Distinct roles of Bdnf I and Bdnf IV transcript variant expression in hippocampal neurons

Summary:

Brain-derived neurotrophic factor (Bdnf) plays a critical role in brain development, dendritic growth, synaptic plasticity, as well as learning and memory. The rodent Bdnf gene contains nine 5’ non-coding exons (I-IXa), which are spliced to a common 3’ coding exon (IX). Transcription of individual Bdnf variants, which all encode the same BDNF protein, is initiated at unique promoters upstream of each non-coding exon, enabling precise spatiotemporal and activity-dependent regulation of Bdnf expression. Although prior evidence suggests that Bdnf transcripts containing exon I (Bdnf I) or exon IV (Bdnf IV) are uniquely regulated by neuronal activity, the functional significance of different Bdnf transcript variants remains unclear. To investigate functional roles of activity-dependent Bdnf I and IV transcripts, we used a CRISPR activation (CRISPRa) system in which catalytically-dead Cas9 (dCas9) fused to a transcriptional activator (VPR) is targeted to individual Bdnf promoters with single guide RNAs (sgRNAs), resulting in transcript-specific Bdnf upregulation. Bdnf I upregulation is associated with gene expression changes linked to dendritic growth, while Bdnf IV upregulation is associated with genes that regulate protein catabolism. Upregulation of Bdnf I, but not Bdnf IV, increased mushroom spine density, volume, length, and head diameter, and also produced more complex dendritic arbors in cultured rat hippocampal neurons. In contrast, upregulation of Bdnf IV, but not Bdnf I, in the rat hippocampus attenuated contextual fear expression. Our data suggest that while Bdnf I and IV are both activity-dependent, BDNF produced from these promoters may serve unique cellular, synaptic, and behavioral functions.

INTRODUCTION

Brain-derived neurotrophic factor (BDNF) is a major regulator of nervous system function. Signaling through the TrkB receptor, BDNF facilitates neuronal differentiation, proliferation, and survival, as well as axonal and dendritic growth, synapse formation, and maturation [1–3]. BDNF is critical for various forms of synaptic plasticity [4–7], and is implicated in many behavioral functions, including multiple types of learning and memory. A significant challenge in studying BDNF function is the complex regulation of the Bdnf gene. Bdnf consists of nine 5’ non-coding exons, designated as I – IXa, each containing a unique promoter from which transcription is initiated [8, 9]. These non-coding exons are spliced to a 3’ common-coding exon (IX) [10], such that all Bdnf transcript variants produce the same BDNF protein.

Transcriptional and epigenetic regulation at Bdnf promoters tightly controls temporal, spatial, and activity-dependent Bdnf expression. Exposure to a wide variety of stimuli, which engage different epigenetic elements and transcription factors, drives transcription of individual Bdnf variants from their upstream promoters [11, 12]. For example, cis-elements within Bdnf promoter I facilitate AP-1-dependent transcription [11] while dynamic changes in DNA methylation control transcription from Bdnf promoter IV following defeat stress and antidepressant treatment [12]. Transcript-specific regulation of Bdnf occurs across multiple rodent models and in response to seizures [13, 14], ischemia [15], and stress [16]. Moreover, production of multiple Bdnf transcripts contributes to BDNF regulation by controlling its activity-dependent and brain region-specific expression, subcellular localization patterns, and its local translation and transcript stability [17, 18].

Despite the known functional roles of BDNF protein, how individual Bdnf transcripts differentially contribute to BDNF-dependent cellular, circuit, and behavioral functions remains unclear. To address this question, we established a transcript-specific CRISPR activation (CRISPRa) system capable of selectively upregulating Bdnf variants I and IV from their endogenous promoters using a catalytically-dead Cas9 (dCas9) protein fused to a strong tripartite transcriptional activator, VPR, comprised of VP64-p65-Rta [19, 20]. While both Bdnf I and IV promoters are responsive to neuronal activity [8, 13], their activation kinetics differ, and they respond to different transcriptional and epigenetic signaling machinery. Here, we used our recently developed CRISPRa system to better understand the functional significance and individual contribution of Bdnf transcript I versus IV on BDNF-dependent functions. We show that Bdnf I upregulation in hippocampal neurons was associated with gene expression changes responsible for dendritic growth, while Bdnf IV upregulation was associated with protein catabolism. Upregulation of Bdnf I, but not IV, caused changes in dendritic spine dynamics in hippocampal neurons, while upregulation of Bdnf IV (but not I) in the rat hippocampus impacted contextual fear expression. These data indicate that although Bdnf I and IV transcripts encode the same protein, selective overexpression of these transcripts causes distinct downstream effects on gene expression, cellular morphology, and behavior.

