Abstract

Neural induction represents the first stage in the formation of the vertebrate nervous system from embryonic ectoderm. Fibroblast Growth Factors (FGFs), initially identified for their mitogenic and angiogenic roles in bovine brain extracts, are now known to have many developmental roles in particular that of neural induction, comprising of a family of 22 FGFs.

Spemann and Mangold (1924) pioneered the study of neural induction through the identification of the organizer. Early work in amphibians suggested that neural fate was instructed by signals from Spemann’s organiser or dorsal mesoderm. Over a decade ago, the default model proposed that neural induction was the direct consequence from inhibition of bone morphogenetic proteins (BMPs) found in Xenopus laevis, not taking into consideration neural induction in avian embryos. Consequently many experimental studies, in the chick, subsequent to this finding conflicted the idea that BMP inhibition was the only necessary step required suggesting that FGFs were required at an earlier stage prior to BMP inhibition.

Much controversy has surrounded the role of FGFs in neural induction but now it is widely accepted to have a role in both amphibians and amniotes.

Fibroblast Growth Factors in neural induction

Structure and Function: FGFs broken down

Fibroblast Growth Factors (FGFs) regulate a vast array of developmental processes, including, limb development, neural induction and neural development (Böttcher and Niehrs, 2005). FGFs play an important role in development of an organism by regulating cellular differentiation, proliferation and migration and are involved in tissue-injury repair (Itoh and Ornitz, 2004). The early FGFs, FGF1 and FGF2 (also known as acidic and basic FGF, respectively) were first discovered from bovine brain and pituitary extracts and identified for their mitogenic and angiogenic activities (Gospodarowicz et al., 1974). Additionally, a number of family members were found revealing a total of 22 FGFs in humans ranging from 17 to 34 kDa in molecular mass in vertebrates. The nomenclature extends to FGF23 but in humans FGF19 is the equivalent to mouse Fgf15 (Ornitz and Itoh, 2001). Also the FGFs have been organised into seven subfamilies based on sequence comparisons.

FGFs show conservation through species, especially across the vertebrate species in gene structure and amino-acid sequence. FGF sequences are yet to be found in unicellular organisms such as yeast (Saccharomyces cerevisiae) and bacteria (Escherichia Coli) (Itoh and Ornitz, 2004). Interestingly, an Fgf-like gene has been encoded in the nuclear polyhedrosis virus genome (Ayres et al., 1994). In protostomes, there are far fewer FGFs in contrast to vertebrates, as two (let-756 and egl-17) have been found in Caenorhabditis elegans and three (branchless, pyramus and thisbe) in Drosophila (Mason, 2007).

Most FGFs have amino-terminal signal peptides (Fig. 1 (a)) and are secreted from cells. FGFs 9, 16 and 20 lack this signal peptide but nevertheless are still secreted (Ornitz and Itoh, 2001). FGF1 and FGF2 lack these signal sequences and are secreted by non-canonical pathways, however they can be found on the cell surface and within the extracellular matrix. Golfarb (2005) suggests that FGFs 11-14 do not interact with FGF receptors (FGFRs) and are not secreted but instead localise to the cell nucleus.

Fig. 1 (above) illustrates the structural features of the FGF polypeptide (a). A signal sequence (shaded grey) can be seen here within the amino terminus and is present in most FGFs.

All FGFs contain a core region (Fig. 1 (a)) containing around 120 amino acids of which 6 are identical amino acids residues and 28 are highly conserved (Goldfarb, 1996). The black boxes (numbered 1 to 12) represent the location of β strands within the core. The three dimensional structure of FGF2 (b) can also be seen where the heparin binding region (yellow) includes residues between β1 and β2 strands and in β10 and β11 strands.

FGFs have a high affinity for heparan sulfate proteoglycans (HSPG) and require heparan sulphate to activate one of four transmembrane receptor tyrosine kinases (FGFR1-4) in all vertebrates. FGFR5 has been identified recently, however most action is mediated via FGFR1-4 (Powers et al., 2000). FGFRs are membrane associated class IV receptor tyrosine kinases (RTKs). The FGFR tyrosine kinase receptors (Fig. 2 B) include 3 immunoglobulin (Ig) domains and a heparin binding sequence which requires heparan sulphate to be activated (McKeehan et al., 1998). HSPG are low affinity receptors that are unable to transmit a biological signal but act as co-factors for activation and regulation of an interaction between FGFs and FGFRs.

Fig. 2 (above) illustrates a two dimensional generic FGF (A) and a FGFR (B) protein. The structure of a FGF (A) coincides with that of Fig. 1, containing a signal sequence in the amino-terminus and the conserved core region containing HSPG and receptor-binding sites. The main features of FGFRs (B) include 3-Immunoglobulin domains, an acidic box (AB) which lies between IgI and IgII, heparin-binding domain, Cell Adhesion Molecule (CAM)-homology domain, transmembrane domain and a split tyrosine kinase enzyme domain for catalytic activity and binding of adaptor proteins. The Ig domains in the extracellular region of a FGFR are required for FGF binding and regulate binding affinity and ligand specificity.

