Address correspondence and reprint requests to Dr Tadahiro Numakawa, Division of Protein Biosynthesis, Institute for Protein Research, Osaka University, Suita, 3–2 Yamadaoka, Osaka 565–0871, Japan. E-mail: firstname.lastname@example.org
Brain-derived neurotrophic factor (BDNF) has been reported to play an important role in neuronal plasticity. In this study, we examined the effect of BDNF on an activity-dependent synaptic function in an acute phase. First, we found that short-term treatment (10 min) with BDNF enhanced depolarization-evoked glutamate release in cultured cortical neurons. The enhancement diminished gradually according to the length of BDNF treatment. The BDNF-enhanced release did not require the synthesis of protein and mRNA. Both tetanus toxin and bafilomycin abolished the depolarization-evoked glutamate release with or without BDNF, indicating that BDNF acted via an exocytotic pathway. Next, we investigated the effect of BDNF on intracellular Ca2+. BDNF potentiated the increase in intracellular Ca2+ induced by depolarization. The Ca2+ was derived from intracellular stores, because thapsigargin completely inhibited the potentiation. Furthermore, both thapsigargin and xestospongin C inhibited the effect of BDNF. These results suggested that the release of Ca2+ from intracellular stores mediated by the IP3 receptor was involved in the BDNF-enhanced glutamate release. Last, it was revealed that the enhancement of glutamate release by BDNF was dependent on the TrkB-PLC-γ pathway. These results clearly demonstrate that short-term treatment with BDNF enhances an exocytotic pathway by potentiating the accumulation of intracellular Ca2+ through intracellular stores.
Neurotrophins were originally identified as a family of neurotrophic factors that regulate neuronal survival and differentiation in the CNS (Barde 1990; Lindholm et al. 1993; Davies 1994). It was reported that brain-derived neurotrophic factor (BDNF), one of the neurotrophins, played an important role in excitatory transmission (Lo 1995; Thoenen 1995; Lu and Figurov 1997). For example, BDNF-knockout mice showed a severe impairment of hippocampal long-term potentiation (LTP) (Korte et al. 1995; Patterson et al. 1996). The impairment was restored by incubation with BDNF for a few hours (Patterson et al. 1996) or by gene transfer (Korte et al. 1995). BDNF-deficient mice showed a significant reduction in the number of vesicles docked in the presynaptic active zone of hippocampal neurons (Pozzo-Miller et al. 1999). In cultured cortical neurons, BDNF treatment for several days enhanced the depolarization-evoked release of glutamate (Takei et al. 1997), and increased the amplitude of postsynaptic currents (Sherwood and Lo 1999). These results suggested that BDNF potentiated both pre- and postsynaptic transmission in the chronic phase.
We previously demonstrated that BDNF induced an accumulation of intracellular Ca2+ (Sakai et al. 1997; Numakawa et al. 2001). We also reported that BDNF directly induced a transient release of glutamate and aspartate from cultured cortical and cerebellar neurons through a non-exocytotic pathway in a manner dependent on intracellular Ca2+ (Takei et al. 1998; Numakawa et al. 1999, 2000). These results suggested that BDNF directly influenced excitatory transmission. Here, we investigated the effect of BDNF on an activity-dependent synaptic function. We found that BDNF treatment for 10 min enhanced the depolarization-evoked glutamate release via exocytosis in cultured cortical neurons, and that the enhancement required the activation of the TrkB-PLC-γ pathway. Furthermore, it was revealed that the augmentation of intracellular Ca2+ through intracellular stores was essential for the potentiation of the depolarization-evoked release mediated by BDNF. Our results demonstrated that BDNF enhanced an exocytotic mechanism in the acute phase, in which Ca2+ derived from intracellular stores was involved.
Materials and methods
Primary dissociated cultures were prepared from the cerebral cortex of P2 rats (Wister ST; SLC, Shizuoka, Japan) as reported previously (Hatanaka et al. 1988; Numakawa et al. 2001). Briefly, cells were gently dissociated with a plastic pipette after digestion with papain (90 U/mL, Worthington Biochemical Corp., Lakewood, NJ, USA) at 37°C. The dissociated cells were plated at a final density of 5 × 105 cells/cm2 on polyethyleneimine-coated 12- and 24-well plates (3.8 and 2 cm2 surface area/well, respectively; Corning, Acton, MA, USA). For the Ca2+ imaging experiment, cells were plated on cover glass, as described later. The culture medium consisted of 5% precolostrum newborn calf serum, 5% heat-inactivated horse serum, 1% rat serum and 89% of a 1 : 1 mixture of Dulbecco's modified Eagle's medium (DMEM) and Ham's F-12 medium containing 15 mm HEPES buffer, pH 7.4, 30 nm Na2SeO3 and 1.9 mg/mL NaHCO3. After 24 h, the medium was replaced. The cultures were maintained for 5–7 days.
