An upstream enhancer and MEF2 transcription factors fine-tune the regulation of the Bdnf gene in cortical and hippocampal neurons

Abstract

The myocyte enhancer factor (MEF2) family of transcription factors, originally discovered for its pivotal role in muscle development and function, has emerged as an essential regulator in various aspects of brain development and neuronal plasticity. The MEF2 transcription factors are known to regulate numerous important genes in the nervous system, including brain-derived neurotrophic factor (BDNF), a small secreted neurotrophin responsible for promoting the survival, growth, and differentiation of neurons. The expression of the Bdnf gene is spatiotemporally controlled by various transcription factors binding to both its proximal and distal regulatory regions. While previous studies have investigated the connection between MEF2 transcription factors and Bdnf, the endogenous function of MEF2 factors in the transcriptional regulation of Bdnf remains largely unknown. Here, we aimed to deepen the knowledge of MEF2 transcription factors and their role in the regulation of Bdnf comparatively in rat cortical and hippocampal neurons. As a result, we demonstrate that the MEF2 transcription factor-dependent enhancer located at −4.8 kb from the Bdnf gene regulates the endogenous expression of Bdnf in hippocampal neurons. In addition, we confirm neuronal activity-dependent activation of the −4.8 kb enhancer in vivo. Finally, we show that specific MEF2 family transcription factors have unique roles in the regulation of Bdnf, with the specific function varying based on the particular brain region and stimuli. Altogether, we present MEF2 family transcription factors as crucial regulators of Bdnf expression, fine-tuning Bdnf expression through both distal and proximal regulatory regions.

Myocyte enhancer factor 2 (MEF2) transcription factors belong to the MADS box transcription factor superfamily, which in vertebrates consists of four paralogous members (MEF2A-D) (1, 2, 3). While the initial identification of MEF2 factors was predominantly associated with myogenesis (4, 5), subsequent research has demonstrated their pleiotropic roles within the central nervous system, where MEF2A, MEF2C, and MEF2D are mainly expressed (3, 6). It has become increasingly evident that MEF2 transcription factors have a significant role in synaptogenesis, neuronal survival, differentiation, and plasticity (7). All MEF2 proteins have a highly conserved N-terminus, containing the MADS-box and MEF2 domain, and a divergent C-terminus important for transactivation. The MADS-box and MEF2 domains mediate dimerization and binding of the MEF2 homo- or heterodimers to DNA consensus site (C/T)TA(A/T)₄TA(G/A), known as MEF2 response element (MRE) (8, 9, 10). After binding, MEF2 transcription factors can activate or repress transcription of their target genes by recruiting coregulators like histone acetyltransferases or deacetylases, respectively (11, 12, 13, 14).

One of the target genes of MEF2 transcription factors is Bdnf (6, 10, 15, 16), a neurotrophin that has an important role in the developing and mature organism by modulating neuronal survival, development, and plasticity (17). Rodent Bdnf gene contains eight 5′ non-coding exons (I-VIII) and one 3′ protein-coding exon (IX), with each exon regulated by its distinct promoter (18). The Bdnf transcripts are generated by splicing one of the non-coding exons (I-VIII) together with the coding exon (IX) (18, 19). In general, Bdnf transcripts of the first cluster (exon I, II, III) have a nervous system-specific expression pattern, whereas Bdnf exon IV- and VI-containing transcripts are expressed in both neural and non-neural tissues (18, 19, 20). The spatiotemporal expression of Bdnf gene is ensured by various transcription factors binding to its distinct promoters (16, 21, 22, 23) and through the use of different enhancer regions (24, 25, 26).

The role of MEF2 family transcription factors in the regulation and signaling of brain-derived neurotrophic factor (BDNF) has been previously demonstrated in numerous studies (6, 10, 15, 21, 27). For example, the BDNF-activated ERK5-MEF2 signaling pathway has been shown to induce the survival of newly generated cerebellar granule neurons (27) and embryonic cerebral neurons (28). Also, silencing of Mef2a and Mef2d using RNA interference reduced activity-dependent induction of Bdnf in rat hippocampal neurons (10). In rat cortical neurons, knockdown of Mef2d strongly increased activity-dependent expression of the Bdnf exon I-containing transcripts, while knockdown of Mef2a or Mef2c reduced neuronal activity-dependent expression of Bdnf exon I- or Bdnf exon IV-containing transcripts, respectively (6). Moreover, it has been shown that MEF2D needs the presence of CREB to bind Bdnf promoter IV in mouse cortical neurons (15), and we have previously shown the binding of MEF2A and MEF2C to Bdnf promoter IV by in vitro DNA pulldown assay in rat cortical tissue nuclear lysates (16). Finally, a region ∼4.8 kb upstream of the Bdnf promoter I encompasses MRE, binds MEF2D transcription factor, and has been shown to regulate the neuronal activity-dependent transcription of Bdnf promoter I-driven reporter constructs in hippocampal neurons (10) but not in cortical neurons (6).

Although the role of MEF2 in the regulation of Bdnf has been described in several previous studies, the endogenous activity of the −4.8 kb MEF2-dependent enhancer and the contribution of MEF2 family transcription factors to the major Bdnf transcripts has not been addressed. Here, we aimed to comparatively study the role of MEF2 transcription factors in the regulation of Bdnf expression in cortical and hippocampal neurons. To achieve this, we used CRISPR/dCas9 system to show for the first time that the −4.8 kb MEF2 transcription factor-dependent enhancer regulates the endogenous expression of the first cluster of Bdnf transcripts in cultured hippocampal neurons. To elucidate the activity of the −4.8 kb enhancer in vivo, we used kainic acid-injected animals and detected robust activity-dependent transcription of −4.8 kb eRNAs in the hippocampus. Finally, RNA interference-mediated knockdown of MEF2 family members in both cultured cortical and hippocampal neurons revealed an interplay between MEF2 transcription factors in the regulation of Bdnf expression through promoter regions and the −4.8 kb enhancer. Together, these findings demonstrate how various members of the MEF2 family of transcription factors exert distinct effects on the expression of Bdnf depending on stimuli and brain regions.

