Screening inducers of neuronal BDNF gene transcription using primary cortical cell cultures from BDNF-luciferase transgenic mice

Brain-derived neurotrophic factor (BDNF) is a key player in synaptic plasticity, and consequently, learning and memory. Because of its fundamental role in numerous neurological functions in the central nervous system, BDNF has utility as a biomarker and drug target for neurodegenerative and neuropsychiatric disorders. Here, we generated a screening assay to mine inducers of Bdnf transcription in neuronal cells, using primary cultures of cortical cells prepared from a transgenic mouse strain, specifically, Bdnf-Luciferase transgenic (Bdnf-Luc) mice. We identified several active extracts from a library consisting of 120 herbal extracts. In particular, we focused on an active extract prepared from Ginseng Radix (GIN), and found that GIN activated endogenous Bdnf expression via cAMP-response element-binding protein-dependent transcription. Taken together, our current screening assay can be used for validating herbal extracts, food-derived agents, and chemical compounds for their ability to induce Bdnf expression in neurons. This method will be beneficial for screening of candidate drugs for ameliorating symptoms of neurological diseases associated with reduced Bdnf expression in the brain, as well as candidate inhibitors of aging-related cognitive decline.


Results construction of a screening assay using primary cultures of Bdnf-Luc mouse cortical cells.
We have previously demonstrated that induction of Bdnf transcription could be detected by measuring luciferase activity and in vitro bioluminescence imaging using primary cultures of cortical cells prepared from Bdnf-Luc mouse embryos 19,20 . Here, we used these cultured cortical cells to screen for potential inducers of Bdnf transcription in neurons. We prepared primary cultures of cortical cells from Bdnf-Luc and wild-type mice at embryonic day 16.5 using 96-well culture plates (Fig. 1a,b). To induce membrane depolarization, cultured cells (at 13 days in vitro [DIV]) were treated with high concentrations (25 mM) of KCl for 6 h (Fig. 1b) (which evokes Ca 2+ influx into neurons mainly via L-type voltage-dependent Ca 2+ channels [L-VDCC] and subsequently activates Bdnf transcription 21 ), then luciferase activity was measured in each well. Compared with controls (5 mM KCl), we detected higher luciferase activity in primary cultures of Bdnf-Luc cortical cells treated with 25 mM KCl (Fig. 1c). In contrast, we barely detected luciferase activity in primary cultures of cortical cells prepared from wild-type mouse embryos in the absence or presence of high KCl concentrations (Fig. 1c).
Next, we used commercially available neurotransmitter libraries (ENZO Life Sciences, Inc.; Supplementary Table S1) to examine the ability of each compound to induce Bdnf expression. Cultured cells prepared from Bdnf-Luc mice were treated with each test compound for 6 h, then luciferase activity was measured in each well. We compared luciferase activities of vehicle (DMSO) controls with each compound. Compounds that increased luciferase activity by more than 2-fold were defined as active. Among seven neurotransmitter libraries (dopaminergic, adrenergic, serotonergic, cholinergic, histaminergic, metabotropic glutamatergic, and GABAergic; Supplementary Table S1), active compounds were particularly detected in the dopaminergic library (Fig. 2a). We also identified several active compounds from the adrenergic library ( Supplementary Fig. S1a) but rarely from other libraries ( Supplementary Fig. S1b-f). In the dopaminergic library, most active compounds were classified as dopamine agonists and dopamine D 1 agonists. Meanwhile, active compounds in the adrenergic library were agonists for the adrenaline β receptor. These results corresponded with our previous study on G s -coupled dopamine D 1 -and adrenaline β receptor-mediated Bdnf transcription via the N-methyl-D-aspartate receptor (NMDAR)/calcineurin/CREB-regulated transcriptional coactivator 1 (CRTC1)/CREB pathway 19 . SKF38393 and isoproterenol have been previously shown to activate Bdnf transcription 19 , and were included in the dopaminergic and adrenergic libraries used in this study (Fig. 1a [No. 34] and Supplementary Fig. S1a [No. 4], respectively). We chose one compound from the dopaminergic library, A68930 (a selective D 1 agonist; Fig. 1a [No. 33]), to determine whether it induced endogenous Bdnf expression in neurons. Using primary cultures of rat cortical cells, we found that endogenous Bdnf mRNA expression levels were increased by A68930 (Fig. 