Cells were stained with anti-MAP2 (a gift from Dr H. Murofushi, The University of Tokyo), and anti-TrkB (Santa Cruz Biotechnology, Santa Cruz, CA, USA) antibodies. Briefly, cells were fixed in 4% paraformaldehyde containing 0.05% polyoxyethylene (10) octylphenyl ether (Triton X-100) at room temperature (25°C) for 20 min, then incubated with anti-MAP2 antibody diluted 1 : 5000 with a 0.01% Triton X-100 solution overnight at 4°C. For staining with anti-TrkB antibody, cells were fixed in 4% paraformaldehyde at room temperature for 20 min and then incubated with anti-TrkB antibody diluted 1 : 1000 with a solution containing 0.05% Triton X-100. For visualizing, we used a Vectastain ABC kit (Vector Laboratories, Burlingame, CA, USA) and 0.02% (w/v) 3,3′-diaminobenzidine tetrahydrochloride and 0.1% (w/v) (NH4)Ni(SO4)2 dissolved in 0.05 m Tris-HCl buffer, pH 7.6, containing 0.01% (v/v) H2O2.
Detection of amino acid neurotransmitters
The amounts of amino acids released into the buffer from the cultured neurons were measured as described previously (Takei et al. 1998; Numakawa et al. 1999, 2001). Briefly, the amounts released into the assay buffer (modified HEPES-buffered Krebs' Ringer solution: KRH buffer containing 130 mm NaCl, 5 mm KCl, 1.2 mm NaH2PO4, 1.8 mm CaCl2, 10 mm glucose, 1% bovine serum albumin, 25 mm HEPES, pH 7.4) from the cultured cortical neurons were measured by high performance liquid chromatography (HPLC; BAS, Tokyo, Japan) and fluorescence detector (CMA280: BAS, Tokyo). Before samples were collected, the cultured medium was replaced with serum-free medium for 3–5 h. Then, cultured neurons were washed three times with KRH buffer. The Ca2+-free solution was prepared without CaCl2 and with EGTA (3 mm). The Na+-free solution was made by replacing NaCl with LiCl. Before performing the experiment in Ca2+- or Na+-free solution, cultured neurons were washed three times with the same solution. Fractions were collected by the batch method every 1 min, into tubes on ice, and filtered to remove cell debris with 0.22 µm membranes. Amino acids in the assay buffer were treated with o-phthalaldehyde and 2-mercaptoethanol. The chemical reaction was run for 5 min at 12°C. Samples were injected into the HPLC system and analyzed using a fluorescence monitor (excitation wavelength, 340 nm emission wavelength, 445 nm). Cycloheximide (CHX; 1 µm) (Sigma, St Louis, MO, USA) and actinomycin D (Act D; 0.1 µm) (Sigma) were applied to the cultured neurons for 60 and 30 min before the application of BDNF, respectively. K252a (200 nm) (Alexis Corp., San Diego, CA, USA) was applied to the neurons for 30 min before the application of BDNF. Tetanus toxin (TeNT; 10 nm) (List Biological Laboratories Inc., Campbell, CA, USA) and bafilomycin (Baf; 125 nm) (Sigma) were applied to the cells for 4 h and 1 h before the BDNF, respectively. Cadmium chloride (CdCl2; 100 µm) (Sigma) and tetrodotoxin (TTX; 0.1 µm) (Latoxan, Valence, France) were applied to the cells for 30 min before the application of BDNF. BAPTA-AM (30 nm) (Research Biochemicals Inc., Natick, MA, USA), thapsigargin (Thap; 100 µm) (Research Biochemicals) and xestospongin C (Xest C; 1 µm) (Calbiochem, Darmstadt, Germany) were applied to the cells for 30 min before the application of BDNF. U73122 (5 µm)(Sigma), U0126 (10 µm) (Promega, Madison, WI, USA) and LY294002 (10 µm) (Calbiochem) were applied to the cells 5 min before BDNF. When the effects of these drugs on depolarization-evoked glutamate release were examined, cultured neurons were washed three times and samples of the basal and depolarization-evoked release were collected using KRH buffer containing these drugs. Cells were treated with BDNF by bath application at a concentration of 100 ng/mL. Depolarization was induced with High-K+ solution (KCl; 50 mm) or 4-aminopyridine (4AP; 4 mm) (Sigma), a K+-channel blocker. The results of this assay were presented as follows; amount of glutamate induced by depolarization/basal release. We confirmed that the basal release was not affected by any drugs used in the assay. All experiments for amino acid analysis were performed three to seven times, and data were confirmed to be reproducible.