Results

The enhancer located 4.8 kb upstream of Bdnf exhibits neuronal activity-dependent transcription

Previous studies have demonstrated that a putative enhancer located ∼4.8 kb upstream of the Bdnf exon I transcription start site promotes neuronal activity-dependent transcription from Bdnf promoter I in transient expression assays in hippocampal (10) but not in cortical neurons (6). Since different laboratories conducted these experiments using distinct protocols, we decided to comparatively investigate the impact of −4.8 kb enhancer on Bdnf promoter I activity in both cultured rat hippocampal and cortical neurons. For that, we conducted luciferase reporter assays using reporter constructs that were used previously (6, 10) and treated the neurons for 6 h with KCl to mimic neuronal activity. Our results showed that luciferase activity controlled by Bdnf promoter I without the −4.8 kb enhancer region decreased ∼40% (p = 0.00153) in untreated and ∼34% (p = 0.00712) in KCl-treated hippocampal neurons compared to luciferase activity driven by Bdnf promoter I with the −4.8 kb enhancer region (Fig. 1A). Notably, this effect was less evident and statistically not significant in untreated and KCl-treated cortical neurons, where the luciferase activity driven by Bdnf promoter I decreased ∼20% (p = 0.09951) and ∼24% (p = 0,07329) without the −4.8 kb enhancer, respectively (Fig. 1A). These results are in good agreement with previous findings (6, 10), and indicate that in heterologous context the −4.8 kb enhancer region potentiates the basal and KCl-induced activity of Bdnf promoter I specifically in hippocampal neurons.

Figure 1.

The −4.8 kb enhancer exhibits enhancer-related characteristics in hippocampal neurons and adult hippocampal tissue.A, schematics of luciferase constructs used for transfection. Red box represents mouse −4.8 kb enhancer region, white box mouse genomic region upstream of Bdnf promoter I, and mouse Bdnf promoter I (arrow indicates TSS) and yellow box indicates luciferase coding region. At 5 DIV, rat primary cortical and hippocampal neurons were transfected with the indicated constructs. At 6 DIV, tetrodotoxin (TTX) was added to neurons to inhibit spontaneous neuronal activity. At 7 DIV, transfected neurons were treated with KCl (25 mM) and D-APV (5 μM) for 6 h or left untreated (CTRL). Bdnf promoter I activity was measured using luciferase assay. Luciferase levels measured from rat cortical or hippocampal neurons transfected with Bdnf pI-luc(-6143) and left untreated were taken as 1. Error bars indicate SEM (n = 4). For statistical analysis, two-tailed paired t test was preformed between the luciferase levels in cells transfected with Bdnf pI-luc(-6143) and Bdnf pI-luc(-4495) upon the respective treatment. B, the Integrative Genomics Viewer tool was used to characterize the −4.8 kb enhancer region based on previously conducted ATAC-seq data (29), where rat cultured hippocampal, cortical and striatal neurons were treated with 10 mM KCl (1 h KCl) or vehicle (Neurobasal medium). The first cluster of rat Bdnf exons is shown as blue boxes (exons) and horizontal lines connecting boxes (introns). Orange-colored tracks represent data from hippocampal neurons (HC), blue-colored tracks indicate data from cortical neurons (CTX) and purple-colored tracks show data from striatal (STR) neurons. Light red vertical line represents the MRE-sequence within the −4.8 kb enhancer. The conservation track shows conservation across 100 vertebrates (phyloP) and the genomic alignment of the −4.8 kb MRE sequence (CTATTTATAG) in rat, mouse and human is indicated by light red shading. C, UCSC Genome Browser tool (https://genome.ucsc.edu/) was used to describe the −4.8 kb enhancer region based on previously conducted (31) NeuN+ ATAC-seq data in different human brain regions: (HC) hippocampus; (MDT) mediodorsal thalamus; (PVC) primary visual cortex; (AMY) amygdala; (PUT) putamen; (NAC) nucleus accumbens; (DLPFC) dorsolateral prefrontal cortex; (OFC) orbitofrontal cortex; (VLPFC) ventrolateral prefrontal cortex; (ACC) anterior cingulate cortex; (STC) superior temporal cortex; (ITC) inferior temporal cortex; (PMC) primary motor cortex; (INS) insula. The first exon (exon I) of human BDNF is shown as blue box, light red vertical line represents the MRE-sequence 4.8 kb upstream of BDNF exon I. D, Adult (3 months old) male Wistar rats were used for intracerebroventricular injection of kainic acid (KA). The experimental animals were randomly divided into four groups: ICV-injected with saline solution (CTRL) and ICV-injected with KA and sacrificed at the indicated time after the onset of seizures (2 h, 4 h or 8 h kA). The mRNA levels of Fos, −4.8 kb enhancer RNA, total Bdnf and Bdnf exon I- and IV-containing transcripts were measured from the rat hippocampus (HC), cerebral cortex (CTX) and cerebellum (CB). The expression levels were measured using RT-qPCR and normalized to Hprt1 mRNA levels. The results from individual animals are depicted with dots and the average of respective enhancer RNA or mRNA levels in two to three animals is shown with the column. The average level of respective transcript measured from the CTRL group was set as 1. Error bars indicate SEM (n = 2–3 animals). Statistical significance was calculated with two-tailed unpaired unequal variance t test relative to the expression level of the respective transcript levels in the control animals. (A, D) ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

