Psychiatric risk gene transcription factor 4 preferentially regulates cortical interneuron neurogenesis during early brain development

Yuanyuan Wang, Liya Liu, Mingyan Lin

1State Key Laboratory of Reproductive Medicine, Nanjing Medical University, Nanjing, Jiangsu 211166, China;

2Department of Neurobiology, School of Basic Medical Sciences, Nanjing Medical University, Nanjing, Jiangsu 211166, China.

Abstract

Keywords: transcription factor 4, hMGEOs, cortical interneuron, ChIP-seq, neurogenesis, schizophrenia

Introduction

Transcription factor 4 (TCF4) is a member of the basic helix-loop-helix (bHLH) transcription factors(TFs) family that recognizes an E-box sequence(CANNTG) as homo- or hetero-dimers with tissuespecific bHLH TFs[1]. TCF4 is abundantly expressed in several tissues including the brain, and is thought to be involved in neural development processes such as cell proliferation, differentiation, and synaptic formation[2]. In addition, TCF4 deficiency, observed in animal model studies, is causal in various cognitive defects[3]. Several whole genome association studies have also revealed that TCF4 is one of the most reproducible susceptibility genes in neurodevelopmental disorders, including schizophrenia (SCZ), Pitt-Hopkins syndrome, autism, and bipolar disorder[4]. This suggests there is a link between TCF4 perturbation and neurodevelopmental disorders which must be investigated.

Systematic analyses of expression pattern of TCF4 in the developing and adult mouse brain showed that TCF4 is abundant in germinal regions at early developmental stages i.e., embryonic day 11.5(E11.5)–E13.5[2]. TCF4 expression is also associated with cortical interneuron production timing in medial ganglionic eminence (MGE)[2,5]. At E13.5 and after,TCF4 is abundant in the neocortex, corpus callosum,anterior commissure, and hippocampus, and is crucial for brain formation during murine embryonic development[2,6]. In addition, ablation of TCF4 during mouse development causes neocortical disorganization, which recapitulates structural brain abnormalities present in Pitt-Hopkins syndrome patients. This evidence suggests that TCF4 may also have a function in normal brain development and neuronal differentiation during embryogenesis[6].

Human neuroblastoma SH-SY5Y cell line, one of the most widely adopted cellular system to model mouse neurodevelopment, has the characteristics of neural stem cells and can differentiate from neuroblast-like state to mature neurons[7]. Two recent studies investigated the regulatory role of TCF4 in SH-SY5Y and found that TCF4 target genes were enriched in functional clusters such as ion transportation, signal transduction, and nervous system development[8–9]. However, findings from SHSY5Y studies are unlikely to reflect the relevant cell type-specific role of TCF-4 in neurodevelopment.Fortunately, the revolutionary technology of singlecell transcriptomics has successfully described the landscape of cell types throughout early human brain development. This foundational research helps to identify the most relevant cell types of TCF4[10–11].

To understand the link between TCF4 and psychiatric disorders (PSD), we first conducted an integrative transcriptomics analysis on public bulk RNA-Seq and single-cell RNA-seq (scRNA-seq) data of the fetal brain to determine the spatiotemporal expression pattern of TCF4. We then performed chromatin immunoprecipitation-sequencing (ChIPseq) on human medial ganglionic eminence-like organoids (hMGEOs) to depict the gene regulatory network of TCF4. To further explore the regulatory mechanism of TCF4 and its relevance to PSD, we conducted an advanced multi-omic analysis integrating landscapes of histone modification, coexpression patterns, and profiles of genetic risks of PSD. Our study suggested interaction with FOS/JUN might determine the functional specificity of TCF4 in hMGEOs, highlighting a potential link between dysregulation of interneuron development induced by TCF4 perturbation and PSD.

Materials and methods

Cell culture

The human induced pluripotent stem cell lines(NC3-1, passage 13; ihtc-03, passage 16) were presented by Dr. Yan Liu's laboratory at Nanjing Medical University, China. All stem cell lines were maintained on vitronectin-coated plates (Life Technologies, USA) with Essential eight medium(Life Technologies), and changed daily at 37 °C in 5%CO2. Cells were passaged every 5 days through ethylenediaminetetraacetic acid digestion (Lonza,USA).

The human neuroblastoma SH-SY5Y cell line at passage 4 was a gift from Dr. Jun Gao's laboratory(Nanjing Medical University). SH-SY5Y cells were cultured in DMEM/F12 medium (Gibco, USA)supplemented with 10% (v/v) fetal calf serum (Gibco)at 37 °C in 5% CO2.