MATERIALS AND METHODS

Animals and cell culture

All experiments were approved by the University of Alabama at Birmingham Institutional Animal Care and Use Committee (IACUC). Male Sprague Dawley rats (90- to 120-day old, 250–350 g) were co-housed in pairs on a 12/12 h light/dark cycle with ad libitum food and water. Timed pregnant dams were individually housed until embryonic day 18 for hippocampal culture harvest [20, 21]. Cells were maintained in complete Neurobasal media for 11 – 14 d in vitro (DIV) with half media changes at DIV1, 4–5, 8–9, and 12.

CRISPR construct design and delivery

CRISRP/dCas9 and sgRNA construct design was previously described and all constructs are available on Addgene [20, 22]. Lentiviruses were produced in a sterile, BSL-2 environment [22]. Physical viral titer was determined using Lenti-X RT-qPCR Titration kit (Takara), and only viruses > 1 × 10¹⁰ GC/ml or > 1 × 10¹² GC/ml were used for cell culture or in vivo experiments, respectively. Stereotaxic surgery was performed as previously described [20]. Bilateral infusions of 1.5 μl (0.5 μl sgRNA and 1 μl dCas9-VPR viruses in sterile 1X PBS, each hemisphere) were directed to dorsal CA1 (AP: −3.3 mm, ML: ±2.0 mm, DV −3.1 mm from bregma). Additional details can be found in supplemental methods.

Spine morphology imaging and analysis

Automated image analysis was performed with Neurolucida 360 (2.70.1, MBF Biosciences, Williston, Vermont), as previously described [23]. After deconvolution, image stacks were imported into Neurolucida 360, the full dendrite length was traced with semi-automatic directional kernel algorithm. A blinded experimenter manually confirmed all assigned points and made any necessary adjustments. Each dendritic protrusion was automatically classified as a dendritic filopodium, thin spine, stubby spine, or mushroom spine [24]. Reconstructions of branched structure analysis were collected in Neurolucida Explorer (2.70.1, MBF Biosciences, Williston, Vermont). Spine density was calculated as the number of spines per 10 μm of dendrite length.

Behavior

Contextual fear conditioning (CFC) was conducted as described [25]. Rats were placed into the training chamber with a metal floor grid (Med Associates) and allowed to explore for 7 min, during which three electric shocks (1 s, 0.5 mA each) were administered every 2 min. Memory was tested at 1 h, 24 h, and 7 days after training. An open field arena (43 × 43 cm; Med Associates) was used to assess locomotor and anxiety behavior, as previously described [25]. One week following CFC, rats were placed in the open field arena and allowed to explore for 30 min. Distance traveled (in cm) and time spent in the center (s) were tracked and quantified using automated video tracking software (CinePlex Studio, Plexon Inc).

Additional details for all procedures can be found in supplemental methods.

RESULTS

Transcription of Bdnf variants I and IV is regulated by activity

Bdnf promoters respond differently to neuronal activity [10, 11, 13, 18]. Therefore, we first examined activity-dependent Bdnf transcript expression patterns in response to neuronal stimulation. DIV 11 rat hippocampal primary neuronal cultures were depolarized with 25 mM potassium chloride (KCl) for 1 – 4 hours and Bdnf transcripts were measured with RT-qPCR (Fig. 1a–b; primer sequences in Table S1). Expression of Bdnf I, II, IV, and VI, as well as total Bdnf IX, was altered by KCl depolarization, with peak changes occurring 3 hours post-stimulation (Fig. 1c). Two-way ANOVA revealed a significant main effect of KCl treatment, time point and interaction between the two for Bdnf I, II, IV, VI and IX transcripts. Sidak’s multiple comparisons post-hoc test showed that Bdnf I and IV are significantly increased 3 hours post-stimulation (p < 0.0001), which was also reflected by significant upregulation of total Bdnf IX at this time point (p < 0.0001) (Fig. 1c). Interestingly, Bdnf V did not respond to KCl stimulation at any time point as indicated by the lack of main effect of KCl treatment, time point, or interaction (p = 0.7595). Additionally, Bdnf III expression was significantly decreased at all time points, as indicated by the significant main effect of KCl treatment (p < 0.0001), but not time point (p = 0.6056). We did not detect transcript variants VII and VIII [10, 26]at baseline or after KCl stimulation (data not shown), and therefore these transcripts were not measured in subsequent experiments.

