Elevated EGR1 binding at enhancers in excitatory neurons correlates with neuronal subtype-specific epigenetic regulation

Abstract

Background

Brain development and neuronal cell specification are accompanied by epigenetic changes that enable the regulation of diverse gene expression patterns. During these processes, transcription factors interact with cell-type-specific epigenetic marks, binding to unique sets of cis-regulatory elements in different cell types. However, the detailed mechanisms through which cell-type-specific gene regulation is established in neurons remain to be explored.

Results

In this study, we conducted a comparative histone modification analysis between excitatory and inhibitory neurons. Our results revealed that neuronal cell-type-specific histone modifications are enriched in super enhancer regions that contain abundant EGR1 motifs. Further CUT&RUN assay confirmed that excitatory neurons exhibit more EGR1 binding sites, primarily located in enhancers. Integrative analysis demonstrated that EGR1 binding is strongly correlated with various epigenetic markers of open chromatin regions and is linked to distinct gene pathways specific to neuronal subtypes. In inhibitory neurons, most genomic regions containing EGR1 binding sites become accessible during early embryonic stages, whereas super enhancers in excitatory neurons, which also host EGR1 binding sites, gain accessibility during postnatal stages.

Conclusions

This study highlights the crucial role of transcription factor binding, such as EGR1, to enhancer regions, which may be key to establishing cell-type-specific gene regulation in neurons.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12915-025-02357-x.

Background

Millions of neurons in the mouse brain interact with each other to perform essential functions, including locomotion control, sensation, and memory. These neurons can be broadly classified into two categories: excitatory and inhibitory neurons, each with distinct morphologies, connectivity, and electrophysiological properties [1]. During embryonic development, neural stem cells in the ventricular zone begin differentiating into excitatory neurons around embryonic day 9.5 (E9.5) and into inhibitory neurons in the adjacent ganglionic eminences around E12.5 [2, 3]. Both kinds of neurons migrate to various brain regions, establish synaptic connections, and mature to fulfill specific functions during postnatal development. Excitatory neurons release neurotransmitters such as glutamate, which bind to receptors on postsynaptic neurons, increasing neural activity. In contrast, inhibitory neurons release gamma-aminobutyric acid (GABA) or glycine, reducing the likelihood of postsynaptic neurons firing an action potential [4]. The balance between excitatory and inhibitory inputs is crucial for normal brain function and behavior. The specification of these neuronal classes is primarily governed by the precise regulation of gene expression networks that enable neurons to generate distinct neurotransmitters, ion channels, and proteins essential for synaptic structure formation.

Epigenetic mechanisms, including DNA methylation and histone modifications, are essential for the precise regulation of cell-type-specific gene expression patterns. In the developing mouse brain, there is a dramatic increase in genomic regions exhibiting cell-type-specific DNA methylation [5, 6]. In neural cells, hypomethylated regions are often enriched with active histone modification markers and are correlated with increased chromatin accessibility, facilitating the initiation and progression of RNA transcription [7]. Using nuclei isolated through affinity purification, a previous study has reported highly distinctive epigenomic landscapes across different types of neocortical neurons [1]. In particular, at enhancer-promoter functional domains of cell fate determining genes, chromatin structures, and histone modifications undergo substantial changes during neuronal differentiation [8].

Transcription factors (TFs) are essential regulators of gene expression and play critical roles in cell-fate determination. They can bind to promoters and cooperate with transcription complex to activate transcription, or bind to enhancers to elevate cell-type-specific gene expression. After neuronal induction by pro-neuronal transcription factors, the differentiation of excitatory and inhibitory neurons is controlled by a complex network of TFs that operate in a spatiotemporal manner to ensure gene activation at specific stages during brain development. For instance, Neurogenin 2 (Ngn2) and Achaete-scute homolog 1 (ASCL1, also known as Mash1) are critical for the differentiation of functional excitatory neurons [9], while the LIM homeobox (Lhx) and distal-less (Dlx) family members are critical for the differentiation of inhibitory neurons [10]. As neurons mature, the regulation of gene expression becomes increasingly important for synaptic plasticity and remodeling. Transcription factors coordinate with epigenetic modifications at their binding sites to regulate gene expression in a temporal manner [11]. Additionally, key transcription factors mediate changes in the epigenetic landscape at their binding sites, thereby influencing gene expression. For instance, EGR1, a transcription factor involved in memory formation, recruits the DNA demethylase TET1 to remove methylation marks at its binding sites and activate downstream genes [12]. While our understanding of transcriptional regulation during early neuronal development is growing, the role of transcription factors in contributing to neuro-epigenetic diversity remains poorly understood.

Due to the low proportion of inhibitory neurons, histone modification maps for inhibitory neurons were not obtained in the previous study [1]. However, the recently developed CUT&RUN technique, which requires only a small number of input cells, allows for the exploration of epigenetic and transcription factor regulation in inhibitory neurons. In this study, we isolated excitatory and inhibitory neuronal nuclei from the adult mouse brain to perform an epigenetic comparison between these two types of neurons. Using histone modification maps, we identified enhancer regions as the most prominent genomic elements exhibiting distinct histone modifications between excitatory and inhibitory neurons. These enhancers were found to contain abundant binding sites for EGR1, a transcription factor associated with neuronal function. Further CUT&RUN assay suggests that EGR1 plays a key role in facilitating distinct gene regulation patterns in excitatory and inhibitory neurons.

Results

Generation of histone modification maps for excitatory and inhibitory neurons

To obtain nuclei from excitatory and inhibitory neurons separately, we bred Emx1-IRES-Cre knock-in (Emx1^IREScre) mice with the Sun1-tagged mice (Fig. 1A). The Emx1^IREScre mice express Cre recombinase in excitatory neurons and glial cells originating from the Emx1-expressing lineage, but not in GABAergic inhibitory neurons [13]. The Sun1-tagged mice contain a floxed STOP cassette, the removal of which allows the expression of nuclear membrane protein SUN1 with its C-terminus fused to a superfolder GFP (sfGFP) [1]. As such, excitatory neurons, but not inhibitory neurons, are tagged with the SUN1-sfGFP fusion protein in the resulting Sun1 f/f | Emx1-Cre (+) mice. To separate excitatory neurons from glial cells, we further stained nuclei with a fluorescent antibody targeting NeuN, a pan-neuronal marker [14]. With the combination of fluorescent signals from NeuN-immunostaining and sfGFP, we were able to remove glia cells (NeuN − and GFP +) and isolate excitatory (NeuN + and GFP +) and inhibitory (NeuN + and GFP −) neuronal nuclei in high purity via flow cytometry (Fig. 1B). During single nuclei suspension preparation, we used a 30% iodixanol solution to eliminate cell debris based on the differing densities of nuclei and debris. This step ensured a clean single nuclei suspension with minimal cell debris, which was verified through both microscope examination and FACS-sorting check (Additional file 1: Fig. S1). The purity of the post-sorted nuclei suspension was confirmed by FACS, which showed the expected fluorescence signals (GFP and PE) in a portion of the sorted nuclei (Additional file 1: Fig. S2).