Development of human medial ganglionic eminence-like organoids

HMGEOs were generated following the protocol developed by Liu et al[12], who established a system of directed differentiation for forebrain γ-aminobutyric acid (GABA) interneurons using human induced pluripotent stem cells. After four weeks of differentiation using an established system, more than 90% of the cells become NKX2-1 and FOXG1 expressing MGE progenitors[12].

Stem cells were detached by dispase (Life Technologies) to form embryoid bodies (EB) and then cultured in the neural induction medium (NIM),containing 490 mL DMEM/F12 medium, 5 mL of minimum essential medium non-essential amino acids(MEM-NEAA) (Gibco), 5 mL of N2 supplement(Gibco) for 7 days. Half the NIM medium were changed out every day from day 1 to day 6. After floating for 7 days, EBs were attached on vitronectincoated surfaces. Rosette structures could be formed during the period from day 10 to day 15. Half the NIM medium were changed out every other day from day 10 to day 15. On day 16, neuroepithelialcontaining rosette clones were detached and neuroepithelial cells gradually formed neurospheres.Neurospheres were continuously floated in NIM, and changed half the NIM medium every other day. For ventral differentiation, 500 nmol/L smoothened agonist (SAG, Millipore, Germany) was added from day 10 to day 40. From day 0 to day 10, BMP inhibitor DMH1 (Tocris Bioscience, UK) and TGF-β inhibitor SB431542 (StemGent, USA, Cat. No. 04-0010) were added. Properly developed neurospheres were collected on day 26 –40 for follow-up ChIP experiments. The protocol for generating dorsal cortical-like organoids was largely the same except that SAG was not added.

Immunostaining

First, hMGEOs were fixed with 4%paraformaldehyde (Sangon Biotech, China) for 2 hours in an Eppendorf tube before washing them with phosphate buffered saline (PBS, Beyotime, China) 3 times. Then, organoids were soaked with 20% sucrose(Sangon Biotech) in PBS overnight at 4 °C and then with 30% sucrose in PBS after organoids sinking to the bottom of Eppendorf tube. Organoids were embedded at an optimal cutting temperature compound and cryosectioned at 10 μmol/L prior to immunostaining.

For immunohistochemistry, we blocked and permeabilized each section in 1% Triton (Biolink,China) and 5% donkey serum (Millipore) in PBS before incubating at 4 °C overnight in primary antibody and then in secondary antibody diluted in 5%donkey serum for 1 hour at 20 °C. After performing three 10 minutes washes in PBS, coverslips were mounted for fluorescent imaging using an Eclipse 80i Fluorescence Microscope.

Western blotting

TCF4 expression in hMGEOs and human dorsal cortical-like organoids (hCOs) was assessed by Western blotting. Organoids were lysed in RIPA buffer (Beyotime) containing protease (Millipore) and a protease inhibitor cocktail (Millipore), then centrifuged at 12 000 g for 5 minutes to collect the protein supernatant. Proteins were quantified by BCA Protein Quantitation Kit (Beyotime). After the quantification, proteins were loaded onto gels(Beyotime) separated by SDS-PAGE with 100 V electrophoresis. Then, proteins were transferred onto polyvinylidene fluoride membranes (PVDF)membranes (Millipore) at 300 mA for 2 hours and blocked with 5% (w/v) nonfat dried milk for 2 hours at room temperature. Primary anti-TCF4 (Santa Cruz Biotechnology, USA, Lot No. sc-393407X; dilution 1:1000) antibody was incubated overnight at 4 °C.Anti-GAPDH (Bioworld, China; dilution 1:1000) was used as an internal reference. Then, the membranes were washed with 8× PBST (Beyotime) solution 5 times for 10 minutes and incubated in HRPconjugated IgG secondary antibody (Biosharp, China;1:5000 dilution) on a shaker for 2 hours at room temperature. After the incubation, the secondary antibody was decanted, and the membranes were again washed 5 times with 1× PBST for 10 minutes.The enhanced chemiluminescence system (Tanon,China) was used detection of the protein bands.

Single-cell RNA-seq re-analysis of human medial ganglionic eminence-like organoids and corticallike organoids

Single-cell expression matrices of human medial ganglionic eminence-like organoids (Day 30 and Day 72, H1 human ES cells, and human iPSC 1090) and cortical-like organoids published by Xiang et al were re-analyzed. Expression matrix was processed with Seurat (version 3.1.5)[13]. The criteria to select cells for subsequent analysis were as follows: unique molecular identifiers per cell ＞500, detected genes ＞300, and a mitochondrial transcript proportion ＜0.3.

