A discrete period of FGF-induced Erk1/2 signalling is required for vertebrate neural specification

Neural specification is a fundamental developmental process during which cells embark on the neural differentiation programme. This is indicated by the onset of expression of genes characteristic of neuroepithelium, which include members of the SoxB family of transcription factors(Pevny et al., 1998; Rex et al., 1997; Wood and Episkopou, 1999). There is growing evidence that FGF signalling is a conserved initiator of vertebrate neural development (Bertrand et al., 2003; Delaune et al.,2005; Launay et al.,1996; Streit et al.,2000; Wilson et al.,2000); however, the molecular mechanism underlying this step has yet to be addressed in higher vertebrates. Activation of FGF receptors (FGFRs)can initiate transduction via three major intracellular pathways: classical MAPK (Erk1/2; also known as Mapk3 and Mapk1, respectively), PI3K (also known as Pik3r1 - Mouse Genome Informatics) and phospholipase C gamma (PLCγ;Plcg), the last two of which can activate protein kinase C proteins (PKCs),which can in turn stimulate Erk1/2 signalling(Schonwasser et al., 1998). There is piecemeal evidence that implicates all three of these pathways in the induction of neural tissue in vertebrate embryos. Most of this comes from work in the frog embryo, in which signalling via Ras (upstream of Erk1/2 and PI3K)is required for the induction of posterior neural tissue(Delaune et al., 2005; Ribisi, Jr et al., 2000),whereas PKC activators can also turn on neural genes via an unknown mechanism(Otte et al., 1988).

To monitor the neural specification process in ES cells, we differentiate the 46C line under defined, serum-free, monolayer culture conditions dependent on FGF signalling (Ying and Smith,2003; Ying et al.,2003). In 46C ES cells, the enhanced green fluorescent protein(EGFP) reporter is targeted to the endogenous Sox1 locus, which drives expression in neural progenitors and allows rapid and accurate quantification of the differentiation by fluorescence microscopy or flow cytometry (Ying et al., 2003). We cross-reference our in vitro findings in vivo using the chick embryo, a well-studied model organism for the process of neural induction in higher vertebrates (Stern, 2005a; Streit and Stern, 1999). Here,early (preneural) genes that identify potential neural tissue include a different SoxB gene, Sox3, whereas, in this vertebrate embryo, the later-expressed Sox2 gene is a marker of definitive neural tissue(Rex et al., 1997).

ES cells were grown on gelatine-coated plastics (Greiner and Nunc) in Glasgow modified Eagle's medium (GMEM) containing 10% foetal calf serum, 0.1%modified Eagle's medium (MEM) non-essential amino acids, 2 mM glutamine (all Life Technologies), 0.1 mM 2-Mercaptoethanol (Sigma) and 100 Units/ml recombinant human leukaemia inhibitory factor (LIF; prepared in-house). Differentiation was performed as previously described(Ying and Smith, 2003; Ying et al., 2003) and media was changed every 2 days in all experiments unless stated otherwise. PD173074(kindly provided by Pfizer) was used at 250 nM and PD184352 (a gift from Philip Cohen, Medical Research Council Protein Phosphorylation Unit,University of Dundee, UK) at 3 μM. For Fig. 1C, 46C cells were differentiated in either N2B27 or N2B27 that also contained: 3 μM PD184352;0.1 μg/ml noggin/Fc chimera (R&D Systems); PD184352 plus noggin; 10 ng/ml Bmp4 (R&D Systems); or Bmp4 plus noggin, for 3 days before flow cytometric analysis.

Cells at the appropriate differentiation stage were trypsinised and transferred into flow cytometry tubes containing phosphate-buffered saline(PBS) +5% foetal calf serum. TO-PRO3 (Molecular Probes) was included to discriminate non-viable cells. For cell cycle analysis, cells were fixed in 70% ethanol, treated with RNase A and stained with propidium iodide.

Cells were differentiated in 9 cm plates in N2B27 media containing 1/5000 dimethylsulfoxide (DMSO; control) or the inhibitors shown(Fig. 1) and lysed daily. Total protein (10-30 μg) was used for western blotting (standard protocols) and blots were analysed with phospho-specific primary antibodies (Cell Signalling)at 1/1000 dilution in 2.5% bovine serum albumin (BSA) in Tris-buffered saline with 0.1% Tween-20 (TBST) and appropriate peroxidase-conjugated secondary antibodies (Jackson, 1/2500), followed by chemiluminescent detection. Rabbit anti-phospho-Smad1 (S214) antibody was from E. De Robertis (University of California, LA, USA) and used as previously described(Pera et al., 2003).

