Keywords

6.1 Introduction

Fibroblast growth factors (FGFs) exert their cellular effects by interacting with FGF receptors (FGFRs) in a complex with heparan sulphate (HS) [1]. FGFRs, a class of receptor tyrosine kinase (RTK), dimerise and undergo transphosphorylation of the kinase domain upon ligand binding [2], leading to the recruitment of adapter proteins and initiating downstream signalling.

The extended FGF family is composed of 22 members, varying in size from 17 to 34 kDa. All members share a conserved 120 amino acid sequence and show 16–65 % sequence homology [3]. However, only eighteen FGFs signal via FGFR interactions (FGF1–10 and 16–23), while FGF11–14, which lack a signal peptide, act in an intracellular manner. Thus, many consider the FGF family to comprise only 18 members. Furthermore, although they are numbered from 1 to 23, FGF15 is the mouse ortholog of human FGF19. Each ligand binds to FGFRs with varying specificity; some are promiscuous, for example FGF1, and bind to multiple receptors, while others, like FGF7, bind only to one receptor isoform [4] (Fig. 6.1).

Fig. 6.1
figure 1

FGFR isoform ligand specificity. Each receptor isoform has varying affinities for the FGF family of ligands. This variation is dictated by alternative splicing of the receptor. Binding affinities taken from [4, 5]

Full size image

There are seven signalling receptors, encoded by four FGFR genes, FGFR1–4 [6]. FGFRs 1–3 have highly conserved intron/exon boundaries [4] (Fig. 6.2).

Fig. 6.2
figure 2

Schematic representation of FGFR1–4. FGFR1–4 contain a variety of defined structural domains, some of which are highly conserved across the receptors and their individual isoforms. The receptors are encoded by genes found on chromosomes 8, 10, 4 and 5, respectively. Three isoforms, termed a, b and c, exist for FGFR1 and 2; only isoforms b and c function in a signalling capacity. FGFR3 has two isoforms, b and c, while only one isoform of FGFR4 exists. These isoforms are generated through alternative splicing. Only one isoform of each receptor is shown. Amino acid residue numbers are indicated at the top of each panel, including a 21–22 amino acid signal sequence (UniProt accession: P11362, P21802, P22607, P22455, respectively)

Full size image

Alternative splicing of exons 8 and 9, encoding IgIII of FGFR1–3, results in translation of two distinct isoforms capable of signal transduction. These isoforms are termed IIIb and IIIc, depending on which exons are spliced out (Fig. 6.3). This third Ig loop encodes the ligand binding domain; alternative splicing of this region is responsible for ligand binding specificity (Fig. 6.1). A third isoform exists for FGFR1 and 2, termed IIIa. This variant results in a truncated, secreted protein, which is unable to transduce a signal and may have an autoinhibitory role in FGF signalling, possibly by sequestering ligands [8]. FGFR4 is distinct in that it has only one isoform, homologous to the IIIc variant of FGFR1–3 [9].

Fig. 6.3
figure 3

FGFR structure, control of ligand specificity and receptor autoinhibition via alternative splicing. Each receptor monomer is comprised of an extracellular domain including three Ig loops, IgI, IgII and IgIII (also referred to as D1, D2 and D3, respectively), an acid box in the IgI–IgII linker region (represented by a white box), a transmembrane domain and an intracellular split kinase domain. Disulphide bonds are present in each Ig loop. IgI and the acid box are involved in autoinhibition of the receptor, while IgII and IgIII are involved in ligand binding. The HS-binding site lies in IgII, indicated in green. Ligand binding specificity is generated by alternative splicing of the IgIII domain. The first half of IgIII is encoded by an invariant exon (IIIa), which is spliced to either exon IIIb or IIIc (represented in blue and red, respectively), both of which splice to the exon that encodes the transmembrane domain (TM) region. Epithelial tissues predominantly express the IIIb isoform and mesenchymal tissues express IIIc. FGFR4 is expressed as a single isoform that is paralogous to FGFR-IIIc. An additional alternative splicing event can occur leading to the deletion of exons coding for IgI and/or the acid box/linker region. This leads to loss of receptor autoinhibition [7]

Full size image

Receptor expression is generally cell type specific, for example IIIb and IIIc isoforms of FGFR1 and 2 are expressed in epithelial and mesenchymal cells, respectively [10, 11]. However, as shall be discussed later, this cell type specificity can change when FGFRs are associated with diseases such as cancer.

