Embryonic induction was first described by Hans Spemann in the early 20th century. In his most famous experiment, Spemann grafted the dorsal blastopore lip from a gastrula-staged amphibian embryo onto the ventral side of a differently pigmented host and found that the small graft itself gave rise only to a bit of notochord; however, it induced the surrounding host tissue to form a second axis with a well-patterned neural tube and axial mesoderm. The central role for cell–cell signalling during embryonic development was thus established. Further studies describing the capacity of the vegetal hemisphere of urodele embryos to induce mesoderm in the adjacent marginal zone were carried out by Peter Nieuwkoop in the 1960s and 1970s. He showed that animal pole explants taken from a blastula-stage embryo (“animal caps”) cultured on their own give rise to surface ectoderm, however, when combined with vegetal cells will form mesoderm. The mesoderm is derived solely from the animal cap cells and not the vegetal cells, indicating that the vegetal cells were providing a signal to change the fate of the responding animal cap cells to mesoderm. A model well supported by embryological experiments is the three-signal model presented by Jonathan Slack (Slack et al., 1987) and is shown in Figure 5. In this model, there are two signals from the vegetal pole, a ventral signal (VV) and a dorsal signal (DV), which respectively induces extreme ventral and extreme dorsal mesoderm in the overlying marginal zone during blastula stages. A third, dorsalising signal (O) is derived from the dorsal marginal zone (Spemann’s organiser), which imparts the full range of mesodermal fates along the dorsoventral axis of the mesoderm during gastrula stages. Although the demonstration that this third interaction is mediated by the interaction of opposing gradients of BMP ligands and BMP antagonists perhaps suggests that early induction and patterning of the mesoderm is best described by a four-signal model.
Two important papers in 1987 demonstrated for the first time that a purified growth factor could mimic the mesoderm-inducing activity of the vegetal pole. Slack et al. (1987) purified bFGF from bovine brain and used it to induce mesoderm in animal caps. They found that low concentrations of bFGF (FGF2) induced mesoderm that had the same character as mesoderm induced by the VV signal. This work also showed that the endogenous mesoderm-inducing signal emanating from vegetal cells could be blocked by heparin, supporting the notion that FGF, a known heparin binding protein, could possibly be one of the endogenous mesoderm-inducing signals. Kimelman and Kirschner (1987) provided further evidence for this by cloning bFGF from a Xenopus oocyte library and showing that it is expressed during amphibian development. Much more recent analyses have shown that in the Xenopus tropicalis genome, there are genes coding for 20 FGF ligands and at least 13 of these are expressed during embryonic development (Lea et al., 2009).
Clear evidence that FGF signalling plays a role in mesoderm formation was presented in another landmark paper from the Kirschner lab (Amaya et al., 1991) where an FGFR lacking its intracellular kinase domain was overexpressed in Xenopus embryos. Because FGF signalling requires receptor dimerisation and subsequent cross-phosphorylation of tyrosine residues, mutant receptors were able to partner with the normal endogenous receptors, providing a true dominant negative method to block all FGF signalling. This resulted in the loss of much of the mesoderm (Amaya et al., 1991). Further work using this powerful tool demonstrated that there is a subset of mesodermal genes that absolutely require FGF signalling for their expression during gastrula stages (Amaya et al., 1993; Isaacs et al., 1994; Schulte-Merker and Smith, 1995), and notable among these genes is the pan-mesodermal marker Xenopus brachyury (Xbra) (Smith et al., 1991). An essential role for FGF signalling in mesoderm formation and maintaining mesodermal gene expression has also been described in mammals, birds, and fish (Ciruna and Rossant, 2001; Griffin et al., 1995; Mathieu et al., 2004).
Although these studies indicate an essential role for FGF in mesoderm formation, there is no good evidence that FGF plays a role as an endogenous, vegetally localised mesoderm-inducing factor as had been envisaged by the early papers described above (Kimelman and Kirschner, 1987; Slack et al., 1987). The data on the whole support a model where FGF signalling is required in the marginal zone for it to respond to the vegetally localised mesoderm-inducing factors (see Isaacs, 1997). The endogenous mesoderm-inducing factors are widely accepted to be members of the TFGβ family of signalling molecules, Xenopus nodal related (Xnr1, 2, and 4) (Agius et al., 2000; Kofron et al., 1999) and Vg1 (Birsoy et al., 2006; Weeks and Melton, 1987). The signal transduction pathway downstream of Xnrs and Vg1 is often activated in experiments using the commercially available ligand activin. Using this approach, mesoderm induction by activin was found to require a functional FGF signalling pathway. When FGF signal transduction is inhibited using the dominant negative FGFR (or mutant forms of Ras or Raf), mesoderm induction by activin is blocked and a subset of mesodermal genes is not expressed (Cornell and Kimelman, 1994; LaBonne and Whitman, 1994).
