Regulation of the ERK signalling pathway in the developing mouse blastocyst

Signalling between cells is crucial for tissue differentiation and is required during the earliest steps of embryo differentiation. During mouse preimplantation, three cell lineages are produced from two successive differentiation events. The first differentiation between outer and inner cells is essentially driven by the Hippo signalling pathway leading to the production of trophectoderm (TE) and inner cell mass (ICM) cells, respectively (Menchero et al., 2016; Nishioka et al., 2009). During blastocyst formation around embryonic day (E)3.5, epiblast (Epi) and primitive endoderm (PrE) cells appear within the ICM in a ‘salt-and-pepper’ pattern evidenced by the reciprocal and complementary expression pattern of the transcription factors NANOG and GATA6 (Chazaud et al., 2006; Guo et al., 2010; Kurimoto et al., 2006; Plusa et al., 2008). Numerous experiments have shown that the FGF/ERK pathway balances this binary cell fate, as high signals induce cells into PrE whereas blocking the pathway leads ICM cells toward an Epi fate (Bessonnard et al., 2017; Chazaud et al., 2006; Feldman et al., 1995; Kang et al., 2013, 2017; Krawchuk et al., 2013; Molotkov et al., 2017; Nichols et al., 2009; Ohnishi et al., 2014; Saiz et al., 2016; Yamanaka et al., 2010). Interestingly, FGF4, the sole FGF ligand detected at this stage (Ohnishi et al., 2014), is transcribed and secreted from Epi cells only (Frankenberg et al., 2011; Guo et al., 2010; Kurimoto et al., 2006; Ohnishi et al., 2014), converting neighbouring cells into PrE, illustrated by expression of the PrE markers SOX17 and GATA4 (Frankenberg et al., 2011). Fgfr2 is specifically expressed in PrE cells (Guo et al., 2010; Kurimoto et al., 2006; Ohnishi et al., 2014); however, Fgfr1, which is equally expressed in Epi and PrE cells, is the initial and main receptor transducing the FGF activity required for PrE specification (Kang et al., 2017; Molotkov et al., 2017).

Despite the increasing data demonstrating the importance of the FGF/ERK pathway during blastocyst development, expression of activated phosphorylated ERK (pERK) has not been described yet. FGF target genes have been recently examined using Fgfr1/2 mutant embryos (Kang et al., 2017). Although Dusp4 is preferentially detected in PrE cells, as would be expected with a pERK activity required in this cell type, other genes, such as Etv4/5, which are ETS-related factors, and Sprouty (Spry)2/4, which are negative regulators of the pathway, are expressed in both cell types (Kang et al., 2017). It was then proposed that a low pERK activity could be present also in Epi cells, where it would activate negative-feedback loops involving ETVs and SPRYs to protect the Epi state (Kang et al., 2017).

Here, we show that ERK phosphorylation begins around E3.0-E3.25 days, concomitant with the initiation of Epi/PrE specification, and is present in ICM progenitor or PrE cells. This activation of the pathway is respectively increased and lost in Klf5 and Fgf4 mutants, which is in accordance with the described phenotypes (Azami et al., 2017; Kang et al., 2013; Krawchuk et al., 2013; Ohnishi et al., 2014). We also found that DUSP4 accumulation tightly follows pERK activity, whereas, surprisingly, ETV5 accumulation is not directly regulated by the MAPK pathway but clearly depends on NANOG expression. Our results lead to a novel model of FGF/ERK regulation in the preimplantation embryo.

The FGF/ERK signalling pathway plays pivotal roles in the segregation of Epi and PrE cells, yet there is currently no reliable method to detect pERK during preimplantation stages, thereby hampering the understanding of how pERK regulates the fundamental process of preimplantation development. We have developed two new methods to detect ERK phosphorylation that are sufficiently sensitive to detect signals in preimplantation embryos. Briefly, to preserve the integrity of the phosphate conjugate we add phosphatase inhibitors during the fixation step and permeabilisation is carried out either with Triton X-100 or proteinase K (see Materials and Methods). Two working methods (Fig. S1A) provide alternative solutions to observe other labelling in co-staining experiments. Phosphorylation of ERK is evident in the cytoplasm and mitotic nuclei can retain a strong labelling (Fig. S1B,C) (Shapiro et al., 1998). Cytoplasmic localisation of pERK has been reported previously in early mouse embryos (Corson et al., 2003) and, interestingly, it was suggested that this subcellular labelling is reminiscent of differentiation rather than proliferation events (Marenda et al., 2006; Michailovici et al., 2014).

