Mitochondrial Free Ca²⁺ Levels and Their Effects on Energy Metabolism in Drosophila Motor Nerve Terminals

Abstract

Mitochondrial Ca²⁺ uptake exerts dual effects on mitochondria. Ca²⁺ accumulation in the mitochondrial matrix dissipates membrane potential (_ΔΨ_m), but Ca²⁺ binding of the intramitochondrial enzymes accelerates oxidative phosphorylation, leading to mitochondrial hyperpolarization. The levels of matrix free Ca²⁺ ([Ca²⁺]_m) that trigger these metabolic responses in mitochondria in nerve terminals have not been determined. Here, we estimated [Ca²⁺]_m in motor neuron terminals of Drosophila larvae using two methods: the relative responses of two chemical Ca²⁺ indicators with a 20-fold difference in Ca²⁺ affinity (rhod-FF and rhod-5N), and the response of a low-affinity, genetically encoded ratiometric Ca²⁺ indicator (D4cpv) calibrated against known Ca²⁺ levels. Matrix pH (pH_m) and _ΔΨ_m were monitored using ratiometric pericam and tetramethylrhodamine ethyl ester probe, respectively, to determine when mitochondrial energy metabolism was elevated. At rest, [Ca²⁺]_m was 0.22 ± 0.04 μM, but it rose to ∼26 μM (24.3 ± 3.4 μM with rhod-FF/rhod-5N and 27.0 ± 2.6 μM with D4cpv) when the axon fired close to its endogenous frequency for only 2 s. This elevation in [Ca²⁺]_m coincided with a rapid elevation in pH_m and was followed by an after-stimulus _ΔΨ_m hyperpolarization. However, pH_m decreased and no _ΔΨ_m hyperpolarization was observed in response to lower levels of [Ca²⁺]_m, up to 13.1 μM. These data indicate that surprisingly high levels of [Ca²⁺]_m are required to stimulate presynaptic mitochondrial energy metabolism.

Introduction

Mitochondria provide most of the ATP that fuels a variety of cellular activities. Changes in cellular activities result in an increase in ATP demand and are usually associated with elevations in cytosolic ADP and calcium concentrations ([Ca²⁺]_i) (1,2), both of which are believed to modulate mitochondrial ATP synthesis. These regulatory mechanisms appear to be preserved in presynaptic nerve terminals, where the synchronization of ATP utilization with mitochondrial ATP production is crucial for sustaining synaptic transmission (3,4). An ADP stimulatory influence on energy metabolism has been well documented in isolated mitochondria and a number of neuronal preparations (1,2,5). The widely used ADP/ATP ratio reflects the energetic status of the cell and determines the amount of ADP available for phosphorylation by the F₁-F₀-ATP synthase (5). However, simulations of mitochondrial metabolism show that changes in the cytosolic ADP/ATP ratio and phosphate levels alone are not sufficient to explain mitochondrial metabolism stimulation (6). The calcium sensitivity of elements of mitochondrial energy metabolism provides an additional mechanism for stimulating ATP synthesis. An elevation in mitochondrial matrix Ca²⁺ levels ([Ca²⁺]_m) increases mitochondrial metabolism in motor nerve terminals in situ (7), although the [Ca²⁺]_m levels responsible for this stimulatory effect have not been quantified.

Quantification of [Ca²⁺]_m levels that stimulate presynaptic mitochondrial energy metabolism in situ necessitates estimation of [Ca²⁺]_m along with several independent measures of mitochondrial energy metabolism, such as matrix pH (pH_m) and mitochondrial membrane potential (_ΔΨ_m). The accuracy of any physiologically relevant [Ca²⁺]_m estimate measured with Ca²⁺ indicators depends on the indicator’s affinity, specificity of targeting to the matrix, environmental sensitivity, and accuracy of the calibration in situ, as well as the ability to replicate in vivo [Ca²⁺]_i transients. Either chemical or genetically encoded Ca²⁺ indicators (GECIs) can be used to measure [Ca²⁺]_m. GECIs offer subcellular specificity of targeting, but they are vulnerable to pH changes, and with the exception of aequorin, they have slow kinetics, nonlinear responses to Ca²⁺, and a low dynamic range (8). Most chemical Ca²⁺-indicators perform well in the areas in which GECIs are deficient, but they can be difficult to load with specificity, and few ratiometric chemical Ca²⁺ indicators are available with a Ca²⁺ affinity suitably low for measuring [Ca²⁺]_m. Nonratiometric imaging with dyes is particularly problematic, as their calibration generally requires permeabilization of the inner mitochondrial membrane to control [Ca²⁺]_m, which inevitably leads to dye loss (9).

Exploiting the advantages of chemical Ca²⁺ indicators and GECIs in the same preparation limits negative influences imposed by their respective disadvantages, providing a greater degree of confidence in [Ca²⁺]_m estimates. The Drosophila larval neuromuscular preparation is amenable to the application of chemical Ca²⁺ indicators, and it is genetically tractable, allowing the expression of GECIs. Further, the endogenous firing rates and accompanying changes in [Ca²⁺]_i have been quantified (7,10). Here, we use two different chemical Ca²⁺ indicators (rhod-5N, K_d ∼ 320 μM, and rhod-FF, K_d ∼ 19 μM) and adopt an analytical approach to calculate [Ca²⁺]_m from the ratio of their responses under similar conditions with no requirement for permeabilization. Two ratiometric GECIs, TN-XXL and D4cpv (K_d ∼ 0.8 μM and K_d ∼ 60 μM, respectively, in vitro) were used to estimate minimum and maximum [Ca²⁺]_m, respectively. A genetically encoded pH indicator (GEpHI; ratiometric pericam) and a mitochondrial potentiometric probe, tetramethylrhodamine ethyl ester (TMRE), were used to report elevation of mitochondrial energy metabolism. Both chemical and GECIs revealed that a high level of [Ca²⁺]_m (∼26 μM) was required to stimulate presynaptic mitochondrial energy metabolism.