SCTransform normalization[13]was applied to each Seurat object to control confounding sources of variations such as sequencing depth and mitochondrial fraction. Integration was performed to correct batch effect. Visualization of transcriptomic profiles were conducted by uniform manifold approximation and projection (UMAP).

Expression matrices were summarized by the top 10 principal components. The Louvain modularity optimization algorithm was implemented to iteratively group cells into clusters. Cell clusters were annotated to known biological cell types using canonical cell marker genes.

Chromatin immunoprecipitation-sequencing analysis

ChIP assay was conducted on hMGEOs derived from ihtc-03/NC3-1 and SH-SY5Y cell lines.According to the manufacturer's instructions, libraries of ChIP DNA were prepared using the ChIP Kit(Millipore, Lot No. 17-10086). ChIP assay was performed using anti-TCF4 antibody (Santa Cruz Biotechnology, USA, Lot No. sc-393407X; dilution 1:1000) and normal mouse IgG (Millipore, Lot No.17-10086; dilution 1:1000). We followed ENCODE guidelines for anti-TCF4 antibody validation[14].

To evaluate ChIP enrichment efficiency, we performed real-time quantitative PCR (qPCR) using AceQ qPCR SYBR Green Master Mix (Vazyme,China) for TCF4 binding sites associated with the genes of interests (SYPL1 [intergenic], CHRNB4[intergenic, distal enhancer], OPRD1 [intron], and RNU5F-1 [intergenic]) in SH-SY5Y. To evaluate ChIP enrichment efficiency in hMGEOs, we performed qPCR for TCF4 binding sites associated with the genes of interests (SYT10 [intron], SEMA3E[distal intergenic], CNTNAP2 [distal intergenic], and BRINP3 [intron]). Values were normalized using the ΔΔCt method. GAPDH was used as an internal reference. Primers were manufactured by Genscript(China). Primer sequences used for qPCR were listed in Supplementary Table 1 (available online).

The sequencing of precipitated DNA of hMGEOs using Illumina Hiseq X-10 (2X150). Sequenced reads were mapped to the human genome (hg38) using BWA software (version 0.7.5a-r405)[15]. Only the uniquely mapped reads were retained for further analyses. Peak calling for TCF4 in ihtc-03/NC3-1 was carried out using PeakSeq (version v.1.1, with options"target_FDR 0.05, max_Qvalue 0.05") on the TCF4 ChIP file against the input file and further retains the peak with input reads ≥5 as the significant enrichment peak[16]. Genome-wide signal coverage tracks were computed using DeepTools (version 3.3.0,bamCoverage), and visualized in the Integrative Genome Browser (IGV, version 2.8.0)[17–18]. Genomic features to peaks were annotated by ChIPseeker(version v1.20.0)[19]. The target genes of TCF4 binding sites were annotated with the Genomic Regions Enrichment of Annotations Tool (GREAT),version 4.0.4, using default parameters[20]. De novo motifs of ChIP-seq peaks were searched by homer(version v4.11.1) using default parameters[21].Heatmaps of binding signals across multiple genomic locations were drawn by deepTools (computeMatrix command on multiple bigwigs and plotHeatmap),version 3.3.0. GO enrichment analysis and result visualization of TCF4 target genes were performed using enrichplot (version 1.6.1) with R using default parameters[22].

Upstream regulators analysis

To infer the candidate upstream regulators of the target genes of TCF4 binding sites, we performed upstream regulators analysis by iRegulon, a plugin in Cytoscape, using default parameters[23]. The corresponding normalized enrichment score (NES)and the number of regulated genes for each inferred upstream regulator were obtained from iRegulon and visualized in R software.

TCF4 co-occurring motif combinations analysis

We first extracted the coordinates of the classical motif of TCF4 hit regions within the TCF4 binding sites using MEME mast software and then extended them by 50 bp in both directions[24]. These regions were considered likely regions harboring TCF4 and potential co-factors. We conducted SIOMICS (with options "-e 0.00005 -c 0.05") analysis on these extended regions to identify co-occurring motif combinations of TCF4[25]. TCF4 binding sites in SHSY5Y were obtained from a previous report[8], and the TCF4 co-occurring motif combinations in SH-SY5Y were identified as above.

