GATA3-dependent cellular reprogramming requires activation-domain dependent recruitment of a chromatin remodeler

Abstract

Background: Transcription factor-dependent cellular reprogramming is integral to normal development and is central to production of induced pluripotent stem cells. This process typically requires pioneer transcription factors (TFs) to induce de novo formation of enhancers at previously closed chromatin. Mechanistic information on this process is currently sparse.

Results: Here we explore the mechanistic basis by which GATA3 functions as a pioneer TF in a cellular reprogramming event relevant to breast cancer, the mesenchymal to epithelial transition (MET). In some instances, GATA3 binds previously inaccessible chromatin, characterized by stable, positioned nucleosomes where it induces nucleosome eviction, alters local histone modifications, and remodels local chromatin architecture. At other loci, GATA3 binding induces nucleosome sliding without concomitant generation of accessible chromatin. Deletion of the transactivation domain retains the chromatin binding ability of GATA3 but cripples chromatin reprogramming ability, resulting in failure to induce MET.

Conclusions: These data provide mechanistic insights into GATA3-mediated chromatin reprogramming during MET, and suggest unexpected complexity to TF pioneering. Successful reprogramming requires stable binding to a nucleosomal site; activation domain-dependent recruitment of co-factors including BRG1, the ATPase subunit of the SWI/SNF chromatin remodeling complex; and appropriate genomic context. The resulting model provides a new conceptual framework for de novo enhancer establishment by a pioneer TF.

Fig. 1

Chromatin domain bound by GATA3 becomes more accessible together with increased levels of enhancer-like histone marks. a Genomic localization of GATA3 across three breast cancer cell lines. UCSC Genome Browser snapshot showing the mapped read coverage of GATA3 ChIP-seq in MDA-MB-231 (blue), MCF7 (green), and T47D cells (red). The top motif identified by HOMER de novo motif analysis is represented at the right with P values. b Representative genomic loci acquired DNA accessibility. The GATA3-binding sites displayed: (top) significant increased level or (bottom) de novo peaks of ATAC-seq, H3K4me1, and H3K27ac signals after ectopic expression of GATA3. The other example loci are shown in Additional file 1: Figure S1. Metagene profiles of normalized ATAC-seq (c), H3K4me1 ChIP-seq (d), and H3K27ac ChIP-seq (e) are shown to compare the average signal levels centered on GATA3 ChIP-seq peaks in control and GATA3 expressing cells. The GATA3 peaks are divided into TSS-proximal (less than or equal to 1 kb) or distal (more than 1 kb) peaks based on the distance

Fig. 2

Identification and characterization of ‘Pioneer’ sites bound by GATA3. a Read density heatmaps showing the signal enrichment of GATA3 ChIP-seq, ATAC-seq, H3K4me1, and H3K27ac ChIP-seq. Classification from G1 to G4 was carried out based on ATAC-seq signal changes between control and GATA3 expressing cells. The number of peaks in each category is reported below the group label. Each row indicates a 10 kb window centered on a GATA3 binding site. The scale of read density after normalization (see Additional file 1: Table S6) is indicated at the bottom right. Metagene profiles of normalized ATAC-seq (b), H3K4me1 ChIP-seq (c), and H3K27ac ChIP-seq (d) in G1 are shown for the comparison of the average signal levels in control and GATA3 expressing cells. Metagene profiles in the other groups are shown in Additional file 1: Figure S2. e The signal enrichment of GATA3 ChIP-seq in each group is shown as a box-and-whisker plot. The top and bottom whiskers show 5th and 95th percentile, respectively. The horizontal line and cross mark the median and mean. *P <0.0001, Mann-Whitney test. Metagene plots of normalized tag density of GATA3 ChIP-seq in G1 (f) and G3 (g) performed with or without thermal treatment at 37 °C. The metagene plot in G2 and the quantitative analysis are represented in Additional file 1: Figure S2