Methods

Fly stocks

Flies were maintained at room temperature on cornmeal media supplemented with baker’s yeast. Third-instar Drosophila larvae of both sexes were used for all of the experiments. The enhancer-trap strain w¹¹¹⁸; P[w⁺, OK6:GAL4] was used to drive expression of upstream-activating-sequence (UAS)-controlled reporter transgenes in motor neurons. The other fly stocks/transgenes used in this study included UAS-GFP, UAS-2mt8RP, UAS-GCaMP1.3, UAS-2mt8D4cpv, and UAS-2mt8TN-XXL.

Generation of flies transgenic for mitochondrially targeted Ca²⁺ indicators

cDNA for D4cpv was provided by Roger Tsien, cDNA for TN-XXL was provided by Dirk Reiff (Max Planck Institute for Biological Cybernetics, Tübingen, Germany) and cDNA for superecliptic pHluorin was provided by Gero Miesenbock (Centre for Neural Circuits and Behaviour, Oxford University, Oxford, United Kingdom). GECIs were targeted to the matrix of mitochondria by fusing a tandem repeat of the first 36 aa of subunit VIII of human cytochrome oxidase (COX) (11) to the N-terminus of each GECI. The fused cDNAs were then cloned into a P-element vector (pUAST) and injected into w¹¹¹⁸ Drosophila embryos by Rainbow Transgenic Flies (Newbury Park, CA). Strains containing homozygous transgenes were obtained after determining the chromosome harboring the transgene and outcrossing all other chromosomes.

Dye loading and fluorescence imaging

Drosophila larvae were filleted and pinned to a Sylgard bath in Schneider’s solution (Sigma Aldrich, Saint Louis, MO). The segmental nerves were cut from the ventral ganglion before dye loading. Rhod-FF and rhod-5N stocks (2 mM in each case) in Pluronic/dimethylsulfoxide solution (Invitrogen, Carlsbad, CA) were diluted in Schneider’s solution to a final concentration of 1 μM and added to larval preparations. After 20 min of incubation at room temperature, larvae were washed with fresh Schneider’s solution for 15 min. For imaging, Schneider’s solution was replaced with hemolymph like solution 6 (HL6) (12) containing 2 mM CaCl₂ and 7 mM L-glutamic acid and the nerves were drawn into a stimulation pipette. In the case of TMRE, larval preparations with severed nerves were incubated in Schneider’s solution containing 50 nM TMRE at room temperature for 20 min. The solution was then replaced with HL6 supplemented with 50 nM TMRE and the nerves were drawn into a stimulation pipette. TMRE-loaded larval preparations were stimulated at various frequencies with the dye present in the bath to minimize TMRE leakage. Dye and GECI fluorescence changes were imaged in MN13-Ib axonal terminals in abdominal segment 4, as described previously in Chouhan et al. (7). Briefly, nerve terminals were visualized using wide-field microscopy on a BX51WI microscope (Olympus, Center Valley, PA) equipped with a 100× (1.0 NA) water-immersion objective. Nerve stimulation at various frequencies was delivered via Master-8 stimulator (AMPI, Jerusalem, Israel) and a Digitimer (Brooksville, FL) model DS2A Mk.II. Images were captured with an Andor (Belfast, United Kingdom) iXon+ DU-860D EMCCD camera using Andor iQ 1.8 acquisition software and saved as 14-bit (128 × 128 pixels) TIF files. Image processing was performed with either Andor iQ or ImageJ software package. For quantification of _ΔΨ_m hyperpolarization, background-subtracted TMRE fluorescence was averaged over 2 s (∼6 frames) at 15 s after onset of a stimulus.

Indicator calibrations

Dissected larval preparations expressing mito-D4cpv or mito-TN-XXL were permeabilized for 20 min with HL6 solutions containing 50 μM ionomycin, 10 μM carbonyl cyanide m-chlorophenyl hydrazone (CCCP) and various known calcium concentrations and imaged by collecting fluorescence from mitochondria in various nerve terminals. Calibration solutions with free calcium concentrations <5 μM were prepared by mixing various volumes of two HL6 solutions: one containing 5 mM EGTA and the other 1 mM CaCl₂, and adjusting pH to 7.3 at 24°C. Free calcium concentration was calculated using Ca-EGTA Calculator v1.2, http://www.stanford.edu/∼cpatton/CaEGTA-NIST.htm. Calibration solutions with free calcium concentrations >5 μM were made by adding various volumes of 1 M CaCl₂ to HL6 solution. EYFP/ECFP fluorescence ratios were plotted against free calcium concentration and fitted with a four-parameter sigmoid equation in SigmaPlot 10 (Systat Software, Evanston, IL).

Mitochondrially-targeted-ratiometric-pericam (mito-RP)-expressing preparations were permeabilized for 20 min with internal solution (120 mM KCl, 0.5 mM KH₂PO₄, 10 mM succinate, 10 mM Bis-Tris propane) containing 100 μM digitonin and 10 μM CCCP adjusted to the various pH values. Fluorescence values (excitation at 490 nm) were normalized to the resting (nonpermeabilized) fluorescence and plotted against pH. The values were fitted to a Boltzmann function using SigmaPlot 10.

Mitochondrial calcium concentration quantification with fluorescent dyes

Inability to accurately measure F_max and F_min values for nonratiometric fluorescent dyes in mitochondria precludes their use in mitochondrial calcium concentration ([Ca²⁺]_m) quantification. However, [Ca²⁺]_m can be estimated by comparing relative fluorescence responses of two fluorescent dyes with distinct K_d values without the need for measuring F_max and F_min values. The theory for this approach is presented in the Appendix.