Multiple alternative splicing that generates a range of FGFR1-4 receptor isoforms with transformed ligand binding properties provides diversity (Olsen et al., 2006). For example, FGF2 interacts with all four receptors FGFR1-4 whereas FGF7 only interacts with the FGFR2 IIIb isoform (a splice variant of FGF2; expressed in epithelial cells). Ligand-receptor binding specificity is affected by alternative splicing particularly in the C-terminal region of the third immunoglobulin loop in FGFR1-3 which produces IIIb or IIIc isoforms (Mason, 2007). Table 1 (below) illustrates the specificity of the FGF ligands for particular FGFR isoforms. This table is useful yet evidence from in vitro may appear misleading as in vivo involves influence from co-factors such as HSPG (Mohammadi et al., 2005).

Table 1 (above) shows there are seven FGFR isoforms (FGFR1b; FGFR1c; FGFR2b; FGFR2c; FGFR3b; FGFR3c and FGFR4) that FGF1 through to FGF23 variously bind. Alternative mRNA splicing of FGFR1-3, particularly in the carboxy-terminal half of the third extracellular immunoglobulin loop (Ig-domain III), derives the b and c isoforms. HSPGs are necessary co-factors in activation of FGFRs by FGFs and evidence has found the ternary complex to comprise of FGF-FGFR-HSPG in a 2:2:1 ratio (Mohammadi et al., 2005). The co-binding of HSPG prevents proteolysis and thermal denaturation (Itoh and Ornitz, 2004). HSPG binding of FGF induces dimerization of FGFR, followed by transphosphorylation of receptor subunits, initiating an intracellular signalling cascade.

FGF signalling: It’s a cellular game

Following formation of the FGF-HSPG-FGFR complex several downstream signalling pathways are activated (Fig. 3 below). This includes three pathways, the Ras/Mitogen-activated protein kinase (MAPK) pathway, Phosphoinositide 3-kinase (PI3K)/ Akt pathway and phospholipase C- (PLC )/ Ca2+/ protein kinase C (PKC) pathway. These pathways are mediated via docking proteins (such as FGF receptor substrate (FRS) and Grb2 in the Ras/MAPK pathway) that recruit downstream enzymes. The Ras/MAPK pathway (Fig. 3) is initiated via Grb2 (a docking protein) where its SH2 domain binds to the tyrosine phosphorylated FRS2 in response to activation of the FGFR receptor (Kouhara et al., 1997). Grb2 binds to SOS (son of sevenless; a guanine nucleotide exchange factor) via a SH3 domain on the Grb2 molecule. This Grb2-SOS complex activates SOS which promotes the dissociation of GDP from Ras so it is able to bind GTP for its activation. Activated Ras activates RAF (MAPKKK) which is normally held in a closed conformation by the 14-3-3 protein. Once activated, RAF phosphorylates and activates mitogen-activated and extracellular signal-regulated kinase (MEK (MAPKK)) which in turn phosphorylates ERK1/2 (MAPK). MAPK then translocates into the nucleus to phosphorylate specific transcription factors of the Ets family which in turn activate expression of FGF target genes. In addition, it is also evident from Fig. 3 that active ERK itself can antagonise FRS activity.

Activation of the PI3K/Akt pathway (Fig. 3) is by binding of Gab1 (Grb2-associated-binding protein 1) to FRS2 indirectly via Grb2. In the presence of Gab1, activation of PI3K stimulates the Akt pathway which suggests FGFs have anti-apoptotic effects in the developing nervous system (Mason, 2007). In addition, PI3K can bind to a phosphorylated tyrosine residue of FGFR directly. The third way in which the PI3K/Akt pathway is activated is by activated Ras inducing membrane localisation of the PI3K catalytic subunit.

PLC- /Ca2+/PKC pathway is also activated when a tyrosine residue is autophosphorylated in the carboxy terminal of the FGFR. PLC- hydrolyses phosphatidylinositol to produce inositol trisphosphate (IP3) and diacylglycerol (DAG) which stimulates calcium release and activates PKC, respectively. PKC has also been found to activate the Ras/MAPK pathway independent of Ras but dependent on c-Raf (Ueda et al., 1996). Fig. 3 also indicated that the final activated components, of the three signalling pathways mentioned, translocate into the nucleus to activate specific transcription factors of the Ets family (particularly Ets1, Pea3, and Erm) which activate expression of FGF target genes and in turn these feedback (Fig, 4) to regulate intracellular signalling (Dailey et al., 2005).