Psychiatric disorders risk gene enrichment analysis

To determine whether TCF4 targets in hMGEOs were enriched with PSD risk genes, we intersected the identified targets with the risk gene list of de novo variants of PSD from Fromer et al[26], and performed enrichment analysis using Fisher's exact one-sided test.

TCF4 and target genes expression correlation analysis

We calculated cell type-specific mean expression within each gestational week based on scRNA-Seq data from human embryonic prefrontal cortex[10].

To analyze correlations between TCF4 and target genes across gestational weeks in cortical interneuron,Pearson correlation analysis was conducted using mean values for TCF4 at nine gestational weeks in cortical interneuron and mean expression values of each target gene of all gestational weeks in cortical interneuron. Target genes with a correlation P-value less than 0.05 and a correlation coefficient greater than 0 were positively correlated with TCF4 expression. Conversely, target genes with a P-value less than 0.05 and a correlation coefficient less than 0 were negatively correlated with TCF4 expression.

Statistical analysis

ChIP-qPCR data were reported as mean ± standard deviation (SD). Statistical analysis was performed with a Student's t-test in R software. Other statistical tests performed were listed in the respective figure legends or sections of methods.

Data availability

All public data were available from Gene Expression Omnibus (GEO). The single cell expression dataset from the human embryonic brain was downloaded under the accession number GSE104276[10]and GSE103723[11]. The single cell expression dataset from human medial ganglionic eminence-like organoids (Day 30 and Day 72, H1 human ES cells, and human iPSC 1090) and corticallike organoids were downloaded under the accession number GSE98201[27]. The H3K27ac and H3K4me3 ChIP-seq dataset of the human fetal brain were downloaded under the accession number GSE63634[28]. The two TCF4 ChIP-seq dataset were downloaded under the accession number GSE96915[8]and GSE112704[9].

Results

Validation of TCF4 antibody

The anti-TCF4 antibody used met the specificity and sensitivity quality control criteria for ChIP antibody in ENCODE guidelines[14]and detected the two TCF4 isoforms in SH-SY5Y, TCF4-B (72 kDa)and TCF4-A (55 kDa) (Fig. 1A). The previous study by Forrest et al have shown that TCF4 regulates gene expression of SYPL1, CHRNB4, OPRD1, and RNU5F-1DNA in SH-SY5Y, which was confirmed by our findings that TCF4-bound DNA fragments within or nearby the SYPL1, CHRNB4, OPRD1, and RNU5F-1 were significantly enriched in our ChIP experiment(Fig. 1B).

Spatiotemporal expression pattern of TCF4

To investigate the spatiotemporal expression pattern of TCF4 in the human brain, we leveraged the transcriptional profile of the whole human brain in the BrainSpan data (http://www.brainspan.org/). We found the expression of TCF4 was largely restricted to the prenatal stages of the prefrontal cortex (Fig. 2A),suggesting that TCF4 may play a role in the prefrontal cortex during prenatal development.

To identify the most relevant cell type of TCF4 function during neurodevelopment, we re-analyzed scRNA-Seq data from human embryonic prefrontal cortex at gestational weeks (GW) 8 to 26 to quantify the expression level of TCF4 in each cell type[10]. The results showed that TCF4 was preferentially expressed in cortical interneuron during early neurodevelopment(Fig. 2B). We confirmed this finding by using another set of high-resolution scRNA-Seq data from the entire human cortex at post-conceptional weeks 22 to 23 (22 and 23 W)[11](Fig. 2C). Collectively, these results implied that TCF4 might mainly be involved in interneuron development during early neurodevelopment.

Fig. 1 Validation of anti-TCF4 and the efficiency of chromatin immunoprecipitation assay. Chromatin immunoprecipitation(ChIP) assay was performed in SH-SY5Y cells. A: Western blotting of immunoprecipitates with anti-TCF4 antibody (IP) and normal mouse IgG (IgG), supernatant after extraction of Protein A/G bead-antibody/chromatin complex (Sup), and nuclear lysates(Input). B: The efficiency of ChIP assay was evaluated by the fold enrichment of TCF4-bound DNA fragments within or nearby genes of interest (i.e., SYPL1, CHRNB4, OPRD1, and RNU5F-1) in SHSY5Y cells. The negative (IgG) sample was used as a control. Data were reported as mean±SD (n=3). Statistical analysis was performed by Student's t-test. ***P＜0.001.