Figure 1.

Bdnf transcripts I and IV are most responsive to activity. a, Bdnf gene structure illustrating non-coding exons (I-IXa) and common-coding exon (IX) (left). Schematic of the Bdnf transcript variants containing non-coding exons I or IV spliced to the common-coding exon IX with short or long 3’ untranslated regions (UTRs) (right). b, Rat hippocampal neuron culture preparation. c, Expression of Bdnf I and IV transcript variants, as well as total Bdnf IX, peaks 3 h after depolarizing rat hippocampal cultures with 25 mM KCl (n = 6, two-way ANOVA; 3 h Bdnf I p < 0.0001; 3 h Bdnf IV p < 0.0001; 3. Bdnf IX p < 0.0001; Sidak’s multiple comparisons test). Expression of other Bdnf transcripts is up- or downregulated at different time points after KCl (n = 6, two-way ANOVA; 3 h. Bdnf II p = 0.0141; 3 h Bdnf III p < 0.0001; 3 h Bdnf VI p = 0.0185; Sidak’s multiple comparisons test). d, Stimulation of rat hippocampal cultures with 5 μM Gabazine, 25 mM KCl, 100 ng/ml recombinant BDNF, chemical LTP (200 nM NMDA; 50 μM forskolin; 0.1 μM rolipram), and 10 μM AMPA for 3 h upregulated the expression of select Bdnf transcript variants, with Bdnf I and Bdnf IV transcripts being most responsive (n = 6–12, one-way ANOVA; 5 μM gabazine F(7, 48) = 118.8, p < 0.0001; 100 ng/ml BDNF F(7, 88) = 99.66, p < 0.0001; 10 μM AMPA F(7, 48) = 9.947, p < 0.0001; chemical LTP F(7, 48) = 6.594, p < 0.0001). Blockade of neuronal activity with 1 μM tetrodotoxin (TTX) or 10 μM MK801 for 3 h. inhibited Bdnf expression, affecting mostly Bdnf I and Bdnf IV transcripts (n = 8, one-way ANOVA; 1 μM TTX F(7, 56) = 1.142, p = 0.3505; 10 μM MK801 F(7, 56) = 1.461, p = 0.2002). Data in panel c are expressed as mean ± SD. Individual comparisons; *p < 0.05, ^**p < 0.01, ^***p < 0.001, ^****p < 0.0001.

Because activity-dependent changes in Bdnf transcription vary in response to neuronal stimulation and inhibition protocols [10, 11, 13], we quantified the expression of Bdnf variants in response to multiple treatment protocols (Fig. 1d). DIV 11 rat hippocampal neuron cultures were treated with the GABA_A receptor antagonist gabazine (5 μM), KCl (25 mM), recombinant BDNF (100 ng/ml), AMPA (10 μM), the sodium channel blocker TTX (1 μM), and the NMDA receptor antagonist MK801 (10 μM), as well as a chemical LTP protocol (200 nM NMDA, 50 μM forskolin and 0.1 μM rolipram). RNA was isolated 3 hours after stimulation, and RT-qPCR revealed that patterns of Bdnf transcript expression were unique for different treatment protocols. For example, gabazine stimulation produced the strongest upregulation of total Bdnf IX levels (5.5-fold), with a 20-fold upregulation of Bdnf I and 5-fold upregulation of Bdnf IV (Fig. 1d). While recombinant BDNF and chemical LTP treatments only upregulated Bdnf I 4.6-fold and 2.8-fold, respectively, AMPA stimulation upregulated Bdnf IV by 5-fold. These results demonstrate that the pattern of stimulus-dependent Bdnf transcript variant expression is unique to different environmental stimuli and that Bdnf variants I and IV are most highly regulated by activity.