Fig. 1.

Summary of CUT&RUN datasets across neuronal subtypes and histone modifications. A Diagrammatic sketch showing experimental design. B Purity of the sorted nuclei for excitatory and inhibitory neurons. C Numbers of reproducible peaks between biological replicates for each histone modification in excitatory and inhibitory neurons. D Genome coverage of each histone modification in excitatory and inhibitory neurons. E The Pearson’s correlation coefficients among histone modifications, neuronal subtypes, and biological replicates. F The distribution of histone modification peaks in annotated genome. Signal intensity of histone modifications surrounding TSSs in excitatory (G) and inhibitory neurons (H). I Histone modification signal at well-known neuronal marker genes, including pan-neuron marker Snap25, excitatory neuron markers Emx1 and Neurod6, and inhibitory neuron markers Prox1 and Reln. RPKM values in 10-bp bins are shown in each panel

Using purified excitatory and inhibitory neuronal nuclei from adult mouse brains, we constructed CUT&RUN libraries to assess histone modifications, including H3K27ac, H3K4me3, H3K4me1, and H3K27me3 (Fig. 1A), with fragment sizes peaking at 168 bp to 175 bp (Additional file 1: Fig. S3A). Highly reproducible peaks (Additional file 2: Table S1) were identified between biological replicates, with Pearson’s R correlations ranging from 0.81 to 0.92 (Additional file 1: Fig. S3B). Additionally, the control IgG signals remained at background levels across all identified histone modification peaks, confirming the antibody specificity and the high quality of our datasets (Additional file 1: Fig. S3C–F). In both types of neurons, the active chromatin marker H3K27ac and the promoter marker H3K4me3 generated approximately 30,000 to 50,000 peaks, while over 100,000 peaks were identified for the enhancer marker H3K4me1 and the repressive marker H3K27me3 (Fig. 1C). The peaks identified for these histone modifications cover around 1.4% to 8.9% of the mouse genome (Fig. 1D). Excitatory neurons tend to have more genomic regions covered by H3K4me1 peaks while the inhibitory neurons host more repressive domains with H3K27me3 peaks. Clustering analysis of histone modification peaks confirmed the strong correlations between biological replicates (Fig. 1E). In addition, the active chromatin marker H3K27ac and enhancer marker H3K4me1 were clustered together for both neuronal types with strong positive correlations, while negative correlations were observed between the repressive marker H3K27me3 and the rest three histone markers. We extended the clustering analysis to include a ChIP-seq dataset for excitatory neurons (Additional file 3: Table S2) published in a previous study [1]. This comparison revealed strong correlations for H3K4me1 and H3K27ac and moderate correlations for H3K27me3 and H3K4me3 between the CUT&RUN data generated in this study and the ChIP-seq data from the earlier study (Additional file 1: Fig. S4).

We next examined the genomic distribution of peaks for all four histone modification markers (Fig. 1F). Not surprisingly, the active and repressive markers showed striking differences in genome distribution. H3K27me3 peaks were predominantly located in intergenic regions, while a significant fraction of H3K27ac and H3K4me3 peaks were found in promoter and 5′UTR regions. As expected, the promoter marker H3K4me3 exhibited strong intensity around transcription start sites (TSSs), followed by H3K27ac. In contrast, H3K4me1 and H3K27me3 peaks were relatively depleted around TSSs (Fig. 1G and H). To further explore the distribution of histone modification signals, we focused on a small set of genes known as neuronal markers. As shown in Fig. 1I, active histone markers were observed at the promoter region of the pan-neuron gene Snap25 in both excitatory and inhibitory neurons. Strong signals for active markers were observed in excitatory neurons surrounding the TSSs of the excitatory neuron marker genes Emx1 and Neurod6. In inhibitory neurons, strong signals for active histone markers were observed for the TSSs of Prox1 and Reln genes, which are the markers for inhibitory neurons.

Comparative histone modification analysis reveals Egr1 as a critical transcription factor in neuronal specification

Previous studies indicated that neuronal cell specification is accompanied by substantial changes in epigenetic signatures [15, 16]. For the four histone modifications, we next determined their differential peaks between the two neuronal subtypes (Additional file 1: Fig. S5A). The percentages of differential peaks between excitatory and inhibitory neurons were found to be higher for the active chromatin marker H3K27ac and the enhancer marker H3K4me1, compared with the other two markers (Fig. 2A). This result indicates that the primary epigenetic differences between excitatory and inhibitory neurons are likely concentrated in enhancer regions. According to the chromatin states inferred from combinations of four histone modifications, we annotated the genome into eight distinct functional regions (Fig. 2B). The regions annotated as active promoters were strongly enriched with H3K27ac and H3K4me3, while active enhancers exhibited more H3K27ac and H3K4me1 peaks. As expected, these histone-modification-based functional annotations closely correlated with genomic annotations derived from the distribution of known genes. For instance, active promoters annotated with histone modifications were found to be overlapped with TSSs, while active enhancers, weak enhancers, and weak active domain were enriched in intergenic, intron, and 3′UTR regions (Fig. 2C).

Fig. 2.

Histone modification differences between excitatory and inhibitory neurons. A Heatmaps showing the difference between excitatory and inhibitory neurons for four histone modifications. Normalized z-scores were plotted in 4-kb windows centered at peaks. B Eight chromatin states annotated with combinatory histone codes. C Enrichment of chromatin states in annotated genome. D Identification of super enhancers in neuronal subtypes. Each dot represents for an enhancer, which is classified into typical enhancer or super enhancer according to the H3K27ac signal. E Examples showing histone modifications in super enhancers. F Images downloaded from the VISTA database showing the reporter gene expression in transgenic mouse embryos under the control of selected enhancers. G Scatterplot showing the proportion changes for TFs’ motifs at super enhancers of excitatory neurons compared with inhibitory neurons. Motifs with p value (determined by binomial test) less than 1e − 65 were colored in red and blue for excitatory and inhibitory neurons, respectively