Since interneurons primarily originate from the medial ganglionic eminence (MGE)[29–31], we tested if the hMGEO was suitable for studying the role of TCF4 in fetal interneurons by leveraging single RNA seq data of hMGEOs and hCOs[27]. It showed that TCF4 was dominantly expressed in interneurons marked by NKX2-1 and GAD1 in hMGEOs (Fig. 3A and C), while TCF4 could be detected in both interneuron and radial glia in hCOs (Fig. 3B and D).Western blotting revealed that TCF4 was more enriched in hMGEOs compared to hCOs (Fig. 3E). In sum, these results suggested the hMGEO was a more desirable model for studying TCF4 and further supported a preferential role of TCF4 in interneurons.

Fig. 2 TCF4 is preferentially expressed in cortical interneurons. A: The normalized gene expression of TCF4 at different developmental time points and brain regions in BrainSpan Atlas of the Developing Human Brain RNA-Seq data. The color scale bar on the bottom illustrates the relative expression level of TCF4 across all-time points and brain regions. Red denotes high expression, and blue denotes low expression. RPKM: reads per kilobase per million mapped reads. pcw: post-conceptional weeks. B: The normalized expression level of TCF4 in 6 cell types of the human embryonic prefrontal cortex at GW 8 to 26. The empty box represents an expression value of 0 or no samples. TPM: transcripts per kilobase of exon model per million mapped reads. C: The normalized expression level of TCF4 in 29 sub cell types from the entire human cortex at 22 to 23 weeks post-conception (22 W and 23 W). GW: gestational weeks; In: cortical interneurons;Ex: excitatory neurons; Npc: neural progenitor cells; Astro: astrocytes; Micro: microglia; Cajal: Cajal-Retzius cells; Opc: oligodendrocyte progenitor cells; Blood: Blood cells. In1-8: eight cortical inhibitory neurons sub-clusters; Ex1-4: four excitatory neurons sub-clusters; NSC1-2: two neural stem cell sub-clusters; Astro1-2: two astrocytes sub-clusters; Endo1-2: two endothelial cells sub-clusters; Micro1-3: three microglia sub-clusters; immune1–4: four immune cells sub-clusters.

The functional role and regulatory pattern of TCF4

Therefore, we developed the hMGEOs following the protocol described previously[12]. On day 40 of differentiation, we observed a high level of transcription factor NKX2-1 expression in hMGEOs(Fig. 4A and B). We hypothesized that the genes regulated by TCF4 in hMGEOs could provide functional insight into the role of TCF4 in neurodevelopmental disorders including schizophrenia. To this end, we carried out a ChIP-seq analysis to define the genomic targets of TCF4.Intriguingly, the TCF4-A was barely detected in hMGEOs (Fig. 3E and Fig. 4C). As expected, DNA fragments from the TCF4 binding sites associated with neurogenesis genes such as SYT10, SEM3EA,CNTNAP2, and BRINP3 were significantly enriched in the hMGEOs ChIP assay (Fig. 4D).

We identified 5916 TCF4 peaks and found that the most enriched de novo motif was in high agreement with the classical motif of the bHLH transcription factor families (Fig. 5A). To better characterize the functional role and regulatory pattern, we focused on TCF4 peaks containing the classical motif of TCF4(resulting in 2648 TCF4 binding sites) in the subsequent analysis. We first annotated binding sites and identified target genes using GREAT[20]and found that TCF4 binding sites were enriched at distal genomic regions rather than proximal ones (Fig. 5B and C). In addition, by leveraging publicly available data on active histone modification (H3K27ac and H3K4me3) in the fetal brain at 12 weeks of embryonic development, we showed that the TCF4 binding sites were significantly enriched with the enhancer marker H3K27ac (Fisher's exact P-value＜4.4×10−10, odds ratio=1.74) of the early fetal forebrain, but not with the promoter marker H3K4me3 (Fisher's exact Pvalue=0.93, odds ratio=0.74) (Fig. 5D). These results indicated that TCF4 was bound primarily at the enhancers in hMGEOs during neurodevelopment.