Until now, research focusing on the −4.8 kb enhancer has been limited to luciferase assays. To study the −4.8 kb region as an enhancer in endogenous context, we initially analyzed previously obtained data of assay for transposase-accessible chromatin using sequencing (ATAC-seq) in rat hippocampal, cortical and striatal neurons (29) to study the chromatin state within the −4.8 kb region (Fig. 1B). The position of the −4.8 kb enhancer was determined by the MRE sequence previously described in Flavell et al. 2008 (10) and located at the center of the enhancer. The −4.8 kb enhancer exhibited stronger ATAC-seq signals in both untreated and KCl-treated rat hippocampal neurons than in cortical neurons. In striatal neurons the ATAC-seq signals within the Bdnf and −4.8 kb region were undetectable, which aligns with previous findings demonstrating that the mRNA levels of Bdnf are not detectable in rat striatum (Fig. 1B) (20, 30). As described previously (10), the −4.8 kb region is highly conserved in vertebrates and the −4.8 kb MRE element has identical sequence among rat, mouse and human, suggesting functionality of the −4.8 kb region also in other vertebrates (Fig. 1B). Next, to determine whether the −4.8 kb enhancer could be a regulatory region in humans, we used published ATAC-seq conducted in neurons from different human brain regions (31). Consistent with the ATAC-seq findings from rat neurons, the −4.8 kb region revealed open chromatin, particularly in human hippocampal neurons (Fig. 1C).

While open chromatin is widely associated with regulatory elements, it does not indicate a functional enhancer by default. In addition to chromatin accessibility, most of the active enhancers require the binding of RNA polymerase and transcription factors, which leads to the transcription of non-coding enhancer RNAs (eRNAs) (32, 33). It has been previously shown that eRNA levels correlate with the target gene expression (29, 33) and that the expression of Bdnf is robustly induced in vivo upon kainic acid (KA) treatment in different brain regions (30). Therefore, we hypothesized that −4.8 kb enhancer RNAs are also induced in vivo upon KA-induced neuronal activation. To test this hypothesis, rats were injected intracerebroventrically (ICV) with KA, a glutamate receptor agonist used as a robust model of epileptic seizures (34), or saline solution as a control. Rats were sacrificed 2 h, 4 h or 8 h after the onset of epileptic seizures and −4.8 kb eRNA together with Bdnf mRNA levels were analyzed using RT-qPCR from hippocampus, cerebral cortex and cerebellum, a brain region where Bdnf levels are not induced after KA treatment (30). As a positive control of a neuronal activity regulated gene expression (35, 36), we measured the levels of Fos mRNA, that were induced after KA treatment in vivo in all analyzed brain regions (Fig. 1D). In addition, a strong, up to ∼24-fold upregulation for −4.8 kb eRNAs was seen in rat hippocampus and ∼3-5-fold induction in cerebral cortex, whereas in cerebellum the −4.8 kb eRNA levels were not induced upon KA treatment. Similarly, the mRNA levels of total Bdnf, Bdnf exon I and Bdnf exon IV were induced in the hippocampus and cortex but not in the cerebellum (Fig. 1D). Altogether, these results demonstrate that the −4.8 kb region exhibits enhancer-related characteristics, such as open chromatin and stimulus-dependent transcription, particularly in the hippocampus.

CRISPR/dCas9 mediated targeting of the −4.8 kb region unveils a hippocampus-specific enhancer of the Bdnf gene

Enhancers can interact with and regulate one or several promoters located at considerable distances from their genomic location (37, 38). The −4.8 kb enhancer has been associated with the nearest regulatory region, Bdnf promoter I, in heterologous context (6, 10). Since the regulation of Bdnf expression involves several distinct promoters acting as independent regulatory hubs (18, 30), it is crucial to clarify the function of enhancer with the respect to each Bdnf promoter. Additionally, it is important to establish the connection between the enhancer and its target gene in the endogenous context. To study whether the alteration of the −4.8 kb enhancer affects the levels of different transcripts of Bdnf in an endogenous context, we used lentivirus-mediated CRISPR/dCas9 system to either repress (Fig. 2A) or activate (Fig. 2F) the enhancer in rat primary hippocampal and cortical neurons. Cultured neurons were treated for 3 h with KCl to mimic neuronal activity, and the levels of the −4.8 kb enhancer RNA and the Bdnf mRNAs were measured using RT-qPCR.

Figure 2.

Targeting the −4.8 kb enhancer with CRISPR/dCas9 system affects the expression of endogenous Bdnf in rat hippocampal and cortical neurons. (A–J) At 0 DIV, rat hippocampal (B, D, G, I) and cortical neurons (C, E, H, J) were infected with negative control gRNA (neg gRNA) or with the pool of 3 gRNAs covering the −4.8 kb enhancer region (−4.8 kb gRNA) and dCas9-KRAB (A–E) or VP64-dCas9-VP64 (F–J) coding lentiviruses. At 7 DIV, tetrodotoxin (TTX) was added to neurons to inhibit spontaneous neuronal activity. At 8 DIV, neurons were treated with KCl (25 mM) and D-APV (5 μM) for 3 h or left untreated. Expression of the −4.8 kb eRNAs (B–C, G–H), total Bdnf (Bdnf coding exon) and different Bdnf transcripts (D–E, I–J) were measured using RT-qPCR. For every experiment, the expression level of the respective RNA in cells transduced with dCas9-KRAB or VP64-dCas9-VP64 and negative gRNA-coding lentiviruses and left untreated was taken as one. The dots indicate results from biological replicates (n = 3), with error bars showing SEM. For statistical analysis, two-tailed paired t test was performed between the RNA levels in cells transduced with lentiviruses encoding −4.8 kb enhancer-specific gRNAs and negative control gRNA and either dCas9-KRAB or VP64-dCas9-VP64 upon the respective treatment, #p < 0.1, ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