Since super enhancers are crucial in defining cell identity [17], we further identified these regions for both excitatory and inhibitory neurons using the functional genome annotations (see Methods; Fig. 2D and Additional file 4: Table S3). For example, strong H3K27ac and H3K4me1 signals were observed in the super enhancers nearby the Bcl11a gene in excitatory neurons but were depleted in the corresponding genomic regions of inhibitory neurons (Fig. 2E). In contrast, this pattern was reversed for the super enhancer identified near the Unc5b gene. Previous studies reported that Bcl11a is required for neuronal morphogenesis [18] and controls the migration of cortical projection neurons [19], while Unc5b plays a key role in the regulation of interneuron migration to the cortex [20]. It is noteworthy that within the super enhancers identified for the Bcl11a and Unc5b genes respectively, the activities of two enhancers, hs957 and mm1663, have been validated in transgenic mouse embryos using LacZ reporters [21] (Fig. 2F). The epigenetic states of cis-regulatory elements have an effect on the transcription factor binding and consequently regulate the expression of target genes [22, 23]. To explore the transcription factors under the influence of differential histone modifications determined in the two neuronal subtypes, we summarized and compared the TF motif frequencies within the super enhancers of excitatory and inhibitory neurons. Between the two neuronal subtypes, Egr1, Rfx1, Rfx2, and Mef2c have more motifs identified in super enhancers of excitatory neurons, while Nf1, Zeb2, Tcf4, and Thrb have more potential binding sites in the super enhancers of inhibitory neurons (Fig. 2G). Top in the ranking, Egr1 is an immediate early response gene involving in learning and memory [24]. Our previous study showed that EGR1 binding sites are enriched in the genomic regions hypomethylated in excitatory neurons in mouse frontal cortex [12]. Parallel analysis was performed on active promoters. Pax7, Brn2, Chop, and Oct11 are with motifs enriched in active promoters of excitatory neurons, while Sp5, Boris, Klf6, Klf1, and Znf416 have more motifs identified in active promoters of inhibitory neurons (Additional file 1: Fig. S6). The difference in TF motif frequencies between the promoters of the two neuron types was less striking compared to the differences observed in super enhancers.

EGR1 favors super enhancer regions and has more binding sites detected in excitatory neurons

To explore the binding preferences of EGR1 in excitatory and inhibitory neurons, we generated EGR1 CUT&RUN libraries for both neuronal types using sorted nuclei mentioned previously (Fig. 1A and B). A total of 24,473 reproducible EGR1 peaks were identified in excitatory neurons, while 10,162 were identified in inhibitory neurons (Fig. 3A and B; Additional file 5: Table S4). Interestingly, we found that EGR1 binding sites in inhibitory neurons were predominantly located in promoter, 5′UTR, intron, and intergenic regions, while in excitatory neurons, the binding sites were more frequently distributed in intron and intergenic regions (Fig. 3C). We then annotated the EGR1 binding sites according to their chromatin states. In both types of neurons, EGR1 binding sites were depleted in repressed domains but enriched in active chromatin regions. In inhibitory neurons, 68.9% of EGR1 bindings sites were associated with active promoters, whereas in excitatory neurons, this number dropped to 33.5% (Fig. 3D). In contrast, approximately 42.2% of EGR1 binding sites in excitatory neurons were associated with enhancers. These results were further supported by aggregate analysis using four kinds of histone modifications, which revealed that all three active histone markers were enriched at the EGR1 binding sites in excitatory neurons (Fig. 3E). In inhibitory neurons, only H3K27ac and H3K4me3 were enriched at the EGR1 binding sites. To validate the motif analysis, we examined the proportion of super enhancers containing EGR1 peaks. In both neuronal types, EGR1 binding was significantly enriched at super enhancers compared to randomly shuffled genomic regions (Fig. 3F). For excitatory neurons, 65.5% of super enhancers contain at least one EGR1 peak, while this number dropped to 36.0% in inhibitory neurons. For instance, multiple EGR1 binding sites were located in the super enhancers surrounding the Fhl2 and Cacng3 genes in excitatory neurons, and the Prox1 and Calb2 genes in inhibitory neurons (Fig. 3G). Collectively, these results suggest EGR1 binding is neuronal cell type specific, and in the excitatory neurons, EGR1 binds more frequently to super enhancers.

Fig. 3.

Detection of EGR1 binding sites in excitatory and inhibitory neurons. A Correlations between biological replicates for EGR1 CUT&RUN datasets generated from excitatory and inhibitory nuclei, respectively. B Numbers of EGR1 binding sites identified in excitatory and inhibitory neurons. C Distribution of EGR1 binding sites in genome annotated with gene structure. D Distribution of EGR1 binding sites across various chromatin states. E Average histone modification signals around EGR1 binding sites in excitatory and inhibitory neurons. F Fraction of super enhancers overlapped with EGR1 binding sites compared with those in random selected genomic regions. Statistical significance is determined by permutation test; the comparisons with p value less than 1e − 3 are labeled with ***. G Examples showing EGR1 binding sites and histone modifications around super enhancers

We further explored the differences in EGR1 binding between excitatory and inhibitory neurons by using a two-fold change threshold for differential binding (Additional file 1: Fig. S5B). Among the 26,347 total EGR1 peaks, 6400 were classified as EXC-predominant, 1930 as INH-predominant, and the remaining 18,017 peaks were shared between both neuronal types (Fig. 4A). These 18,017 pan-neuronal EGR1 binding sites displayed similar epigenetic features across both neuronal subtypes, including chromatin accessibility, DNA methylation profile (Additional file 3: Table S2), and histone modifications. Specifically, the three active histone markers—H3K27ac, H3K4me1, and H3K4me3—were observed at pan-neuronal EGR1 binding sites. In contrast, for EXC-predominant EGR1 peaks, stronger signals of H3K27ac and H3K4me1 were observed in excitatory neurons. For INH-predominant peaks, stronger signals of H3K27ac and H3K4me3 were observed in inhibitory neurons (Fig. 4A). These results suggest that EXC-predominant EGR1 peaks may function as enhancers, while INH-predominant peaks may play a role in promoter activity.

Fig. 4.

Epigenetic signatures of EGR1 binding sites. A Epigenetic marks plotted in 4-kb windows centered at EGR1 binding sites. B Selected examples showing neuronal subtype-predominant EGR1 binding, chromatin states, and epigenetic signatures. C Correlations between EGR1 binding and various epigenetic markers. D Fraction of EGR1 binding sites located in the TAD boundary of various cell types compared with randomly selected regions via 1000 times shuffle. Statistical significance is determined by permutation test; the p values less than 1e − 3 are labeled as *** at the top of each comparison. E Contact maps of multiple neuronal subtypes at 50-kb resolution showing EGR1 binding sites in excitatory neuron-specific TAD boundary at Satb2 gene locus. F EGR1 binding signal and histone modifications at Satb2 genes adjacent to an excitatory neuron-specific TAD boundary showing in E. The chromatin state tracks are colored according to the annotation in Fig. 2B

To further explore the association between EGR1 binding and other epigenetic markers, we re-analyzed ATAC-seq and MethylC-seq data for excitatory and inhibitory neurons generated in a previous study [1] (Additional file 3: Table S2). We found that strong EGR1 binding sites were associated with high chromatin accessibility and low DNA methylation levels (Fig. 4A and B), supporting the idea that EGR1 binds primarily to active chromatin regions. We then calculated Pearson’s correlation coefficients between EGR1 binding and various epigenetic markers (Fig. 4C). EGR1 binding exhibited a negative correlation with DNA methylation (− 0.63), consistent with its ability to recruit the DNA demethylase TET1 to its binding sites [12]. Strong correlations were also observed between EGR1 binding and active epigenetic markers, with the strongest correlation to H3K27ac (Pearson’s r = 0.59).