CRISPRa targeting of Bdnf I or Bdnf IV drives expression of the respective transcript and corresponding BDNF protein production

We recently developed a CRISPR activation (CRISPRa) system that successfully targets Bdnf I and Bdnf IV transcripts [20]. We deployed this system to better understand whether Bdnf transcripts possess unique biological functions (Fig. 2a). Single guide RNAs (sgRNAs) were designed complementary to 19–20 nucleotide sequences within Bdnf promoters I and IV less than 50 base pairs (bp) upstream from the transcription start sites (TSS) (Fig. 2b; Table S1). dCas9-VPR and sgRNAs were packaged into high-titer (10¹¹ GC/ml) lentiviral vectors and co-transduced into primary rat hippocampal cultures at DIV 4–5 (Fig. 2c). Robust transgene expression was verified by immunocytochemistry (ICC) with antibodies specific for mCherry (to visualize sgRNA) and FLAG (to visualize dCas9-VPR) at DIV 11 (Fig. S1).

Figure 2.

CRISPRa induction of Bdnf I or IV transcript variants upregulates BDNF protein levels and causes distinct changes in downstream gene expression. a, Schematic of the CRISPR/dCas9-VPR system and associated plasmids. b, Schematic of CRISPRa targeting of Bdnf promoters I and IV. c, Experimental timeline and viral transduction groups. d, Western blot for BDNF protein after CRISPRa targeting of Bdnf I, Bdnf IV, and Bdnf I & IV transcripts revealed a significant increase in BDNF protein levels in all conditions as compared to the non-targeting lacZ control (n = 7, one-way ANOVA, F(3, 24) = 6.696, p = 0.0019; Dunnett’s multiple comparisons test). e, Venn diagram illustrating 664 differentially expressed genes (DEGs) from RNA-seq after Bdnf I sgRNA targeting (red) and 2,842 DEGs after Bdnf IV sgRNA targeting (blue), as compared to lacZ sgRNA control. Right, correspondence heatmaps indicating the overlap between DEGs after Bdnf I upregulation and Bdnf IV upregulation. Simple linear regression analysis indicated that expression changes between manipulations were at least partially overlapping (significant derivation from zero, p < 0.0001 for all comparisons), with varying effect sizes (indicated by R² values). All data are expressed as mean ± SEM. Individual comparisons; ^**p < 0.01.

We previously confirmed selective upregulation of Bdnf transcripts I or IV with virtually no off-target CRISPRa effects [20]. CRISPRa produced a twofold increase in mature BDNF protein levels after either Bdnf I, Bdnf IV or Bdnf I & IV targeting as compared to a non-targeting negative control, lacZ gRNA (Fig. 2d). These data suggest that upregulation of either Bdnf transcript is equally capable of producing mature BDNF protein.

Bdnf I or Bdnf IV upregulation with CRISPRa causes unique gene expression changes

BDNF plays a key role in regulating synaptic plasticity by modulating downstream signaling cascades [11, 27, 28]. However, whether manipulation of distinct Bdnf transcripts differentially impacts downstream gene expression profiles has not been explored. We previously reported that CRISPRa of Bdnf I or Bdnf IV in hippocampal neurons resulted in 259 common differentially expressed genes (DEGs) when compared to a lacZ control (Fig. 2e) [20]. In agreement with Bdnf’s role in activity-dependent signaling cascades, we found both that Bdnf I and Bdnf IV upregulation resulted in regulation of immediate early genes (Fos, Arc, Egr1 and Egr3). However, we identified many DEGs that were unique to CRISPRa conditions - 405 DEGs specific to Bdnf I upregulation and 2,583 DEGs specific to Bdnf IV upregulation, compared to lacZ control (Fig. 2e) [20]. To determine whether these DEGs are unique to the upregulation of each Bdnf transcript or whether there is a correspondence between DEGs in one condition and the same genes in the other condition, we directly compared log₂ fold change values of Bdnf I only DEGs to the same genes in the Bdnf IV condition (Fig. 2e, top heatmap) and, vice versa, Bdnf IV only DEGs to the same transcripts in the Bdnf I sgRNA condition (Fig. 2e, bottom heatmap). Linear regression of log₂ fold change values between conditions revealed that there was a low (yet significant, p < 0.0001) correlation between Bdnf I and Bdnf IV manipulations. These data indicate that independent upregulation of either Bdnf I or Bdnf IV produces both unique as well as overlapping sets of gene expression changes.

