A novel intergenic enhancer that regulates Bdnf expression in developing cortical neurons

Summary Brain-derived neurotrophic factor (BDNF) promotes neuronal differentiation and survival and is implicated in the pathogenesis of many neurological disorders. Here, we identified a novel intergenic enhancer located 170 kb from the Bdnf gene, which promotes the expression of Bdnf transcript variants during mouse neuronal differentiation and activity. Following Bdnf activation, enhancer-promoter contacts increase, and the region moves away from the repressive nuclear periphery. Bdnf enhancer activity is necessary for neuronal clustering and dendritogenesis in vitro, and for cortical development in vivo. Our findings provide the first evidence of a regulatory mechanism whereby the activation of a distal enhancer promotes Bdnf expression during brain development.


INTRODUCTION
The brain-derived neurotrophic factor (BDNF) gene encodes a neurotrophin with critical roles in brain development and functions, ranging from neuronal survival and differentiation during early development, to long-term potentiation and synaptic plasticity in the adult brain. 1,2 Reduced BDNF expression has been implicated in a host of neurological diseases, including neuropsychiatric disorders such as schizophrenia, 3 stress 4 and depression 5 ; neurodegenerative diseases including Huntington's 6,7 and Alzheimer's disease 8 ; and neurodevelopmental disorders such as Rett syndrome 9 and attention deficit hyperactivity disorder. 10 Conversely, enhanced BDNF expression is linked to the neuroprotective effects of environmental enrichment, 11,12 exercise, 13,14 and anti-depressants. 2,15 Given the myriad functions identified for BDNF, understanding the regulation of the BDNF gene in neurons during brain development and disease is of paramount importance.
The rodent and human structure of the BDNF gene is complex, consisting of multiple 5 0 exons, each containing its own promoter and 5 0 untranslated region (5 0 UTR), that are alternatively spliced to a universal coding exon 16-18 ( Figure S1A). Despite being translated into identical proteins, Bdnf mRNA variants exhibit specific expression patterns and physiological effects. [19][20][21][22][23][24][25][26] For example, disruption of exon I or II, but not IV or VI, enhances male aggression 19 and impairs female maternal care. 20 Our current understanding of Bdnf transcriptional control is mostly centered on the distinct role of each promoter, however its regulation through distal elements remains unclear.
Enhancers are short regions of regulatory DNA, whose activity promotes the expression of their target gene(s). 27 Combinations of enhancer elements confer spatially and temporally regulated gene expression profiles. 28 In linear chromosomal distance, enhancers are often located far from the genes that they control, although within the three dimensional (3D) nuclear space they become proximal through enhancer-promoter looping. 29 Enhancer-promoter proximity can be critical for appropriate gene expression and is supported by the genome architecture of the region. [30][31][32][33][34] Interactions can occur in the context of topologically associated domains (TADs), which are megabase-sized regions of DNA that interact more frequently within themselves than with the surrounding regions. 35 Genome topology and gene activation is also affected by nuclear compartmentalization, and the position of the gene with respect to nuclear landmarks is important. Putative enhancers for Bdnf have been identified based on 3D proximity to the gene and H3K27ac occupancy. 36 An intronic enhancer regulating both basal and stimulus-dependent expression of Bdnf was recently found for transcripts expressed from promoters I-III. 37 Sub-TAD level loops often reflect contacts occurring between gene promoters and enhancers, 35 so we reasoned that enhancers for the Bdnf gene could be found within the sub-TAD identified here. To this end, we performed 4C-seq, a technique that identifies chromatin regions making contact with a specific 'viewpoint' sequence. 48 A viewpoint designed at Bdnf promoter I identified two regions of interaction in NPCs and PMNs, and in cortical neurons (Figure 2A). The first interaction site is internal to the Bdnf gene, around exon VIII. The second is an intergenic region located around 170 kb downstream of Bdnf and around 5 kb upstream of the Lin7c gene (a distal interacting site, hereby referred to as Bdnf Enh170 ; Figure 2A). The profile appeared the same irrespective of the Bdnf expression levels in the cell type (Figure 2A), consistent with the HiC analysis ( Figure S1B). To confirm Bdnf interaction with the distal interacting site, we designed a viewpoint spanning Bdnf Enh170 and performed 4C-seq in cortical neuronss. We identified the reciprocal interaction from Bdnf Enh170 to Bdnf, with a peak at exon VIII, suggesting that this site anchors the loop ( Figure 2B). Interactions were negligible from Bdnf Enh170 to sequences upstream of the Bdnf gene ( Figure 2B).
