Brain-derived Neurotrophic Factor (BDNF)-induced Mitochondrial Motility Arrest and Presynaptic Docking Contribute to BDNF-enhanced Synaptic Transmission*♦

Background: Appropriate mitochondrial transport and localization are highly associated with neuronal activities. Results: BDNF induces mitochondrial motility arrest via Miro1 binding with Ca2+, which facilitates BDNF-enhanced neurotransmitter release. Conclusion: BDNF-regulated mitochondrial motility is essential for BDNF-enhanced synaptic transmission. Significance: These results provide novel insights into the mechanistic link between BDNF-dependent mitochondrial motility and BDNF-mediated synaptic transmission. Appropriate mitochondrial transport and distribution are essential for neurons because of the high energy and Ca2+ buffering requirements at synapses. Brain-derived neurotrophic factor (BDNF) plays an essential role in regulating synaptic transmission and plasticity. However, whether and how BDNF can regulate mitochondrial transport and distribution are still unclear. Here, we find that in cultured hippocampal neurons, application of BDNF for 15 min decreased the percentage of moving mitochondria in axons, a process dependent on the activation of the TrkB receptor and its downstream PI3K and phospholipase-Cγ signaling pathways. Moreover, the BDNF-induced mitochondrial stopping requires the activation of transient receptor potential canonical 3 and 6 (TRPC3 and TRPC6) channels and elevated intracellular Ca2+ levels. The Ca2+ sensor Miro1 plays an important role in this process. Finally, the BDNF-induced mitochondrial stopping leads to the accumulation of more mitochondria at presynaptic sites. Mutant Miro1 lacking the ability to bind Ca2+ prevents BDNF-induced mitochondrial presynaptic accumulation and synaptic transmission, suggesting that Miro1-mediated mitochondrial motility is involved in BDNF-induced mitochondrial presynaptic docking and neurotransmission. Together, these data suggest that mitochondrial transport and distribution play essential roles in BDNF-mediated synaptic transmission.

Appropriate mitochondrial transport and distribution are essential for neurons because of the high energy and Ca 2؉ buffering requirements at synapses. Brain-derived neurotrophic factor (BDNF) plays an essential role in regulating synaptic transmission and plasticity. However, whether and how BDNF can regulate mitochondrial transport and distribution are still unclear. Here, we find that in cultured hippocampal neurons, application of BDNF for 15 min decreased the percentage of moving mitochondria in axons, a process dependent on the activation of the TrkB receptor and its downstream PI3K and phospholipase-C␥ signaling pathways. Moreover, the BDNF-induced mitochondrial stopping requires the activation of transient receptor potential canonical 3 and 6 (TRPC3 and TRPC6) channels and elevated intracellular Ca 2؉ levels. The Ca 2؉ sensor Miro1 plays an important role in this process. Finally, the BDNF-induced mitochondrial stopping leads to the accumulation of more mitochondria at presynaptic sites. Mutant Miro1 lacking the ability to bind Ca 2؉ prevents BDNFinduced mitochondrial presynaptic accumulation and synaptic transmission, suggesting that Miro1-mediated mitochondrial motility is involved in BDNF-induced mitochondrial presynaptic docking and neurotransmission. Together, these data suggest that mitochondrial transport and distribution play essential roles in BDNF-mediated synaptic transmission.
Mitochondria are important organelles with multiple functions, including ATP production, the maintenance of intracellular Ca 2ϩ homeostasis, and the regulation of apoptosis. In neurons, proper mitochondrial trafficking and distribution along neurites are involved in synaptic plasticity modulation such as presynaptically neurotransmitter release, postsynaptically membrane trafficking of ion channels, and dendritic spine maturation (1,2).
Recent work has begun to reveal the mechanisms underlying mitochondrial motility. In vivo and in vitro studies have identified several kinesin family members, including KIF5A/B/C, KIF1B␣, and KLP6, that are involved in mitochondrial anterograde transport (3)(4)(5)(6)(7). In addition to the motor proteins, intracellular signaling pathways that regulate mitochondrial movement have also been investigated. It has been reported that nerve growth factor (NGF) affects mitochondrial motility by its downstream phosphoinositide 3-kinase (PI3K) signaling pathway (8). Edelman and co-workers (9,10) have demonstrated that Akt-glycogen synthase kinase 3␤ (GSK3␤) is associated with axonal mitochondrial transport stimulated by serotonin and dopamine. Calcium has been demonstrated to be one of the critical factors that regulate mitochondrial transport (11)(12)(13). Recently, a complex composed of KIF5, Milton/TRAK, and Miro1 was reported to mediate mitochondrial transport along microtubules; Miro1 as the Ca 2ϩ sensor could regulate mitochondrial anterograde and retrograde movement in both axons and dendrites (13)(14)(15)(16).
BDNF 3 is broadly expressed in the central nervous system and plays critical roles in regulating neuronal survival, development, and synaptic plasticity. The actions of BDNF are dictated by two classes of receptors on the cell surface, TrkB and p75. Upon activation of TrkB receptors, several signaling pathways are activated, including MAPK, PI3K, and phospholipase-C␥ (17). It has been reported that in addition to regulating synaptic plasticity, BDNF can stimulate brain mitochondrial metabolism by increasing the efficiency of respiratory coupling and ATP synthesis (18). However, the mechanism by which BDNF regulates mitochondrial function is still unclear. In particular, it remains unknown whether BDNF can regulate mitochondrial transport and distribution in neurons.
In this study, we investigated the effects of BDNF on mitochondrial motility in primary hippocampal neurons. We report that BDNF could induce more mitochondrial docking at presynaptic sites, an effect mediated by the elevation of intracellular Ca 2ϩ through PI3K and PLC␥ signaling pathways and transient receptor potential canonical (TRPC) channels in a Miro1-dependent manner. More importantly, BDNF-enhanced synaptic transmission is prevented by mutant Miro1 lacking the ability to bind Ca 2ϩ , indicating an important role for presynaptically accumulated mitochondria on neurotransmission.