DEGs in Bdnf I vs. Bdnf IV CRISPRa conditions are associated with unique gene ontology terms and neuropsychiatric disorders

Correspondence analysis (Fig. 2e) indicated at least partial overlap between Bdnf I or Bdnf IV CRISPRa targeting. However, closer inspection revealed that many genes were also oppositely regulated by these manipulations, suggestive of distinct transcriptional consequences of isoform-specific upregulation. To more definitively identify transcripts that were uniquely regulated following either Bdnf I or Bdnf IV CRISPRa, we next performed a new DESeq2 analysis to directly compare these two CRISPRa-targeting conditions (Bdnf I vs. Bdnf IV sgRNAs). This analysis revealed 1,483 genes selectively enriched after Bdnf I upregulation and 1,943 genes selectively enriched after Bdnf IV upregulation (Fig. 3a; Tables S2, S3). Gene ontology (GO) analysis identified some functional overlap between Bdnf I vs. IV - regulated genes, such as neuron differentiation, axon development, neuron projection development, and mRNA processing (Fig. 3b), suggesting that these cellular processes depend on expression of both Bdnf I and IV transcripts. However, each set of Bdnf I vs. IV - regulated genes was also associated with unique GO terms. Bdnf I-regulated genes belonged to GO categories associated with transcription, dendrite development, neurogenesis and differentiation of specific brain areas and cell types (Fig. 3b, top). Bdnf I-regulated genes were responsible for dendrite development, such as Shank1, Shank2, Ntrk2, and Nrep [29], and activity-dependent gene expression, such as Mecp2 [30] (Table S4).

Figure 3.

RNA-seq directly comparing Bdnf I vs. Bdnf IV sgRNA-targeting with CRISPRa revealed unique DEGs associated with different gene ontology terms and neuropsychiatric disorders. a, Volcano plot showing DEGs detected by DESeq2 in dCas9-VPR Bdnf IV vs. Bdnf I - targeted conditions. Standard cutoff point is represented by the horizontal dotted line (adjusted p < 0.05). Bdnf I - enriched (1,483 genes, red) and Bdnf IV - enriched (1,943 genes, blue) genes are indicated. b, Top significant gene ontology (GO) terms for Bdnf I - enriched and Bdnf IV - enriched genes, illustrating some overlapping (axon development, neuron projection development, neuron differentiation) but mostly unique GO terms for each gene set. c, DEGs after Bdnf I or Bdnf IV upregulation with CRISPRa overlapping clinical gene sets in the Harmonizome database.

Bdnf IV-regulated genes were associated with key cellular processes such as protein catabolism and metabolism, and regulation of translation (Fig. 3b, bottom). The top GO terms associated with Bdnf IV upregulation were ubiquitin-dependent protein catabolism and modification of proteins by small protein conjugation, such as ubiquitination, with specific enriched genes like Ubc and other ubiquitin-proteasome pathway-related genes (Table S5). Bdnf IV-regulated genes were known synaptic plasticity and neurodegeneration-associated marker genes, such as App, Sod1, Grin1, and Drd2 (Fig. 3a,b). These data imply that upregulation of distinct Bdnf transcripts may be associated with specific intracellular pathways that have physiological significance for cell-type and brain-area specific processes.

Bdnf expression is frequently altered in neurocognitive and neuropsychiatric disorders [31, 32]. Therefore, we investigated whether identified DEGs after Bdnf I vs. Bdnf IV upregulation overlapped with MESH term-defined clinical gene sets within the Harmonizome database (Harmonizome_CTD Gene-Disease Associated Dataset) [33]. We identified that genes differentially expressed after Bdnf I and Bdnf IV upregulation were associated with psychiatric disorders, including depression, anxiety, bipolar disorder, schizophrenia and mood disorders, as well as neurocognitive diseases, such as Alzheimer’s and Parkinson’s (Fig. 3c). In addition, we found that some neurological disorders were uniquely associated with DEGs in either Bdnf I or Bdnf IV-manipulated conditions. For example, Angelman, Rett and Asperger syndromes were uniquely associated with genes differentially expressed after Bdnf I upregulation (Table S6), while Huntington’s disease, Down syndrome, phobic disorders and frontal lobe epilepsy were associated with DEGs after Bdnf IV upregulation (Table S7) (Fig. 3c). While these neurocognitive disorders may have shared gene dysregulation profiles, unique Bdnf transcripts may have differential roles in disease etiology.

Bdnf I upregulation with CRISPRa increases mushroom dendritic spine density, length, and head diameter, and increases the complexity of dendritic arbors

Bdnf enhances dendritic branch length and number [3, 34, 35] and promotes synaptic spine formation and maturation [2, 36], while disrupting BDNF expression decreases spine density [37]. Previous studies reported that selective disruption of BDNF production from distinct promoters has opposing effects on CA1 and CA3 dendritic arbors and dendritic spine morphology [38]. Therefore, we investigated whether CRISPRa upregulation of Bdnf I and Bdnf IV contributes to dendritic spine density and arborization.