To verify the loop of Bdnf to the distal interacting site and assess its frequency more quantitatively, DNA-FISH was performed using fosmids encompassing a) the Bdnf Enh170 and Lin7c gene, b) the Bdnf gene, and  iScience Article c) a control downstream region located the same distance from Bdnf Enh170 as Bdnf (170 kb). Measuring the distance between these probes in pairwise combinations revealed that in PMNs, Bdnf Enh170 was closer to, and exhibited more frequent interactions with, the Bdnf probe compared to the downstream probe (Figure 2C). Importantly, Bdnf Enh170 and the Bdnf gene regions were in closer proximity in PMNs than in NPCs, and co-localization frequency increased during differentiation ( Figure 2C). Thus, although the looping profiles are similar at the population level (Figure 2A), single cell analysis indicated more frequent interactions between Bdnf Enh170 and Bdnf taking place in cells where Bdnf expression is high ( Figure 2C). The use of a reciprocal combination of probe labels further supported this conclusion ( Figure S2). Our findings are in accordance with previous studies showing that chromosome conformation capture technologies usually capture proximity of enhancers and promoters, whereas DNA-FISH can detect direct interactions between genomic regions. 49,50 Bdnf Enh170 bears many characteristics typical of enhancers To assess whether Bdnf Enh170 exhibits the characteristics of an enhancer, we first analyzed publicly available data. Sensitivity to DNase I is a feature of active chromatin regions including promoters and enhancers. 51 ENCODE DNase I hypersensitivity data showed peaks at Bdnf Enh170 in brain ( Figure 3A), but not in other tissues, which was similar to the pattern of DNase I hypersensitivity observed at Bdnf promoters ( Figure S3A). A dataset using an alternative chromatin accessibility assay named Assay for Transposase-Accessible Chromatin with Sequencing (ATAC-seq 52 ), also identified open chromatin at Bdnf Enh170 in microdissected hippocampal dentate gyri (not shown).
We then investigated other enhancer hallmarks at Bdnf Enh170 using ChIP-seq datasets generated by our and other laboratories (Table S1 53-55 ). Chromatin modifications, such as H3K4me1 and H3K27ac, predict enhancer function genome-wide. 56,57,58 The histone acetyltransferase CBP (CREB Binding Protein) is an enhancer regulator which catalyses the addition of H3K27ac. 59,58 The transcriptional coactivator Mediator interacts with cohesin to regulate enhancer-promoter looping. 60 Enhancer sites recruit multiple transcription factors. 28 We identified a strong peak of the enhancer epigenetic signatures H3K27ac and H3K4me1 at Bdnf Enh170 in basal and depolarized cortical neurons ( Figure 3B). CBP and the MED23 Mediator subunit were also found to bind to Bdnf Enh170 in cortical neurons ( Figures 3A and S3B), together with the transcription factors MEF2, CREB and TBR1 ( Figures 3A and S3B). The transcription factors and coactivators show a double peak at Bdnf Enh170 , coinciding with a double peak of DNase I hypersensitive sites.
Enhancers are transcribed in many cell types, including neurons. 53,54,61 In some instances, the enhancer RNA (eRNA) has functional roles, such as interacting with Negative Elongation Factor, 62 CBP, 63 or RNA Polymerase II (RNAPII 53 ), or affecting 3D contacts. 64 In other systems, eRNA transcription may contribute to the maintenance of the transcriptional machinery or the opening of the chromatin. 65,66 Regardless of mechanism, the production of eRNAs is considered a critical feature of active enhancers. We therefore sought to determine whether transcriptional activity could be detected from Bdnf Enh170 . eRNAs are lowly expressed and unstable, and conventional RNA-seq databases may not show transcription at enhancer sites. Methods that detect nascent RNA such as Genome Run On with sequencing (GRO-seq) are better suited for detecting eRNAs, because they map transcripts actively engaged with RNAPII. 67 Analysis of GRO-seq data from Reelin-stimulated cortical neurons, 54 showed that RNA is transcribed bidirectionally from the Bdnf Enh170 region ( Figure 3A). As expected, the Lin7c gene exhibited bidirectional RNA production at the active promoter. 67,68 To validate the sequencing data, qRT-PCR was performed on NPCs and PMNs using primers that generate amplicons within the region of GRO-seq enrichment ( Figure 3A, sites A and B). Because eRNA are transcribed at very low levels, cells were treated with the transcriptional inhibitor DRB (5,6-dichloro-1-beta-D-ribofuranosylbenzimidazole) to determine background levels. We found that Bdnf Enh170 was transcribed in PMNs, at levels significantly higher than either in NPCs or in DRB-treated PMNs ( Figure 3C). A region just downstream of the enhancer (À4.0 kb from the Lin7c transcriptional start site) showed no increase in transcription from NPC to PMN, and no sensitivity to DRB in PMN ( Figure 3C), confirming that the eRNAs are not a continuation of Lin7c promoter antisense transcripts.   69 ), and lentivirus was generated either in combination with no targeting (NT) guide RNA (gRNA), or a gRNA targeted to Bdnf Enh170 (Enh g1 or Enh g2 ). NPCs were infected with CRISPRi lentivirus expressing a puromycin resistance cassette and allowed to differentiate in vitro; neurons were selected for two days before harvesting PMNs ( Figure 4A). Expression of Bdnf Enh170 eRNA after differentiation was significantly decreased in the presence of enhancer-targeted gRNAs ( Figure 4B). Bdnf Enh170 inhibition caused a significant reduction of Bdnf total mRNA (measured in the universal exon) confirming that that it is a functional Bdnf enhancer ( Figure 4C). Analysis of different Bdnf isoforms indicated that Bdnf Enh170 inhibition resulted in lower expression of most Bdnf variants, with significant effects on isoforms expressing exon IV, V, VI, VIII or IXa ( Figure 4D). We did not see a reduction in Lin7c mRNA or antisense promoter transcription at À2.0 kb ( Figure 4E), confirming that the CRISRPi inhibitory effect at the enhancer does not spread into the adjacent Lin7c promoter.
To further investigate the link between enhancer and variant transcription, we performed these experiments using dCas9 fused to the transcriptional activator VP64. 70 The targeting of CRISPRa (activator) virus complexes to Bdnf Enh170 using the same Enh g1 increased eRNA transcription ( Figure 4F). Bdnf total mRNA showed an increase after Bdnf Enh170 activation, albeit not statistically significant ( Figure 4G). Bdnf variants as a group showed increased expression, with significance seen for exon V ( Figure 4H). No changes were observed for Lin7c mRNA or antisense transcription ( Figure 4I). These experiments confirm that Bdnf Enh170 is a bona fide enhancer of developmental Bdnf expression.