The presence of exogenous FGF has been shown to increase the number of PrE cells (Yamanaka et al., 2010); we found that culturing embryos with recombinant FGF between E2.5 and E3.75 increased the number of pERK-labelled cells as well as its signal intensity, with almost all ICM cells exhibiting an elevated pERK signal compared with the control (Fig. S2A). Interestingly, an increase of pERK staining could be observed after only 10 min of FGF treatment (Fig. S2B). To further validate these techniques, we collected E2.5 embryos and cultured them until E3.75 with MEK or FGFR+MEK inhibitors, which block the pathway upstream of ERK phosphorylation (Nichols et al., 2009; Yamanaka et al., 2010). The pERK signal was totally downregulated in the presence of inhibitors (Fig. S2C), confirming the specificity of our methods. Thus, we established for the first time two protocols to detect pERK during preimplantation development.

We analysed ERK phosphorylation during preimplantation development from the one-cell stage, thus before the reported Fgf4 expression that starts in a subset of the inner cells between E3.0 and E3.25 (Guo et al., 2010; Ohnishi et al., 2014). Cells expressing pERK were first detected around E3.0 (Fig. 1A). The number of labelled cells increased with time, exhibiting graded pERK levels (Fig. 1A-C). A few TE cells were marked (Fig. S1B) and this is consistent with required pERK activity for TE proliferation (Nichols et al., 2009). Then, we determined the identity of pERK-labelled ICM cells by examining NANOG and GATA6 levels at each developmental stage (Fig. 1B,D). Cell counts indicated that very few cells were labelled between E3.0 and E3.25 (Fig. 1C), around two to three per embryo, with a relatively weak staining that was present in uncommitted cells that are doubly positive for NANOG and GATA6 (Bessonnard et al., 2014, 2017; Plusa et al., 2008; Saiz et al., 2016) (Fig. 1D). Subsequently, the number of cells with a strong pERK staining gradually increased (Fig. 1C) and they progressively corresponded to PrE cells solely labelled by GATA6 (Fig. 1D). Thus, ERK phosphorylation essentially occurs in PrE-specified cells as well as in a subset of uncommitted cells, implying induction of their specification into PrE cells. However, not all PrE cells were labelled by pERK whereas all PrE cells require FGF/ERK activity for their specification (Chazaud et al., 2006; Kang et al., 2017; Molotkov et al., 2017; Nichols et al., 2009; Yamanaka et al., 2010). This low frequency of labelled cells may reflect a transient nature of pERK labelling.

After cell specification into Epi and PrE, little is known about MAPK activity in the three cell lineages during their further differentiation. From E4.0, several FGF ligand genes are expressed in the Epi (Fgf4, Fgf5, Fgf8, Fgf15, Fgf18) and in the PrE (Fgf3, Fgf17), and can be transduced by Fgfr1, Fgfr2, Fgfr3 or Fgfr4 (Ohnishi et al., 2014). Moreover, PDGF signalling has been shown to be important for PrE cell survival (Artus et al., 2013) and could act through the MAPK pathway. Accordingly, pERK was detected in PrE cells at E4.5 (Fig. 1E) and their derivatives at E6.5 (Fig. S3A). Consistent with a previous report (Corson et al., 2003), strong pERK signals were detected in the extra-embryonic ectoderm of E6.5 embryos (Fig. S3A).

In FGF/ERK-impaired embryos, NANOG levels are maintained in the E4.5 Epi whereas they decay in wild-type (WT) embryos (Kang et al., 2013; Molotkov et al., 2017; Nichols et al., 2009). This suggests that FGF4 enables Epi maturation by downregulating NANOG levels. Accordingly, we could observe ERK phosphorylation in some Epi cells at E4.5 (Fig. 1E), most likely through the transduction of FGFR1, the only FGF receptor expressed in this tissue at this stage (Ohnishi et al., 2014). Similarly, mouse embryonic stem (ES) cells, which are closely related to the E4.5 Epi (Boroviak et al., 2014), express high levels of Fgf4, and the FGF pathway is necessary for their exit of pluripotency and differentiation (Hamilton and Brickman, 2014; Kalkan et al., 2017; Kunath et al., 2007; Molotkov et al., 2017; Schröter et al., 2015; Stavridis et al., 2010). Heterogeneous labelling was observed among the cells (Fig. S3B) (Deathridge et al., 2019; Hamilton and Brickman, 2014), possibly due to heterogeneous levels of NANOG between cells (Chambers et al., 2007) that are controlling Fgf4 secretion, and also owing to possible transient ERK phosphorylation. In the presence of the MEK inhibitor, pERK staining was strongly diminished in ES cells (Fig. S3B).

