Abstract
Neurotrophins like BDNF have a key role in the proper functioning of the central nervous system, influencing numerous processes like memory formation and behavior. An imbalance in BDNF levels can lead to a wide range of diseases, including depression and neurodevelopmental disorders. While the potential therapeutic effects of BDNF are well-recognized, there is a knowledge gap in understanding the mechanisms governing BDNF expression levels. Here, we focused on the regulation of Bdnf gene expression in response to different stimuli, specifically studying the effects of neuronal activity and BDNF-TrkB signaling on Bdnf transcription in cultured neurons from rats of either sex. We used in vitro DNA pulldown combined with mass spectrometry to determine transcription factors that interact with the Bdnf promoters upon different stimuli and validated numerous known regulators, such as USF and AP1 family, and novel candidate regulators using reporter assays. We show that the USF family of transcription factors is specifically recruited after membrane depolarization, whereas the AP1 family participates in Bdnf regulation only after BDNF-TrkB signaling. We further describe ATF2, MYT1L, and EGR family as novel regulators of Bdnf expression by demonstrating their direct binding to Bdnf promoters using chromatin immunoprecipitation assays both in vitro and in vivo, showing their functional role in Bdnf gene expression and ultimately identifying their regulatory cis-elements in Bdnf promoters. Furthermore, our results show competition between ATF2, CREB, and AP1 family in regulating Bdnf levels. Collectively, our results provide insight into the regulation of Bdnf expression upon different stimuli.
Significance Statement
Membrane depolarization and neurotrophin BDNF (brain-derived neurotrophic factor) signaling via its receptor TrkB (tropomyosin receptor kinase B) are critical processes for proper neuronal functions. Here, we studied how these two stimuli regulate the expression of two main Bdnf transcripts—Bdnf exon I- and IV-containing transcripts. Our results reveal a remarkable overlap of regulators that are recruited to both Bdnf promoters I and IV and after both stimuli. Overall, our results shed light to the complex regulation of Bdnf expression highlighting a dynamic interplay of cooperative and competitive mechanisms among transcription factors. Understanding the regulatory mechanisms beyond single transcription factors and considering the combinatorial effects could pave the way for specifically modulating Bdnf levels as therapeutic interventions.
Introduction
Neurotrophins are secreted signaling molecules that play crucial roles in neuronal physiology (Chao, 2003; Skaper, 2018). Within the central nervous system, BDNF functions in many processes, such as neuronal differentiation, synapse formation and synaptic plasticity, formation of long-term memory, maturation of neuronal circuits, and modulation of behavior (Park and Poo, 2013; Wang et al., 2022). Altered BDNF levels are linked to several central nervous system diseases, including depression, bipolar disorder, schizophrenia, addiction, and neurodevelopmental disorders (Autry and Monteggia, 2012). Furthermore, various antidepressants and mood stabilizers, prescribed for neuropsychiatric disorders, affect BDNF signaling (Casarotto et al., 2021; Wang et al., 2022). Elevating BDNF levels has been shown to alleviate symptoms in mouse models of Rett syndrome (Chang et al., 2006), Huntington disease (Xie et al., 2010), Alzheimer's disease (Choi et al., 2018), Smith–Magenis syndrome (Javed et al., 2022, 2023), and Prader–Willi syndrome (Queen et al., 2022, 2023). While the significance of BDNF in the central nervous system and its therapeutic potential are widely recognized, gaps in understanding the BDNF regulation and signaling have hampered the advancements in the field.
Here, we aimed to explore stimulus-induced regulation of Bdnf gene to better understand how cells express Bdnf and hence contribute to the development and maintenance of the nervous system. The Bdnf gene contains multiple promoters that direct the expression of distinct 5′ exons, giving rise to transcripts encoding the same protein (Aid et al., 2007; Pruunsild et al., 2011). Many transcription factors have been shown to regulate Bdnf expression in response to various stimuli (reviewed in West, 2008; West et al., 2014), with cAMP response element-binding protein (CREB) being the best-described regulator of activity-dependent Bdnf gene expression (Shieh et al., 1998; Tao et al., 1998; Tabuchi et al., 2002; Hong et al., 2008; Benito et al., 2011; Pruunsild et al., 2011; Palomer et al., 2016; Tai et al., 2016). In this study, we compared the regulation of Bdnf gene expression after KCl treatment and BDNF treatment. KCl treatment is a widely used method to mimic neuronal activity in vitro by depolarizing the neuronal membrane, causing Ca2+ influx via L-type voltage-sensitive calcium channels, and initiating signaling cascades and gene expression patterns similar to those after neuronal activity in vivo (reviewed in Rienecker et al., 2020). Membrane depolarization that exceeds threshold potential initiates a cascade of events at axon terminals, culminating in neurotransmitter release into the synaptic cleft. Next, released neurotransmitters bind to receptors on the postsynaptic membrane, leading to membrane depolarization, intracellular signaling, and modulation of gene expression, thereby completing the synaptic signaling process. BDNF is secreted in response to neuronal activity and, through binding to its receptor TrkB, initiates gene expression programs that strengthen synapses and neuronal circuits (Park and Poo, 2013). It has been well described that BDNF signaling also induces the expression of Bdnf itself in an autocrine or paracrine manner, forming a self-amplifying autocrine loop (Wibrand et al., 2006; Yasuda et al., 2007; Zheng and Wang, 2009; Nakajima et al., 2015; Tuvikene et al., 2016; Esvald et al., 2020). The use of these two stimuli allows us to investigate the molecular mechanisms underlying membrane depolarization and BDNF-TrkB signaling in a controlled in vitro environment.
Neuronal activity and BDNF-TrkB signaling induce different sets of genes (Ibarra et al., 2022), induce Bdnf expression with different dynamics (bell-shaped curve vs continuous increase in expression levels), and convey Bdnf induction (at least partially) via different transcription factors (Pruunsild et al., 2011; Tuvikene et al., 2016, 2021; Esvald et al., 2020, 2022). Furthermore, Bdnf induction requires de novo protein synthesis after BDNF-TrkB signaling (Esvald et al., 2020) but depends less on protein synthesis after neuronal activity (Hughes et al., 1993; Lauterborn et al., 1996; Sano et al., 1996; Castrén et al., 1998; Tao et al., 1998). Thus, we hypothesized that different transcription factors are involved in the induction of Bdnf after each stimulus.
Utilizing a hypothesis-free in vitro DNA pulldown approach followed by mass spectrometry (Tuvikene et al., 2021; Esvald et al., 2022), we identified transcription factors binding to Bdnf promoters I and IV upon membrane depolarization and after BDNF-TrkB signaling. Our findings corroborate previous studies and introduce novel regulators of Bdnf gene. Notably, we show that in the regulation of Bdnf gene, the USF family participates specifically after membrane depolarization, while the AP1 family regulates Bdnf expression only after BDNF-TrkB signaling. We further investigated a selection of the novel candidate regulators using coexpression with Bdnf promoter constructs in cultured neurons. We then focused on the ATF2, MYT1L, and EGR family transcription factors in the regulation of Bdnf gene. We validated their binding to Bdnf promoters using chromatin immunoprecipitation assay, determined their functional effect using overexpression, or impaired their function with dominant-negative proteins or knockdown of gene expression with CRISPR interference or RNA interference system. Finally, we identified ATF2-regulated cis-elements and novel functional cis-elements for EGR and MYT1L transcription factors in Bdnf proximal promoters. For ATF2, we further show competition with CREB to binding CRE elements in Bdnf promoters and competition with AP1 family members to dimerize with cJUN and bind AP1-2 element in Bdnf promoter I. Collectively, our data reveal complex interplay between transcription factors in regulating the levels of Bdnf after different stimuli.
Materials and Methods
Rat primary cortical neuron culture
All animal procedures were performed in compliance with the local ethics committee. Cortical neuron cultures were generated from Sprague Dawley rat male and female pups of the same litter at embryonic days 20–21 as previously described in Esvald et al. (2020, 2022) with minor modifications. Neurons were plated at density 75,000 neurons/cm2 in Neurobasal A medium (NBA, Invitrogen) containing 1× B27 supplement (Invitrogen), 1 mM ʟ-glutamine (Invitrogen), and 100 μg/ml Primocin (InvivoGen) or in Dulbecco's modified Eagle’s medium (DMEM with high glucose, PAN Biotech) containing 10% fetal bovine serum (FBS, PAN Biotech), and the media was changed to supplemented NBA 2 h after plating. Neurons were grown in supplemented NBA at 37°C in 5% CO2 environment. At 2 days in vitro (DIV) and at 5 DIV, half of the medium was replaced with fresh supplemented NBA. To inhibit the proliferation of non-neuronal cells, 10 μM 5-fluoro-2′-deoxyuridine (FDU, Sigma-Aldrich) was added to the media from 2 DIV.
In vitro DNA pulldown and mass spectrometry
Cortical neurons were grown on 145 mm culture dishes (Greiner). At 7 DIV, 24 h before cell lysis, 1 μM tetrodotoxin (TTX, Tocris Bioscience) was added to the neurons, and at 8 DIV the neurons were treated for 2 h with 100 ng/ml recombinant mature BDNF protein (Icosagen, catalog #P-105-100) or with a mixture of 25 mM KCl (AppliChem) and 5 μM d-(2R)-amino-5-phosphonovaleric acid (d-APV, Cayman Chemical Company).
The nuclear lysates were prepared with high salt extraction as previously described (Lahiri and Ge, 2000; Wu, 2006) with minor modifications (Tuvikene et al., 2021; Esvald et al., 2022). Briefly, after treatment, neurons were washed one time with phosphate-buffered saline (PBS), lysed in 12 ml ice-cold cytoplasmic lysis buffer [10 mM HEPES, pH 7.9 (adjusted with NaOH), 10 mM KCl, 1.5 mM MgCl2, 0.5% NP-40, 300 mM sucrose, 1× cOmplete Protease Inhibitor Cocktail (Roche), and phosphatase inhibitors as follows: 5 mM NaF (Fisher Chemical), 1 mM β-glycerophosphate (Acros Organics), 1 mM Na3VO4 (ChemCruz), and 1 mM Na4P2O7 (Fisher Chemical)] on ice for 10 min and collected with a cell scraper. The suspension was transferred to a 15 ml tube and gently vortexed, and the nuclei were pelleted with centrifugation at 2,600 × g at 4°C for 1 min. The cytoplasmic fraction was aspirated, and the nuclear pellet was resuspended in 200 µl ice-cold nuclear lysis buffer [20 mM HEPES, pH 7.9, 420 mM NaCl, 1.5 mM MgCl2, 0.1 mM EDTA, 2.5% glycerol, 1× cOmplete Protease Inhibitor Cocktail (Roche), and phosphatase inhibitors] and transferred to a new Eppendorf tube. To extract proteins from the nuclei, the suspension was rotated at 4°C for 30 min and to clear the lysate centrifuged at 10,400 × g at 4°C for 10 min. The supernatant was transferred to a new Eppendorf tube, and the protein concentration was measured using BCA Protein Assay Kit (Pierce). The lysates were aliquoted for single use, snap-frozen in liquid nitrogen, and stored at −80°C.