In cultured neurons the expression of −4.8 kb eRNA increased ∼4-6-fold in hippocampal (Fig. 2, B and G) and ∼2-3-fold in cortical neurons (Fig. 2, C and H) upon KCl treatment, showing the −4.8 kb region as an activity-dependent enhancer in vitro. To repress the enhancer, we used catalytically inactive Cas9 (dCas9) fused with the Krüppel associated box (KRAB) domain (dCas9-KRAB) coupled with 3 distinct gRNAs covering the −4.8 kb enhancer region. Repression of the −4.8 kb enhancer diminished the KCl-dependent expression of −4.8 kb eRNAs in hippocampal (Fig. 2B) but not in cortical neurons (Fig. 2C). Similarly, the repression of the −4.8 kb enhancer decreased the levels of the first cluster of Bdnf transcripts (Bdnf exon I-, IIc- and III-containing transcripts) in hippocampal neurons, where the strongest ∼4-fold decrease was observed for Bdnf exon I-containing transcripts after KCl treatment (Fig. 2D). Repression of the enhancer neither affected the other studied transcripts of Bdnf (Bdnf exon IV-, VI- and IX-containing transcripts) in hippocampal neurons (Fig. 2D) nor the expression levels of Bdnf mRNAs in cortical neurons (Fig. 2E).

Next, we used tandem repeats of transactivator VP16 domain fused with dCas9 (VP64-dCas9-VP64) together with 3 different gRNAs covering the −4.8 kb enhancer to induce the transcription from the enhancer. The activation of the −4.8 kb enhancer increased the −4.8 kb eRNA levels in untreated hippocampal (Fig. 2G) and cortical neurons (Fig. 2H). Activation of the −4.8 kb enhancer also increased Bdnf levels in untreated and KCl-treated hippocampal (Fig. 2I) and cortical neurons (Fig. 2J), where the first cluster of Bdnf transcripts was upregulated ∼2-3-fold, indicating that the transcriptional activation by the CRISPR activator complex artificially induced the transcription from the −4.8 kb enhancer and upregulation of Bdnf expression. Although the −4.8 kb eRNA levels increased in both KCl-treated cultured cortical neurons (Fig. 2, C and H) and after KA treatment in adult cortical tissue (Fig. 1D), the repression of −4.8 kb enhancer had no effect on the −4.8 eRNA (Fig. 2C) and Bdnf levels in cultured cortical neurons (Fig. 2E). Therefore, the role of the −4.8 kb enhancer in the regulation of Bdnf in cortical neurons remains to be elucidated. We cannot rule out that the −4.8 kb enhancer regulates Bdnf expression in specific cortical cell type(s) or depending on developmental stage. Collectively, here we demonstrate for the first time that in the endogenous context the region 4.8 kb upstream of Bdnf is crucial for enhancing the neuronal activity dependent Bdnf expression in hippocampal neurons.

The MEF2 family transcription factors regulate Bdnf via both proximal and distal regulatory regions

In addition to the −4.8 kb enhancer region, MEF2 transcription factors have been shown to regulate Bdnf promoter I and IV (10, 15), but the contribution to all the major promoters of Bdnf has not been addressed. Therefore, next we investigated the role of different endogenous MEF2 transcription factors in the regulation of Bdnf transcripts comparatively in cultured cortical and hippocampal neurons. For this, we used the RNAi-mediated knockdown of Mef2a, Mef2c or Mef2d transcription factors. Mef2b was excluded as it has been shown to have low expression in cultured cortical and hippocampal neurons (6). Rat hippocampal (Fig. 3A) or cortical (Fig. 3B) neurons were transduced with lentiviruses encoding Mef2a, Mef2c, Mef2d or negative control shRNA. As the MEF2 transcription factors are known to function upon stimuli, for example, neuronal activity and after BDNF-TrkB-signaling (39), neurons were left untreated or treated for 3 h with KCl or BDNF to induce membrane depolarisation or TrkB-signaling, respectively.

Figure 3.

Silencing of Mef2 transcription factors alters the transcription from different Bdnf promoters and −4.8 kb enhancer in cortical and hippocampal neurons.A and B, at 0 DIV rat hippocampal (A) or cortical (B) neurons were infected with lentiviruses encoding Mef2 shRNAs (Mef2a, Mef2c or Mef2d) or negative control shRNA (neg shRNA). At 7 DIV, tetrodotoxin (TTX) was added to neurons to inhibit spontaneous neuronal activity. At 8 DIV neurons were treated with 25 mM KCl together with 5 μM D-APV or 50 ng/ml BDNF for 3 h or left untreated (0 h). The mRNA expression levels of different Mef2 family members, Bdnf transcripts and −4,8 kb eRNA were measured using RT-qPCR. The expression levels in cells transduced with negative shRNA-encoding lentiviruses and left untreated was set as 1. All biological replicates (n = 3) are shown as dots. Error bars indicate SEM. Asterisks indicate statistical significance calculated using two-tailed paired t test between the respective transcript levels measured from neurons transduced with neg shRNA and indicated Mef2 shRNA at respective treatment, #p < 0.1, ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001. To simplify the understanding of the results, we arranged the heatmap coloring based on the relative fold-change, which was calculated against the respective negative control shRNA treatment type. On color scale, red demonstrates increase, and blue shows decrease in the expression level. Note that the color scale has been capped for both hippocampal and cortical results to ensure comparability (i.e., fold-changes >5 in both heatmaps are shaded in the darkest red or blue).

First, we analyzed the efficiency of the knockdown tools by measuring the mRNA levels of Mef2 family transcription factors. Silencing of Mef2a, Mef2c or Mef2d decreased the mRNA levels of all the respective Mef2 family members in untreated as well as in KCl- or BDNF-treated hippocampal and cortical neurons. Mef2a and Mef2c shRNAs silenced the target genes with 80 to 90% efficiency (Fig. 3, A and B), silencing of Mef2d shRNA in untreated and treated cortical neurons was slightly weaker, resulting in 50 to 70% reduction of Mef2d mRNA (Fig. 3B).