6.2 FGF:FGFR:HS Complex

Heparin, used in vitro as the model heparan sulphate (HS), is a member of the HS family of proteoglycans (HSPGs) and has been used to establish the necessity of HS binding in FGF:FGFR:HS complex formation [12]. This acidic molecule resembles the highly sulphated saccharide chains of HS [13]. Upon binding to FGFs/FGFRs, HS saccharide chains induce a conformational change. The length of the saccharide chain is important in FGF–FGFR interactions. Ornitz and colleagues reported interaction of a dodecasaccharide with both high- and low-affinity heparin-binding sites of ligands and showed that octasaccharides, thought to be the smallest saccharides with biological activity in FGF–FGFR interactions, could only engage the low-affinity binding sites of the ligand [14]. However, others have postulated that smaller chains, including hexasaccharides and disaccharides, may have biological activity [13, 15]. The heparin-binding residues found in the IgII loop of FGFRs (Fig. 6.2) are highly conserved [16], while heparin-binding residues of FGFs are diverse. Because of this, different FGFs require various HS sulphation patterns and/or length of chains for their optimum activity. Variability of HS sulphation patterns and length across cell types has an effect on FGF–FGFR interactions and may be a mediator of the biological activity of FGFRs [13–15, 17].

Another highly sulphated glycosaminoglycan (GAG), chondroitin sulphate (CS), is also able to interact with FGFs and FGFRs to promote complex formation. Studies have shown that insufficient synthesis of GAGs, which are assembled in the Golgi, impairs FGF/FGFR signalling capabilities [18]. The sulphation pattern and chain length of GAGs is so variable that there may be tissue- and even cell-specific GAG chains with varying specificities for ligands and receptors [19]. The difference in these chains may be of particular importance in the regulation of FGF/FGFR signalling. Work by a number of groups has also shown that variations in GAG sequences capable of interacting with FGFs and FGFRs can both inhibit or facilitate FGF signalling [17, 20–23].

A widely accepted model of FGFR interactions [16] proposed a complex of FGF:FGFR:HS in a 2:2:2 ratio (Fig. 6.4a). Two independent FGF:FGFR:HS ternary complexes are formed in a 1:1:1 ratio via HS binding to both receptor and ligand. They bind via receptor interactions, as well as interactions between the ligand in one complex and the receptor in another, thus forming a stable, symmetrical dimer. Direct ligand–ligand interactions are not observed. This FGF-FGFR complex can only be formed in the presence of HS. In summary, stabilisation of the dimer is through the following interactions: receptor–ligand, receptor–HS, ligand–HS and receptor–receptor.

Fig. 6.4
figure 4

Alternative FGF:FGFR:HS models. The basic structure of the FGF:FGFR complex comprises two receptor molecules and two ligands. Two models are presented which differ in the number of heparan sulphate proteoglycans (HSPGs) they contain. Dimerisation occurs upon ligand binding, leading to autophosphorylation of the kinase domain and subsequent downstream signalling. (a) FGF:FGFR:HS 2:2:2 model. First presented by Schlessinger et al. [16], and taken to be the most biologically relevant, this model proposes a symmetrical dimer utilising two HSPGs which bind ligands, bringing them into close proximity with the receptor, facilitating dimerisation. The HS chains also bind to the heparan-binding site of the IgII loop of the receptor to form a complete, active molecule capable of autophosphorylation and subsequent phosphorylation of signalling molecules. (b) FGF:FGFR:HS 2:2:1 model. Proposed by Pellegrini et al. [24], this dimer is formed using only one HSPG which binds both ligands necessary for each receptor monomer. Dimerisation occurs with the HS chain binding to both ligands and receptors, leading to signalling cascade activation

Full size image

A second model [24] proposed FGF:FGFR:HS complex formation in a 2:2:1 ratio (Fig. 6.4b). Crystal structure analysis of FGFR2–FGF1 interactions showed a central heparin molecule linking two ligands and two receptor molecules. In this model, each ligand binds to a receptor monomer with heparin interacting with both ligands but only one receptor molecule. Two 1:1 FGF:FGFR complexes are joined to form a dimer via interactions with one HS chain.

