Genomic Glucocorticoid Action in Embryonic Mouse Neural Stem Cells

Abstract

Prenatal exposure to synthetic glucocorticoids (sGCs) reprograms brain development and predisposes the developing fetus towards potential adverse neurodevelopmental outcomes. Using a mouse model of sGC administration, previous studies show that these changes are accompanied by sexually dimorphic alterations in the transcriptome of neural stem and progenitor cells (NSPCs) derived from the embryonic telencephalon. Because cell type-specific gene expression profiles tightly regulate cell fate decisions and are controlled by a flexible landscape of chromatin domains upon which transcription factors and enhancer elements act, we multiplexed data from four genome-wide assays: RNA-seq, ATAC-seq (assay for transposase accessible chromatin followed by genome wide sequencing), dual cross-linking ChIP-seq (chromatin immunoprecipitation followed by genome wide sequencing), and microarray gene expression to identify novel relationships between gene regulation, chromatin structure, and genomic glucocorticoid receptor (GR) action in NSPCs. These data reveal that GR binds preferentially to predetermined regions of accessible chromatin to influence gene programming and cell fate decisions. In addition, we identify SOX2 as a transcription factor that impacts the genomic response of select GR target genes to sGCs (i.e., dexamethasone) in NSPCs.

1. INTRODUCTION

Antenatal administration of synthetic glucocorticoids (sGC) ameliorates infant complications of premature delivery by mimicking the endogenous surge of cortisol that peaks during late gestation^1,2. However, potential adverse neurodevelopmental consequences of antenatal sGC exposure include short-term cortical architectural changes in mice, sheep, and non-human primates, and long-term behavioral or cognitive impairments in humans^3–9. Mouse models of sGC administration revealed that a single dose of the glucocorticoid receptor (GR) agonist Dexamethasone (Dex) at embryonic day E14.5 altered neural stem and progenitor cell (NSPC) proliferation and differentiation in vivo, while in-vitro experiments with primary embryonic NSPCs demonstrated robust and sex-specific changes in gene expression following acute Dex treatment^7,10.

Cerebral cortical development represents a dynamic period of chromatin remodeling that directs neurogenesis and other cell fate decisions, with subpopulations of cells within neuronal or glial lineages expressing distinct global chromatin signatures^11,12. This is a flexible epigenetic process guided by specialized enzymatic complexes and transcription factors^11–14. Furthermore, gene-regulatory enhancer regions upon which transcription factors act are often located far from their target promoters, highlighting the need for genome wide assessment of chromatin architecture to fully understand how developmental programming of gene expression is established and/or maintained¹⁵.

Transcriptional regulation by GCs is driven by GR, which associates directly or indirectly with DNA to activate or repress target genes¹⁶. Both chromatin accessibility and histone modifications play a major role in determining de-novo genomic GR occupancy^17–19. However, some GR binding sites occur in less permissive chromatin or genomic sites lacking distinct histone modifications^17,20,21. Importantly, cell type-enriched co-factors or coregulators create unique GR occupancy patterns in different cell types by enhancing the receptor’s ability to associate directly with specific DNA sequences (i.e., glucocorticoid response elements or GREs), or recruiting GR indirectly to genomic sites occupied by other transcription factors^22,23.

While antenatal exposure to sGC reprograms the neurodevelopmental trajectories in the fetal forebrain, the molecular signatures that direct basal and hormone-induced genomic GR action in NSPCs have not previously been elucidated. In this study we characterized the chromatin landscape and GR cistrome of embryonic mouse NSPCs to determine whether robust acute sGC-induced alterations in gene expression are accompanied by changes in chromatin accessibility and GR distribution at regulatory genomic sites that control NSPC fate.

2. METHODS

2.1. Animals and Cell Culture

Cortical NSPC cultures derived from the C57BL/6 mouse embryonic telencephalon (E14.5) were generated in Neurocult media supplemented with epidermal growth factor (EGF), fibroblast growth factor-1 (FGF-1), penicillin streptomycin, and proliferation growth supplement (StemCell Technologies), in accordance with the technical manual provided by StemCell Technologies and the ethical and Environmental Health & Safety practices required by the University of Pittsburgh IACUC and the National Institutes of Health^10,24,25. Cultures from each embryo were kept separate, and fetal sex was determined by digesting tail tissue overnight in 200μl of nonionic detergent buffer with 1.2μl of proteinase K at 56°C. Afterwards, samples were heat inactivated at 95°C for 10 minutes and isolated DNA subjected to PCR analysis to detect the Y chromosome Sry gene¹⁰.

For drug treatment after the third passage (P3), 100nM Dexamethasone (Dex), a saturating concentration for GR binding that has been used extensively to characterize genome-wide responses in NSPCs, or an ethanol vehicle, was added directly to the culture media for 4 hours^7,10,25. Cells were removed from the culture media and collected immediately afterwards for analyses. ATAC-seq samples were technical replicates (n=3) derived from one female embryo, while biological replicates (n=3) combining neurospheres from multiple female embryos per replicate were used for GR ChIP-seq analysis.

2.2. Omni ATAC-seq

A detailed protocol and library assessment for Omni ATAC-seq is described by Corces et al., and is outlined in Supplementary Table 1²⁶. In brief, 100,000 cells with >90% viability were lysed in ice-cold buffer (10mM Tris-HCl pH 7.4, 10mM NaCl, 3mM MgCl₂, 0.015% NP-40). Nuclei isolated by centrifugation were resuspended in transposase reaction mix containing a 1:20 dilution of Transposase (TDE1) in 1X TD buffer (Illumina) and incubated at 37°C for 30 minutes with mixing. Fragmented product was purified using a DNA Clean and Concentrator-5 kit (Zymo Research) according to the manufacturer’s instructions. DNA fragments underwent 3–5 cycles of preamplification to incorporate Illumina adapter sequences. qPCR was carried out on 10% of the pre-amplified product to determine if further amplification cycles were required. The amplified library was purified using Ampure beads (Beckman Coulter). Following quality and quantity assessment with Qubit (Invitrogen) and HSD1000 Tape Station assay (Agilent), libraries were diluted to 10pM, pooled, and sequenced with NextSeq 500 (Illumina). Flowcells for the NextSeq 500 were seeded with 1.8pM denatured library for automated cluster formation and 2 × 75 paired end sequencing, approximately 50 million reads per sample.