Imaging of intracellular Ca2+
Cells were cultured for 5–7 days on cover glasses (Matsunami Glass Ind. Ltd, Osaka, Japan) coated with polyethyleneimine. Before the loading with Fluo-3AM (Molecular Probes, Eugene, OR, USA), a Ca2+ indicator, cells were washed three times with KRH buffer, and incubated at 37°C for 1 h with 10 µm Fluo-3AM diluted in KRH buffer. The cells were washed three more times with KRH buffer, and placed on cover glasses on an inverted microscope (TMD-300, Nikon, Tokyo, Japan). The fluorescence intensity was monitored using a confocal laser microscope (RCM 8000, Nikon). Neurons were irradiated with an excitation blue light beam (488 nm) produced by an argon ion laser at a scanning frequency of 1/30 s. The emitted fluorescence was guided through a × 40 water-immersion objective to a pinhole diaphragm at 520 nm using a diachronic mirror. The fluorescence intensity was detected with one of two photo multipliers. The intensity of emission from each neuron targeted was scanned at 1/30 s intervals with a monitor video enhancer. Cells were treated with BDNF by bath application at a concentration of 100 ng/mL. Thapsigargin (100 µm) was added 30 min prior to the application of BDNF. Depolarization was induced with High-K+ solution (KCl; 50 mm) or 4AP (4 mm). Ca2+ imaging experiments were performed at room temperature.
BDNF enhanced depolarization-evoked glutamate release in cultured cortical neurons
To examine the possibility that an activity-dependent glutamate release is regulated by BDNF in short-term, we analyzed whether BDNF enhances the depolarization-evoked release of glutamate in cultured cortical neurons. Depolarization was induced using High-K+ solution (KCl; 50 mm) or 4-aminopyridine (4AP; 4 mm), known as a K+ channel blocker. BDNF treatment for 10 min enhanced both 4AP- and High-K+-induced glutamate release (Fig. 1a). Because the effect of BDNF on the 4AP-evoked release was greater, we used 4AP to induce depolarization in this study. Next, we investigated the time-course of the BDNF effect. Cultured neurons were treated with BDNF for 10, 30, 60, or 240 min. As shown in Fig. 1(b), BDNF significantly enhanced the 4AP-evoked release 2.45-fold after 10 min treatment. The BDNF-mediated enhancement of glutamate release, however, decreased gradually on longer treatment. The enhancement was maintained for 60 min after the application of BDNF. However, no enhancement was observed at 4 h. We have previously reported that the BDNF-enhanced glutamate release required up-regulation of synaptic proteins in the chronic phase (Takei et al. 1997). Then, we examined whether a novel protein and/or mRNA is required for the BDNF effect in the acute phase. Neither cycloheximide (CTX; 1 µm), a protein synthesis inhibitor, nor actinomycin D (Act D; 0.1 µm), an RNA synthesis inhibitor, influenced the effect (Figs 1c and d). These results suggested that BDNF enhanced the release of glutamate without any translation or transcription.
The effect of BDNF is through TrkB, a high affinity BDNF receptor. Then, we confirmed the expression of TrkB receptors in cultured cortical neurons. We performed immunocytochemistry using anti-TrkB antibody and anti-MAP2 antibody as a neuronal marker. Almost all of the cells in the culture were MAP2-positive (Fig. 2a, left). In sister cultured neurons, most cells expressed TrkB in somas and weakly in dendrites (Fig. 2a, right). Therefore, TrkB was expressed in cultured cortical neurons. Next, we examined the effect of K252a, a membrane-permeable Trk kinase inhibitor (Berg et al. 1992). The effect of BDNF on the release was completely inhibited by K252a (200 nm) (Fig. 2b). In addition, TrkB-IgG (3 µg/mL), which prevents BDNF from associating with TrkB (Shelton et al. 1995), also blocked the effect of BDNF (data not shown). These results suggested that BDNF enhanced the depolarization-evoked glutamate release via the activation of TrkB in cultured cortical neurons.