Next, we studied the effect of MEF2-silencing on −4.8 kb eRNA expression (Fig. 3, A and B). In KCl-treated hippocampal neurons, silencing of Mef2c decreased the level of −4.8 kb eRNA ∼2.5-fold, while silencing of Mef2d increased −4.8 kb eRNA levels ∼2-fold (Fig. 3A). Surprisingly, silencing of Mef2d increased −4.8 kb eRNA levels ∼6-fold in BDNF-treated cortical neurons, while silencing of other Mef2 family members had no effect on the level of −4.8 kb eRNAs in treated cortical neurons (Fig. 3B). These results indicate that MEF2D represses the −4.8 kb enhancer in cortical and hippocampal neurons, and that MEF2C acts as a transcriptional activator of this enhancer in KCl-treated hippocampal neurons.

Silencing of Mef2a had a tendency to decrease the levels of total Bdnf and almost all Bdnf transcripts in untreated and BDNF-treated hippocampal neurons (Fig. 3A), while silencing of Mef2a did not have any effect on the levels of Bdnf mRNA in cortical neurons (Fig. 3B). Silencing of Mef2c decreased the expression of Bdnf in untreated, KCl- and BDNF-treated hippocampal neurons, with the highest decrease, ∼6-7-fold, seen for Bdnf exon I-containing mRNA in untreated and KCl-treated neurons (Fig. 3A). Although silencing of Mef2c increased the levels of first cluster of Bdnf transcripts in KCl-treated cortical neurons, the expression levels of the second cluster of Bdnf transcripts and total Bdnf were decreased (Fig. 3B). Since Bdnf exon IV-containing transcripts are the major source of Bdnf mRNA in the cortical neurons (40), the small increase in the transcripts with exons from the upstream cluster did not influence significantly the total Bdnf mRNA levels. Silencing of Mef2d had the most drastic effects on the expression of Bdnf. In hippocampal neurons the Bdnf exon I-containing transcripts were upregulated ∼6-fold after KCl-treatment and Bdnf exon IV-containing transcripts were downregulated regardless of treatment (Fig. 3A). In cortical neurons the silencing of Mef2d led to significant upregulation of Bdnf expression both in untreated and BDNF-treated, but not in KCl-treated neurons (Fig. 3B).

Collectively, we determined that among the MEF2 family of transcription factors, MEF2C and MEF2D primarily govern the expression of Bdnf in both hippocampal and cortical neurons. Although we cannot definitively ascertain whether the observed effects of silencing are direct or indirect, our findings strongly suggest that the MEF2 family transcription factors have a pleiotropic role in the complex regulation of Bdnf gene, acting through both proximal and distal regulatory regions in distinct ways in cortical and hippocampal neurons.

Discussion

In this study, we show how the family of MEF2 transcription factors exerts multifaceted control over the expression of its target gene Bdnf. This regulation involves distinct mechanisms operating in response to different stimuli and brain regions. While previous research has shown the −4.8 kb region to enhance Bdnf promoter I in heterologous context (6, 10), our findings show for the first time that in endogenous context the −4.8 kb enhancer regulates Bdnf in hippocampal neurons. Notably, we provide evidence of the −4.8 kb enhancer activity in rodents in vivo and describe the −4.8 kb enhancer in an open chromatin conformation in human hippocampal neurons. Finally, we determine that after silencing the MEF2 family members, MEF2C and MEF2D have the most significant impact on the transcriptional regulation of Bdnf in both cultured hippocampal and cortical neurons. Altogether, our results highlight the intricate interplay between MEF2 family transcription factors in the regulation of BDNF expression.

We demonstrate that the −4.8 kb enhancer regulates the first cluster of Bdnf transcripts, containing Bdnf exons I, IIc, and III, after neuronal activity in hippocampal neurons. The first cluster of Bdnf exons is known to exhibit a highly neuron-specific expression pattern and is suggested to be coregulated as a single cluster of promoters (18, 19, 20, 21). We have previously described an intronic enhancer region located downstream of the Bdnf exon III (+3 kb enhancer) which plays a crucial role in BDNF-TrkB signaling- and neuronal activity-induced expression of the first cluster of Bdnf transcripts in neurons (25). Although further studies are needed to determine the precise looping interactions, it is plausible that the −4.8 kb and +3 kb enhancer regions could synergistically facilitate the neuronal activity-dependent expression of the first cluster of Bdnf exons in hippocampal neurons. This aligns well with the hypothesis proposed in Tuvikene et al. 2021 (25), suggesting that the +3 kb enhancer functions as an anchor site for chromatin interactions between the Bdnf (promoters I, II, II) and the respective enhancers.

The hippocampus and cerebral cortex are the brain regions where BDNF is highly expressed (18, 19, 20). In addition to transcription factors involved in Bdnf regulation in both these brain regions (e.g., CREB, NPAS4), numerous transcription factors regulate individual promoters of Bdnf depending on the brain region (16, 21). Although all Bdnf transcripts encode the same protein, it is known that transcript-specific modulation of Bdnf causes differential downstream cascades and phenotypes (41, 42). For example, in the hippocampus the upregulation of Bdnf exon I-containing transcripts increases the complexity of dendritic spine morphology, while upregulation of Bdnf exon IV-containing transcripts reduces fear expression (42). In contrast, mutations in Bdnf promoter IV impair the neuronal activity-dependent expression of BDNF important for the proper development of inhibitory synapses in the cerebral cortex (15). A differential engagement of enhancer regions could ensure this distinctive proportion of Bdnf transcripts in cortical and hippocampal neurons. Accordingly, the MEF2-dependent −4.8 kb enhancer is responsible for the increase in Bdnf exon I-containing transcripts upon neuronal activity and could thereby contribute to the complexity of dendritic spines in the hippocampus. Conversely, and as shown in the current study, the −4.8 kb enhancer is likely dispensable for the regulation of Bdnf exon IV-containing transcripts in cortical neurons, which comprise the vast majority of Bdnf transcripts needed there to refine synaptic connectivity. Moreover, the ability of MEF2 transcription factors to cooperate with cell type-specific transcription factors to activate tissue-specific enhancers (43) might explain the differential activation of regulatory regions crucial for BDNF expression in the cortex and hippocampus.