2.3. Dual Cross-Linking Chromatin Immunoprecipitation for Next Generation Sequencing (ChIP-seq)

A detailed protocol is described by Rollins and Rogatsky et al.,²⁷. In summary, Dynabeads (ThermoFisher) were washed three times with 1% BSA in PBS on ice and resuspended with 1–10ug of antibody against GR (ThermoFisher PA1–511A) overnight at 4°C. Approximately 5 × 10⁶ cells, as a single cell suspension, were incubated in 0.2 mM DSG-PBS (ProteoChem #c1104) for 30 minutes at room temperature (RT) prior to a 10-minute incubation at RT in a 1:16 diluted solution of 16% methanol-free formaldehyde and fixing buffer (10X prepared at 500mM HEPES-KOH, pH7.5, 1M NaCl, 10mM EDTA, pH 8.0, 5 mM EGTA, pH 8.0). Cross-linking was terminated by addition of 2.5M glycine for 5 minutes at RT, and following multiple washes in PBS, cell pellets were flash frozen in liquid nitrogen and stored at −8°C until further processing. Fixed cells were incubated in lysis buffer (50 mM HEPES-KOH, pH 7.5,140 mM NaCl, 1mM EDTA, pH 8.0, 10% glycerol, 1% NP-40, 0.25% Triton X-100 containing protease and phosphatase inhibitors) for 10 minutes at 4°C and spun down at 600g for 10 minutes at 4°C to isolate cell nuclei. Following several washes (10mM Tris-HCl, pH 8.0, 0.2 mM NaCl, 1 mM EDTA, pH 8.0, 0.5 mM EGTA, pH 8.0), nuclei were resuspended in shearing buffer (0.1% sodium dodecyl sulfate (SDS), 10mM EDTA, pH 8.0, 50 mM Tris, pH 8.0), incubated for 15 minutes on ice, and sonicated using a Diagenode Bioruptor Pico water bath sonicator. Following sonication, nuclear lysates were cleared by centrifugation at 14,000 g × 20 minutes at 4°C and transferred to a DNA LoBind (Sigma) 1.5mL tube. Nuclear lysates were incubated with the antibody-conjugated Dynabeads overnight at 4°C. The next day, the Dynabeads antibody-conjugated complexes were washed eight times with modified RIPA buffer containing protease and phosphatase inhibitors, and finally washed once with TE (10mM Tris, pH 8.0 and 1 mM EDTA, pH 8.0) containing 50mM NaCl. DNA was eluted into TE containing 0.5% SDS with Proteinase K at 55°C for 1.5 hours. Cross linking was reversed by incubation at 65°C for a minimum of 6 hours. Purified DNA was purified using the QIAquick PCR Purification Kit (Qiagen) according to manufacturer’s directions. 2uL of eluted DNA was set aside for qPCR analysis, and 2uL was used for quality analysis using a Qubit fluorometer and Agilent Bioanalyzer. Remaining eluted DNA was stored at −20°C until NextGen sequencing using an Illumina HiSeq 2500 system.

2.4. ATAC-seq Data Analysis

Quality control of raw sequenced reads, which determines total number of reads, read length, and GC content among other measures, was performed using FASTQC. After sequencing, adaptors were removed using Trim Galore! with --paired –nextera parameters. Paired-end reads without adapter sequences were aligned against the mm10 reference genome using bwa mem with default parameters. Mitochondrial reads were removed using a python script removeChrom. PCR artifacts and duplicates were removed using MarkDuplicates, available in the Picard toolkit. A python script, SAMtoBED, was used to convert the read alignments (BAMs) into paired-end BED format (BEDPE) for downstream peak calling. ATAC-Seq peak regions of each sample were called using MACS2 with parameters -f BEDPE -g mm. ATAC-seq specific quality metrics were evaluated with ATACseqQC R package²⁸. An average of ~56 million paired-end sequencing reads per sample were procured, after removal of mitochondrial reads and mark duplicates, and the average fraction of reads in peaks (FRiP) was 0.06 (Supplementary Table 2). The insert size analysis revealed that approximately 50% of all sequenced reads map the content of <1 nucleosome (<150bp) (Supplementary Figures 1–6). This agrees with the expected results as described by Buenstoro et al., though we did not observe an abundance of nucleosome multimer-sized fragments beyond a dimer²⁹. ATAC-seq peaks were called in each of three technical replicates using MACS2 (n=3, p<0.05) (Supplementary Table 3), and consensus peaks occurring in at least 2 replicates were used in all successive analyses (Supplementary Tables 4–5). A heat map of read densities shows a symmetric distribution of mapped reads relative to transcriptional start sites (TSSs) (Supplementary Figure 7).

The DiffBind (DBA) R package was used for identifying differential sites between two experimental groups³⁰. An experimental design sample table, which includes a set of peaks and associated metadata, was generated as described in the DiffBind manual. The experiment DBA object is constructed through dba() using this sample table. Deseq2 results were extracted from DiffBind, with default normalization of full library sizes of the samples. Reads are modeled in a generalized linear model framework of a two-group comparison. We identified significantly different peaks with FDR<0.05 in two comparison groups (Supplementary Table 6). A merging function was used to find all overlapping peaks from our GR ATAC-seq data and published H3K27ac ChIP-seq data (GEO104686) and derive a single set of unique genomic intervals covering all the supplied peaks (Supplementary Tables 7–9)³¹. dba.count() with option minOverlap=0.5, summits=200, was then used to take the alignment files and compute count information for each of the peaks/regions in the consensus set. To determine whether open chromatin regions from the ATAC-seq analysis correlated with transcription, overlap between ATAC-seq peaks with −5kb/+3kb from the TSS of significantly Dex-upregulated genes (log FC>0.8, p<0.05; n=300) was computed by bedtools intersect (Supplementary Tables 10–12). These genes were previously identified using RNASeq data¹⁰.

2.5. GR ChIP-seq Data Analysis

Quality control of raw sequenced reads, which determines total number of reads, read length, and GC content among other measures, was performed using FASTQC. Adaptors were removed from the sequenced reads using Trim Galore! with -q 20 --stringency 2 parameters. Single-end reads were then aligned against the mm10 reference genome using bowtie2 with default parameters. ChIP-seq peaks were called using MACS2 with parameters -B --SPMR --qvalue .05 --gsize mm --nomodel --extsize 200. An An input sample of DNA, which was cross-linked and sonicated but not immunoprecipitated with the GR antibody, was used as background for MACS2. Strand cross correlation metrics show significant clustering of enriched DNA sequence tags at the locations bound by the protein of interest (Supplementary Figure 8). A standard peak list was established by merging the union of all peaks which occur in any replicates within a treatment condition. Using the same parameters described for ATAC-seq, the DiffBind (DBA) R package was used for identifying differential sites between two groups³⁰. Hypergeometric Optimization of Motif Enrichment (HOMER v4.10.3) package annotatePeaks.pl was used for performing peak annotation. Known motif enrichment analysis was carried out by findMotifsGenome.pl (HOMER), searching for motifs of 8, 10, and 12 bp in length within the ± 200 bp flanking regions of the peak summits. p values were calculated by comparing the enrichments within the target regions and those of a random set of regions (background) generated by HOMER. Overlap between two feature files was determined with the intersect -wa -wb function in bedtools v.2.27.1. For example, the overlap between ATACseq differential peaks with H3K27ac consensus peaks are determined by bedtools intersect -wa -wb -filenames -a atacseq.peaks -b H3K27ac.consensus.peaks.