EXPERIMENTAL PROCEDURES
Reagents and Antibodies-Human recombinant BDNF was purchased from PeproTech (Rocky Hill, NJ). BAPTA-AM and K-252a were obtained from Calbiochem. Other inhibitors and activators were obtained from Sigma. Antibodies were purchased as follows: mouse anti-Miro1 antibody from Abnova (Taipei, Taiwan); rabbit anti-Kinesin 5AϩBϩC antibody from Abcam (Cambridge, MA); rabbit anti-VDAC antibody from Cell Signaling (Danvers, MA); mouse anti-GAPDH antibody from Upstate (Billerica, MA); rabbit anti-TrkB antibody from Millipore (Billerica, MA); mouse monoclonal anti-SV2 antibody from the Developmental Studies Hybridoma Bank (Iowa City, IA); horseradish peroxidase-conjugated goat anti-mouse or -rabbit IgG antibodies used for Western blot from Calbiochem; and Alexa Fluor 488-or 594-or Cy5-conjugated goat anti-mouse or rabbit IgG heavy and light chains (HϩL) used for immunofluorescent staining from Invitrogen. Restriction enzymes were purchased from Fermentas (Hanover, MD). Fluo4-AM, JC-1, FM 1-43 AM, trypsin, and cell culture reagents were purchased from Invitrogen. The ATP assay kit and mitochondrial isolation kit were purchased from Beyotime Institute of Biotechnology (Haimen, Jiangsu, China). All other reagents were from Sigma unless otherwise noted.
Plasmids Construction and siRNA-The pDsRed2-mito, which encodes mitochondrial targeted DsRed fluorescent protein, was purchased from Clontech. The coding region of rat Miro1 was amplified from rat cDNA and subcloned into pcDNA 3.1-flag (Invitrogen) and pEGFP-C2 (Clontech) expression vectors. The EF domain mutant rat Miro1 (E221K and E341K) construct was generated by means of a three-step PCR. Rab5 was constructed from rat cDNA and inserted into pEGFP-N1 (Clontech). The target sequences of TRPC3, TRPC5, and TRPC6 siRNAs were from previous studies (19 -21), as follows: TRPC3 5Ј-GGCTGCGCATTGCCATAAA-3Ј, TRPC5 5Ј-CACTCTTCGCGATATCGAA-3Ј, and TRPC6 5Ј-GAACGGCCTCATGATTATT-3Ј. All the siRNAs were subcloned into the pSuper-mRFP vector (OligoEngine, Seattle) according to the manufacturer's protocol, and the efficiency was confirmed by real time PCR. Dominant-negative con-structs of TRPC3, TRPC5, and TRPC6 were kindly provided by Yizheng Wang (Neurobiology Institute, Shanghai, China). All of the constructs were confirmed by DNA sequencing to exclude potential mutations introduced by PCR.
Neuronal Cultures and Transfection-Cultures of hippocampal neurons were prepared from timed Sprague-Dawley rat brains as described previously (22). In brief, hippocampi dissected from embryonic day 18 rats were placed into ice-cold Hanks' balanced saline solution, digested with 0.05% trypsin/ EDTA at 37°C for 10 min, and triturated in DMEM/F-12 medium containing 10% fetal bovine serum. Neurons were seeded onto coverslips coated with 0.1 mg/ml poly-D-lysine in 6-well plates at a density of 3-8 ϫ 10 5 and then cultured in Neurobasal medium containing 2% B27 supplement and 0.5 mM L-glutamine. Neurons were cultured in an incubator at 37°C with 5% CO 2 at 95% humidity for 12-14 days before treating with 50 ng/ml BDNF or other experiments, and the medium was half-changed every 3-4 days. The concentration of 50 ng/ml is a saturated dose for neurons and has been customarily used for the studies of the effects of BDNF on neuronal dendritic growth and synaptic activities (23,24). We used this concentration for all experiments. The neurons were transfected with Lipofectamine 2000 (Invitrogen) following the manufacturer's instructions at 8 -10 days in vitro (DIV) and used 2-4 days after transfection.
Mitochondrial Isolation and Western Blot-Mitochondria were isolated from cultured neurons at 12-14 DIV with the mitochondrial isolation kit from Beyotime Co. following the manufacturer's instructions. Briefly, cultured neurons in 10-cm dishes were starved in Neurobasal medium without B27 for 8 h and then treated with BDNF (50 ng/ml) for 15 min at 37°C. After that, neurons were harvested in mitochondrial isolation buffer with PMSF and transferred to a grinder for crushing until the solution became homogeneous. The mitochondria were isolated by differential centrifugation at 600 ϫ g and 3000 ϫ g for 10 min at 4°C, respectively, and stored in mitochondrial lysis buffer. For Western blotting, the mitochondrial proteins were boiled with 4ϫ sample buffer for 5 min and separated by SDS-PAGE. After transfer to a polyvinylidene difluoride (PVDF) membrane (Bio-Rad), the proteins were probed with the noted primary antibodies and the corresponding HRP-conjugated secondary antibodies. Finally, the Western blot signals were detected using the ECL detection kit (Millipore) and quantified with Quantity One (Bio-Rad).