6.3 Signalling Pathways

Upon dimerisation, reciprocal phosphorylation of the tyrosine kinase domains of the receptors occurs. These phosphorylated receptors are then able to act as docking sites for intracellular proteins, leading to activation of signalling cascades (Fig. 6.5) [25–27]. This autophosphorylation occurs in a specific order; ‘first-phase’ phosphorylation increases the catalytic activity of the kinase after ligand binding, while ‘second-phase’ phosphorylation creates phosphotyrosine-binding sites for docking molecules containing Src homology-2 (SH2) and phosphotyrosine-binding (PTB) domains [25, 28]. From this, four signalling pathways can be activated: MAP Kinase (MAPK), PI3K/AKT, PLCγ and STAT [25]. The key difference between FGFRs in signalling is the strength of their tyrosine kinase activity; their target proteins are the same [29].

Fig. 6.5
figure 5

FGF:FGFR-induced downstream signalling. Ligand-receptor binding induces four signalling cascades: MAPK, PI3K/AKT, PLCγ and STAT. These pathways comprise a series of phosphorylation events, culminating in the regulation of target genes, which dictate cellular processes, for example proliferation and migration

Full size image

The lipid-anchored adapter protein FRS2 plays an integral role in the MAPK and PI3K/AKT pathways. FRS2α binds to the receptor via its PTB domain [30, 31] and undergoes phosphorylation. GRB2, another adapter molecule, is then recruited to FRS2α. From this point, two FGF-induced signalling pathways can be activated:

6.3.1 Phosphoinositide-3 Kinase

GRB2/FRS2α binds to and phosphorylates GAB1 via the SH3 domain of GRB2 [32]. This FRS2α/GRB2/GAB1 complex recruits PI3K via the SH2 domain of its p85 subunit. Activated PI3K produces phosphatidyl-inositol (3, 4, 5)-trisphosphate (PIP3), resulting in activation of the AKT pathway. Anti-apoptotic signalling, as well as cell growth and proliferation, is then initiated [32].

6.3.2 Mitogen-Activated Protein Kinase

Activation of the mitogen-activated protein kinase (MAPK) pathway results in mitogenic activity and cell survival [33]. The MAPK pathway is initiated by RAS binding to the FGFR/FRS2α/GRB2/SOS complex. RAS then recruits and phosphorylates RAF, leading to phosphorylation of MEK (MAPK/ERK kinase) and subsequent phosphorylation and activation of MAPK [33]. MAPK, also known as Extracellular Signal-Regulated Kinase (ERK), is then able to activate transcription factors in the nucleus, for example c-MYC, and influence the cell cycle.

The PLCγ and STAT pathways are mediated through other mechanisms.

6.3.3 Phospholipase C γ

Autophosphorylation of FGFR residue Tyr766 in FGFR1 creates a specific binding site for the SH2 domain of phospholipase Cγ (PLCγ), leading to tyrosine phosphorylation of PLCγ [34]. Recruitment of PLCγ is aided by PIP3, generated in response to PI3K stimulation [35]. Activation of PLCγ leads to cleavage of phosphotidyl-inositol-4, 5-bisphosphate (PIP2) into the second messengers inositol trisphosphate (IP3) and diacylglycerol (DAG) [36]. IP3 then releases calcium stores from the endoplasmic reticulum (ER) [37]. Calcium ions, along with DAG, then activate protein kinase C (PKC). PKC is then able to phosphorylate RAF and activate the MAPK pathway.

6.3.4 Signal Transducer and Activator of Transcription

The STAT family of cytoplasmic transcription factors can be activated by non-receptor tyrosine kinases, the Janus Kinases (JAK), leading to cell proliferation, differentiation or apoptosis [38]. Upon FGFR dimerisation and autophosphorylation, JAKs are phosphorylated by the receptor, forming a FGFR/JAK complex. This acts as a docking site for STATs, which are in turn tyrosine phosphorylated in their SH2 domain [39]. STAT dimers form and translocate to the nucleus, where they bind to gamma-activated site (GAS) enhancers to activate or repress gene transcription [39].

6.4 Regulation of FGF Signalling

Regulation of FGF signalling is critical to ensure a balanced response to receptor stimulation. This occurs largely through four mechanisms:

6.4.1 Receptor Internalisation

CBL, a multidomain protein that possesses an intrinsic ubiquitin ligase activity [40], binds to the FRS2α/GRB2 complex via the SH3 domain of GRB2 and the proline-rich region of CBL. Recruitment of CBL to FRS2α leads to ubiquitination of both FGFR and FRS2α and therefore attenuation of FGFR-mediated signalling.