The biotinylated Bdnf promoter regions (described in Esvald et al., 2022) were produced using primers (Microsynth) listed in Tuvikene (2024) and HOT FIREPol polymerase (Solis BioDyne). PCR products were purified using DNA Clean and Concentrator-100 kit (Zymo Research) according to the manufacturer's recommendations with 1:5 ratio of PCR solution and DNA binding buffer. The concentration of the DNA was measured with NanoDrop 2000 spectrophotometer (Thermo Scientific).
In vitro pulldown was performed as described in Tuvikene et al. (2021) and Esvald et al. (2022) with the following minor modifications: 75 pmoles of biotinylated DNA was bound to 50 µl Pierce Streptavidin Magnetic beads; 340–400 µg of nuclear proteins was used for the pulldown; the eluted proteins were transferred to a Protein LoBind Tube (Eppendorf) and snap-frozen in liquid nitrogen. Mass spectrometric analysis was performed with nano-LC-MS/MS using Q Exactive Plus (Thermo Scientific) at Proteomics core facility at the University of Tartu, Estonia, as described in Mutso et al. (2018) using label-free quantification instead of SILAC and Rattus norvegicus reference proteome for analysis. The following mass spectrometric analysis and custom R-script are described in Tuvikene et al. (2021). Gene ontology categories were accessed from www.geneontology.org at 16.03.2020. All proteins detected in mass spectrometry are listed in Extended Data Figure 1-1 and promoter-specific transcription factors (defined as at least a 1.5-fold enrichment to one promoter respective to the other) are listed in Extended Data Figure 1-2.
Transfection and luciferase reporter assay
All used plasmids are listed in Tuvikene (2024) with references to Addgene plasmid number, publications if previously published, or a coding sequence for newly generated constructs. All coding sequences were cloned to pRRL backbone under the control of human PGK promoter. For studying ATF2 mechanism, pQM vectors expressing CREB, FOS, or JUN under PGK promoter were transfected along with pLV-hUBc-FLAG-ATF2 plasmid or empty pLV vector as control. Mutagenesis of the cis-elements in reporter plasmids was carried out as described in Pruunsild et al. (2011), and the generated mutations are listed in Tuvikene (2024).
For transfection and luciferase reporter assays, cortical neurons were grown on 48-well cell culture plates (Greiner). At 6–7 DIV, neurons were transfected using 200 ng of DNA and Lipofectamine 2000 (Invitrogen) at a DNA-Lipofectamine 2000 ratio 1:3 in prewarmed NBA media without supplements. The amount of DNA was divided as follows: (1) 100 ng of Bdnf promoter construct, 80 ng of effector plasmid, and 20 ng of a normalizer plasmid pGL4.83-mPGK-hRLuc, (2) in the experiments using reporters with mutated cis-elements 180 ng of Bdnf promoter-encoding plasmid and 20 ng of pGL4.83-mPGK-hRLuc plasmid, or (3) for studies of ATF2 mechanism, 90 ng of Bdnf promoter, 45 ng of either effector plasmids, and 20 ng of a normalizer plasmid pGL4.83-mPGK-hRLuc. The neurons were transfected for 3–4 h on an orbital shaker at 37°C in 5% CO2 environment. Transfection was terminated and 1 μM TTX (Tocris Bioscience) was added by replacing the medium back to the collected media that was kept in the incubator for the transfection time.
At 7 DIV for ATF2 mechanism studies or at 8 DIV for all other reporter experiments, neurons were treated for 8 h with recombinant mature BDNF (50 ng/ml, PeproTech, catalog #450-02, or 100 ng/ml, Icosagen, catalog #P-105-100) or with a mixture of 25 mM KCl (AppliChem) and 5 μM d-APV (Cayman Chemical Company) and lysed with passive lysis buffer (Promega). Luciferase signals were measured with Dual-Glo Luciferase Assay System (Promega) using GENios Pro plate reader (Tecan).
Background-corrected Firefly luciferase signals were normalized to background-corrected Renilla luciferase signals. If the transfection was performed in duplicate wells, the averages of duplicate wells were calculated. Data was log-transformed, mean-centered, and autoscaled before statistical analysis.
Lentivirus-mediated experiments
All used overexpression and dominant-negative constructs, guide RNA (gRNA) targeting sequences for CRISPR, and short hairpin RNA (shRNA) targeting sequences are listed in Tuvikene (2024). For CRISPR interference and activation experiments, the EGFP-encoding sequence was removed from pLV-hUbC-dCas9-KRAB-T2A-GFP and pLV-hUbC-VP64-dCas9-VP64-T2A-GFP (published in Tuvikene et al., 2021), and the generated pLV-hUbC-dCas9-KRAB and pLV-hUbC-VP64-dCas9-VP64 plasmids were used for lentivirus production. For co-IP experiments, pLV-3xFLAG-ATF2 or empty pLV vectors were used.
Lentiviral particles were produced as follows: HEK293FT cells (Thermo Fisher Scientific) were grown on cell culture dishes or flasks and at 80–90% confluency transfected with lentiviral transfer vector, psPAX2, and pVSVG plasmids at ratios 2:1.5:1, respectively. Cells were transfected using polyethyleneimine (PEI, Sigma) at DNA:PEI ratio of 1:2. Lentivirus was produced in HEK293FT cells growing in lentiviral production media [DMEM with high glucose (PAA Laboratories), 10% FBS (Pan Biotech), 1 mM sodium-pyruvate, 1× MEM Nonessential Amino Acids (Capricorn Scientific), 20 mM HEPES, pH 7.4] and the media was collected the two following days to obtain secreted viral particles. Purification of viral particles was performed using Speedy Lentivirus Purification solution (Abm) and precipitated viral particles were resuspended in 1× PBS, aliquoted, snap-frozen in liquid nitrogen, and stored at −80°C.
The relative titers were calculated based on provirus incorporation rate in cortical neurons using qPCR. The Woodchuck hepatitis virus posttranscriptional regulatory element (WPRE) and unrelated region (URR) primers are listed in Tuvikene (2024). For functional experiments, equal amounts of control and effector-encoding viral particles were used.
For experiments using lentiviral transductions, cortical neurons were grown on 12-well cell culture plates (Greiner) and transduced at the day of plating (0 DIV). Exceptionally for A-ATF2, ATF2, Zn-EGR3, and EGR3 overexpression, the neurons were transduced at 2 DIV (Figs. 4C, 6D, respectively).
RNA extraction and RT-qPCR
For the analysis of mRNA levels, the neurons were treated at 7 DIV with 1 μM TTX (Tocris Bioscience) and at 8 DIV for 3 h with 50 ng/ml recombinant BDNF (PeproTech, catalog #450-02) or with a mixture of 25 mM KCl (AppliChem) and 5 μM d-APV (Cayman Chemical Company).
Neurons were lysed in RLT lysis buffer with 1% β-mercaptoethanol (VWR Life Sciences), and RNA was purified with Mini Spin columns (EconoSpin) and RNeasy Mini Kit (Qiagen) according to the recommendations from the manufacturer with on-column digestion of genomic DNA using RNase-Free DNase Set (Qiagen). RNA concentration was measured with BioSpec-Nano (Shimadzu) or NanoDrop 2000c (Thermo Scientific) spectrophotometer. Equal amounts of RNA were taken to synthesize complementary DNA (cDNA) using SuperScript IV Reverse Transcriptase (Thermo Fisher Scientific) and oligo-dT20 primer and random hexamer primers. qPCR was performed using HOT FIREPol EvaGreen qPCR Mix Plus (Solis BioDyne) and LightCycler 480 Instrument II (Roche). Primers used for qPCR are listed in Tuvikene (2024). All mRNA expression levels were normalized to HPRT1 expression levels, and the results were log-transformed, mean-centered, and autoscaled before statistical analysis.
Western blot
For Western blot, the neurons were grown on 24-well plates (Greiner). At 7 DIV 1 μM TTX was added, and at 8 DIV the neurons were lysed in 1× Laemmli buffer (without bromophenol blue and β-mercaptoethanol) and lysates were heated at 95°C for 5 min. The protein concentration was measured with Pierce BCA Protein Assay Kit (Thermo Scientific) according to the manufacturer's protocol. Next, bromophenol blue and β-mercaptoethanol (VWR Life Sciences, 5% final concentration) were added and the lysate was heated again at 95°C for 5 min and finally centrifuged at 16,000 × g for 1 min to remove insoluble material.
A total of 20 µg of total protein was analyzed using SDS-PAGE and transferred to a PVDF membrane using Trans-Blot Turbo Transfer system (Bio-Rad). Membranes were blocked for at least 1 h at room temperature in 5% skimmed milk in TBST buffer (1× Tris-buffered saline, pH 7.4, and 0.1% Tween 20). For Western blot, all antibodies were diluted in 2% skimmed milk in TBST and incubated overnight at 4°C. The following antibodies were used: anti-ATF2 (1:1,000, Cell Signaling Technology, catalog #35031S, lot #1), anti-EGR1 (1:1,000, Cell Signaling Technology, catalog #4153S, lot #5), anti-MYT1L (1:1,000, Millipore, catalog #ABE2915-100ug, lot #Q3323437), and anti-FLAG M2-HRP antibody (1:5,000, Sigma-Aldrich, #A8592). The secondary antibody anti-mouse IgG conjugated with horseradish peroxidase (HRP; Thermo Scientific, 1:5,000) or anti-rabbit IgG-HRP (Thermo Scientific, 1:5,000) was diluted in 2% milk in TBST and incubated for at least 1 h at room temperature. For probing multiple proteins from the same membrane, the previous signal was quenched by incubating the membrane with 30% H2O2 for 20 min at 37°C, followed by washing three times with TBST buffer, reblocking, and finally reprobing with anti-V5 antibody (1:5,000, Thermo Fisher Scientific, #R960-25).
Chemiluminescence signal was produced with SuperSignal West Femto or Atto Maximum Sensitivity Substrate (Thermo Scientific) and measured using ImageQuant Las 4000 imaging system (GE Healthcare Life Sciences). For loading control, all membranes were stained with Coomassie solution (0.1% Coomassie Brilliant Blue R-250 Dye, 25% ethanol, 7% acetic acid), followed by washes with destaining solution (30% ethanol, 10% acetic acid) and rinsing with tap water. The membranes were imaged using ImageQuant Las 4000 (GE Healthcare Life Sciences).