Immunofluorescence Staining and Image Processing-To analyze the localization of mitochondria in neuronal synaptic sites, neurons at 8 -10 DIV were transfected with mitoDsRed2 to label mitochondria. Two to 4 days after transfection, neurons were fixed with 4% paraformaldehyde, permeabilized with 0.4% Triton X-100, and then blocked with 10% goat serum. Presynaptic sites were labeled with SV2 antibody (recognizing synaptic vesicles) followed by Alexa Fluor 488-conjugated or Cy5conjugated goat anti-mouse IgG. The dendrites and axons can be clearly distinguished; mitoDsRed2-labeled axons were shown as long thin processes, whereas the dendrites were far shorter and thicker and had a different branching morphology. Axonal mitochondrial localization at presynaptic sites was measured as follows: circular regions of interest with a radius of 1 m from the SV2 puncta were placed around each identified puncta, and mitochondria occupying at least half of the radius of a region of interest in the same axon were considered to be colocalized with SV2 puncta. Fluorescence images were acquired with a Nikon Eclipse TE 2000-U microscope. All of the images were collected using a 60ϫ oil immersion objective lens and processed with MetaMorph software (Universal Imaging Corp., West Chester, PA). All the images were acquired under the same conditions. To avoid errors due to subjective intention, an investigator blind to the experimental conditions analyzed the images using ImageJ.
To examine the surface expression of TrkB in neurons expressing Miro1 and Miro1 KK , hippocampal neurons were cotransfected with pEGFP-N1 and pcDNA3.1 FLAG-Miro1/ Miro1 KK at a ratio of 1:5. We confirmed that neurons expressing GFP also express Miro1 or Miro1 KK (data not shown). Two days after transfection, the neurons were fixed with 4% paraformaldehyde in PBS for 15 min. Then the neurons were rinsed three times in PBS and blocked in 10% goat serum for 1 h at room temperature. Surface TrkB was stained using rabbit anti-TrkB antibody (Millipore, 1:300), which recognizes the extracellular domain of TrkB, in blocking buffer for 1 h at room temperature. Surface TrkB was subsequently detected by incubation with Alexa Fluor 594-conjugated goat anti-rabbit IgG. Acquired images were analyzed using ImageJ software.
Live Imaging-Neurons were cultured on 35-mm glass bottom culture dishes at a density of 5 ϫ 10 5 /dish (MatTek, Ashland, MA). For image acquisition, dishes were placed in a liveimaging station mounted on a Nikon Eclipse TE 2000-U microscope in an atmosphere of 37°C and 5% CO 2 , and all the photos were collected using a 40ϫ oil immersion objective lens and processed by MetaMorph software. Live images were captured every 3 s for 5 min before and after BDNF or other inhib-itors were applied. Analyses were performed on 150 m of axon at least 50 m away from the cell body. Kymographs were generated with ImageJ software and show the axon along the x axis and time across the y axis. The length of the kymographs is 5 min, and time increases down the formed photos. Mitochondrial motility was assessed by counting the percentage of moving mitochondria during an imaging period (5 min unless stated otherwise). Mitochondrial velocity was measured using the MTrackJ plugin of ImageJ.
FM 1-43 Staining-FM 1-43 staining was carried out as described previously with a few modifications (25)(26)(27). For labeling active presynaptic boutons, the neurons were stimulated by superfusing with the hyperkalemic solution (31.5 mM NaCl, 90 mM KCl, 2 mM CaCl 2 , 2 mM MgCl 2 , 25 mM HEPES, 30 mM glucose) containing FM 1-43 (5 M) for 45 s. During the loading period, the dye was rapidly taken up into presynaptic terminals as the released vesicles underwent endocytosis. The extracellular solution was then switched to normal saline solution (119 mM NaCl, 2.5 mM KCl, 2 mM CaCl 2 , 2 mM MgCl 2 , 25 mM HEPES, 30 mM glucose). ADVASEP-7 (1 mM) was introduced for 60 s at 1 and 6 min for enhanced removal of the dye from the external medium. A schematic of the procedure is provided in Fig. 8A. An image was then taken to record the loading of FM 1-43 in the synaptic boutons (I). The cultures were then exposed to multiple 15-s applications of hyperkalemic bath solution (without FM 1-43 dye) to evoke repeated cycles of exocytosis, which facilitate the release of the dye trapped in the active vesicles. After 30 min of repeated cycles of exocytosis, another image (II) was taken. The difference in the number of FM 1-43-labeled boutons in the axonal regions between I and II was calculated to indicate the number of active presynaptic boutons.
To study BDNF-induced neurotransmitter release, after loading and unloading of the FM 1-43 dye, the cultures were exposed to human recombinant BDNF (50 ng/ml) for 15 min, and the procedure of staining and destaining was then repeated. A schematic of the procedure is provided in Fig. 8A. To measure the plasticity changes in presynaptic terminals stimulated by BDNF, we compared the differences in changes of FM 1-43 fluorescence intensity along the GFP-labeled axons before (I and II) and after treatment with BDNF (III and IV). To block possible recurrent excitation and induction of active-dependent plasticity, 40 M D-2-amino-5-phosphonovalerate (NMDA receptor antagonist) and 20 M 6-cyano-7-nitroquinoxaline-2,3-dione (the non-NMDA receptor antagonist) were included in the hyperkalemic solution. All images were taken using a Nikon Eclipse TE 2000-U microscope with 40ϫ oil immersion objective lens and analyzed with ImageJ software. The individual puncta were identified based on a fluorescence intensity that exceeded a threshold set above background and a diameter between 0.5 and 3 m. To avoid objective errors, all images were subjected to exactly the same acquisition and analysis conditions. The investigator was blind to the experimental conditions.
To study the effects of Miro1 on BDNF-induced synaptic function, we transfected neurons with pEGFP-N1 and pcDNA3.1 FLAG-WT Miro1 or pCDNA3.1-Miro1 KK in the proportion of 1:5 at 9 DIV, such that neurons expressing GFP also expressed WT Miro1 or Miro1 KK . The active synaptic boutons and the BDNF-induced neurotransmitter release were examined as described above at 12-14 DIV.