6.4.2 Receptor Cleavage

Numerous growth factor receptors undergo ectodomain shedding, a process known to downregulate signalling. Ectodomain shedding, or S1 cleavage, is a process of proteolytic cleavage either within or near the membrane by members of the metalloprotease (MMP) and A Disintegrin And Metalloproteinase (ADAM) family [41]. Induction of this cleavage occurs in response to receptor activation [42]. Cleavage within the transmembrane domain by γ-secretase, known as S2 cleavage, often follows. Together, these cleavage events are known as Regulated Intramembrane Proteolysis [43].

FGFR1 is cleaved by MMP2 [44], and FGFR2 can be targeted by ADAM9 or 15 [45, 46]. Cleavage of both receptors leads to attenuation of signalling via two main mechanisms: downregulation of the number of active receptors at the cell surface and generation of a soluble extracellular domain able to compete with membrane-bound receptors for ligand binding [47]. Interestingly, FGFR1 also can be cleaved intracellularly by the serine protease Granzyme B. Although this was reported initially as a means of cytotoxic T lymphocytes inducing target cell apoptosis [48], the process is hijacked in cancer cells to allow nuclear trafficking of the C-terminus of the receptor, which acts to regulate transcription of a pro-migratory gene signature [49].

FGFR3 is unique in that the S1 cleavage occurs in an endosomal compartment, where it is cleaved by an as yet unknown protease, rather than involving a member of the ADAM family [41]. S2 cleavage via γ-secretase then occurs, generating a soluble intracellular domain capable of trafficking to the nucleus. Here, the nuclear FGFR3 fragment may be responsible for novel interactions in addition to the well-established downstream signalling pathways of receptor activation.

6.4.3 Induction of Negative Regulators

The first identified negative regulator of FGFRs was sprouty (SPRY) [50], one of a family of four proteins. SPRYs are thought to act through one of two mechanisms. Firstly, they may interact with GRB2, interrupting the FRS2α/GRB2 complex and therefore decreasing signal transduction [51]. Alternatively, SPRY–RAF interactions may occur, preventing RAF phosphorylation and therefore inhibiting MAPK signalling [52].

MAPK signalling can also be inhibited by Sprouty-related Enabled/vasodilator-stimulated phosphoprotein Homology 1 Domain-containing proteins (SPRED1 and 2) [53]. SPRED proteins prevent RAF activation of MEK by forming a complex between RAS and RAF. Co-localisation of SPRED2 with the protein Neighbor of BRCA1 (NBR1) results in sequestration of FGFR and lysosomal degradation [54].

Similar Expression to FGF (SEF) proteins also negatively regulate FGF signalling via a number of mechanisms: targeted inhibition at or downstream of MEK [55]; inhibition of RAS activation, which also inhibits the PI3K pathway [56]; direct interaction with FGFR and subsequent inhibition of FGFR and FRS2α phosphorylation [56–58]; and blockage of ERK/MEK dissociation, where SEF acts as a spatial regulator of phospho-ERK migration to the nucleus [59].

Another mechanism of negative regulation is via direct phosphorylation of MAPK pathway proteins. For example, SOS and RAF are substrates of MAPK. Phosphorylation of SOS by MAPK disrupts interactions between SOS and GRB2. This decreases recruitment of SOS to the membrane and results in diminished RAS activation [60]. MAPK also phosphorylates RAF, reducing RAF kinase activity and therefore decreasing MEK and MAPK phosphorylation [61]. Induction of the MAPK pathway can also lead to attenuation of the PI3K/AKT pathway. Activation of MAPK leads to GAB1 phosphorylation. This decreases PI3K recruitment to GAB1, in turn reducing AKT pathway activation [62].