Fig. 3 scRNA-Seq analysis showed the expression pattern of TCF4 in hMGEOs and hCOs. A: UMAP plot of 6 identified clusters in hMGEOs after 30 and 72 days of in vitro differentiation. Each color represents one cluster. B: UMAP plot of 8 identified clusters in hCOs after 30 and 72 days of in vitro differentiation. Each color represents one cluster. C: Feature plot showing the expression patterns of TCF4 and cell type-specific markers in hMGEOs. The colors from light grey to blue represent expression values from low to high. D: Feature plot showing the expression patterns of TCF4 and cell type specific markers in hCOs. The colors from light grey to blue represent expression values from low to high. E: Western blotting showing the relative abundance of TCF4 protein between human cortical-like organoids differentiated from ihtc-03 cell line on day 33 and human medial ganglionic eminence-like organoids differentiated from ihtc-03 cell line on day 33. GAPDH was used as an internal control. CN: cortical excitatory neuron; Cyc: cyclic cells; EPC: ependymocyte; HMC: high metabolic cells; IN: interneuron; Inter: intermediate cells; RG: radial glia; Sub: Subplate. NKX2-1: medial ganglionic eminence marker;GAD1: interneuron marker; TOP2A: cyclic cells marker; GFAP: astrocytes marker. TBR1: cortical excitatory neurons marker. UMAP:uniform manifold approximation and projection. hMGEOs: human medial ganglionic eminence-like organoids. hCOs: human cortical-like organoids.

Fig. 4 TCF4 ChIP assay in hMGEOs. A: Bright-field microscopic images of ihtc-03-derived hMGEOs at different stages. Top-left: A colony of human induced pluripotent stem cells (iPSCs). Scale bar: 250 μm. Top-right: spherical embryoid bodies (EB)-like structures on day one. Scale bar: 100 μm. Middle-left: EB on day seven. Scale bar: 250 μm. Middle-right: The rosette structure developed from EB on day 12. Scale bar: 100 μm. Bottom-left: The rosette structure developed from EB on day 16. Scale bar: 100 μm. Bottom-right: Neuroepithelial spheres with rosette-containing colonies on day 25. Scale bar: 250 μm. B: Immunofluorescence of NKX2-1 positive cells in day-40 hMGEOs. NKX2-1 (red): medial ganglionic eminence marker NKX2-1 staining. HO (blue): Hoechst staining. Scale bar: 35 μm. C: Western blotting of immunoprecipitates with anti-TCF4 (IP), supernatant after extraction of Protein A/G bead-antibody/chromatin complex (Sup),and nuclear lysates (Input). D: The efficiency of ChIP assay was evaluated by the fold enrichment of TCF4-bound DNA fragments within or nearby SYT10, SEMA3E, CNTNAP2, and BRINP3 from ihtc-03-derived hMGEOs on day 33. The negative (IgG) sample was used as a control. Data were reported as mean±SD (n=3). Statistical analysis was performed by Student's t-test. **P＜0.01; ***P＜0.001. NS:Neuroepithelial spheres.

To determine whether TCF4 could activate or repress the transcription of target genes in hMGEOs,we carried out the Pearson correlation analysis between the expression level of TCF4 and 3572 target genes in cortical interneuron based on scRNA-Seq data of the prefrontal cortex during brain development[10]. It turned out that a disproportionate amount of target genes was positively correlated with the level of TCF4 expression in cortical interneuron(Fig. 5E). These findings, coupled with the observation of the coexistence of TCF4 binding with active enhancer histone modifications, indicated that TCF4 mainly activated transcription distally in hMGEOs during neurodevelopment.

Fig. 5 TCF4 primarily activates the transcription of genes associated with neurogenesis by binding to distal enhancers in hMGEOs.A: De novo motif of TCF4 predicted by homer software in hMGEOs ChIP-seq (top) and the classical motif of TCF4 from the JASPAR database (bottom). B: Bar plot showing the location of TCF4 binding sites relative to the transcription start site of the target genes in hMGEOs, estimated by GREAT software. C: Pie chart showing the genomic annotation of TCF4 binding sites in hMGEOs, estimated by the ChIPseeker R package. D: Heatmap showing the intensities of ChIP-seq signals of TCF4 in ihtc-03-derived hMGEOs, TCF4 in NC3-1-derived hMGEOs, H3K27ac in fetal brain and H3K4me3 in fetal brain around TCF4 binding sites. The signal intensity was measured in count-per-millon (CPM). The 12 000 bp flanking region around peak centers were shown, and each row represents a distinct peak. E:Density plot showing the distribution of the Pearson correlation coefficients (P-value＜0.05) of expression between TCF4 and target genes based on fetal brain prefrontal cortex scRNA-Seq data[10]. Positive (or negative) values indicate positive (or negative) correlation between TCF4 and target genes. F: Enrichment map displaying the functional clusters among the top 500 TCF4 target genes. The dot size and color represent the gene count and the p.adjust value of the enriched biological processes. G: Scatter plot showing the normalized enrichment score of the potential upstream transcription factors and the corresponding number of downstream targets, predicted by iRegulon on the top 500 TCF4 target genes. The sizes of the points were proportional to the scaled normalized enrichment score. H: Schematic showing the relative location of the top one co-occurring motif combination within TCF4 binding sites, predicted by SIOMICS.

