Analysis of chromatin accessibility uncovers TEAD1 as a regulator of migration in human glioblastoma

Abstract

The intrinsic drivers of migration in glioblastoma (GBM) are poorly understood. To better capture the native molecular imprint of GBM and its developmental context, here we isolate human stem cell populations from GBM (GSC) and germinal matrix tissues and map their chromatin accessibility via ATAC-seq. We uncover two distinct regulatory GSC signatures, a developmentally shared/proliferative and a tumor-specific/migratory one in which TEAD1/4 motifs are uniquely overrepresented. Using ChIP-PCR, we validate TEAD1 trans occupancy at accessibility sites within AQP4, EGFR, and CDH4. To further characterize TEAD's functional role in GBM, we knockout TEAD1 or TEAD4 in patient-derived GBM lines using CRISPR-Cas9. TEAD1 ablation robustly diminishes migration, both in vitro and in vivo, and alters migratory and EMT transcriptome signatures with consistent downregulation of its target AQP4. TEAD1 overexpression restores AQP4 expression, and both TEAD1 and AQP4 overexpression rescue migratory deficits in TEAD1-knockout cells, implicating a direct regulatory role for TEAD1-AQP4 in GBM migration.

Fig. 1

Chromatin accessibility analysis in acutely isolated human GSCs and NSPCs. a FACS plot showing isolation strategy of GBM and neural germinal matrix (GM) populations used for ATAC-seq preparation. See also Supplementary Fig. 1. b Differential ATAC-seq peak analysis between the union of E+GSC and E+NSPCs (n = 7) and E–GBM (n = 3) defines 5020 significantly enriched, developmentally shared peaks in E+GSCs. c Functional enrichment analysis of developmentally shared E+GSC peaks using GREAT. Illustrated are top significantly enriched gene ontology (GO) terms and examples of the corresponding annotated genes, which relate to stem-cell maintenance and proliferation. d Differential ATAC-seq peak analysis between E+GSC (n = 4) and the union of E–GBM and NSPCs (n = 9) defines 10,509 significantly enriched peaks unique to tumor E+GSCs. e Functional enrichment analysis of GSC tumor-specific peaks using GREAT. Illustrated are top significantly enriched GO terms and examples of the corresponding annotated genes, which relate cell migration, cell-substrate assembly, and EGFR signaling

Fig. 2

TEAD is the top selectively enriched motif at GSC-specific open chromatin and TEAD1 is its most highly expressed family member across GBMs a, b Homer de novo motif discovery outlines the 20 most highly enriched TF motifs at chromatin accessibility regions defined by the GSC tumor-specific (a) and developmentally shared (b) differential ATAC-seq peak analyses (motifs in bold show selective enrichment in only one peak set). The TEAD motif (with highest scores for TEAD4 and TEAD1) is the top, selectively enriched motif within differential GSC tumor-specific peaks (in red). See also Supplementary Data 1. c Bar graph of rld-normalized gene expression values for all significantly and uniquely enriched GSC tumor-specific TF motifs, generated from parallel RNA-seq data in E+GSC and E−GBM populations. TEAD1 is the only highly expressed gene (top 25th percentile), which is differentially overexpressed in E+GSCs (*p = 0.0496 for TEAD1, **p = 0.006 for ZNF416, n = 3. Bars represent mean ± SEM). d Violin plot of TEAD1-4 expression in TCGA GBM RNA-seqV2 data (n = 150) shows that TEAD1 is the most highly expressed TEAD family member, followed by TEAD2, TEAD3, and TEAD4. Expression shown is log2(normalized counts + 1), normalized as detailed in methods. p-Values from one-sided Wilcoxon matched pairs test. Bar represents median value. e Bar graph of rld-normalized gene expression values for TEAD1-4 derived from RNA-seq E + GSC data (***p < 0.001; n = 3. Bars represent mean ± SEM)

