Mapping transcription factor occupancy using minimal numbers of cells in vitro and in vivo

Abstract

The identification of transcription factor (TF) binding sites in the genome is critical to understanding gene regulatory networks (GRNs). While ChIP-seq is commonly used to identify TF targets, it requires specific ChIP-grade antibodies and high cell numbers, often limiting its applicability. DNA adenine methyltransferase identification (DamID), developed and widely used in Drosophila, is a distinct technology to investigate protein-DNA interactions. Unlike ChIP-seq, it does not require antibodies, precipitation steps, or chemical protein-DNA crosslinking, but to date it has been seldom used in mammalian cells due to technical limitations. Here we describe an optimized DamID method coupled with next-generation sequencing (DamID-seq) in mouse cells and demonstrate the identification of the binding sites of two TFs, POU5F1 (also known as OCT4) and SOX2, in as few as 1000 embryonic stem cells (ESCs) and neural stem cells (NSCs), respectively. Furthermore, we have applied this technique in vivo for the first time in mammals. POU5F1 DamID-seq in the gastrulating mouse embryo at 7.5 d post coitum (dpc) successfully identified multiple POU5F1 binding sites proximal to genes involved in embryo development, neural tube formation, and mesoderm-cardiac tissue development, consistent with the pivotal role of this TF in post-implantation embryo. This technology paves the way to unprecedented investigation of TF-DNA interactions and GRNs in specific cell types of limited availability in mammals, including in vivo samples.

Figure 1.

Optimization of DamID-seq in mouse embryonic stem cells (ESCs) and comparison with ChIP-seq. (A) φC31 integrase–mediated cassette exchange system used to generate the cell lines in this study. (POI) Protein of interest; (ORF) open reading frame. (B) The optimized DamID-seq workflow. (C) POU5F1 DamID-seq tracks generated from 10⁶ ESCs and POU5F1 ChIP-seq tracks generated from 10⁷ ESCs (Buecker et al. 2014). The bars below each track represent ChIP-seq (black) and DamID-seq (blue) statistically significant peaks; y-axis represents read counts per million of Dam-POU5F1 (Dam-subtracted) and POU5F1 ChIP-seq (input-subtracted), respectively. (D) Peak overlap between three different published POU5F1 ChIP-seq data sets (Marson et al. 2008; Whyte et al. 2013; Buecker et al. 2014). (E) Overlap between POU5F1 DamID-seq peaks and the union of ChIP-seq peaks. Since DamID-seq peaks are larger than ChIP-seq peaks (see Supplemental Fig. S6B) and can contain multiple ChIP-seq peaks within, the number of the overlapping peaks are shown separately for each technology. (F) Motif enrichment analysis (Heinz et al. 2010) of the POU5F1-bound peaks identified only by DamID-seq, ChIP-seq, or both technologies. (G) Gene ontology (GO) enrichment analysis of POU5F1 peaks identified only by DamID-seq, ChIP-seq, or both using GREAT (McLean et al. 2010).

Figure 2.

Characterization of POU5F1 DamID-seq and ChIP-seq peaks. (A) For each POU5F1 DamID/ChIP-seq overlapping and specific peaks, the signal intensities from POU5F1 DamID-seq (Dam; Dam-POU5F1 separately), POU5F1 ChIP-seq, DNase-seq, and H3K4me3, H3K4me1, H3K27Ac, and H3K27me3 ChIP-seq are represented. (B) CpG methylation levels of POU5F1 DamID/ChIP-seq overlapping and specific peaks. The ChIP-seq–specific peaks were extended to the same average size of DamID-seq peaks as to avoid biases due to different size of the peaks. (C) Distribution of POU5F1 DamID/ChIP-seq overlapping and specific peaks in the genome. (D) Expression level of genes with DamID-seq–specific (384 genes), overlapping (739 genes), and ChIP-seq–specific peaks (2372 genes) in their promoter regions (±3 kb from TSS). (E) Expression changes of genes in D, 24 h after knockdown of Pou5f1 in ESCs (King and Klose 2017). (F) DamID/ChIP-seq overlapping and specific peaks including mouse ChIP-seq HOT regions (Wreczycka et al. 2017).

Figure 3.

