Motile Axonal Mitochondria Contribute to the Variability of Presynaptic Strength

Abstract

One of the most notable characteristics of synaptic transmission is the wide variation in synaptic strength in response to identical stimulation. In hippocampal neurons, approximately one-third of axonal mitochondria are highly motile and some dynamically pass through presynaptic boutons. This raises a fundamental question: Can motile mitochondria contribute to the pulse-to-pulse variability of presynaptic strength? Recently, we identified syntaphilin as an axonal mitochondrial docking protein. Using hippocampal neurons and slices of syntaphilin knockout mice, we demonstrate that the motility of axonal mitochondria correlates with presynaptic variability. Enhancing mitochondrial motility increases the pulse-to-pulse variability, while immobilizing mitochondria reduces the variability. By dual-color live imaging at single-bouton levels, we further show that motile mitochondria passing through boutons dynamically influence synaptic vesicle release, mainly by altering ATP homeostasis in axons. Thus, our study provides new insight into the fundamental properties of the CNS to ensure the plasticity and reliability of synaptic transmission.

INTRODUCTION

The effects of synaptic variability on neuronal or circuit activity are increasingly recognized. Some degree of variability may be necessary for signal processing in flexible or adaptive systems (Hessler et al., 1993; Murthy et al., 1997; Zador, 1998). A fundamental question is how the variation in synaptic strength arises. In the past decade, numerous studies focused on the structural and stochastic properties of molecular events underlying synaptic variability (see reviews by Atwood & Karunanithi, 2002; Stein et al., 2005; Marder & Goaillard, 2006; Branco & Staras, 2009; Ribrault et al., 2011). These events include opening voltage-dependent Ca²⁺ channels and spacing between the channels and the Ca²⁺-sensor; fusion modes and kinetics of synaptic vesicles (SVs); content and size of SVs; single versus multiple SV release; and number of receptors at synapses (Liu et al., 1999; Karunanithi et al., 2002; Conti & Lisman, 2003; Edwards, 2007; Young & Neher, 2009). Most of these factors are the basis for marked heterogeneity in synaptic transmission from neuron-to-neuron or from synapse-to-synapse. However, it is not known which dynamic process at axonal terminals contributes to pulse-to-pulse variability at single-bouton levels in response to identical synaptic stimulation.

Mitochondria maintain synaptic transmission by producing ATP and buffering Ca²⁺. Mitochondrial loss from synapses inhibits synaptic transmission due to insufficient ATP supply (Verstreken et al., 2005). Axonal mitochondria undergo dynamic and bidirectional transport and motile mitochondria can become stationary or pause at synapses and move again in response to physiological changes (see review by Sheng & Cai, 2012). In hippocampal neurons, approximately one-third of axonal mitochondria are motile (Kang et al., 2008). This raises a question: Can motile mitochondria influence SV release, thereby contributing to the pulse-to-pulse variability of presynaptic strength? Our recent study identified syntaphilin (SNPH) as a “static anchor” docking mitochondria (Kang et al., 2008). SNPH targets the mitochondrial outer-membrane, is sorted to axons via its axon-sorting sequence, and immobilizes axonal mitochondria via its microtubule (MT)-binding domain. SNPH-MT interaction is stabilized by dynein light chain LC8 (Chen et al., 2009). Deleting murine snph recruits the majority of axonal mitochondria into motile pools while over-expressing SNPH abolishes their motility. Thus, snph mice provide us with a unique genetic tool to address whether changes in axonal mitochondrial motility compromise the variability of presynaptic strength.