Bdnf Enh170 regulates activity-dependent Bdnf expression in cortical neurons
Bdnf gene expression is increased after neuronal stimulation as well as during differentiation. To investigate whether Bdnf Enh170 contributed to activity-dependent Bdnf induction, we first investigated whether Bdnf Enh170 is transcribed in response to neuronal activation. Primary cortical neurons were depolarized with KCl, and eRNA levels were assessed with qRT-PCR. Bdnf Enh170 eRNA increased concomitant with Bdnf mRNA ( Figure S4A). The activity-dependent Activator Protein-1 (AP-1) transcription factors JUN and FOS were recruited to Bdnf Enh170 in response to neuronal depolarization ( Figure S4B), which is consistent with transcription factors encoded by early response genes, like Fos and Jun, controlling the expression of late response genes, such as Bdnf. 71 To assess whether the enhancer was required for activity-dependent Bdnf induction, CRISPRi (dCas9-KRAB) lentivirus was generated either in combination with no targeting gRNA (NT) or targeted to the putative enhancer region (Enh g1 , Enh g2 ), as before. CRISPRi lentivirus was added to primary cortical neurons on the same day as dissociation, and the media was changed the following day. After 5 DIV, the media was supplemented with puromycin to select lentiviral-transduced cells, and then cells were depolarized with KCl at DIV 7 for 3h before harvesting ( Figure S4C). Expression of Bdnf Enh170 eRNA after neuronal activation was significantly decreased in the presence of enhancer-targeted guide RNAs ( Figure S4D). Although we did not detect a significant reduction of total Bdnf mRNA after Bdnf Enh170 inhibition ( Figure S4E), analysis of different Bdnf isoforms indicated that Bdnf Enh170 inhibition resulted in lower expression of Bdnf variants as a group ( Figure S4F), confirming its role as an enhancer. Significant effects of Bdnf Enh170 inhibition were seen on variants containing exon II and V ( Figure S4F). We did not see a reduction in Lin7c mRNA or antisense promoter transcription at À2.0 kb ( Figure S4G), confirming that the CRISRPi inhibitory effect at the enhancer does not spread into the adjacent Lin7c promoter. The effects of enhancer inhibition on Bdnf  iScience Article expression were more subtle after depolarization than during differentiation, and the variants primarily affected were different. This is consistent with the hypothesis that Bdnf gene expression may depend on several enhancers, which are activated in a combinatorial manner depending on physiological context.

Bdnf Enh170 regulates Bdnf-dependent dendritogenesis in cortical neurons
We next sought to study whether Bdnf Enh170 promoted the physiological functions of Bdnf in cortical neurons after stimulation. Bdnf expression is necessary for activity-dependent dendritogenesis, 72,73 a critical process for neuronal growth at later stages of development. To study how Bdnf Enh170 promotes activity-dependent dendritogenesis, cortical neurons were transfected with plasmids encoding dCas9-KRAB-MECP2, 74 with a plasmid encoding the same gRNAs used to knockdown expression in the lentiviral system (BPK1520 vector: NT, Enh g1 , Enh g2 ) and a GFP-encoding plasmid. dCas9-KRAB-MECP2 is a potent repressor in neurons, 75 and because of the single cell nature of the assays it was important to ensure that a strong inhibition was taking place at individual loci. Neurons were maintained either in basal or depolarizing conditions (50 mM KCl, 48 h), and GFP-positive, non-overlapping neurons were analyzed. Quantitative hybridization chain reaction (HCR) RNA-FISH with Bdnf and Lin7c probes confirmed that in NT cortical neurons, we could detect an increase in Bdnf and Lin7c mRNA after depolarization ( Figure 5A), as expected. Inclusion of a gRNA targeting Bdnf Enh170 decreased Bdnf but not Lin7c total mRNA levels in stimulated neurons ( Figure 5A), further confirming that Bdnf Enh170 enhances activity-dependent Bdnf expression.
Dendritic tracing and Sholl analysis showed that, as expected, depolarization induced a significant increase in dendritic complexity in control neurons transfected with dCas9-KRAB-MECP2 only (NT, Figure 5B). KCldependent dendritic branching was substantially reduced when neurons were transfected with dCas9-KRAB-MECP2 targeted to Bdnf Enh170 (Enh g1 or Enh g2 ), with no effect on basal arborization ( Figure 5B). To assess whether the effect of Bdnf Enh170 inhibition was rescued by Bdnf, dendritogenesis was assessed in neurons expressing a vector encoding either the Bdnf coding sequence (pBdnf) or an empty control vector (EV), and co-transfected with CRISPRi vectors (NT or Enh g2 ). Depolarization of cortical neurons in the presence of EV increased neuritic arborization, which was reduced by Bdnf Enh170 inhibition ( Figure S5). Co-transfection of pBdnf rescued the branching defects close to the soma, although it did not fully reinstate the branching in distal dendrites ( Figure S5). Together these findings indicate that Bdnf Enh170 regulates Bdnf expression to promote dendritogenesis.

