BRACHYURY and CDX2 Mediate BMP-Induced Differentiation of Human and Mouse Pluripotent Stem Cells into Embryonic and Extraembryonic Lineages

Summary BMP is thought to induce hESC differentiation toward multiple lineages including mesoderm and trophoblast. The BMP-induced trophoblast phenotype is a long-standing paradox in stem cell biology. Here we readdressed BMP function in hESCs and mouse epiblast-derived cells. We found that BMP4 cooperates with FGF2 (via ERK) to induce mesoderm and to inhibit endoderm differentiation. These conditions induced cells with high levels of BRACHYURY (BRA) that coexpressed CDX2. BRA was necessary for and preceded CDX2 expression; both genes were essential for expression not only of mesodermal genes but also of trophoblast-associated genes. Maximal expression of the latter was seen in the absence of FGF but these cells coexpressed mesodermal genes and moreover they differed in cell surface and epigenetic properties from placental trophoblast. We conclude that BMP induces human and mouse pluripotent stem cells primarily to form mesoderm, rather than trophoblast, acting through BRA and CDX2.

F. FGF2, 20 ng/ml; A, ActivinA, 100 ng/ml). Samples were immunostained for BRA (i) or SOX17 (ii) and the nuclear marker DAPI. B. Representative fluorescent images of undifferentiated H9 hESCs. Samples were immunostained for BRA or SOX17 and the nuclear marker DAPI as indicated. C. qPCR analysis of H9 hESCs grown for one day in CDM + BMP4 (B, 10 ng/ml) plus or minus FGF2 (F, 20ng/ml) or the FGF signalling inhibitor SU5402 (SU, 10 µM). D. Western blot analysis of pJNK1, pJNK2 and p38 expression in H9 hESCs grown for one day in CDM supplemented with BMP4 (10 ng/ml) plus or minus FGF2 (20 ng/ml) and a gradient of Activin/Nodal signalling (absence, 10 or 100 ng/ml), where 'Sb' refers to SB431542 treatment, '-' refers to no Activin A addition and '10' or '100' refer to the amounts of Activin A in ng/ml

Human ESC and Mouse EpiSC Culture in Chemically Defined Conditions
For differentiation, H9 hESCs were grown with or without FGF2 (20 ng/ml, zebrafish, Hyvönen) for up to 7 days. Early mesoderm differentiation was induced by FGF2 + LY294002 + BMP4 (10 ng/ml), named FLyB; early endoderm differentiation was induced by FGF2 + LY294002 + ActivinA (100 ng/ml), named FLyA. Proximal streak mesoderm differentiation was induced by treating cells with FLyB for 36h followed by FGF2 + BMP4 (50ng/ml) for 3.5 days, while mid/distal-streak mesoderm differentiation was induced by treating cells with FLyB for 36 hours followed by FGF2 + Ly (1 μM) for 3.5 days containing conditions. Trophoblast-like differentiation was induced by treating cells with BMP (10ng/ml) + Sb for 5 or 7 days. The hESC line HUES9 was grown as previously described (Cowan et al., 2004) or differentiated in CDM-PVA as for H9 hESCs. All experiments were repeated at least twice on different passages of cells to ensure that the patterns of gene expression described were reproducible.

Human ESC knockdown lines
Stable knockdowns of BRA and CDX2 were carried out with pLKO.1-shRNA vector (Sigma) by Lipofectamine transfection following manufacturer's instructions. A Scrambled pLKO.1-shRNA vector (Sigma) was used as a control. Stable clones were screened by PCR and immunostaining and the percentage of knockdown was 13 determined by comparison to expression in the scrambled control transfected lines.

Mouse Late Epiblast Layers Dissection and Culture Conditions
Late epiblast layers were dissected manually from E6.5 129S2 mouse embryos as previously described (Brons et al., 2007). For differentiation, epiblast layers were grown overnight (up to 24 h) in CDM + Activin + FGF and then moved to CDM supplemented with combinations of the growth factors / inhibitors described above for up to 7 days. All differentiation experiments were repeated at least twice using late epiblast layers from different litters.

Human placental mesenchyme extraction and culture
Human placental mesenchyme for culture was isolated by negative selection performed

RNA extraction, cDNA synthesis and amplification of mouse late epiblast layer explant cultures
Amplified cDNA was purified using a PCR-purification kit (Qiagen) and following manufacturer's instructions. Ethanol precipitation of the DNA was then performed 14 to further purify the cDNA samples.

RNA extraction and Quantitative Polymerase Chain Reaction
Each sample was run in duplicate and normalised to Porphobilinogen Deaminase (PBGD) in the same run. Primer sequences can be found in Tables S1 and S2. Error bars on all QPCR graphs represent standard deviation from three independent biological replicates. Student's t tests (two-tailed assuming nonequal variance) were performed; superscripts indicate that all three replicates showed significantly different gene expression (*p ≤ 0.05; + p≤0.01) compared to the day zero undifferentiated sample group, except where indicated.

Immunobblot analysis
Membranes were incubated overnight with primary antibody at a concentration of 1:1000 or 1:5000 (pERK1/2, p38, pAKT, pJUN1/2 and pJUNK were from Cell Signalling; NANOG and α-TUBULIN were from Abcam, BRA was from R&D Systems). Secondary antibody incubations (1:10000, HRP-conjugated rabbit anti-goat, goat anti-mouse or goat anti-rabbit from Sigma) were performed at room temperature for 1h30min. Lastly, Dura-West substrate (Thermo Scientific), prepared according to the manufacturer's instructions, was used to develop the immunobblots prior to X-ray film exposure.

Immunofluorescence
Primary antibody incubations were performed over-night at 4°C as follows:

Flow Cytometry of Intracellular Proteins
Antibodies for intracellular detection were added directly conjugated to a fluorofore: 16

Flow Cytometry of Extracellular Proteins
Unlabelled primary antibodies were added and detected with fluorochrome-conjugated anti-mouse IgG or IgM (Sigma-Aldrich). Free secondary antibody-binding sites were blocked with murine immunoglobulin, before staining with directly conjugated mAbs. The

Microarray Analysis
Bead level data from all hybridizations was background corrected, using default parameters of the RMA algorithm (Irizarry et al., 2003), and summarized using the beadarray package for the Bioconductor suite of software for the R statistical programming environment. Processed sample expression profiles were quantile normalized using the limma package for Bioconductor prior to analysis of differential regulation between sample groups with the moderated t-statistic of the same package (Wettenhall and Smyth, 2004). In order to reduce errors associated with multiple hypothesis testing on such a scale, the significance p-values obtained were converted to corrected q-values using the false discovery rate (FDR) method of (Storey & Tibshirani, 2003), as implemented in the qvalue package for Bioconductor. Differential regulation between two sample groups was deemed significant at a threshold of q < 0.01 (FDR 1%). Heatmaps of gene expression were created by importing subsets of processed microarray data (for details see Supplementary Information) into the Java TreeView package (Saldanha, 2004).