These results indicate that FGF is required in the response to mesoderm induction and, indeed, the earliest activation of dpERK can be visualised in blastula-stage embryos in the dorsal marginal zone and animal hemisphere (Branney et al, 2009); this is the tissue responding to mesoderm induction at this time. Later, the activation of dpERK can be visualised in the ring of mesoderm around the closing blastopore in Xenopus and in the germ ring during epiboly in zebrafish (Figure 5). While other factors signal through RTKs to activate dpERK, this early MAPK activity has been shown to be completely dependent on FGF signalling (Christen and Slack, 1999), so Figure 6 gives a good picture of FGF activity in the mesoderm. The expression of the ligands Fgf3, Fgf4, Fgf8, and Fgf20 is found in the early mesoderm (Branney et al., 2009; Christen and Slack, 1997; Isaacs et al., 1995; Lea et al., 2009; Lombardo et al., 1998) and Fgf4 has been shown to be expressed as a direct response to activin, even in the presence of protein synthesis inhibitors (Fisher, 2002). These data point to a model for mesoderm induction where Nodal signalling from the vegetal pole induces mesoderm; and among the earliest genes activated in the nascent mesoderm are Fgf4 and Fgf8, which are required to activate and maintain the expression of a subset of genes important in specifying mesoderm identity (Fletcher and Harland, 2008).
Brachyury is a T-box transcription factor that is transcribed in response to FGF signalling in the absence of protein synthesis indicating that it is a direct and early response. In addition, Fgf4 can be transcriptionally activated by Xbra itself, while Xbra expression requires FGF. This points to an autocatalytic regulatory loop in the early mesoderm where FGF signalling induces the expression of Xbra, which in turn feeds back to maintain the expression of Fgf4 (Isaacs et al., 1994; Schulte-Merker and Smith, 1995). More recently, the initial activation of Xbra has been shown to require FGF signalling: the earliest expression of Xbra is lost in embryos treated with SU5402. However, the expression of FGF4 and FGF8 does not fully depend on the presence of Xbra and are most likely activated by Nodal signalling (Fletcher and Harland, 2008).
FGF8 is expressed when the zygotic genome is activated at the mid-blastula transition in Xenopus and is present in two alternatively spliced forms, FGF8a and FGF8b (Branney et al., 2009; Fletcher et al., 2006). FGF8a was shown to have little mesoderm-inducing activity (Christen and Slack, 1997) and thought to be more important for neural patterning (Hardcastle et al., 2000). By generating an antisense morpholino oligo that specifically targeted FGF8a, Fletcher et al. (2006) showed that FGF8a is essential for posterior neural cell fate and that FGF8b is the major isoform important for mesoderm formation.
Another important transcriptional target of FGF signalling is Egr1 (early growth response 1) (Branney et al., 2009) that codes for a zinc finger transcription factor and, like Xbra, is expressed throughout the marginal zone (Panitz et al., 1998). A detailed study into the regulation of Egr1 has shown that the ETS-box transcription factor Elk-1 is phosphorylated by MAPK and interacts with SRF to form a complex that binds to regulatory sequences upstream of Egr1 and, in this way, FGF signalling directly activates the transcription of Egr1 (Nentwich et al., 2009). Egr1 itself represses the transcription of Xbra, but activates the expression of MyoD (Nentwich et al., 2009), another immediate early target of FGF signalling (Fisher, 2002). Nentwich et al. (2009) demonstrate that this effector of FGF signalling, Egr1, can act to promote transcription of one gene and repress transcription of another gene, both of which are known to be positively regulated by FGF signalling. These data point to the complexity of the gene networks downstream of signalling that manage to orchestrate cell fate decisions during development.
A detailed picture of FGF-dependent mesodermal gene expression is now available. A large-scale analysis of genes normally expressed in early gastrula-stage embryos was compared to those expressed in sibling embryos in which FGF signalling was inhibited using the dominant negative FGFR (Branney et al., 2009). Another study compared gene expression in control mesoderm versus that in mesoderm in which FGF signalling was inhibited with the FGFR inhibitory drug SU5402 (Chung et al., 2004). These transcriptomic approaches have identified many genes already known to depend on FGF, such as Xbra and Cdx4, which internally validated the analysis, as well as a large number of new genes including a novel MAP kinase phosphotase (MKP) called DUSP5 (Branney et al., 2009). It is clear from these analyses that a subset of the genes transcriptionally activated by FGF signalling are those coding for factors known to down-regulate FGF signalling, such as the MKPs and the Sproutys. Transcriptionally activating negative regulators of FGF signalling may be a mechanism to balance the positive feedback loop working through Xbra, which is also a transcriptional target of FGF signalling. These findings suggest that both positive and negative feedback loops act to fine-tune the level of FGF signalling during development (Branney et al., 2009).
In addition, Branney et al. (2009) identified a set of genes expressed in the Organiser as targets of FGF signalling, including noggin, chordin, and goosecoid (among others). The notion that FGFs induce Organiser genes is in stark contrast with the original view of FGFs acting as part of the ventral vegetal (VV) signal in the three-signal model (Figure 5). However, in support of the microarray data, these genes were also found to be sensitive to inhibition of FGF signalling in other studies using SU5402 (Delaune et al., 2005; Fletcher and Harland, 2008). In addition, the level of pSMAD1 is significantly increased in embryos in which FGF signalling is blocked; this indicates an increase in BMP signalling in these embryos and is consistent with the down-regulation of genes expressed in the Organiser that code for BMP inhibitors (Branney et al., 2009).