BDNF enhanced the exocytotic pathway
We have previously reported that BDNF elicited a transient glutamate release via a non-exocytotic pathway (Takei et al. 1998; Numakawa et al. 1999). The BDNF-induced release was via a glutamate transporter (Numakawa et al. 2001). Therefore, the release through the glutamate transporter might contribute to the enhancement of glutamate release mediated by BDNF. In a previous report, we showed that the transient release induced by BDNF lasted less than 5 min (Takei et al. 1998; Numakawa et al. 1999). In the present study, we confirmed that application of BDNF for 10 min did not affect the basal release before depolarization (data not shown), indicating that the BDNF-enhanced release evoked by depolarization occurred via a distinct mechanism. Then, to examine whether the BDNF-enhanced glutamate release was via exocytosis, we performed a series of experiments as follows. First, to investigate the involvement of a change in membrane potential in this phenomenon, the Na+ dependency of the release was examined. 4AP did not induce a release of glutamate in an extracellular Na+-free solution. A 4AP-evoked release also did not occur in cultured neurons treated with BDNF [relative values (4AP-evoked release/basal release) with and without BDNF were 0.87 ± 0.44 and 0.83 ± 0.21, respectively; p > 0.1]. TTX (0.1 µm), a voltage-dependent Na+ channel blocker, also abolished the 4AP-induced glutamate release with or without BDNF treatment [relative values (4AP-evoked release/basal release) with and without BDNF were 1.11 ± 0.50 and 0.96 ± 0.45, respectively; p > 0.1]. Furthermore, we examined the Ca2+ dependency of the 4AP-evoked release, because Ca2+ influx is known to be an important step in exocytosis. In the extracellular Ca2+-free solution containing EGTA (3 mm), glutamate release was not induced by 4AP with or without BDNF treatment [relative values (4AP-evoked release/basal release) with and without BDNF were 1.18 ± 0.32 and 1.20 ± 0.25, respectively. p > 0.1]. Cadmium chloride (CdCl2; 100 µm), a voltage-dependent Ca2+ channel blocker, also abolished the 4AP-evoked glutamate release [relative values (4AP-evoked release/basal release) with and without BDNF were 0.93 ± 0.14 and 1.22 ± 0.15, respectively; p > 0.1]. These results suggested that Na+ and Ca2+ influx were essential for both the 4AP-evoked release and BDNF-enhanced release.
To further investigate the BDNF-enhanced release that occurred via synaptic vesicles, we used tetanus toxin (TeNT) which was expected to block the exocytosis by cleaving vesicle-associated membrane proteins (VAMPs) (Schiavo and Montecucco 1995). As shown in Fig. 3(a), TeNT (10 nm) completely blocked the 4AP-evoked release with or without BDNF treatment. In addition, Baf (10 nm), which causes the depletion of glutamate in synaptic vesicles (Araque et al. 2000), completely blocked the 4AP-evoked glutamate release with or without BDNF treatment. These results strongly suggested that BDNF enhanced the exocytotic pathway.
Ca2+ increase through intracellular Ca2+ stores was essential for BDNF-enhanced glutamate release
The next question was how BDNF enhanced the depolarization-evoked release. Recent studies showed that BDNF induced an increase in intracellular Ca2+ (Li et al. 1998; Kang and Schuman 2000). Then, we investigated the possibility that a BDNF-mediated release of Ca2+ from intracellular stores was involved in the BDNF-enhanced release of glutamate. Treatment with BDNF for 10 min significantly potentiated both the 4AP- and High-K+-induced increase in Ca2+(Fig. 4a). and the potentiation was maintained for 1 min. This phenomenon was observed in both soma and dendrites. We confirmed that BDNF did not influence the basal level of intracellular Ca2+ before depolarization (data not shown). Thapsigargin (100 µm), which induced a rapid release of Ca2+ from intracellular stores and caused depletion of Ca2+ (Thastrup et al. 1990), completely inhibited the BDNF-induced potentiation (Fig. 4b). BAPTA-AM, a cell-permeable Ca2+ chelator, also inhibited the potentiation by BDNF (data not shown). These results suggested that the BDNF-mediated potentiation of the increase in Ca2+ was through the use of intracellular stores.