Mitochondrial Membrane Potential Assay-Mitochondrial membrane potential was measured in live neurons with the use of the fluorescent probe JC-1. With higher mitochondrial membrane potentials, JC-1 forms red fluorescent aggregates in mitochondria, with a broad excitation spectrum and a maximum emission at 590 nm. However, at lower membrane potentials, JC-1 exists as fluorescent green monomers with emission at 527 nm when excited at 490 nm. Neurons at 12-14 DIV were loaded with 2 g/ml JC-1, incubated at 37°C for 30 min, then washed twice with medium, and imaged in conditioned culture medium. JC-1 fluorescent signals in live neurons were collected using a Carl Zeiss LSM-780 confocal microscope with a Plan Apochromat 63ϫ oil immersion objective lens (NA 1.40) (Microstructural Platform of Shandong University). JC-1 was excited at 488 nm (green) and 561 nm (red), and its corresponding fluorescent emission was collected using bandpass (BP 490 -550 nm) and long path emission filters (LP 590 nm) for monomer and aggregate signals, respectively. Three to four fields from each culture dish were selected and imaged. For mitochondrial membrane potential measurement, the JC-1 fluorescent intensity, defined as the red to green ratio, was calculated using ImageJ.
ATP Production Assay-The neuronal ATP level was measured using the ATP assay kit from Beyotime Co. Briefly, the lysate of cultured neurons was collected and centrifuged at 10,000 ϫ g for 10 min at 4°C, and then the ATP level in the supernatant was examined by mixing the supernatant with luciferase reagent. The emitted light was measured using a microplate luminometer. The protein level was determined using a BCA protein assay kit (Thermo Scientific, Rockford, IL). The ATP level was normalized to the protein concentration of each sample.
Intracellular Ca 2ϩ Measurement-The changes of intracellular calcium concentration were examined by Fluo4-AM staining. Neurons were incubated with 2 M Fluo4-AM in Neurobasal medium for 30 min at 37°C, followed by incubation for another 30 min after rinsing with Hanks' balanced saline solution. Ca 2ϩ imaging was performed using a spinning disk confocal microscope (Cell Observer SD) with a 40ϫ oil immersion objective (Microstructural Platform of Shandong University). Cells were excited with a laser at 488 nm, and the intensity of the fluorescence was collected at 525 nm as the Fluo4-AM signal. The images were collected before and after BDNF treatment, and the fluorescent intensity was analyzed with ImageJ. To measure the calcium in the axonal terminals, neurons were transfected with red fluorescent protein vector to label the neurites, and the axons and dendrites could be distinguished by their morphology. The axon terminals could be found along the axon extending from the soma, and the Fluo4-AM-labeled axonal tips were analyzed before and after BDNF application.
Statistical Analysis-The results of more than three independent experiments were compiled, and they were analyzed using Student's t test or one-way analysis of variance followed by post hoc tests. Data are shown as the mean Ϯ S.E. Statistical significance was determined with *, p Ͻ 0.05; **, p Ͻ 0.01.

BDNF Treatment Induces Reduced Mitochondrial Motility in
Axons-To measure axonal mitochondrial movement, we carried out live cell microscopy of rat neurons expressing mitoDsRed2 (labeling mitochondria without affecting their movement). Cultured hippocampal neurons at 12-14 DIV were bath-incubated with 50 ng/ml BDNF for 15 min, and mitochondrial motility in axons was examined. The dendrites and axons can be distinguished; the mitoDsRed2-labeled axons were visible as long thin processes, although the dendrites were far shorter and thicker and had a different branching morphology. Kymographs were created with the axon along the x axis and the time along the y axis. Stationary and moving mitochondria can be seen as straight lines and diagonal lines, respectively. As shown in Fig. 1, A-C, BDNF treatment significantly reduced axonal mitochondrial motility for both anterograde and retrograde movement. At basal status, the percentage of moving mitochondria in axons was ϳ13.3% for anterograde and 12.9% for retrograde over a 5-min period, similar to a previous report (13). However, BDNF treatment caused a significant decrease in the percentage of mobile mitochondria in axons for both directions (Fig. 1, A-C). Although the percentage of moving mitochondria was decreased, the velocity of the moving mitochondria was not affected by BDNF treatment (Fig. 1D). Thus, BDNF treatment increased the fraction of mitochondria in the stationary phase rather than altering the speed of moving mitochondria. To determine whether BDNF could affect the movement of organelles other than mitochondria, we also investigated the effect of BDNF on the motility of endosomes, which were labeled by GFP-tagged Rab5. No changes in the percentage of moving endosomes were observed upon BDNF treatment ( Fig. 1, E-G), indicating that the effect of BDNF on mitochondrial motility is specific.
Because it has been demonstrated that BDNF can increase mitochondrial oxidative coupling and the efficiency of the ATP synthesis (18,28), we further investigated whether BDNF-induced mitochondrial stopping was associated with mitochondrial membrane potential and ATP production. Mitochondrial membrane potential was examined with JC-1 staining. As shown in Fig. 2, A and B, no significant changes were observed after BDNF treatment for 15 min. However, extending the BDNF treatment to 24 and 72 h significantly increased the mitochondrial membrane potential. Creatine has been reported to stimulate mitochondrial respiration and membrane potential (2,29). The neurons were treated with creatine for 18 h as a positive control (Fig. 2, A and B). ATP production was not affected by 15 min and 24 h of BDNF treatment, but lengthening the BDNF treatment to 72 h resulted in a slight but significant increase in ATP production (Fig. 2C). Thus, the above results suggest that BDNF-induced decreased mitochondrial motility precedes increases in mitochondrial functions such as mitochondrial membrane potential and ATP production.