Chick embryo manipulations and in situ hybridisation were performed as described (Eblaghie et al.,2003). Anti-EGFP antibody was from Jackson (1/100).

Quantitative PCR (Q-PCR) was performed on a Bio-Rad iCycler with i-Script SYBR green (Bio-Rad) on cDNA synthesised (using Improm-II; Promega) from 1μg total RNA (Nucleospin RNA II; Macherey-Nagel). The primer sequences and precise conditions for Q-PCR are available on request.

We first established the activation status of the three main signalling cascades regulated by FGF using western blotting on lysates of 46C ES cells grown in the presence of serum and LIF, and of ES cells at different stages of serum-free monolayer differentiation (days 1-4). Over this time period, the majority of ES cells progressively differentiate to Sox1-EGFP-positive neural progenitors under the influence of autocrine and/or paracrine FGF(Fig. 1A)(Ying et al., 2003). After an initial drop in activation following overnight culture in serum-free N2B27 media, phospho-Erk1/2 and phospho-PKB(S473) (PKB is also known as Akt - Mouse Genome Informatics; a key mediator of PI3K signalling) levels increased as differentiation proceeded, while phospho-PLCγ1(Y783) (PLCγ1 is also known as Plcg1 - Mouse Genome Informatics) levels gradually decreased(Fig. 1B). This indicates that all three pathways are active during neural specification and may play a role in the initiation of this process.

In order to determine which pathways are activated in response to endogenous FGF during the period of neural specification, we employed the specific FGF receptor inhibitor PD173074(Mohammadi et al., 1998). 46C cells were grown under differentiation conditions in the presence of PD173074 and analysed daily by flow cytometry for Sox1-EGFP, as well as by western blotting. At a concentration sufficient to inhibit the emergence of Sox1-expressing cells (Fig. 1A), PD173074 prevented the activation of Erk1/2 with no inhibitory effect on the phosphorylation of PKB or PLCγ1(Fig. 1B). This result indicates that Erk1/2 is downstream of FGF signalling in this context and that this transduction cascade is likely to be responsible for initiating neural specification in murine ES cells. To test this hypothesis, we interfered specifically with Erk activity. In the presence of the specific MEK (Mdk)inhibitor PD184352 (Davies et al.,2000), Erk1/2 phosphorylation is lost in a dose-dependent manner(Fig. 1B and see Fig. S1 in the supplementary material) concurrently with a reduction in the proportion of Sox1-EGFP cells assayed by flow cytometry(Fig. 1A, data not shown) (see also Lowell et al., 2006; Kunath et al., 2007). A similar effect was observed when Erk1/2 activation was inhibited by overexpression of the Erk1/2 phosphatase MKP3 (also known as Dusp6 - Mouse Genome Informatics; see Fig. S1 in the supplementary material)(Eblaghie et al., 2003).

Previous work has revealed that Erk1/2 can act to inhibit BMP signalling(Kretzschmar et al., 1997; Kretzschmar et al., 1999; Pera et al., 2003) and that FGF signals lead to Bmp4 and Bmp7 repression in the embryo(Wilson et al., 2000). However, the negative effect of FGFR inhibition on neural specification of ES cells cannot be rescued by addition of the BMP inhibitors noggin and chordin(Ying et al., 2003),suggesting that it is not exerted by a derepression of this anti-neurogenic pathway. Experiments in the early chick embryo also suggest that FGF acts independently of BMP activity in neural induction (see Stern, 2005b). To determine whether Erk1/2 signalling acts by interfering with BMP signalling during neural specification of ES cells, we assessed whether the addition of noggin could prevent the PD184352-mediated reduction in Sox1 expression. Inhibition of Erk1/2 did not result in increased Smad C-terminal (activating)or reduced linker (inhibitory) (Kretzschmar et al., 1997; Pera et al.,2003; Sater et al.,2003) phosphorylation under monolayer differentiation conditions(Fig. 1D). Moreover, PD184352 treatment continued to suppress neural specification when Bmp4 signalling was exogenously blocked with noggin (Fig. 1C,D), so we conclude that FGF and Erk can act independently of attenuating BMP in this context. On the contrary, Bmp4 induced non-neural differentiation of ES cells is also inhibited by PD184352 (M.P.S. and K.G.S.,unpublished) (Kunath et al.,2007).