Chromatin immunoprecipitation assay
For ChIP assays, cortical neurons were grown on 10 cm or 145 mm cell culture dishes (Greiner). At 7 DIV 1 μM TTX was added and at 8 DIV neurons were treated for 2 h with 50 ng/ml recombinant BDNF (PeproTech, catalog #450-02) or with a mixture of 25 mM KCl (AppliChem) and 5 μM d-APV (Cayman Chemical Company).
Neurons were fixed directly in the growth media with final concentration of 1% formaldehyde (Cell Signaling, catalog #12606) by adding 1/9th volume of 10× cross-linking solution (containing 11% formaldehyde, 100 mM NaCl, 0.5 mM EGTA, 50 mM HEPES, pH 8.0) and incubating for 10 min on the orbital shaker. The following protocol to prepare nuclear lysates was performed as described in Esvald et al. (2022) (developed based on Vierbuchen et al., 2017) using L1 and L2 buffers. Here, instead of L3, the nuclear pellet was resuspended in high SDS lysis buffer [1% SDS, 10 mM EDTA, 50 mM Tris-HCl, pH 8.0, 1× cOmplete protease inhibitor cocktail (Roche)] for sonication. Chromatin was sonicated using the BioRuptor Pico device (Diagenode) with sonication beads (Diagenode, catalog #C01020031) for 10 cycles (30 s on and 30s off). Protein concentration was measured with Pierce BCA Protein Assay Kit (Thermo Scientific) and 250–300 µg of nuclear protein was used per immunoprecipitation (IP). Then, 5 µl of anti-ATF2 (Cell Signaling Technology, catalog #35031S, lot #1), anti-EGR1 (Cell Signaling Technology, catalog #4153S, lot #5), or anti-MYT1L antibody (Millipore, catalog #ABE2915-100µg, lot #Q3323437) was added to the lysate and rotated at 4°C. The next day, preblocked (with BSA) 50 µl Dynabeads Protein G (Invitrogen, #10003D) was used per IP to capture the antibody-chromatin complexes. The dilution of SDS for IP, wash, and elution steps were performed as previously described (Esvald et al., 2020). RNase, proteinase K treatment, and subsequent purification of the DNA with QIAquick PCR Purification Kit (QIAGEN) were also performed as previously described in Esvald et al. (2020, 2022).
For ChIP assay using tissue samples, three cortices from 8-d-old or two hippocampi from adult (8-month-old) Sprague Dawley rats of either sex per replicate were dissected and snap-frozen in liquid nitrogen. Tissues harvested from rats of different litters were considered biological replicates. The ChIP assay was performed following the protocol described in Esvald et al. (2022), with a minor modification—the lysate was diluted twofold with dilution buffer [1% Triton X-100, 150 mM NaCl, 2 mM EDTA, 20 mM Tris-HCl, pH 8.0, 1× cOmplete protease inhibitor cocktail (Roche)] before IP.
The binding of a transcription factor to specific DNA regions was analysed with qPCR. All qPCR reactions were performed in triplicates with 1× LightCycler 480 SYBR Green I Master kit (Roche) on LightCycler 480 PCR Instrument II (Roche). The primers measuring binding to Bdnf promoters are listed in Tuvikene (2024). DNA enrichment was calculated relative to the respective DNA levels in input sample. For statistical analysis, data was log-transformed, mean-centered, and autoscaled. Data is presented as binding relative to unrelated region (URR).
Co-immunoprecipitation
Primary cortical neurons were grown on a 145 cm dish and transduced at 0 DIV with lentiviruses encoding either 3× FLAG-ATF2 or control lentiviral particles (pLV). At 7 DIV, 1 µM TTX (Tocris Bioscience) was added, and at 8 DIV the neurons were treated for 2 h with a mixture of 25 mM KCl (AppliChem) and 5 μM d-APV (Cayman Chemical Company) or with 100 ng/ml recombinant mature BDNF protein (Icosagen, catalog #P-105–100).
Cells were washed with ice-cold 1× PBS and lysed in 1.2 ml of co-IP lysis buffer [(20 mM Tris-HCl, pH 7.5, 400 mM NaCl, 1 mM EDTA, 0.5% NP-40, 1× cOmplete Protease Inhibitor Cocktail (Roche), and phosphatase inhibitors 5 mM NaF (Fisher Chemical), 1 mM β-glycerophosphate (Acros Organics), 1 mM Na3VO4 (ChemCruz), and 1 mM Na4P2O7 (Fisher Chemical)]. The cells were collected with a cell scraper, lysate was incubated at 4°C for 30 min while rotating, and sonicated three times for 2 s with Torbeo Ultrasonic probe sonicator (36810-series, Cole-Parmer). Next, lysates were centrifuged at 16,000 × g at 4°C for 15 min to remove cell debris, supernatant was transferred to a new tube, and an aliquot for input sample was taken. Protein concentration was measured using Pierce BCA Protein Assay Kit (Thermo Scientific).
For IP, 100 µl of Dynabeads Protein G (Invitrogen #10003D) per sample was taken, washed three times with 0.05% Tween 20–1× PBS solution, and then incubated with 10 µg of mouse anti-FLAG (Sigma, #F1804) antibody in 0.05% Tween 20–1× PBS at 4°C for 4 h while rotating. For antibody immobilization, beads were washed three times with 0.05% Tween 20–1× PBS, one time with 0.2 M triethanolamine solution, and cross-linked with freshly made 5.4 mg/ml dimethyl pimelimidate in 0.2 M triethanolamine solution at room temperature for 30 min. Next, the beads were washed with 50 mM Tris-HCl, pH 7.5, for pH normalization, three times with 0.05% Tween 20–1× PBS solution and resuspended in IP buffer [20 mM Tris-HCl, pH 7.5, 200 mM NaCl, 1 mM EDTA, 0.5% NP-40, 1× cOmplete Protease Inhibitor Cocktail (Roche), and phosphatase inhibitors as above].
Protein lysates were adjusted to equal protein quantity and concentration of 1 mg/ml, and FLAG-cross-linked Dynabeads were added and incubated overnight at 4°C with rotation. The next day, beads were washed three times with IP buffer and finally eluted with 0.1 M glycine, pH 2, at room temperature for 10 min while rotating. The eluate was transferred to a new Protein LoBind tube and 0.17 M Tris-HCl, pH 8.5, was added to normalize pH. The eluates were snap-frozen in liquid nitrogen and stored at −80°C.
For Western blot, Laemmli buffer with final concentration of 5% β-mercaptoethanol was added to input and eluate samples and analyzed using the following primary antibodies: anti-cJUN (1:1,000, Cell Signaling Technology, catalog #9165), anti-cFOS (1:1,000, Cell Signaling Technology, catalog #2250), anti-CREB (1:1,000, Cell Signaling Technology, catalog #4820), and anti-FLAG (1:5,000, Sigma, catalog #F1804).
Electrophoretic mobility shift assay
HEK293 cells were grown on 6 cm dishes in MEM (Thermo Fisher Scientific) supplemented with 10% fetal bovine serum (Pan Biotech), 100 U/ml penicillin, and 0.1 mg/ml streptomycin (Thermo Fisher Scientific). The cells were transfected ∼24 h after seeding at 50–70% confluency with 6 µg DNA using polyethylenimine (PEI) at 1:2 ratio. For cell lysis, HEK293 cells were washed with ice-cold 1× PBS and lysed in 150 µl sonication buffer [20 mM HEPES, pH 7.9, 25% glycerol, 500 mM KCl, 1.5 mM MgCl2, 0.4 mM EDTA, 1 mM EGTA, 5 mM DTT, 0.5 mM PMSF, 1× cOmplete Protease Inhibitor Cocktail (Roche), and 1× PhosSTOP Phosphatase Inhibitor Cocktail (Roche)]. The cells were collected with a cell scraper, incubated on ice for 15 min, and then sonicated for 3 s using a Torbeo Ultrasonic probe sonicator (36810-series, Cole-Parmer). Lysates were centrifuged at 16,000 × g at 4°C for 15 min. Supernatant containing native protein lysate was transferred to a new tube, aliquoted for single use, snap-frozen in liquid nitrogen, and stored at 80°C. Overexpression of the transcription factors was verified using Western blot.
To generate EMSA probes, sense and antisense oligonucleotides were annealed in annealing buffer (50 mM NaCl, 1 mM EDTA, 1× PNK-A buffer, 10 μM sense and antisense oligonucleotides). Sequences for all used oligos are listed in Tuvikene (2024). For annealing, the oligos were incubated at 95°C for 5 min and then slowly cooled down at rate 1°C per min.
For electrophoretic mobility shift assay (EMSA), 2 μl of native protein lysate and 1 pmol annealed biotinylated oligo were used. EMSA reaction was performed in 20 µl, containing the following: 1 μg poly(dI-dC) (Sigma-Aldrich), 0.1 mg/ml BSA (Thermo Fisher Scientific), 1× cOmplete Protease Inhibitor Cocktail (Roche), 1× binding buffer (10 mM HEPES, pH 7.5, 0.5 mM EDTA, 2 mM MgCl2, 0.05% NP-40, 4% Ficoll-400). The reactions were incubated at room temperature for 20 min. For competition EMSAs, 10 pmol of unlabeled oligos was added to the reaction 5 min before adding the biotinylated probe. For supershift assays, 0.5 µl of ATF2 antibody (Cell Signaling Technology, catalog #35031) or 0.5 µl of cJUN antibody (Cell Signaling Technology, catalog #9165) was added to the reaction and incubated for 30 min.
Electrophoresis was performed in 0.5× TBE buffer using a 4.5% nondenaturing polyacrylamide gel containing 0.25× TBE, 0.01% NP-40, and 2.5% glycerol. Before loading the samples, the gel was prerun at 100 V for 10 min, then the samples were loaded, and the gel was run at 100 V for ∼55 min. Proteins were then transferred to a nylon membrane using wet transfer at 380 mA for 40 min in fresh 0.5× TBE solution and proteins were fixed to the membrane with UV light. Next, membrane was incubated in streptavidin-HRP (1:10,000, Columbia Biosciences, #HRP-2212) solution in 5% SDS-1×PBS at room temperature for 30 min. Finally, membrane was washed one time with 0.5% SDS–1× PBS, one time with 0.1% SDS–1× PBS, and three times with 1× PBS, and chemiluminescence signal was produced with SuperSignal West Atto Ultimate Sensitivity Substrate (Thermo Scientific) and measured using ImageQuant Las 4000 imaging system (GE Healthcare Life Sciences).
Experimental design and statistical analysis
Power calculations were not done, necessary sample size was not estimated, and number of biological replicates was chosen based on what is generally acceptable in the field. Samples were not randomized, and the analysis was not blinded. The number of independent biological replicates (cultures obtained from different litters) is written in the figure legends. Statistical significance was calculated using two-tailed paired t test on log-transformed, mean-centered, and autoscaled data (according to Willems et al., 2008). All hypotheses that were statistically analyzed were specified before conducting the experiments, and thus p values were not corrected for multiple comparisons to preserve statistical power. For graphical representation, transformed and scaled means and means ± standard error of the mean (SEM) were back-transformed to linear scale with error bars representing upper and lower limits of back-transformed mean ± SEM. Back-transformed means and all individual data points are shown in each figure.