Next, we examined whether the BDNF-enhanced release of glutamate required the release of Ca2+ from intracellular stores. Thapsigargin completely abolished the effect of BDNF (Fig. 4c). BAPTA-AM (30 µm) partially inhibited the effect [relative values (4AP-evoked release/basal release) with and without BDNF were 1.88 ± 0.83 and 2.82 ± 0.47, respectively; p > 0.1]. Furthermore, we tested the effect of Xest C (1 µm), a cell-permeable IP3 receptor antagonist (Gafni et al. 1997). The effect of BDNF was blocked by Xest C (Fig. 4c), suggesting that the Ca2+ derived from intracellular stores in a process mediated by the IP3 receptor was essential for the BDNF-enhanced glutamate release.
BDNF enhancement of glutamate release required the activation of the TrkB-PLC-γ pathway
We showed that the BDNF-enhanced glutamate release was dependent on the release of Ca2+ from intracellular stores, which was mediated by the IP3 receptor. It was known that BDNF could induce the activation of mitogen-activated protein kinase (MAPK), phosphatidylinositol-3 kinase (PI3K) and phospholipase C-γ (PLC-γ) through TrkB. Because the activation of PLC-γ induces the generation of IP3, the effect of BDNF might be dependent on the activation of the TrkB-PLC-γ pathway. Then, we examined the role of PLC-γ in the BDNF-enhanced glutamate release. U73122 (10 µm), a PLC-γ inhibitor (Smith et al. 1990), inhibited the BDNF effect in a dose-dependent manner (Fig. 5). It was suggested that the activation of TrkB-PLC-γ was essential for the BDNF-mediated enhancement of glutamate release. In contrast, neither U0126 (10 µm), a MAPKK inhibitor (Favata et al. 1998), nor LY294002 (10 µm), a PI3K inhibitor (Vlahos et al. 1994), blocked the effect of BDNF (data not shown). These results suggested that BDNF enhanced the exocytotic pathway in a PLC-γ-dependent manner.
Studies have suggested that BDNF enhances spontaneous synaptic transmission. BDNF, as well as NT-3, enhanced the frequency of spontaneous postsynaptic currents at neuromuscular junctions (Lohof et al. 1993; Stoop and Poo 1996; He et al. 2000; Yang et al. 2001), and rapidly increased not only the amplitude of impulse-evoked excitatory postsynaptic currents (EPSCs) but also the frequency of spontaneous EPSCs in the visual cortex (Carmignoto et al. 1997). In cultured hippocampal neurons, the application of BDNF acutely increases the frequency of miniature EPSC (Levine et al. 1995; Li et al. 1998; Sherwood and Lo 1999). Also, we previously reported that BDNF elicited a rapid and transient release of glutamate via a non-exocytotic pathway from cultured cortical neurons (Takei et al. 1998) and cerebellar granule neurons (Numakawa et al. 1999). These results suggested that BDNF directly induced excitatory synaptic transmission.
BDNF also potentiated activity-dependent synaptic efficacy in the chronic phase. For example, treatment with BDNF for 5 days increased the depolarization-evoked release of glutamate in cultured cortical neurons and this effect was caused by a rise in the level of synaptic proteins (Takei et al. 1997). At the Schaffer collateral-CA1 synapses, BDNF and NT-3 produce a long-lasting (2–3 h) enhancement of the initial slope of field EPSP, which requires a local protein synthesis, although it was not clear whether an RNA synthesis was involved (Kang and Schuman 1996). It has been reported that treatment with BDNF for 3 h enhanced glutamate release via exocytosis, which requires the synthesis of protein, but not RNA, in cultured cortical neurons (Bradley and Sporns 1999). In the present study, we found that BDNF treatment for 10 min enhanced the depolarization-evoked glutamate release, and the effect did not require translation and transcription, suggesting that the increase in levels of synaptic proteins did not cause the enhancement of release in the acute phase. Therefore, a novel mechanism might be involved in the effect of BDNF.