2b). Corresponding with a previous finding that dopamine D 1 receptor activation induced Bdnf transcription via the NMDAR/calcineurin pathway 19 , A68930-induced Bdnf transcription was completely blocked by the NMDAR antagonist D-2-amino-5-phosphonovaleric acid (APV) and calcineurin inhibitor FK506 (Fig. 2c). We also chose several compounds, propylnorapomorphine (Fig. 1a Supplementary Fig. S1b), only slightly affected mRNA levels ( Supplementary Fig. S2d). Focusing on these compounds, we found that changes in endogenous Bdnf mRNA expression levels were strongly correlated with those in luciferase activity ( Supplementary Fig. S2e, r = 0.931, p < 0.0001). Furthermore, we identified three active compounds, all of which were GABA A receptor antagonists, from the GABAergic library ( Supplementary Fig. S1f www.nature.com/scientificreports www.nature.com/scientificreports/ [No. 29,34,and 35]). This corresponds with a previous result showing that a GABA A receptor antagonist blocked inhibitory neurotransmission and evoked neuronal excitation-induced Bdnf transcription in mature neurons 22 . Taken together, we showed that compounds activating Bdnf transcription in neuronal cells can be screened using primary cultures of Bdnf-Luc cortical cells in a 96-well format.
Identification of herbal extracts that induce Bdnf expression. Next, we determined whether the current screening assay could be used to identify active crude extracts. We determined the ability of a series of extracts prepared from herbal medicines to induce Bdnf transcription in neurons. We used a library consisting of 120 herbal extracts (Supplementary Table S2). Cultured Bdnf-Luc cortical cells were treated with herbal extracts at a final concentration of 500 μg/mL for 6 h (Fig. 3a), 24 h (Fig. 3b), or 48 h (Fig. 3c) (Fig. 3c). We chose a subset of active extracts and confirmed that they significantly increased the endogenous expression of Bdnf  19,20 . (b) Schematic for validating the screening assay system. In 96-well culture plates, Bdnf-Luc mouse (Tg) cortical cells and wild-type (Wt) mouse cortical cells were seeded into wells along lines A-D and lines E-H, respectively. At 13 DIV, cells in wells of columns 2-11 were treated with a high concentration (final 25 mM) of KCl for 6 h. Cells in the wells of columns 1 and 12 were treated with PBS for 6 h. (c) Luciferase activity of each well (left) and the average of luciferase activity (right). Means ± SEM (n = 8 (5 mM KCl), or 40 (25 mM KCl)), ****p < 0.0001 vs. 5 mM KCl (unpaired t-test). mRNA in cultured rat cortical cells (Fig. 4a We particularly focused on the effect of the extract prepared from Ginseng Radix (GIN), and found that GIN increased levels of endogenous Bdnf mRNA (Fig. 4b), as well as luciferase activity (Fig. 4c), in a dose-dependent manner. Changes in endogenous mRNA levels were strongly correlated with those in luciferase activity (Fig. 4d, r = 0.991, p < 0.01). We also examined the effect of GIN on levels of exon-specific Bdnf mRNA, and found higher induction levels of exon I and exon IV-containing Bdnf mRNA (Fig. 4e), both of which are major components of activity-regulated Bdnf transcription in neurons 23,24 .
As shown in Fig. 3, several extracts reduced luciferase activity. We demonstrated that expression levels of Luciferase and endogenous Bdnf mRNA were reduced after addition of the transcription inhibitor actinomycin D ( Supplementary Fig. S4a, t 1/2 = 5.21 h [Bdnf], 3.67 h [Luciferase]), in cultured Bdnf-Luc cortical cells. In contrast, luciferase activity did not significantly alter after actinomycin D treatment ( Supplementary Fig. S4b), suggesting that luciferase protein was stably expressed in the cells. Therefore, reduced luciferase activity by these extracts was not caused by the repression of Bdnf transcription. www.nature.com/scientificreports www.nature.com/scientificreports/ We found that some compounds and extracts increased luciferase activity (Figs 2a and 3, Supplementary  Fig. 1). To exclude the possibility that increased luciferase activity reflected increased stability of luciferase protein and/or the enhancement of luciferase translation, we used actinomycin D. We found that high concentrations (25 mM) of KCl, norepinephrine, and GIN significantly increased luciferase activity using a screening assay ( Supplementary Fig. S4c), which agreed with our earlier results (Figs 1c and 3a, Supplementary Fig. S1a). However, these increases were not observed in the presence of actinomycin D ( Supplementary Fig. S4c), suggesting that increased luciferase activity mainly reflected the activation of Bdnf transcription.