To date, the binding of MEF2 transcription factors has been linked to the −4.8 kb enhancer in hippocampus (10) and Bdnf promoter IV in cerebral cortex (6, 15, 16). The MEF2 transcription factor binding is not always restricted to the high affinity MRE site. For instance, MEF2 activates the transcription from Bdnf promoter IV through the CaRE1 element (6). In addition to proximal promoters, widespread binding of MEF2 transcription factors has been observed in distal regulatory regions (10, 43). Moreover, MEF2D binding to enhancers that lack the consensus MRE site has been shown to play a crucial role in photoreceptor cells, where MEF2D in cooperation with cell-specific transcription factors regulate the genes critical for photoreceptor functioning (43). Therefore, it is important to consider that our MEF2 knockdown results may arise from a combination of regulatory mechanisms, i.e., binding of the MEF2 transcription factors both to proximal and distal regulatory regions of Bdnf gene and also to the regulatory regions of transcription factors regulating Bdnf.

While MEF2A, MEF2C, and MEF2D are widely expressed in the central nervous system, the proportion of each MEF2 family member varies significantly depending on the particular brain region (3, 6). For example, the expression of MEF2C in the hippocampus is highly restricted to the dentate gyrus (3, 44), where MEF2C is responsible for the development and morphology of dendritic spines (45, 46). Similarly, conditional deletion of MEF2C, but not MEF2A and MEF2D, is linked with altered memory formation and synaptic plasticity in the hippocampus (45, 47, 48), and MEF2C-related aberrations have been implicated in the development of autism-spectrum disorders (ASDs) (7). Recently it was shown that deletion of Mef2c specifically in adult-born dentate granule cells impaired contextual fear memory and caused deficits in social interactions, behavioral patterns associated with ASD (46). Although we have not addressed the exact region or subpopulation of neurons in the hippocampus in our studies, silencing of Mef2c had a profound effect, resulting in a significant reduction not only in the levels of the −4.8 kb eRNA but also in the overall expression of Bdnf mRNA in both basal levels and in KCl-treated hippocampal neurons. Further work on MEF2-regulated Bdnf expression in specific hippocampal subregions on single-cell level is needed. Nevertheless, our results describing Bdnf as a crucial downstream mediator of MEF2C could help to decipher the signaling cascade important for the proper synaptic development in the hippocampus.

Previous genome-wide studies have suggested that the redundancy and compensation among MEF2 family members can complicate the studies of MEF2 transcription factors (10, 48, 49). It is tempting to speculate that the observed effects on Bdnf regulation may also result from the intricate interplay between MEF2 family members. For example, when Mef2d is silenced, MEF2C might compensate for its absence, potentially enhancing the transcription of Bdnf. Studies in mouse cerebellar granule neurons have demonstrated that conditional MEF2D knockout induces compensatory genomic occupancy by MEF2A, but only at specific MEF2D sites (50). Although beyond the scope of this study, it would be interesting to study the endogenous binding of all the MEF2 transcription factors to Bdnf proximal and distal regulatory regions in wild-type and MEF2 knockout models in hippocampal and cortical neurons to confirm the direct participation in the transcriptional regulation.

From the MEF2 family members investigated in this study, silencing of Mef2a leads to the mildest effects when compared to the knockdown of Mef2c and Mef2d. It has been shown that brain-specific Mef2a knockout mice have no abnormalities, Mef2a-Mef2d double knockout mice have deficits in motor coordination but otherwise remain viable, and Mef2a-Mef2c-Mef2d triple knockout mice have decreased body weight together with neuronal apoptosis and postnatal lethality (48). Conversely, homozygous MEF2A knockout mice die during the first postnatal week due to the cardiac dysfunction (51), while MEF2D knockout animals appear to be vital and fertile, but exhibit problems in the circadian system as well as maturation and survival of photoreceptors (43, 52, 53). The described phenotypic peculiarities indicate that MEF2A might have a strong redundancy or it does not play a central role in the neural system, as seen for the regulation of Bdnf gene in cortical neurons.

In agreement with previous results (6), the silencing of Mef2d led to a significant increase in the mRNA levels of Bdnf. It is known that MEF2 transcription factors can, in addition to positive effects, repress their target genes (39). For example, in muscle cells, HDAC4 has been shown to form a repressor complex with MEF2D to control the regulation of genes (12, 13). Similarly, MEF2D in complex with HDAC4 could repress the expression of Bdnf gene in rat hippocampal and cortical neurons. We cannot rule out the possibility that MEF2D is crucial to refine the proper levels of BDNF, as abnormally increased levels of BDNF have been observed in the hippocampus and cortex of patients with temporal lobe epilepsy (54, 55). Whether the increase in Bdnf levels following the silencing of Mef2d indicates the effect of compensation or a way to refine the proper expression levels of BDNF, is a question that needs to be elucidated in the future.

In conclusion, the fine-tuning of Bdnf gene by MEF2 transcription factors illustrates distinct but robust regulatory mechanisms in the cortex and hippocampus. However, numerous exciting research questions are still awaiting exploration. For example, how do various MEF2 family members differentially control Bdnf transcription? How do the Bdnf promoters interact with the −4.8 kb enhancer and are the interactions different across distinct cell types? How do the outcomes of the current study relate to the regulation of Bdnf at the single cell level? Future studies of this intricate relationship will hopefully unlock important insights, clarifying not only the regulation of Bdnf gene but also adding new knowledge on the functioning of the brain.