The range of calcium concentrations that can be reliably estimated by this method depends on the performance of the detection system used, the loading conditions, and the K_d values for fluorescent indicators. The standard deviation of fluorescent noise (deflections of the signal from the mean, measured from a region of interest) for rhod indicators under loading conditions described above is <0.7% of F_max:

$$\sigma =\frac{{\sigma }_{\text{noise}}}{{\text{F}}_{\text{max}}}\le 0.7\text{\%}.$$

Thus, fluorescent values that are different by at least three standard deviations for fluorescent noise (3σ_noise) are resolved with probability >99.7%. Rearranging Eq. 2 from the Appendix gives an estimate for the range of calcium concentrations measured by each indicator:

$$\left[{\left[{\text{Ca}}^{2+}\right]}_{\min };{\left[{\text{Ca}}^{2+}\right]}_{\max }\right]=\left\{\frac{3\sigma }{1-3\sigma }\times {\mathrm{K}}_{\mathrm{d}};\frac{1-3\sigma }{3\sigma }\times {\mathrm{K}}_{\mathrm{d}}\right\}\approx \left\{3\sigma \times {\mathrm{K}}_{\mathrm{d}};\frac{1}{3\sigma }\times {\mathrm{K}}_{\mathrm{d}}\right\}.$$

Assuming that rhod-FF and rhod-5N intramitochondrial K_d values are 1.5-fold higher than their in vitro estimates, the minimum and maximum [Ca²⁺]_m that can be detected using these two indicators together are 10 μM and 1350 μM, respectively.

The mean ± SE for [Ca²⁺]_m values was calculated using error propagation theory. Partial derivatives of [Ca²⁺] with respect to R^FF and R^5N for Eqs. 4 and 5 in the Appendix were used to calculate the SE,

$$\mathrm{S}{\mathrm{E}}_{\left[\mathrm{C}{\mathrm{a}}^{2+}\right]}^2={\left({\frac{\partial \left[\mathrm{C}{\mathrm{a}}^{2+}\right]}{\partial {\mathrm{R}}^{\text{FF}}}|}_{\overline{{\text{R}}^{\text{FF}}},\overline{{\text{R}}^{5\text{N}}}}\right)}^2\times \mathrm{S}{\mathrm{E}}_{R^{FF}}^2+{\left({\frac{\partial \left[\mathrm{C}{\mathrm{a}}^{2+}\right]}{\partial {\mathrm{R}}^{5\mathrm{N}}}|}_{\overline{R^{\text{FF}}},\overline{{\text{R}}^{5\text{N}}}}\right)}^2\times \mathrm{S}{\mathrm{E}}_{{\text{R}}^{5\text{N}}}^2,$$

where $\overline{{\text{R}}^{\text{FF}}}$ and $\overline{{\text{R}}^{5\text{N}}}$ are the mean values of R^FF and R^5N.

Results

Cytosolic Ca²⁺ is actively taken up by mitochondria where it stimulates mitochondrial energy metabolism. To estimate [Ca²⁺]_m levels in presynaptic mitochondria at various [Ca²⁺]_i levels, we either loaded fluorescent chemical Ca²⁺ indicators or expressed ratiometric GECIs in mitochondria of motor neurons (MNs) in Drosophila larvae. Mitochondrial Ca²⁺ uptake was examined in the presynaptic mitochondria of the MN that innervates body wall muscle fiber 13 (MN13-Ib) with big boutons (Fig. 1 A). Nerve stimulation at 80 Hz for 2 s resulted in a rapid increase in [Ca²⁺]_i, as reported by GCaMP3 fluorescence (Fig. 1 B, upper). The rise in [Ca²⁺]_i was accompanied by rapid mitochondrial Ca²⁺ uptake and matrix pH (pH_m) alkalinization followed by a slow return of both [Ca²⁺]_m and pH_m to baseline. Removal of Ca²⁺ from the extracellular medium abolished changes in pH_m (Fig. 1 B, middle). Inhibition of the mitochondrial ADP/ATP exchanger with 50 μM bongkrek acid had no effect on pH_m (Fig. 1 B, lower), suggesting that mitochondrial Ca²⁺ acts as the principal driver of mitochondrial alkalinization.

Figure 1.

Imaging of mitochondrial pH and Ca²⁺ changes in Drosophila larval motor neuron terminals. (A) Images of a motor neuron forming a terminal with type-Ib big boutons on the surface of muscle fiber 13. Top to bottom: phase-contrast image; boutons containing cytosolic (cyto.) GFP; mitochondria (mito.) loaded with rhod-FF; merge of cyto. and mito. (B) Changes in cytosolic Ca²⁺, [Ca²⁺]_i (blue line; F/Fo, GCaMP3), mitochondrial matrix Ca²⁺, [Ca²⁺]_m (black traces; (2Fo − F)/Fo, mito-RP excitation at 420 nm), and pH_m (red traces, F/Fo, mito-RP excitation at 490 nm) obtained upon 80 Hz nerve stimulation for 2 s at 10 s in normal (upper), Ca²⁺-free (middle), and 50 μM bongkrek acid (BA) containing HL6 solution (lower).