To understand functional clusters of genes regulated by TCF4, gene ontology enrichment analyses were performed on TCF4 targets. These target genes revealed a strong enrichment for neurogenesis events,including neural precursor cell proliferation, apoptotic process, and telencephalon development. (Fig. 5F).Notably, TCF4 target genes were significantly enriched for risk genes of neurodevelopmental disorders, including SCZ, autism and intellectual disability (Table 1), suggesting TCF4 perturbation could be a major contributor to PSD by regulating many other risk genes at the early stage of neurodevelopment.

Interaction with other factors can influence the regulatory specificities of TCF4 in different cell types.To explore this possibility, first conducting iRegulon analysis to search for upstream candidate regulators of TCF4 target genes in hMGEOs. Among the target genes of TCF4 in hMGEOs, DEAF1, FOS/JUN,NANOS, TCF12, and ETS1 were the most likely upstream regulators (Fig. 5G). These candidate upstream regulators may be the downstream targets regulated by TCF4 or the factors cooperating with TCF4. To further infer the interacting factors with TCF4, SIOMICS[25]analysis was performed. Notably,transcription factor FOS/JUN involved in neuronal plasticity, neural network formation and immune response were the top TFs co-occurring with TCF4 in the hMGEOs (Fig. 5H), suggesting that FOS/JUN might cooperate with TCF4 to regulate interneuron development.

Difference in the predicted role of TCF4 between hMGEOs and SH-SY5Y

Our study shared less than half of TCF4 targetgenes with those identified in two previous ChIP-seq experiment of TCF4 in SH-SY5Y[8–9](Supplementary Fig. 1A and B, available online). Enrichment analysis showed TCF4 target genes in hMGEOs were featured by biological processes involved in neurogenesis (Fig.5F), while target genes in SH-SY5Y converged on biological events related to the capacity of exploring the environment such as response to external stimuli,neuron projection and migration (Supplementary Fig.1C and D, available online). We also noticed a difference in the annotation of TCF4 binding sites, as TCF4 bound with both active enhancers (overlapping with fetal brain H3K27ac histone modification in Forrest et al: Fisher's exact test P-value＜2.2×10−16,odds ratio=7.28; in Xia et al: Fisher's exact test Pvalue＜＜2.2×10−16, odds ratio=28.40) and promoters(overlapping with fetal brain H3K4me3 histone modification in Forrest et al: Fisher's exact test Pvalue＜2.2×10−16, odds ratio=7.05; in Xia et al: Fisher's exact test P-value＜2.2×10−16, odds ratio=11.60) in SHSY5Y while it showed preferential bindings with active enhancers in hMGEOs (Supplementary Fig.1E–H, available online). Upstream analysis further suggested a distinct set of TFs co-occurring with TCF4 in SH-SY5Y, including members of bHLH family such as TWIST2 and NEUROD2(Supplementary Fig. 1I–L, available online).

Table 1 Enrichment of TCF4 target genes for psychiatric risk genes

Discussion

Large-scale human genomic studies have led to the identification of an increasingly long list of risk regions and genes associated with PSD, including TCF4, one of a few genes robustly implicated in the genetic aetiology of these diseases[4]. This genomic landscape offers unprecedented advantages for the illumination of disease mechanisms but also presents challenges. A key concern is a highly pleiotropic biology encoded by the risk genes. Great efforts have been made to prioritize cells that are fundamental to the genesis of PSD[32], suggesting reductive targets for experimental modeling. In this study, we showed TCF4 is preferentially expressed in cortical interneurons during early neural development,indicating the perturbed biology by TCF4 genetic variants at this particular spatiotemporal point would play an assignable role in neurodevelopmental disorders. Note that the observed expression pattern of TCF4 is consistent with findings in two recent studies revealing abundant expression of TCF4 in migrated interneurons in cortical development[2,33], further corroborating the link between TCF4 and cortical interneurons.