Results
Determining novel regulators of Bdnf promoters using in vitro DNA pulldown
To find novel stimulus-specific regulators of Bdnf, we performed in vitro DNA pulldown assays coupled with mass spectrometry. This method has previously successfully identified the binding of transcription factor hnRNPD0B to an obesity-related single nucleotide polymorphism in the BDNF gene (Mou et al., 2015), revealed regulators of an intronic enhancer region of the Bdnf gene (Tuvikene et al., 2021), and pinpointed brain region-specific regulators of Bdnf promoters (Esvald et al., 2022).
Here, we focused on Bdnf promoters I and IV as loss of different Bdnf transcripts have distinct consequences (Hong et al., 2008; Hill et al., 2016; Maynard et al., 2016, 2018; McAllan et al., 2018; Hallock et al., 2020; You et al., 2020). Furthermore, not all neurons express Bdnf exon I and IV simultaneously (Maynard et al., 2020), implying cell-type and promoter-specific regulatory mechanisms. Our previous results have shown that after BDNF-TrkB signaling Bdnf promoter I and IV have distinct regulators, i.e., AP1 and CREB family transcription factors, respectively (Tuvikene et al., 2016; Esvald et al., 2020). Therefore, we hypothesized that different transcription factors regulate the activity of Bdnf promoters after different stimuli.
We performed in vitro DNA pulldown of nuclear lysates of cortical neurons treated with either KCl or BDNF (Fig. 1A). In in vitro DNA pulldown experiments performed with KCl- and BDNF-treated neurons (and respective untreated control neurons), we detected 235 and 266 different proteins, respectively. The majority of the identified factors were general nucleic acid-binding proteins, for example, histones (Extended Data Fig. 1-1), and ∼25% of all the identified proteins were classified as transcription factors based on gene ontology annotation. In total, we identified 58 and 64 transcription factors in experiments with KCl- and BDNF-treated neurons, respectively. As expected, we detected a major known regulator of Bdnf expression, cyclic AMP response element-binding protein (CREB), binding to both Bdnf promoters I and IV but not in a differential manner. This is in agreement with published data as it is known that CREB can regulate the activity of both Bdnf promoters I and IV after neuronal activity (Shieh et al., 1998; Tao et al., 1998; Tabuchi et al., 2002; Hong et al., 2008; Benito et al., 2011; Pruunsild et al., 2011; Palomer et al., 2016; Tai et al., 2016). We also detected a strong preference of upstream stimulatory factor (USF) family for Bdnf promoter IV (Extended Data Fig. 1-2), although it has been reported to regulate both Bdnf promoters I and IV (Tabuchi et al., 2002; W. G. Chen, et al., 2003b; Pruunsild et al., 2011). In agreement with our previous results (Esvald et al., 2022), we determined the binding of BCL11A to Bdnf promoter I and RAI1, SATB2, and TCF4 to Bdnf promoter IV.
Determining transcription factors that bind Bdnf promoter I and IV in vitro. A, Schematic representation of the in vitro DNA pulldown assay to determine transcription factors binding to Bdnf promoters I and IV (schematics created with BioRender.com). Cortical neurons were treated at 8 days in vitro (DIV) for 2 h with KCl or BDNF. Native nuclear proteins were extracted and incubated with the biotinylated Bdnf promoter regions bound to the streptavidin-conjugated magnetic beads. Unbound proteins were washed and DNA with bound factors was eluted from streptavidin. The bound proteins were identified with LC-MS/MS (Extended Data Fig. 1-1). B, Transcription factors that showed promoter-specific binding to rat Bdnf promoters I and IV. Specific binding was defined based on a label-free quantification signal ratio of 1.5 or greater between the two Bdnf promoters in the respectively treated lysate (Extended Data Fig. 1-2). Red asterisks mark transcription factors that have a predicted binding site in the respective Bdnf promoter region according to JASPAR database (Castro-Mondragon et al., 2022).
Figure 1-1
Lists of all peptides, proteins and transcription factors detected in mass-spectrometric analysis in in vitro DNA pulldown experiments. Download Figure 1-1, XLSX file.
Figure 1-2
Calculations of relative enrichments of the detected transcription factors on Bdnf promoters in in vitro DNA pulldown experiments. Download Figure 1-2, XLSX file.
To identify regulators specific to distinct Bdnf promoters, we calculated a preference index, measuring the relative enrichment of transcription factor on one promoter compared with the other. We set a threshold of 1.5, meaning that we considered factors enriched at least 1.5-fold more on one promoter than the other as specific to that promoter (Fig. 1B, exact ratios presented in Extended Data Fig. 1-2). Of the Bdnf promoter I-specific regulators, activator protein 1 (AP1) family members were detected on promoter I, with a preference index of 5–25 for FOS and 2.6–2.8 for JUN after treatment. AP1 family members have previously been recognized as regulators of Bdnf promoter I after BDNF-TrkB signaling (Tuvikene et al., 2016). Intriguingly, our analysis revealed many C2H2 zinc finger transcription factors binding to Bdnf promoter I, including members of the early growth response (EGR) family (preference index ∼3), Specificity protein (SP) family (detected exclusively on promoter I), and Krüppel-like factors (KLF; preference indices varying from 1.7 to 11.7). At the same time, we detected a substantial number of heterogeneous nuclear ribonucleoproteins (hnRNPs) binding to Bdnf promoter IV (preference indices varying from 3 to 70). Many of the determined transcription factors have a predicted binding site in respective Bdnf promoter region (Fig. 1B) based on JASPAR database (Castro-Mondragon et al., 2022). Collectively, our in vitro DNA pulldown experiments were successful in not only corroborating known regulators but also in suggesting novel regulators of Bdnf expression.
Validating the functionality of novel regulators of Bdnf transcription with reporter assays
We next selected a subset of transcription factors that bind to Bdnf promoters in the in vitro promoter pulldown assay for further investigation. To investigate their roles, we generated vectors for either overexpression of the transcription factors or their dominant-negative proteins. These vectors were then cotransfected with Bdnf promoter constructs into cortical neurons. The used dominant-negative proteins, A-FOS, A-ATF2, and A-USF, contain an acidic extension and function by dimerizing with the respective leucine zipper transcription factor, thereby inhibiting DNA binding through interaction with the basic region (Olive et al., 1997; Ahn et al., 1998; Qyang et al., 1999).
In untreated neurons the most pronounced effects were observed after overexpression of EGR1, EGR3, KLF12, ZNF148, and MAX, which increased the activity of both Bdnf promoters I and IV in the luciferase reporter assay (Fig. 2). In contrast, the overexpression of PURA, PURB, and DLX1 decreased the activity of both Bdnf promoters in untreated neurons. Notably, overexpressing A-USF and MYT1L resulted in divergent effects on Bdnf promoters I and IV in untreated neurons, with A-USF and MYT1L enhancing the activity of promoter I while reducing the activity of promoter IV.
Transcription factors determined with the in vitro DNA pulldown regulate the activity of Bdnf promoters I and IV in reporter assay. Rat cultured cortical neurons were transfected at 7 DIV with rat Bdnf promoter I and promoter IV constructs along with vectors for overexpression of dominant-negative proteins (denoted with A in A-ATF2, A-FOS, and A-USF) or overexpression of different transcription factors. At 8 DIV, neurons were left untreated (0 h), treated with 25 mM KCl (with 5 μM d-APV), or 50 ng/ml BDNF for 8 h. The promoter activity was determined by measuring luciferase activity. Data is shown as relative to the luciferase activity in respectively treated cells transfected with respective promoter construct and EGFP. Numbers above the columns indicate the average of four to eight independent experiments (n = 4–8) and all data points are shown with dots. Error bars represent SEM. Statistical significance was calculated relative to the activity of respective promoter in respectively treated cells transfected with EGFP (black asterisks) or between the activity of different Bdnf promoters in respectively treated neurons overexpressing the same effector (red asterisks). # ≤ 0.1, *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001 (paired two-tailed t test).
In depolarized neurons, the overexpression of EGR1 and MAX significantly increased the activity of Bdnf promoter I, with EGR1 also increasing Bdnf promoter IV activity (Fig. 2). Conversely, overexpression of PURB and DLX1 decreased the activity of both Bdnf promoters after membrane depolarization. Remarkably, following membrane depolarization, the overexpression of A-FOS, A-ATF2, and A-USF had contrasting effects on the activity of Bdnf promoters—A-FOS and A-ATF2 inhibited the activity of promoter I, while increasing that of promoter IV, and A-USF slightly increased the activity of promoter I and decreased the activity of promoter IV.
Following BDNF treatment, almost all studied factors influenced the activity of Bdnf promoters I and IV. The biggest effect was observed with the overexpression of EGR1, KLF12, ZNF148, and MAX, which all increased the activity of both Bdnf promoters. In contrast, overexpression of PURB and DLX1 again reduced the activity of both promoters (Fig. 2). Overexpression of A-FOS, A-ATF2, A-USF, and EGR3 had an opposite effect on the activity of Bdnf promoters after BDNF treatment. Specifically, A-FOS, A-ATF2, and EGR3 decreased the activity of promoter I while enhancing the activity of promoter IV, whereas A-USF increased the activity of promoter I and decreased that of promoter IV.
Overall, the majority of the studied transcription factors affected the activity of Bdnf promoters in the same direction following both KCl and BDNF treatments. This consistency could be explained by shared downstream signaling pathways, such as the MAPK/ERK pathway, serving as a convergence point for both stimuli. Notable exceptions were observed with EGR3, KLF12, HMG20A, and MYT1L—overexpression of these factors decreased the respective promoter activity after KCl treatment but increased it following BDNF treatment. Interestingly, we did not identify any transcription factors that exhibited the opposite effect, i.e., decreasing the promoter activity after BDNF treatment and increasing it following KCl treatment. Collectively, our findings suggest that the transcription factors identified in the in vitro DNA pulldown assays can regulate the activity of Bdnf promoters, exhibiting both stimulus- and promoter-specific responses.
Investigating the regulators of Bdnf transcription in the endogenous context
We have previously analyzed some of the investigated factors in the endogenous context after BDNF-TrkB signaling and reported that AP1 family regulates the expression of Bdnf exon I (Tuvikene et al., 2016), while ATF2, USF family, CEBPα, nor CEBPβ did not have a significant role in regulating total Bdnf levels after BDNF-TrkB signaling (Esvald et al., 2020). Notably, of these transcription factors, ATF2 and AP1 family have been reported to participate in the regulation of Bdnf intronic enhancer region in reporter assays (Tuvikene et al., 2021). To investigate the role of these factors in Bdnf gene regulation at the transcript level after membrane depolarization and to compare their role after BDNF-TrkB signaling, we used lentivirus-mediated overexpression of dominant-negative factors for these transcription factors.