Ginsenosides or gintonin do not participate in activation of Bdnf transcription. Next, we sought to identify active compounds from GIN. Ginsenosides are well-known active compounds of Ginseng Radix 25 . Here, we examined eight types of ginsenosides (specifically, ginsenoside Rb1, Rb2, Rc, Rd, Re, Rf, Rg1, and Rg2), and determined the activity of these compounds using a screening assay. Cultured Bdnf-Luc cortical cells were treated with each ginsenoside for 6 h, then luciferase activity was measured in each well. Although GIN increased luciferase activity in a dose-dependent manner (Fig. 4c), no individual ginsenoside significantly increased luciferase activity (Fig. 5a), suggesting that individual ginsenosides do not affect Bdnf expression. In contrast, gintonin, a lysophosphatidic acids (LPA)-protein complex mainly containing LPA C 18:2 , has previously been isolated as an active component of Ginseng Radix 26 . LPA receptors are G protein-coupled receptors (GPCR) such as G i/o , G 12/13 , G q/11 , and G s -coupled receptors. We previously reported that stimulation of G s -and G q -coupled receptors activates Bdnf transcription via a NMDAR/Ca 2+ /calcineurin/CRTC1/CREB-dependent pathway 19 , suggesting that GIN-regulated Bdnf expression may be regulated by gintonin-regulated activation of G q/11 and/ or G s -coupled LPA receptors. However, the LPA 1/2 agonist oleoyl-L-α-lysophosphatidic acid did not affect Bdnf mRNA expression (Fig. 5b). Additionally, the LPA 1/3 antagonist, Ki 16425, did not suppress GIN-induced Bdnf expression (Fig. 5c). We also examined whether a series of ginsenosides with or without LPA 1/2 agonist could affect Bdnf expression using a screening assay. We chose major ginsenosides, ginsenoside Rb1, Rc, Rd, Re, and Rg1, in an extract of Ginseng Radix (Supplementary Table S2, No. 36, refer to database URL). These ginsenosides and/or LPA 1/2 agonist at different concentrations were added to cultured Bdnf-Luc cortical cells for 6 h, then  (a) Cells were treated with 500 μg/mL GIN and total RNA was prepared at the indicated time. Means ± SEM (n = 3), **p < 0.01, ***p < 0.001, and ****p < 0.0001 vs. water at the same time point (two-way ANOVA with Tukey's multiple comparisons test). (b) Cells were treated with different concentrations of GIN, and total RNA was prepared 3 h after treatment. Means ± SEM (n = 3), ****p < 0.0001 vs. 0 μg/mL (one-way ANOVA with Dunnett's multiple comparisons test). (c) Bdnf-Luc cortical cells were seeded into a 96-well culture plate and cultured for 13 days. Cells were then treated with different concentrations of GIN, and luciferase activity in each well was measured 6 h after treatment. Means ± SEM (n = 3), *p < 0.05 and ****p < 0.0001 vs. 0 μg/mL (one-way ANOVA with Dunnett's multiple comparisons test). (d) Relationship between changes in endogenous Bdnf mRNA expression levels (b) and those in luciferase activity (c) was analysed using a correlation coefficient test. Statistical analysis was performed by Pearson's correlation coefficient test. (e) The effect of GIN on the expression of 5′ exon-specific Bdnf mRNA in cultured rat cortical cells. Cells were treated with 500 μg/mL GIN and total RNA prepared 3 h after the treatment. Means ± SEM (n = 3), **p < 0.01, ***p < 0.001, and ****p < 0.0001 vs. water (unpaired t-test). N.D.: not detected. www.nature.com/scientificreports www.nature.com/scientificreports/ www.nature.com/scientificreports www.nature.com/scientificreports/ luciferase activity was measured. As shown in Fig. 5d, luciferase activity was not affected by a combination of ginsenosides with or without oleoyl-L-α-lysophosphatidic acid, whereas it was significantly increased by GIN. We also prepared a methanol eluate fraction with enriched ginsenosides 27 from a water extract of Ginseng Radix, using a Diaion HP-20 column chromatography ( Supplementary Fig. S5a,b). A screening assay showed that luciferase activity was unaffected by this methanol eluate fraction but was increased by water extract ( Supplementary  Fig. S5c), supporting our finding of a lower effect of individual ginsenoside on luciferase activity (Fig. 5a). Taken together, GIN was shown to activate endogenous Bdnf expression in neurons, yet individual ginsenosides and/or activation of LPA receptors by gintonin does not participate in its induction.