Fig. 3

GATA3 induces nucleosome reorganization by targeting nucleosomal DNA. Metagene plots of normalized tag density of MNase-seq in G1 (a), G2 (b), and G3 (c) categories. Tag densities were normalized by the total counts in a +/- 1 kb window centered on the midpoint of GATA3 binding sites, then smoothed with a moving average (N  = 5). d The model of GATA3 bound nucleosome. GATA3 DBD and nucleosome core particle are represented in yellow and white, respectively. In the zoomed image, double-stranded DNA co-crystallized with GATA3 DBD is highlighted in cyan. The modified sequences are also indicated below. e Reconstituted nucleosome by salt dialysis method. The GATA consensus motif positive and negative DNAs were used to reconstitute mononucleosomes. f Purified recombinant GATA3 DBD. Purified proteins was analyzed by SDS-PAGE followed by staining with Coomassie. g Gel shift assay showing nucleosome binding of GATA3 DBD. The consensus motif containing nucleosome was used in lanes 1 to 4. The motif-mutated nucleosome was used in lanes 5 to 8. Nucleosomes (at a final concentration of 0.2 μM) were incubated with GATA3 DBD for 30 min at room temperature, and analyzed on native-polyacrylamide gel. The protein concentration of GATA3 DBD is as follows: 0 μM (lanes 1, 5), 0.5 μM (lanes 2, 6), 1 μM (lanes 3, 7), and 2 μM (lanes 4, 8)

Fig. 4

N-terminal truncation mutant binds but fails to open chromatin. a Schematic representation of wild-type GATA3 and N-terminal domain lacking (TA1del) mutant. b Genomic distribution of TA1del mutant. ChIP-seq with antibody against GATA3 or Ty1 is shown in purple or dark blue, respectively. ChIP-seq signals in wild-type GATA3 expressing cells were represented in the top and middle panels, while ChIP-seq signals in TA1del mutant expressing cells were shown in the bottom panel. c Venn diagram showing the peak overlap between wild-type GATA3 and TA1del mutant. d The top motif identified by de novo motif analysis is represented with P values. ChIP-seq peaks of TA1del mutant were analyzed by HOMER software. e Representative genomic loci showing that TA1del mutant expressing cells do not exhibit increased levels of ATAC-seq, H3K4me1 and K27ac signals as compared with wild-type GATA3 expressing cells. f Metagene profiles of normalized ATAC-seq signals showing the average tag density in TSS-proximal or -distal (>1 kb) peak flanking regions. The x-axis indicates the distance from the midpoint of TA1del mutant ChIP-seq peaks

Fig. 5

Transactivation domain is required for chromatin reprogramming. a Read density heatmaps showing the signal enrichment of Ty1 ChIP-seq, ATAC-seq, H3K4me1, and K27ac ChIP-seq in control, wild-type, and mutant GATA3 expressing cells. The same classification was utilized, but only the sites where both wild-type and mutant were localized are represented. The number of peaks in each category is reported. Each row indicates a 10 kb window centered on a GATA3 binding site. The scale of read density after normalization is indicated at the bottom right. Metagene profiles of normalized ATAC-seq (b), H3K4me1 ChIP-seq (c), and H3K27ac ChIP-seq (d) in G1 are shown for the comparison of the average signal levels in control, wild-type GATA3, and TA1del mutant expressing cells. Metagene profiles in the other groups are shown in Additional file 1: Figure S5. Metagene profiles of normalized MNase-seq signal density in G1 (e), G2 (f), and G3 (g) are shown, respectively. Top panels indicate the comparison between control and wild-type GATA3 expressing cells. Bottom panels indicate the comparison between control and TA1 del mutant expressing cells. Only overlapped peaks between wild-type and mutant GATA3 were used for the metagene analyses (b-g)

Fig. 6

BRG1 interacts with GATA3, and is recruited at GATA3-bound chromatin domains. a Ty1-GATA3 was immunoprecipitated from MDA-MB-231 nuclear extracts with Ty1 antibody. Immunoblots were performed with the indicated antibodies. b Endogenous GATA3 was immunoprecipitated from T47D nuclear extracts with GATA3 antibody. Immunoblots were performed with the indicated antibodies. c Representative genomic locus showing the BRG1 enrichment in control, wild-type, and mutant GATA3 expressing cells. The highlighted region displays increased signals of BRG1 ChIP-seq in wild-type GATA3 expressing cells. d Read density heatmaps showing the signal enrichment of BRG1 ChIP-seq signals. The same classification and common peaks used in Fig. 6a are represented. e Metagene profiles of normalized BRG1 ChIP-seq tag density in G1, G2, and G3 are shown