Bdnf Enh170 influences neuronal differentiation and cortical development in vivo
Bdnf and its main receptor TrkB play a critical role in mouse cortical development, chiefly by regulating neuronal progenitor proliferation 76 and neural migration. 76,77 We next investigated whether Bdnf Enh170 may promote these developmental functions of Bdnf. Initial experiments performed on NPCs in culture indicated that inhibition of Bdnf Enh170 affected PMN cluster formation, quantified by measuring nuclei-nuclei distance (Figures S6A and S6B). PMN dispersion was reversed by co-infection with a lentiviral vector encoding Bdnf ( Figure S6C), indicating that Bdnf is necessary for neuron-neuron interaction, cell migration, or adhesion properties in vitro. The expression of markers of neuronal differentiation such as Map2, NeuN, and Nestin was unchanged (not shown).
Finally, we investigated whether Bdnf Enh170 could affect cortical development in vivo. The mouse cortex is formed in a characteristic inside-out manner, with deep layers generated first and more superficial layers generated later. 78 Neurons are generated in the ventricular zone (VZ) and initially populate the deeper layers of the cortex, whereas neurons born at later developmental stages must cross the deeper layers of the cortex to form the upper layers. To ask whether Bdnf Enh170 affected neuronal cell migration, we employed in utero intracerebroventricular injection with electroporation. Locked Nucleic Acids (LNAs, Qiagen) were used to specifically target Bdnf Enh170 eRNA for degradation, because of the toxicity of large CRISPRi plasmids in vivo. We identified an LNA that significantly reduced the levels of Bdnf Enh170 eRNA in PMNs (LNA Enh ; Figure 6A). Control LNA Neg or LNA Enh were electroporated together with a GFP expression plasmid into E13. iScience Article inhibition ( Figures 6B and 6C). Taken together, these data demonstrate that Bdnf Enh170 is necessary for cortical development and that knockdown of its eRNA results in neuronal migration defects.
During the development of the cortex, cortical neurons use multipolar migration to move from their birthplace, and then establish polarity to enable bipolar radial migration. 79 Once they reach their position in the cortex, they develop axons and dendrites and form connections. 79 Because of the interlinked nature of radial migration and neurite outgrowth, 78 and the importance of Bdnf for dendritic tree  6E). Knockdown of Bdnf Enh1 eRNA using LNA Enh increased the circularity of neurons within the cortical plate ( Figures 6D and 6E), suggesting that Bdnf Enh170 influenced neuron shape and neurite branching. This effect was rescued by expression of the Bdnf coding region ( Figures 6D and 6E).