2.6. Proximity Ligation Assay

A detailed protocol for the Duolink^® proximity ligation assay (PLA) (Sigma Cat no. DUO92102) is described by Alam³². After the third passage, NSPCs derived from the embryonic telencephalon (E14.5) were plated in 24-well cell culture plates (Falcon) containing glass coverslips coated with Poly-D-Lysine (Sigma P1524) and Laminin (Corning^™ via Fisher CB-40232, Mouse). 24 hours after plating, cells were treated with vehicle or Dex for 4 hours as previously described, then fixed in a 10% neutral buffered formalin solution (ThermoScientific REF9990244) for 30 minutes at 4°C¹⁰. The Duolink^® PLA protocol was followed, using 80ul reaction volumes. In brief, cells on coverslips were incubated with Duolink^® blocking solution for 1 hour in a heated humidity chamber at 37°C with gentle shaking, then overnight with a SOX2 anti-rabbit antibody (1:500) (Abcam ab97959) and a GR anti-mouse antibody (1:500) (Invitrogen MA1–510) at 4°C with gentle shaking. To control for non-specific PLA-probe activity, the GR anti-mouse antibody (Control^GR) or the SOX2 anti-rabbit antibody (Control^SOX2) were omitted from the overnight incubation, followed by the complete experimental protocol. The next day, cells were washed 2 × 5 minutes in Wash Buffer A and incubated with mouse MINUS and rabbit PLUS PLA probes at a 1:5 dilution in Duolink^® antibody diluent for 1 hour in a preheated humidity chamber at 37°C. Coverslips were washed in Wash Buffer A as previously described and incubated with ligase, at a 1:40 dilution in 1X ligation buffer, for 30 minutes in a preheated humidity chamber at 37°C. Coverslips were washed again and incubated with polymerase at a 1:80 dilution in 1X amplification buffer for 100 minutes in a preheated humidity chamber at 37°C. Coverslips were washed 2 × 10 minutes in Wash Buffer B. Nuclei were stained with Dapi in PBS (1:10,000) for 1 minute, washed a final time with PBS, then mounted on glass coverslips and stored at 4°C protected from light.

2.7. Confocal Microscopy and PLA Quantification

Each experimental group of cells was plated on 4 coverslips of a 24-well plate, and 6 images were taken per slide per embryo for a total of 24 images per embryo per treatment group (for n=3, 72 images per treatment group). Using an FV1000 Olympus confocal microscope, PLA probes were visualized at 60X magnification in the Alexa Fluor 594 channel (excitation wavelength 543nm, emission wavelength 618nm), with variable voltage detector levels (<650), minimal gain (<2) and no photobleaching (offset <12) (lens parameters: sampling speed 2.0 us/Pixel, 12bits/pixel, 640 pixels total). Nuclei were visualized with the same parameters in the DAPI channel. To quantify PLA probe photoactivation, each nucleus was outlined and defined as an individual region of interest (ROI), and the average intensity profile was determined for the ROIs. Overall average intensity is defined as the average ROI per cell per image per embryo. For downstream analyses, probe-positive cells are defined as a ROI with an average intensity profile >100, due to baseline levels of background photoexcitation in the Alexa Fluor 594 channel present in the negative control samples (average background intensity profile ~59). To compare the average abundance of PLA probe photoactivation between all four experimental groups, One-way ANOVA test was performed to quantify significant differences among group means. Multiple comparisons were performed when necessary. For comparisons between two experimental groups, an unpaired two-tailed t-test was used.

2.8. Generation of SOX2 KO NSPC Cultures and Microarray Gene Array

Sox2^loxP and Sox2^loxPΔneo alleles and mice were generated as described by Favaro et al., 2009³³. Neurosphere cultures of SOX2 knockout (KO) NSPCs and control non-deleted wild-type NSPCs (C57BL/6) derived from the fetal telencephalon were established at postnatal day (P0) as previously described^33–35. The establishment, maintenance, and expansion of SOX2 ablated neurospheres has been found to be optimal with cerebral cortical tissue derived from PO mice^33,36. It is important to note that SOX2 KO NSPC cultures retain their self-renewal potential for up to 7–10 passages (~30 days), followed by a decline in proliferation capacity³³. After the third passage, biologically distinct replicates of both sexes were treated in-vitro with 100nM Dex for 4h and processed for microarray gene expression analysis (n=6 wild-type; n=7 SOX2 KO).The RNeasy micro kit (Qiagen) was used to extract total RNA. RNA integrity was determined by both the absorption ratio (260/280) of ≥1.8, as well as the RNA integrity value of ≥8.0 measured by a Bioanalyzer 2100. 100ng of purified total RNA underwent transcription in-vitro via the MessageAmp Premier Enhanced assay protocol (Thermo Fisher Scientific). The diversity of cRNA was confirmed by Nanodrop, which generates one electrophoretogram per reaction with sample integrity, yield, and size diversity against a Universal Human Reference RNA. Following purification and amplification, 15 μg of biotin-labeled cRNA was fragmented and hybridized to the Affymetrix Mouse Clariom S^™ array in accordance with the manufacturers protocol (ThermoFisher Scientific). After 18 hours, the arrays were washed and stained on an Affymetrix Fluidics Station, then immediately scanned using a Scanner 3000 after hybridization.

Microarray Data Analysis

Of the 22,206 genes measured, transcriptome analysis software identified those with a p-value <0.05, considered as having a significant change in expression in Dex-treated groups compared to vehicle controls. Significant (p<0.05) and robust fold changes (−1.5 ≤ FC ; 1.5 ≥FC ) induced by Dex within WT or SOX2 KO groups were compared to reveal how SOX2 ablation alters the direction and intensity of gene response. All microarray gene expression data was uploaded to IPA (QIAGEN Inc., https://www.qiagenbio-informatics.com/products/ingenuity-pathway-analysis), containing the gene identifiers and their corresponding expression, fold-change, and p-value, among other common metrics. Using the “build” and “overlay” pathway functions, the gene identifiers were then sorted by fold change and mapped to the corresponding gene object in the “glucocorticoid receptor signaling” and “FGF signaling” canonical signaling pathways, determined in the Ingenuity Pathway Knowledge Base (IPKB).

2.9. Data and Code Availability

ATAC-seq and GR ChIP-seq, as well as microarray data are openly available via the Gene Expression Omnibus resource (GEO: GSE175850 and GSE222392).

3. RESULTS

3.1. Genome-wide Open Chromatin Profiling of Embryonic Cortical NSPCs Reveals Putative Enhancer Regions Involved in the Transcriptional Response to Dex.

To characterize the acute transcriptomic response to a clinically relevant sGC that results in altered cell fate outcomes in primary embryonic NSPC cultures and in the developing mouse brain, we modeled single-dose sGC administration during mid-gestation by treating primary embryonic NSPCs (E14.5) with 100nM Dex or an ethanol vehicle for 4 hours and determined by ATAC-seq the global chromatin landscape (Figure 1A)^7,10. Accessible regions of chromatin were highly correlated between vehicle-treated (n=26,326) and Dex-treated NSPCs (n=28,798) (Pearson’s correlation analysis, mean r =0.97, Supplementary Figure 9). Accessible chromatin was distributed densely nearby the TSS (±2kb) (Supplementary Figure 7), and may have gene regulatory potential as they are physically permissive to sequence-specific transcription factors and protein complexes necessary for the initiation of transcription^29,37. However, the majority of ATAC-seq peaks were detected within intronic or intergenic regions of the genome (Supplementary Figure 10). These may represent enhancer, promoter, or insulator regions that are physically permissive to interactions with gene regulatory chromatin-binding factors in NSPCs^29,37. Similar findings are reported in models of human forebrain development^12,37,38. To predict which ATAC-seq peaks occur at active transcriptional enhancers, defined as distal genomic regions denoted by distinct histone modifications associated with an active role in transcriptional regulation, we utilized publicly available H3K27ac ChIP-seq data obtained in untreated E14.5 mouse NSPC cultures^{19,31,39–41}. The isolation of cortical NSPCs, culture preparation, and culture maintenance used for this data set closely resembled ours, and based upon studies conducted in MM.1S, T47D and A1–2 cells, acute Dex exposure does not change the vast majority of the H3K27ac cistrome, though it may create some hormone-dependent regions with increased transcriptional regulation capability^42,43. ~75.7% (n=19,919) and ~78.7% (n=22,673) of ATAC-seq peaks in vehicle-treated or Dex-treated NSPCs, respectively, overlap with H3K27ac ChIP peaks and thus, likely correspond to promoter or enhancer regions (Figure 1B).