We thus report that the ERK pathway is active in the three cell lineages derived from Epi, PrE and TE.

Fgf4 is specifically expressed in Epi cells (Frankenberg et al., 2011; Guo et al., 2010; Kurimoto et al., 2006; Ohnishi et al., 2014) and acts as a paracrine factor to differentiate neighbouring cells (Frankenberg et al., 2011). Fgf4 targeted inactivation leads to an absence of PrE cells and pan-expression of Epi markers in the ICM (Kang et al., 2013; Krawchuk et al., 2013). ERK phosphorylation was not detected in Fgf4 knockout embryos (Fig. 2A) demonstrating that FGF4 is the sole secreted signal activating the ERK pathway at this stage of development.

Our previous results showed that Klf5 is essential for blastocyst development (Ema et al., 2008). Indeed, through direct binding to Fgf4 regulatory elements, KLF5 prevents precocious expression of Fgf4 at E3.0-E3.25 to initiate cell differentiation at the right time and to ensure a proper balance between Epi and PrE cells (Azami et al., 2017). Consistent with our previous study, Klf5^−/− embryos at E3.25 had a stronger pERK signal and in more cells compared with that of WT embryos (Fig. 2B).

We have previously shown (Frankenberg et al., 2011) that Fgf4 transcription in Epi cells requires the presence of NANOG (Fig. 2E). As expected, in Nanog mutant embryos, the lack of FGF4 led to an absence of pERK (Fig. 2C).

Removing one or two alleles of Gata6 increased the number of Fgf4-expressing cells (Fig. 2E) as more cells produce NANOG (Bessonnard et al., 2014). As a consequence, pERK staining could be observed in a higher number of ICM cells in Gata6^−/− embryos (Fig. 2D), owing either to transcriptional activation of Fgf4 by NANOG, or to a lack of GATA6 repression (Fig. 2E). It is interesting to note that ICM cells in Gata6^−/− embryos have an Epi identity and are nevertheless able to transduce FGF signalling through FGFR1, which is expressed in these cells (Fig. 2E). This ERK activity is reminiscent of what is found in E4.5 WT Epi cells or in ES cells.

ETVs and DUSP4 are considered as direct targets of the MAPK/ERK pathway (Neben et al., 2017; Patel and Shvartsman, 2018) and are present in the blastocyst (Guo et al., 2010; Kang et al., 2017; Kurimoto et al., 2006). We carried out single-cell RNA analyses, and we subdivided the samples between Fgf4-expressing (Fgf4⁺) and Fgf4-non expressing (Fgf4⁻) populations as Fgf4 was identified as the Epi marker with the earliest binary expression (Guo et al., 2010; Ohnishi et al., 2014). Transcripts of Dusp4 were barely detected at the 16-cell stage and later were preferentially expressed in Fgf4⁻ cells identified as PrE cells (Fig. 3A). By contrast, Etv5 expression was enriched in Fgf4⁺ Epi cells as early as E3.25 (Fig. 3A; see Fig. S4D for RNA-seq data analysis, data taken from Posfai et al., 2017). Etv4 expression started from E3.5, thus after the beginning of Epi/PrE specification (Fig. 3A, Fig. S4D). At E3.75, when the NANOG/GATA6 mutually exclusive pattern is established, Etv4 was enriched in Fgf4⁻ PrE cells whereas Etv5 was enriched in Fgf4⁺ Epi cells (Fig. 3A).

At E3.25, the protein DUSP4 was observed in some outside cells corresponding to TE (Fig. S4A) as reported for its RNA localisation (Posfai et al., 2017), consistent with an active FGF pathway to promote TE growth (Nichols et al., 2009). Later, DUSP4 localised to PrE cells (Fig. 3B,C) but could also be found in undifferentiated cells, labelled by both GATA6 and NANOG, especially at the beginning of blastocyst formation (Fig. 3C). DUSP4 could not be detected in Epi cells (Fig. 3C). Double labelling with pERK and DUSP4 antibodies showed a high correlation of expression (Fig. 3B). However, some pERK-only marked cells could be observed, probably reflecting an early activation of the pathway just preceding a visible DUSP4 translation. As well, DUSP4-only labelled cells were found, illustrating a plausible transient nature of ERK activation. As expected, the number of DUSP4-labelled cells was enhanced by FGF2 treatment and no signal could be detected upon MEK inhibitor administration (Fig. S4B).