We observed a major effect of A-ATF2 and A-FOS overexpression on the basal levels of Bdnf transcripts (Fig. 3). Remarkably, overexpression of A-ATF2 led to a 48-fold increase in the basal levels of Bdnf exon I transcripts and enhanced the KCl- and BDNF-induced levels of Bdnf exon I transcripts by ∼2.8–3.5-fold. A-FOS overexpression resulted in a 10-fold and fivefold increase in the basal levels of Bdnf exon I and exon IV transcripts, respectively. Furthermore, overexpression of A-FOS remarkably decreased the BDNF-induced levels of Bdnf exon I transcripts, but not after KCl treatment, and did not alter the stimulus-induced levels of Bdnf exon IV transcripts. In contrast, overexpression of A-USF decreased the levels of Bdnf exon I and exon IV transcripts only after membrane depolarization and not after BDNF-TrkB signaling. The overexpression of A-CEBPβ slightly increased the basal levels of Bdnf (although not statistically significantly) but did not impact the expression of Bdnf mRNA levels after stimuli. Collectively, our results confirm that the AP1 family regulates BDNF-TrkB signaling-dependent expression of Bdnf exon I, and USF family regulates the levels of both Bdnf exon I- and exon IV-containing transcripts after KCl treatment. Importantly, our results show that the effects of AP1 and USF families are stimulus specific, with AP1 family involved only after BDNF treatment and USF family only after KCl treatment.
AP1 family, USF family, and ATF2 are involved in the regulation of Bdnf gene transcription. Rat cultured cortical neurons were transduced with lentiviral particles encoding the indicated dominant-negative transcription factors or EGFP as a control. At 8 DIV, the neurons were treated with 25 mM KCl (with 5 μM d-APV) or 50 ng/ml BDNF for 3 h. The expression levels of Bdnf transcripts containing either exon I or IV and total Bdnf mRNA levels were measured using RT-qPCR. The expression level of the respective transcripts in untreated neurons transduced with EGFP-encoding lentiviruses was set as 1. All data points are shown as dots and each biological replicate is denoted with the same color as shown in the legend. Numbers above the columns indicate the average of three independent experiments (n = 3). Error bars indicate SEM. Statistical significance was calculated relative to the levels of the respective transcripts in EGFP-overexpressing cells at respective treatment. #p ≤ 0.1, *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001 (paired two-tailed t test).
ATF2 is a novel regulator of basal and stimulus-dependent expression of Bdnf exon I and IV transcripts
Next, we wanted to further investigate the role of ATF2 (activating transcription factor 2, also known as CRE-BP1) in Bdnf gene regulation. First, we investigated whether ATF2 directly binds Bdnf promoter regions in the endogenous context using chromatin immunoprecipitation (ChIP) assay (Fig. 4A). Our ChIP-qPCR revealed that ATF2 binds both Bdnf promoter I and IV and that the binding of ATF2 slightly increased after both stimuli. We also noted a stronger enrichment at Bdnf promoter I compared with Bdnf promoter IV following BDNF treatment (Fig. 4A). To determine whether ATF2 protein levels are increased upon stimuli, which could explain the increase in binding to Bdnf promoters, we performed Western blot assay. We determined that ATF2 protein levels were stable after BDNF treatment but rather exhibited a slight decrease after KCl treatment (Fig. 4B, similar effect was seen in two independent experiments), while the mRNA levels showed slight drop mainly after BDNF treatment (Fig. 4D,E, left panel). We also noted a slight decrease in the apparent molecular weight of ATF2 after both stimuli (Fig. 4B), suggesting the presence of posttranslational modifications. ATF2 phosphorylation on multiple sites in response to stimulus by different kinases (reviewed in Watson et al., 2017), such as JNK (Gupta et al., 1995; Kirsch et al., 2020), MAPK kinases (Ouwens et al., 2002; Kirsch et al., 2020), or protein kinase C (Yamasaki et al., 2009), has been previously described. It is plausible that membrane depolarization and TrkB signaling results in posttranslational modifications of ATF2; however, the exact modifications are currently not known and require further studies. Our results imply that the increased binding of ATF2 to Bdnf promoters after stimulus is rather a sign of changed affinity of ATF2 to DNA or increased accessibility of Bdnf promoters and is not caused by increased expression levels of ATF2.
ATF2 regulates basal and stimulus-dependent expression levels of Bdnf. Rat cultured cortical neurons were treated at 8 DIV for the time indicated with 25 mM KCl (and 5 μM d-APV) or 50 ng/ml BDNF. A, ChIP-qPCR assay using anti-ATF2 antibody. Data is shown as enrichment to Bdnf promoter I and IV (pI and pIV, respectively) relative to binding to the unrelated region (URR) in untreated cells. B, Western blots verifying the overexpression of V5-ATF2 and dominant-negative protein A-ATF2 and the effect of CRISPR interference (using dCas9-KRAB) and activation (using VP64-dCas9-VP64) on ATF2 protein levels. Coomassie staining is shown as loading control. C–E, The neurons were transduced with lentivirus particles encoding (C) V5-ATF2, A-ATF2, or EGFP as control; (D) dCas9-KRAB together with lentiviruses encoding either negative guide RNA (neg gRNA) or a guide RNA directed to the Atf2 promoter region (ATF2 gRNA); or (E) VP64-dCas9-VP64 together with neg gRNA or ATF2 gRNA. The expression levels of Atf2 and Bdnf transcripts containing either exon I or IV and total Bdnf mRNA levels were measured using RT-qPCR. The expression level of respective transcripts in untreated neurons transduced with EGFP-encoding lentivirus particles or untreated neurons transduced with negative guide RNA with dCas9-KRAB or VP64-dCas9-VP64 were set as 1. All data points are shown as dots and each biological replicate is denoted with the same color as shown in the legend. Numbers above the columns indicate the average of three or four independent experiments (A, n = 3; C, n = 4, D, E, n = 3). Error bars indicate SEM. Statistical significance was calculated relative to the binding to URR (black asterisks) or between Bdnf promoters (red asterisks) in respectively treated cells in ChIP-qPCR assay (A) and in RT-qPCR experiments relative to the levels of respective transcripts in EGFP-overexpressing cells (C), neurons transduced with negative guide RNA with dCas9-KRAB (D) or VP64-dCas9-VP64 (E) at respective treatments. # ≤ 0.1, *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001 (paired two-tailed t test).
Next, we performed lentivirus-mediated overexpression of ATF2 and its dominant-negative form, A-ATF2 (Fig. 4B,C). Interestingly, overexpression of ATF2 did not affect the expression of Bdnf transcripts (Fig. 4C). In contrast, we found that overexpression of A-ATF2 increased the basal levels of Bdnf exon I transcripts ∼41-fold and KCl- and BDNF-induced levels ∼12- and ∼2-fold, respectively (Fig. 4C). Furthermore, overexpression of A-ATF2 increased basal levels of Bdnf exon IV transcripts ∼5-fold, and KCl- and BDNF-induced levels ∼1.3-fold and ∼1.2-fold, respectively. Finally, to elucidate the specific role of endogenous ATF2, we employed CRISPR interference and activation systems to specifically modulate ATF2 expression levels (Fig. 4B,D,E). Silencing of Atf2 decreased Atf2 mRNA levels ∼3-fold and increased the expression levels of Bdnf exon I transcripts ∼29-fold and Bdnf exon IV transcripts ∼3-fold in untreated neurons (Fig. 4D). Atf2 knockdown also increased stimulus-induced levels of Bdnf exon I ∼2.5-fold (Fig. 4D). Our results demonstrate that the effects of A-ATF2 on Bdnf gene regulation were specific to ATF2. Finally, the CRISPR activation of Atf2 increased Atf2 mRNA levels threefold but did not affect basal nor stimulus-induced levels of Bdnf transcripts (Fig. 4E). Collectively, our data show that ATF2 is a novel regulator of Bdnf transcription, affecting the expression of Bdnf exon I more than exon IV.
ATF2 regulates the activity of Bdnf promoters as homodimers via CRE sites or as heterodimers with cJUN via AP1 site
ATF2 belongs to the ATF/CREB basic leucine zipper family of transcription factors and forms homodimers or heterodimers with cJUN (Hai and Curran, 1991; Lau and Ronai, 2012) or CREB (Abdel-Hafiz et al., 1993; Lau and Ronai, 2012). To elucidate which combinations of AP1 family members (cJUN or cFOS) or CREB family member CREB could regulate the activity of Bdnf proximal promoters I and IV with ATF2, we first performed luciferase reporter assays and overexpressed these transcription factors in cortical neurons. Our results show that ATF2 overexpression alone does not have remarkable effect on the promoter activity (Fig. 5A), in agreement with our results showing that ATF2 overexpression does not notably affect Bdnf mRNA levels (Fig. 4E). In contrast, ATF2 overexpression significantly potentiated the activity of both Bdnf promoter I and IV in untreated cells when coexpressed with cJUN, up to ∼3.4-fold. Slight synergistic effect was also observed with concurrent overexpression of ATF2 and cFOS, while a slight negative effect was seen in untreated cells when ATF2 was overexpressed along with CREB. Interestingly, after both membrane depolarization and BDNF-TrkB signaling, ATF2 and cJUN together had a much weaker potentiating effect on promoter I activity than cJUN alone, whereas ATF2 and cJUN together had a stronger potentiating effect on Bdnf promoter IV than cJUN alone.