Figure 1. Mapping Open Chromatin by ATAC-seq in Embryonic Cerebral Cortical NSPCs.

A) Schematic diagram overview of bioinformatic processing of vehicle-treated or Dex-treated NSPCs (E14.5). The black dotted line represents the cut-off threshold (p<0.05) for called ATAC-seq peaks (top) or ChIP-seq peaks (bottom) (boxed in). Genomic regions are categorized as a) constitutively accessible and GR-bound in both treatment conditions, b) differentially accessible and GR-bound in Dex-treated NSPCs only, c) constitutively accessible but not GR-bound in either condition, or d) constitutively accessible in both treatment conditions but GR-bound only in the presence of Dex. Multi-Omics Data Integration (right) allows visualization of (i) peaks proximal to the TSS of a Dex-regulated gene, (ii) peaks enriched for a glucocorticoid responsive element (GRE) or SOX TF motif, and (iii) peaks with a H3K27ac histone modification.

B) Percentage of consensus ATAC-seq peaks in vehicle (75.7%, n=19,919) or Dex-treated (73.6%, n=21,182) cortical NSPCs (E14.5), which overlap with the promoter/enhancer mark H3K27ac detected in cortical NSPCs (E14.5) from an independent study (Gene Expression Omnibus accession no. 104686).

C) Percentage of Dex-induced target genes, identified by RNA-seq (log FC>0.8, n=300), with at least one consensus TSS-proximal (−5 kb/+3kb) ATAC-seq peak in vehicle- or Dex-treated cortical NSPCs (E14.5) (Vehicle, 91.33%, n=274) (Dex 92.7%, n=278).

D) Constitutive and Dex-induced ATAC-seq peaks in the proximal (−5kb/+3kb) regulatory region of the Dex-induced Hif3α gene. TSS is marked by a green line. The red and blue brackets indicate constitutive and induced peaks, respectively.

We utilized published gene expression data collected in NSPCs to assess the chromatin landscape surrounding Dex-regulated genes, focusing on ATAC-seq peaks located proximal (−5kb/+3 Kb) to the TSS of Dex-induced genes (log FC>0.8, n=300)¹⁰. This −5kb/+3kb TSS-proximal range was chosen because it includes promoter and surrounding non-promoter regions bound by RNA Pol-II in the mouse brain genome⁴⁴. Thus, this range spans the region of Dex-regulated genes that are likely to be closely associated with actively transcribing RNA Pol-II. In both vehicle and Dex treatment groups, a vast majority (274 and 278, respectively) of the 300 highly Dex-induced genes had one or more regions of proximally located accessible chromatin (Figure 1C, Supplementary Table 10–11). A small subset of these genes underwent a change in chromatin structure surrounding their TSS in response to Dex (4 hour), with 11 proximally located accessible regions appearing only in Dex-treated NSPCs and 1 appearing only in vehicle-treated NSPCs (ATAC Peak ID #8) (Supplementary Table 12).

These results suggest that for the vast majority of GR-inducible genes in mouse NSPCs, the chromatin landscape surrounding their TSS is not significantly changed within 4 hours of Dex treatment. Furthermore, promoters of most Dex-inducible genes are already in an accessible chromatin state prior to Dex exposure. Our data agrees with studies conducted in other cell types which report ~95% of de novo GR binding occurring in pre-established open regions of chromatin^17,45. In line with this, only 95 ATAC-seq peaks (~0.24% of total peaks) were present only in Dex-treated NSPCs (p<0.05) with 12 of those located within −5kb/+3 Kb of a Dex-induced gene (Hif3a, Raver2, Cables1, Lpp, Rgs20, Bcl2l1, Ptk2b, Pknox2, Smox, Phyhd1, Ahcyl1). For example, Hif3a is a significantly and robustly Dex-upregulated gene with a Dex-inducible accessible region of chromatin located −5kb/+3kb from the TSS (i.e., within its first intron; Figure 1D, blue bracket) in addition to a constitutively open chromatin region within the same intron (Figure 1D, red bracket).

3.2. GR Binding and the Chromatin Landscape in Embryonic Cortical NSPCs

ChIP-seq was performed to identify GR-bound regions of DNA in NSPCs treated with 100nM Dex or vehicle for 4 hours (Figure 1A). This revealed that 941 sites are bound by GR in vehicle-treated NSPCs, whereas Dex addition resulted in a ~236.0% increase in GR binding sites (n=3162). To relate GR binding pattern to chromatin accessibility, we overlayed the ATAC-seq and GR ChIP-seq profiles which revealed that ~92.2% (n=2915) and ~79.6% (n=2519) of GR binding occurred in constitutively accessible (i.e., pre-accessible) chromatin in vehicle- or Dex-treated NSPCs, respectively, as opposed to a minority of GR binding which occurred at inaccessible chromatin (Figure 2A). Indeed, the genomic distribution of GR ChIP-seq peaks is similar to that of the ATAC-seq peaks, as over ~70% of GR binding occurred in intronic or intergenic regions (Supplementary Figure 11).

Figure 2. GR Binding and the Chromatin Landscape in NSPCs.

A) Quantification of GR ChIP-seq peaks that occur in vehicle- (Pre-Dex, n=941) or in Dex-treated NSPCs (Post-Dex, n=3162). Within these groups, ChIP-seq peaks occurred at (i) weak or inaccessible regions of chromatin, or (ii) pre-accessible sites of chromatin.

B) Quantification of GR ChIP-seq peaks (y-axis) that occur in (i) accessible regions of chromatin, which have an H3K27Ac mark (Ac. H3K27ac+), (ii) accessible regions of chromatin that lack an H3K27ac mark (Ac. H3K27ac-), or (iii) inaccessible regions of chromatin (x-axis) in vehicle- or Dex-treated NSC. Genomic localization of H3K27ac was delineated in cortical NSPCs (E14.5) from an independent study (Gene Expression Omnibus accession no. 104686). Fisher’s exact test of independence between category (i) and (ii) (***p<0.01).

C) ATAC-seq and GR ChIP-seq peaks located in a H3K27ac+ distal enhancer region of DNA (frame view chr18:16,552,637–63,558,00). Horizontal rows from top to bottom: ATAC-seq peaks in triplicate of vehicle- or Dex-treated NSPCs (blue), ChIP-seq peaks in triplicate of vehicle- or Dex-treated NSPCs (pink). Genomic regions with a ChIP-seq peak for the H3K27ac promoter/enhancer mark detected in cortical NSPCs (E14.5) from an independent study (GEO104686) are indicated by the black bracket (top). The RefSeq horizontal row indicates location of any protein coding regions in the murine (mm10) reference genome (bottom). The nearest gene to this position is ~100 kb away (out of frame).