Numerous data demonstrate that the three-dimensional (3D) genome structure plays an important role controlling the interaction of genomic DNA with transcription factors to achieve gene expression regulation [8, 25, 26]. A previous study reported that EGR1 motif is more abundant at pyramidal glutamatergic neuron-specific contacts compared with dopaminergic neurons [27]. To assess the relationship between EGR1 binding and 3D chromatin conformation, we re-analyzed aggregated single-cell diploid chromatin conformation capture (Dip-C) data to infer the topologically associated domain (TADs) for three types of excitatory neurons (cortical layer 2–5 pyramidal cells, cortical layer 6 pyramidal cells, and hippocampal pyramidal cells) and interneurons [28] (Additional file 3: Table S2). We identified that 10.5% and 14.6% of EGR1 binding sites in excitatory and inhibitory neurons, respectively, were located within TAD boundary, significantly higher than randomly selected genomic regions (Fig. 4D). For instance, at the Satb2 gene locus (which is a determinant for upper-layer neuron specification) and its upstream region, EXC-predominant EGR1 binding sites were observed at the excitatory-specific TAD boundary associated with multiple active epigenetic markers (Fig. 4E and F). Collectively, our results demonstrate that EGR1 binding is highly correlated with active epigenetic markers, and it may contribute to the establishment of cell-type-specific chromatin structures.

Differential EGR1 binding is associated with distinct gene pathways and expression program in the two neuronal subtypes

To explore the functional relevance of differential EGR1 binding in excitatory and inhibitory neurons, we inferred the EGR1 target genes by utilizing genomic annotation and 3D chromatin structure. Genes were defined as EGR1 target genes if they contained EGR1 motif-binding sites in the promoter, gene body, and within the same TAD (Fig. 5A). With EXC- and INH-predominant EGR1 peaks, 1688 and 502 genes were identified as EGR1 target genes, respectively. To explore the expression of Egr1 and its target genes, we generated RNA-seq data for excitatory and inhibitory nuclei. We found that Egr1 was strongly expressed in excitatory neurons, showing a fold change of 1.75 compared to inhibitory neurons (Fig. 5B). Since EGR1 may recruit TET1 to remove the methylation marks and activate downstream genes [12], it may serve as a positive regulator of its target genes. As expected, genes with EXC-predominant EGR1 peaks showed slightly but significantly higher expression levels in excitatory neurons compared to those in inhibitory neurons, and vice versa for INH-predominant EGR1 target genes (Fig. 5C and D). To further explore the regulatory functions of EGR1 in these two neuronal subtypes, we performed the GO enrichment analysis for genes associated with EXC- and INH-predominant EGR1 binding sites. Despite the distinct genes associated with each type of EGR1 binding, both gene sets enriched for synapse-related functions such as “regulation of synapse organization” and “regulation of synapse structure or activity” (Fig. 5E and F). Additionally, genes with EXC-predominant EGR1 binding sites were enriched in pathways like “axonogenesis” and “regulation of membrane potential,” highlighting the role of EGR1 in the functional and structural properties of neurons. In summary, our findings suggest that EGR1 plays a critical role in neuronal specification and is involved in a range of regulatory functions that are crucial for the distinct characteristics of excitatory and inhibitory neurons. The differential EGR1 binding in these subtypes is likely important for maintaining their unique functional roles in synaptic plasticity, axonogenesis, and neuronal activity regulation.

Fig. 5.

Functional characterization of neuronal-subtype predominant EGR1 binding sites. A Illustration of EGR1 target gene inference. B Expression of Egr1 in excitatory and inhibitory neurons. Expression of EXC- (C) and INH-predominant (D) EGR1 binding sites associated genes. Statistical significance was determined by t-test; the p values less than 1e − 3 are labeled as *** at the top of each comparison. GO enrichment analysis for EXC- (E) and INH-predominant (F) EGR1 binding sites associated genes

Establishment of EGR1 regulatory networks differs in two neuronal subtypes during brain development

Since EGR1 binding is strongly correlated with active epigenetic markers (Fig. 4C), the changes in chromatin accessibility of EGR1 binding sites could reflect dynamic EGR1 binding. To illustrate how the epigenetic landscape of EGR1 binding sites was established during brain development, we made use of single-nucleus ATAC-seq (snATAC-seq) datasets generated with developing mouse brains from E12.5 to P56 [29, 30]. We focused on excitatory and inhibitory neuronal populations for our analyses (Fig. 6A). The aggregated ATAC-seq data revealed the opening of chromatin at the pan-neuronal gene Snap25 in both types of neurons, while cell-type-specific genes, such as Neurod6 and Dlx5, were accessible only in their respective neuronal populations (Additional file 1: Fig. S7A), confirming the specificity of these datasets. We next calculated the number of accessible EGR1 peaks across the developmental stages. Although this number increased in both neuronal types as development progressed, a considerable fraction of INH-predominant EGR1 peaks were activated at early stages. In contrast, EXC-predominant EGR1 peaks gain accessibility more gradually, particularly accelerating during postnatal stages (Fig. 6B). Aggregated snATAC-seq data revealed that most pan-neuron and almost all INH-predominant EGR1 peaks became accessible before P0, whereas most EXC-predominant EGR1 peaks gained strong signals only in postnatal stages (Fig. 6C). For example, EXC-predominant EGR1 peaks surrounding Herc6 and Fhl2 become accessible only after P0 in excitatory neurons, while INH-predominant EGR1 peaks surrounding Calb2 and Kcnh2 in inhibitory neurons became accessible earlier, in the embryonic stages (Fig. 6D and E). These results suggest that the accessibility of neuronal cell-type-specific EGR1 peaks may be established at different time points during brain development.

Fig. 6.

Chromatin accessibility of EGR1 binding sites in excitatory and inhibitory neurons during brain development. A Single-nucleus ATAC-seq (snATAC-seq) data analysis to show the UMAPs of excitatory and inhibitory nuclei during brain development. B Number of accessible EGR1 binding sites during brain development. The top panel shows the number of accessible EGR1 binding sites in excitatory and inhibitory neurons, respectively. The bottom panel displays the differential number of accessible EGR1 binding sites between excitatory and inhibitory neurons. C Heatmap showing chromatin accessibility of EGR1 binding sites during brain developments. Examples to show the chromatin accessibility of EXC- (D) and INH-predominant (E) EGR1 binding sites during brain development

To further understand the establishment of EGR1 regulatory gene network, we explored the expression of Egr1 together with its target genes in excitatory and inhibitory neurons across developmental stages. We utilized single-cell RNA-seq (scRNA-seq) datasets generated from E12.5 to P60 mouse brains [31–33]. To explore the dynamic gene expression for the two neuronal types during brain development, aggregated analysis of scRNA-seq data generated for each stage was performed (Fig. 7A and B). For example, Snap25 was expressed in both neuronal populations and dramatically increased in postnatal stages, while Neurod6 and Dlx5 were expressed in excitatory and inhibitory neurons, respectively (Additional file 1: Fig. S7B). Despite these “omics” data were generated by different labs, the successful data integration enables us to provide a continuous view of brain gene expression and chromatin accessibility from embryonic stage E12.5 to postnatal P56.