DISCUSSION
The Bdnf gene has a complex genomic structure . In mice, it comprises at least nine 5 0 untranslated exons, each containing a promoter that is alternatively spliced to a common translated coding exon. Despite intensive scrutiny, the role of most promoters in regulating Bdnf expression remains unclear. Here, we identify Bdnf Enh170 as a novel intergenic enhancer that influences Bdnf expression during cortical development and in response to neuronal depolarization. Bdnf Enh170 can increase the expression of many Bdnf isoforms during neuronal differentiation ( Figures 4F-4I). Its inhibition significantly regulates the expression of at least five Bdnf 5 0 isoforms during differentiation ( Figure 4D), and at least two in response to depolarization ( Figure S4F). Bdnf Enh170 bears most known enhancer hallmarks, including binding of CBP and Mediator, iScience Article chromatin accessibility, specific histone modifications, and transcription (Figures 3 and S3). Analysis of genome topology revealed that Bdnf gene and enhancer are located within a sub-TAD which is bounded by CTCF and cohesin (Figures S1B and S1C). Bdnf activation correlates with increasing frequency of enhancer-promoter co-localization ( Figures 2C and S2) and movement of the genomic region away from the nuclear periphery ( Figure 1D). Bdnf Enh170 inhibition altered neuronal clustering ( Figure S6), dendritic branching ( Figures 5B and S5), and cortical development ( Figure 6), demonstrating its key role in regulating Bdnf functions in vitro and in vivo.
We explored the 3D genome architecture of the Bdnf genomic region using HiC and 4C-seq and described, for the first time, a sub-TAD of increased interaction frequency that includes the Bdnf gene and the downstream region including Bdnf Enh170 and the neuronal gene Lin7c, with CTCF and cohesinpositive boundaries within the Bdnf and Lin7c genes (Figures 2A, 2B, S1B, and S1C). Neither the sub-TAD boundaries nor the enhancer-promoter loop sites changed during neuronal differentiation (Figures 2A and S1B), suggesting that the boundaries are pre-wired in NPCs. Topological structure has been shown to precede gene activation in many model systems, 83-87 and preconfigured loops prime genes for transcriptional activation. 88 Single cell imaging showed an increase in co-localization of Bdnf Enh170 with the Bdnf promoter during the activation of the gene ( Figures 2C and S2), suggesting that the sub-TAD organization may facilitate enhancer-promoter interaction and transcription.
Importantly, the enhancer-promoter loop that we identified with 4C-seq was also found in a recent study that examined the topology of the Bdnf genomic region using 5C technology in cortical neurons. 36 In addition to constitutive loops, the authors also described loops which form in response to depolarization. 36 Future investigations will clarify the functional significance of these loops on Bdnf expression, and the interplay with the intergenic enhancer characterized in our study.
To explore the promoter selectivity of Bdnf Enh170 we performed gain and loss of function experiments. CRISPRa experiments ( Figures 4F-4I) demonstrated that all Bdnf variants can be regulated by Bdnf Enh170 . However, different Bdnf variants were downregulated to different extents by CRISPRi, which may be influenced by the expression level and stability of each isoform, as well as the physiological context. During differentiation, Bdnf Enh170 inhibition markedly affected isoforms containing exon IV, V, VI, VIII, and IXa ( Figure 4), whereas in response to depolarization, the strongest effect of Bdnf Enh170 inhibition was on exon V and exon II-expressing isoforms ( Figure S4F). These findings demonstrate that Bdnf regulation is dependent on physiological context and suggests the existence of additional enhancers or modulators of enhancer activation.
A recently described Bdnf intronic enhancer 37 has been shown to promote transcription only from Bdnf promoters I, II, and III in response to neuronal activation. 37 In addition to different promoter-selectivity, Bdnf Enh170 inhibition decreased total Bdnf mRNA expression during differentiation of cultured neurons ( Figure 4C), which is in contrast to the effect observed on the inhibition of the intronic enhancer, 37 further confirming their distinct role in regulating Bdnf expression.
We found that Bdnf Enh170 inhibition has significant physiological consequences for neuronal differentiation and development both in vitro and in vivo. In addition to promoting the formation of neuronal clusters (Figure S6) and dendritogenesis ( Figures 5 and S5), Bdnf Enh170 is necessary for mouse cortical neuron development in vivo ( Figure 6). Inhibition of Bdnf Enh170 expression disrupts neuronal migration, which was restored by co-electroporation with a vector expressing Bdnf ( Figures 6B and 6C). The effect on neuronal migration could be linked to the neurite extension phenotype seen in vitro ( Figures 5 and S5) because neuronal polarity and morphology are tightly interlinked with migration. 78 Moreover, Bdnf Enh170 inhibition caused a loss of bipolar morphology in neurons that reached the cortical plate, which was restored by coelectroporation with a vector expressing Bdnf ( Figures 6D and 6E).
Inhibition of the Bdnf receptor TrkB has been shown to reduce neuronal progenitor proliferation in the VZ and neuronal migration. 76,77 Our results however indicate that Bdnf Enh170 affects principally the expression of Bdnf during neuronal radial migration, with little or no effect on cell proliferation. A possible explanation is that neuronal progenitor proliferation may depend on Bdnf encoded by mRNA variants that are not regulated by Bdnf Enh170 . Investigation of the importance of the intronic enhancer 37 and other putative Bdnf enhancers 36 in vivo will elucidate the complex regulation of Bdnf during cortical development. iScience Article Transcriptional regulation of the BDNF gene has important implications for the pathogenesis of many neurological disorders. Bdnf Enh170 was identified in mouse but is conserved in the human genome, where it is in a similar orientation and position relative to the BDNF and LIN7C genes (not shown). In humans, an antisense transcript that regulates BDNF expression, BDNF-AS, runs from immediately upstream of the LIN7C transcriptional start site through the entire intergenic region and the BDNF gene. 18,89 Further investigation will address whether, as for the mouse gene, multiple enhancers regulate the human BDNF gene, determining distinct spatiotemporal expression patterns that may be perturbed in neurological disorders.

Limitations of study
Most experiments for this study were performed using ex vivo cultured and differentiated cortical neurons, which may compromise the relevance to in vivo differentiation processes. The CRISPRi studies were done with two independent guide RNAs to reduce enhancer RNA expression, however off-target effects cannot be ruled out. Although the efficiency of the LNA GapmeRs at reducing enhancer RNA expression was demonstrated, the effect on Bdnf mRNA expression could not be verified because of low transfection, and was not verified in vivo where the LNAs were used. As for all assays based on microscopy, quantification of fluorescence based on images may be less accurate than other methods.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

Lead contact
Further information and requests for resources and reagents directed to, and will be fulfilled by, Antonella Riccio (a.riccio@ucl.ac.uk).

RNA isolation and reverse transcription
For transcriptional inhibition, 50 mM of the RNAPII inhibitor DRB (Merck) was added to culture medium for 1h. RNA was isolated from neuronal cultures using TRIzol (Thermo Fisher Scientific) according to the manufacturer's instruction. RNA was treated with the TURBO DNA-free kit (Thermo Fisher Scientific) before being reversed-transcribed in a 20 mL reaction volume containing random hexamers, RiboLock RNAse inhibitor and RevertAid (Thermo Fisher Scientific). qRT-PCR reactions (20 mL) contained 10 mL SYBR Select Master Mix (Thermo Fisher Scientific) and 0.25 mM primers (sequences shown in Table S3) and were performed on a BioRad CFX qPCR machine.

DNA Fluorescence In Situ Hybridization (FISH)
DNA-FISH experiments were performed as described 53 with some modifications. Cells were fixed for 10 min in 4% PFA (paraformaldehyde, TAAB) in PBS, followed by permeabilization for 10 min in 0.5% Triton X-100 in PBS. iScience Article mixed immediately prior to addition to the coverslip for hybridization. DNA was counterstained with 4 0 ,6diamidino-2-phenylindole (DAPI). Coverslips were washed and mounted in ProLong Gold (Thermo Fisher Scientific). Confocal images of neuronal nuclei were acquired using a Leica SPE3 confocal microscope for lamina association, or an SP8 confocal microscope for double DNA-FISH. Images were analyzed using Fiji software. Probe coordinates were identified using the 3D Objects Counter tool on hyperstacks of individual nuclei (ensuring only 1 or 2 foci per cell). For double DNA-FISH analysis, the separation of the probe coordinates (distance AB) from each channel were calculated using the formula: For measurements of probe to nuclear periphery, the distance from the centre of the FISH signal to the closest point of the nuclear edge, identified using DAPI staining, was quantified.