D) Enrichment of non-random consensus TF binding motifs in our ChIP-seq reads relative to random background, determined by the HOMER motif algorithm. Glucocorticoid responsive element; GRE. Androgen responsive element; ARE. Progesterone receptor; PGR. Nuclear factor-1; NF-1. Androgen receptor; AR.

E) Percentage of ChIP-seq peaks enriched for GR or SOX TF motifs in cortical NSPCs exposed to vehicle or Dex in-vitro. The number of ChIP-seq peaks are indicated at the bottom of each chart. GRE; glucocorticoid responsive element.

F) Percentage of ChIP-seq peaks enriched for a GRE full-site or GRE half-site using the HOMER motif enrichment algorithm. Fisher’s Exact test (***p<0.0001).

Our results agree with findings in other cell types which show open chromatin in distal genomic regions as a biological prerequisite for the majority of GR binding^17,18,45. This led us to investigate whether GR binding occurs preferentially in regions marked by the presence of histone H3K27ac, a marker of active enhancers¹⁹. We utilized our chromatin accessibility data, along with publicly available H3K27ac ChIP-seq data, to characterize GR binding in (i) H3K27ac-positive accessible regions (H3K27ac+), (ii) H3K27ac-negative accessible regions (H3K27ac-), or (iii) inaccessible regions of chromatin (Figure 2B)³¹. A Fisher’s test of independence revealed that GR binding in vehicle vs. Dex-treated NSPCs is associated with H3K27ac presence (***p<0.01). An example of GR binding at H3K27ac+ accessible chromatin is depicted in Figure 2C.

Next, we aimed to identify TF binding motifs that underlie each GR ChIP-seq peak to delineate genomic regions that are (i) bound by GR, or (ii) occupied by potential GR pioneering factors or cofactors. The HOMER motif discovery algorithm was used to determine enrichment of non-random consensus TF binding motifs in our ChIP-seq reads relative to random background (Figure 2D). The Homer algorithm classified GR ChIP-seq peaks as occurring at a canonical GRE if they were enriched (p<1E-100) for a motif containing an inverted palindromic sequence with two consensus ‘AGAACA’ motifs separated by 3 nucleotides^16,21,46,47. This analysis showed that although a significant fraction of GR binding events occurred even in the absence of added Dex (Figure 2A), only ~4% (n=41) of GR ChIP-seq peaks contain a canonical GRE in vehicle-treated NSPCs (Figure 2E, top). These data suggest that in the absence of added GCs, GRs in primary NSPC cultures may have the capacity to occupy genomic regions that do not contain the classic palindromic sequence of a GRE. Possible rationale for these instances is provided in the discussion. In contrast, with the addition of Dex, there is a ~20-fold increase in the total number of GR ChIP-seq peaks enriched for a GRE (n=823, 26% of total) (Figure 2E, bottom). Since the Homer algorithm only predicts canonical GREs, the percentage of GR-ChIP peaks directly occupying DNA in Dex-treated NSPCs (i.e. 26%) may be an underestimate since GR binding sites with minor variations in the canonical GRE sequence will not be detected^48,49. Nonetheless, these data suggest that Dex-induced GR binding to the genome is facilitated through direct GRE binding in NSPCs.

The second-most significantly enriched motif within the GR ChIP-seq peaks belonged to the SOX family of transcription factors (TFs). We combined GR ChIP-seq peaks enriched ((p<1E-10)) for any SOX TF motif expressed in E14.5 cortical NSPCs (SOX2, SOX3, SOX4, SOX6, SOX9, SOX10, SOX15) because their consensus binding motifs are similar in sequence and position weight matrix^10,50,51. We observed enrichment of SOX TF binding motifs in ~33% (n=308) of GR ChIP-seq peaks in vehicle-treated samples (Figure 2E, top). In contrast, with the addition of Dex, there is a ~4-fold increase in the number of GR ChIP-seq peaks enriched for a SOX motif (n= 1319, 42% of total). (Figure 2E, bottom). In 88% of instances where a ChIP-seq peak was enriched for both a GRE and SOX motif in Dex-treated NSPCs, the motifs were located less than 100 nucleotide bases apart (Supplementary Figure 12). These data suggest that in addition to direct GRE binding, Dex-induced genomic GR activity preferentially occurs at sites that are co-occupied or closely associated with the binding of SOX TFs. Interestingly, a genome-wide sequencing study in wild-type (WT) mouse NSPCs found that over 90% of SOX2-bound sites occur in H3K27ac+ nucleosomes, with an enrichment (p<0.001) of SOX2 binding in distal H3K27ac-positive sites when compared to 1,000 random sets of H3K27ac+ genomic loci³⁶. Using this SOX2 ChIP-seq data set, we determined that in vehicle- and Dex-treated NSPCs, 86% and 83% of the genomic sites, respectively, predicted to host a close association of GR and SOX (Figure 2E) are bona-fide SOX2 binding sites (Supplementary Figure 13). This suggests that the HOMER motif enrichment algorithms successfully identified SOX TF binding sites in close association with GR ChIP-seq peaks, and that the majority of these sites are likely H3K27ac+ active enhancer regions.

Studies show that higher oligomeric states drive stronger gene regulatory responses to GR activation compared to lower oligomeric states⁵². To gain a deeper mechanistic understanding of the GR-DNA interactome in NSPCs, we utilized HOMER to measure enrichment of half-site GRE motifs sites containing a single hexamer consensus sequence within GR ChIP-seq peaks, thought to be bound by GR in a lower oligomeric state^46,47. This revealed that in the absence of Dex, genomic GR binding occurs at half-site GREs, with ~36.9% of GR ChIP-seq peaks enriched for at least one half-site GRE in NSPCs (Figure 2F). Interestingly, the addition of Dex did not drastically change the overall percentage of GR ChIP-seq peaks enriched for a half-site GRE, but it did cause a ~20-fold increase in the overall number of full-site GREs bound at 4 hours (Figure 2F). In addition, a Fisher’s Exact test of independence reveals that there is a significant relationship between the size of the enriched motif (i.e. half-site vs. full-site) and treatment group (***p<0.001).

3.3. Dex Treatment Changes Chromatin Accessibility in Cortical Embryonic NSPCs

Since 95 ATAC-seq peaks (p<0.05) were differentially affected by vehicle- vs. Dex-treatment, we characterized the directionality of each structural alteration to further understand the role of GR binding in chromatin landscape remodeling. 85 ATAC-seq peaks were identified that are more accessible after exposure to Dex, whereas 10 are rendered less accessible by Dex (p<0.05). These data suggest that within the Dex-responsive chromatin of NSPCs, GR binding results primarily in increases, but not decreases, in chromatin accessibility (Figure 3A). The HOMER motif discovery algorithm reveals ~84% (n=71) and ~46% (n=39) of differential peaks with Dex-increased accessibility were enriched for at least one GRE, or a GRE and a SOX TF binding motif, respectively, relative to random background sequences (p<1e-38). In contrast, differential peaks with decreased accessibility lacked both GR binding and GRE enrichment, but 5 are enriched for a SOX TF binding motif (Figure 3B) (p<0.05). Thus, Dex-dependent increases in chromatin accessibility may be facilitated by genomic GR binding at GREs with a close association to SOX TFs, whereas decreases in chromatin accessibility occur at sites not bound by GR and, perhaps, are SOX TF-affiliated genomic regions (Figure 3C).