In Nanog mutants, DUSP4 was poorly detected in the ICM (Fig. 3D), in accordance with the absence of Fgf4 in these embryos (Fig. 2E; Frankenberg et al., 2011). By contrast, DUSP4 labelling was observed in some cells in Gata6^−/− embryos (Fig. 3D). This result shows that DUSP4 does not need GATA6 presence to be expressed but relies essentially on pERK signalling activity. Alternatively, this expression in the Epi of Gata6^−/− embryos could correspond to cells exiting pluripotency, as ES cells in transition from naïve pluripotency express DUSP4 (Kalkan et al., 2017). Indeed, we have previously shown that Epi specification is more precocious in these mutants (Bessonnard et al., 2014), and this most likely leads also to an earlier Epi transition to a mature state.

ETV5 is an ETS-related transcription factor involved in the transduction of the MAPK pathway (Herriges et al., 2015; Sharrocks, 2001) and is expressed in the blastocyst ICM (Guo et al., 2010; Kang et al., 2017). By immunofluorescence, we showed that from E3.5 ETV5 is preferentially observed in Epi cells as it colocalises with NANOG mainly (Fig. 4A, Fig. S4C), confirming RNA expression data (Fig. 3A); however, some PrE cells also expressed the protein (Fig. S4C). Surprisingly, ETV5 labelling was lost upon FGF2 administration and was present in more cells upon administration of MEK and FGFR inhibitors (Fig. 4A). Thus, ETV5 does not behave like a MAPK signalling reporter but instead acts as an Epi marker. Strikingly, ETV5 levels were drastically downregulated in Nanog^−/− embryos and the protein was present throughout the ICM of Gata6^−/− embryos (Fig. 4B). Therefore, our data clearly indicate that ETV5 is not positively regulated by RTK signals but by the presence of NANOG during blastocyst formation.

FGF signalling, through FGFR and ERK, controls PrE differentiation in the ICM of the mouse blastocyst (Brewer et al., 2016; Chazaud et al., 2006; Feldman et al., 1995; Kang et al., 2013, 2017; Krawchuk et al., 2013; Molotkov et al., 2017; Nichols et al., 2009; Saiz et al., 2016; Yamanaka et al., 2010). However, the spatiotemporal expression of active, phosphorylated ERK has never been reported during mouse preimplantation development. We show here that ERK phosphorylation begins at the onset of Epi/PrE differentiation, around E3.25, when Fgf4 transcripts start to be elevated in some ICM cells (Ohnishi et al., 2014). Thus, as soon as Fgf4 is expressed at sufficient levels, ERK phosphorylation can be observed. This phosphorylation is present essentially in uncommitted cells, doubly labelled by NANOG and GATA6, and in PrE cells solely marked by GATA6. This is consistent with the known pERK activity of converting uncommitted cells into PrE (Bessonnard et al., 2017; Saiz et al., 2016) and our data show here that ERK phosphorylation can be seen in PrE cells up to at least E4.5. At this stage, pERK is also observed in some Epi cells, when they are differentiating further through a formative phase (Smith, 2017). Interestingly, NANOG is maintained in the ICM of Fgf4 and in Fgfr1/Fgfr2 mutants at E4.5 (Kang et al., 2013; Molotkov et al., 2017), indicating that Epi maturation requires FGF signalling, which is consistent with the observed ERK phosphorylation.

DUSP phosphatases have been shown to be transcriptional ERK targets in several models and inhibit the FGF pathway in a negative-feedback loop through dephosphorylation of ERK (Amit et al., 2007; Brewer et al., 2016; Chu et al., 1996; Eblaghie et al., 2003; Ekerot et al., 2008; Lemmon and Schlessinger, 2010; Niwa et al., 2007). We show that DUSP4 accumulation tightly follows ERK phosphorylation, and depends on it. In WT embryos, DUSP4 could be considered as a PrE marker; however, its expression in Gata6^−/− embryos, which do not contain PrE cells, indicates that this protein is linked to pERK activity rather than PrE identity and is thus a reliable ERK activity reporter. Co-labelling experiments of pERK with DUSP4 suggest transient activation of ERK whereas DUSP4 presence is more robust. Transient or sustained ERK activities have been observed downstream of RTK signalling to control different cellular responses. These differences in the output, despite sharing signalling intermediates, can be explained in part by differences in feedback mechanisms (Lemmon and Schlessinger, 2010). The rather transient nature of the pERK labelling observed during preimplantation suggests the presence of negative-feedback loops that could be mediated through DUSP4.