Since an increase in the intracellular Ca2+ level is essential for exocytosis, we examined whether BDNF affected the transient Ca2+ increase induced by depolarization. We found that it potentiated the Ca2+ increase. It has been reported that BDNF could induce the release of Ca2+ from intracellular Ca2+ stores. For example, the BDNF-induced increase in Ca2+ is mediated by release from intracellular stores in cultured hippocampal neurons (Li et al. 1998). Kang and Schuman (2000) have reported that BDNF- and NT-3-induced synaptic potentiations are mediated by intracellular Ca2+ signaling. These results suggested that BDNF rapidly potentiates synaptic transmission by modulating Ca2+ mobilization through intracellular stores. Then, we examined whether the BDNF-potentiated Ca2+ increase was derived from intracellular stores. It was revealed that BDNF did potentiate the increase through the use of intracellular stores, and that the potentiation was required for the BDNF-enhanced glutamate release. We have recently reported that BDNF itself induced a transient increase in Ca2+ derived from intracellular stores for a few seconds in cultured cerebellar granule neurons (Numakawa et al. 2001). In this system, we confirmed that BDNF directly induced a Ca2+ increase and that there was a return to the basal level after treatment for 10 min (data not shown).
How did the Ca2+ potentiation induced by BDNF enhance the glutamate release? The formation of the synaptic core (SNARE) complex, containing the v-SNARE vesicle-associated membrane protein (VAMP or synaptobrevin), t-SNARE syntaxin 1 and synaptosomal-associated protein of 25 kDa (SNAP-25), constituted a crucial step in synaptic vesicle fusion at the nerve terminals (reviewed by Sudhof 1995). Several studies reported that Ca2+ affected the affinity among synaptotagmin, syntaxin and SNAP-25 (Chapman et al. 1995, 1996; Leveque et al. 2000; Verona et al. 2000; Littleton et al. 2001). Leveque et al. (2000) suggested that synaptotagmin phosphorylated by Ca2+/calmodulin-dependent protein kinase II (CaMKII) potentiated the interaction with syntaxin and SNAP-25. Therefore, the BDNF-potentiated Ca2+ increase may facilitate the formation of the SNARE complex.
In the present study, it was suggested that an IP3 receptor-mediated release of Ca2+ from intracellular stores was involved in the BDNF-enhanced release. It is known that the generation of IP3 is induced by activated PLC-γ which was downstream of TrkB. Then, we examined the possibility that PLC-γ activation was essential for the BDNF enhancement of glutamate release. We confirmed that the activation of PLC-γ was required for the enhancement. In Xenopus motorneurons, PI3K and PLC-γ/IP3 receptor pathways were both necessary and sufficient to mediate the NT-3-induced potentiation of transmitter release (Yang et al. 2001). In contrast, it was reported that BDNF treatment for 10 min augmented glutamate release from synaptosomes prepared from rat and mouse cerebral cortex and that the phenomenon required the activation of TrkB-MAPK pathway (Jovanovic et al. 2000). Our results were different in that the BDNF-enhanced glutamate release was dependent on the TrkB-PLC-γ pathway, not on MAPK. Therefore, it was suggested that neurotrophins elicited biological effect in different signaling pathways dependent on the cell population or experimental conditions.
It was reported that a rise in intracellular Ca2+, which was regulated in an activity-dependent manner, caused production of BDNF (Shieh et al. 1998; Tao et al. 1998), suggesting that BDNF expression was regulated by synaptic activity. Recently, we reported that BDNF was locally and rapidly released at synaptic sites in an activity-dependent manner (Kojima et al. 2001). On the other hand, many studies reported that BDNF potentiated spontaneous and activity-dependent synaptic transmission (Lohof et al. 1993; Levine et al. 1995; Stoop and Poo 1996; Takei et al. 1997, 1998; Li et al. 1998; Numakawa et al. 1999; Sherwood and Lo 1999). Therefore, released BDNF in an activity-dependent manner might be important for synaptic function as a positive regulator. Here, we found that BDNF enhanced exocytotic glutamate release, that is, BDNF might potentiate the following synaptic activity. Thus, synaptic efficacy was regulated by not only BDNF itself but also BDNF-potentiated activity.
This work was supported in part by Grants-in-Aid for Scientific Research from the Ministry of Education, Science, Sports, Culture and Technology of Japan, and from CREST of the Japan Science and Technology Cooperation. TN and SY are JSPS Postdoctoral Fellows. We thank the Regeneron Pharmaceutical Co. for donating BDNF, and Dr Hiroyuki Nawa (Niigata University) for valuable discussions with regard to this study.