We also used a compound library consisting of 96 types of herbal medicine-derived compounds (Supplementary Table S3), and examined the ability of each compound to induce Bdnf expression in primary cultures of Bdnf-Luc cortical cells. Cultured Bdnf-Luc cortical cells were treated with each compound for 6 h, then luciferase activity was measured in each well. We found that six compounds increased luciferase activity ( Supplementary Fig. S6). Among these active compounds, aconitine (No. 1), hypaconitine (No. 56), and mesaconitine (No. 81) are components of Aconiti Radix Processa, the extract of that was identified as active using our screening assay (Fig. 3a [No. 94]). Aconitum alkaloids reportedly bind the α subunit of voltage-dependent Na + channels and inhibit their inactivation, resulting in the induction of membrane depolarization in neurons 28 . Considering a previous report showing activity-dependent Bdnf transcription in neurons 21 , this suggests that the extract prepared from Aconiti Radix Processa may induce activity-dependent Bdnf expression mediated by Aconitum alkaloids.
Using a luciferase-based promoter assay, we found that GIN activated Bdnf promoter IV (Bdnf-pIV; Fig. 6b), which is a core promoter involved in activity-regulated Bdnf transcription 21,29 . Furthermore, activation of the promoter was barely observed when CRE (also known as CaRE3 21,29 ) on Bdnf-pIV was mutated (Fig. 6b). Taken together with the results obtained from a series of inhibitors (Fig. 6a), CREB-mediated transcription evoked mainly by L-VDCC/Ca 2+ /CaMKs appears to be involved in the activation of Bdnf transcription by GIN.
We next examined whether CREB-dependent transcription was enhanced by GIN. It is well known that CREB phosphorylation at serine 133rd and nuclear translocation of the CREB coactivator CRTC1 contributes to the activation of CREB-dependent transcription by neuronal activity 30 . Here, we found that GIN increased the number of neurons with phosphorylated CREB at serine 133rd and nuclear CRTC1 (Fig. 6c,d). These results indicate that GIN activates Bdnf transcription via Ca 2+ signal-mediated CREB-dependent transcription.
We also comprehensively analysed gene expression profiles regulated by GIN in cultured rat cortical cells. We chose transcripts with fold-change values of greater than or less than 2 (upregulated and downregulated, respectively). We found that 63 and eight transcripts were significantly upregulated and downregulated, respectively, by treatment of the cells with GIN for 3 h (Supplementary Table S4). We also found that GIN increased the expression of Bdnf and other genes encoding plasticity-related factors such as Nr4a1, Nr4a2, and Homer1 (Supplementary Table S4a) 31 .
Finally, we examined whether luciferase activity would be measured in other cell types. It has been reported that membrane depolarization induces Ca 2+ influx via L-VDCCs and subsequently increases Bdnf expression in primary cultures of cerebellar granule cells 32,33 . Here, we prepared primary cultures of cerebellar granule cells from Bdnf-Luc mice brain using a 96-well culture plate, then cultured cells at 7 DIV were treated with high concentration (25 mM) of KCl for 6 h. We found that luciferase activity was significantly increased by KCl, and the increases were completely abolished in the presence of nicardipine (Supplementary Fig. S7). Thus, it is strongly suggested that inducers of Bdnf expression could be identified in other cell types.