We further explored the relationship between pH_m, mitochondrial membrane potential (_ΔΨ_m), and [Ca²⁺]_m at different stimulation frequencies. A combination of two mitochondrial fluorescent chemical Ca²⁺ indicators with different K_d values, such as rhod-FF (K_d = 19μM) and rhod-5N (K_d = 320μM), allows estimation of [Ca²⁺]_m levels without necessitating indicator calibrations after each experiment. Although rhod-FF and rhod-5N cannot be used in the same preparation for the real-time [Ca²⁺]_m measurements, due to the overlap in their spectral properties, different Ca²⁺ indicator combinations are possible, such as mag-fluo-4 and rhod-5N. However, from our experience, adequate loading of both indicators at the same time is difficult and often leads to inconsistent results. We therefore used individual preparations loaded with either rhod-5N or rhod-FF that were then stimulated for 2 s by a pair of different frequencies (stimulating frequency followed by a reference frequency) separated by an 8-s interval. Stimulations for 5 s did not produce a larger increase in the fluorescence of either dye, indicating that equilibrium for [Ca²⁺]_m is reached quickly (data not shown). Fig. 2 A shows changes in rhod-FF and rhod-5N fluorescence at 42 and 80 Hz. Background-corrected intensities, calculated as the average fluorescence signal 2–3 s after the cessation of each stimulus train, were used to calculate the ratios of response amplitudes (stimulating frequency/reference frequency) at 30, 42, 60, and 80 Hz for both Ca²⁺ indicators. Rhod-FF ratios of 30/42 Hz, 0.66 ± 0.03 (n = 6); 42/60 Hz, 0.83 ± 0.02 (n = 9); and 42/80 Hz, 0.76 ± 0.03 (n = 10) were higher than the corresponding ratios for rhod-5N: 30/42 Hz, 0.52 ± 0.13 (n = 6); 42/60 Hz, 0.70 ± 0.05 (n = 11); and 42/80 Hz, 0.57 ± 0.03 (n = 7), which is consistent with rhod-FF having a higher affinity for Ca²⁺. The 60/80-Hz ratios could not be reliably measured, since stimulation at 60 Hz led to an insignificant calcium uptake at 80 Hz, suggesting Ca²⁺-dependent inhibition of Ca²⁺ uptake at higher [Ca²⁺]_m. Thus, calculation of [Ca²⁺]_m values for 60-Hz and 80-Hz stimuli from the ratios with 60 Hz and 80 Hz only as reference frequencies might result in their underestimation. The determined ratios and the published in vitro K_d values for these Ca²⁺-indicators were used to derive absolute values for [Ca²⁺]_m. The detailed procedure for the calculation of [Ca²⁺]_m is fully described in the Appendix to this article. The calculated [Ca²⁺]_m values ranged from 8.4 μM to 29.5 μM for the frequencies used. However, these [Ca²⁺]_m estimates are likely to be significantly underestimated due to the differences in K_d values for free and cell/organelle-internalized Ca²⁺-indicators (13). The shaded columns in Fig. 2 B show the effects of varying rhod-FF and rhod-5N affinities between 1 × K_d and 1.5 × K_d (assuming F_min ∼ 0) on solution spaces for [Ca²⁺]_m at 42- and 80-Hz stimulation. It has been estimated that the intramitochondrial rhod-5N K_d value is at least 1.5 times higher (∼470 μM) than its in vitro K_d (14). To account for this, we calculated [Ca²⁺]_m using higher (1.5×) estimates for Ca²⁺ affinities of the Ca²⁺ indicators (see Table 1).

Figure 2.

Quantifying mitochondrial matrix Ca²⁺ concentrations with rhod-FF and rhod-5N indicators. (A) Fluorescent responses of rhod-FF (black) and rhod-5N (red) normalized to 80 Hz recorded from presynaptic mitochondria in MN13-Ib axonal terminals stimulated with 42- and 80-Hz electrical pulse trains for 2 s at 10 and 20 s. Dotted lines show the 42/80-Hz ratio values for both indicators. (B) Graphical simulation of rhod-FF and rhod-5N fluorescence (F/F_max, F_min = 0) at different Ca²⁺ concentrations ([Ca²⁺]) for a range of K_d values for both indicators (thick, 1 × K_d, to thin, 1.5 × K_d). Dotted lines show the rhod-FF and rhod-5N fluorescence intensities that give 42/80-Hz ratios similar to those in A and also result in single numerical solutions for [Ca²⁺] for rhod-FF and rhod-5N at 42 and 80 Hz. Dotted areas are the solution spaces for [Ca²⁺] at 42 and 80 Hz for various K_d.

Table 1.

Equilibrium values of [Ca²⁺]_m at rest and during nerve stimulation

	Rhod-FF/Rhod-5N	Mito-D4cpv	Mito-TN-XXL
Rest	—	—	0.22 ± 0.04
30 Hz	13.1 ± 3.9	—	—
42 Hz	24.3 ± 3.4	27.0 ± 2.6	—
60 Hz	35.4 ± 10.0	—	—
80 Hz	44.3 ± 6.4	62.9 ± 4.9	—

Nerves were stimulated at various frequencies, and equilibrium values were derived from measurements using Ca²⁺ indicators rhod-FF/5N, D4cpv, and TN-XXL. Values are shown as the mean ± SE (μM).

We sought independent estimates for [Ca²⁺]_m values by using ratiometric GECIs targeted to the mitochondrial matrix, although they have a more limited dynamic range than fluorescent chemical Ca²⁺ indicators. D4cpv is a low-affinity fluorescence-resonance-energy-transfer-based (enhanced yellow fluorescence protein (EYFP) (cpVenus)/enhanced cyan fluorescent protein (ECFP) pair) GECI wherein the calmodulin/peptide pair was redesigned to change its Ca²⁺ affinity and also to prevent perturbation by wild-type calmodulin (15). We expressed mitochondrially targeted D4cpv (mito-D4cpv) in Drosophila larval MNs. Larval preparations expressing mito-D4cpv were permeabilized with solutions containing 50 μM ionomycin, 10 μM CCCP, and various known Ca²⁺ concentrations to calibrate the GECI. Ratios of EYFP/ECFP fluorescence intensities were plotted against log[Ca²⁺], and fitted with a four-parameter Hill’s model (Fig. 3 A). The apparent Ca²⁺ affinity for mito-D4cpv (K_{d, app} = 90.3 μM) was higher than the reported affinity of cytosolic D4cpv (K_d, app = 64 μM) (16). The other parameters derived from the model were Hill’s coefficient, nH = 1.24, and R_min and R_max of 2.48 and 7.34, respectively.

Figure 3.

Measuring mitochondrial matrix Ca²⁺ concentrations with the genetically encoded Ca²⁺ indicator mito-D4cpv. (A) Calibration curve (four-parameter Hill’s fit, R² = 0.95) showing EYEP(cpVenus)/ECFP ratio of mito-D4cpv as a function of the logarithm of Ca²⁺ concentration (μM). (B) Changes in EYFP (red line) and ECFP fluorescence (black line) and their ratio (dotted blue line) in mitochondria of M13-1b axonal terminals stimulated with 42- and 80-Hz electrical pulse trains for 2 s at 10 and 20 s.