Rat hippocampal primary cultures were transduced with CRISPRa constructs on DIV 4–5 and transfected with a construct encoding Lifeact-GFP on DIV 12, a fluorescently-tagged small peptide which binds to intracellular f-actin [39] (Fig. 4a). Cells were fixed at DIV 14 and 63X high-resolution confocal laser scanning microscopy z-stack images were acquired for dendritic spine analysis. After maximum-intensity image deconvolution, three-dimensional digital reconstruction models of dendrites were quantified based on length and shape (Fig. 4b). Mushroom spine density was significantly increased after Bdnf I and Bdnf I & IV upregulation, as compared to a lacZ control (Fig. 4c, mushroom). Furthermore, mushroom spine volume, length and spine head diameter were increased specifically after Bdnf I, but not Bdnf IV, upregulation (Fig. 4d).

Figure 4.

Mushroom spine density, volume, length and head diameter as well as dendritic complexity are significantly increased after Bdnf I, but not Bdnf IV, upregulation. a, Experimental design timeline. b, Representative maximum-intensity high-resolution confocal microscope images after deconvolution (left) and corresponding three-dimensional digital reconstruction models of dendrites (right) after Bdnf I, Bdnf IV or Bdnf I & IV targeting with dCas9-VPR. Scale bar, 5 μm. Colors in digital reconstructions correspond to dendritic classes: green represents mushroom spines; blue, thin spines; brown, stubby spines; and yellow, filopodia. c, Dendritic spine density of thin, stubby, mushroom and filopodia spines, as well as all combined spines (total), illustrating a significant increase in mushroom spine density after CRISPRa targeting of Bdnf I and Bdnf I & IV, compared to a lacZ control (n = 15–24 dendrites from three independent cell cultures, one-way ANOVA, F(3, 71) = 3.878, p = 0.0126). d, The cumulative frequency distributions for mushroom spine volume (left), length (center), and head diameter (right) plotted for each Bdnf or lacZ dCas9-VPR-targeting condition. Bdnf I and Bdnf I & IV upregulation with CRISPRa significantly increases mushroom spine volume and length (n = 225–388 spines, one-way ANOVA, F(3, 1223) = 16.42, p < 0.0001) while Bdnf I upregulation significantly increases mushroom head diameter (n = 225–388 spines, one-way ANOVA, F(3, 1223) = 5.127, p = 0.0016). e, Schematic of the Sholl analysis using concentric circles starting at the soma in 5 μm increments (outlined by Lifeact-GFP fluorescence). f, Dendritic length in each concentric segment, as a function of the radius from the soma, illustrating that CRISPRa upregulation of Bdnf I and Bdnf I & IV significantly increases proximal dendrite length (n = 20–29 dendritic branches from three individual cultures, two-way ANOVA, F(50,4800) = 220.7, p < 0.0001). g, The number of dendritic intersections as a function of the radius from the soma, illustrating that CRISPRa upregulation of Bdnf I and Bdnf I & IV significantly increases the number of proximal dendritic intersections (n = 20–29 dendritic branches from three individual cultures, two-way ANOVA, F(49,4704) = 271.2, p < 0.0001). All data are expressed as mean ± s.e.m. Individual comparisons, *p < 0.05, ^**p < 0.01, ^****p < 0.0001.

To explore whether upregulation of Bdnf I or IV influences dendrite complexity, rat hippocampal primary neurons were transduced with CRISPRa constructs and transfected with Lifeact-GFP (Fig. 4b). Cells were fixed at DIV 14 and 20X high-resolution confocal laser scanning microscopy z-stack images were acquired. Neuronal 3D reconstructions for Sholl morphometry analyses with concentric circles were performed (Fig. 4e). CRISPRa upregulation of Bdnf I, but not Bdnf IV, significantly increased dendritic length (Fig. 4f) and the number of dendritic intersections at the 25–75 μm radius from the cell body, as compared to a lacZ control (Fig. 4g). These data illustrate that Bdnf I upregulation increases the length and complexity of dendritic arbors.