Because neural induction in the chick embryo is also dependent on FGF signalling (Stern, 2005b; Streit et al., 2000; Wilson et al., 2000), we hypothesised that if Erk signalling is part of a fundamental mechanism that mediates neural specification it must be required in the embryo as well as in ES cells. Next, we therefore interfered with Erk1/2 activity that is detected in the prospective neural plate at Hamburger and Hamilton (HH) stage 3-3+(Eblaghie et al., 2003; Lunn et al., 2007) by overexpressing MKP3. Strikingly, MKP3 caused a downregulation of the preneural marker Sox3 in transfected cells(Fig. 2A), indicating a conserved requirement for Erk1/2 signalling for neural specification.

The finding that the activation of the other two pathways (PI3K-PKB and PLCγ1) is not FGF dependent during ES cell neural specification does not exclude the possibility that they play a role in this process. In order to investigate a possible role of PDK1 (the kinase linking PI3K activity to the activation of PKB and several other kinases) in neural specification, we examined neural differentiation in PDK1-deficient ES cells(Williams et al., 2000). These cells can not activate PKB, p90RSK (also known as Rps6ka2), SGK or p70S6 kinase (also known as Rps6kb) (Williams et al., 2000) and have lower levels of at least six PKC isoforms(Balendran et al., 2000), but they can still activate Sox1 (Fig. 1D) and readily generated nestin-expressing neural progenitors following monolayer differentiation (see Fig. S1 in the supplementary material). Furthermore, beads soaked in the PI3K inhibitor LY294002 failed to interfere with Sox3 expression when implanted in the chick prospective neural plate (Fig. 2B). These findings are consistent with the phenotype of mice lacking PDK1 or with inactivating mutations in key PDK1 residues that mediate PKB or S6K activation, which have forebrain defects, but still all form some neural tissue (Lawlor et al.,2002; McManus et al.,2004). Taken together, these findings demonstrate that the PI3K pathway is not necessary for neural specification.

Activation of the Erk1/2 pathway has previously been linked to ES cell differentiation (e.g. Chen et al.,2006), and treatment of ES cells with the MEK inhibitor PD098059 has been shown to maintain the expression of the pluripotency marker Oct4(also known as Pou5f1 - Mouse Genome Informatics) under embryoid body differentiation conditions (Burdon et al.,1999). Furthermore, the morphology of 46C cells grown in N2B27+PD184352 for 3 days appeared similar to that of ES cells(Fig. 3C), so next we sought to determine whether treatment with PD184352 inhibits ES cell differentiation completely. ES cells were plated in N2B27+PD184352 for 2 days then analysed for the expression of the pluripotency determinant Nanog(Chambers et al., 2003; Mitsui et al., 2003) and the primitive-ectoderm marker Fgf5. ES cells express high levels of Nanog but do not express Fgf5; however, the latter is quickly upregulated during differentiation and is thought to be one of the earliest differentiation markers of ES cells(Rathjen et al., 1999),although its activation has not been shown to be an irreversible commitment step. Treatment with PD184352 during differentiation in N2B27 prevented the onset of Sox1 expression but could not inhibit the transcription of Fgf5, as assayed by quantitative (QPCR-PCR; Fig. 3A) or immunocytochemistry(see Fig. S1 in the supplementary material) and had no significant effect on the cell cycle profile (see Fig. S2 in the supplementary material). Expression of Nanog (Fig. 3A) was also maintained, suggesting that, in the absence of Erk1/2 signalling, ES cells cannot progress past a primitive-ectoderm-like stage [this is also observed in Erk2-null ES cells(Saba-El-Leil et al., 2003; Kunath et al., 2007)]. The effect of PD184352 is readily reversible; when the inhibitor-containing medium was removed 2 days after plating and replaced with fresh media the cells resumed their differentiation and expressed Sox1-EGFP within the subsequent 24 hours (Fig. 3B and data not shown). Taken together, these results suggest that ES cell differentiation(beyond the upregulation of Fgf5) requires Erk signalling.

