Human embryos surplus to IVF requirements were donated to research after informed patient consent, with approval of local ethics committees and the UK Human Fertilisation and Embryology Authority (Research licence R0178). Embryos were obtained from Bourn Hall Clinic, Cambridgeshire. Embryos were thawed using EmbroThawTM Kit (EMF40_T, Fertipro) according to the manufacturer's instructions. Embryos were cultured to day 3 in EmbryoAssist™ (12140010, Origio), after which they were changed to BlastAssist® (12160010, Origio) until the appearance of a cavity, upon which they were transferred to N2B27 medium (SCS-SF-NB-02, Stem Cell Sciences) (^{Nichols and Ying, 2006; Ying and Smith, 2003}) for further culture. Embryos were cultured in 50 μl drops of medium under mineral oil (MINOIL500, Fertipro) and changed to fresh medium every 2 days. Embryos were cultured in a humidified atmosphere at 5% oxygen, 7% carbon dioxide and 37 °C. Embryos were exposed to inhibitors from day 3 of development.

Culture medium was supplemented with inhibitors in the combinations specified in the text: PD0325901, a Mek inhibitor with an IC₅₀ value in the 20–50 nM range in human cells (^{Ciuffreda et al., 2009}), 0.5 μM or 1 μM; Chir99021, a GSK3β inhibitor with an IC₅₀ value of 10 nM in human cells (^{Ring et al., 2003}), 3 μM (both synthesised in the Division of Signal Transduction Therapy, University of Dundee, Dundee, UK) and PD173074 (Sigma), a pan-FGF receptor inhibitor with an IC₅₀ value in the range of 5–22 nM for human FGFR1–5 (^{Byron et al., 2008; Grand et al., 2004; Mohammadi et al., 1998; Skaper et al., 2000; St Bernard et al., 2005; Stavridis et al., 2007}), 100 nM.

The strains of mice used in this study were MF1 outbred, and F1 hybrids between C57BL/6/Ola and CBA/Ca. The strain of rat used was Sprague–Dawley (SD). Animals were maintained by in-house breeding on a lighting regime of 14 hours light and 10 hours darkness with food and water supplied ad libitum. Prior to caging with stud males, female mice were oestrous selected by visual inspection of the vagina (^{Champlin et al., 1973}). Female rats were selected for oestrus using a vaginal impedance monitor (Muromachi Kikai, Tokyo, Japan, MK-11)(^{Koto et al., 1987}). Detection of a copulation plug the following morning was used to confirm successful mating; the resulting embryos were then considered to be E0.5. For all experiments, the embryos from at least two females were pooled and randomly assigned to experimental groups. Mouse and rat embryos were flushed from oviducts using M2 medium at E2.5 and E3.5 respectively. Mouse embryos were cultured in BlastAssist® for 2 days before being moved to N2B27 as expanded blastocysts. Rat embryos were cultured in BlastAssist® for 1 day, after which they were switched to N2B27. Embryos were cultured at 5% oxygen, 7% carbon dioxide and 37 °C.

Embryos were processed for immunohistochemistry as described previously (^{Nichols et al., 2009}). Briefly, they were fixed in 4% paraformaldehyde in PBS for 15 minutes, then rinsed in PBS containing 3 mg/ml polyvinylpyrolidone (PBS/PVP; P0930, Sigma), permeabilised in PBS/PVP containing 0.25% Triton X-100 (23,472-9, Sigma) for 30 minutes and blocked in blocking buffer, which comprised PBS containing 0.1% BSA, 0.01% Tween 20 (P1379, Sigma) and 2% donkey serum. Primary antibodies were Oct4 (C-10 sc-5279, Santa Cruz Biotech, Santa Cruz, CA, USA), Nanog (ab21624 or ab21603, Abcam, Cambridge, UK), Gata4 (C-20 sc-1237, Santa Cruz), Gata6 (R&D Systems, AF1700) and Sox17 (R&D Systems, AF1924). Antibodies were diluted 1:100 (Oct-4 was used at 1:200) in blocking buffer and embryos were incubated in the appropriate antibody solution at 4 °C overnight. They were rinsed three times in blocking buffer for 15 minutes each, and incubated in secondary antibody solution for 1 hour at room temperature. Secondary antibodies raised in Donkey labelled with Alexa fluorophores (Invitrogen, Paisley, UK) were diluted 1:500 in blocking buffer (anti-mouse IgG_2b for Oct4:647; anti-rabbit IgG for Nanog:488; anti-goat IgG for Gata4, Gata6 and Sox17:555). Embryos were then rinsed three times in blocking buffer, incubated briefly in increasing concentrations of Vectashield (H-1200, Vector Labs, Peterborough, UK) before mounting on glass slides in small drops of concentrated Vectashield with DAPI, and subsequently sealed with nail varnish.

Embryos were imaged on either a Leica TCS SP5 confocal microscope or Revolution XD Confocal System (Andor). Reconstructions of three-dimensional images from sections and cell counts were performed using Leica or Imaris software.

Probability (P) values were established using Student's t-test for comparison between two samples.

Human embryos were previously thawed and cultured in standard IVF medium to the blastocyst stage (day 6–7) at 20% oxygen. Blastocyst formation rates were low (17.5%, 25/143) and many blastocysts did not progress beyond the initial cavitation stages. The use of 5% oxygen has been reported to be beneficial for blastocyst formation rates (^{Dumoulin et al., 1999; Waldenstrom et al., 2009}). By reducing the concentration of oxygen in the culture regime to 5%, our blastocyst formation rates increased significantly (35.1%, 67/191), with many embryos progressing to day 7 and older. Therefore, for all subsequent experiments a humidified atmosphere of 5% oxygen, 7% carbon dioxide and 37 °C was adopted.