Mito-D4cpv was imaged in MN13-Ib axonal terminals to estimate [Ca²⁺]_m values at 42 and 80 Hz. Fig. 3 B shows typical fluorescent responses seen in the EYFP and ECFP channels, and their ratio. Stimulation at 42 and 80 Hz for 2 s increased the EYFP/ECFP ratio to 3.37 ± 0.09 (n = 9) and 4.38 ± 0.11 (n = 8), respectively. Fitting those values into the Hill’s equation gives [Ca²⁺]_m values of 27.0 ± 2.6 μM and 62.9 ± 4.9 μM for 42 and 80 Hz, respectively. Both mito-D4cpv and fluorescent chemical Ca²⁺ indicators (1.5 × K_d) produced comparable estimates for [Ca²⁺]_m at 42 Hz, but mito-D4cpv gave a higher value at 80 Hz, see Table 1. This supports our earlier assumption and previous estimates that K_d for rhod-5N/FF in mitochondria would be at least 1.5 times higher than the values suggested from in vitro measurements.

[Ca²⁺]_m at rest was also estimated with the high-affinity GECI TN-XXL (8) (K_{d, app} ∼ 1.1 μM; Fig. S1 A in the Supporting Material), which we targeted to the mitochondrial matrix (mito-TN-XXL). In resting [Ca²⁺]_m measurements, the nerves were severed at least 30 min before imaging. The resting [Ca²⁺]_m in MN13-Ib presynaptic mitochondria was 220 ± 43 nM (n = 8; Table S1), which is higher than our estimate of resting [Ca²⁺]_i (32 ± 6 nM) in these neurons. Comparable resting [Ca²⁺]_m values have been reported from other cells (15,17,18). Plotting resting and stimulated [Ca²⁺]_m values against our [Ca²⁺]_i estimates derived from fura-dextran measurements yields a linear relationship (Fig. 4 A), which suggests that mitochondria do not start taking up Ca²⁺ until [Ca²⁺]_i exceeds 200–250 nM.

Figure 4.

Interrelation between mitochondrial matrix Ca²⁺, cytosolic Ca²⁺, and mitochondrial matrix pH. (A) Graph showing a linear relationship between the volume average cytosolic Ca²⁺ concentration, [Ca²⁺]_i, and mitochondria matrix Ca²⁺ concentration, [Ca²⁺]_m. (B) (Left) Graph depicting the equilibrium mitochondrial matrix pH associated with different mitochondrial matrix Ca²⁺ concentrations, [Ca²⁺]_m. (Right) Plot showing differential changes in mitochondrial matrix pH associated with ∼10 μM step increase in mitochondrial matrix Ca²⁺ concentration. Error bars indicate the mean ± SE.

To quantify pH_m changes associated with Ca²⁺ uptake by mitochondria, we used mito-RP (10). The 490-nm excitation/ 520-nm emission wavelengths report pH_m changes relatively independent of Ca²⁺ (19). Mito-RP was calibrated with standard pH solutions containing 100 μM digitonin and 10 μM CCCP, producing pK_a ∼ 8.61 (Fig. S1 B). The matrix pH level in MN13-Ib mitochondria at rest was estimated to be ∼7.80 (Table S1). Nerve stimulation caused a rapid transient change in pH_m followed by a plateau above baseline (Fig. 1 B), with the latter used as a measure of the equilibrium pH_m. pH_m showed a biphasic relationship with increasing [Ca²⁺]_m levels. [Ca²⁺]_m accumulation below ∼20 μM led to mitochondrial matrix acidification ([Ca²⁺]_m(30 Hz) ≈ 13.1 μM, pH_m = 7.74 ± 0.02, n = 9), whereas [Ca²⁺]_m above that level alkalinized the matrix ([Ca²⁺]_m(80 Hz) ≈ 44.3 μM, pH_m = 7.95 ± 0.03, n = 13) (Fig. 4 B). These changes in pH_m were only minimally affected by concurrent acidification in the cytosol (Fig. S2), permitting a correlation between [Ca²⁺]_m and pH_m values. Plotting of differential changes in pH_m against [Ca²⁺]_m shows that maximal mitochondrial alkalinization per unit of Ca²⁺ is achieved at close to [Ca²⁺]_m ∼ 26 μM, or 42 Hz stimulation, which is the average rate at which MN13-Ib MN fires in situ (10).

Although pH_m changes are important for the control of overall mitochondrial metabolic activity, and they generally mirror changes in _ΔΨ_m, proton gradient comprises only ∼20% of the proton motive force required for ATP synthesis (20). Therefore, we also sought to determine _ΔΨ_m changes to evaluate mitochondrial capacity for ATP synthesis at different [Ca²⁺]_m levels. To measure _ΔΨ_m, we employed TMRE in nonquench mode, a rhodamine-based cationic mitochondrial probe that follows the Nernstian distribution in mitochondria. Stimulation at 30 Hz for 2 s resulted in _ΔΨ_m depolarization followed by recovery to baseline (Fig. 5) in MN13-Ib mitochondria (_ΔF/Fo (30 Hz) = 0.9 ± 0.7%, n = 8). Higher-frequency stimulations gave rise to an after-stimulus _ΔΨ_m hyperpolarization with a peak around 15 s after the onset of a stimulus train similar to the NAD(P)H-stimulation-evoked increase seen in these terminals (10). _ΔΨ_m increases became nearly saturated after 60 Hz (_ΔF/Fo (42 Hz) = 4.1 ± 1.1%, n = 7; _ΔF/Fo (60Hz) = 8.4 ± 0.9%, n = 8; and _ΔF/Fo (80 Hz) = 9.3 ± 1.1%, n = 5), which parallels pH_m changes. Thus, both pH_m and _ΔΨ_m measurements suggest that mitochondrial metabolism in Drosophila larval nerve terminals is most responsive at [Ca²⁺]_m levels in the range∼20–30 μM.

Figure 5.

Imaging of mitochondrial membrane potential changes (_ΔΨ_m) with TMRE probe. (Upper) Representative traces of TMRE fluorescent changes ((F − Fo)/Fo = _ΔF/Fo) in response to 2-s stimuli at 30 and 42 Hz in mitochondria of MN13-Ib terminals. (Lower) Quantification of after-stimulus _ΔΨ_m increases for different stimulation frequencies. Multigroup comparisons were carried out using analysis of variance with Tukey’s posttest. Error bars indicate the mean ± SE.