RESULTS AND DISCUSSION

Mitochondrial Motility Correlates with the Pulse-to-Pulse Variation of EPSC Amplitudes

By using cultured hippocampal neurons from snph^+/+ and snph^−/− mouse littermates, we first examined whether axonal mitochondrial motility influences mean amplitudes of excitatory postsynaptic currents (EPSCs) and their pulse-to-pulse variation. Representative and superimposed traces of 30 sweeps (Figures 1A and 1B) in snph^−/− neurons showed a larger fluctuation in EPSC amplitude under 0.05 Hz stimulation compared with snph^+/+ neurons. Elevated expression of EGFP-SNPH, but not EGFP-SNPHΔMT, a docking loss-of-function mutant, further reduced the variation found in wild-type (WT) neurons. Deleting snph robustly increased axonal mitochondrial motility (76 ± 20%, n = 17, p < 0.01) (Figure 1C) relative to WT neurons (38 ± 16%, n = 15). Conversely, over-expressing SNPH abolished axonal mitochondria motility (0.3 ± 1%, n = 19, p < 0.01). While axonal mitochondrial motility did not impact the mean EPSC amplitudes, it significantly influenced the variability of pulse-to-pulse EPSC amplitudes (Figure 1D). The coefficient of variation (CV) of EPSC amplitudes was larger in snph^−/− neurons (CV = 0.546 ± 0.011, p = 0.017) compared to snph^+/+ neurons (CV = 0.301 ± 0.012). In contrast, elevated SNPH expression further reduced the CV (0.141 ± 0.017, p = 0.028) found in WT neurons. As controls, expressing SNPHΔMT or GFP had no effect on mitochondrial motility (35 ± 9% and 35 ± 12%, p > 0.05, respectively) and on CV values (0.347 ± 0.011, p = 0.713; 0.326 ± 0.006, p = 0.941, respectively) relative to WT neurons. Pearson correlation analysis revealed a significant positive correlation (r = 0.988, p = 0.002) between the motility of axonal mitochondria and the CV values of pulse-to-pulse EPSC amplitudes.

Figure 1. Axonal Mitochondrial Motility Correlates with the Pulse-to-Pulse Variation of EPSC Amplitudes.

(A-C) Single trace (A) and superimposed traces of 30 sweeps (B) at 0.05-Hz stimulation and kymographs of axonal mitochondrial motility (C) from hippocampal neurons of snph^+/+ or snph^−/− mouse littermates, or WT neuron pairs, where presynaptic neurons expressing EGFP-SNPH, EGFP-SNPHΔMT, or GFP forming synapses with untransfected postsynaptic neurons. Mitochondria were labeled with DsRed-Mito (red). Neurons were transfected at DIV7-8; EPSCs were recorded 3 days after transfection. In kymographs, vertical lines represent stationary organelles; oblique lines or curves to the right indicate anterograde transport.

(D) Axonal mitochondrial motility influences the pulse-to-pulse synaptic variability. 30 traces from each neuron pair were averaged and data in the same group were pooled to calculate the average mean peak EPSC amplitude (left). The CV of pulse-to-pulse EPSC amplitudes was analyzed from 30 sweeps (right).

(E-G) Increasing mitochondrial motility enhanced synaptic fluctuations in acute hippocampal slices. Sample traces of the middle 30 EPSC events from 90 to 120 (E) and superimposed 200 EPSC events (F) and mean peak EPSC amplitude and relative CV values (G) from snph^+/+ or snph^−/− littermates (postnatal 3-5 weeks). Each recording was evoked by 0.05-Hz stimulation via the Schaffer collateral pathway. 200 sweeps form each slice were averaged and the data in the same group were pooled to calculate the mean peak EPSC amplitude.

Data were collected from the total number of neurons (D) or slices (G) indicated within bars. Data sets (D) were analyzed by the non-parametric Kruskal-Wallis test (p=0.029) comparing the five groups. The Dwass-Steel-Critchlow-Fligner post hoc analysis was applied for multi-group comparison. The p values on top of bars (D) are pair-wise comparisons to snph^+/+ neurons. Data sets were also log₁₀ transformed to normalize distribution, followed by one-way ANOVA analysis and the Dunnett test for multi-group CV comparison of (D, p=0.027). Data sets (G) were analyzed using the non-parametric the Mann-Whitney test for two groups. Error bars: SEM. Scale bars in C: 10 μm.