ATF2 regulates the activity of Bdnf proximal promoter regions as a heterodimer with cJUN transcription factor. A, Rat cultured cortical neurons were transfected at 6 DIV with rat Bdnf promoter I and promoter IV constructs along with expression vectors for ATF2, cJUN, cFOS, or CREB. At 7 DIV, neurons were left untreated (0 h), treated with 25 mM KCl (with 5 μM d-APV), or 100 ng/ml BDNF for 8 h. The promoter activity was determined with luciferase reporter assay. Data is shown relative to the promoter activity in EGFP-transfected neurons. Numbers above the columns indicate the average of three independent experiments (n = 3) and all data points are shown with dots. Error bars represent SEM. Statistical significance was calculated between the activity of Bdnf promoter in respectively treated neurons overexpressing the indicated transcription factor with or without ATF2 (Ctrl and ATF2, respectively). B, Co-immunoprecipitation (co-IP) assay from rat cultured cortical neurons transduced with FLAG-ATF2-encoding or control lentiviruses and treated at 8 DIV for 2 h with 25 mM KCl (and 5 μM d-APV) or 100 ng/ml BDNF. Co-IP was performed with anti-FLAG antibody and representative Western blot figures using anti-cJUN, anti-cFOS, and anti-CREB antibodies are shown. C, The indicated transcription factors were overexpressed in HEK293 cells and native protein lysates were prepared for electrophoretic mobility shift assay (EMSA). Western blots confirm the expression of the studied transcription factors. D, Schematical representation of Bdnf promoter I and IV, indicating the positions of known cis-elements (shown with boxes). Identical nucleotides between the human and rat promoter regions are shown as dots. DNA sequences that were used as probes in EMSA are highlighted with dark green and named after the core binding site. E, F, EMSA assessing the binding of ATF2, cJUN, cFOS, and CREB to Bdnf promoter I and IV probes. The specific complex is denoted with an arrow on the right. G, Competition EMSA using the indicated probes, with or without a 10-fold excess of unlabeled wild-type oligo (10× WT) or oligo with mutated core site (10× mut) for competition. H, Supershift assays using probes bound by ATF2, with native lysates overexpressing ATF2 and c-JUN, and incubated with anti-ATF2 and anti-c-JUN antibodies. Supershifted complexes containing ATF2 homodimers and ATF2-cJUN heterodimers are shown with arrows (black and gray arrow, respectively). #≤ 0.1, *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001 (paired two-tailed t test). AP1, AP1 family-binding element; BHLHB2-RE, BHLHB2 response element; CaRE, calcium response element; CRE, cAMP response element; NFkB-RE, NFkB response element; UBE, USF-binding element; PasRE, NPAS4-binding element.
While luciferase assays reflect the overall transcriptional activation capabilities of the studied transcription factors, they do not mechanistically demonstrate whether ATF2 forms heterodimers or merely competes for binding to the same regulatory elements. To address this, we next studied the dimerization partners of ATF2 using co-immunoprecipitation (co-IP) assays. We transduced cortical neurons with FLAG-ATF2-encoding lentiviruses, treated the neurons with KCl or BDNF, and performed co-IP using anti-FLAG antibody (Fig. 5B). Our co-IP results show that ATF2 interacts with cJUN, whereas we detected no interaction with cFOS or CREB.
We next sought to determine which cis-elements in Bdnf promoters are bound by ATF2. ATF2 is generally known to bind the CRE element (Maekawa et al., 1989; Castro-Mondragon et al., 2022) and can compete with CREB for DNA binding (Niwano et al., 2006). However, together with cJUN, ATF2 can bind CRE and AP1 elements (Hai and Curran, 1991). Therefore, we hypothesized that the regulatory sites involved could be AP1 or CRE-elements in Bdnf promoters. To test this, we performed electrophoretic mobility shift assays (EMSA). We overexpressed ATF2, cJUN, cFOS, and CREB in HEK293 cells, prepared native protein lysates (Fig. 5C), and assessed their binding to known AP1 sites (AP1-1 and AP1-2) and two CRE sites (one in each promoter; Fig. 5D) in EMSA. Our results show that ATF2 binds to the AP1-2 but not AP1-1 site in Bdnf promoter I, and CRE sites in both promoter I and IV, with the strongest shift detected using the promoter I CRE probe (Fig. 5E,F). Notably, ATF2 binding to AP1-2 probe increased when coexpressed together with cJUN and slightly decreased to both CRE sites-containing probes in the presence of CREB (Fig. 5E). We then performed competition assays and confirmed that ATF2 specifically binds to the AP1-2 and CRE sites and not the flanking DNA regions (Fig. 5G). Finally, we investigated whether the complexes contain ATF2 homodimers or heterodimers with cJUN by using ATF2 and cJUN overexpression protein lysates and performed supershift assays with anti-ATF2 and anti-cJUN antibodies on AP1-2 and CRE site-containing probes (Fig. 5H). Both ATF2 and cJUN antibodies caused supershift on all the used probes, suggesting that both ATF2 homodimer and ATF2-cJUN heterodimers bind these elements. To unequivocally demonstrate that ATF2–cJUN heterodimers bind to the studied probes, we performed supershift assays using both antibodies simultaneously. This resulted in an even more supershifted band corresponding to the heterodimer complex. Interestingly, our supershift assay results suggest that heterodimer formation is more favorable at the AP1-2 site, whereas the CRE elements in promoters I and IV preferentially bind ATF2 homodimers. Our results indicate that the increased binding of ATF2 to Bdnf promoters following stimulation in ChIP assays (Fig. 4A) is likely driven by the formation of higher-affinity ATF2-JUN dimers compared with ATF2 homodimers. Collectively, our results show that ATF2 forms heterodimers with cJUN to regulate the activity of Bdnf promoter I via AP1-2 and CRE sites and Bdnf promoter IV via the CRE element.
MYT1L is a novel repressor of Bdnf exon I transcription
Based on our in vitro DNA pulldown assay and luciferase reporter experiments, we next focused on MYT1L as a possibly stimulus- and promoter-specific regulator of Bdnf. MYT1-like (MYT1L or neural zinc finger 1, NZF1), MYT1 (or NZF2), and MYT3 (or NZF3 or ST18) are three CCHHC-type zinc finger transcription factors that form the MYT family (reviewed in J. G. Kim and Hudson, 1992; Y. Jiang et al., 1996; Yee and Yu, 1998; Besold and Michel, 2015) in which expression levels in the brain peak during late embryogenesis and the levels of MYT1L remain the highest (reviewed in Matsushita et al., 2014; J. Chen et al., 2022). MYT1L is known for repressing the expression of non-neuronal genes in neurons (Mall et al., 2017) and MYT1-mediated repression of target genes promotes neurogenesis (Vasconcelos et al., 2016). The repressor functions of MYT1 and MYT1L are mediated by recruitment of SIN3B and histone deacetylases (Romm et al., 2005; Mall et al., 2017; J. Chen et al., 2023). While the role of MYT1L as a transcriptional repressor is well established, some evidence suggests it may also function as a transcriptional activator. Specifically, increased expression of certain genes has been observed after MYT1L knockdown (Kepa et al., 2017; J. Chen et al., 2021), and reporter assays have shown that the N-terminal part of MYT1L (but not MYT1) can activate gene expression (Manukyan et al., 2018). However, no clear mechanism for this activation has been described, and the majority of gene expression and histone modification data support MYT1L primary role as a transcriptional repressor.
It has been shown that overexpression of MYT1 increases total Bdnf mRNA levels in the hippocampus (Bahi and Dreyer, 2017) and decreases Bdnf and Ntrk2 (TrkB) mRNA levels in the nucleus accumbens (Chandrasekar and Dreyer, 2010). In addition, MYT1 seems to bind Bdnf +3 kb enhancer region and Bdnf promoter VI in neural stem cell line NS5 (data from Vasconcelos et al., 2016). Loss of MYT1L function causes numerous phenotypes, including obesity, hyperactivity, and intellectual disability (De Rocker et al., 2015; Blanchet et al., 2017; J. Chen et al., 2021; Coursimault et al., 2022; Wöhr et al., 2022; Weigel et al., 2023), resembling those associated with abnormal BDNF levels. Here, we determined MYT1L binding to Bdnf promoter I by our in vitro DNA pulldown assay (Fig. 1) and showed that overexpression of MYT1L increased the activity of only Bdnf promoter I in untreated neurons and after BDNF-TrkB signaling, but not after membrane depolarization (Fig. 2). Furthermore, MYT1 and MYT1L bind the same cis-element, AAGTT (Vasconcelos et al., 2016; Mall et al., 2017), and a deletion in the Bdnf promoter I region that increases the basal activity of Bdnf promoter I (Tabuchi et al., 2002) contains two MYT-family binding consensus elements. Based on these notions, we hypothesized that MYT1L functions as a repressor of the activity of Bdnf promoter I.
To study the role of MYT1L, we first investigated MYT1L binding to Bdnf promoters in endogenous context with ChIP-qPCR (Fig. 6A). The ChIP-qPCR analysis showed that MYT1L was bound to Bdnf promoter I statistically significantly more than to Bdnf promoter IV both at basal conditions and after both stimuli (Fig. 6A), in agreement with our hypothesis that MYT1L is a regulator of Bdnf promoter I. We next determined that MYT1L protein levels remained stable after stimuli (Fig. 6B), despite the mRNA levels slightly decreasing after BDNF-TrkB signaling (left panel of 6C, 6D).
MYT1L is a novel repressor of Bdnf promoter I activity. Rat cultured cortical neurons were treated at 8 DIV for the time indicated with 25 mM KCl (and 5 μM d-APV) or 50 ng/ml BDNF. A, ChIP-qPCR assay using anti-MYT1L antibody. Data is shown as enrichment to Bdnf promoter I and IV (pI and pIV, respectively) relative to binding to the unrelated region (URR) in untreated cells. B, Western blots verifying the functionality of MYT1L knockdown and overexpression. Coomassie staining is shown as loading control. C, D, The neurons were transduced with lentivirus particles encoding (C) shRNAs silencing Myt1l expression or scrambled shRNA (scr shRNA) as a control; and (D) overexpression of MYT1L or EGFP as control. The expression levels of Myt1l and Bdnf transcripts containing either exon I or IV and total Bdnf mRNA levels were measured using RT-qPCR. The expression level of respective transcripts in untreated neurons transduced with EGFP-encoding lentivirus particles or untreated neurons transduced with scrambled shRNAs-encoding or EGFP-encoding lentivirus particles were set as 1. E, On the left, schematical representation of Bdnf promoter I with relevant cis-elements. Identical bases between the human and rat promoter regions are shown as dots. On the right, luciferase reporter assay with wild-type Bdnf promoter I (pI WT) or with Bdnf promoter constructs where respective MYT-binding element was mutated (MYT m1 and m2, shown in blue in the schematics on the left). The activity of Bdnf promoters was measured with luciferase reporter assay and data is shown relative to the activity of wild-type promoter in untreated neurons. All data points are shown as dots and each biological replicate is denoted with the same color as shown in the legend. Numbers above the columns indicate the average of three to five independent experiments (A, n = 3; C, n = 5, D, E, n = 4). Error bars indicate SEM. Statistical significance in ChIP-qPCR assay was calculated relative to binding to URR (black asterisks) or between Bdnf promoters (red asterisks) in respectively treated cells (A), in RT-qPCR experiments relative to the levels of respective transcripts in neurons transduced with scr shRNA or EGFP-encoding lentiviruses (C, D), and in luciferase reporter assay relative to the activity of the wild-type promoter in respectively treated cells (E). #≤ 0.1, *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001 (paired two-tailed t test). AP1, AP1 family-binding element; CRE, cAMP response element; UBE, USF-binding element; PasRE, neuronal PAS domain protein 4 (NPAS4)-binding element bHLH-PAS transcription factor response element.