Fig. 3

CRISPR-Cas9 ablation of TEAD1/4 inhibits migration in primary GBM cells. a Western immunoblot confirms population knockout of TEAD1 and TEAD4 after CRISPR-Cas9-mediated gene ablation. b Cell growth analysis reveals significantly decreased proliferation in TEAD1KO cells at 48–72 h, compared to sham (n = 3; 48 h: **p = 0.008; 72 h: *p = 0.01. Bars represent mean ± SEM). c Neurosphere (NS) assays show no difference in sphere number (day 6; n = 3 wells, multiple NS per well. Bars represent mean ± SEM). d Neurosphere (NS) assays show decreased sphere size in TEAD1 knockout, compared to sham. (day6; n = 3 wells, multiple NS per well; TEAD1KO: **p = 0.002. Dots represent individual NS and lines delineate mean). e Transwell invasion assays show decreased percent cell invasion in TEAD1KO and TEAD4KO cells, compared to sham (24 h; n = 3 wells; TEAD1KO: **p = 0.002; TEAD4KO: **p = 0.001. Bars represent mean ± SEM). On right, representative images of transwell invasion chamber membranes are shown. f, g Spheroid migration assays show decreased area of confluent cell migration (dispersion) at 36 h in TEAD1KO and TEAD4KO cells, compared to sham (f, PDL substrate), with partial rescue of migratory deficits in TEAD1KO cells after TEAD1 overexpression (OE) (g, laminin + PDL substrate) (n = 3 wells, with multiple NS per well; TEAD1KO: ***p = 0.0001; TEAD4KO: ***p = 0.00007; TEAD1OE 1: ***p = 0.00058; TEAD1OE 2 (1/10 dilution of TEAD1OE 1): ***p = 0.00015. Bars represent mean ± SEM). On right, migration area marked by red dash line in representative spheroids is shown. All experiments in a–g are performed on G-13063 cells. Scale bars = 75 μM. See also Supplementary Fig. 5

Fig. 4

In vivo infiltration is highly impaired in TEAD1KO GBM xenografts. a Representative immunofluorescent histology of tumor engraftment and core growth near the injection site in sham and TEAD1KO cells, 3.5 months after orthotopic xenotransplantation. b Representative immunofluorescence images of TEAD1 expression, confirming the stable loss of TEAD1 in TEAD1KO xenografts. Insets represent the entire TEAD1 immunofluorescence image with DAPI co-labeling. c Quantification of cell proliferation in sham and TEAD1KO cells at 3.5 months post xenotransplantation (n = 3 animals per condition; Ki67+ cells counted out of all HNA+ cells in 11 serial histological sections. Bars represent mean ± SEM). d Quantification of migratory tumor spread in sham and TEAD1KO cells at 3.5 months post xenotransplantation (n = 3 animals per condition; ***p = 0.0003 for both. Area = sum of areas with tumor spread in 11 serial coronal sections. Volume = average area × the distance between the first and the last serial section examined. Bars represent mean ± SEM). Schematic of the analyzed mouse brain serial coronal sections is depicted on the right. e Representative example of infiltrative tumor spread along the corpus callosum in sham and TEAD1KO cells at 3.5 months post xenotransplantation. HNA: human nuclear antigen, CC: corpus callosum, v: ventricle, a anterior, p: posterior. Scale bar = 50 μM