Discussion

We have determined presynaptic values of [Ca²⁺]_m in Drosophila larval MNs by employing both chemical and genetically encoded Ca²⁺ indicators (GECIs). [Ca²⁺]_m was empirically derived from the ratio of rhod-5N fluorescence responses to two [Ca²⁺]_i transients of different amplitudes, relative to the ratio of rhod-FF responses to comparable [Ca²⁺]_i transients. Our calculations reveal that [Ca²⁺]_m rises to 24.3 μM when the MN fires close to its typical in vivo rate for 2 s. Low-affinity matrix-targeted ratiometric GECI D4cpv corroborated this [Ca²⁺]_m estimate (27.0 μM). Our estimates of elevated [Ca²⁺]_m were refined using the high-affinity matrix-targeted ratiometric GECI TN-XXL to estimate resting [Ca²⁺]_m. Measures of change in mitochondrial energy metabolism (i.e., changes in pH_m and _ΔΨ_m) were collected using matrix-targeted GEpHI ratiometric pericam and the mitochondrial cationic dye TMRE, respectively. We found that pH_m did not alkalinize and _ΔΨ_m did not increase significantly at stimulation frequencies below the endogenous firing rate of MN, which suggests that mitochondrial energy metabolism is well integrated with presynaptic activity. The fact that neither pH_m nor _ΔΨ_m showed an appreciable increase until [Ca²⁺]_m reached ∼26 μM also indicates that the least Ca²⁺-sensitive elements of oxidative phosphorylation might be responsible for setting the overall Ca²⁺ sensitivity of mitochondrial energy metabolism in situ.

We interpret the [Ca²⁺]_m levels measured in this study to represent fluctuations in a physiological range. The average peak firing rate of MN13-Ib during fictive locomotion is 42.4 ± 1.6 Hz (10), but it can fire at up to 80 Hz for brief periods. When driven at 80 Hz for 2 s, [Ca²⁺]_m rises (on average) to ∼54 μM. Longer stimulation trains and stimulation rates exceeding 80 Hz do not yield [Ca²⁺]_m estimates significantly higher than 54 μM, although higher stimulus rates result in higher [Ca²⁺]_i (data not shown). D4cpv reported a higher [Ca²⁺]_m level when mitochondria were permeabilized in the presence of 1 mM Ca²⁺ relative to 100 μM Ca²⁺ (Fig. 3 A), indicating that [Ca²⁺]_m levels >54 μM would have been reported had they been achieved during intense nerve stimulation. As mitochondrial proteases could be activated by high levels of [Ca²⁺]_m (21), we considered the possibility that [Ca²⁺]_m may achieve toxic levels, possibly as a consequence of measurements being done ex vivo. Our highest estimates of [Ca²⁺]_m are below those made using low-affinity Ca²⁺ indicators in either neuronal or nonneuronal (HeLa) cultured mammalian cells, where estimates peaked between 100 μM and near-millimolar levels (14,15,22–24), but the toxicity of such levels is rarely addressed. Two observations argue against [Ca²⁺]_m levels of ∼54 μM leading to toxic and irreversible effects in this study. First, the distribution and gross morphology of groups of presynaptic mitochondria remained essentially unchanged for up to 6 h after nerve stimulation at 80 Hz ex vivo. Second, mitochondrial pH_m and [Ca²⁺]_m responses (measured with chemical Ca²⁺ indicators or GECIs) remained remarkably robust and consistent for up to 6 h poststimulation ex vivo.

Although our [Ca²⁺]_m estimates are below the highest levels reported from cells in culture, they are substantially above the only other [Ca²⁺]_m estimate made at presynaptic nerve terminals; an estimate of ∼1 μM from lizard MNs (25). The reasons for the discrepancy in the values between lizard and Drosophila are not obvious. One possibility is that the use of a rhod-based Ca²⁺ indicator (rhod-5N) in lizard MN terminals limited [Ca²⁺]_m levels or damaged mitochondria, as reported for the effects of rhod-2 and rhod-FF (26), which appears unlikely, as we also relied upon rhod-5N in addition to rhod-FF, and their use did not yield estimates significantly different from those based on our low-affinity GECI D4cpv. Second, Drosophila MN terminals are glutamatergic, whereas lizard MN terminals are cholinergic, but this fundamental difference offers no immediately evident explanation. Third, terminals between the preparations may differ in their cytosolic levels of Mg²⁺, adenine nucleotides, inorganic phosphate, Na⁺, and H⁺, all of which have been shown to influence Ca²⁺ uptake and/or release (27), but few of which have been quantified in either lizard or Drosophila MN terminals. Last, Paolo Bernardi and colleagues reported that Drosophila mitochondria in permeabilized S₂R⁺ cells display Ca²⁺ transport systems that match their mammalian equivalents (and presumably lizards), but they have a unique selective Ca²⁺ release channel (28).

Our estimate of resting [Ca²⁺]_m (∼220 nM) is well above [Ca²⁺]_i resting levels (∼32 nM), but very close to the [Ca²⁺]_i threshold at which presynaptic mitochondria begin to take up Ca²⁺ from the cytosol, >200–250nM. However, the significance of this coincidence is not clear. [Ca²⁺]_i rises to 350 ± 34 nM at the endogenous firing rate of the MN, and [Ca²⁺]_m rises to ∼26 μM. Firing at 80 Hz, [Ca²⁺]_i level reaches 551 ± 23 nM, and [Ca²⁺]_m reaches ∼54 μM within 2 s. Such a rapid accumulation of large amounts of Ca²⁺ by mitochondria might indicate their exposure to Ca²⁺ microdomains. We have previously established that the mitochondria in Drosophila MN terminals do not take up Ca²⁺ from the endoplasmic reticulum (10), although we cannot dismiss the possibility that mitochondria have access to voltage-gated calcium-channel microdomains at the plasma membrane.