Chromatin immunoprecipitation
Chromatin immunoprecipitation (ChIP) experiments were performed as described previously 53 with some modifications. To crosslink proteins with DNA, the medium was removed from neuronal cultures, and crosslinking buffer (0.1 M NaCl, 1 mM EDTA, 0.5 mM EGTA and 25 mM HEPES-KOH, pH 8.0) containing 1% formaldehyde was added for 10 min at room temperature. The cross-linking reaction was stopped by adding glycine to a final concentration of 125 mM. Cells were rinsed three times with ice-cold PBS containing protease inhibitor cocktail and 1 mM PMSF, collected by scraping and centrifuged at 3,000 rpm at 4 C for 10 minutes. Cell pellets were transferred to 1.5 mL tubes and lysed with 20 cell pellet volumes (CPVs) of buffer 1 (50 mM HEPES-KOH, pH 7.5, 140 mM NaCl, 1 mM EDTA, pH 8.0, 10% glycerol, 0.5% NP-40, 0.25% Triton X-100 and complete protease inhibitor cocktail) for 10 min at 4 C. Nuclei were pelleted by centrifugation at 3,000 rpm for 10 min at 4 C, incubated with 20 CPVs of buffer 2 (200 mM NaCl, 1 mM EDTA, pH 8.0, 0.5 mM EGTA, pH 8.0, 10 mM Tris-HCl, pH 8.0, and complete protease inhibitor cocktail) for 10 min at RT and re-pelleted. 4 CPVs of buffer 3 (1 mM EDTA, pH 8.0, 0.5 mM EGTA, pH 8.0, 10 mM Tris-HCl, pH 8.0, and complete protease inhibitor cocktail) were added to the nuclei, and sonication was carried out by applying 20 pulses, 30 seconds each, at 30 seconds intervals. Insoluble materials were removed by centrifugation at 14000 rpm for 10 min at 4 C, the supernatant was transferred to a new tube, and the final volume of the nuclear lysate was adjusted to 1 ml by adding buffer 3 supplemented to give 150 mM NaCl, 1% Triton X-100, 0.1% sodium deoxycholate in the final chromatin sample. 50 mL of the 1 mL chromatin samples was saved for an Input, whereas the remaining fraction was incubated with 5 mg Rad21 (Abcam ab992, RRID:AB_2176601) antibody and 50 mL Dynabeads (Thermo Fisher Scientific; washed once) and rotated overnight at 4 C. Beads were pelleted and washed with: low-salt buffer (0.1% SDS, 1% Triton X-100, 2 mM EDTA, 20 mM Tris-HCl, pH 8.0, 150 mM NaCl), high-salt buffer (0.1% SDS, 1% Triton X-100, 2 mM EDTA, 20 mM Tris-HCl, pH 8.0, 500 mM NaCl) and LiCl buffer (0.25 M LiCl, 1% IGEPAL CA630, 1% deoxycholic acid (sodium salt), 1 mM EDTA, 10 mM Tris, pH 8.1) and twice with TE buffer (10 mM Tris-HCl, pH 8.0, 1 mM EDTA). For each wash, beads were incubated for 10 min at 4 C while rotating. Immunoprecipitated DNA was eluted by adding elution buffer (0.1 M NaHCO 3 pH 8.0, 1% SDS) and incubating at 65 C, 5 min and then rotating at RT for 15 min. Crosslinking was reversed by adding 10 mL 5M NaCl and incubating the samples at 65 C overnight. DNA was purified using PCR purification columns (Qiagen), quantified using the Qubit high sensitivity assay, and subjected to qPCR using the same amount of DNA in immunoprecipitated and input PCRs. Primer sequences are shown in Table S3. The protocadherin HS5 region was used as a positive control 95 and a region on chromosome 5 was used as a negative control.

