The transcriptional foundation of pluripotency

Abstract

A fundamental goal in biology is to understand the molecular basis of cell identity. Pluripotent embryonic stem (ES) cell identity is governed by a set of transcription factors centred on the triumvirate of Oct4, Sox2 and Nanog. These proteins often bind to closely localised genomic sites. Recent studies have identified additional transcriptional modulators that bind to chromatin near sites occupied by Oct4, Sox2 and Nanog. This suggests that the combinatorial control of gene transcription might be fundamental to the ES cell state. Here we discuss how these observations advance our understanding of the transcription factor network that controls pluripotent identity and highlight unresolved issues that arise from these studies.

Fig. 1.

Nanog, Oct4 and Sox2 protein domains. Each protein is divided into domains, either real or putative. DNA-binding domains are shown in green and regions with reported trans-activating potential in orange. (A) Nanog can be divided into N-terminal and C-terminal halves. The N-terminal half contains a DNA-binding homeodomain (HD) and an N-terminal domain (ND). The C-terminal half contains a dimerisation domain (blue) referred to as the tryptophan repeat (WR), in which every fifth residue is a tryptophan (Mullin et al., 2008), that separates C-terminal domain 1 (CD1) from C-terminal domain 2 (CD2). (B) Oct4 has DNA-binding domains comprising a POU-specific DNA-binding domain (POU_S) and a POU-homeodomain (POU_HD), each of which can interact independently with DNA (Herr and Cleary, 1995), as well as transactivation domains located N-terminal (N-TAD) or C-terminal (C-TAD) to the POU domain. (C) Sox2 is a High mobility group (HMG) family member and has a single HMG DNA-binding domain and a transactivation domain (TAD). The size of each protein is indicated in amino acid residues (aa). Drawings are not to scale.

Fig. 2.

The Oct/Sox DNA motif and the ternary structure of Oct-Sox-DNA. (A) The DNA sites recognised by Oct4 and Sox2 are both non-palindromic and might be expected to exist in four distinct relative orientations. However, one of these orientations predominates at sites that have been validated by reporter assays. (B) Nuclear magnetic resonance (NMR) analysis of Oct1 bound to DNA shows two conformations. (a) The sequence of a composite Oct/Sox site is illustrated. (b) After binding of Oct1 to the DNA, the binary complex exists in two conformations; the POU homeodomain (HD) contacts DNA directly in both, but in only one is the POU-specific DNA-binding domain (S) also in direct contact with DNA. (c) DNA binding by Sox2 via its HMG DNA-binding domain (HMG) provides stabilising side-chain interactions that lock POU_S onto the DNA. (C) The ternary structure of Oct1 (red) and Sox2 (green) bound to the Hoxb1 regulatory element (blue DNA). The backbone position of residues, the side chains of which provide stabilising interactions, are highlighted (yellow). Reprinted with permission from Williams et al. (Williams et al., 2004).

Fig. 3.

Affinity-based methods for identifying interacting proteins. (A,B) Modification of a transcription factor (TF) using epitope tags. (A) The TF open reading frame is modified by in-frame fusion to an epitope tag encoding the so-called Flag peptide against which antibodies are commercially available. (B) A 16-residue tag (BIO) added to the TF is a substrate for biotinylation when expressed in cells that are engineered to express the E. coli BirA gene, which encodes a biotin ligase. (C) Affinity purification and partner identification strategy. Nuclear extracts are prepared from ES cells that express an epitope-tagged (Ep) form of a TF under conditions in which binding of TF-interacting proteins to the TF is maintained. Protein complexes are then incubated with an affinity reagent (anti-Flag IgG or streptavidin) that is immobilised on a solid support. After washing to remove contaminating proteins, the captured proteins are separated by electrophoresis, excised from the gel, digested with protease and peptides identified by mass spectrometry. PAGE, polyacrylamide gel electrophoresis.

Fig. 4.

Global chromatin immunoprecipitation (ChIP) methods. A comparison of the (A) ChIP-seq and (B) ChIP-chip approaches. (A) ES cells that express genes (green) encoding the protein to be analysed are treated with a reagent that cross-links chromatin-associated proteins to DNA. Chromatin is then prepared from the cells and sonicated to reduce the average size of the DNA fragments. Various chromatin-associated proteins (coloured shapes) are bound to the DNA (double wavy lines), including the protein of interest, which is shown in green bound to its target DNA (also in green). The sonicated chromatin is incubated with an antibody against the protein of interest and the antibody-chromatin complexes are then collected after incubation with anti-Ig immobilised on a solid support, centrifugation and washing. This enriches for both the target protein relative to other chromatin-associated proteins and its associated target DNAs (shown in green). After reversal of cross-links, the DNA is purified and analysed by Solexa sequencing and the signals corresponding to binding sites mapped to the genome (see Fig. 5). (B) ES cells that express the biotin ligase gene BirA are transfected with an expression construct that encodes a tagged version (blue) of the protein of interest (green; see Fig. 3B). Chromatin is then prepared from the BirA line and the derivative line expressing the protein of interest and is processed as in A, except that sonicated chromatin is incubated with streptavidin coupled to a solid phase. Following the collection and washing of streptavidin-coupled chromatin, protein-DNA cross-links are reversed and the purified DNA is hybridised to a microarray of promoter fragments that typically extend from +8 to -2 kb relative to the transcription initiation site of a subset of genes.

Fig. 5.

Overlap in data between large-scale global localisation studies. (A) Workflow for ChIP-chip and ChIP-seq studies and comparison of outputs. (B) Overlap of target genes obtained for biotinylated c-Myc, Nanog, Oct4 and Sox2 from the studies of Kim et al. (blue) (Kim et al., 2008) and Chen et al. (yellow) (Chen et al., 2008b). To facilitate a comparison of data from the two studies, the mapping of target genes to peak binding positions was recalculated from the reported peak locations. Peaks were assigned to the nearest gene, as annotated in the Ensembl v46 mouse genomic data. Target peaks were only considered if they were in genomic regions measured by both the promoter array and the ChIP-seq study. The number of genes with ChIP-seq peaks within 10 kb of the transcription initiation site but not mapped to the microarray are shown as excluded counts.

Fig. 6.

Illustration of an active promoter in ES cells. Pluripotency transcription factors are shown bound to an enhancer sequence upstream of the transcription initiation site of an associated gene (indicated by the arrow). Two monomeric subunits of Nanog are illustrated to reflect the fact that Nanog is active in dimeric form (Mullin et al, 2008; Wang et al, 2008). Contact between enhancer-bound transcription factors and the RNA polymerase II (Pol II) complex machinery occurs through a bridging interaction with the p300 transcriptional co-activator.