Molecular Tuning of the Axonal Mitochondrial Ca2+ Uniporter Ensures Metabolic Flexibility of Neurotransmission

Abstract

The brain is a vulnerable metabolic organ and must adapt to different fuel conditions to sustain function. Nerve terminals are a locus of this vulnerability, but how they regulate ATP synthesis as fuel conditions vary is unknown. We show that synapses can switch from glycolytic to oxidative metabolism, but to do so, they rely on activity-driven presynaptic mitochondrial Ca²⁺ uptake to accelerate ATP production. We demonstrate that, whereas mitochondrial Ca²⁺ uptake requires elevated extramitochondrial Ca²⁺ in non-neuronal cells, axonal mitochondria readily take up Ca²⁺ in response to small changes in external Ca²⁺. We identified the brain-specific protein MICU3 as a critical driver of this tuning of Ca²⁺ sensitivity. Ablation of MICU3 renders axonal mitochondria similar to non-neuronal mitochondria, prevents acceleration of local ATP synthesis, and impairs presynaptic function under oxidative conditions. Thus, presynaptic mitochondria rely on MICU3 to facilitate mitochondrial Ca²⁺ uptake during activity and achieve metabolic flexibility.

Keywords: ATP; Ca2+ imaging; MCU; MICU3; metabolic flexibility; metabolism; mitochondria; spiking HEK cells; synapse; synaptic transmission.

Figure 1.. Activity-driven mitochondrial ATP production in nerve terminals.

(A) Average normalized traces of presynaptic ATP, measured with the luminescent reporter Syn-ATP. Neurons were electrically stimulated with 600 action potentials (AP) at 10 Hz (black bar) before and after a 5-minute treatment with oligomycin A in the presence of lactate and pyruvate but with no glucose. The grey dashed line represents the normalized ATP value of 1, except for the case in which oligomycin A is added, where it represents the linear fitting to the points prior to stimulation. (B) Change in normalized ATP levels immediately after stimulation (n= 8 cells). Average change in ATP ± SEM: control, −0.02 ± 0.02; oligomycin, −0.51 ± 0.07. (C) Endocytosis time constants (tau) of vGLUT1-pH in neurons stimulated in the presence of glucose or lactate and pyruvate. n= 21-24 cells. Average tau (sec) ± SEM; glucose, 4.0 ± 0.3; lact+pyr, 4.8 ± 0.5. (D) Sample vGLUT1-pH traces in response to 100 AP at 10 Hz (black bar) before and 5 minutes after oligomycin A treatment in the presence of lactate and pyruvate. (E) Fractional retrieval was calculated as the fraction of vGLUT1-pH signal remaining after 2 times the endocytic constant of control (red dashed line in D) after stimulation (0 and 1 denote complete and no endocytosis, respectively). See STAR methods. n = 8 cells. Average fractional retrieval ± SEM: control, 0.2 ± 0.02; oligomycin A 0.8 ± 0.2. Error bars are SEM. **p<0.01, ***p<0.001, Mann-Whitney U test. n.s. not significant. The box-whisker plot represents median (line), 25th–75th percentile (box), and min-max (whisker).

Figure 2.. MCU is required for mitochondrial Ca²⁺ uptake and activity-driven ATP production.

(A) Average traces of mito^4x-GCaMP6f (A) showing mitochondrial Ca²⁺ uptake in control and MCU KD axons stimulated with 20 AP at 20 Hz. (B) Peak responses of Mito^4x-GCaMP6f (ΔF/F) following stimulation. n = 19-39 cells. mean ΔF/F ± SEM: control, 0.41 ± 0.05; MCU KD (shRNA-1), 0.04 ± 0.02; MCU KD (shRNA-2), 0.19 ± 0.05. Henceforth, MCU KD refers to MCU KD with shRNA-1. Error bars are SEM. ****p<0.0001, Kruskal-Wallis H Test. (C, D) Normalized presynaptic ATP traces (C) in control and MCU KD neurons stimulated with 600 AP at 10 Hz. The grey dashed line represents the normalized ATP value of 1. (D) Change in normalized ATP levels immediately after stimulation. n= 9-21 cells. Average change in ATP ± SEM: control, −0.04 ± 0.02; MCU KD, −0.17 ± 0.03. Error bars are SEM. ***p<0.001, Mann–Whitney U test. (E, F) Sample semi-log plot of vGLUT1-pH traces (E) following stimulation with 600 AP at 10 Hz in control (black) and MCU KD neurons (red) supplied with lactate and pyruvate. Blue trace shows the response of the same MCU KD neuron shown in red but in the presence of media containing glucose. (F) Fractional retrieval block calculated as described in STAR methods. n = 9-18 cells. Average fractional retrieval ± SEM: control, 0.16 ± 0.02; MCU KD (lact+pyr), 0.40 ± 0.09; MCU KD (glucose), 0.15 ± 0.05. Unless otherwise indicated, all experiments were performed in the presence of lactate and pyruvate. *p<0.05, **p<0.01, ***p<0.001, Kruskal-Wallis H Test, and Wilcoxon matched-pairs test for comparison between MCU KD (lact+pyr) and MCU KD (glucose). The box-whisker plot represents median (line), 25th–75th percentile (box), and min–max (whisker).

