Differences in the epigenetic and reprogramming properties of pluripotent and extra-embryonic stem cells implicate chromatin remodelling as an important early event in the developing mouse embryo

ES (OS25), TS (B1) and XEN (IM8A1) cells selectively express genes characteristic of either the ICM, trophectoderm or primitive endoderm

Reverse transcription polymerase chain reaction (RT-PCR) was used to assess the relative abundance of different mRNA transcripts in ES (OS25, [23]), TS (B1, [21]) and XEN (IM8A1, [22]) cell lines. Consistent with previous reports, ES cells expressed Oct4, Nanog, Sox2, Fgf4, Rex1 [24] and Esrrb [25], TS cells expressed Cdx2, Eomes, Esrrb and Hand1 [21], and XEN cells selectively expressed Gata4, Gata6, Foxa2 and Hnf4 [22] (Figure 1B and Additional file 1). Rex1 transcripts were detected in all three cell types but were most abundant in ES cells; Sox2 transcripts were detected in both ES and TS cells; Eomes transcripts were detected in all three cell types but were most abundant in TS cells. These data show that each of the stem cell lines displays a different profile of gene expression, in line with previous studies [21, 22, 24] and with their different origins. At the level of specific genes, however, there is considerable overlap in expression between the cell lines.

ES, TS and XEN cell lines have similar but distinct replication timing profiles

In order to directly compare the epigenetic profiles of extra-embryonic stem cell lines with those of pluripotent cell lines, we initially assessed the replication timing of a panel of developmental genes in OS25, B1 and IM8A1 cell lines. Genes include those that encode transcription factors regulating the specification of germ layers in the embryo [19], as well as those encoding transcription factors that are important for the biology of early embryonic ICM, TE, PrE and EPI lineages. Replication was assessed using a previously established assay [26, 27] in which asynchronous cells are pulse-labelled with 5-bromo-2-deoxyuridine (BrdU), fractionated according to cell-cycle stage (see Additional file 2, part A) and the relative abundance of newly synthesized locus-specific DNA is compared between successive cell cycle fractions using quantitative PCR. Although the exact relationship between chromatin structure and replication timing is not fully understood, early replication is a characteristic of 'accessible' and highly acetylated chromatin while late replication is a feature of heterochromatic domains and some repressed genes [28]. Consistent with this, α-globin a constitutively early replicating gene, was detected in S1 fractions isolated from ES, TS and XEN cells (Additional file 2, part B top panel), while Amylase 2.1, a late replicating control, was detected in S3 and peaked in the S4 fractions in all three cell types (Additional file 2, part B middle panel) [19]. Detection of similar levels of BrdU-labelled Gbe DNA in cell cycle fractions that were 'spiked' with a constant amount of Drosophila BrdU-labelled DNA (Additional file 2, part B lower panel), confirmed an equivalent recovery of immuno-precipitated DNA in all analyses shown. The replication times of candidate genes were determined from at least two independent experiments, scored according to a peak abundance of locus-specific DNA (in G1/S1 [early], S2 [middle-early], S2 and S3[middle], S3 [middle-late] or S4/G2 [late]) and the results were colour-coded to facilitate comparison (see Figure 2, as previously described [19, 27]).

Most of the genes analysed replicated in early (or middle-early) S-phase in all three stem cell lines (44, 44, 45 out of 58 genes in ES, TS and XEN cells, respectively, Figure 2A). These included a subset of ICM-associated genes (Oct4, Nanog and Fgf4) expressed by ES cells, as well as genes associated with TE (Cdx2 and Hand1) and PrE (Gata4 and Hnf4). In addition, many genes that are not thought to be expressed at significant levels in any of these cell types, for example Math1, Scl and Myog, replicate early in all three embryonic stem cell lines. These data suggests that in TS and XEN cells, many developmental regulator genes remain 'accessible' - as reflected by the prevalence of early replicating loci - similar to that reported previously for ES cells [19]. Overall, the replication timing profiles of ES and TS cells were similar (41/58) or identical (32/58), while XEN cells showed a greater disparity. This is illustrated by a delayed replication of several pluripotency-associated genes in XEN cells (for example, Rex1 and Sox2) and the early replication of PrE-associated genes Gata6 and Foxa2 (Figure 2B) and is in keeping with the idea that some tissue-specific genes may replicate earlier when transcriptionally active [29, 30]. Similarly, Pem and Psx1, which encode factors required for extra-embryonic lineages, replicated later in ES cells as compared to TS and XEN cells and the replication of Pl1, a TE-specific factor, was selectively advanced in TS cells (Figure 2B). These results were confirmed by analysing additional independent TS and XEN cell lines (Additional file 1) that were derived from mice carrying floxed Dicer alleles [31]. Comparing TS^B1 and TS^Dicerfx/fx or XEN^IM8A1 and XEN^Dicerfx/fx (Figure 2 and Additional file 1), as well as numerous different ES cell lines [19, 32], confirmed that the replication timing profiles of different embryonic and extra-embryonic cell lines were robustly preserved.