4C-seq
4C-seq experiments were performed as described previously. 46 To crosslink proteins with DNA, the medium was removed from neuronal cultures, and crosslinking buffer (0.1 M NaCl, 1 mM EDTA, 0.5 mM EGTA and 25 mM HEPES-KOH, pH 8.0) containing 1% formaldehyde was added for 10 min at room temperature. The cross-linking reaction was stopped by adding glycine to a final concentration of 125 mM. Cells were rinsed three times with ice-cold PBS containing protease inhibitor cocktail and 1 mM PMSF, collected by scraping and centrifuged at 3,000 rpm at 4 C for 10 minutes. Cell pellets were lysed in 10 mL lysis buffer (10 mM Tris pH 8.0, 10 mM NaCl, 0.2% NP40 supplemented with protease inhibitor cocktail and 1 mM PMSF) on ice for 20 min. Nuclei were collected by centrifugation (1800 rpm, 5 min, 4 C), resuspended in 1.23 DpnII buffer and transferred to Protein LoBind tubes (Eppendorff). SDS was added to 0.3% final concentration and nuclei were incubated 1h at 37 C in thermomixer shaking at 900 rpm (30s on, ll OPEN ACCESS iScience 26, 105695, January 20, 2023 21 iScience Article 30s off). Triton X-100 was added to 2% final concentration and nuclei were incubated 1h 37 C in a thermomixer shaking at 900 rpm (30s on, 30s off). 750 Units of DpnII (NEB) was added and incubated overnight at 37 C in a thermomixer shaking at 900 rpm (30s on, 30s off). The next day, the DpnII buffer was replaced with fresh 1.23 DpnII buffer supplemented with 0.3% SDS and 2% Triton X-100 and another 750 Units of DpnII and incubated overnight at 37 C in thermomixer shaking at 900 rpm (30s on, 30s off). Samples of undigested and DpnII-digested DNA was reverse crosslinked and run on a gel to confirm that most DNA fragments were <3 kb after digestion.
Nuclei were centrifuged (1800 rpm, 3 min) and washed twice with 13 T4 DNA ligase buffer before resuspending in 100 mL 13 T4 DNA ligase buffer with 1600 Units T4 DNA ligase (NEB). In nucleo ligation was carried out overnight at 16 C without shaking before confirming that high molecular weight products were obtained. Samples were then reverse crosslinked in the presence of proteinase K overnight at 65 C before phenol:chloroform extraction and ethanol precipitation. DNA was quantified using Qubit high sensitivity assays (Thermo Fisher Scientific). 6-10 mg of DNA was digested with 120 Units Csp6I enzyme (Thermo Fisher Scientific) [3-5 Csp6I digests per sample] overnight at 37 C in thermomixer shaking at 900 rpm (30s on, 30s off). After confirmation that Csp6I-digested products were <3 kb, Csp6I was heat inactivated at 65 C for 20 min before phenol:chloroform extraction and ethanol precipitation. DNA was resuspended in 6 mL total volume to allow proximity ligation by 1600 Units T4 DNA ligase (NEB) overnight at 16 C. Samples were purified by phenol:chloroform extraction and ethanol precipitation, followed by PCR purification columns (Qiagen), before quantitation using with Qubit high sensitivity assays (Thermo Fisher Scientific).
4C-seq libraries were generated using Expand Long Template polymerase (Roche) and primers designed using the 4C-seq primer database 48 (Table S3). Forward primers were generated with the Illumina p1 sequence (AATGATACGGCGACCACCGAGATCTACACTCTTTCCCTACACGACGCTCTTCCGATCT), a two-nucleotide barcode to allow multiplexing of samples, and then the primer sequence. Reverse primers were generated with the Illumina p2 sequence: CAAGCAGAAGACGGCATACGAGATCGGTCTCGGCATTCCTGCTGAACCGCTCTTCCGATCT).
6-10 PCRs were set up per sample to generate library diversity. PCRs were run using the following program: 3 min 94 C; then 29 cycles of 10s 94 C, 1 min 55 C, 3 min 68 C; then 10 min 68 C. PCR products were purified using the High Pure PCR product purification kit (Roche). Libraries were quantified with Qubit high sensitivity assays, assessed using the Agilent Tapestation, and run on an Illumina MiSeq (MiSeq Reagent Kit v3, 150-cycle). 4C-seq data analysis and normalization was performed as described. 48

Lentiviral addition to cultured neurons
Lentivirus was added to NPC cultures at DIV1, and half the media was changed at DIV2 as usual. When half of the media was changed at DIV5, the new media was supplemented with 1.0 mg/mL puromycin dihydrochloride (Merck) (final conc on cells 0.5 mg/mL) to select for transduced PMN.
For cortical neuron cultures, media was changed from plating media to neurobasal media 2h after plating, and then lentivirus was added 2h later (all DIV 0). The next day, all the media was changed. Half of the media was changed again at DIV 5, when it was supplemented with 2.0 mg/mL puromycin dihydrochloride (final conc on cells 1.0 mg/mL) to select for transduced neurons. HCR probe sets targeting the coding sequences of Bdnf (B1 initiator, 20 split-initiator probes) and Lin7c (B3 initiator, 30 split-initiator probes) were purchased from Molecular Instruments. Experiments were performed based on the manufacturer's protocol for mammalian cells. 97 All reagents and materials used were RNAse-free. In brief, transfected cells were fixed with 4% paraformaldehyde (TAAB) at room temperature for 10 mins followed by permeabilization in 70% ethanol for 3h at 4 C. Cells were washed 2 3 5 mins in 2x SSCT Buffer (2x SSC +0.1% Tween20) and pre-hybridized in Probe Hybridization Buffer for 30 minsat 37 C. Cells were then incubated with 1.2 pmol of each probe overnight at 37 C in a humidified chamber. Excess probes were washed off 4 3 5 min with Probe Wash Buffer at 37 C, followed by 2 3 5 min washes in 5x SSC Buffer at room temperature. Pre-amplification was performed in Amplification buffer for 30 minsat room temperature. 18 pmol of each fluorescent hairpin amplifier (B1h1/B1h2 Alexa Fluor 647 and B3h1/ B3h2 Alexa Fluor 594) were snap cooled in separate tubes by heating for 90 sat 95 C in a pre-warmed thermocycler and allowed to cool in the dark for 30 min. After pre-amplification, buffer was removed from cells and replaced with cooled hairpins mixed in Amplification Buffer. To enable quantitative HCR imaging, amplification performed for 45 min in the dark at room temperature. Excess hairpins were washed 5 3 5 min in 5x SSCT Buffer, followed by 10 min incubation in 1 mg/mL DAPI in 1x PBS. Cells were mounted in Pro-Long Gold antifade mountant (#P36930, Thermo Fisher Scientific) and cured overnight. Negative controls without probes and without amplification were captured for each repeat.