We next examined synaptic fluctuation in acute hippocampal slices from snph^+/+ and snph^−/− mouse littermates by recording 200 EPSC events evoked by 0.05 Hz stimulation in Schaffer collateral synapses. Enhancing axonal mitochondrial motility by deleting snph robustly increased EPSC fluctuation (Figures 1E and 1F). The pulse-to-pulse variability (Figure 1G) was significantly increased in snph^−/− mice (CV = 0.479 ± 0.098, p = 0.037) relative to WT littermates (CV = 0.246 ± 0.031). Due to its specific targeting to axonal but not dendritic mitochondria (Kang et al., 2008), SNPH influences pulse-to-pulse EPSC variability likely through a presynaptic mechanism. This was confirmed in cultured neuron pairs where only presynaptic neurons expressing SNPH were selected for recording (Figures 1A and 1B). The kinetics of EPSCs does not show statistical differences in the half-width, rise and decay times and slopes of EPSCs between snph^+/+ and snph^−/− hippocampal slices (see Extended Results in Supplemental Information), indicating that the variation is unlikely from asynchronous SV release. Thus, our results support that axonal mitochondrial motility impacts pulse-to-pulse variability of presynaptic strength.

Given the facts that (1) EPSCs were averaged through summation of the currents from multiple synapses between neuron pairs and (2) mitochondria dynamically pass by or pause at synapses, we assume that mitochondrial motility patterns on individual boutons are highly variable. To test this hypothesis, we further identified five motility patterns of axonal mitochondria at boutons during a 16-min recording time: stationary mitochondria sitting out of synapses (54.07 ± 2.53%) or within synapses (16.29 ± 1.66%); moving mitochondria passing through synapses (14.77 ± 1.58%) or pausing at synapses for a short (<200 sec, 7.01 ± 1.29%) or a longer time period (>200 sec, 8.30 ± 1.52%) (Figure S1 and Movie S1). Our findings are consistent with a previous study in cortical neurons (Chang et al., 2006). Thus, variable patterns of mitochondrial motility at individual boutons are likely one of the main sources contributing to the variability of EPSC amplitudes, which was further supported by genetic manipulation: either immobilizing axonal mitochondria by over-expressing SNPH or increasing their motility by deleting snph in neurons (Figures 1A-1D).

Mitochondrial Motility Influences SV Release at Single-Bouton Levels

Our previous study showed that increased mitochondrial motility in snph^−/− neurons has no impact on quantal size (Kang et al., 2008). To examine presynaptic variability at single-bouton levels, we examined two dynamic events by dual-color live imaging of axonal mitochondria labeled with DsRed-Mito and SV release using synapto-pHluorin (spH). All DsRed-Mito signals along axons were co-labeled by cytochrome c (see Extended Results in Supplemental Information). spH is a SV-targeted and pH-sensitive GFP whose fluorescence is quenched by intraluminal acidic pH, but increases upon exocytosis when it is exposed to the neutral pH of the extracellular medium. Thus, changes in spH fluorescence reflect the relative strength of SV release and recycling (Sankaranarayanan & Ryan, 2000). Because of technical limitation in resolving single action potential (AP)-induced spH responses, we instead used trains of stimulation (20 Hz, 10 sec) to provide a sufficient signal-to-noise ratio (Sun et al., 2010). We reasoned that if variability arises from moving mitochondria, single-bouton responses to each train of stimulation should display different spH ΔF_peak in correlation with mitochondrial distribution and motility.

We first determined whether mitochondrial motility impacts SV release during four successive train stimulation with 100-sec intervals by examining three patterns of mitochondrial distribution and motility: (1) boutons with and without a stationary mitochondrion (Figure 2A); (2) boutons with a mitochondrion moving out (Figure 2B); and (3) boutons with a mitochondrion passing by (Figure 2C) during four trains of stimulation. SV release in response to the first train was unaffected by the presence or absence of a presynaptic mitochondrion. SV release remained stable at boutons with a stationary mitochondrion (Figures 2A-2C), while SV release was depleted starting at the second train at boutons lacking a mitochondrion (Figure 2A). Interestingly, motile mitochondria significantly impact SV release, which was quickly reduced when a mitochondrion moved out of the boutons (Figure 2B). Conversely, a mitochondrion passing by boutons enhanced SV release (Figure 2C). Analysis of relative ΔF_peak values revealed a larger variation in SV release during repeated stimuli at boutons when a mitochondrion is moving out or passing by (Figures 2E and S2C). Consistently, mitochondrial motility also influenced SV release in response to a shorter stimulation train (10 Hz, 5 sec with 100-sec interval) (Figures S2A and S2B).