Most of the proteins produced function as feedback inhibitors (as seen in Fig. 4), including Sprouty (Spry), Sef and MAP Kinase phosphatase 3 (MKP3) which modulate particularly the Ras/Erk pathway at different levels (Mason, 2007). In contrast, stimulation of the fibronectin leucine-rich transmembrane type III (XFLRT3) protein causes FGF signalling to be positively regulated (Böttcher et al., 2003).

Sprouty (Spry) was one of the first identified feedback regulators of the FGF pathway. Thisse and Thisse (2005) found Spry to antagonise FGF Signalling by gain and/or loss of function experiments in mouse. Spry acts at the level of Raf and/or Grb2 (Fig. 4). Gain and/or loss of function experiments in zebrafish demonstrated that Sef antagonises FGF signalling (Fig. 4) acting at level of MEK and ERK (Tsang et al., 2002). Mouse studies have suggested that FGFR signalling is required for Dusp6 transcription which codes for MKP3 (Ekerot et al., 2008). From this study it was also found that MKP3 acts as a negative regulator of ERK activity (as seen in Fig. 4). Sef and XFLRT3 are located at the membrane (Fig. 4) and carry out antagonising actions with FGFR directly.

FGF signalling can be regulated at different levels, from the membrane all the way down to the level of phosphorylation of MAPK and it is important also to know that FGFs have been detected in the nucleus (Mason, 2007). Most of the downstream target genes as described earlier are feedback inhibitors (Spry, Sef and MKP3) but FGF signals are also known to interact with many other important pathways such as transforming growth factor-β (TGF-β), Hedgehog (HH), Notch and Wnt (Gerhart, 1999). Therefore, in conjunction with these, FGFs are responsible for development of most organs of the vertebrate body. In the nervous system, FGFs have been implicated to play a role in early developmental processes, such as neural induction, patterning and proliferation (Umemori, 2009).

Neural induction: The Default Model

Spemann and Mangold (1924) pioneered the study of neural induction, which is defined as the process by which naive ectodermal cells aquire a neural fate. Their work involved demonstrating that tissue from the dorsal lip of the frog Xenopus laevis blastopore could induce a second ectopic nervous system (Fig. 5 above left) when implanted onto the ventral side of a host gastrula embryo. The second ectopic nervous system was host derived indicating that the graft was important in determining cell fate. This region, located on the dorsal side of an amphibian embryo, was named the Spemann organizer as it could direct the neighbouring ectodermal cells to form nervous system instead of epidermis.

Although the organizer (group of dorsal mesodermal cells) was found to be present in many species (Hamburger, 1988) it was the Xenopus laevis which gave an insight into the molecular events involved in neural induction in vertebrates (Hemmati-Brivanlou et al., 1994). This was particularly because amphibians were found to be ideal experimental models for the study of neural induction as neurulation initiated within twelve hours after fertilisation (Weinstein and Hemmati-Brivanlou, 1997).

It was implied that signals from the organizer provide instructions to the ectoderm to form neural tissue therefore for many decades the view was that the ‘default’ state of the ectoderm was to produce epidermis. The first challenges to this model came from studies making use of dissociated cell cultures (Sato and Sargent, 1989). It was found that when animal caps were cultured intact that epidermis formed but neural tissue arose from animal caps that had been dissociated for prolonged periods (as seen in Fig. 6 below). This led to the idea that intact tissue may block the formation of neural tissue by presence of neural inhibitors which are diluted out when the tissue is dissociated. Recent research has found that the default nature of the ectoderm is to produce neural tissue that requires inhibition of a neural inhibitor from the ectoderm.

Before considering the process of neural induction I would like to take a step back and describe the three germ layers of the embryo. Following fertilisation, the zygote undergoes stages of cleavage to eventually form a gastrula with three germ layers (in triploblastic animals) usually only visible in vertebrate animals. The Germ layers will eventually give rise to all of the animal’s organs through a process known as organogenesis. The three layers include, the ectoderm (outermost), endoderm (innermost) and mesoderm (which is between the ectoderm and endoderm) layers. The Endoderm gives rise to the lung, thyroid and pancreas. The mesoderm forms the skeleton, skeletal muscle, the urogenital system, heart and blood. The outermost layer, the ectoderm which is of concern here, gives rise to the epidermis and nervous system. It is at gastrulation that the vertebrate ectoderm is competent to differentiate into neural tissue or epidermis. Unless told otherwise, the default nature of the ectoderm is to produce neural tissue and this was outlined as the default model.