Fig. 7.

Single-cell RNA-seq analysis of EGR1 target genes during mouse brain development. A UMAPs showing the excitatory and inhibitory neurons across mouse brain development. B UMAPs showing the neurons in each developmental stage. C Egr1 expression in excitatory and inhibitory neurons during brain development. D Average expression of cluster 5 of EXC-predominant EGR1 peaks associated genes during brain development. E Average expression of cluster 3 of INH-predominant EGR1 peaks associated genes during brain development. Examples to show the expression of genes associated with EXC- (F) and INH-predominant (G) EGR1 peaks during brain development. H Sketch to show the co-expression between Egr1 and its target genes. I Jaccard index between Egr1 and its target genes during brain development in excitatory and inhibitory neurons

The expression of Egr1 itself increased during brain development in both neuronal types, especially in postnatal stages. Notably, the expression of Egr1 in excitatory neurons was over two-fold higher than in inhibitory neurons at P21 and P60 (Fig. 7C). To examine the effect of EGR1 on its target genes, we performed clustering analysis for genes associated with EXC- and INH-predominant EGR1 binding sites (Additional file 1: Fig. S7C and D). Many EGR1 target genes exhibited expression patterns distinct from Egr1, indicating that additional regulatory mechanisms may also participate in the regulation of these genes. Interestingly, 271 genes associated with EXC-predominant EGR1 peaks and 61 genes associated with INH-predominant EGR1 peaks showed expression patterns synchronized with Egr1 during brain development (Fig. 7D and E). For example, the expression of Fhl2 and Herc6 in excitatory neurons, Calb2 and Kcnh2 in inhibitory neurons were synchronized with Egr1 (Fig. 7F and G). Compared to the 61 INH-predominant EGR1-associated genes, 271 EXC-predominant EGR1-associated genes were more enriched in neuronal functions, such as “regulation of membrane potential” and “regulation of metal ion transport” (Additional file 1: Fig. S7E and F). Additionally, we examined the co-expression relationship between these genes and Egr1 across neuronal types during brain development using the Jaccard index (Fig. 7H). While the Jaccard index increased across brain development in both neuronal subtypes, EXC-predominant EGR1 target genes in excitatory neurons exhibited higher expression levels than those in inhibitory neurons, particularly during postnatal stages, and vice versa for INH-predominant EGR1 targets (Fig. 7I). In conclusion, these findings suggest that Egr1 plays a critical role in regulating gene expression during neuronal specification. It is involved in both excitatory and inhibitory neuronal development, contributing to the regulation of a subset of genes that are critical for the distinct functional roles of these neurons in the developing brain.

Discussion

Current understanding of brain epigenetic regulatory network remains limited, particularly regarding the link between epigenetic programming and neuronal specification. Key questions remain about how distinct epigenetic signatures associated with specific neuronal subtypes contribute to functional diversity. While a small number of datasets have been generated to explore the roles of histone modification and DNA methylation in controlling chromatin loops mediated by transcription factors in a cell-type-specific manner, much remains unknown. For example, Mo et al. generated a comprehensive epigenome dataset for excitatory and inhibitory neurons, including methylomes, ATAC-seq data, and ChIP-seq data for histone modifications [1]. Their comparative analysis provided a link between epigenomic diversity with the functional and transcriptional complexity of neurons. Due to the low proportion of inhibitory neurons, only histone modification maps for excitatory neurons were obtained at that time. Recently, single-cell epigenetics technologies help in gaining insight into the cell-type-specific gene regulatory programs. Zhu et al. developed a Paired-Tag method for joint profiling of histone modifications and transcriptome in single cells and applied it to frontal cortex and hippocampus of adult mice to produce cell-type-resolved maps of chromatin state and transcriptome [16]. Despite these advances, the specific mechanisms by which gene regulation is achieved in neuronal subtypes remain largely unclear.

In this study, we provided genome-wide chromatin state maps with high-coverage CUT&RUN data of four kinds of histone modifications in both excitatory and inhibitory neurons. Our histone modification data for excitatory neurons strongly align with previously published ChIP-seq data from Mo et al. The comparative analysis revealed that differential peaks for H3K4me1 and H3K27ac were more prevalent than for H3K4me3 and H3K27me3, suggesting that cell-type-specific histone modifications are primarily enriched in enhancers. While previous studies have highlighted the functional importance of enhancers in brain cell types and examined its relationship with disease-risk [34, 35], our motif analysis of super enhancer regions in excitatory and inhibitory neurons identified EGR1 as a key transcription factor potentially involved in neuronal specification.

Although we have successfully generated the first neuronal subtype-specific EGR1 binding profiles and comprehensive histone modification maps, several limitations of our CUT&RUN dataset and analysis process need to be considered. First, our data were generated using FACS-sorted NeuN + nuclei, which inherently excludes NeuN − neurons, as previously documented by Kumar and Buckmaster [36]. Second, our study utilized whole mouse brain tissue (excluding the olfactory bulb) to maximize the yield of neuronal nuclei, which may obscure potential brain region-specific difference. Furthermore, integrating multiple public datasets from different brain regions may introduce inconsistencies due to variations in experimental conditions and regional-specific differences. For instance, the snATAC-seq data obtained from previous studies employed Dlx5 + and Hex5 − as markers for inhibitory neurons, whereas our own RNA-seq data demonstrated relatively lower expression levels of Hex5 in inhibitory neurons, highlighting potential discrepancies across different studies. Lastly, current study lacks experimental validation by utilizing samples obtained from Egr1 knockout mice due to technical and resource constraints. The generation and analysis corresponding CUT&RUN data from knockout models represent an important direction for our future research.

Our previous work demonstrated that EGR1 recruits the DNA demethylation enzyme TET1 to activate downstream gene expression during postnatal brain development [12]. Interestingly, DNA demethylation mediated by EGR1 is largely limited to excitatory neurons, though the mechanisms that enable cell-subtype-specific functions of EGR1 are not yet fully understood. In this study, our EGR1 CUT&RUN data suggest that EGR1 may play distinct roles in gene expression regulation between the two neuron types and is involved in the formation of super enhancers in excitatory neurons. Cell-type-specific EGR1 binding is associated with diverse epigenetic features, including DNA methylation, chromatin accessibility, histone modifications, and 3D genomic conformation. This interplay between transcription factors and epigenetic marks may also apply to other neurodevelopmental processes and cell types.