Activation of PI3K and PLC␥ Signaling Pathways Is Required for BDNF-induced Mitochondrial Stopping-Because BDNF treatment induced more mitochondria into the stationary pool, we next investigated the mechanisms underlying this phenomenon. We found that K252a, a widely used inhibitor of Trk kinases, could completely block BDNF-induced mitochondrial stopping regardless of the movement direction, although the p75 inhibitor tat-PEP5 had no effect (Fig. 3A; the retrogrademoving mitochondria result is similar to anterograde-moving mitochondria and is not shown), indicating that the effect of BDNF on mitochondrial movement relies on TrkB kinase activity. Next, we used pharmacological inhibitors to investigate which signaling pathway downstream of TrkB activation was involved. A specific PLC␥ inhibitor, U73122, abolished BDNFinduced mitochondrial stopping, whereas its inactive derivative U73343 had no effect (Fig. 3B). Pretreatment with the PI3K inhibitor LY294002 also abolished BDNF's effects on mitochondrial motility, whereas inactivating the MAPK with U0126 had no effect (Fig. 3B). These results suggest that activation of the PI3K and PLC␥ signaling pathways is required for BDNFinduced mitochondrial stopping. A direct downstream effector of PLC␥ is inositol 1,4,5-triphosphate (IP 3 ), which induces internal Ca 2ϩ release by activating the IP3 receptor (IP 3 R). Blocking IP 3 R with xestospongin C or 2APB abolished BDNFinduced mitochondrial stopping (Fig. 3B). Another downstream effector of PLC␥ is diacylglycerol (DAG), which then activates protein kinase C. The membrane-permeable DAG analog 1-oleoyl-2-acetyl-sn-glycerol (OAG) could not induce mitochondrial motility arrest, but when OAG was applied together with BDNF, a reduced percentage of moving mitochondria was observed, suggesting that DAG is not involved in BDNF-induced mitochondrial motility arrest (Fig. 3B). As observed in neurons pretreated with K252a, both anterograde and retrograde movements were blocked by inhibitors of PI3K, PLC␥, and IP 3 R. The figures show only the results for anterograde-moving mitochondria. Thus, the results suggest that BDNF-induced mitochondrial stopping depends on PI3K and PLC␥-IP 3 R activation but not on the PLC␥-DAG signaling pathway. Elevation of Intracellular Ca 2ϩ Is Necessary for BDNF-induced Mitochondrial Stopping-It has been known that BDNF induces elevation of intracellular Ca 2ϩ (19,30), and Ca 2ϩ has been reported to be an important regulator of mitochondrial movement (13,16). We hypothesize that Ca 2ϩ plays an important role in BDNF-regulated mitochondrial motility. Indeed, Fluo4-AM imaging revealed that BDNF induces significant elevation of intracellular Ca 2ϩ , which reached high levels by 15 min and could remain at this level even after 30 min (Fig. 4A). The increased intracellular Ca 2ϩ concentration could be suppressed by pretreatment with K252a, U73122, LY294002, or xestospongin C (Xest C) (TrkB, PLC␥, PI3K, and IP 3 R inhibitors, respectively) (Fig. 4B), which is consistent with our previous result that these inhibitors could block the effect of BDNF on mitochondrial motility. The above result also suggests that IP 3 R-regulated Ca 2ϩ release from the internal store is required for BDNF-induced intracellular Ca 2ϩ elevation. The membrane-permeable Ca 2ϩ buffer BAPTA-AM and the extracellular Ca 2ϩ depletion reagent EGTA also blocked BDNF-induced intracellular Ca 2ϩ elevation (Fig. 4B), suggesting Ca 2ϩ influx is required for BDNF-induced intracellular Ca 2ϩ elevation. Therefore, BDNF-induced elevation of intracellular Ca 2ϩ depends on both IP 3 R-mediated intracellular Ca 2ϩ release from internal stores and extracellular Ca 2ϩ influx. Next, we investigated whether the intracellular Ca 2ϩ indeed mediates the BDNF-induced stopping of mitochondria and which Ca 2ϩ channel is involved in this process. Preincubation with BAPTA-AM (2 M) or EGTA (2 mM), which blocked the BDNF-induced intracellular Ca 2ϩ elevation, abolished the effect of BDNF on mitochondrial mobility in both directions (Fig. 4C, retrograde movement not shown). Therefore, the elevated Ca 2ϩ is essential for the reduction in mitochondrial motility induced by BDNF. Pretreatment with nifedipine or -conotoxin MVIIC, which are inhibitors of L-and P/Q-type voltage-dependent Ca 2ϩ channels, respectively, had no effects on BDNF-induced mitochondrial stopping. Applying a higher concentration of ryanodine (100 M) to block ryanodine-sensitive internal Ca 2ϩ channels also showed no effect on BDNF-FIGURE 2. Transient BDNF treatment has no effect on mitochondrial membrane potential or ATP production. A and B, representative images of JC-1 staining of control (Con) neurons and neurons treated with BDNF or creatine (A). Scale bar, 20 m. Mitochondrial membrane potential was not changed upon BDNF treatment for 15 min, but it increased significantly after BDNF treatment for 72 h. Creatine treatment for 18 h was used as a positive control (B). C, no changes in ATP production were observed after transient BDNF treatment, but ATP production increased significantly after BDNF treatment for 72 h. Values are shown as the mean Ϯ S.E., **, p Ͻ 0.01, Student's t test.
induced mitochondrial stopping (Fig. 4C). Thus, voltage-dependent Ca 2ϩ channels and internal ryanodine channels are unlikely to contribute to BDNF-induced mitochondrial stopping.
