Miro1 is a calcium sensor for glutamate receptor-dependent localization of mitochondria at synapses

Abstract

Energy use, mainly to reverse ion movements in neurons, is a fundamental constraint on brain information processing. Trafficking of mitochondria to locations in neurons where there are large ion fluxes is essential for powering neural function. Mitochondrial trafficking is regulated by Ca2+ entry through ionotropic glutamate receptors, but the underlying mechanism is unknown. We show that the protein Miro1 links mitochondria to KIF5 motor proteins, allowing mitochondria to move along microtubules. This linkage is inhibited by micromolar levels of Ca2+ binding to Miro1. With the EF hand domains of Miro1 mutated to prevent Ca2+ binding, Miro1 could still facilitate mitochondrial motility, but mitochondrial stopping induced by glutamate or neuronal activity was blocked. Activating neuronal NMDA receptors with exogenous or synaptically released glutamate led to Miro1 positioning mitochondria at the postsynaptic side of synapses. Thus, Miro1 is a key determinant of how energy supply is matched to energy usage in neurons.

Figure 1

Altering Miro1 Expression Affects Mitochondrial Mobility via an Interaction with KIF5 Neurons were transfected with mtdsred2 as well as Miro1 GFP or shRNAi to Miro1, or mtdsred2 alone, 2–3 days before being imaged at DIV 12–14. (A–C) Static image of a dendrite at time = 0 in mtdsred2-transfected cell (A), mtdsred2- and Miro1-transfected cell (B), and mtdsred2- and Miro1 RNAi-transfected cell (C). (A′–C′) Kymographs showing increased mitochondrial movement in a neuronal dendrite upon Miro1 expression (B′) and decreased mitochondrial movement upon RNAi expression (C′) compared to controls (A′). Height, 2 min (time increases down the page); scale bar, 10 μm. (D) Percentage of mitochondria moving in dendrites of control cells (n = 8 dendrites, 295 mitochondria), cells expressing Miro1 (n = 8 dendrites, 144 mitochondria), cells expressing Miro1 RNAi (n = 9 dendrites, 569 mitochondria), and cells expressing a scrambled control RNAi (n = 7 dendrites, 412 mitochondria). Error bars here and throughout represent the standard error of the mean. p values compare with the control bar. (E) Western blot showing specific knockdown of Miro1 in cultured cortical neurons using Miro1 RNAi. Actin used as a control for loading shows no change. (F) Average velocity of moving mitochondria in dendrites of control cells, cells expressing Miro1, and cells expressing Miro1 RNAi. (G) Schematic of the primary structure of Miro1. (H–I′) Static images and kymographs showing mitochondrial movement through dendrites transfected with Miro1 GFP and transduced with either control (9E10) antibody (H) or a function-blocking KIF5 motor (SUK4) antibody (I). Height, 2 min; scale bar, 10 μm. (J) Percentage of mitochondria moving in dendrites of control 9E10-transduced cells (n = 8 dendrites, 386 mitochondria) and SUK4-transduced cells (n = 11 dendrites, 211 mitochondria). (K) COS7 cells were transfected with KIF5C and either an empty vector or Miro1. Mitochondria were isolated, and blots of the mitochondrial fraction of the cells, and of the total cell lysate, were probed for KIF5C, Miro1, and (as a reference mitochondrial protein) mtHsp70. In the mitochondrial fraction the amount of KIF5C (and of Miro1) is increased by Miro1 expression (compared to empty vector), but the total level of KIF5C is unaltered. (L) Quantification of four replicates of the data in (K), with the levels of KIF5C and Miro1 normalized to the level of the reference protein mtHsp70 in the mitochondrial fraction.