In light of our findings regarding the cell-type-specific regulatory functions of Egr1, we believe that our results may be further extended beyond brain development. As an IEG, EGR1 can be induced quickly in both excitatory and inhibitory neurons upon neuronal activation [37, 38]. Previous studies have demonstrated that other well-known IEGs, such as FOS, bind to neuronal activity-induced enhancers in cortical neurons [39] and play critical roles by interacting with cell-type-specific enhancers [40]. Such a mechanism could trigger downstream regulatory cascades, such as the recruitment of TET1 to facilitate DNA demethylation in a cell-type-specific manner. Altogether, our findings underscore the cell-type-specific functions of IEGs, pointing to a promising direction for future research.

Conclusions

Our comprehensive histone modification and CUT&RUN data across two different neuronal types demonstrate that EGR1 plays a central role in regulating neuronal specification through epigenetic mechanisms. These findings offer valuable insights into the processes that drive neuronal subtype specification and lay the groundwork for future research into cell-type-specific transcriptional regulation in the nervous system.

Methods

Mice

Mice were maintained and bred in a 12-h light/dark cycle under standard pathogen-free conditions. The Sun1 mice (strain #: 021039) and Emx1-IRES-Cre mice (strain #: 005628) were obtained from Jackson Laboratory. Crude DNA was extracted from tail biopsies using Direct PCR tail lysis buffer supplemented with Proteinase K solution and genotyped by PCR according to the Jackson Laboratory’s protocols. Adult (8 weeks) male mouse whole brain (excluding olfactory bulb) samples were used for experiments. All mouse strains used in this study are C57BL/6 background.

Nuclei isolation

Mice were euthanized by inhalation of carbon dioxide (CO2). Cervical dissociation was further performed and the whole brain tissues were rapidly dissected. Nuclei preparation was adapted from previous publications [41, 42]. Briefly, the mouse brain tissue was Dounce homogenized in NE buffer (0.32 M sucrose, 10 mM Tris–HCl pH 8.0, 5 mM CaCl2, 3 mM MgCl2, 1 mM DTT, 0.1 mM EDTA, 0.1% Triton X-100, 1 × Proteinase Inhibitor Cocktail), incubated on ice for 10 min, filtered through 70-μm cell strainer (Miltenyi Biotec, cat# 130–098-462), and spun down at 1000 g for 5 min at 4 °C. The supernatant was removed and the pellet nuclei were further purified using a 30% iodixanol cushion and centrifuged at 8000 g for 20 min at 4 °C. The cell debris on the top of the supernatant were aspirated, and the purified nuclei were pelleted at the bottom of the tube.

Nuclei staining and FACS sorting

The purified nuclei were resuspended in PB buffer (1xPBS with 1% BSA, 1 × Proteinase Inhibitor Cocktail) and incubated with mouse anti-NeuN-PE antibody (Sigma, cat# FCMAB317PE) for 1 h at 4 °C. The stained nuclei were washed twice with PB buffer, resuspended in PB buffer, and subjected to FACS sorting procedures using the BD FACS ARIA Flow Cytometer with gating for GFP and PE signals. Both excitatory neuronal nuclei (GFP + and NeuN +) and inhibitory neuronal nuclei (GFP − and NeuN +) were collected for downstream experiments.

Cleavage under targets and release using nuclease (CUT&RUN)

CUT&RUN was performed using the CUTANA CUT&RUN Kit (EpiCypher, Cat# 14–1048) as previously described [43]. Briefly, the ConA Beads were washed twice and resuspended in Bead Activation Buffer. The FACS-sorted nuclei were pelleted and resuspended in Wash Buffer and mixed with the ConA Beads. The nuclei-bead slurry was incubated on a tube rotator for 10 min at room temperature (RT), allowing the nuclei absorbed to the beads. The nuclei/beads conjugates were resuspended in 50 μL of Antibody Buffer (Wash Buffer with 0.01% Digitonin and 2 mM EDTA) containing 2 μg of H3K27ac antibody (abcam, cat# ab4729), or H3K4me1 antibody (Active Motif, cat# 39,498), or H3K4me3 antibody (Active Motif, cat# 39,060), or H3K27me3 antibody (Active Motif, cat# 39,055), or EGR1 antibody (Santa Cruz, cat# sc101033) and incubated in a tube nutator overnight at 4 °C. The next morning, the nuclei/beads conjugates were washed twice in 200 μL of Cell Permeabilization Buffer (Wash Buffer with 0.01% Digitonin), resuspended in 50 μL of Cell Permeabilization Buffer, and 2.5 μL pAG-MNase (20 × stock) was added. The nuclei/beads conjugates were incubated for 10 min at RT, followed by two washes in 200 μL of Cell Permeabilization Buffer, and resuspension in 50 μL of Cell Permeabilization Buffer. Tubes were chilled on ice, 1 μL of 100 mM calcium chloride was added, and the tubes were nutated for 2 h at 4 °C. Then, 33 μL of Stop Buffer and 1 μL of Spike-in DNA (0.5 ng/μL) were added to each tube. The tubes were incubated for 10 min at 37 °C and placed on a magnet stand until slurry cleared. The supernatant containing CUT&RUN enriched DNA fragments was collected in 1.5-mL tubes and DNA purification was performed using the DNA Cleanup Columns provided in the kit following the manufacturer’s instructions.

Construction and sequencing of CUT&RUN libraries

Libraries for CUT&RUN samples were prepared using the NEBNext Ultra II DNA Library Prep Kit for Illumina (NEB, cat# E7645S) following the manufacturer’s instructions. Briefly, the CUT&RUN enriched DNA fragments were end-repaired and dA-tailed, and ligated to DNA adaptors. After purification with Ampure XP beads (Beckman Coulter, cat# A63880), PCR amplification was performed to enrich adaptor-ligated DNA fragments. Molar concentration of the finished libraries was estimated using a combination of Qubit dsDNA HS assay kit (Thermo Fisher, cat# Q32854) on Qubit 3.0 Fluorometer (Thermo Fisher, cat# Q33218) and Agilent DNA D1000 Screen Tape (Agilent, cat# 5067–5582) on 4150 TapeStation System (Agilent, cat# G2992AA). Individually indexed libraries were pooled and sequenced on Novaseq 6000 platform with paired end 150-bp mode.

RNA extraction

The FACS-sorted neuronal nuclei were pelleted and resuspended in 1 mL TRIzol. After TRIzol/chloroform phase separation, the aqueous phase was collected in a new 1.5-mL tube, and an equal volume of 100% ethanol was added. The mixture was loaded onto the RNA Clean&Concentrator-5 columns. After washing once with RNA Wash Buffer, DNase I solution was added into the column and incubated for 15 min at room temperature to remove any residual DNA. After washing once with RNA Prep Buffer and twice with RNA Wash Buffer, RNA was eluted from the column with RNase-free H2O. RNA concentrations were quantified with Qubit RNA HS assay kit (Thermo Fisher, cat# Q32852) on Qubit 3.0 Fluorometer (Thermo Fisher, cat# Q33218).