ETS-related molecules such as ETV4 and 5, are transcriptionally activated by the FGF pathway in several different developmental processes and tissues (Aulehla et al., 2008; Liu et al., 2003; Raible and Brand, 2001; Roehl and Nüsslein-Volhard, 2001). ETS factors are often involved in positive- or negative-feedback loops of ERK signalling (Herriges et al., 2015; Reginensi et al., 2011), being either activators or repressors of transcription and potentially regulating negative effectors of the pathway, such as SPROUTYs (SPRYs) (Hacohen et al., 1998; Herriges et al., 2015; Neben et al., 2017; Sharrocks, 2001). Etv4, Etv5 and Spry4 have been reported to be expressed in both Epi and PrE cells during blastocyst formation, whereas Spry2 is preferentially expressed in the Epi (Boroviak et al., 2015; Kang et al., 2017; Morgani et al., 2018). Moreover, Etv4 and Spry4 respond to FGF/ERK signalling (Kang et al., 2017; Morgani et al., 2018). It was then proposed that these ETVs and SPRYs would suppress or dampen ERK activity through a FGF negative-feedback loop, to desensitise Epi cells from their own FGF4 secretion (Kang et al., 2017). However, we found that pERK is almost undetectable in blastocyst Epi cells, implying either that Etv4 and Spry4 are only transcribed in precursor cells with a high RNA stability, or that these genes can respond to very low levels of pERK, or that their expression depends also on other factors. Data extraction from single-cell RNA-seq analyses (Posfai et al., 2017) show that only Spry4, but not Etv4, is expressed in precursor cells at the 16-cell stage (Fig. S4D). Thus, SPRY4, under FGF stimulation, could tame the FGF/ERK pathway at least in the precursor cells.

Spry2 is particularly interesting because its transcripts are differentially enriched between Epi and PrE cells (Boroviak et al., 2015; Kang et al., 2017). Indeed, FGF signalling must not be blocked in PrE cells. SPRY2 protein is present in ES cells (Christoforou et al., 2016), and, similar to its transcripts (Fig. S4), the levels of this protein increase between the morula and blastocyst stages (Gao et al., 2017). In Fgfr mutants, Spry2 expression is maintained at E3.5 and then downregulated by the late blastocyst stage (Kang et al., 2017), indicating that its initial expression is independent of FGF signalling. ETS-binding sites have been characterised in the human SPRY2 promoter (Ding et al., 2003), and they are conserved in the mouse. Thus, SPRY2 is a good candidate downstream of ETVs to block FGF signalling in the Epi, as early as the 16-cell/E3.25 stage (Fig. S4D); however, we cannot exclude the possibility that other negative factors, such as SPREDs or RSKs (Mühl et al., 2015; Neben et al., 2017; Nett et al., 2018), might be at play. We show here that Etv5 transcripts, like those of Spry2, and protein are enriched in EPI cells. Surprisingly, ETV5 levels are not positively regulated by the FGF/ERK pathway but depend on the presence of NANOG, likely through direct activation as NANOG binds to the Etv5 locus in ES cells (Boyer et al., 2005; Loh et al., 2006; Murakami et al., 2016). Thus, this reveals a novel potential mechanism by which NANOG activates the expression of FGF4 and ETV5 at the same time, on one hand to induce paracrine PrE cell differentiation, and on the other hand to protect Epi cells against autocrine/paracrine FGF4 activity. This mechanism of direct activation of ETV5 by NANOG would seem to be a more efficient way to protect the identity of Epi cells compared with activation of ETV5 through a negative-feedback loop found in other systems (Herriges et al., 2015). Indeed, ETV5 is already present when Epi cells are subjected to autocrine/paracrine FGF4 activity.