Interestingly, the neural-associated genes Sox1 and Neurod that are not expressed by any of the embryonic stem cell lines, showed clear differences in replication timing between ES, TS and XEN cells (Figure 2B, lower panel). Sox1 replication was advanced in ES cells while Neurod replicated early in XEN cells. Although unexpected, these results suggest underlying changes in the chromatin context of these genes in the stem cell lines. In the case of Neurod, although the transcription factor is known to function in neuronal development, it has also been shown to have an important role in the development of specialized cell types arising from the gut endoderm [33]. Despite being derived from the EPI and not from the PrE, gut endoderm cells have morphological and functional similarities to visceral endoderm cells [34]. The advanced replication of Neurod in XEN cells might therefore reflect changes in transcriptional competence at the locus that is associated with an affiliation to the 'endoderm' lineage.

Chromatin profiling of gene promoters in stem cell lines

The chromatin profile of important regulator genes was compared between embryo-derived stem cell lines using chromatin immunoprecipitation (ChIP) in order to evaluate the abundance of specific histone modifications that are associated with either active (H3K4me2, H4ac and H3K9ac) or repressed (H3K27me3 and H4K20me3) chromatin. For these analyses primers were designed to recognize the promoter region (up to 600 kb upstream the transcriptional start site) of each candidate gene; genes that are known to be abundantly expressed by each cell type were used as positive controls for 'active' chromatin marks. Pericentric heterochromatin (γ-satellite repeats) provided controls for H4K20me3 immuno-precipitations, H3K27me3 was validated by analysing known bivalent loci in ES cells [19], and the abundance of modified histones was calculated relative to histone H3.

As anticipated, the promoters of many genes that are overtly expressed in ES cells (shown in bold, Figure 3, upper panel), as well as many bivalent genes (including Eomes, Fgf5, Foxd3, Mash1, Math1, Sox1 and Tbx15) were enriched for H3K4me2, and/or H3K9ac and H4ac at their promoters [19, 20]. Exceptions included the promoters of Tpbpa and Pl1, two markers of differentiated trophoblast. In ES cells histone H3K27me3, a modification catalyzed by polycomb repressor complex 2 (PRC2), was abundant at the promoters of genes that were either not expressed or expressed at low levels, including TS-associated genes (Cdx2, Eomes, Pem, Psx1), PrE-associated genes (Gata6, Foxa2) and genes that are normally expressed by subsets of differentiated tissue (such as Mash1, Math1 and Neurod) (shown in purple in Figure 3). Some silent late-replicating genes showed only low levels of H3K27 trimethylation (Tpbpa and Pl1) suggesting that these genes, in contrast to bivalent genes, are not developmentally 'poised' in ES cells and may, therefore, require extensive chromatin-remodelling for correct developmental expression. Levels of promoter H4K20me3 (shown in red in Figure 3), a mark associated with mammalian pericentric heterochromatin [35], were modest in ES cells with the exception of Sox2 (an observation that is likely to reflect the fact that OS25 cells carrying Sox2 as a transgene). In TS (B1) and XEN (IM8A1) cell lines, in contrast to ES (OS25), H4K20me3 was detected at the promoters of many genes and was particularly enriched at several silent genes in XEN cells (Sox2, Foxd3, Sox1) (Figure 3, see Additional file 3 and Figure 1A for expression data). Taken as a whole these ChIP analyses suggest that, although the promoters of many development regulator genes are co-marked with histone modifications associated with active (acetylated, H3K4me2) and repressive (H3K27me3) chromatin in both ES and TS cells, this is not the case in XEN cells. Rather, in XEN cells histone marks that characterize accessible chromatin genes tend to be restricted to genes that are productively expressed at high (Gata6, Foxa2, Pem, Psx1) or moderate levels (Eomes, Fbx15, Rex1, Tbx15). The exception to this generalization is Math1 (lower panel of Figure 3 and Additional file 3), a promoter that is enriched for H3K4me2 in ES, TS and XEN cells and, therefore, appears to retain a bivalent (or poised) configuration. Although we do not currently know the cause or significance of this single observation, collectively our data suggest that the chromatin structure of many genes is different between the stem cell lines, supporting earlier proposals that epigenetic reprogramming occurs in embryonic and extra-embryonic lineages during early mouse development [17, 18].