Figure 3. Differential Chromatin Accessibility in Response to Dex in Embryonic Cortical NSPCs.

A) Points representing an ATAC-seq differential peaks between vehicle- vs. Dex-treated samples (n=95). Fold change of normalized average ATAC-seq peak read intensity (Dex-Veh) (x-axis); p-value (p<0.05) (y-axis).

B) HOMER motif analyses predict enrichment of GRE and/or SOX TF motifs (x-axis) under (i) all differential ATAC-seq peaks (black), differential ATAC-seq peaks induced by Dex (blue), or differential ATAC-seq peaks attenuated by Dex (pink). HOMER motif predictions were called using a (p<0.05) cut-off for significance.

C) A model of TSS-proximal TF motifs with facilitate the dynamic chromatin response to Dex in embryonic cortical NSPCs. Each vertical bar represents a cluster of ATAC-seq reads, with a set of bars representing a single ATAC-seq peak. The vertical height of each bar represents intensity of ATAC-seq reads at a single genomic location.

D) Peak ID of differential ATAC-seq peaks (p<0.05), which occur within the −5kb/+3kb region of a Dex-regulated gene, or at a Dex-regulated gene TSS. Asterisks indicate ATAC-seq peaks with a Dex-induced decrease in chromatin accessibility. RNA-seq data previously published by Frahm et al., 2018¹⁰. RNA-seq log Veh/Dex fold change.

To determine if Dex preferentially alters the chromatin structure of genomic regions with a particular functional role in gene regulation, we grouped each of the 95 differential ATAC-seq peaks according to their genomic location. Eleven of them were located intergenically within −5kb/+3kb of the TSS for a Dex-regulated gene (log FC>0.8, p<0.05; n=300), and the majority (n=10) had increased accessibility in response to Dex. An additional five differential regions were mapped directly to a TSS. At three of these TSS locations, the TSS belongs to a GR-occupied, Dex-regulated gene (i.e., Fam107a, Hsd17b4, Phyhd1) (Figure 3D). These data suggest that although most ATAC-seq peaks occur far away from a gene body, Dex can also regulate the glucocorticoid transcriptome by altering the chromatin landscape surrounding TSS-proximal regions.

3.4. GR Binding at Differentially Accessibly Chromatin in Embryonic Cortical NSPCs

Approximately 95% (n=90) of the Dex-responsive dynamic chromatin occurred in intronic or intergenic genomic regions (Figure 4A), with the majority located outside of the −5kb/+3kb TSS-proximal range (n=79). To test for a potential association between a dynamic chromatin state and functional noncoding genomic variants specific to GCs, we quantified the number of differential ATAC-seq peaks which overlapped with previously described H3K27ac+ ChIP-seq peaks³¹. Sixty-four (~67%) of differential ATAC-seq peaks overlapped with a H3K27ac+ ChIP-seq peak, indicating that most Dex-induced changes in the chromatin landscape occur within active enhancer regions. These changes in chromatin accessibility strongly correlated with direct genomic GR binding, evidenced by the enrichment of 51 H3K27ac+ differential ATAC-seq peaks for the GRE motif, and an overlap with a GR ChIP-seq peak. However, some chromatin dynamics occurred independent of genomic GR binding because 13 H3K27ac+ differential ATAC-seq peaks lacked GR ChIP-seq peaks (Figure 4B). Thus, Dex-activated GR alters the chromatin landscape of enhancer regions predominantly via direct DNA binding but can exert some effects on chromatin structure via other TFs secondary to or independent of direct association with DNA.

Figure 4. Hormonal Regulation of Chromatin Accessibility and Glucocorticoid Receptor Binding in Genomic Enhancer Regions.

A) Genome annotation of differential ATAC-seq peaks in vehicle- and Dex-treated NSPCs (n=95). TSS; Transcriptional start site.

B) Percentage of differential ATAC-seq peaks (x-axis) which overlap with a ChIP-seq peak for H3K27ac and/or GR. Genomic localization of H3K27ac was delineated in cortical NSPCs (E14.5) from an independent study (Gene Expression Omnibus accession no. 104686).

3.5. Validation of GR-SOX2 Proximity in NSPCs

To validate our bioinformatics-informed prediction of a close association between GR and a SOX factor, we performed a PLA to identify proximity (<40nm) between GR and SOX2 in vehicle- or Dex-treated NSPCs (Dex; 100nM) (Figure 5A). SOX2 was chosen because it and GR share many target genes in NSPCs, with over 145 genes significantly altered by both SOX2 ablation or Dex exposure in vitro^10,36. The overall average PLA probe intensities indicating instances of GR-SOX2 proximity (<40nm) in vehicle-treated NSPCs and Dex-treated NSPCs were similar (overall average intensity = 123.5 and 124.4, respectively), while negative control groups for antibody-specific PLA-probe activity had significantly reduced amounts of PLA probe photoactivation (Control^GR and Control^SOX2; overall average intensity = 56.9 and 60.8, respectively) (p<0.001) (Figure 5B). Because PLA probes were detected above baseline intensity values (>100) in 62.8% and 40.4% of vehicle or Dex-treated NSPCs (i.e., probe-positive cells), respectively (Figure 5C), we determined whether the abundance of PLA probes was changed by Dex in this probe-positive cell population. While overall average PLA probe intensity varied more between the vehicle and Dex-treated probe-positive cell populations (165.4 and 202.5, respectively) compared to the entire cell population (123.5 and 124.4, respectively), it was not significantly changed by Dex in the probe-positive cell population (Figure 5D). Together, these biochemical data support our bioinformatic-informed prediction of GR and SOX2 interacting within NSPC nuclei to coordinate genomic responses to GR activation.

Figure 5. GR-SOX2 Proximity in NSPCs.

A. Detection of GR and SOX2 proximity in-vitro, using adherent, proliferating NSPCs treated with either vehicle or Dex for 4 hours. Blue; Dapi staining of nuclei. Red; PLA probes indicating GR-SOX2 proximity (<40nm). Negative control groups for antibody-specific PLA-probe activity (Control^GR and Control^SOX2) had minimal fluorescence of PLA probes. Images shown at 60X magnification contain a 10μm scale bar. Pseudo-color increased post-processing for visual enhancement of publication images. Asterisk indicates the cell shown on the magnified inset, upper right corner.

B. Overall average PLA probe intensity per cell per image per embryo in vehicle- (n=1,178 cells, 24 images) or Dex-treated (n=1,256 cells, 24 images) NSPCs (n=3 biological replicates), with all cells per image included. Control groups; Control^GR and Control^SOX2. One-way ANOVA test reveals a significant difference among group means (p<0.001). The single asterisk and bar indicate a significant difference between two groups (p<0.05) (F_3,8 = degrees of freedom for the numerator (DFn=3) or denominator (DFd=8) of the F ratio (F=11.26).