Upregulation of Bdnf IV, but not I, decreases freezing during contextual fear learning

BDNF plays a critical role in many forms of learning and memory, including the acquisition and extinction of contextual fear memories [40–43]. Therefore, we next investigated whether Bdnf I and IV differentially impact contextual fear learning. We bilaterally infused high-titer (10¹² GC/ml) lentiviruses expressing dCas9-VPR and sgRNAs for either Bdnf I, IV, I & IV or lacZ into the CA1 subregion of the dorsal hippocampus of adult rats (Fig. 5a). After a 12-day recovery period to allow for the stable integration and expression of the virally-delivered transgenes (Fig. 5b), rats underwent contextual fear conditioning (CFC) (Fig. 5c) as previously described [25]. Rats were handled for two days immediately prior to training and then were exposed to the training chamber for 2 min, followed by three 1 sec, 0.5 mA shocks delivered every 2 min, with a final exploration period of 1 min after the final shock. Rats were re-exposed to the same fear conditioning chamber 1 hour, 24 hours, and 7 days after the training for 7 min and time spent freezing was recorded.

Figure 5.

CRISPRa-mediated induction of Bdnf IV, but not Bdnf I, decreases time spent freezing following contextual fear conditioning (CFC). a-b, Bilateral lentiviral infusions targeting dorsal CA1 of the hippocampus were performed in adult male rats (n = 7–8 rats/condition). Two weeks after stereotaxic viral infusions, CFC and open field tests were carried out, followed by transcardial perfusions and immunohistochemistry (IHC) to verify sgRNA (mCherry, red) and dCas-VPR (FLAG, green) expression. Cell nuclei were stained with DAPI. Scale bar, 500 μm. Schematic of the target region was adapted from Paxinos and Watson. c, Schematic of fear acquisition (training) timeline. d, Comparison of freezing behavior at different timepoints before and after CFC. There were no significant differences between groups at baseline (prior to CFC). Animals in the Bdnf IV CRISRPa condition exhibited a significant decrease in time spent freezing as compared to Bdnf I CRISPRa condition when tested immediately after CFC (n = 7–8, one-way ANOVA; F(3,26) = 3.948). While no significant differences were observed one hour following CFC, Bdnf IV and Bdnf I & IV CRISPRa rats exhibited significantly lower time spent freezing when tested 1 d or 7 d after CFC compared to the lacZ controls (n = 7–8, one-way ANOVA; 1 d test: F(3,26) = 4.421; 7 d test: F(3,26) = 3.169). All data are expressed as mean ± s.e.m. Individual comparisons, *p < 0.05 and ^**p < 0.01. e, There were no significant differences in the total distance traveled or time spent in the center of the arena in the open field test.

Rats in all conditions had low baseline levels of freezing prior to shock presentation (Fig. 5d, left), and time spent freezing increased following training. Rats in the Bdnf IV-upregulated condition spent significantly less time freezing compared to rats in the Bdnf I-upregulated condition during the last minute of training after the final shock (Fig. 5d). While not statistically significant, Bdnf IV-upregulated rats spent less time freezing compared to lacZ and Bdnf I rats, when tested 1 hour after conditioning (Fig. 5d). When tested at 24 hours and 7 days, there was a significant reduction in time spent freezing in the Bdnf IV, as well as Bdnf I & IV-upregulated conditions (Fig. 5d). To rule out that changes in time spent freezing were due to changes in activity or anxiety, we quantified total locomotion and center time in an open field test 7 days following the last fear conditioning test. There were no significant differences in total distance traveled or time spent in the center of the arena between any of the Bdnf-targeting conditions or the lacZ control (Fig. 5e). Taken together, these data suggest that upregulation of Bdnf IV, but not I, decreases expression of contextual fear.

DISCUSSION

Prior efforts to study the role of BDNF in the brain includes infusion of recombinant BDNF [41, 44–46], region- and cell type-specific conditional knockdowns [16, 47], and mouse models lacking Bdnf production from distinct promoters [48]. To expand our understanding of Bdnf gene regulation in neuronal plasticity and behavior, we developed and validated tools for precise manipulation of transcript-specific Bdnf expression using CRISPRa/i from the endogenous loci [20, 49]. Using CRISPRa, we investigated distinct roles of activity-dependent Bdnf transcripts I and IV in cultured hippocampal neurons and in the rat dorsal hippocampus. Our findings in vitro suggest that overexpressing Bdnf I vs IV has differential effects on downstream gene expression as well as on dendrite length and dendritic spine morphology. In vivo, we find that overexpressing Bdnf I vs IV differentially impacts fear expression in response to CFC.