Figure 2

Ca²⁺-Dependent Binding of Miro1 to KIF5 Motors In Vitro and In Situ (A–E) The kinesin subunits KIF5A, KIF5B, or KIF5C, and TRAK2 fragments (476–700 or 1–700) were translated in vitro and labeled with ³⁵S-methionine. The resulting protein was subjected to pull-down assay with GST-Miro1 or GST-Miro1-ΔEF or with GST alone as control, separated by SDS-PAGE, and radioactivity was detected using a phosphor storage screen. Input represents 10% of input used in each experiment. (A) Pull-down assay demonstrating that Miro1 binds directly to ³⁵S-labeled KIF5A, KIF5B, and KIF5C (with no other proteins in the system). No added Ca²⁺ present. GST alone did not bind KIF motors. (B) Repeat of (A) in different [Ca²⁺]. High [Ca²⁺] inhibits Miro1 binding to KIF5C, and this inhibition is absent when Miro1's EF hands are mutated. (C) Pull-down assay demonstrating that Miro1 binds directly to TRAK2 (amino acids 476–700), with no other proteins in the system, in a Ca²⁺-independent manner. (D) Pull-down assay with only KIF5C, Miro1, and TRAK2 (amino acids 1–700) in the system. As in (B) and (C), Miro1 binds to KIF5C in a Ca²⁺-dependent manner, but binds to TRAK2 in a Ca²⁺-independent manner. (E) As in (D), but using the EF hand mutant of Miro1: binding of Miro1-ΔEF to KIF5C is Ca²⁺ independent. (F) Coimmunoprecipitation showing that binding of endogenous Miro1 and KIF5 motors is Ca²⁺ sensitive. (G and H) Binding of Miro1 to KIF5 is sensitive to calcium in the physiological range, and this sensitivity is occluded by a mutation in the EF hand domains of Miro1 (Miro1 ΔEF). Solubilized brain homogenates were subjected to a GST pull-down assay with either GST-Miro1 WT or GST-Miro1-ΔEF in the presence of varying concentrations of [Ca²⁺]_free as shown (G). Equal protein loading was visualized using Ponceau stain. The GST-Miro1 WT-KIF5 interaction was blocked by calcium, while the GST-Miro1 ΔEF interaction showed little calcium sensitivity. (H) shows quantification of data as in (G), as a function of [Ca²⁺]_free (normalized to data in zero Ca²⁺; n = 3 for 1 μM, 2 for 5, 50 μM, 1 for 15 μM). Curves have the forms (with c = [Ca²⁺], K = K_m of site): K/(c + K) if there is only one EF hand binding Ca²⁺ to inhibit the interaction (black line); {K/(c + K)}² if two EF hands of identical K_m can bind Ca²⁺ and the interaction with KIF5 is inhibited if one or both of them bind Ca²⁺ (red); and 1 − {c/(c + K)}² if both EF hands have to bind Ca²⁺ to block the interaction (green). Best-fit K_m values are shown in the inset. (I–K) Glutamate evokes dissociation of KIF5 from mitochondria in brain slices. (I) After fractionation of brain slice lysate, essentially all cytochrome c is in the mitochondrial fraction, and none in the vesicle fraction (probed with antibody to the vesicle protein SV2). (J) Glutamate treatment of brain slices (100 μM for 10 min, with 1 μM glycine) decreases the amount of KIF5 in the mitochondrial fraction. (K) Quantification of data as in (J) on eight slices from six rats. (L) Schematic mechanism to explain the data. Miro1 links mitochondria to KIF5 motors, and a glutamate-evoked rise of [Ca²⁺] dissociates Miro1 from KIF5. TRAK2 remains bound to Miro1 even when [Ca²⁺] is high.

Figure 3

Miro1 Mediates Glutamate-Dependent Mitochondrial Stopping (A–F) Neurons transfected with Miro1 WT and mtdsred2, Miro1 ΔEF and mtdsred2, or mtdsred2 alone were treated with 30 μM glutamate and 1 μM glycine for 10 min. Dendrites were imaged at times −3 to −1 min and 10–12 min with respect to the start of glutamate application. Kymographs (A′–F′) show mitochondrial movement in control (A and B), Miro1 WT-transfected (C and D), and Miro1 ΔEF-transfected (E and F) dendrites before (A, C, and E) and after (B, D, and F) glutamate treatment. Height, 2 min (time increases down the page); scale bars, 10 μm. (G) Percentage of mitochondria moving before glutamate treatment in dendrites of control cells (n = 8 dendrites, 354 mitochondria), Miro1 WT-transfected cells (n = 8 dendrites, 257 mitochondria), and Miro1 ΔEF-transfected cells (n = 8 dendrites, 253 mitochondria). (H) Percentage of mitochondria moving after glutamate treatment, normalized to the percentage of mitochondria moving before treatment.

Figure 4

Activation of Glutamate Receptors Recruits Mitochondria to Synapses by a Mechanism Depending on the EF Hands of Miro1 (A) Coculture of neurons expressing either synaptophysin GFP or mtdsred2. Time-lapse microscopy shows that upon glutamate stimulation (starting at time = 0) mitochondria (red), which initially are spread out throughout the process, move to synaptophysin-positive synaptic zones (actual GFP fluorescence is shown in the top panel and is schematized as green bar in lower panels). Scale bar, 2 μm. (B) Neuronal dendrites transfected with mtdsred2 alone, or mtdsred2 plus Miro1 WT, or Miro1 ΔEF before (mtdsred2 alone) and after glutamate treatment. Cells were fixed and stained with the presynaptic marker SV2 (green). Arrows show mitochondria in contact with presynaptic zones. Scale bar, 5 μm. (C) Cumulative distribution of the distance of mitochondria from the nearest synapse in untransfected cells without (gray line, n = 129 mitochondria in 6 dendrites) and with glutamate application (black line, n = 84 mitochondria in 6 dendrites). p values compare measurements with Monte Carlo simulation of random mitochondrial and synaptic distribution in 10,000 dendrites (dotted red line). (D) Cumulative distribution in Miro1 WT-transfected cells without (gray line, n = 45 mitochondria in 4 dendrites) and with glutamate (black line, n = 53 mitochondria in 6 dendrites) compared to Monte Carlo simulation of random mitochondrial and synaptic distribution (dotted line). (E) Cumulative distribution in Miro1 ΔEF-transfected cells without (gray line, n = 48 mitochondria in 4 dendrites) and with glutamate (black line, n = 113 mitochondria in 8 dendrites) compared to Monte-Carlo simulation of random mitochondrial and synaptic distribution (dotted line). (F) Frequency distribution of mean intersynapse distance from measured values (black line, n = 34 dendrites) and simulations (dotted line). (G) Mean distance of mitochondria along dendrite from SV2-positive clusters without glutamate treatment (untransfected controls, n = 213 mitochondria; Miro1 WT, n = 98 mitochondria; Miro1 ΔEF, n = 161 mitochondria, p values compared with the Monte-Carlo simulation with randomly positioned synapses and mitochondria [right bar] are 0.20, 0.31, and 0.15, respectively). (H) Mean distance of mitochondria from SV2-positive clusters after glutamate treatment, normalized to nontreated values, numbers of mitochondria were as follows: control, 218; wild-type, 57; EF hand mutant, 199.