How ERK activity is controlled during Epi/PrE specification has remained unclear. With these novel data we propose (Fig. 4C) that in precursor cells of the early ICM, expressing both NANOG and GATA6, pERK activity remains very low or absent due to (1) KLF5 presence that limits the number of Fgf4 transcripts (Azami et al., 2017) and (2) possible inhibition of the FGF pathway by a negative-feedback mechanism through SPRY4 (Kang et al., 2017; Morgani et al., 2018). Then at the beginning of Epi specification, NANOG levels become predominant over GATA6 levels, enabling the transcription of both Fgf4 and Etv5. FGF4 is secreted and binds to FGFR1, which is equally expressed among ICM cells. In Epi cells, ETV5 is likely to protect the cells from the surrounding FGF4 by activating negative regulators of the FGF pathway, such as SPRY2, to attenuate or block the pERK signals. In PrE cells, NANOG is not expressed and therefore ETV5 is absent. This enables efficient FGF/ERK pathway activity, illustrated by DUSP4 expression, to differentiate the cells into PrE by maintaining a high expression of GATA6 and inducing downstream PrE factors such as SOX17 and GATA4.

In conclusion, our results show that NANOG is key to binary cell differentiation by activating within the same cell the differentiation paracrine cues and a protective cell-autonomous mechanism.

Preimplantation embryos obtained from natural matings [CD1 (ICR) background] were collected from oviducts or uteri by flushing with M2 medium (Sigma-Aldrich). Embryos were cultured in the presence of MEK inhibitor (PD0325901; 0.5 µM) (Selleck), FGFR inhibitor (PD173074; 100 nM) (Selleck), FGF2 (Cell Guidance Systems) or FGF4 (1 µg/ml) (R&D)+heparin(1 µg/ml) (Sigma-Aldrich) in KSOM+AA medium (Millipore) in 4-well plates at 37°C, 5% CO₂. All experiments were replicated at least three times.

Fgf4^tm1.2Mrt (Sun et al., 2002), obtained from MMRRC at UC Davis, were crossed with Ayu1-Cre mice, which express Cre recombinase from the zygote to generate the deletion of Fgf4 allele. Fgf4^+/− mice were then intercrossed to obtain Fgf4 mutant embryos. Nanog and Gata6 mutant embryos were obtained as previously described (Bessonnard et al., 2014; Frankenberg et al., 2011; Mitsui et al., 2003; Sodhi et al., 2006) through natural matings. Klf5 knockout embryos were generated as previously described (Azami et al., 2017). All mice were maintained on the CD1 (ICR) background. Embryo genotyping was performed after confocal imaging, as described previously (Azami et al., 2017; Bessonnard et al., 2014; Frankenberg et al., 2011). Experiments were performed in accordance with French/EU and Japanese guidelines for the care and use of laboratory animals. Experimental procedures were approved by the Animal Care and Use Committee of Shiga University of Medical Science and methods were carried out in accordance with the approved guidelines (approval number: 2016-11-8, 2018-10-1). All animal work was conducted according to French and European directives for the use and care of animals for research purposes and was approved by the local ethics committee, C2E2A (Comité d’Ethique pour l’Expérimentation Animale en Auvergne).

Immunostaining of embryos was performed as described previously (Azami et al., 2017). Briefly, embryos were fixed in 4% paraformaldehyde (PFA) in 1×PBS for 15-30 min at room temperature (RT). Permeabilisation was performed in 0.5% Triton X-100 in PBS for 15 min and embryos were incubated in blocking solution [1% donkey serum, 0.1% bovine serum albumin (BSA) and 0.01% Tween 20 in PBS] for 1 h at RT. Then, embryos were treated with primary antibodies in blocking solution overnight at 4°C. After washing with 0.1% Tween 20 in PBS three times, embryos were treated with secondary antibodies (made in donkey, Jackson Laboratories, 1/300) in blocking solution for 1 h at RT. Nuclei were stained with Hoechst 33342 (Thermo Fisher) or DAPI and cell membranes with phalloidin (1/1000, Jackson ImmunoResearch). To visualise DUSP4, a tyramide signal amplification (TSA) reaction was carried out according to manufacturer's instructions (Thermo Fisher or Perkin Elmer). Double immunostaining for pERK and DUSP4, e.g. with antibodies from the same host, was performed by a sequential immunostaining of pERK and then DUSP4. Briefly, after the TSA reaction with the first rabbit antibody, the horseradish peroxidase (HRP) was inactivated by a treatment of 5 min 3% H₂O₂ in PBS. Then, the second rabbit primary antibody staining was carried out as usual (as control, the second primary antibody was not added to ascertain the inactivation of the first HRP). Primary antibodies are listed in Table S1.