While upregulation of Bdnf I and IV similarly increased BDNF protein levels (Fig. 2d), there were differential effects on gene expression (Fig. 2e). These results support the possibility that each splice variant is associated with unique downstream molecular signaling pathways. Bdnf I-regulated genes were associated with dendrite development, neurogenesis and differentiation of specific brain areas and cell types (Fig. 3b, top), results that support the finding that mushroom spine phenotypes (increased densities, spine volume, length and head diameter) were increased and that more complex dendritic arbors developed in response to Bdnf I, but not Bdnf IV, upregulation (Fig. 4c–g). Moreover, genes enriched after Bdnf I upregulation in the “dendrite development” GO category (including Cdkl5, Mecp2, Disc1, Shank1, Shank2, and Ntrk2) (Table S4) were also identified within the Harmonizome clinical gene set database to be associated with Angelman, Rett and Asperger syndromes (Fig. 3c), which are associated with dendrite and dendritic spine dysgenesis [50, 51]. These findings provide novel insights on the role of Bdnf I dendritic spine development in neurodevelopmental disease regulation.

In contrast, Bdnf IV-regulated genes belonged to GO categories that control basic cellular functions, such as protein catabolism and metabolism (Fig. 3b, bottom). Interestingly, Bdnf IV upregulation was strongly associated with enrichment of genes encoding proteins in the ubiquitin-proteasome pathway (UPP), which mediates regulated degradation of intracellular proteins [52, 53]. Genes encoding ubiquitin, Ubc, proteasomal alpha subunits, Psma1-6, proteasomal beta subunits, Psmb3, 4, 7, 9, and proteasomal ATPases, Psmc2-5, were enriched after Bdnf IV upregulation (Table S5). The UPP is necessary for induction of the transcription and translation-dependent late phase of LTP (L-LTP), as well as learning and memory consolidation [53, 54], and chemically blocking the UPP prevents induction of L-LTP and Bdnf transcription [55, 56]. While further investigation is needed, it is plausible that upregulating Bdnf IV triggers an autoregulatory loop engaging further transcription of UPP components and enhanced UPP activity.

Both Bdnf I and Bdnf IV-regulated genes were associated with similar neuropsychiatric and neurodegenerative disorders, including depression, anxiety, Alzheimer’s and Parkinson’s diseases, disorders where a relationship with Bdnf dysregulation is well-documented [57, 58]. However, only Bdnf IVregulated genes were associated with Huntington’s disease (App, Drd2), Down syndrome (Sod1), phobic disorders (Casp3), and frontal lobe epilepsy (Chrna4, Chrnb2), also previously tied to Bdnf dysregulation [31, 59]. Together with the RNA-seq data, these findings suggest that although the downstream transcriptional signatures caused by Bdnf I and Bdnf IV upregulation are similar, there are many unique downstream transcriptional readouts associated with each splice variant that may be important for neuronal function and disease progression.

While the importance of Bdnf expression in CFC has been well established, contributions of individual Bdnf transcripts is not clear. Mutant mice in which production of BDNF from promoter IV is disrupted show enhanced fear and resistance to fear extinction [42], which can be rescued by selective over-expression of BDNF in hippocampal neurons that project to the prefrontal cortex [43]. Supporting these previous studies, our data show that upregulating Bdnf IV, but not I, in dorsal CA1 of the rat decreases fear expression 24 hours and 7 days after CFC. In our study, Bdnf IV-targeted animals trended towards lower freezing levels even 1 hour after fear conditioning, and hence further studies are necessary to disentangle whether the decreases in fear expression at later time points are due to a general decrease in expression of fear, deficits in fear memory acquisition, or enhanced fear extinction. Further studies deciphering the role of the Bdnf IV transcript in the regulation of learning and memory and CFC are necessary and, are now possible, due to the novel CRISPR tools described here and in previous manuscripts [49, 60–62].

In summary, our study is the first to report the functional impacts of manipulating different transcript variants of Bdnf both in rat hippocampal neurons and in rat hippocampus. CRISPRa upregulation of Bdnf I promotes formation of mature dendritic spines and extended dendritic arbors, while upregulation of Bdnf IV mediates expression of contextual fear. Future studies delineating distinct roles for Bdnf transcript variants in the nervous system will increase our understanding of Bdnf’s involvement in synaptic plasticity, neuropsychiatric and neurodegenerative diseases, as well as our ability to modulate Bdnf gene expression for therapeutic applications.