Discussion
Here, we developed a screening assay to identify inducers of Bdnf transcription using primary cultures of cortical cells prepared from Bdnf-Luc mouse embryos. In this transgenic mouse line, the firefly luciferase gene has been introduced into the translation start site of Bdnf using a BAC clone containing the entire mouse Bdnf gene. Therefore, in contrast to the construct in which a truncated promoter region is fused to a reporter gene, luciferase expression from BAC-based Bdnf-Luc is predicted to reflect endogenous Bdnf expression. In fact, our previous study showed that changes in Bdnf expression could be successfully evaluated by measuring luciferase activity as well as a bioluminescence signal in cultured cortical cells 19,20 . In this study, we successfully detected membrane depolarization-induced increases in luciferase activity, with high signal/noise ratios, using primary cultures of Bdnf-Luc cortical cells in 96 well-format culture plates. Using this method, we identified compounds and herbal extracts from a series of libraries that activated Bdnf transcription in neurons.
Rc, Rd, Re, and Rg1) in GIN (Supplementary Table S2, No. 36 (refer to database URL)) were mixed (final concentrations of each ginsenoside; 10, 50, or 100 μM). Mixed ginsenosides (G-Mix) was added into Bdnf-Luc cortical cells at 13 DIV with or without LPA (final concentration; 0, 10, 50, or 100 μM). GIN was used as a positive control (right). Luciferase activity in each well was measured 6 h after addition. Means ± SEM (n = 6-8). ****p < 0.0001 vs. water (unpaired t-test). www.nature.com/scientificreports www.nature.com/scientificreports/ www.nature.com/scientificreports www.nature.com/scientificreports/ We defined compounds that increased luciferase activity by more than 2-fold as active ones, and presume that this threshold is necessary to identify inducers of Bdnf mRNA expression levels in neurons. For example, extracts prepared from Ginseng Radix (No. 9, fold-induction value = 2.10) and Zanthoxyli Piperiti Pericarpium (No. 76, fold-induction value = 2.01) increased endogenous Bdnf mRNA expression levels. In contrast, serotonin, which did not increase luciferase activity (fold-induction value = 1.29 [10 nM], 1.13 [100 nM], 1.32 [1000 nM], respectively), only slightly affected the mRNA levels at any concentrations. However, this threshold was not sufficient for identifying Bdnf inducers. We found that 100 nM propylnorapomorphine and 100 nM cabergoline increased luciferase activity (fold-induction value; 100 nM propylnorapomorphine = 2.96, 100 nM cabergoline = 2.42, respectively), yet these compounds at the same concentration did not significantly affect endogenous Bdnf mRNA expression levels. This might reflect a time lag between changes in endogenous Bdnf transcription and those in luciferase activity. For instance, it has been reported that activation of Bdnf transcription peaked at 1 h after depolarization 21 whereas luciferase activity increased gradually and peaked at approximately 7 h after depolarization 20 . In support of this, time-course changes in luciferase activity after the addition of herbal extracts did not always correspond with those in endogenous Bdnf mRNA expression levels in our study. We suggest that this time lag should be further investigated and that endogenous Bdnf expression levels should be examined to determine if they are increased by active agents obtained in the current screening assay.
The most important aspect of our screening assay is that activation of Bdnf transcription could be evaluated in a primary neuronal culture. Previously, Jaanson et al., (2014) developed a similar screening assay in stable HeLa cell 34 . They also used a BAC clone and replaced the coding region of the BDNF with the Renilla luciferase-EGFP gene. Accordingly, changes in reporter gene expression were predicted to synchronise with those of endogenous Bdnf expression. However, this strategy enabled screening for regulators of Bdnf expression in HeLa cells, but not in neuronal cells, yet it is questionable whether active compounds identified in HeLa cells have the ability to induce Bdnf expression in neuronal cells. Ishimoto et al., (2012) also developed a novel screening assay to identify CREB activators in HEK293T cell lines 35 . They also used the same herbal extract library that we used in our study. However, none of the extracts that activated CREB in HEK293T cells could activate Bdnf expression in our current screening assay. Moreover, although the addition of GIN increased CREB phosphorylation in cultured cortical cells, GIN was not observed to activate CREB in the screening assay using HEK293T cells 35 . These results also suggest that intracellular signalling pathways and cellular responses differ between primary neuronal cells and established cell lines. Alternatively, using our current method, we can screen for inducers of Bdnf expression in neuronal cells and other cell types if primary cultures are prepared from tissues of interest. In support of this, we prepared primary cultures of cerebellar granule cells from Bdnf-Luc mice using a 96-well culture plate, and found that membrane depolarization increased luciferase activity, which agreed with previous reports regarding the activity-dependent Bdnf expression in cultured cerebellar granule cells 32,33 .