The Default model of vertebrate neural induction, discovered over a decade ago in Xenopus, proposed that in the presence of bone morphogenetic protein (BMP), a signalling molecule of the TGF-β superfamily, causes the ectoderm to give rise to an epidermal cell fate (Stern, 2006; Muñoz-Sanjuan and Brivanlou, 2002). In support of this model, consistent with the idea that BMP activity inhibits neural fates, animal caps which had been injected with RNA encoding effectors of BMP4 (Smad 1/5 or Msx1) neuralization did not occur. Conversely, it was found that inhibition of BMP activity in the ectoderm is essential for a neural fate which forms the basis of the default model of neural induction. Inhibition of BMP is achieved through direct binding of BMP antagonists emitted from the organizer (Wilson and Hemmati-Brivanlou, 1997). These BMP antagonists include chordin (Sasai et al., 1995), noggin (Lamb et al., 1993) and follistatin (Hemmati-Brivanlou et al., 1994) which bind to BMPs extracellularly to prevent its interaction with its own receptor (Hemmati-Brivanlou and Melton, 1997). These molecules have direct neural activity which means they induce formation of neural tissue in the ectoderm without forming mesoderm.

It was initially believed that these molecules acted as ligands to bring about neural tissue formation. Experiments found that there was conservation through species, identifying that chordin was homologous to the short gastrulation (sog) gene found in Drosophila which has been shown to antagonize the BMP homologue decapentaplegic (dpp) (Wharton et al., 1993), suggesting that these molecules might act as inhibitors rather than inducers and that these inhibitory mechanisms have been conserved from arthropods through to vertebrates. It was experiments (Fig. 6) showing that dissociated ectodermal explants would become neural tissue in absence of ‘inducing’ signals from the organizer (Sato and Sargent, 1989). Evidence found that neural induction resulted from inhibition of the TGF-β pathway as expression of dominant-negative activin receptor gave rise to neural fates in amphibian ectoderms (Hemmati-Brivanlou and Melton, 1994). It was found that chordin, noggin, follistatin and molecules such as Cerberus and Xnr3 (Xenopus nodal related 3) bound to BMP in the extracellular space inhibiting its action (Hemmati-Brivanlou and Melton, 1997) leading to the much debated default model of neural induction.

Neural Induction: FGFs get it started

Support for the default model still remains, mainly in Xenopus, but other work (especially in chick and mouse) suggests a more complex mechanism (Streit et al., 1998). It has been established that the BMP pathway is involved in determining ectodermal cell fate (Wilson and Hemmati-Brivanlou, 1997) but it still remains to be proved conclusive if BMP inhibition is required for neural induction alone or if other pathways act separately or with BMP inhibition.

In the chick embryo it has been found that naive epiblast cells do not respond to BMP antagonists until previous exposure to organizer signals for five hours (Streit et al., 1998). Striet et al. (2000) grafted an organizer to observe the genes induced in the epiblast within this time period. A gene ERNI (early response to neural induction) was identified as a coiled coil domain with a tyrosine phosphorylation site and found to be expressed throughout the region that later contributes to the nervous system at pre-primitive streak stages (Hatada and Stern, 1994). Striet et al. (2000) findings made ERNI the earliest known marker after a response to organizer signals, prior to even Sox3 (induced by the node in 3 hours (Streit and Stern, 1999)).

FGFs are becoming more evident that they have a major role in neural induction as it has been shown to begin before gastrulation, before BMP antagonists even appear (Wilson et al., 2000). In the chick, it has been found that FGFs have the role of blocking BMP signalling and promoting neural differentiation (Wilson et al., 2000). In ascidians, FGF signalling is the main mechanism of neural induction with BMP antagonism playing a role in later development (Lemaire et al., 2002). In frogs and fish, in contrast, FGFs do not have a certain role in neural induction and is believed their primary role is BMP inhibition (Pera et al., 2003).

Exposure of the chick epiblast to an implanted organiser for around 5 hours induces Sox3 (an early neural plate marker) (Stern, 2005). After removal of the implanted organiser, chordin can be used to stabilise it (Striet et al., 1998) which implies that before the ectoderm can respond to BMP antagonists it must be exposed to 5 hours of signals from the organizer. During these 5 hours, several genes become activated such as, ERNI (early response to neural induction) which becomes active after 1 hour (Streit et al., 2000) and Churchill (Chch) after about 4 hours (Sheng et al., 2003). These are both induced by FGF and not BMP inhibition, indicating the importance of FGFs in early neural induction. Churchill which is expressed in the neural plate inhibits brachyury, a transcription factor, which as a result suppresses mesoderm formation by preventing cell ingression.

In the chick, FGF8 is expressed in the hypoblast, prior to gastrulation before Hensen’s node appears (the chick equivalent to the organizer) indicating that neural induction is in fact able to begin before gastrulation. This is important because ERNI and Sox3 mark neural induction and require FGF signalling (Stern, 2005). Streit et al. (2000) found that FGF8 coated beads induce ERNI as efficiently as the node within 1-2 h without inducing brachury and also the expression of Sox3. These results indicate FGFs to be possible early signals in neural induction. It is FGF8 which has been identified as the best candidate because it is expressed in the anterior part of the str