10⁶ NSC SOX2 DamID-seq in comparison with ChIP-seq. (A,B) Union of two published NSC SOX2 ChIP-seq data sets (A) (Mateo et al. 2015; Mistri et al. 2015) and its overlap with 10⁶ NSC SOX2 DamID-seq peaks (B). (C) Motif enrichment analysis (Heinz et al. 2010) of the SOX2-bound peaks identified only by DamID-seq, ChIP-seq, or both technologies. (D) GO enrichment analysis of SOX2 peaks identified only by DamID-seq, ChIP-seq, or both using GREAT (McLean et al. 2010). (E) For each SOX2 DamID/ChIP-seq overlapping and specific peaks, the signal intensities from SOX2 DamID-seq (Dam; Dam-SOX2 separately), SOX2 ChIP-seq, DNase-seq, and H3K4me3 and H3K27me3 ChIP-seq are represented. (F) Distribution of SOX2 DamID/ChIP-seq overlapping and specific peaks in the genome. (G) DamID/ChIP-seq overlapping and specific peaks including mouse ChIP-seq HOT regions (Wreczycka et al. 2017).

Figure 4.

10⁴, 10³ ESC POU5F1 DamID-seq and 10⁴, 10³ NSC SOX2 DamID-seq. (A) POU5F1 DamID-seq tracks from 10⁶/10⁴/10³ ESCs and POU5F1 ChIP-seq track from 10⁷ ESCs (Buecker et al. 2014). (B) Overlaps of 10⁶/10⁴/10³ ESC POU5F1 DamID-seq peaks. (C) Read counts of peaks in 10⁶ ESC DamID-seq (blue) and those identified by 10⁴ (gray) and 10³ (green) ESC DamID-seq. (D) Percentage of ESC super enhancers (Whyte et al. 2013) containing POU5F1 DamID-seq peaks using different number of cells. (E) Motif enrichment in the 10³ ESC POU5F1 DamID-seq peaks. (F) GO enrichment analysis of 10³ ESC POU5F1 DamID-seq peaks using GREAT (McLean et al. 2010) (G) SOX2 DamID-seq tracks from 10⁶/10⁴/10³ NSCs and SOX2 ChIP-seq tracks generated from 5 × 10⁶ NSCs (Mateo et al. 2015). (H) Overlaps of 10⁶/10⁴/10³ NSC SOX2 DamID-seq peaks. (I) Motif enrichment in the 10³ NSC SOX2 DamID-seq peaks. (J) GO enrichment analysis of SOX2 DamID-seq peaks from 10³ NSCs.

Figure 5.

POU5F1 DamID-seq with 7.5-dpc epiblasts. (A) DamID-seq samples were prepared from each Dam/Dam-POU5F1–expressing 7.5-dpc chimeric embryo generated via morula aggregation, excluding the PGC-containing region. Nanog-GFP confirms contribution of Dam/Dam-POU5F1 ESCs, while the reporter expression is limited to posterior. (B) Read counts and coefficient of variation of the epiblast POU5F1 DamID-seq peaks. Peaks with high read counts (log₂ > 4.5) and low standard deviation (<75%) indicated in red were used for further analyses. (C) The merged epiblast POU5F1 DamID-seq tracks (top) generated from six Rosa26-Neo-Dam-Pou5f1 and six Rosa26-Neo-Dam embryos. Blue bars indicate selected confident peaks from B.

Figure 6.

Analysis of POU5F1 target genes in the 7.5-dpc epiblasts. (A) Motif enrichment analysis (Heinz et al. 2010) of the epiblast POU5F1 DamID-seq peaks. (B) GO enrichment analysis of POU5F1 DamID-seq peaks from 7.5-dpc epiblasts using GREAT (McLean et al. 2010). (C) Differentially expressed genes between ESCs and EpiSCs (Tesar et al. 2007). Red triangles represent the POU5F1 binding peak–associated genes in 7.5-dpc epiblasts. (D) Expression levels of the 7.5-dpc epiblast POU5F1 binding peak–associated genes in ESCs and post-implantation epiblasts (Kojima et al. 2014). Genes whose expression significantly changed 24 h after Pou5f1 deletion in the 7.5-dpc mouse embryo (DeVeale et al. 2013) are indicated in color. (CAV) Epiblast of cavity; (PS) prestreak; (LMS) late mid streak; (LS) late streak; (OB) no bud; (EB) early bud; (LB) late-bud. (E,F) POU5F1 binding peaks identified by DamID-seq (E) and expression changes of the development-related six genes upon Pou5f1 deletion in 7.5-dpc embryos (F) (DeVeale et al. 2013).