Next, we used RNA interference with lentivirus-mediated overexpression of short hairpin RNAs (shRNAs, adapted from Mall et al., 2017) to silence Myt1l expression (Fig. 6B,C). Silencing decreased Myt1l mRNA levels ∼3-fold (Fig. 6C) and both Myt1l shRNAs increased the basal levels of Bdnf exon I and IV mRNAs ∼3-6-fold (Fig. 6C). Myt1l silencing also increased KCl-induced levels of both Bdnf exon I and IV transcripts and increased the levels of Bdnf exon IV transcripts after BDNF-TrkB signaling. Interestingly, the two Myt1l shRNAs affected the BDNF-induced levels of Bdnf exon I in the opposite manner, probably accounting for off-target effects of one of the shRNAs. Finally, lentivirus-mediated overexpression of MYT1L (Fig. 6B) increased basal levels of both Bdnf exon I and IV transcripts but did not affect stimulus-induced levels of Bdnf transcripts (Fig. 6D). Surprisingly, the changed levels of MYT1L affected both Bdnf exon I and IV transcripts in a similar manner, even though our ChIP-qPCR revealed much higher binding to Bdnf promoter I.
Finally, we studied the potential MYT-family cis-regulatory elements in Bdnf promoter regions. As we hypothesized that MYT1L is a regulator of Bdnf promoter I and binds the aforementioned elements that are located in a region which deletion increases the activity of the promoter, we mutated these potential cis-elements in Bdnf promoter I reporter construct (Fig. 6E, blue). As expected, mutating these sites increased the activity of Bdnf promoter I at basal levels and after both KCl and BDNF treatment in luciferase reporter assay (Fig. 6E). Collectively, we have determined two MYT1L cis-elements in Bdnf promoter I, and our results show that MYT1L represses the levels of Bdnf exon I-containing transcripts.
EGR family is a novel regulator of stimulus-dependent expression of Bdnf exon I and IV transcripts
Of known stimulus-dependent transcription factors, we identified both AP1 family members and multiple members of the early growth response (EGR) family in the in vitro DNA pulldown experiment (Fig. 1). Additionally, we showed that the overexpression of EGR1 and EGR3 had a remarkable effect on the activity of Bdnf promoters in the luciferase reporter assay (Fig. 2). Therefore, we decided to further focus on the EGR family as a potential stimulus-dependent regulator of Bdnf expression. The EGR family members EGR1 (or Krox-24, NGFI-A, Zif268), EGR2 (or Krox-20), EGR3, and EGR4 (or NGFI-C) are all inducible C2H2 zinc-finger transcription factors (reviewed in Herdegen and Leah, 1998). It is known that BDNF-TrkB signaling rapidly induces the expression of EGR family members (Calella et al., 2007) and EGR family contributes remarkably to BDNF-TrkB signaling-induced transcription (Ibarra et al., 2022). It has been reported that the induction of Bdnf exon IV and VI mRNAs after electroconvulsive seizures in the hippocampus is impaired in Egr3 knock-out animals (Meyers et al., 2018). Moreover, EGR1 is involved in glucocorticoid receptor-signaling repression of Bdnf exon IV and VI mRNA levels (H. Chen et al., 2017, 2019).
First, we studied EGR1 binding to Bdnf promoters with ChIP-qPCR (Fig. 7A). EGR1 was bound to both Bdnf promoters similarly after BDNF treatment and enriched more to Bdnf promoter IV than to promoter I after KCl treatment (Fig. 7A). In agreement with previous literature, Egr1 mRNA and protein levels were almost nonexistent in untreated neurons and were remarkably induced after stimulus, notably more after BDNF-TrkB signaling than after membrane depolarization (Fig. 7B).
EGR family is a novel regulator of stimulus-dependent expression of Bdnf exon I and IV-containing transcripts. Rat cultured cortical neurons were treated at 8 DIV for the time indicated with 25 mM KCl (and 5 μM d-APV) or 50 ng/ml BDNF. A, ChIP-qPCR assay using anti-EGR1 antibody. Data is shown as enrichment to Bdnf promoter I and IV (pI and pIV, respectively) relative to the binding to the unrelated region (URR) in untreated cells. B, Western blots verifying the overexpression of EGR1 and dominant-negative proteins for EGR1 (Zn-EGR1) and EGR3 (Zn-EGR3). Coomassie staining is shown as a loading control. C, D, The neurons were transduced with lentiviral particles encoding (C) EGR1 or ZnEGR1 or EGFP as control; or (D) EGR3 or ZnEGR3 or EGFP as control. The expression levels of Bdnf transcripts containing either exon I or IV and total Bdnf mRNA levels were measured using RT-qPCR. E, On the left, schematic representation of Bdnf promoters I and IV with the described regulatory elements. Identical bases between the human and rat promoter regions are shown as dots. On the right, luciferase reporter assay with wild-type Bdnf promoter (pI WT or pIV WT) or with Bdnf promoter constructs where respective EGR-binding elements were mutated (EGR m1, m2, m3, shown in orange on the schematics on the left). The activity of Bdnf promoters was measured with luciferase reporter assay and data is shown relative to the activity of respective wild-type promoter in untreated neurons. The data of the wild-type promoter is the same as depicted in Figure 6. All data points are shown as dots and each biological replicate is denoted with the same color as shown in the legend. Numbers above the columns indicate the average of three to five independent experiments (A, n = 3; C, n = 5; D, E, n = 4). Error bars indicate SEM. Statistical significance in ChIP-qPCR assay was calculated relative to the binding to URR (black asterisks) or between Bdnf promoters (red asterisks) in respectively treated cells (A), in RT-qPCR experiments relative to the levels of respective transcripts in neurons transduced with EGFP-overexpressing cells (C, D), and in luciferase reporter assay relative to the activity of the respective wild-type promoter in respectively treated cells (E). #≤ 0.1, *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001 (paired two-tailed t test). AP1, AP1 family-binding element; BHLHB2-RE, BHLHB2 response element; CaRE, calcium response element; CRE, cAMP response element; NFkB-RE, NFkB response element; UBE, USF-binding element; PasRE, NPAS4-binding element.
Next, we performed lentivirus-mediated overexpression of EGR1 and the dominant-negative variant of EGR1, ZnEGR1, which consists only of the EGR1 zinc-finger domains responsible for binding DNA (Levkovitz and Baraban, 2001; Fig. 7B,C). The overexpression of EGR1 slightly increased the basal levels of Bdnf exon I mRNAs and increased KCl-induced levels of Bdnf exon I and IV mRNAs by ∼2- and ∼1.4-fold, respectively (Fig. 7C). At the same time, overexpression of EGR1 did not further increase the BDNF-induced expression levels of Bdnf exon I and IV transcripts. Interestingly, overexpression of ZnEGR1 had a stronger effect than overexpressing EGR1 and increased basal levels of Bdnf exon I mRNAs ∼6-fold and basal levels of Bdnf exon IV mRNAs ∼3-fold. Overexpression of ZnEGR1 also increased both KCl- and BDNF-induced levels of Bdnf exon I mRNAs ∼1.5-1.8-fold, slightly decreased the KCl-induced levels of Bdnf exon IV mRNAs, and did not affect BDNF-induced levels of Bdnf exon IV mRNAs. Overexpression of EGR3 remarkably increased the basal levels of both Bdnf exon I and exon IV transcripts, ∼9- and ∼3-fold, respectively, but also increased KCl-induced levels of Bdnf exon I (Fig. 7D). Overexpression of a dominant-negative form of EGR3, ZnEGR3, containing only the zinc-finger domains of EGR3 (Levkovitz et al., 2001) decreased KCl-induced levels of both Bdnf exon I and IV mRNAs but decreased only Bdnf exon IV mRNA levels ∼1.8-fold after BDNF treatment (Fig. 7D).
Finally, we set out to find the cis-elements responsible for the EGR-dependent induction of Bdnf promoters. We mutated three predicted EGR response elements in Bdnf promoter I (Fig. 7E). Mutating the EGR site 3 increased the basal activity of Bdnf promoter I, mutating EGR sites 2 and 3 increased KCl-induced activity, and mutating EGR sites 1 and 3 increased BDNF-induced activity of the promoter I (Fig. 7E). In Bdnf proximal promoter IV we predicted one EGR response element (Fig. 7E) and mutating this site increased the activity of the promoter in both untreated neurons and after KCl and BDNF treatment (Fig. 7E). Notably, the identified EGR-binding site in Bdnf promoter IV partially overlaps with Ca-response element 1 (CaRE1), which is the binding site for MEF2 and CaRF. Interestingly, while most of the previously described regulatory sites in Bdnf promoters are well conserved between murine and human, the EGR1 binding sites in promoter I show notable differences. Collectively, we have identified numerous novel EGR cis-elements regulating the activity of Bdnf promoters, and our results show that EGR family is a novel regulator of Bdnf stimulus-dependent transcription.
ATF2, MYT1L, and EGR1 bind to Bdnf promoters in vivo
To extend our in vitro findings and investigate whether the transcription factors ATF2, MYT1L, and EGR1 could also regulate Bdnf expression in vivo, we assessed their binding to Bdnf proximal promoters in both young and adult rat brains using ChIP-qPCR (Fig. 8). For ChIP analysis, we selected the cortex of postnatal day 8 rats, consistent with our in vitro DNA pulldown experiments, and the adult hippocampus, where Bdnf expression levels are known to be high (Esvald et al., 2023).
ATF2, MYT1L and EGR1 bind Bdnf promoters in vivo. ChIP-qPCR assays were conducted using postnatal day 8 rat cortex (CTX) or adult rat hippocampus (HC) tissues with antibodies against ATF2, MYT1L, and EGR1. Data is shown as relative enrichment of each transcription factor compared with the enrichment at unrelated region (URR). All data points are shown as dots and each replicate is denoted with the same color as shown in the legend. Numbers above the columns indicate the average of three independent experiments (n = 3). Error bars indicate SEM. Statistical significance was calculated relative to the binding to URR (black asterisks) or between Bdnf promoters (shown in red). NS, not significant, *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001.
Our results show that ATF2 and EGR1 bind similarly to both Bdnf promoters in young and adult animals (Fig. 8). Notably, MYT1L shows a significantly higher binding to Bdnf promoter I in young postnatal animals and to a similar degree to both Bdnf promoter I and IV in adult animals. This implies that the role of MYT1L in Bdnf regulation changes either during development or differs between brain regions, consistent with previous findings highlighting age and brain region-dependent effects of MYT1L haploinsufficiency on gene expression and neuronal functions (S. Kim et al., 2022). This is further supported by recent findings that MYT1L does not bind Bdnf locus in embryonic mice (E14), but in adult mice preferentially binds to Bdnf promoter I, with much weaker binding to promoter IV (data from J. Chen et al., 2023). Published CUT&RUN experiments have further revealed that in vivo MYT1L binds broadly across Bdnf second cluster without a distinct binding peak, in contrast to a well-defined binding peak in Bdnf promoter I, followed by broader binding across the first cluster of exons (J. Chen et al., 2023). Collectively, our results confirm the binding of ATF2, MYT1L, and EGR1 to Bdnf promoters in vivo.