Next, we sought to determine the duration of Erk1/2 signalling required for the progression of neural differentiation. The addition of PD184352 to ES cells at the time of plating robustly blocked the normal emergence of Sox1-EGFP-expressing cells after 3 days of culture, whereas application of the inhibitor at 17 or 24 hours after plating resulted in a substantial proportion of Sox1-EGFP-positive cells at the same time point(Fig. 3C and data not shown). This suggests that differentiation of at least some of the ES cells commences during this time and that a relatively short exposure to Erk signalling is sufficient for neural specification.

In the chicken embryo, implantation of PD184352-soaked beads into the prospective neural plate at HH stage 3 resulted in the downregulation of the Erk1/2-target Ets genes Erm (also known as Etv5) and Pea3 (also known as Etv4) and of the preneural marker Sox3 (similar to the results observed when MKP3 was overexpressed; Fig. 2), whereas a border marker, Dlx5, was locally upregulated. Exposure to PD184352 at this stage also prevented the onset of Sox2 expression(Fig. 2B). This suggests that Erk1/2 signalling is required for the maintenance of an early preneural state and for the progression to a committed neural state, marked by the continued expression of Sox3 and the onset of Sox2 expression. However, when implanted at HH stage 4 (late primitive streak stage),PD184352-soaked beads no longer inhibited Sox3 or Sox2expression (Fig. 2B, right),identifying an early and discrete requirement for Erk signalling in neural specification. This falls within the period during which the chick epiblast is competent to respond to endogenous neural-inducing signals(Storey et al., 1992; Streit et al., 1997) and prompted us to characterise further the apparently analogous period of Erk activity required for neural specification in ES cells.

To refine the window of Erk1/2 requirement during ES cell differentiation,we decided to allow the cells to progress to the epiblast-like Fgf5-expressing state in the presence of PD184352 and then determine which period of Erk1/2 activity is required to commit these cells to their differentiation programme. To this end we plated 46C ES cells in N2B27+PD184352 and allowed them to differentiate for 2 days, by which time cells expressed Fgf5 and Nanog but had negligible levels of Sox1-EGFP (Fig. 3A). The media was then removed and, after two washes, replaced with media with or without inhibitor for varying lengths of time (see Fig. 4B,C). At the end of this period, cells were washed again and either lysed for western blotting analysis or incubated in fresh media plus PD184352 for 15 hours before flow cytometry analysis for Sox1-EGFP (Fig. 4A). Western blotting analysis shows that phosphorylated Erk1/2 first appeared around 5 minutes after PD184352 withdrawal, with levels peaking around 30 minutes after its removal and remaining elevated for a further 3 hours before dropping to background levels by 9 hours(Fig. 4B), indicating that the inhibitor was washed off the cells efficiently. Flow cytometry analysis showed that a period of 9 hours in the absence of PD184352 resulted in a significantly higher proportion of Sox1-EGFP-positive cells compared with inhibitor-treated controls (Fig. 4C). This effect decreased with shorter periods of incubation in the absence of inhibitor, but a statistically significant increase in the proportion of Sox1-EGFP-positive cells compared to the N2B27+PD184352 condition was maintained until the 1 hour time-point. Taken together, these results indicate that Erk1/2 activity sustained over a period of approximately 1 hour is sufficient to drive neural specification in ES cell-derived primitive-ectoderm cells.

In summary, our experiments define a conserved period during cell fate determination in mouse ES cells and in the chick embryo when Erk1/2, but not PKB, activity is required for neural specification. Additionally, we show that, in ES cells, this requirement is independent of BMP signal transduction and determine the minimum duration of sustained Erk1/2 activity required to control the transition to a committed neural fate.

This work was supported by a MRC Collaborative Career Development Fellowship in Stem Cell Research funded by the BBSRC (G113/18; M.P.S.), a BBSRC studentship (J.S.L.) and a grant from the MRC to K.G.S. (G0301013;K.G.S., B.J.C.). We would like to thank Pam Halley for expert technical help with in situ hybridisation; Pfizer for PD173074; P. Cohen for PD184352; Eddy De Robertis for phospho-Smad1 (S214) antibody; Austin Smith and Tilo Kunath for discussion of unpublished data; and Simon Arthur for comments on the manuscript.