Our observations that pH_m and _ΔΨ_m did not rise significantly in response to [Ca²⁺]_m as high as 13.1 μM raise the question of which Ca²⁺-binding targets in the matrix may be essential for the elevations seen in pH_m and _ΔΨ_m. The Ca²⁺-binding targets with a matrix locus are well defined: pyruvate dehydrogenase (PDH), NAD⁺-isocitrate dehydrogenase (ICDH), 2-oxoglutarate dehydrogenase, and the F₁-F₀-ATP synthase in the inner mitochondrial membrane. Ca²⁺ stimulates the activity of the F₁-F₀-ATP synthase, possibly through a number of mediators, but with an estimated apparent Ca²⁺ affinity of ∼1 μM (29). PDH supplies the Krebs cycle with acetyl-CoA and has an apparent Ca²⁺ affinity of ∼0.75 μM (30,31). 2-oxoglutarate dehydrogenase within the Krebs cycle has a higher apparent Ca²⁺ affinity than PDH in a high ADP/ATP environment (∼0.28 μM (31)). The Ca²⁺ affinity of ICDH, however, is at least an order of magnitude higher than that of any other dehydrogenases and estimated at 5 and 41 μM, depending on the ADP/ATP environment (31,32). The dehydrogenase Ca²⁺ affinity measurements above were made in tolulene-permeabilized mitochondria and likely reflect in situ Ca²⁺ affinities. Measurements made on the purified enzymes in solution generally support those reported here. However, Ca²⁺ affinity estimates on purified enzymes might be underestimated, as the use of chelation agents such as EGTA interferes with other ionic species (33). Evident in these estimates is the significant gap between the Ca²⁺ affinity of ICDH and those of the three other matrix Ca²⁺ targets. Our observations that [Ca²⁺]_m values of several tens of μM are required to elevate pH_m and _ΔΨ_m are consistent with ICDH stimulation being required to increase mitochondrial energy metabolism, and also with the theory of the three dehydrogenases acting in series. Acidification observed at low [Ca²⁺]_m levels may be due to the couptake of phosphate with Ca²⁺. Low [Ca²⁺]_m levels are incapable of stimulating ICDH, which is necessary to counteract the influx of protons in the form of H₂PO₄⁻. The high Ca²⁺ affinity of ICDH may also provide an explanation for why an elevation in mitochondrial metabolism is not observed in lizard MN terminals (34), where [Ca²⁺]_m peaks at only ∼1 μM (25).

Our analytical approach for estimating [Ca²⁺]_m, using data from two different Ca²⁺ indicators with distinct Ca²⁺ affinities, captures some of the benefits of low-affinity chemical Ca²⁺ indicators, such as their high dynamic range and relatively linear responses, but sidesteps their disadvantage of dye loss during membrane permeabilization and calibration. Mito-D4cpv provides for a conventional ratiometric imaging, but its biphasic Ca²⁺ response curves can be problematic, and constructing an in situ calibration curve also can be demanding (Fig. 3 A). The genetics required to introduce GECI transgenes into mutant backgrounds for analysis are also less convenient than using topically applied chemical Ca²⁺ indicators. However the use of mito-D4cpv, and mito-TN-XXL, is particularly valuable for making ratiometric measurements, because even if some protein is lost during permeabilization, which is unlikely, ratiometric estimates of [Ca²⁺]_m remain unaffected.

Conclusions

In this study, we provide estimates of presynaptic [Ca²⁺]_m from individually identified MN terminals in situ, from which we were able to obtain measures of changes in mitochondrial energy metabolism. The use of complementary Ca²⁺ indicators targeted to the matrix provided a robust estimate of [Ca²⁺]_m. The magnitude of the changes in [Ca²⁺]_m, and the [Ca²⁺]_m values at which pH_m and _ΔΨ_m change are surprisingly consistent with what is known about the relative affinities of Ca²⁺-responsive elements of oxidative phosphorylation, even though most of those data were gleaned from nonneuronal cell types outside of their cellular context.

Appendix: Estimation of Absolute Ca²⁺ Levels from the Relative Responses of Two Ca²⁺-Sensitive Indicators

Fluorescence of a Ca²⁺ indicator inside the cell can be approximated as

$$\text{F}\sim {\epsilon }_{\mathrm{I}}\times {\text{q}}_{\mathrm{I}}\times \left(1-\text{f}\right)+{\epsilon }_{\text{CaI}}\times {\text{q}}_{\text{CaI}}\times \text{f}+{\text{F}}_{\text{background}},$$ (1)

where ε_I and ε_CaI are the extinction coefficients, and q_I and q_CaI are the quantum yields of the Ca²⁺-free and Ca²⁺-bound forms of the indicator, respectively. f is the fraction of Ca²⁺-bound indicator ([CaI]/[I]_total). For typical BAPTA-based fluorescent indicators with a single Ca²⁺ binding site, f is equal to

$${\text{Ca}}^{2+}+\mathrm{I}\stackrel{{\text{K}}_{\text{d}}}{\iff }\text{CaI}$$

$${\text{K}}_{\text{d}}=\frac{\left[{\text{Ca}}^{2+}\right]\times \left[\text{I}\right]}{\left[\text{CaI}\right]}\text{;}\left[\text{CaI}\right]+\left[\text{I}\right]={\left[\text{I}\right]}_{\text{total}}$$

$$\text{f}=\frac{\left[\text{CaI}\right]}{{\left[\text{I}\right]}_{\text{total}}}=\frac{\left[{\text{Ca}}^{2+}\right]}{\left[{\text{Ca}}^{2+}\right]+{\text{K}}_{\text{d}}}.$$ (2)