RNA-seq library preparation

RNA-seq libraries were prepared using the NEBNext Ultra II Directional RNA Library Prep Kit for Illumina (NEB, cat# E7760S) following the manufacturer’s instructions. Briefly, rRNAs were depleted from the neuronal nuclei RNA samples using NEBNext rRNA Depletion Kit (NEB, cat# E6350S). The rRNA-depleted RNA samples were fragmented, followed by first strand cDNA synthesis and second strand cDNA synthesis. After purification with Ampure XP beads (Beckman Coulter, cat# A63880), the double-stranded cDNA samples were end-repaired, dA-tailed, and ligated with adaptors. After purification with Ampure XP beads (Beckman Coulter, cat# A63880), PCR amplification for 12 cycles were performed to enrich adaptor-ligated DNA. The PCR products were purified by Ampure XP beads. The concentrations of the RNA-seq libraries were quantified by Qubit dsDNA HS assay kit (Thermo Fisher, cat# Q32854) on Qubit 3.0 Fluorometer (Thermo Fisher, cat# Q33218) and the peak distributions were measured by Agilent DNA D1000 Screen Tape (Agilent, cat# 5067–5582) on 4150 TapeStation System (Agilent, cat# G2992AA). The indexed RNA-seq libraries were pooled and sequenced on Novaseq 6000 platform with paired end 150-bp mode.

CUT&RUN data analysis

For all reads derived from CUT&RUN libraries, sequencing adapters and low-quality bases were first trimmed with cutadapt (v1.18, https://github.com/marcelm/cutadapt/) and trim_galore (v0.5.0, https://www.bioinformatics.babraham.ac.uk/projects/trim_galore/). The retained reads were aligned to mouse genome (mm10) using bowtie2 (v2.3.5) [44] in pair-end mode with option “-N 1 -L 25.” PCR duplications were removed using picard with the option “REMOVE_DUPLICATES = true” (v2.25.0, https://broadinstitute.github.io/picard/). Non-redundant reads were further filtered for minimal mapping quality (MAPQ ≥ 30) using samtools view (v1.12) [45] with option “-q30.”

Peak calling for histone modifications was performed by MACS2 (v2.2.5) [46] using option “-p 0.05” for H3K27ac and H3K4me3 and “–broad -p 0.05” for H3K4me1 and H3K27me3. The reproducible peaks between biological replicates were further identified following irreproducible discovery rate (IDR, v2.0.4.2) framework [47] with parameters “–rank signal.value.” Stricter parameters were adopted for EGR1 cun&run datasets peak calling to generate the highly reliable transcription factor binding sites, with option “-p 0.005” for MACS2 and “–rank signal.value –idr-threshold 0.02” for IDR framework. The spurious peaks commonly present in CUT&RUN samples demonstrated by previous studies [48, 49] were downloaded and excluded from the peak regions of all CUT&RUN samples in this study.

The clustering analysis and correlation among neuronal subtypes, markers, and biological replicates

The correlation coefficient between samples was calculated as follows: the RPKM value was generated on a 1-kb window base, the signal score was then summed within each 5-kb window for the entire genome and was compared across different samples. Pearson correlation coefficient was used for all analyses and hierarchical clustering was adopted for clustering analysis.

Detection of neuronal-subtype predominant histone modification peaks and EGR1 binding sites

DESeq2 (v1.30.1) [50] was adopted to perform differential peak analysis between excitatory and inhibitory neurons for H3K27ac, H3K4me3, H3K4me1, H3K27me3, and EGR1, separately. For each marker, firstly, a union peak list of excitatory and inhibitory neurons was generated by bedtools merge (v2.30.0) [51]. The read count in the union peak regions were then calculated and normalized by “count” function in the DESeq2. Finally, the differential peak regions were determined by “result” function in the DESeq2, with thresholds “padj ≤ 0.05” and “FoldChange ≥ 2” or “FoldChange ≤ 0.5.”

Annotation of chromatin states

ChromHMM (v1.23) [52] was adopted to annotate the chromatin states. In brief, BinarizeBam function was first used to divide the mouse genome into 200-bp non-overlapped bins and convert the signal in bam file to binary data in 200-bp bins for each histone modification marker in two neuronal subtypes, respectively. The biological replicates were merged and considered as one sample. LearnModel function was then used to train the prediction model by integrating the four histone modification markers and assign the 200-bp bins into multiple chromatin states. The 8-state model was selected, since it presented maximum number of chromatin states with distinct histone modification marker combinations. The 8 chromatin states were labeled based on their combinations of histone modifications.

Identification of super enhancers

Ranking of super enhancer (ROSE) [17] was used to identify super enhancers. Genomic regions annotated as “active enhancer,” “weak enhancer,” and “strong active domain” were selected and merged as an enhancer pool. The enhancers in the pool located within 12.5 kb from each other were merged and then ranked by the H3K27ac signal. The point with the tangent slope equals to 1 was selected as the inflection point to classify super enhancers and typical enhancers. Enhancers above the point were defined as super enhancers and the rest were defined as typical enhancers.

Motif analysis

Homer software (v3.12) [53] was applied to perform motif analysis. “findMotifsGenome.pl” function was used to search all the motifs in each genomic sequence for super enhancers of excitatory and inhibitory neurons, respectively. For each motif, the percentage of enhancers containing this motif was calculated for excitatory and inhibitory neurons, separately. The binomial test was used to determine the statistical significance of the percentage difference in two neuronal subtypes for each motif.

Re-analysis of MethylC-seq, ATAC-seq, ChIP-seq, and Dip-C data

MethylC-seq and ATAC-seq data of three sorted neuronal subtypes from adult mouse neocortex were downloaded from previous study [1], including excitatory (EXC) neurons, parvalbumin (PV) expressing fast-spiking interneurons, and vasoactive intestinal peptide (VIP) expressing interneurons (Additional file 3: Table S2), each sample with two biological replicates. The data of PV and VIP neurons were merged as inhibitory neurons.

For MethylC-seq datasets, sequence adapters and low-quality bases were filtered with cutadapt and trim_galore. The retained reads were aligned to mouse genome (mm10) using bismark (v0.24.2) [54] with default parameters, PCR duplications were removed using deduplicate_bismark module embedded in bismark software, and genome-wide cytosine methylation report was generated by using bismark_methylation_extractor module in bismark software. The CpG dinucleotides covered by at least 10 reads were retained for calculating the methylation level at EGR1 binding sites and visualization in the genome browser. Within each EGR1 binding region, the methylation level was calculated as the ratio of methylated CpG dinucleotides to the total number of CpG dinucleotides present in the region.