Visualisation of phosphorylated ERK (pERK) was performed following two protocols.

Embryos collected from uterus or oviducts were precultured in KSOM+AA medium at 37°C, 5% CO₂. They were fixed in 8% PFA together with phosphatases inhibitors: NaF (25 mM, Sigma-Aldrich), sodium orthovanadate (2 mM, NEB) and β-glycerophosphate (50 mM, Santa Cruz Biotechnology) for 10 min at RT. They were permeabilised 5 min in 0.5% Triton X-100/PBS and non-specific sites were blocked with 10% fetal bovine serum/0.1% Triton X-100/PBS for 30 min. Primary antibodies were added to the blocking solution overnight at 4°C. After three washes, secondary antibodies coupled to peroxidase were added and TSA reaction was carried out according to manufacturer's instructions (Thermo Fisher or Perkin Elmer).

Embryos collected from uterus or oviducts were precultured in KSOM+AA medium at 37°C, 5% CO₂ and fixed in 4% PFA with PhosSTOP (Roche) for 20 min on ice. Embryos were permeabilised and activated with 10 ng/ml proteinase K (ProK) (Roche) in 0.1% Tween 20/PBS at RT for 3-5 min, depending on the developmental stage. The ProK reaction was stopped in 2 mg/ml glycine-PBS for 2 min at RT and embryos were washed in 0.1% Tween 20/PBS. Blocking was performed in 0.5% donkey serum/0.5% casein/0.5% BSA in PBS for 30 min at RT. Then, embryos were treated with the primary antibody in blocking solution overnight at 4°C. After washing with 0.1% Tween 20 in PBS three times, embryos were treated with HRP-conjugated secondary antibody in blocking solution for 1 h at RT. TSA (Thermo Fisher) was used to detect pERK signal.

Leica TCS SP8, SP5 or SPE confocal microscopes were used to acquire the fluorescent images of immunostained embryos as described previously (Azami et al., 2017; Gasnier et al., 2013). Look up table (LUT) visualisation of fluorescence intensity was processed using Fiji. The ‘thermal’ range of LUT indicates high levels of expression in red and low levels in blue. Unless otherwise indicated, images are displayed as single z-sections representative of the embryo. Cell counts were performed as previously described (Bessonnard et al., 2014).

Isolated ICM cells (after immunosurgery; Nagy, 2003) and morulae (16°C) were incubated for 10 min in 1×TrypLE Express Enzyme (Gibco) at 37°C and cells were isolated by repeated mouth pipetting using pulled capillaries of serially smaller diameter openings. Each single cell was collected in 5 µl of 2× Reaction Mix (Invitrogen, CellsDirect One-Step qRT-PCR Kit) and stored at −80°C or processed immediately. Genotyping of embryos was carried out using TE cells from immunosurgery with whole genome amplification (WGA) using the REPLI-g kit (Qiagen) followed by PCR genotyping as above.

cDNA from the desired genotypes was preamplified with 18 cycles. Quality and genotype checks were performed by qPCR for each cell with housekeeping gene primers (Rps17 and Rpl30) and primers detecting the mutations. The analysed cells originate from at least three different embryos in each category.

Single-cell qPCR on a Fluidigm Biomark system (GENTYANE facility) was carried out on 48.48 or 96.96 Dynamic Arrays, according to the manufacturer's instructions. Cells with absent or low Ct values for the housekeeping genes Rps17 and Rpl30 were removed from the analysis (5%). Ct values were normalised with the mean of housekeeping genes Rps17 and Rpl30 using the 2-ΔCt method. In box plots, values are relative to the mean of all WT E3.25 cells to be able to compare between genotypes and stages. Cells were considered in Fgf4^– subpopulations when CT values were above 35 and all other cells were placed in Fgf4⁺ subpopulations. Primers used for this study are listed in Table S2.

For the box plots, the edges of the box represent 25th and 75th quartiles. The median is represented by the central line. The whiskers extend to 1.5 times the interquartile range (25th to 75th percentile). Cells are plotted individually in single-cell experiments.

Statistical tests were performed using Graphpad software. Statistical significance was assessed using the Wilcoxon-Mann–Whitney test (non-parametric) on expression levels. Unless otherwise indicated, the phenotypes are fully penetrant.

Images were acquired and treated at the CLIC imaging facility (Clermont-Fd). C.C. thanks C. Pléver for genotyping the mice.