Discussion
In our study, we hypothesized that Bdnf transcriptional autoregulation and membrane depolarization recruit distinct transcription factors to drive Bdnf expression. Based on the induction of different sets of genes (Ibarra et al., 2022), different dynamics of Bdnf expression, and usage of at least partially distinct transcription factors for Bdnf induction upon different stimuli (Pruunsild et al., 2011; Tuvikene et al., 2016; Esvald et al., 2020, 2022), we set out to determine regulators specific to each of the stimulus. Surprisingly, our results showed that many transcription factors affected the activity of Bdnf promoters in the same direction after both BDNF and KCl treatment in luciferase reporter assays, with no remarkable preference for one stimulus. Additional experiments indicated that ATF2, MYT1L, and EGR1 bind to Bdnf promoters and regulate transcription after both stimuli. Collectively, our findings suggest a considerable overlap in Bdnf gene regulation and limited stimulus specificity.
Based on our results, notable exceptions are the AP1 and USF families, which we report for the first time as stimulus-specific regulators of Bdnf gene expression. Specifically, we identified the USF family acting primarily after membrane depolarization, whereas the AP1 family functions predominantly after BDNF-TrkB signaling. We also show that EGR1 is more inducible after neurotrophin signaling, suggesting that the EGR family inherently participates more in Bdnf regulation after BDNF-TrkB signaling than after membrane depolarization. This result aligns well with the RNA-seq results published by Ibarra et al. (2022). Future research is needed to determine whether AP1 and USF families are recruited exclusively after BDNF-TrkB and membrane depolarization, respectively, or also after other stimuli, such as cAMP signaling, to induce Bdnf expression. The purpose of having only a few stimulus-specific regulators among numerous broadly acting regulators remains unclear. As Bdnf is induced by a myriad of stimuli (West, 2008; West et al., 2014), we speculate that there are additional, yet undescribed regulators that are recruited exclusively after particular stimuli to provide induction of Bdnf expression in specific contexts.
In our search for novel Bdnf regulators, by using in vitro pulldown experiments, we identified numerous transcription factors that are broadly expressed in the brain, whereas Bdnf itself is not expressed in all brain regions and neuronal populations. In addition to the key transcription factors necessary for Bdnf induction, enhancer regions and chromatin accessibility and state are also crucial. For instance, although TrkB is highly expressed and can activate signaling in striatal inhibitory neurons, BDNF-TrkB autoregulation is not functional in these cells. In striatum Bdnf mRNA expression levels are extremely low and the existing BDNF protein there is transported from other brain regions (Hofer et al., 1990; Timmusk et al., 1994; Altar et al., 1997; Conner et al., 1997; Baquet et al., 2004). Recent studies, including single cell-sequencing datasets, further confirm that in inhibitory neurons Bdnf mRNA is mostly undetectable and Bdnf transcription is not activated in these neurons by neuronal activity-mediated membrane depolarization (Spiegel et al., 2014; Zeisel et al., 2018). Furthermore, ATAC-seq data from cultured neurons shows that the Bdnf locus is not open in striatal neurons (Carullo et al., 2020), even though many of the transcription factors, such as CREB, AP1, and USF family, that regulate Bdnf expression in excitatory neurons are also present in inhibitory neurons (Zeisel et al., 2018). Interesting open questions for future studies are whether Bdnf locus is actively repressed in inhibitory neurons and how Bdnf locus is regulated during development.
The proximal promoters of Bdnf are considered to fine-tune Bdnf expression in a tissue-specific manner (Timmusk et al., 1993; Nakayama et al., 1994; Esvald et al., 2023) and contribute to a variety of behavioral functions (Hong et al., 2008; Hill et al., 2016; Maynard et al., 2016, 2018; McAllan et al., 2018; Hallock et al., 2020; You et al., 2020). Therefore, we expected to identify Bdnf regulators with strong promoter specificity. Surprisingly, our luciferase reporter assays showed that most of the tested transcription factors altered the activity of both Bdnf promoters I and IV, with only minor specificity. Furthermore, our ChIP assays showed that ATF2, EGR1, and MYT1L bind to both studied Bdnf promoters both in vitro and in vivo, and we identified novel EGR-binding elements both in Bdnf promoter I and in promoter IV. Furthermore, although MYT1L showed stronger binding to promoter I, it also influenced Bdnf exon IV transcript levels. As MYT1L exhibits broad binding across both clusters of Bdnf exons (I–III and IV–VII; J. Chen et al., 2023), it is plausible that MYT1L interacts with RNA polymerase II, histones, or other transcription factors or regulates Bdnf exon IV levels via additional unknown enhancer region. Collectively, our findings and reports by others (West et al., 2014) indicate a remarkable overlap of transcription factors recruited to distinct Bdnf promoters. We propose that the use of the same transcription factors by distinct Bdnf promoters enables simultaneous transcription from multiple Bdnf promoters using a limited set of transcription factors, thereby amplifying stimulus-dependent Bdnf expression levels. For example, promoter I contains three consecutive EGR binding sites, and promoter IV harbors at least one, illustrating how multiple cis-elements can cooperate to drive robust Bdnf induction following BDNF-TrkB signaling.
Many of the functional cis-elements in Bdnf promoters can bind multiple transcription factors, suggesting a model of competitive binding. For example, the AP1-2 site of Bdnf promoter I can bind either AP1 family homo- and heterodimers or ATF2-cJUN heterodimers, and the CRE sites can bind CREB or ATF2 homodimers or ATF2-cJUN heterodimers. Our findings suggest that the AP1 family heterodimers are the most potent in regulating Bdnf promoter I activity, followed by ATF2-cJun dimers and then CREB or ATF2 homodimers. Which of these complexes ultimately form depends on their relative availability in the cell, thereby determining the extent of Bdnf induction. This model of competitive binding is consistent with our observation that disrupting ATF2 function, either by impairing DNA binding or via knockdown, leads to a marked increase in Bdnf expression—particularly of transcripts containing exon I, possibly due to increased CREB and AP1 occupancy at the promoter.
Similarly, the identified EGR cis-elements are GC-rich sequences that can be recognized also by other zinc finger transcription factors, such as SP1-like and KLF transcription factors (Kaczynski et al., 2003). In our in vitro DNA pulldown assays, we detected binding of KLF12, KLF13, KLF16, SP3, and SP4, presumably to the identified EGR elements, suggesting that various transcription factors compete for DNA binding. A similar competitive mechanism has been described for SP family transcription factors binding to the Adh5 promoter (Kwon et al., 1999). Furthermore, as some SP1-like/KLF family members function as activators and others as repressors (Kaczynski et al., 2003), a small number of cis-elements can switch between repressing and activating roles, depending on the cellular context.
Several important cis-regulatory elements in Bdnf promoters overlap, forming composite elements with unique binding properties (Jolma et al., 2015), ensuring specific gene expression programs. For instance, the AP1-1 site overlaps with the PasRE in promoter I, and therefore NPAS4 binding prevents AP1 family binding, rendering the AP1 family TrkB signaling-specific regulator of Bdnf promoter I (Pruunsild et al., 2011; Tuvikene et al., 2016). In promoter IV, CaRF and MEF2 sites (i.e., CaRE1; Tao et al., 1998; Lyons et al., 2012) and the herein discovered novel EGR site overlap. In stimulated neurons, decreased Bdnf exon IV mRNA levels caused by dominant-negative ZnEGR3 and, in contrast, increased promoter IV activity after mutating EGR site, could be explained by competition at the CaRE1 site. These overlapping elements could enable a binary switch in which binding of one factor sterically blocks the binding of others.
For Bdnf promoter IV, several repressors have been described, including MeCP2 (Martinowich et al., 2003; W. G. Chen, et al., 2003a), BHLHB2 (X. Jiang et al., 2008), and NFATc4 (Ding et al., 2018). However, only one repressor, RE1-silencing transcription factor [REST or neuron-restrictive silencer factor (NRSF)], has been described for Bdnf promoter I (and promoter II; Timmusk et al., 1999; Hara et al., 2009). In this study, we identified KLF family members and MYT1L as novel repressors of Bdnf promoter I. Our findings, including the effects of MYT1L silencing and mutating the MYT1L binding sites, support a repressive role for MYT1L. In contrast, overexpressing MYT1L increased endogenous Bdnf levels in unstimulated neurons, an outcome we currently cannot explain. We note that MYT1L has a broad effect on neuronal gene expression, influencing both early development and maturation (Kepa et al., 2017; Mall et al., 2017; J. Chen et al., 2021, 2023; Weigel et al., 2023), and thus it is possible that overexpressing MYT1L alters the neuronal state and maturation, indirectly causing Bdnf upregulation. However, we observe direct MYT1L binding to Bdnf promoters in cultured cortical neuron and in the brain in vivo, suggesting that MYT1L can play both repressive and activating roles for Bdnf depending on the cellular context. Such a switch could involve interactions with cofactors or posttranslational modifications that reduce the affinity of MYT1L for corepressors, consistent with the “ready-set-go” model (J. Chen et al., 2022).
In addition to recruiting coactivators and/or corepressors, transcription factors can strengthen or inhibit the functions of each other (Lambert et al., 2018). For example, certain KLF, SP, and EGR family members increase DNA accessibility (Zhao et al., 2022), and AP1 family facilitate chromatin opening by recruiting chromatin remodelers (Vierbuchen et al., 2017). Hence, by supporting the binding of other transcription factors, these transcription factors could increase the activity of promoters beyond their own transactivation capabilities. Conversely, some transcription factors can decrease the activity of others, e.g., MeCP2 represses CREB on Bdnf promoter IV (Tai et al., 2016). Additionally, MYT1L has been shown to co-occupy DNA with activity-dependent transcription factors, potentially neutralizing their effect on gene expression (J. Chen et al., 2023). Future investigations will be necessary to determine the extent to which Bdnf regulators function synergistically, competitively, or redundantly.
Footnotes
This work was supported by Estonian Research Council Grant PRG805 and PRG2583, European Union through the European Regional Development Fund (Project No. 2014-2020.4.01.15-0012), European Commission and Estonian Research Council (ERA-NET NEURON Cofund2 programme grant GDNF UpReg), and Horizon 2020 Framework Programme (H2020) EU734791. We also thank the “TUT Institutional Development Program for 2016–2022” Graduate School in Clinical Medicine, which received funding from the European Regional Development Fund under program ASTRA 2014-2020.4.01.16-0032 in Estonia. We thank Epp Väli for technical assistance, Käthy Rannaste for cloning, Andreas Remmel for producing the first set of lentiviruses encoding Atf2-targeting gRNAs, and Marko Susi for contribution in the early stages of this study with the EGR family in the regulation of Bdnf expression.
The authors declare no competing financial interests.
- Correspondence should be addressed to Tõnis Timmusk at tonis.timmusk{at}taltech.ee.