For ATAC-seq datasets, quality control was performed using the same strategy with MethylC-seq datasets. The retained high-quality sequences were aligned to mouse genome (mm10) using bowtie2 (v2.3.5) with parameter “-N 1 -L 25.” The average RPKM value in non-overlapped 10-bp bins were calculated and used for visualization in the genome browser.

ChIP-seq of H3K27ac, H3K4me1, H3K4me3, and H3K27me3 for sorted excitatory neurons from adult mouse brain were downloaded from previous study [1]. Quality control was performed using the same pipeline with MethylC-seq and ATAC-seq datasets. The retained high-quality sequences were aligned to mouse genome (mm10) using bowtie2 (v2.3.5) with parameter “-N 1 -L 25.”

Aggregated scDip-C contact matrix for cortical layer 2–5 pyramidal cells and interneurons was downloaded from GEO datasets with accession GSE146397. HiCExplorer (v3.6) [55] was adopted for data analysis, hicNormalize function was used to normalize the contact matrix, and hicFindTADs function was used to detect TAD.

RNA-seq data analysis

For RNA-seq datasets of excitatory and inhibitory nuclei, adapters and bases of low quality were trimmed using fastp software (v0.23.4) [56] and the remaining reads were mapped to the mouse genome (mm10) by hisat2 (v2.2.1) with bowtie2 (v2.3.5) and quantified by stringtie (v2.2.1) [57] to achieve the expression level of each gene. TPM (transcripts per million) values were adopted as the expression levels in this study.

Functional enrichment analysis

GO enrichment analysis was performed by enrichGO function in clusterProfiler package (v4.14.6) [58] with parameter “ont = “BP”; pvalueCutoff = 0.05.” All the genes expressed in excitatory or inhibitory neurons were used as the background.

Analysis of snATAC-seq data

snATAC-seq data from developing mouse brain at E12.5, E13.5, E14.5, E15.5, E16.5, P0, P21, and P56 were downloaded [29, 30] (Additional file 3: Table S2) and re-analyzed following the instructions in previous study [29] (https://github.com/r3fang/snATAC) with slight modification. For the data in each developmental stage, pair-end sequencing reads were aligned to mouse genome (mm10) using bowtie2, non-uniquely mapped and improperly paired alignments were filtered, and PCR duplications and mitochondrial reads were removed. Macs2 software was used to perform peak calling on retained reads and a read count matrix was generated with peaks in the row and cells in the column. The read count matrix in peaks was then converted to the matrix in promoters, by merging the peaks located in same promoter. The promoter read count matrix was used to perform dimension reduction analysis to cluster and assign the cells into known cell types. Cells with reads located in promoter of Neurod6 were defined as excitatory neurons; cells with reads located in promoter of Dlx5 and without reads located in promoter of Hes5 were defined as inhibitory neurons [29]. Finally, the excitatory and inhibitory neurons were merged, respectively, to produce a pseudo-bulk ATAC-seq dataset for each neuronal subtype in each developmental stage.

Analysis of scRNA-seq data

scRNA-seq data from developing mouse brain at E12.5, E13.5, E14.5, E15.5, E16.5, P0, P7, P21, and P60 were downloaded from previous studies [31–33] (Additional file 3: Table S2); a read count matrix with gene in row and cell in column was achieved in each stage. Seurat (v4.3.0) [59] was adopted to perform data analysis. Briefly, the excitatory and inhibitory neurons were selected and retained for downstream analysis according to the annotations in raw datasets. Remaining neurons with potential double droplets or having mitochondrial mRNA loads over 10% were removed. The genes expressed in less than 3 cells and the cells expressed less than 200 genes were filtered in further analysis. The retained expression read count matrix was normalized by NormalizeData function, and top 2000 variable genes were selected from the normalized matrix using FindVariableFeatures function. Dimensional reduction was performed based on normalized expression matrix of top 2000 variable genes, and the top 10 principal components were used to generate the UMAP (Uniform Manifold Approximation and Projection).

Expression clustering analysis

Clustering analysis for EXC- and INH-predominant EGR1 binding sites associated genes was performed by Mfuzz software (v2.60.0) [60]. By default, a fuzzy c-means clustering was performed on expression matrix with gene in row and sample in column. The fuzzifier parameter (m) was estimated by mestimate function in the Mfuzz. The number of clusters (c) was set to 6, since it is the minimum cluster number satisfies the condition that one of the clusters showing similar expression pattern with Egr1.

Abbreviations

CUT&RUN

Cleavage under targets and release using nuclease

GABA

Gamma-aminobutyric acid

TF

Transcription factor

sfGFP

Superfolder green fluorescent protein

GFP

Green fluorescent protein

EXC

Excitatory neuron

INH

Inhibitory neuron

TSSs

Transcription start sites

Dip-C

Diploid chromatin conformation capture

TAD

Topologically associated domain

GO

Gene ontology

snATAC-seq

Single-nucleus assay for transposase-accessible chromatin with high throughput sequencing

scRNA-seq

Single-cell RNA sequencing

MethylC-seq

Methylcytosine sequencing

ChIP-seq

Chromatin immunoprecipitation sequencing

TPM

Transcripts per million

Authors’ contribution

H. X. conceived and designed the study; H. X., M. F., and X. L. supervised the study; B. C., G. C. and M. F. provided and characterized mouse strains; X. X. isolated neurons and constructed CUT&RUN libraries; L. Y., Y. L. and Y. C. performed the bioinformatic analyses; L. Y., X. X., and H. X. interpreted results and wrote the manuscript. All authors discussed the results and edited the manuscript.

Funding

This work was supported by NIH grant NS094574, the National Key Research and Development Program of China (2023YFA1800500), NIH grant MH120498 and ES031521, the Center for One Health Research at the Virginia-Maryland College of Veterinary Medicine and the Edward Via College of Osteopathic Medicine, the Fralin Life Sciences Institute faculty development fund, the National Natural Science Foundation of China (32150006 and 32200350), Yunnan Revitalization Talent Support Program Yunling Scholar Project (XML), Yunnan Fundamental Research Projects (202201AU070208), and the open project of State Key Laboratory of Genetic Resources and Evolution (GREKF22-07), the National Key Research and Development Program of China (2023YFA1800500).

Data availability

All data generated or analyzed during this study are included in this published article, its supplementary information files and publicly available repositories. The datasets supporting the conclusions of this article are available in the NCBI Gene Expression Omnibus (GEO) with the accession number GSE218312 and GSE291158. Publicly available datasets used in this study are summarized in Additional file 3: Table S2. Data analysis scripts used in this study are available on Zenodo repository (https://doi.org/10.5281/zenodo.15876680).

Declarations

Ethics approval and consent to participate

The animal experiments had been approved prior to the study by the Institutional Animal Care and Use Committee (IACUC) of Virginia Tech.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.