Mouse TS, XEN and ES cell lines dominantly reprogram lymphocytes in interspecies heterokaryons but induce the expression of different lineage-associated genes

Spontaneous and experimental cell fusions between ES cells and cells from a range of somatic tissues result in the nuclei of differentiated cells being reprogrammed to express an ES-specific gene expression pattern (so-called dominant reprogramming) [36, 37]. Although the rules of dominance are not fully understood, other stem cell populations, including embryonic germ and embryonic carcinoma cells, can also reprogramme somatic cells towards a pluripotent state in vitro [38, 39]. In order to establish whether extra-embryonic stem cell populations have this capacity and, moreover, whether reprogramming by these extra-embryonic cells induces the de novo expression of different cohorts of genes, we tested the ability of TS and XEN cells in order to reprogramme human B cells using a previously established assay system [40]. Briefly, mouse stem cells and human B-lymphocytes were mixed in a 1:1 ratio and inter-species heterokaryons (cells in which parental nuclei share the same cytoplasm but remain spatially separate) were generated by polyethylene glycol (PEG)-mediated cell fusion. Human B cell reprogramming within these heterokaryons was assessed 1 to 3 days after fusion, by qRT-PCR using primers that were designed (and validated) to specifically amplify human transcripts. Expression of HPRT served as a positive control in these analyses and transcripts derived from the human pluripotency-associated genes (OCT4, NANOG, CRIPTO and REX1), TE (CDX2 and HAND1) and PrE-associated genes were examined in detail (GATA6, FOXA2 and HNF4) (Figure 4). Human B cells (hB) did not express detectable levels of pluripotency-associated transcripts prior to fusion, but following heterokaryon formation with mouse ES cells (hBxES), expression of OCT4, NANOG, REX1 and CRIPTO was initiated and increased up to day 3 (Figure 4, upper panel left). Fusion of hB with mouse TS (B1) or with mouse XEN (IM8A1) cells did not induce the expression of any of the human pluripotency-associated genes tested, including CRIPTO. This could be considered surprising as XEN (IM8A1) cells express high levels of mouse Cripto transcripts (several fold more than ES cells) and both TS and XEN cell lines express mouse Rex1 (Additional file 4). In heterokaryons formed between hB and TS (B1), expression of TE-associated genes (CDX2 and HAND1) was induced, as well as low levels of some PrE-associated genes (GATA6 and FOXA2). Fusions between hBxXEN cells resulted in a rapid and sustained induction of GATA6, FOXA2 and HNF4 (Figure 4). In contrast, mouse lymphocyte-specific transcripts (such as CD19, CD37 and CD45) were not detected throughout these experiments (data not shown) which is in line with the dominance of embryonic and extra-embryonic stem cells in reprogramming. These results collectively show that XEN and TS stem cell lines, like ES cells, retain a capacity to dominantly reprogramme somatic cells, but impose a programme of gene expression that is consistent with their different lineage affiliations.

Cell lines and cell culture

ES cells (OS25) were maintained in an undifferentiated state on 0.1% gelatin (StemCell Technologies, Vancouver, Canada)-coated flasks (Fisher Scientific UK Ltd, Leicestershire, UK) in G-MEM-BHK 21 medium (Invitrogen Ltd, Paisley, UK) supplemented with 10% fetal calf serum (FCS; PAA Laboratories, Gmbh, Pasching, Austria), non-essential amino acids, sodium pyruvate, sodium bicarbonate, antibiotics, L-glutamine, β-mercaptoethanol (Sigma-Aldrich Co Ltd, Gillingham, UK) and ESGRO-LIF (1000 U/ml) (Chemicon/Millipore, Billerica, USA). TS cell lines (B1 and Dicerfx/fx) were cultured in the presence of 70% mitotically inactivated mouse embryo fibroblast cells-conditioned medium and 30% TS medium to which human recombinant Fgf4 (25 ng/ml) (Sigma-Aldrich) and heparin (1 μg/ml) (Sigma-Aldrich) were added. The TS cell medium was RPMI 1640 supplemented with 20% FCS (GlobePharm, Cork, Ireland), sodium pyruvate, β-mercaptoethanol, L-glutamine and antibiotics. XEN cell lines (IM8A1 and Dicerfx/fx) were maintained on 0.1% gelatin-coated flasks in RPMI 1640 supplemented with 20% FCS (GlobePharm), sodium pyruvate, L-glutamine, antibiotics and β-mercaptoethanol. EBV-transformed human B-lymphocyte clones were maintained in RPMI medium supplemented with 10% FCS (GlobePharm), L-glutamine and antibiotics. All cell lines used in this study were subjected to karyotypic analysis to check chromosome number. XEN cell lines routinely contained 40-46 chromosomes consistent with their previously reported aneuploid status [22] while ES and TS cell lines appeared normal.