Figure 2. Mitochondrial Motility Influences SV Release at Single-Bouton Levels.

(A-D) Dual-color live imaging showing the distribution and motility of axonal mitochondria at boutons and corresponding spH traces (A-C) and normalized spH ΔF_peak (D) during four trains of stimulation (20-Hz at 10-sec with 100-sec interval). Note that SV exocytosis remained robust and stable at boutons with a stationary mitochondrion (red arrows/traces), while exocytosis diminished starting at the 2^nd train at mitochondrion-free boutons (green arrows/traces) or when a mitochondrion is moving out of the bouton at the 3^rd train (blue arrows/traces). A mitochondrion passing by bouton during the 3^rd train rescued SV release (purple arrows/traces). (E-G) Mitochondrial motility influences ΔF_peak variability. CV values of the ΔF_peak variation at each bouton during repeated stimulation were quantified and then averaged over all the trains from each group of boutons (E) with (red) or without (green) a stationary mitochondrion, or with a motile mitochondrion (blue). Representative spH traces are shown in Figure S2C. Kymographs (F) and average CV values (G) reflect the trial-to-trial ΔF_peak variation at each bouton in snph^+/+, snph^−/−, or in neurons over-expressing SNPH or Miro1 (Kruskal-Wallis test for comparing all four groups: p<0.0001) (Also see Movie S2-S4).

Data were collected from the number of neurons indicated in parentheses (D) or from the number of boutons indicated within bars (E, G) and analyzed by the non-parametric Kruskal-Wallis test for comparing multiple groups, followed by the Dwass-Steel-Critchlow-Fligner post hoc analysis for all pair-wise comparisons. The Mann-Whitney test was applied for pair-wise comparisons for two groups. Error bars: SEM. Scale bars in A-C: 1 μm; F: 10 μm.

We analyzed CV values of ΔF_peak variation at each bouton during repeated trains and then averaged CVs over all the trains from each group of boutons (Figure 2E, Figure S2C). In boutons with a motile mitochondrion, the average CV value (0.154 ± 0.021) was substantially higher relative to those from boutons with a stationary mitochondrion (0.061 ± 0.009, p < 0.001) or lacking a mitochondrion (0.081 ± 0.010, p < 0.001). Thus, motile mitochondria, when passing by terminals, contribute to larger variations in SV release. We further studied the variability in single-bouton levels with different total axonal mitochondria motility: (1) WT neurons: 38 ± 16%; (2) neurons with elevated SNPH expression: 0.3 ± 1%; (3) snph^−/− neurons: 76 ± 20%; and (4) neurons expressing Miro1: 79 ± 3%. Miro1 is a mitochondrial adaptor for linking kinesin motors (Figure 2F, Movies S2-S4). Enhanced motility of axonal mitochondria in snph^−/− neurons and in neurons expressing Miro1 led to higher CV values of ΔF during stimulation trains: snph^−/− peak (CV = 0.30 ± 0.04, p = 0.010) and Miro1 (CV = 0.33 ± 0.03, p = 0.007) relative to WT neurons (CV = 0.18 ± 0.03) (Figure 2G). In contrast, CV was reduced in neurons over-expressing SNPH (CV = 0.11 ± 0.04, p = 0.035). Pearson correlation analysis revealed significant positive correlation (r = 0.987, p = 0.015) between the motility of axonal mitochondria and the average CV values of ΔF_peak.

As the spH response reflects the balance of exo- and endocytosis, we applied 2 μM bafilomycin A1 (Baf), a reacidification blocker, after the third train to trap SVs in the neutral pH environment, hence ΔF_peak at the fourth train mainly reflects exocytosis. Compared to boutons with a stationary mitochondrion in the presence of Baf, ΔF_peak remained smaller at boutons lacking a mitochondrion, suggesting that the reduced spH response was mainly due to altered SV exocytosis (Figures 3A and 3B). We further compared the size of releasable SV pools at terminals with versus without a mitochondrion in the presence of Baf by eliciting four trains of stimulation. ΔF_peak for the first train were similar at both types of boutons. However, the difference was observed at the second train, and became larger at the third and fourth trains (Figures 3C and 3D), indicating a smaller size of total releasable SVs in terminals without a mitochondrion. Our results are consistent with a previous study showing that depleting mitochondria within axonal terminals results in a faster SV depletion following intensive stimulation (Verstreken, et al. 2005).