As a first step to elucidate the function of TCF4 in such a particular context, we performed ChIP-seq experiment on hMGEOs, focusing on delineating the role of TCF4 as a TF in this study. We found that the major isoform expressing in hMGEOs is TCF4-B,which can activate transcription to a greater extent than many others as a result of possession of two transcriptional activation domains[34]. Indeed, we demonstrated that the predicted target genes with positively correlated expression with TCF4 outnumbered those with the opposite correlation. Not surprisingly, target genes formed functional clusters overrepresented by ontology terms related to interneuron neurogenesis. Intriguingly, target genes exhibited significant overlap with genes previously implicated in schizophrenia, autism and intellectual disability including a number of critical players in the maintenance of E/I balance[35]such as ERBB4,CNTNAP2, NRG1, TSC1, UBE3A, CNTNAP4, and DISC1, supporting a convergent role for TCF4 in modulating the known component of the disease risk mechanism[36]. These results together suggested TCF4 should play a positive role in promoting generation of cortical interneuron, and TCF4 perturbation could contribute to the development of PSD by E/I imbalance due to defective interneuron neurogenesis.

Perhaps one of the most interesting findings emerging from our study is the interaction between TCF4 and non-bHLH proteins. It is widely accepted that TCF4 exerts its regulatory roles through homodimerization or heterodimerization with the classical bHLH proteins, such as the neurogenin or NeuroD family[37]. In neural progenitor cells (NPCs),TCF4 was shown to interact with bHLH TFs, such as NEUROG1/2, ASCL1, and OLIG1/2, to regulate NPCs maintenance and/or differentiation into neurons,oligodendrocytes, and astrocytes during brain development[4]. Forrest et al and Xia et al's studies in SH-SY5Y also supported the dimerization of TCF4 with other bHLH proteins[8–9]. While it is long recognized that other bHLH family members have the potential to cooperate with non-bHLH proteins[38],evidence to support such kind of interaction for TCF4 during neurodevelopment is just beginning to emerge.For example, a recent study provided transcriptomic evidence that TCF4 interacts with non-bHLH proteins like SOX11 in mouse Satb2+ intercortical projection neurons[39]. The predicted interaction between TCF4 with non-bHLH proteins FOS/JUN is not entirely unexpected, as other bHLH proteins, such as MYOD,have been shown to cooperate with FOS/JUN family proteins by binding to regulatory elements adjacent to AP-1 sites[38]. This interaction is further supported by a prior study showing that expression levels of FOS and TCF4 were highly coordinated in human ventral forebrain spheroids-derived GABAergic interneurons[33]. Nevertheless, further experiments are required to verify the interaction between TCF4 and FOS in interneurons.

The demonstrated differences between our ChIPseq data in hMGEOs and two other ChIP-seq data in SH-SY5Y for TCF4 in terms of genomic binding sites, functional enrichment and interacting partners should be taken with caution. These may reflect the cell type-specific roles of TCF4, but could also be attributed to technical issues we were unable to address in the present study, such as lacks of matched epigenomic profiles from hMGEOs for multi-omic analysis, or more likely antibody-specific bias. The antibody used in Forrest et al's and our study can recognize both TCF4-A and TCF4-B in SH-SY5Y,while that used in Xia et al's study can only detect TCF4-B in SH-SY5Y. Thus, the inferred functional specificity of TCF4 in different contexts based on the difference between our ChIP-seq data in hMGEOs and Forrest et al's ChIP-seq data in SH-SY5Y is supposed to suffer less from antibody-specific bias but more from confounding effects of TCF4-A, while the inference based on the difference between our ChIPseq data in hMGEOs and Xia et al's ChIP-seq data in SH-SY5Y would be the opposite.

In conclusion, the identification of genome-wide binding sites for TCF4 in hMGEOs sheds a novel insight into the functional role of TCF4 in cortical development. More importantly, our study provided compelling evidence to support the biological rationale linking TCF4 to the developing cortical interneuron and PSD, and represented several interesting hypotheses to be tested in future neurobiological studies. We hope a better characterization of the connections between TCF4 genetic variants and its pleiotropic biology would eventually turn this gene into potentially druggable targets in treating a range of neurodevelopment disorders.

Acknowledgments

This study was funded by the Natural Science Foundation of the Jiangsu Higher Education Institutions of China (Grant No. 17KJB180009) to M.L., the Natural Science Foundation of Jiangsu Province (Grant No. BK20171062) to M.L., and the National Natural Science Foundation of China (Grant No. 81701320) to M.L. We appreciate comments from the editor and anonymous reviewers.