RT-PCR analysis

RNA extraction from ES, TS, XEN cells and heterokaryons was performed using RNeasy protect mini kit (Qiagen, USA) and RNase-free DNase set (Qiagen) for digestion of residual DNA. Total RNA (2.5 μg) was then reverse transcribed using the Superscript first-strand synthesis system (Invitrogen) and cDNA of interest amplified in a total reaction volume of 50 μL using 500 nM primers, and 1.25 U of HotStarTaq (Qiagen). The PCR cycling conditions were as follows: 95°C for 2 min, 30 cycles 95°C for 30 s, annealing at 60°C or 65°C for 30 s and elongation at 72°C for 2 min, finishing with a step at 72°C for 10 min.

Replication timing assay

BrdU-labelling, ethanol fixation, cell cycle fractionation by flow cytometry and isolation of BrdU-labelled DNA by immunoprecipitation were carried out as previously described [19] with the same BrdU-pulse labelling time for all three stem cell populations (30 min). The abundance of newly replicated DNA in each cell-cycle fraction was determined by real-time PCR amplification.

Real-Time PCR analysis

Real-Time PCR analysis was carried out on a Opticon™ DNA engine (MJ Research, Inc, MA, USA) under the following cycling conditions: 95°C for 15 min, 40 cycles at 94°C for 15 s, 60°C for 30 s, 72°C for 30 s followed by plate read. PCR reactions were performed in a 30 μL reaction volume containing 2× SYBR Green (Qiagen), 1.5 μL of template and 300 nM primers. Each measurement was performed in duplicate. For heterokaryon analysis data were normalised to human GAPDH expression.

Chromatin immunoprecipitation analysis

Exponentially growing ES, TS and XEN cells were processed for ChIP analysis as described previously [19]. 140 μg chromatin was subjected to immunoprecipitation with 5 μL anti-H3K9ac (Upstate Biotechnology, NY, USA), 5 μL anti-H3K4me2 (Upstate), 5 μL anti-H4ac (Upstate), 5 μL anti-H3K27me3 (Upstate), 5 μL anti-H4K20me3 (Upstate), 2.5 μL of a rabbit anti-mouse-IgG antiserum (negative-control) (Dako Inc, CA, USA) and 4 μL of anti-H3 (Abcam, MA, USA). After purification, DNA was resuspended in 80 μL TE solution. Quantification of precipitated DNA was performed using real-time qPCR (quantitative PRC) amplification. Histone's modification levels were normalized against total H3 detected and the ratio of modified-H3 to H3 was denoted as relative enrichment. ChIP experiments were performed twice.

Experimental heterokaryons

Heterokaryons were generated by fusing either mouse ES, TS or XEN cells and human B-lymphocytes using 50% polyethylene glycol, pH 7.4 (PEG 1500, Roche, Hertfordshire, UK). Equal numbers of stem cells and B-lymphocytes were mixed, washed twice in phosphate buffered saline at 37°C and 1 mL of PEG at 37°C was added to the pellet of cells over 60 s followed by an incubation at 37°C for 90 s. Cell mixtures were washed with 10 mL of DMEM at 37°C added over 3 min. After centrifugation the pellet was allowed to swell in complete medium for 3 min before resuspension. In order to eliminate non-fused hB cells Ouabain (10^-5 M) was added to the medium. Proliferating stem cells were eliminated by the addition of 10^-5 M Ara-C 6 h after fusion and then removed after 12 h. Fused cells were cultured under conditions promoting the maintenance of undifferentiated mouse stem cells.