Figure 3. Impact of Presyanptic Mitochondria on the Size of Releasable SV Pools.

(A, B) Representative spH traces (A) and normalized ΔF_peak (B) before and after applying Baf (2 μM) at the fourth train of stimulation.

(C, D) Representative spH traces (C) and step increase of ΔF_peak in response to each train (D) when applying Baf (2 μM) during repeated stimulation.

Data were collected from 10-50 boutons from each neuron and analyzed by the Dwass-Steel-Critchlow-Fligner post hoc analysis for all pair-wise comparisons. The Mann-Whitney test was applied for pair-wise comparisons for two groups. Error bars: SEM.

ATP Homeostasis Is Critical to Maintain SV Release

Mitochondria maintain presynaptic homeostasis by supplying ATP and by buffering Ca²⁺ during intense, prolonged stimulation (Tang and Zucker 1997; Verstreken et al., 2005; Kang et al., 2008). ATP supports synaptic functions including SV mobilization and fusion, and generation of synaptic membrane potentials. To examine the role of axonal mitochondria in maintaining ATP homeostasis during repeated trains of stimulation, we applied an engineered fluorescent ATP sensor Perceval (Berg et al., 2009). The fluorescence intensity ratio (F488nm/F405nm) of Perceval reflects the relative ATP/ADP ratio, thus allowing temporal and spatial detection of physiological changes in cellular ATP levels in live neurons. At resting condition, axonal ATP/ADP ratio was similar with versus without a mitochondrion. However, when neurons were stimulated with repeated trains (20 Hz for 10 sec with 100-sec intervals), ATP/ADP ratio drops significantly, reflecting ATP consumption during neuronal firing and SV release. Within axonal terminal areas (2μm × 2 μm) containing a stationary mitochondrion, ATP/ADP ratio was recovered just before next train of stimulation (Figure 4A). In contrast, within mitochondrion-free areas, ATP/ADP ratio before the second train was slightly lower (p=0.059) but was substantially lower (p<0.007) before the third train, reflecting reduced ATP supplying after each train of stimulation. As intracellular pH values impact the Perceval fluorescent intensity (Berg et al., 2009), we monitored pH in the axonal terminals using the pH dye SNARF-5F. Axonal pH values undergo very minor changes (7.1-7.3) during the stimulation regardless of whether with or without a mitochondrion (Figure 4B). Thus, dynamic changes in the ATP/ADP ratios during trains of stimulation are physiological relevant to the strength of SV release (Figures 2A and 2D). ATP production by presynaptic mitochondria is critical for maintaining SV release. Our results are also consistent with a recent biochemical study in hippocampal synaptosomes, where inhibition of mitochondrial oxidative phosphorylation resulted in a substantial drop in ATP levels accompanied by reduced evoked SV release and unchanged cytosolic calcium levels (Ivannikov et al., 2013). Thus, axonal mitochondrial distribution and motility contribute to the variability of SV release likely by influencing presynaptic ATP homeostasis.

Figure 4. ATP Homeostasis Is Critical to Maintain SV Release.

(A) The cellular ATP/ADP ratio during trains of stimulation. The curve of fluorescence intensity (F488nm/F405nm) (upper panels) reflects the relative ATP/ADP ratio. Normalized F488nm/F405nm ratio (lower panels) was recorded before each train.

(B) Measurement of axonal pH using SNARF-5F during trains of stimulation. 20-50 areas with or without a mitochondrion along each axon were selected for imaging. SNARF-5F was excited at 488 nm and detected around 580 nm and 640 nm during 4 trains of stimulation, as indicated by red bars (upper left). At the end of stimulation, SNARF-5F signal was calibrated using various buffered solutions containing high K/nigericin with varying pH values (blue bar, upper right). Calibrated pH values (lower panel) were averaged from 50 images.