Figure 3.. Axonal mitochondria do not rely on the ER for Ca²⁺ uptake.

(A-D) Average traces of ER-GCaMP6-150 (A), or mito^4x-GCaMP6f (C) in control and neurons treated with the SERCA inhibitor CPA to silence axonal ER Ca²⁺ responses (A, B). Paired comparison of peak ER-GCaMP6-150 (B) or mito^4x-GCaMP6f (D) ΔF/F responses before and after CPA treatment. n = 8-12 cells. Mean peak ER Ca²⁺ response (ΔF/F) ± SEM: control, 0.77 ± 0.16; CPA-treated, 0.04 ± 0.01. Average peak mitochondrial Ca²⁺ response (ΔF/F): control, 1.88 ± 0.38; CPA-treated, 1.82 ± 0.41. Experiments were performed in the presence of glucose. Error bars in traces are SEM. **** P < 0.0001; n.s. = not statistically significant, paired Kolmogorov-Smirnov test. The box-whisker plot represents median (line), 25th–75th percentile (box), and min–max (whisker).

Figure 4.. Lower Ca²⁺ levels are needed to trigger mitochondrial Ca²⁺ uptake in nerve terminals than in non-neuronal cells.

(A-C) Correlation of the mitochondrial matrix-targeted mito^4x-GCaMP6f and the outer membrane-targeted OMM-jRCaMP1b signal (A) in response to a given stimulus in mitochondria of “spiking” HEK cells and neuronal axons. Responses in “spiking” HEK cells were binned across OMM responses for ease of comparison with neurons. (B, C) Averaged traces of matrix (green) Ca²⁺ fluxes in “spiking” HEK cells (B) and neuronal axons (C) in response to 20 AP at 20 Hz in neurons and the equivalent Ca²⁺ increase in the outer membrane (OMM) of HEK cells, measured by quantifying OMM Ca²⁺ signals (red) in both cell types (traces not shown for clarity, but average OMM peak response is denoted with a red dashed line). Experiments were performed in the presence of glucose. Error bars are SEM. **p<0.01, ****p<0.0001, Mann–Whitney U test. The box-whisker plot represents median (line), 25th–75th percentile (box), and min–max (whisker).

Figure 5.. MICU3 shifts the Ca²⁺ sensitivity of presynaptic mitochondrial Ca²⁺ uptake.

(A-C) Average traces of mito^4x-GCaMP6f (B) in neuronal axons (schematic in A) stimulated with 20 AP at 20 Hz (B, top panel) or 100 AP at 100 Hz (B, bottom panel) in control and MICU3 KD neurons. (C) Peak mito^4x-GCaMP6f ΔF/F following stimulation. n = 8-39 cells. Average ΔF/F ± SEM: control (20 AP at 20 Hz) (same as in Fig. 2A and B), 0.41 ± 0.05; MICU3 KD (20 AP at 20 Hz), 0.04 ± 0.02; control (100 AP at 100 Hz), 1.33 ± 0.20; MICU3 (100 AP at 100 Hz), 1.04 ± 0.29; MCU KD (100 AP at 100 Hz), 0.28 ± 0.11. (D-F) Average traces of mito^4x-GCaMP6f (E) in control and MICU3-overexpressing “spiking” HEK cells (schematic in D) stimulated for 10 msec. To facilitate comparison, only HEK cells whose OMM Ca²⁺ signals are equivalent to neurons stimulated with 20 AP at 20 Hz are plotted. (F) Peak mito^4x-GCaMP6f ΔF/F following stimulation. n = 34-43 cells. Average ΔF/F ± SEM: control, 0.03 ± 0.01; MICU3 overexpression, 0.4 ± 0.13. Error bars are SEM. p>0.05, **p<0.01, ****p<0.0001, Mann–Whitney U test. The box-whisker plot represents median (line), 25th–75th percentile (box), and min–max (whisker).

Figure 6.. MICU3 is required for feed-forward regulation of ATP production in nerve terminals.

(A and B) Normalized presynaptic ATP traces (A) in control and MICU3 KD neurons stimulated with 600 AP at 10 Hz supplied with lactate and pyruvate. The grey dashed line represents the normalized ATP value of 1. (B) Change in normalized ATP levels immediately after stimulation. n= 8-21 cells. Average change in ATP; control (same as in Fig. 2C and D), −0.04 ± 0.02; MICU3 KD, −0.25 ± 0.03. ***p<0.001, Mann–Whitney U test. (C and D). Sample semi-log plot of vGLUT1-pH traces (C) following stimulation with 600 AP at 10 Hz in control and MICU3 KD neurons supplied with lactate and pyruvate or glucose. (D) Fractional retrieval block as described in STAR methods. n= 8-21 cells. Average retrieval block ± SEM: Control (same as in Fig. 2E and F), 0.16 ± 0.02; MICU3 KD (lact+pyr), 0.41 ± 0.05; MICU3 KD (glucose), 0.27 ± 0.07. Error bars are SEM. **p<0.01, ***p<0.001, Kruskal-Wallis H Test. The box-whisker plot represents median (line), 25th–75th percentile (box), and min–max (whisker).