Multiple findings have demonstrated that TRPC channels are Ca 2ϩ -permeant channels activated by BDNF (19,30,31); therefore, we next investigated whether they are involved in BDNF-evoked stopping of mitochondria. When primary hippocampal neurons were pretreated with SKF-96365 (5 M), a known nonspecific inhibitor of transient receptor potential channels, BDNF-induced mitochondrial stopping in both directions was completely abolished (Fig. 5, A-E, quantification of anterograde-moving mitochondria is shown), suggesting that TRPC channels participate in BDNF-induced mitochondrial stopping. Among the TRPC channels, TRPC3, TRPC5, and TRPC6 have been widely studied in neuronal systems (19,30,(32)(33)(34). We next examined which TRPC channel contributed to the BDNF-induced mitochondrial stopping. TRPC3, TRPC5, or TRPC6 expression was knocked down by specific short interfering RNAs (siRNAs), and the knockdown efficiency was examined in PC12 cells by real time PCR (Fig. 5F). The effect of TPRC3, -5, or -6 on BDNF-regulated mitochondrial motility was then investigated. Introducing either TRPC3 or TRPC6 siRNAs in neurons abolished BDNF's effect on mitochondrial motility, but TRPC5 siRNA had no effect (Fig. 5G). To exclude off-target effects of the siRNAs, dominant-negative (DN) constructs of TRPC3, TRPC5, or TRPC6 labeled with green fluorescent protein (GFP) were transfected into hippocampal neurons. Transfection of DN-TRPC3 or DN-TRPC6, but not DN-TRPC5, abolished the effect of BDNF on mitochondrial motility (Fig. 5G), further confirming that TRPC3 and TRPC6 are essential in BDNF-induced mitochondrial stopping. Furthermore, BDNF-induced elevation of intracellular Ca 2ϩ was also blocked in neurons with TRPC3 or TRPC6 but not TRPC5 knockdown (Fig. 5H).
Miro1 Mediates BDNF-induced Mitochondrial Motility Arrest-The above results demonstrate that TRPC3 and TRPC6 are essential for BDNF-induced intracellular Ca 2ϩ elevation and mitochondrial stopping. We next investigated how the intracellular Ca 2ϩ regulates mitochondrial motility in response to BDNF. Recently, several studies have demonstrated that Miro1, a mitochondrial outer membrane protein, acts as a Ca 2ϩ sensor to regulate mitochondrial motility (13,16). We hypothesized that Miro1 might also play a role in BDNF-regulated mitochondrial movement. To address this question, we transfected wild type (WT) or mutant Miro1 (Miro1 KK , EFhand mutation resulting in the loss of Ca 2ϩ binding) into cultured hippocampal neurons and analyzed the effects of BDNF on mitochondrial motility in the successfully transfected neurons. Consistent with previous reports, overexpression of either WT or mutant Miro1 caused a significant increase in the percentage of moving mitochondria (Fig. 6, A-G) (13). In neurons expressing the control vector or WT Miro1, BDNF reduced the percentage of anterograde-moving mitochondria by 41.1 Ϯ 4.6% and 45.6 Ϯ 5.0%, respectively (Fig. 6, A-D and G; quantification of retrograde-moving mitochondria was similar and is not shown). However, in neurons expressing Miro1 KK , BDNF lost the ability to cause mitochondria to stop (Fig. 6, E-G), which suggests that Miro1 mediates BDNF-induced mitochondrial stopping by binding with Ca 2ϩ .
Two conflicting models have been proposed for Miro1-Ca 2ϩ -mediated mitochondrial transport. The first proposes that upon Miro1 binding with Ca 2ϩ , KIF5 is released from the Miro1-KIF5 complex on the mitochondrial surface, thereby halting mitochondrial movement (13). The other model proposes that elevated Ca 2ϩ prevents the kinesin-Miro1 complex from binding to microtubules but leaves the complex on the mitochondrial surface, causing the arrest of mitochondrial movement (16). Because BDNF-regulated mitochondrial movement is also dependent on the Miro1-Ca 2ϩ pathway, we investigated which model provided the best fit. We analyzed KIF5 and Miro1 protein levels in cytosolic and mitochondrial fractions from BDNF-or vehicle-treated neurons. The Western blot results revealed that treatment with BDNF for 15 min had no effect on the Miro1 and KIF5 protein levels in either the cytosolic or mitochondrial fractions (Fig. 6, H and I), suggesting that BDNF-induced mitochondrial stopping may be due to the dissociation of the KIF5-Miro1 complex from microtubules rather than to the dissociation of KIF5 from mitochondria.
To confirm the mechanism of the Miro1-dependent BDNFinduced mitochondrial motility arrest, the effects of Miro1/ Miro1 KK on intracellular Ca 2ϩ concentration were examined. Overexpressing Miro1 or Miro1 KK did not change the basal intracellular Ca 2ϩ levels in somas. Moreover, in neurons expressing either Miro1 or Miro1 KK , BDNF could still cause intracellular Ca 2ϩ increase to a degree similar to that in control neurons (Fig. 6J). The Ca 2ϩ changes in axon terminals were similar to those in somas, and it is interestingly to note that BDNF induces higher calcium concentration at the axon terminals compared with that in the neighboring regions as shown in Fig. 6K. The surface TrkB levels in neurons expressing the empty vector, WT Miro1, and Miro1 KK were also examined, and they did not show significant differences (Fig. 6, L and M). These results indicate that Miro1 and Miro1 KK did not affect basal intracellular Ca 2ϩ or BDNF-induced intracellular Ca 2ϩ increase nor did they alter surface TrkB levels, suggesting that Miro1's effects on BDNF-regulated mitochondrial motility depend on its ability to bind Ca 2ϩ .