Figure 5

Stimulation of Neuronal Activity Recruits Mitochondria to Synapses as a Result of Ca²⁺ Entering via Synaptic NMDA Receptors Binding to Miro1's EF Hands (A) Field stimulation increases action potential firing in cultured neurons (recorded in current-clamp mode using the whole-cell patch-clamp technique). Voltage recording before, during, and after two 100 Hz stimulation trains: inset shows activity at a faster timescale. (B) Stimulation of cultures at 100 Hz for 1 s results in a rapid drop in mitochondrial movement that lasts for ∼150 s after stimulation (0.35 ± 0.18 of mitochondria moving 120 s after stimulus normalized to baseline, n = 5 dendrites, 53 mitochondria, p = 0.009). This effect is blocked by D-APV (1.15 ± 0.25 fold increase, relative to control, of mitochondria moving after 120 s, n = 4 dendrites, 71 mitochondria, p = 0.14). (C) Neural activity and synaptic glutamate release evoked by block of GABA-mediated inhibition with bicuculline results in increased stopping of mitochondria. Traces are normalized to 2 min baseline movement before the experiment; values stated below are at maximum response at 150 s. Addition of 50 μM bicuculline decreased movement to 0.54 ± 0.08 (n = 6 dendrites, 166 mitochondria, p = 0.0005) of the control level, whereas the same treatment in zero calcium had no effect (1.10 ± 0.09 fold over control, n = 4 dendrites, 193 mitochondria, p = 0.66). Addition of 50 μM D-APV and bicuculline resulted in a slight increase in movement (1.38 ± 0.05 fold over control, n = 4 dendrites, 200 mitochondria, p = 0.03). (D) The bicuculline effect can be blocked using Miro1 ΔEF. Miro1 WT-expressing cells respond in an equivalent manner to control cells to bicuculline treatment (0.59 ± 0.07 of baseline value, n = 4 dendrites, 128 mitochondria, p = 0.006), whereas Miro1 ΔEF-expressing cells do not respond (1.12 ± 0.06, n = 5 dendrites, 226 mitochondria, p = 0.22). (E) Path of a single mitochondrion along a dendrite before (gray) and during and after (red and black at times shown in [F]) a 1 s period of field stimulation at 100 Hz. Synapse SV2 label shown as green. (F) Distance from the nearest synapse at any given time for the mitochondrion in (E). (G) Mean distance of mitochondria from the nearest synapse in the dendrites of five neurons for field stimulation as in (E) and (F).

Figure 6

Miro1-Mediated Recruitment of Mitochondria to Synapses Occurs Even When Only a Few Synapses Are Active (A) Image of cell axon (colored green for clarity) stimulated with a pipette to evoke local synaptic input to a dendrite (blue). Mitochondria labeled with mtdsred2 appear red (for clarity, a mask has been used to remove mitochondrial fluorescence from most of the image). Boxes show regions over which mitochondrial motility was measured near (black) and away (white) from the synaptic input. Inset shows normalized percentage mitochondrial movement before, during, and after 100 Hz stimulation at t = 0 (two 1 s trains separated by 2 s) in boxed areas near (black) and away from synaptic input (white). (B) Mean normalized percent mitochondrial movement before, during, and after 100 Hz stimulation at t = 0, near to (black, 5 dendrites, 45 mitochondria) and away from (gray, 5 dendrites, 53 mitochondria) the site of synaptic input. (C) Experiment as in (A) and (B) comparing results at sites within 15 μm of synaptic inputs for cells expressing wild-type Miro1 and the Miro1 EF hand mutant (in addition to mtdsred2). Mean data are from 8 dendrites with 79 mitochondria for Miro1-WT cells and 8 dendrites with 69 mitochondria for Miro1-ΔEF cells.