For most of the intensity-based Ca²⁺ fluorescent indicators q_CaI ≫ q_I, whereas the extinction coefficients are comparable (35,36). Thus, Eqs. 1 and 2 can be rewritten as:

$${\text{F}}_{\text{rest}}={\epsilon }_{\text{CaI}}\times {\text{q}}_{\text{CaI}}\times \frac{{\left[{\text{Ca}}^{2+}\right]}_{\text{rest}}}{{\left[{\text{Ca}}^{2+}\right]}_{\text{rest}}+{\text{K}}_{\text{d}}}+{\text{F}}_{\text{background}}$$

$${\text{F}}_1={\epsilon }_{\text{CaI}}\times {\text{q}}_{\text{CaI}}\times \frac{{\left[{\text{Ca}}^{2+}\right]}_{\text{rest}}+\Delta {\left[{\text{Ca}}^{2+}\right]}_1}{{\left[{\text{Ca}}^{2+}\right]}_{\text{rest}}+\Delta {\left[{\text{Ca}}^{2+}\right]}_1+{\mathrm{K}}_{\mathrm{d}}}+{\mathrm{F}}_{\text{background}}$$

$${\mathrm{F}}_2={\epsilon }_{\text{CaI}}\times {\mathrm{q}}_{\text{CaI}}\times \frac{{\left[{\text{Ca}}^{2+}\right]}_{\text{rest}}+\Delta {\left[{\text{Ca}}^{2+}\right]}_2}{{\left[{\text{Ca}}^{2+}\right]}_{\text{rest}}+\Delta {\left[{\text{Ca}}^{2+}\right]}_2+{\mathrm{K}}_{\mathrm{d}}}+{\text{F}}_{\text{background}}.$$

F_rest, F₁, and F₂ are the indicator fluorescence intensities at rest and during stimuli 1 and 2. Δ[Ca²⁺]₁ and Δ[Ca²⁺]₂ are corresponding buffered Ca²⁺ influxes in response to those stimuli. Assuming Δ[Ca²⁺]₁, Δ[Ca²⁺]₂ > [Ca²⁺]_rest, the relative changes in the fluorescence of the indicator can be written as

$$\frac{{\text{F}}_1-{\text{F}}_{\text{rest}}}{{\text{F}}_2-{\text{F}}_{\text{rest}}}=\frac{\text{Δ}{\left[{\text{Ca}}^{2+}\right]}_1\times \left(\text{Δ}{\left[{\text{Ca}}^{2+}\right]}_2+{\mathrm{K}}_{\mathrm{d}}\right)}{\text{Δ}{\left[{\text{Ca}}^{2+}\right]}_2\times \left(\text{Δ}{\left[{\text{Ca}}^{2+}\right]}_1+{\mathrm{K}}_{\mathrm{d}}\right)}.$$ (3)

Comparison of the relative fluorescent responses of two indicators with different Ca²⁺ affinities to a similar set of two stimuli allows for generation of two linear equations that can be solved for Δ[Ca²⁺]₁ and Δ[Ca²⁺]₂. In the case of rhod-FF (FF) and rhod-5N (5N), Eq. 3 can be rewritten as

$$\frac{{\text{F}}_1^{\text{FF}}-{\text{F}}_{\text{rest}}^{\text{FF}}}{{\text{F}}_2^{\text{FF}}-{\text{F}}_{\text{rest}}^{\text{FF}}}=\frac{\text{Δ}{\left[{\text{Ca}}^{2+}\right]}_1\times \left(\text{Δ}{\left[{\text{Ca}}^{2+}\right]}_2+{\mathrm{K}}_{\mathrm{d}}^{\text{FF}}\right)}{\text{Δ}{\left[{\text{Ca}}^{2+}\right]}_2\times \left(\text{Δ}{\left[{\text{Ca}}^{2+}\right]}_1+{\mathrm{K}}_{\mathrm{d}}^{\text{FF}}\right)}={\mathrm{R}}^{\text{FF}}$$

$$\frac{{\text{F}}_1^{5\text{N}}-{\text{F}}_{\text{rest}}^{5\text{N}}}{{\text{F}}_2^{5\text{N}}-{\text{F}}_{\text{rest}}^{5\text{N}}}=\frac{\text{Δ}{\left[{\text{Ca}}^{2+}\right]}_1\times \left(\text{Δ}{\left[{\text{Ca}}^{2+}\right]}_2+{\mathrm{K}}_{\mathrm{d}}^{5\text{N}}\right)}{\text{Δ}{\left[{\text{Ca}}^{2+}\right]}_2\times \left(\text{Δ}{\left[{\text{Ca}}^{2+}\right]}_1+{\mathrm{K}}_{\mathrm{d}}^{5\text{N}}\right)}={\text{R}}^{5\text{N}}.$$

The solutions for these equations are

$$\text{Δ}{\left[{\text{Ca}}^{2+}\right]}_1=\frac{{\text{K}}_{\text{d}}^{\text{FF}}\times {\text{K}}_{\text{d}}^{5\text{N}}\times \left({\text{R}}^{\text{FF}}-{\text{R}}^{5\text{N}}\right)}{{\text{K}}_{\text{d}}^{5\text{N}}\times \left(1-{\text{R}}^{\text{FF}}\right)+{\text{K}}_{\text{d}}^{\text{FF}}\times \left({\text{R}}^{5\text{N}}-1\right)}$$ (4)

$$\text{Δ}{\left[{\text{Ca}}^{2+}\right]}_2=\frac{{\text{K}}_{\text{d}}^{\text{FF}}\times {\text{K}}_{\text{d}}^{5\text{N}}\times \left({\text{R}}^{\text{FF}}-{\text{R}}^{5\text{N}}\right)}{{\text{R}}^{5\text{N}}\times {\text{K}}_{\text{d}}^{5\text{N}}\times \left(1-{\text{R}}^{\text{FF}}\right)+{\text{R}}^{\text{FF}}\times {\text{K}}_{\text{d}}^{\text{FF}}\times \left({\text{R}}^{5\text{N}}-1\right)}.$$ (5)

More accurate estimates of Ca²⁺ concentration for each stimulus can be calculated by estimation of the resting Ca²⁺ concentration with a high-affinity indicator and adding it to the buffered Ca²⁺ influx values found using the equations above, [Ca²⁺] = [Ca²⁺]_rest + Δ[Ca²⁺]. However, in the case of mitochondria, Ca²⁺ concentration at rest is typically several orders of magnitude smaller than after stimulus-induced Ca²⁺ uptake. Thus, the buffered Ca²⁺ influx values represent good quantitative estimates of equilibrium mitochondrial free Ca²⁺ concentration.