(C) Representative spH traces (upper) and normalized spH ΔF_peak (lower) showing reduced SV release at boutons with a stationary mitochondrion under treatment with 4μg/ml oligomycin for 15 min. Note that oligomycin reduced exocytosis at the 3^rd stimulus at boutons with a mitochondrion (red), a phenotype similar to the bouton without a mitochondrion (green).

Data were collected from 11 neurons (A) or 5 neurons (C). 10-50 boutons or axonal areas were imaged for each neuron. Data sets were analyzed by the non-parametric Kruskal-Wallis test for comparing multiple groups. The Mann-Whitney test was applied for pair-wise comparisons for two groups. Error bars: SEM.

To determine whether the ATP production by mitochondria is critical to maintain SV release, we treated neurons for 15 min with 4μg/ml oligomycin, an inhibitor of mitochondrial ATP generation. The spH response to the first train was not affected by oligomycin (Figure 4C). In boutons with a mitochondrion, there is no significant change (p=0.67) in the average ΔF_peak upon the first train between control (1.37 ± 0.04) and oligomycin-treated groups (1.39 ± 0.10). Consistently, in boutons without a mitochondrion, no ΔF_peak change (p=0.95) was observed between control (1.39 ± 0.01) and oligomycin-treated groups (1.38 ± 0.07). Given the fact that oligomycin only blocks new ATP production, after pre-existing ATP was largely depleted beginning at the third train (Figure 4A), ΔF_peak was reduced at boutons when mitochondrial ATP production was blocked (Figure 4C), a phenotype mimicking boutons lacking mitochondria. These results highlight a critical role of mitochondria in maintaining synaptic transmission by supplying ATP.

Reduced spH response in single-bouton levels may represent a different phenomenon as EPSC variability recorded from multiple synapses between paired neurons. Approximately 30% of mitochondria dynamically move along axons (Figure S1 and Movies S1 and S2), EPSC variability during low stimulation frequency (Figure 1) is likely contributed by summation of many small changes in ATP levels at each synapse during a short time scale. Due to insufficient spH signal-to-noise ratio in response to single AP, we are unable to estimate release probability with the imaging approach.

Synaptic mitochondria efficiently buffer presynaptic [Ca²⁺]_i during intensive synaptic activity (Tang and Zucker, 1997). To determine whether mitochondria have any impact on the basal and evoked [Ca²⁺]_i under our stimulation profile, we applied two calcium indicators. First, we applied the YFP/CFP ratio-metric calcium indicator YC3.60 to monitor the basal and evoked [Ca²⁺]_i transients during the four trains of stimulation (20 Hz, 10 sec with 100-sec interval). The YFP/CFP ratio curves reflect the relative [Ca²⁺]_i levels (Figure S3A and S3B). The basal [Ca²⁺]_i is defined by averaging YFP/CFP ratios during a 10-sec period just before each train (Figure S3C), while the evoked [Ca²⁺]_i is reflected by total YFP/CFP value during each train of stimulation (Figure S3D). The YC3.60 ratio-metric imaging shows no significant change in the basal (p=0.413) and evoked [Ca²⁺]_i (p=0.476) with or without a mitochondrion. Alternatively, we confirmed the evoked [Ca²⁺]_i transient by using the single-color genetic variant of GCaMP3 calcium indicator GECO (Genetically Encoded Calcium indicator for Optical imaging). The change in GECO fluorescence intensity over baseline (ΔF/F₀) showed fast rising upon stimulation (Figures S3E and S3F). Normalized peak values of ΔF/F₀ over baseline (p=0.07, Figure S3G) and the total increase in normalized GECO intensity during each train of stimulation (p=0.08, Figure S3H) show no significant difference with or without a mitochondrion. Altogether, our studies consistently show that axonal mitochondria have no detectable impact on the basal and evoked [Ca²⁺]_i under our simulation profile.