C. Average percentage of vehicle-treated (n=726) or Dex-treated (n=529) NSPCs with PLA probes (probe intensity per cell >100) (n=3). Control groups; Control^GR and Control^SOX2. One-way ANOVA test reveals a significant difference among group means (p<0.01). The single or double asterisks and bar indicate a significant difference between two groups (p<0.05 or p<0.01) (F_3,8 = degrees of freedom for the numerator (DFn=3) or denominator (DFd=8) of the F ratio (F=14.79).

D. Overall average PLA probe intensity per cell per image per embryo, only in probe-positive (probe intensity per cell >100, Fig. 5C) NPSCs treated with vehicle (n=726 cells) or Dex (n=529 cells) for 4 hours (n=3). Control groups; Control^GR and Control^SOX2. Unpaired two-tailed t-test does not reveal significance (p>0.05).

3.6. Microarray Gene Expression Analyses Reveals SOX2-dependent Effects of Dex on NSPC Gene Expression

To investigate the functional role of SOX2 in determining transcriptional output following GR activation by Dex, we performed unbiased genome-wide measurements of gene expression by microarray in vehicle or Dex-treated (4h) WT NSPCs derived from the developing mouse telencephalon at postnatal day (P0)(n=6), or NSPCs (P0) derived from the same region of conditionally (at E11.5) SOX2-ablated mice (SOX2 knock out; KO)(n=7). SOX2 ablation did not change GR (encoded by NR3C1) expression (Supplementary Figures 14–15). Dex alters the expression of 429 genes (p<0.05) in WT NSPCs and 901 genes in SOX2 KO NSPCs at 4h, with some genes classified as robustly Dex-upregulated (fold change; FC ≥ 1.5) or robustly Dex-downregulated (FC ≤ 1.5) (Figure 6A; Supplementary Tables 13–18). Within these groups of robustly Dex-regulated genes, a subset of gene induction occurs in both WT and SOX2 KO NSPCs (i.e., Hif3a is Dex-upregulated in both WT and SOX2 KO NSPCs, suggesting that a subset of GR action occurs independently of SOX2 (Figure 6B; ‘Shared’). In contrast, another subset of robustly Dex-regulated genes occurs uniquely in WT NSPCS (n=59) compared to SOX2 NSPCs (i.e. Cspg4 is Dex-upregulated in WT NSPCs but unaffected in SOX2 KO NSPCS (Figure 6B; ‘WT Unique’). This differential gene response in WT vs SOX2 KO NSPCs suggests that the glucocorticoid response of a subset of GR target genes is SOX2-dependent, because SOX2 is required for the Dex-induced transcriptional outcome. Interestingly, ‘Shared’ or SOX2-independent genes which had similar responses to Dex in both WT and SOX2 KO NSPCS were more likely to have a GR binding site within +/− 10kb of the TSS, while SOX2-dependent genes (i.e. WT-Unique) were less likely occur have a GR binding site +/− 10kb relative to their promoters (Supplementary Table 19). These data suggest SOX2 acts at distal enhancers to regulate a subset of GR transcriptional output in NSPCs. The final subset of Dex-responsive genes were measured only in SOX2 KO NSPCs, but showed no significant Dex response in WT NSPCs. These SOX2 KO unique transcriptional outputs are likely due to major alterations in chromatin connectivity, previously shown to occur following SOX2 ablation (Figure 6B; ‘KO Unique’)³⁶. Lastly, a comparison of gene expression patterns in WT vs SOX2 KO NSPCs highlights alterations in the Dex-responsive trending gene responses, determined by fold change of average expression, involved in canonical glucocorticoid signaling pathways (Figure 6C–D), as well as canonical FGF signaling pathways (Supplementary Figure 16), both critical regulators of NSPC pluripotency and neuronal maturation during neurodevelopment^7,10,53.

Figure 6. Gene Expression Profiling of the Dex Transcriptome in WT vs SOX2 KO NSPCs.

6A) Quantification of genes which were robustly upregulated (p<0.05; FC ≥ 1.5), robustly downregulated (p<0.05; FC ≤ −1.5), or moderately regulated (p<0.05; 1.5 < FC > −1.5) following 4h Dex treatment in WT NSPCs (P0; n=7) or SOX2 KO NSPCs (P0; n=7). FC; fold change of average expression value. Total n of significantly regulated genes (p<0.05) per group is displayed at the bottom of each chart.

6B) Quantification of genes which were Dex-upregulated (p<0.05)(left) or Dex-downregulated (p<0.05)(right) in (i) WT NSPCs only when compared to SOX2 KO NSPCs (WT Unique), (ii) both WT and SOX2 KO NSPCs (Shared), or (iii) SOX2 KO NSPCs only when compared to WT NSPCs (KO Unique).

6C-D) Genes involved in canonical glucocorticoid signaling pathways which are upregulated (Red; FC > 0) or downregulated (Green; FC < 0) by Dex at 4h in WT NSPCs (6C) or (6D) SOX2 KO NSPCs.

4. DISCUSSION

Defining the structure and dynamics of the chromatin landscape of embryonic-derived telencephalon-derived NSPCs will further an understanding of how functional coding and noncoding genomic sequences shape early neuronal development in fetuses exposed to sGCs in utero^54,55. Our study revealed that most genomic regions that are accessible to TFs occur in H3K27ac+ noncoding genomic regions, a property that highlights distal enhancer regions as critical influencers of transcriptional output during neurodevelopment^11,56–58. We explored chromatin accessibility as a molecular determinant of genomic GR action in NSPCs and identified 3162 genomic GR binding sites that facilitate gene expression changes and ultimately may contribute to altered cell fate decisions during critical periods of cortical expansion in-utero⁷. In addition, we uncovered several established and unique features of the GR cistrome in NSPCs. Genomic GR binding preferentially occurred at distal H3K27ac+ sites, which are associated with transcriptionally ‘active’ enhancers, but not ‘poised’ enhancers that have potential to be active in transcription, leading us to conclude that GR binds preferentially to active enhancer regions in NSPCs^39,40. This genomic binding pattern is established by an increased presence of GREs at distal non-promoter regions, and cell type-enriched TFs, which guide GR to its genomic targets^59,60. These data corroborate the notion that the GC transcriptome is controlled not only by promoters and cis-regulatory elements, but also through GR-responsive distal regulatory elements. Although the precise mechanisms underlying remote regulation by GR have not been fully elucidated in primary cell cultures, they have been studied in established cell lines^45,61,62. Sequence motifs that serve as recognition sites for the SOX family of TFs were significantly associated with GR binding sites in our NSPC cultures. This is a novel finding that we believe is of potential relevance to NSPC biology, since SOX2 is critically involved in the maintenance of enhancer-promoter interactions in NSPCs in vitro³⁶. SOX2 also shares many target genes with GR in NSPCs^10,36. Thus, SOX2 may contribute to long range genomic interactions that enable or facilitate GR action at regulated promoters^36,50,63. While these data have not been directly studied using NSPCs in-vivo, the GR/SOX2 co-expressing neural cell types in the developing mouse brain include radial glial cells with potential to self-replicate or generate neural and/or glial progeny (primarily localized to the ventricular zone), and neural progenitor populations with neurogenic potential (primarily localized to the subventricular zone and the hippocampus dentate gyrus)^63–66.