Experimental procedures

Rat primary cortical and hippocampal neuron culture

For cultured neurons, all animal procedures were conducted in accordance with European Directive 2010/63/EU, reviewed and approved by the Ministry of Agriculture of Estonia (Permit Number: 45). Animals were maintained under a 12 h light/dark cycle in a humidity (50 ± 10%) and temperature (22 ± 1 °C) controlled room. Rats were group-housed (2–4 animals per cage), using either conventional polycarbonate or H-TEMP polysulfone cages with ad libitum access to water and food. The cultures of rat cortical and hippocampal neurons were generated from embryonic (E20/21) Sprague Dawley rats. Briefly, cortices and hippocampi were dissected and incubated in 1 ml of 0.25% trypsin and 1 mM EDTA solution (Gibco) for 10 min at 37 °C. Next, DNaseI (Roche) and MgSO₄ (Sigma) were added to the trypsinized tissue to a final concentration of 0.5 mg/ml and 12 mM, respectively, and incubated for 10 min at 37 °C. Then, 275 μl of trypsin inhibitor (1%, Gibco), 110 μl of BSA (10%, Pan Biotech) and 50 μl of DNaseI (5 mg/ml, Roche) were added. The tissue was triturated and the resulting suspension was diluted in 1× Hank's balanced salt solution. The solution was centrifuged at 200g for 30 s. The supernatant was centrifuged again at 200g for 6 min. The supernatant was removed, and the precipitation was resuspended in prewarmed DMEM (Corning) containing 10% FBS (Pan Biotech). Plates were pre-treated in 0.1 M borate buffer with poly-L-lysine (0.2 mg/ml, Sigma Aldrich) for at least 1 h at room temperature. Cells were incubated for 2 h at 37 °C and 5% CO2. Then the medium was replaced with new cell medium (Neurobasal-A medium (NBA, Gibco) containing 1 × B27 supplement (Gibco), 100 U/ml penicillin, and 0.1 mg/ml streptomycin (Gibco) or 100 μg/ml primocin (Invivogen) and 1 mM L-glutamine (Gibco)). At 2 days in vitro (DIV), all of the medium was changed and 10 μM mitotic inhibitor 5′-fluoro-2′-deoxyuridine (Sigma-Aldrich) was added to inhibit proliferation of non-neuronal cells.

Intracerebroventricular injection of kainic acid

Adult (3-month-old) male Wistar rats were used for intracerebroventricular injection (ICV) of kainic acid. The animals were housed in a specific pathogen-free environment, three per polypropylene cage in controlled temperature (23 ± 2 °C), humidity (50 ± 5%) and light (12 h light/dark cycle), with ad libitum access to food and water. The experiments were carried out in accordance with the National Institute of Health Guidelines for the Care and Use of Mammals in Neuroscience and Behavioral Research (The National Academics Press), European Communities Council Directive 2010/63/EU revising Directive 86/609/EEC, and national Decree-Law No 26 on March 4, 2014, and were approved by the local Animal Care Committee (OPBA) of University of Palermo, Italy and Ministry of Health, Italy. Up to 12 animals were randomly divided into four experimental groups: control (ICV-injected with saline solution), kainate 2 h, 4 h or 8 h (ICV-injected with kainate solution and sacrificed 2 h, 4 h or 8 h, respectively, after the onset of seizures). The surgical procedure was performed under aseptic conditions. Rats were initially anesthetized with intraperitoneal injection of chloral hydrate (200 mg/kg) and kept anesthetized for the entire duration of the surgical procedure with isoflurane. This procedure allows the early recovery of the animal from anesthesia immediately after the end of the surgery and the early and full manifestation of seizure behavior. Rats were placed in a David Kopf stereotaxic apparatus and received bilateral ICV injections of 0.35 μg/μl of kainate (Merck, dissolved in 0.9% physiological saline), using the following stereotaxic coordinates from the Bregma (according to Ref. (56)): AP  =  0, L  =  1.4, and V  =  4.3. The control group was ICV-injected with 1 μl of 0.9% physiological saline. Injections were performed by 30-gauge injector cannula that was connected by a piece of polyethylene tube to the 10 μl Hamilton syringe. Each injection was performed over 3 min, and following injection, the needle remained in the target location for 3 min to avoid kainate solution reflux along the needle tract and to achieve a proper diffusion of the drug. Animals were sacrificed by decapitation at different time points after the onset of epileptic seizures. Brain was quickly removed for dissection of hippocampus, frontal cortex, and cerebellum. Dissected brain areas were rapidly frozen and stored at −80 °C for later use.

Lentivirus production and transduction

All used lentiviruses were produced in HEK293FT cell line (Thermo Fisher Scientific) on 145 mm cell culture dishes, cultured in DMEM (Pan Biotech) with 10% fetal bovine serum (FBS, Pan Biotech) and 1% penicillin-streptomycin (100 U/ml penicillin, 0.1 mg/ml streptomycin, Gibco) medium. The transfection solution contained 18.22 μg effector, 13.67 μg psPAX2 (Addgene) and 9.11 μg pVSVG (Invitrogen) plasmids with 2:1 PEI to DNA ratio. HEK293FT cells were transfected for ∼16 h at 37 °C and 5% CO2. After that, the transfection solution was replaced with 24 ml lentivirus production medium (DMEM (Pan Biotech), 10% FBS (Sera Plus), 1 mM pyruvate, 1× non-essential amino acids (Gibco), and 20 mM HEPES, pH 7.4). Medium containing lentivirus particles were collected twice with 24 h interval. To purify lentiviruses, the collected medium was centrifuged at 4500g for 5 min at +4 °C and the supernatant was filtered through a 0.45 μm polyvinylidene difluoride filter (Merck). 1/9 volume of Speedy Lentivirus Purification solution (Abm) was added to medium and lentiviral particles were precipitated with centrifugation at 7000g for 1 h at +4 °C. Lentiviral particles were resuspended in phosphate buffered saline (PBS) buffer.

