoplasmic reticulum can influence synaptic transmission at other synapses (Llano et al., 2000;Emptage et al., 2001), our data suggest that mitochondria are the major intracellular calcium store in this mammalian glutamatergic terminal. Acting in concert with mobile calcium-binding proteins (Lee et al., 2000) and plasma membrane calcium pumps (Garcia and Strehler, 1999), mitochondria will shape cytoplasmic calcium transients and serve as an intermediary in calcium export. Our observations demonstrate that mitochondria also function in short-term presynaptic modulation on a millisecond time scale and with much lower presynaptic calcium loads than previously thought.