We found that expression levels of Luciferase and endogenous Bdnf mRNA were similarly reduced after the addition of actinomycin D. In contrast, luciferase activity did not significantly alter under the same conditions, suggesting that luciferase protein is stably expressed even if transcription is inhibited. Consequently, it would be difficult to identify negative regulators of Bdnf transcription using our screening assay, because luciferase protein would be stably expressed and luciferase activity would not be easily reduced even if Bdnf transcription is prevented by negative regulators. Nevertheless, our screening assay allows us to identify inducers of Bdnf transcription in neuronal cells. It is also possible that increases in luciferase activity reflect stabilization of luciferase protein and/or enhanced luciferase translation. However, we confirmed that increases in luciferase activity by depolarization, norepinephrine, and GIN were not observed in the presence of actinomycin D, suggesting a contribution of Bdnf transcriptional activation to increases in luciferase activity.
Although our screening assay allows us to identify Bdnf inducers in neurons, it is difficult to clarify how active agents increase Bdnf transcription. In rodents, Bdnf transcription is controlled by nine distinct promoters reported to be regulated by multiple transcription factors that respond to a series of intracellular signalling pathways 36,37 . Thus, a number of targets exist upstream of Bdnf transcription, although they may not be specific for regulating Bdnf expression. Therefore, it is difficult to construct a screening assay to identify active agents that specifically activate the machinery of Bdnf expression. However, our current assay will be useful for screening agents that are pharmacologically unvalidated compounds or crude extracts, even though the mechanisms of action of these agents on the induction of Bdnf transcription are unclear. In contrast to our screening assay, screenings using genetically engineered biosensors enable the identification of agents that affect specific molecules such as GPCRs. For example, an allosteric biosensor of cAMP has been developed that monitors changes in intracellular signalling evoked by the activation of G s -coupled GPCRs. Using HEK293 cells expressing this allosteric biosensor, Vedel et al., (2015) showed that β 2 adrenergic receptor agonists including salbutamol and metaproterenol increased intracellular cAMP levels in a dose-dependent manner 38 . However, among β 2 adrenergic agonists in the adrenergic library used in this study, only metaproterenol increased luciferase activity. Screening assays focusing on specific targets are useful for identifying active agents with certain mechanisms of action. However, active agents obtained using these methods are not always active for the purpose of identifying Bdnf inducers. Thus, our screening assay is beneficial for the screening of Bdnf inducers in neuronal cells, particularly for identifying unvalidated agents.
Our screening assay identified several herbal extracts that induced Bdnf transcription, and we subsequently confirmed that these extracts induced endogenous Bdnf expression in cultured cortical cells. Although it was difficult to identify negative regulators of Bdnf transcription using our screening assay, we nevertheless identified several extracts that reduced luciferase activity. One possible reason for this is reduction of cell viability based on our use of crude water extracts of herbal medicines, which might include cytotoxic components. Because the luciferase-luciferin reaction that produces bioluminescence is dependent on ATP, a reduction of cell viability would result in lower luciferase activity. In any case, we chose GIN from the active extracts, and demonstrated that it regulates Bdnf transcription via Ca 2+ signal-mediated CREB-dependent transcription.