The impact of GR and SOX interactions on neurodevelopment could extend beyond the potential role of SOX2 as a cofactor or pioneering factor for GR. For example, neuronal population size in the developing cortex is increased by antenatal Dex in vivo (E17.5), accompanied by decreases in SOX2 expression as NSPCs progress towards a committed lineage^7,67. This developmental stage-specific antagonism of SOX2 may have long-term consequences in NSPCs, because adult rats exposed to Dex during the prenatal period display upregulated GR expression, attenuated SOX2 expression, and disordered NSPC function in the hippocampus, accompanied by increased depression susceptibility⁶⁸. Considering that SOX2 deficiency or dysregulation has been associated with neurodevelopmental disorders and genetic disease, these models underscore the importance of understanding how fluctuating GC levels collide with spatiotemporally regulated SOX TF expression to direct NSPC function in embryos and adults^69,70.

A minority subset of Dex-induced GR binding occurred in inaccessible regions of chromatin, pointing at GR’s ability to function as a pioneer factor or nucleosome remodeling protein at a limited number of sites^21,71. Nonetheless, because GR primarily binds to accessible chromatin, many have aimed to understand what additional factors guide GR to specific genomic regions. Possible explanations include the addition of epigenetic modifications of histone H3, which yields a more permissive environment for GR to bind, or the presence of cell type-enriched TFs that define basal chromatin state and facilitate GR occupancy or activity^18,72. These additional factors may impact unliganded vs. ligand-bound GR differently given our observed predominance of GR binding to GRE half-sites in vehicle-treated cells relative to GR enrichment at palindromic GRE sequences in NSPCs treated with Dex. Furthermore, these data may suggest that culture conditions required to maintain NSPCs in our system, in the absence of added GCs, do not promote GR dimerization and engagement of canonical GREs, but rather, allow a low level of monomeric GR to occupy GRE half-sites. This could be in part due to supplementation of the culture media with growth factors (i.e. EGF and FGF-1) that impact GR-based mechanisms via altered phosphorylation states, and regulate GR co-factors (i.e. AP-1, SGK-1) in some cell types^73–76. Lastly, aside from GR binding frequency or differences in subcellular signaling pathways upstream of GR, the discrepancy in transcriptional response that results from basal vs hormone-induced GR binding in NSPCs may be explained by differences in the nucleotide sequences flanking the core GRE that modulate activity downstream of GR binding⁷⁷. Collectively, these data indicate GR may play unique roles during distinct stages of neurodevelopment depending upon the levels of endogenous fetal GCs, which rise in late gestation, or in response to therapeutic antenatal sGCs.

Most Dex-inducible accessible sites occurred in noncoding regions far removed from promoters, preventing us from examining how Dex-induced chromatin remodeling directly impacts transcription. Nonetheless, 17 of the differential peaks which are located proximally to the TSS of Dex-regulated genes are likely to be functionally relevant, because these proximal regions often host pioneer factors, co-activators, and sequence-specific TFs that recruit clusters or condensates of RNA Pol II machinery necessary for transcriptional activation⁷⁸. We also infer functional relevance by examining the identity and role of the genes near Dex-inducible chromatin remodeling. For example, Dex treatment increased chromatin accessibility near 3/10 of the most highly Dex-regulated genes (Fam107a, Hif3a, Ptk2b)²⁶,but decreased accessibility near genes which control lineage specification of NSPCs in the developing telencephalon (Sox6, Rgs20)^79–82.

This characterization of the chromatin landscape of embryonic NSPCs helps integrate chromatin remodeling as a potential risk factor for neurodevelopmental and psychiatric disorders^83–85. Furthermore, our unbiased assessment of genome-wide GR action in embryonic NSPCs may inform future studies that aim to identify novel gene networks that influence the developmental trajectory of newborn neurons. Our understanding of GR action in fetal mouse (E14.5) brain-derived NSPCs is limited by a single acute timepoint resembling the earliest ages that human fetuses are exposed to antenatal glucocorticoids with respect to robust gliogenesis, ongoing neurogenesis, and a period of limited endogenous glucocorticoid exposure. In addition, a single Dex exposure in mice in-vivo (E14.5) is relevant to humans as causes similar brain phenotypes, such as reduced cortical size and cortical surface area, observed in newborn infants exposed to sGCs^7,86,87. However, this interpretation would be better informed by performing these experiments at temporal and repeated intervals, in both male and female-derived NSPCS, to fully understand how sex-specific dynasticity in GR binding and chromatin remodeling directs sexually dimorphic gene expression patterns¹⁸. Ultimately, these data suggest that GC-induced changes in gene expression during development may contribute to a cascade of biological changes associated with risk for various neuropsychiatric conditions in adulthood, though direct links between antenatal GCs and long-term outcomes are still unclear. For example, retrospective population studies associate antenatal GC therapy with higher rates of cerebral palsy in infants⁸⁸ and neurosensory abnormalities and behavioral disorders in children^86,89,90, whereas behavioral alterations and cognitive deficits are observed in adults⁹¹. Strikingly, dysregulation of the hypothalamic-pituitary-adrenal axis is a primary long-term consequence of early sGC exposure or conditional Sox2 ablation, and is also a core pathophysiology of stress-related mental disorders such as depression, anxiety, anorexia nervosa, post-traumatic stress disorder, and schizophrenia^92–98. Similar health outcomes are observed in the offspring of prenatally stressed mothers, independent of postnatal effects, identifying the prenatal period as highly sensitive to perturbations in GC signaling⁹⁹. Lastly, SOX TFs were the only candidate co-factors that guide a subset of GR action, providing a potential mechanism through which GR changes the developmental trajectory of embryonic cortical NSPCs. Our microarray gene expression data supports this possibility, as it demonstrates a subset of GC-regulated genes with differential transcriptional outcomes dependent upon SOX2 presence. Furthermore, the fact that the number of Dex-regulated genes in NSPCs is dramatically increased upon SOX2 ablation suggests that the presence of SOX2 is required to restrict GR access to a subset of developmentally relevant genes perhaps through the regulation of long-range chromatin looping by SOX2. Future studies may unveil associations between a SOX2-dependent GC transcriptome and NSPC cell fate outcomes in the developing brain. Our genomic data may advise human longitudinal follow-up studies to monitor the development of psychiatric diseases with a SOX2-related etiology such as depression and anxiety^68,100, intellectual disability^101,102, epilepsy¹⁰³, schizophrenia and bipolar disorder¹⁰⁴, or motor deficits in children and adults exposed to sGCs antenatally105. In summary, excess GCs have the potential to alter GC transcriptional responses by acting upon a developmentally regulated chromatin landscape to influence the fate of NSPC populations and, ultimately, may contribute to adverse neurologic outcomes later in life⁹⁴.

• Most accessible chromatin regions are H3K27ac+ transcriptional enhancers.

• Dex does not drastically change the chromatin landscape of NPSCs.

• Dex increases GR binding, mostly within accessible chromatin, by 236% in NSPCs.

• GR and SOX2 interact to coordinate the genomic response to sGCs in-vitro.

• SOX2 may be involved in the etiology of sGC-associated neurodevelopmental deficits.