The interplay between Nanog and Gata6 is suggested to be crucial for the epiblast/hypoblast fate decision in mouse embryos (^{Plusa et al., 2008}). Although Nanog protein has been shown to localise to a sub-population of cells in the ICM of human embryos at the mid/late blastocyst stage (^{Cauffman et al., 2009; Hyslop et al., 2005; Kimber et al., 2008}), the identity of the Nanog-negative ICM cells has not yet been investigated. We therefore investigated the expression of the hypoblast markers Gata6, Gata4 and Sox17 in human blastocysts.

We first examined human embryos at 6 days post-fertilisation. Nanog is confined to a few inner cells within the embryo (Fig. 1A). Gata6 and Oct4 are widely expressed throughout the embryo, with levels varying between individual cells, consistent with previous data (^{Cauffman et al., 2005; Cauffman et al., 2006; Chen et al., 2009; Kimber et al., 2008}). Some inner cells exhibit high levels of Nanog and low levels of Gata6 (Fig. 1B, arrows), whilst in other cells both markers are expressed at about the same level (Fig. 1B, arrowheads). This pattern is similar to that observed in the early mouse blastocyst (^{Plusa et al., 2008}).

At day 7 of development, Gata4 and Sox17, both markers of differentiated hypoblast, are restricted to a narrow subset of cells within the embryo (Fig. 1C). Significantly, at this stage of development, this pattern is mutually exclusive with Nanog expression. Some embryos show Gata4 or Sox17-positive cells in what appears to be an epithelial layer overlying the Nanog-positive epiblast on the blastocoelic surface (Figs. 1C and 2C). This mirrors delamination of the hypoblast seen in rodent blastocysts. Oct4 protein is much more restricted to cells of the ICM than in earlier blastocysts, with staining in both the epiblast and hypoblast (Fig. 1C), as is the case with early murine hypoblast (^{Silva et al., 2009}). This may be due to slight differences in the developmental age of the embryos, most likely reflecting variability in their development in vitro. These embryos tended to exhibit reduced total cell number, consistent with this (Fig. 1C, D). These observations suggest that the human embryo at day 7 resembles the mouse embryo at E4.5 when all three embryonic lineages can be distinguished.

FGF/Mek inhibition in mouse preimplantation embryos has a striking effect on lineage segregation (^{Chazaud et al., 2006; Nichols et al., 2009; Yamanaka et al., 2010}). Inhibition of FGF signalling diverts all cells of the ICM to the epiblast fate, bypassing the hypoblast. We were interested in investigating whether the formation of human hypoblast is also dependent on FGF signalling. If so, this would indicate that the segregation of these lineages may be based on a conserved mechanism. We performed experiments with human embryos, adding inhibitors from the 6–8 cell stage. 1 μM PD0325901 is effective to block the formation of hypoblast in murine embryos if applied before its segregation (^{Nichols et al., 2009}). However, all human embryos treated with 1 μM PD0325901 that developed to late blastocysts showed Gata4 expression in a subset of cells, separate from the Nanog-expressing population (Fig. 2A, D). To test whether an alternative downstream pathway for FGF signalling is adopted during human hypoblast induction, we introduced an inhibitor of the FGF receptor, PD173074, in combination with PD0325901. Using this combination, Gata4-positive cells were still identified in all the human embryos investigated (Fig. 2C, D). We examined whether PD0325901 could work synergistically with Chir99021, since these are the components of 2i that effectively maintain naïve pluripotency in cultures of murine ES cells (^{Ying et al., 2008}). Gata4-positive cells were still observed in embryos developed under these conditions (Fig. 2B, D). In all three conditions that target FGF/Erk signalling, human embryos possess Gata4 and Nanog-positive cells that are exclusive to one another. Thus, the segregation of epiblast and hypoblast within the human embryo appears not to be dependent on FGF signalling.

To eliminate the human-specific culture regime as a potential variable in the response of embryos to FGF signalling inhibition, we cultured murine embryos from the 8-cell-stage under identical conditions. Mouse embryos responded as previously reported to 1 μM PD0325901 and 2i (^{Nichols et al., 2009}) (Fig. 2E, G). We also used the combination of PD0325901 and PD173074 that has been shown to block hypoblast formation (^{Yamanaka et al., 2010}). This also prevented formation of Gata4 positive cells. Finally, we cultured rat embryos under these conditions and found very similar responses (Fig. 2F, G). This indicates that the effect of FGF inhibition is not specific to the mouse.

In both mouse and rat embryos, 1 μM PD0325901 reproducibly blocks the formation of the hypoblast. The addition of Chir99021 to PD0325901 (2i) seems to facilitate some hypoblast cells escaping this block, especially in the rat embryo (Fig. 2G). The combination of 0.5 μM PD0325901 and PD173074 reduces the emergence of hypoblast compared to control embryos, but is not as effective in blocking hypoblast as 1 μM PD0325901. Despite the consistency of 1 μM PD0325901 in its effects on mouse and rat embryos, it did not eliminate the formation of Gata4-positive cells in the human embryo. However, it should be noted that culturing human embryos for extended periods in FGF signalling inhibitors resulted in no detectable detrimental effect on the Nanog positive population of cells.