BDNF-induced Increase of Mitochondria in the Stationary Pool Contributes to the Effect of BDNF on Neurotransmitter
Release-Because mitochondria provide energy and calcium buffering functions that are important for neuronal activities such as synaptic transmission, we asked whether BDNF-induced arrest of mitochondrial movement played a role in BDNF-mediated synaptic transmission. We first examined the number and size of synaptic vesicles with SV2 immunostaining. BDNF treatment for 15 min did not change the number and size of synaptic vesicles compared with controls (Fig. 7, A and B). To determine whether BDNF-induced mitochondrial stopping led to altered mitochondrial distribution, neurons expressing mitoDsRed2 were fixed after BDNF treatment for 15 min and then stained with the SV2 antibody. We analyzed mitochondrial localization at presynaptic sites. Axonal mitochondria localized within 1 m from the SV2 puncta were regarded as colocalized with synaptic vesicles. After BDNF treatment, we observed a significantly increased percentage of mitochondria colocalized with SV2 puncta (Fig. 7, C and D), suggesting that BDNF treatment induces the accumulation of more mitochondria at presynaptic sites. Because Miro1 has been shown to mediate mitochondrial motility in response to BDNF, we asked whether BDNF-induced mitochondrial recruitment to presynaptic site depends on Miro1. We examined the effect of BDNF on the colocalization between mitochondria and SV2 in WT Miro1-or Miro1 KK -expressing neurons. As shown in Fig. 7D, WT Miro1 had no effect, but the Miro1 KK mutant blocked the increased colocalization between mitochondrial and SV2 triggered by BDNF. The above results indicate that BDNF could induce more mitochondrial docking at presynaptic sites and that this effect is dependent on Miro1.
We next investigated whether BDNF-induced Miro1-dependent mitochondrial presynaptic docking is associated with BDNF-modulated neurotransmitter release. Previous studies have reported that BDNF could rapidly enhance evoked synaptic transmission and spontaneous glutamate release (27). We first investigated whether WT Miro1 or Miro1 KK affects the basal number of active synaptic boutons, which is an indicator of the number of neurotransmitter release sites. FM 1-43 was used to label the presynaptic sites, as described under "Experimental Procedures." The number of active synaptic boutons per unit length in successfully transfected neurons was quantified. The neurons expressing WT Miro1 or Miro1 KK had similar numbers of active synaptic boutons (Fig. 8, B and C), suggesting that the Miro1 or mutant Miro1 does not affect the basal levels of active synaptic vesicles. We next examined whether Miro1 is involved in BDNF-enhanced neurotransmitter release. Consistent with a previous report, BDNF increased K ϩ -evoked neurotransmitter release in control neurons (Fig. 8, D and E) (35). This effect was blocked by the Miro1 KK mutant but not by WT Miro1 (Fig. 8E), suggesting that Miro1's activity as a Ca 2ϩ sensor mediates BDNF-enhanced neurotransmitter release. Together, these data indicate that Miro1 mediates BDNF-induced mitochondrial presynaptic accumulation and BDNF-enhanced synaptic transmission.

DISCUSSION
This study shows that BDNF increased the proportion of stationary mitochondria, which is dependent on the elevation of intracellular Ca 2ϩ induced by the activation of PI3K and PLC␥-IP 3 pathways and TRPC3/6 channels. Miro1, functioning as a Ca 2ϩ sensor, dissociates KIF5 from microtubules, resulting in more mitochondria recruited into presynaptic sites, and thereby enhancing synaptic transmission.
Our studies provide several insights into the molecular mechanisms and significance of BDNF-regulated mitochondrial motility. To our knowledge, this is the first report to show that BDNF specifically regulates mitochondrial motility. We showed that BDNF-induced mitochondrial stopping is dependent on TrkB receptor tyrosine kinase activity and its downstream PI3K and PLC␥ signaling pathways (Figs. 3 and 4). Previously, Chada and Hollenbeck (8,36) demonstrated that focal stimulation with NGF-coated beads causes a local accumulation of mitochondria in axons and that this is dependent on the activation of PI3K. Given the high structural similarity between NGF and BDNF, it is likely that neurotrophic factors share common signaling pathways to regulate mitochondrial trafficking.
In neurons, the elevation of intracellular Ca 2ϩ through voltage-dependent Ca 2ϩ channels or NMDA receptors has been reported to arrest microtubule-based mitochondrial movement (11,12,37). We showed that, following activation of PI3K and PLC␥, the intracellular Ca 2ϩ concentration was significantly increased by a mechanism dependent on IP 3 and TRPC3/6 channels. Indeed, it has been reported that the activation of PLC␥-IP 3 by BDNF leads to Ca 2ϩ release from internal stores, which then activates store-operated TRPC channels and allows a sustained extracellular Ca 2ϩ influx (30). PI3K has also been reported to induce accumulation of phosphatidylinositol 3,4,5-trisphosphate as a signaling node to asymmetric activation of Akt, leading to potentiation of TRPC channels in the membrane and allowing the influx of Ca 2ϩ (38). It is noteworthy that both PI3K and PLC␥ are required for BDNF-induced Ca 2ϩ elevation and mitochondrial stopping; blocking either of the pathways could abolish BDNF's effects. Similarly, BDNF-induced Ca 2ϩ -dependent growth cone turning requires both PI3K and PLC␥ activation (19). How these two molecules or pathways jointly regulate BDNF's function requires further investigation.