To determine the relative lentivirus titres, neurons were infected with different dilutions of viruses at 0 DIV. At 7 DIV, 0.1 mg/ml DNasel (Roche) with 10 mM MgSO₄ (Sigma) was added to neurons and incubated for 20 min at 37 °C to eliminate plasmid DNA contamination. Next, cells were lysed in proteinase K lysis buffer (0.08 mg/ml proteinase K, 30 mM Tris-HCl (pH 8,0), 1% Tween 20, 0.2% NP40, 1 mM EDTA). The lentiviral titers were analyzed based on provirus incorporation using provirus-specific Woodchuck Hepatitis virus post-transcriptional regulatory element or puromycin resistance (PURO) primers, and unrelated genomic region (for normalization (Supporting information). For functional experiments, equal amounts of lentiviral particles were used to transduce neurons, achieving >95% transduction efficiency. Rat cortical neurons were transduced at 0 DIV. At 7 DIV, tetrodotoxin was added to neurons to inhibit spontaneous neuronal activity. At 8 DIV, cells were treated with 25 mM KCl (PanReac AppliChem) and 5 μM D-(2R)-amino-5-phosphovaleric acid (D-APV, Cayman Chemical), with 50 ng/ml BDNF (Peprotech) or left untreated. After 3 h of treatment, neurons were lysed in 500 μl RLT buffer (Qiagen) containing 1% β-mercaptoethanol (ROTH).

RNA isolation, cDNA synthesis and qPCR

Total RNA from rat cortical neurons or ICV-injected rat brain tissue was isolated using RNeasy Mini Kit (Qiagen) or with RNeasy lipid tissue kit (Qiagen), respectively, with on-column DNase digestion using RNase-free DNase set (Qiagen) according to the manufacturer’s instructions. RNA concentration was measured with a BioSpec-nano spectrophotometer (Shimadzu) or with a Nanodrop 2000c (Thermo Scientific) spectrophotometer. cDNA was synthesized from equal amounts of total RNA using Superscript III or IV Reverse Transcriptase (Thermo Fisher Scientific) with 1:1 mixture of oligo(dT)₂₀ (Microsynth) and random hexamer primers (Microsynth). qPCR was performed in triplicates using 1 × HOT FIREpol EvaGreen qPCR Mix Plus (Solis Biodyne) or 1 × LightCycler 480 SYBR Green I Master (Roche) on LightCycler 480 II Real Time PCR instrument (Roche). All used qPCR primers are shown in Supporting information. Hprt1 mRNA levels were used to normalize gene and enhancer RNA expression.

DNA constructs

The RNA interference (pLKO.1-neg-shRNA, pLKO.1-MEF2A-shRNA, pLKO.1-MEF2C-shRNA, pLKO.1-MEF2D-shRNA) and luciferase plasmids (pGL3-BDNFpI-luc(-6143) and pGL3-BDNFpI-luc(-4495)) were kind gift from A. E. West, Department of Neurobiology, Duke University, North Carolina, United States of America. The gRNA targeting sequences (Supporting information) of the −4.8 kb enhancer were designed using Benchling CRISPR tool (http://www.benchling.com) as described previously in Avarlaid et al. 2024 (24), and cloned into the pRRL-U6-gRNA-hPGK-EGFP plasmid. Briefly, a total of 3 gRNAs targeting the ∼400 bp core region covering the −4.8 kb MRE sequence were used and the gRNA viruses were pooled for the CRISPR/dCas9 experiments. The plasmids (pLV-hUbC-dCas9-KRAB-T2A-GFP, pLV-hUbC-VP64-dCas9-VP64-T2A-GFP) used for CRISPR/dCas9-mediated modulation of the −4.8 kb enhancer were previously described in Tuvikene et al. 2021 (25).

Transfection of cells and luciferase reporter assay

Rat cortical neurons were transfected at 5 DIV using Lipofectamine 2000 (Invitrogen) reagent. The transfection was carried out as duplicates in 48-well cell culture plates with each well containing 180 ng of luciferase reporter plasmid and 20 ng of pGL4.83-mPGK-hRluc normalizer plasmid in unsupplemented NBA with DNA to Lipofectamine ratio of 1:3. The transfection was carried out for 3 to 4 h and was terminated by changing the media back to the supplemented NBA that was previously collected from the neurons.

At 6 DIV, TTX (1 μM, Tocris Bioscience) was added to prevent spontaneous neuronal activity. At 7 DIV, neurons were treated with KCl (25 mM, PanReac AppliChem) and D-APV (5 μM, Cayman Chemical) for 6 h or left untreated. After treatments at 7 DIV the cortical and hippocampal neurons were lysed in 1 × Passive Lysis Buffer (Promega) and luciferase assay was performed using Dual-Glo Luciferase Assay (Promega) system. Luminescence signal was measured using GENios Pro Multifunction Microplate Reader (Tecan). For data analysis, background corrected Firefly luciferase signals were normalized with background corrected Renilla luciferase signals and the averages of duplicates were calculated.

Statistical analysis

Sample size estimation was not performed, and randomization or blinding were not used. All tested hypotheses were specified before conducting the experiments. For cultured primary neurons, biological replicates were cultures obtained from rat pups of different litters. For in vivo experiments, biological replicates were individual animals. For statistical analysis, normalized data was log-transformed, mean centered, and autoscaled. The results were analyzed using Microsoft Excel 365 or RStudio version 2021.09.0 and R version R 4.1. Statistical significance was calculated using a two-tailed paired or unpaired t test, as indicated in figure legends, and p-values were not corrected for multiple comparisons. For graphical representation, data was backtransformed and error bars indicate upper and lower limits of backtransformed means ± SEM.

Data availability

All data are contained within the article.

Supporting information

This article contains supporting information.

Conflict of interest

Jürgen Tuvikene and Tõnis Timmusk were employees of Protobios LLC. The authors declare no competing financial interests.

Reviewed by members of the JBC Editorial Board. Edited by Elizabeth J. Coulson