We next asked whether increased Ca²⁺ influx or altered Ca²⁺ buffering contributes to the variability of spH responses. Under 10 mM [Ca²⁺]_ex, the first ΔF_peak value (1.71 ± 0.02, p<0.001) (Figure S4A) is significantly larger than that under 2 mM [Ca²⁺]_ex (1.40 ± 0.03) (Figure 2). However, mitochondria are still required to maintain SV release during four trains of stimulation (Figure S4A). Conversely, incubation with 100 μM EGTA-AM for 3 min has no detectable impact on SV release at boutons with a stationary mitochondrion or on reduced SV release at boutons without a mitochondrion (Figures S4B and S4C), although the first ΔF_peak value under EGTA-AM (1.25 ± 0.02, p<0.001) is significantly smaller than that under 2 mM [Ca²⁺]_ex (1.40 ± 0.03) (Figure 2). Therefore, altered levels of ATP, but not [Ca²⁺]_i, is likely one of major sources contributing to the presynaptic variability in response to identical stimulation.

In the current study, we simultaneously imaged both axonal mitochondria and SV release from single boutons during trains of stimulation. These procedures allowed us to measure the variation of SV release under different patterns of mitochondrial distribution and motility (see Extended Discussion in Supplemental Information). It is conceivable that the absence of a stationary mitochondrion within an axonal terminal reduces local ATP supply; and that a motile mitochondrion passing through a bouton temporally and spatially influences ATP homeostasis, thus impairing ATP-dependent processes at synaptic terminals including SV pool replenishment (Heidelberger et al., 2002). Although presynaptic mitochondria are crucial to maintain the proper size of total releasable SV pools during trains of stimulation, it is possible that other ATP-dependent processes may collectively contribute to presynaptic variability when mitochondria travel along axons and move into or pass by boutons. Thus, our study revealed, for the first time, that the dynamic movement of axonal mitochondria is one of the primary mechanisms underlying the pulse-to-pulse variability or the trial-to-trial variation of presynaptic strength in the CNS.

EXPERIMENTAL PROCEDURES

(Extended Experimental Procedures can be found in Supplemental Information)

Dual patch-clamp EPSC recordings in cultured hippocampal neurons were made with a whole-cell configuration. After identifying paired neurons, action potentials were detected under current-clamp by injecting 400 pA of current (depolarization); only those neurons with resting membrane potentials around -60 mV were selected for paired recording. The hippocampal slice in the recording chamber was superfused with ACSF (1 ml/min) supplemented with picrotoxin (50 μM) and saturated with 95% O₂/5% CO₂. After obtaining stable currents, 200 sweeps were recorded at a holding potential of -70 mV. 50 μM picrotoxin and 100 μM cyclothiazide were applied in the bath solution to block GABA receptor currents and avoid desensitization of postsynaptic receptors during recording.

Dual-color time-lapse imaging was performed to monitor mitochondrial movement and SV release at presynaptic boutons. Neurons at DIV7-9 were co-transfected with spH and DsRed-Mito using calcium phosphate method, followed by imaging at DIV12-14. 10-50 boutons were imaged for each neuron examined. The engineered fluorescent sensor Perceval was used to monitor the ATP/ADP ratio in live neurons. Perceval was co-transfected with DsRed-Mito and SNPH into neurons at DIV7-9. Axonal areas (2μm × 2 μm) containing or lacking a mitochondrion were selected for imaging. Perceval was excited at 405 nm and 488 nm and detected in the 505–550 nm range. The fluorescence intensities (F488nm/F405nm) reflect the relative cellular ATP/ADP ratio. Intracellular pH values were also monitored with the pH indicator dye SNARF-5F.

Highlights.

• Axonal mitochondrial motility correlates with the pulse-to-pulse EPSC variability

• Motile mitochondria passing by boutons contribute to variation of SV release

• Presynaptic mitochondria maintain SV release during sustained synaptic activity

• Altered ATP homeostasis is one of the primary sources for presynaptic variability

Abbreviation

AP

action potential

Baf

bafilomycin A1

CV

coefficient of variation

DIV

days in vitro

EPSC

excitatory postsynaptic current

GECO

genetically encoded calcium indicator for optical imaging

Mito

mitochondria

MT

microtubule

SNPH

syntaphilin

spH

synapto-pHluorin

SV

synaptic vesicle

Syn

synapse

WT

wild-type