RNA-FISH
Airyscan imaging was performed using a Zeiss LSM900 confocal microscope with a 633 Plan Apochromat objective (NA = 1.4) and Airyscan 2 detector with GaAsp technology. Airyscan optimal settings were used for capture, and images were processed using the Zen Blue 3.4 Airyscan 3D processing module with standard settings. DIC microscopy was also performed on all fields of view captured to verify spot locations within cells. For image analysis, masking using the GFP channel was performed on maximum projections ll OPEN ACCESS iScience 26, 105695, January 20, 2023 23 iScience Article with a single macro for all images. Afterward, RNA spot quantification was performed using batch processing in the FISH-Quant plugin on ImJoy, a hybrid computing platform for deep learning image analysis, with filter sigma = 1.0 and spot detection threshold set at 50.

Dendritogenesis assays
Assays were carried out as described previously. 90 Briefly, 2-3h after plating in 24-well plates, mouse cortical neurons were transfected using Optimem containing 375 ng dCas9-KRAB-MECP2 DNA, 125 ng BPK1520 (Non-targeting, or containing guides targeting Bdnf Enh170 ), a GFP expression vector (200 ng pBIRD ( Figure 5) or 500 ng pCIG vector (EV or pBdnf); Figure S5) and 0.8-1.5 mL Lipofectamine 2000 (Thermo Fisher Scientific). After 2h, the medium was replaced with culture media containing 0.33X B27 (serum starve conditions) with or without 50 mM KCl. Cells were cultured for 48h followed by immunostaining with anti-GFP (Abcam ab13970, RRID:AB_300798, 1:2000). Coverslips were blinded before images of GFP-transfected non-overlapping neurons were obtained using a Zeiss Axio Imager microscope and analyzed in Fiji. For Sholl analysis we used the Simple Neurite tracer plugin, and then samples were deblinded. Clustering of neuronal cells was analyzed in Fiji using maximal z projections of the DAPI channel (each image was of a single neuronal cluster and its surrounding cells; if the edge of another cluster was in the image this was removed before processing). After applying a Gaussian blur filter (Sigma 4.0) to even out the signal, we used the 'Find Maxima' tool to identify each nucleus. The XY coordinates were inputted into R and used to compute the distance between every point and every other point, before the median per image was calculated and samples were deblinded.

LNA transfection
NPC were plated as usual in 6-well plates, and then transfected at DIV5 with 100 nM LNA Neg or LNA Enh in Optimem using 1.5 mL Polyethylenimine (PEImax, Polysciences) and centrifugation (500 xg, 5 min). Media was replaced after 2h with PMN media mixed 1:1 with reserved media from the cells prior to transfection. Cells were harvested at DIV7.

In utero electroporation
In utero intracerebroventricular injections with electroporation were performed essentially as described previously. 38,53 E13.5 pregnant mice were anesthetized with isoflurane in oxygen carrier (Abbot Laboratories), and the uterine horns were exposed through a small incision in the ventral peritoneum. Plasmid DNA solution (1.0-1.5 mg/mL), prepared using an Endo-Free plasmid purification kit (Qiagen), was mixed with 50 mM antisense LNA GapmeR (in vivo-ready, Qiagen) and 0.05% Fast Green (Sigma) and injected through the uterine wall into the lateral ventricles of the embryos using pulled borosilicate needles and a mouth aspirator (Sigma). Five electrical pulses were applied at 40 V (50-ms duration) across the uterine wall at 950 ms intervals using 3-mm platinum Tweezertrodes (Harvard Apparatus) and an ECM-830 BTX square wave electroporator (Harvard Apparatus). The uterine horns were replaced in the abdominal cavity and the abdomen wall, and the skin was sutured. 48h after surgery, pregnant mice were sacrificed, and embryos were subjected to immunofluorescence to assess radial migration. iScience Article Embryonic brains were fixed using 4% PFA in PBS overnight at 4 C. Fixed samples were cryoprotected using 30% sucrose overnight at 4 C. Brains were frozen in Optimal Cutting Temperature (OCT, Sakura) and 12 mm coronal sections were cut using a Leica cryostat. Tissue sections were permeabilized in 0.3% Triton X-100, 10% normal goat serum, 2% BSA in PBS at room temperature for 1h and incubated with chicken anti-GFP (Abcam ab13970, RRID:AB_300798, 1:1000) primary antibodies overnight at 4 C. After three sequential washes with PBS, sections were incubated with goat anti-chicken AlexaFluor-488 (Thermo Fisher Scientific A-11039, RRID:AB_2534096, 1:1000) and 4 0 ,6-diamidino-2-phenylindole (DAPI) for 90 min at RT. Sections were washed with PBS and mounted using Fluoromount-G (Southern Biotechnology). Images were acquired on Leica SP8 confocal microscope at 1024 3 1024 pixel resolution; migration analysis images were acquired with a 203 objective, circularity analysis images were acquired with a 633 objective.

Figure 5
A) Line and error bars, mean number of foci G SEM Each point represents a cell, n = 30 across 3 biological replicates.
Sidak's multiple comparisons test (at distances with significance):