www.nature.com/scientificreports www.nature.com/scientificreports/ An important point that remains unclear in this study is how GIN can upregulate Bdnf expression in neuronal cells. We attempted to identify active compounds contributing to GIN-induced Bdnf transcription by examining the effect of a series of ginsenosides on Bdnf transcription, yet found that no ginsenoside affected transcription. There are at least two possibilities to explain this: first, multiple ginsenosides may cooperatively participate in the induction of Bdnf expression; or second, other components may contribute to Bdnf induction. We found that a mix of major ginsenosides in GIN did not affect luciferase activity. Moreover, the MeOH eluate fraction of Ginseng Radix with enriched ginsenosides also had no effect on luciferase activity, suggesting that multiple ginsenosides are less likely to affect Bdnf induction. Gintonin, an LPA-protein complex that acts on LPA receptors 26 , is a candidate for an active component. However, our present results showed that an LPA 1/2 agonist had no effect on Bdnf mRNA expression, nor did an LPA 1/3 antagonist on GIN-induced Bdnf expression. In this study, we did not examine whether other LPA receptors (G s -, G q -, G i -, or G 12/13 -coupled GPCRs 39 ) such as LPA 4/5 contribute to GIN-induced Bdnf expression. Because our current results also show a major contribution of L-VDCC/Ca 2+ and CaMK pathways to GIN-induced Bdnf expression, it is unlikely that G s -and/or G q -coupled LPA receptor-mediated Bdnf expression is not involved in the induction. We have previously reported that G s -or G q -coupled GPCR-mediated Bdnf induction is mainly dependent on the NMDAR/Ca 2+ /calcineurin/CRTC1/ CREB pathway 19 . If GIN-induced Bdnf expression is caused by the activation of G s -and/or G q -coupled LPA receptors by gintonin, GIN-induced Bdnf expression should be strongly inhibited by NMDAR antagonist or calcineurin inhibitor. We also confirmed that a series of ginsenosides with LPA did not affect Bdnf expression using a screening assay.
Because of known membrane depolarization-induced Bdnf expression in neurons 21 , we examined K + concentrations in GIN solution but detected negligible final concentration of K + in the culture medium (approximately 4 mM K + in 10 mg/mL GIN solution). We previously reported that high concentrations of KCl-evoked membrane depolarization affected the expression of a large number of transcripts in cultured rat cortical cells 40 . In contrast, our microarray analysis showed that GIN likely regulates a limited number of transcripts in neurons. Further investigations are therefore necessary to reveal the cellular and molecular basis underlying GIN-induced Bdnf expression in neuronal cells. It is plausible that GIN exerts its neurotrophic effects via the induction of Bdnf expression in neurons. In support of this, GIN showed beneficial effects on neurological diseases including Alzheimer's disease 41 and depression 42,43 . Further work should examine whether these effects involve BDNF induction. Previous work demonstrated that the administration of ginsenosides such as Rg1 exhibited an antidepressant-like effect in mice 44 and ameliorates memory impairment in Alzheimer's disease model mice 45 . Although these reports show that ginsenoside rescued the reduced expression of BDNF in model animals, it is unclear whether it does so under normal conditions. Our current study is the first to show that GIN, but not active components of GIN such as ginsenoside and gintonin, directly activates Bdnf transcription in neurons. Thus, GIN is expected to have beneficial effects on cognitive and other neuropsychiatric functions in healthy groups as well as those affected by neurological diseases.
Reduced BDNF levels have been reported in some psychiatric disorders and neurodegenerative diseases such as depression and Alzheimer's disease, strongly indicating that BDNF inducers have beneficial effects in these diseases. Further, it has also been reported that higher BDNF expression levels in the brain are associated with slower cognitive decline 46 . Our current screening assay is useful for identifying Bdnf inducers from crude extracts prepared from herbal medicines and chemical compound libraries, and possibly also from natural foods. Therefore, this assay could be used to develop cognitive stabilisers in the future, which could be consumed as part of a daily diet to suppress cognitive impairment related to aging via increased Bdnf expression in the brain. Taken together, our screening assay could contribute to identifying candidate drugs for improving symptoms in neural diseases as well as protective agents of age-related cognitive decline, at the early phase of drug screening. . Each herbal extract used in this study was obtained using a standard method described as follows: each herbal medicine (purchased from Tochimoto Tenkaido (Osaka, Japan)) was extracted in water (10-times volume of herbal medicine) at 100 °C for 50 min, evaporated under reduced pressure, and freeze-dried to obtain a powder extract. Each herbal extract was redissolved in water.  . 1733 and 1809), and were performed in accordance with the Guidelines for the Care and Use of Laboratory Animals of the University of Toyama and Takasaki University of Health and Welfare. Mice were housed under standard laboratory conditions (12 h-12 h/light-dark cycle, room temperature at 22 ± 2 °C) and had free access to food and water. The generation of Bdnf-Luc mice has been described previously 19 . Wild-type littermates were used as control animals (Fig. 1b,c). www.nature.com/scientificreports www.nature.com/scientificreports/ primary cultures. Primary cultures of Bdnf-Luc mouse cortical cells were prepared from transgenic mice at embryonic day 16.5, as described previously 19 . The cerebral cortex was isolated from each embryonic brain, and tissue lysates were briefly prepared from the remaining brain using a passive lysis buffer (Promega, Madison, WI, USA). Next, luciferase activity of each lysate was measured using a luminometer, and transgenic brains were selected on the basis of luciferase activity. Dissociated cells were seeded at 7.6 × 10 4 cells and cultured in poly-L-lysine-coated 96-well culture plates (Greiner Bio-One, Kremsmünster, Austria) with neurobasal medium (Thermo Fisher Scientific, Waltham, MA, USA) containing B27 supplement (Thermo Fisher Scientific), 2 μg/mL gentamicin (Thermo Fisher Scientific), and 0.5 mM glutamine (Thermo Fisher Scientific). In Supplementary Figure S4a, dissociated cells were seeded at 1.8 × 10 6 cells and cultured in poly-L-lysine-coated 6-well plates (AGC Techno Glass, Shizuoka, Japan) for RT-PCR. Half the culture medium was replaced with fresh medium every 3 days.

Animals.
Primary cultures of rat cortical cells were prepared from Sprague-Dawley rats at embryonic day 17 (Japan SLC, Shizuoka, Japan), as described previously 19 . Dissociated cells were seeded at 1.8 × 10 6 cells and cultured in poly-L-lysine-coated 6-well plates (AGC Techno Glass) for RT-PCR, or at 8 × 10 5 cells in poly-L-lysine-coated 12-well plates (AGC Techno Glass) for a reporter assay with neurobasal medium, described above. For immunostaining, cells were seeded at 8 × 10 5 cells and cultured on poly-L-lysine-coated coverslips of 18 mm in diameter (Matsunami Glass Ind., Ltd., Osaka, Japan) in 12-well plates (AGC Techno Glass). Half the culture medium was replaced with fresh medium every 3 days.

Microarray analysis.
Microarray was performed according to previous studies 19,47 using a GeneChip Rat Genome 230 2.0 Array and 3′ IVT Express Kit (Affymetrix, Santa Clara, CA, USA). Primary cultures of rat cortical cells at 13 DIV were treated with 500 μg/mL GIN for 3 h, and total RNA was extracted using the RNeasy Mini Kit and QIAshredder (QIAGEN, Hilden, Germany) for microarray analysis.
Reporter assay. For reporter assays, half the culture medium was replaced with fresh medium at 3, 6, and 10 DIV, then DNA transfection was performed by the calcium/phosphate DNA co-precipitation method at 11 DIV. One hour before the addition of calcium/phosphate/DNA mixture, conditioned medium was removed and kept in a 10% CO 2 incubator at 37 °C, and serum-free Dulbecco's modified Eagle medium (DMEM) with high glucose (Invitrogen, Catalog No.12100046) without antibiotics that had been pre-warmed in a 10% CO 2 incubator at 37 °C was added to the cultured cells. Calcium/phosphate/DNA precipitates were prepared by mixing 200 μL of plasmid DNA (16 μg, pGL4.12-Bdnf-pIV:phRL-TK(int−) = 10:1) in 250 mM CaCl 2 solution with an equal volume of 2 × HEPES buffered saline (42 mM HEPES [pH 7.03], 274 mM NaCl, 9.5 mM KCl, 2.67 mM Na 2 HPO 4 , and 15 mM glucose) and then incubating at room-temperature for 15 min. Next, 95 μL of mixture was added to the cultured cells and incubated in a 10% CO 2 incubator at 37 °C for 15 min. Cells were then washed twice with serum-free D-MEM with high glucose in the absence of any antibiotics that had been pre-warmed in 10% CO 2 incubator at 37 °C. Conditioned medium was then returned to the cells. Detailed information on the plasmid DNA for measuring the activity of Bdnf-pIV was described previously 19 . Two days after DNA transfection, cells