TRPC channels are a family of nonselective Ca 2ϩ -permeable cation channels. They can form functional heteromeric or homomeric channels (39,40) and participate in diverse neuronal functions, including cell survival, spine formation, neurite growth, and growth cone turning (19,30,32,33). TRPC3 and TRPC6 channels have been demonstrated to localize to synapses (31,41), and TRPC3-and TRPC6-mediated Ca 2ϩ influx is critical for BDNF-induced growth cone guidance and survival of cerebellar granule cells (19,32). In our study, we demonstrate that TRPC3 and TRPC6, but not TRPC5, also play critical roles in BDNF-regulated mitochondrial docking (Fig. 5). It has been speculated that by allowing different patterns of Ca 2ϩ influx, diverse TRPC channels may perform distinct regulatory functions in neurons. Indeed, a recent study demonstrated that TRPC5-mediated Ca 2ϩ elevation activates CaMKII␣, whereas TRPC6-mediated Ca 2ϩ elevation activates CaMKIV, resulting in inhibition and promotion of neuronal dendritic growth, respectively (42). In our experiments, it is likely that at presynaptic sites TRPC3 forms heteromers with TRPC6 and is activated by BDNF, allowing Ca 2ϩ influx and resulting in higher intracellular Ca 2ϩ at the synaptic sites, which leads to more mitochondrial docking at synaptic sites regardless of anterograde or retrograde movement of mitochondria. Interestingly, Feng et al. (43) recently reported that TRPC3 localizes not only to plasma membrane but also to the inner mitochondrial membrane and that it contributes to mitochondrial Ca 2ϩ uptake. One study has suggested mitochondrial matrix Ca 2ϩ content to be an intrinsic signal for modulating mitochondrial motility in neurons (44). Thus, we speculate that mitochondrial TRPC3 channels may also contribute to BDNF-mediated mitochondrial motility by mediating mitochondrial Ca 2ϩ levels.
Miro1 is required for BDNF-induced mitochondrial docking at synaptic sites. Miro1 is a mitochondrial outer membrane protein with two EF-hand domains, which bind Ca 2ϩ , and several recent studies have demonstrated that KIF5, Miro1, and TRAK/Milton form a complex that mediates Ca 2ϩ -dependent regulation of mitochondrial motility (13,16,45). We demonstrated that Miro1, in its role as a Ca 2ϩ sensor, regulates BDNFinduced mitochondria recruitment to presynaptic sites. A study using Drosophila has similarly demonstrated that Miro is required for controlling the anterograde transport of mitochondria and their proper distribution within nerve terminals (14). The fact that BDNF-enhanced synaptic transmission could be blocked by mutant Miro1 may be due to the lack of synaptic mitochondria. Two different models of Miro1 regulation of mitochondrial motility have been proposed. One model suggests that elevated Ca 2ϩ causes dissociation of KIF5 from the Miro1-motor complex (13); the other suggests that the KIF5-Miro1 complex dissociates from microtubules in response to elevated Ca 2ϩ (16). Our data favor the model that the KIF5-Miro1 complex dissociates from microtubules upon BDNF treatment, thus causing reduced mitochondrial motility.
It is worth noting that an intact actin cytoskeleton is required for mitochondrial docking induced by NGF (8). We speculate that in addition to Miro1-mediated microtubule-based movement, actin-based mitochondrial movement is also involved in BDNF-induced mitochondrial docking, and myosin motors might be involved in regulating the interaction between mitochondria and actin at synaptic sites upon BDNF treatment.
Finally, our data suggest that the recruitment of more mitochondria for docking at presynaptic sites contributes to BDNFenhanced neurotransmitter release. BDNF-enhanced neurotransmitter release has been previously reported (46,47), and it has been demonstrated that synapsin is phosphorylated by FIGURE 7. BDNF induces mitochondria recruited to presynaptic site dependent on Miro1 binding with Ca 2؉ . A and B, transient application of BDNF had no effect on the size and density of SV2-labeled puncta. C, representative images of presynaptic mitochondria (in red) and SV2-labeled puncta (in green) in neuronal axons with empty vector, WT Miro1, and Miro1 KK expression before and after BDNF treatment. The arrows indicate the colocalization between mitochondria and SV2. Scale bar, 5 m. D, colocalization index (mitochondria colocalized with SV2/total mitochondria) in neurons expressing empty vector, WT Miro1, or Miro1 KK before and after BDNF treatment. Values are shown as the mean Ϯ S.E. from three independent experiments, **, p Ͻ 0.01, Student's t test. Con, control.
MAPK and that Myo6 and GIPC1 are associated with BDNFenhanced synaptic transmission (27,48). Recently, Amaral and Pozzo-Miller (49) reported that BDNF could enhance spontaneous quantal transmitter release through a Ca 2ϩ -dependent mechanism. We provide evidence that Miro1's ability to bind Ca 2ϩ is required for BDNF-enhanced neurotransmitter release. This suggests that Miro1 is one of the downstream effectors of calcium, regulating BDNF-enhanced synaptic transmission most likely via its mediation of mitochondrial docking at presynaptic sites. Such a mechanism is consistent with previous findings that mutant Miro reduces the population of presynaptic mitochondria and causes impaired neurotransmitter release (14). Several studies have demonstrated that mitochondria are required at synapses to help maintain neurotransmission by buffering Ca 2ϩ (1,14). It has also been reported that BDNFenhanced neurotransmitter release is due to increases in the number of vesicles docked at the active zone of individual synapses (50), which is partly dependent on mitochondrial ATP production (51). One recent study has shown that presynaptic mitochondria are essential determinants of synaptic vesicle exocytosis, primarily via ATP synthesis (52). In the case of BDNF-enhanced neurotransmitter release, although an overall increase in ATP production was not observed (Fig. 2C), we speculate that there is an increase in ATP production at presynaptic sites due to more mitochondria localized at the area, thus facilitating synaptic transmission.
In conclusion, our study reveals that BDNF could increase the stationary mitochondrial pool in axons and that this is mediated by elevation of intracellular Ca 2ϩ . Furthermore, the BDNF-induced mitochondrial motility arrest causes more mitochondria to be recruited at presynaptic sites, which plays an important role in BDNF enhanced neurotransmitter release. These findings might provide insights not only into the mechanistic link between BDNF-dependent mitochondrial motility and BDNF-mediated synaptic transmission but also more generally into the fields of mitochondrial transport and neurotransmission.