Fig. 5

Validation of TEAD1-binding targets by chromatin immunoprecipitation. a Density plot of correspondence analysis between chromatin accessibility and gene expression. Plotted on the y-axis is the average rld-normalized gene expression for each gene from all RNA-seq E+GSC sample data. Plotted on the x-axis is the highest ATAC-seq peak associated with the proximal promoter [−5 kb, +3 kb] of the same gene. Color intensity indicates density of the gene population, with red representing higher densities and blue representing lower densities. Strong correspondence is observed between open chromatin peaks and a moderate/high level of gene expression in all E+GSC samples (plot for one representative sample shown here). Several putative TEAD1-target genes of interest are indicated on the plot. b IGV plot of ATAC-seq piled reads at EGFR, AQP4, and CDH4 in four different acutely sorted E+GSCs (D.PROM distal promoter of EGFR, peak180759: chr7:55,000,372–55,001,595; P.PROM proximal promoter to TSS of EGFR, peak180777: chr7:55,085,981–55,088,747). For this IGV representation, reads are centered on the cut site of the Tn5 enzyme, correcting for the 9 bp occupancy of Tn5, and presumed footprints/peak troughs corresponding to TEAD motifs are delineated by downward arrow. c Chromatin immunoprecipitation (ChIP-PCR) in GBM tissues. Significant enrichment of TEAD1 (but not TEAD4) over IgG is seen specifically at differential open chromatin peaks with associated TEAD1 motif within the upstream promoter of EGFR (D. PROM, chr7:55001141 motif start site) (n = 6, *p = 0.011), at CDH4 (chr20:60011935 motif start site) (n = 6, *p = 0.027), and at the proximal promoter of AQP4 (chr18:24444251 motif start site) (n = 6, *p = 0.029). Significant binding above background is not observed at EGFR P.PROM or at chromatin inaccessible regions (IN2 = EGFR intron 2). Enrichment is expressed as fold increase over IgG, after normalization with 10% input, using 2^–ΔΔCt analysis accounting for primer efficiency: (E^IA−S)_sample/(E^IA−S)_IgG. E: primer efficiency; IA: 10% Input-Adjusted Ct; S: sample Ct. Bars represent mean ± SEM. d Western immunoblot illustrates decreased expression of AQP4 and CDH4 in TEAD1KO but not in TEAD4KO G-13063 GBM cells

Fig. 6

TEAD1 regulates expression of EGFR. a Immunofluorescence image depicts observed expression of EGFR in sham and TEAD1-knockout G-13063 xenografts 3.5 months post transplantation. Scale bar = 50 μM. b Western immunoblot depicts marked downregulation of EGFR in vitro after knockout of TEAD1, but not TEAD4, in G-13063 cells, which is partially restored after TEAD1 overexpression for 48 h. On right is a bar graph quantification of immunoblots from three independent experiments (n = 3; TEAD1KO vs. TEAD4KO: **p = 0.004 and **p = 0.003 for TEAD1 and EGFR, respectively; TEAD1KO + TEAD1OE vs. TEAD1KO: *p = 0.01 for TEAD1; p = 0.068 for EGFR in one tail t-test analysis. Bars represent mean ± SEM). c Western immunoblot depicts downregulation of pERK/ERK but not pAKT/AKT in TEAD1KO cells, compared to sham (n = 3 cell lines; pERK/ERK: *p = 0.029; bars represent mean ± SEM). On right are shown representative immunoblot images from one cell line

Fig. 7

TEAD1 regulates GBM migration by modulating AQP4 expression. a Venn diagram depicts the intersection of genes with TEAD motifs and genes consistently downregulated in TEAD1KO cell lines and migration-deficient spheroids (striped area). AQP4 is the only one of 32 genes in this intersection found to be a direct TEAD1 binding target in vivo (highlighted in red). ATAC-seq set contains 2612 peak-annotated genes from E+GSC vs. E−GBM+NSPC differential accessibility analysis. Top RNA-seq set (overall targets) contains all 1648 significantly up or downregulated genes from TEAD1KO vs. Sham differential expression analysis in all samples from four different patient-derived GBM lines and three migration experiments (n = 7; padj. < 0.05; all log2(fold change) values of HGNC-annotated genes included). Bottom RNA-seq set (migratory targets) contains 865 significantly up or downregulated genes from TEAD1KO vs. Sham G-13063 spheroids from three independent migration experiments (n = 3; padj. < 0.05; log2(fold change) >1 or <−1). See also Supplementary Fig. 7. b Spheroid migration assay showing significant reversal of cell dispersion deficit at 30 h in TEAD1 knockout cells after overexpression of CDH11 or AQP4 (G-13063 line; laminin + PDL substrate; n = 3 wells. TEAD1KO + CDH11OE: **p = 0.004; TEAD1KO + AQP4OE: **p = 0.0015. Bars represent mean ± SEM). On right, migration area marked by red dash line in representative spheroids is shown. Scale bar = 75 μM. c Quantification of AQP4 expression by RT-qPCR. AQP4 is significantly upregulated in TEAD1KO cells after TEAD1 overexpression (n = 6, 4 technical and 2 biological replicates; *p = 0.02 TEAD1KO + TEAD1OE vs. TEAD1KO) and is robustly expressed after exogenous lentivirus overexpression. Bars represent mean ± SEM