Alternative internal control mechanisms of FGF signalling exist, including autoinhibition of the receptor [16, 63, 64]. The FGFRs exist in ‘closed’ and ‘open’ conformation equilibrium [7]. The first Ig loop (IgI) and the IgI/IgII linker region containing the acid box, a glutamate, aspartate and serine-rich sequence [6], are responsible for formation of the ‘closed’, autoinhibited state. Spectroscopic investigations have shown the acid box engages in electrostatic interactions with the HS-binding site of the IgII loop, inhibiting receptor–HS interactions and, therefore, receptor activation. This then encourages intramolecular interactions between IgI and the ligand-binding sites of the IgII and IgIII loops, further aiding the acquisition of a closed conformation [65]. Alternative splicing of exons encoding the IgI and/or acid box region leads to enhanced affinity of the receptor for its ligand and HS, increasing downstream signalling [65]. Loss of this region has been implicated in cancer [66, 67]. This mechanism of autoinhibition supports FGF binding specificity of receptors as only specific ligands with high affinity for the receptors will overcome the inhibition and bind to the receptor.

6.4.4 Klotho Interactions

FGFRs can also interact with klotho family proteins. These senescence-related, single-pass transmembrane proteins function as FGF19 subfamily signalling cofactors. FGFs are split into seven subfamilies, with the FGF19 subfamily comprising FGF19, 21 and 23. These endocrine factors regulate metabolic processes [68]. The HS-binding sites of this subfamily differ greatly from other FGFs, reducing their affinity for HS [69, 70]. Because of this, they require Klotho as a cofactor, to signal through FGFRs. Klotho expression is confined to a limited number of tissues [71]. It is able to bind FGFR1c independent of HS binding and convert it into a FGF23 receptor in the kidney [72]. Mutations in klotho proteins or the FGF19 subfamily are associated with diseases including autosomal dominant hypophosphatemic rickets, premature ageing disorders and diabetes [73]. Klotho is able to actively compete with FGF2 for FGFR1c binding, therefore attenuating FGF2 signalling [72].

6.5 FGFRs in Development

The critical role played by FGFR signalling during embryogenesis is highlighted by its conservation throughout evolution, from invertebrates through to higher mammals. There are a number of reviews that provide exquisite detail on FGFR signalling in a wide range of model organisms, including Caenorhabditis elegans [74], Drosophila melanogaster [75] and vertebrates [76–78]. However, we highlight below some key findings in the major model organisms.

6.5.1 Caenorhabditis elegans

C. elegans has just one FGFR, EGL-15, which was identified in mutant screens as a result of its importance in the migration of hermaphrodite sex myoblasts [79]. The EGL-15 receptor is essential for sensing the chemoattractant FGF ligand EGL-17, expressed in the target gonad and vulva [80]. Another FGF ortholog, LET-756, which shows structural homology to the FGF-9 subfamily [81], is essential for larval viability [82]. Further elements of the downstream signalling pathway were elucidated with the identification of a receptor tyrosine phosphatase, CLR-1 [83], and components of the MAPK cascade [84], which are key to FGF signalling in the worm, as they are in other model organisms. Interestingly, the FGF co-receptor Klotho has two functional orthologs in the worm, and these are essential in mediating the longevity and stress resistance effects of EGL-15/EGL-17 signalling [85]. Beyond the scope of this chapter, there are a number of non-canonical FGFR interactions in the worm that are the subject of an elegant review elsewhere [74].

6.5.2 Drosophila melanogaster

The tracheal system in Drosophila has been a key system for the identification of aspects of the FGF signalling pathway over the past 30 years. Breathless—one of two FGFRs in the fly—regulates tracheal branching [86], acting in concert with its cognate FGF ligand, branchless [87], to activate downstream MAPK signalling [75]. Further genetic dissection of the branching process identified Sprouty as a negative regulator of FGFR signalling [50] and described how the Notch pathway interacts with FGFR signalling in controlling cell fate [88], although orthologs of other negative regulators of the FGFR pathway, Sef and XFLRT3, are not present in invertebrates [89].

A further FGFR ortholog, Heartless, was identified by virtue of its pivotal role in mesoderm migration and subsequent specification [90, 91], although its ligands, the FGF-8 orthologs Pyramus and Thisbe, were not identified until much later [92, 93]. Downstream of FGF signalling, a novel adapter, Dof, is critical for activating intracellular signalling [94], in much the same fashion that FRS2 acts in the vertebrate pathway.

6.5.3 Zebrafish

FGF signalling is an important factor in patterning the zebrafish embryo, interacting with signalling by TGF-β superfamily members to regulate mesoderm induction [95]. FGF signalling acts as a posteriorising factor driving trunk and tail development during anterior–posterior patterning [96], regulating downstream T-box family transcription factors Notail and Spadetail [97], and also regulates dorso-ventral patterning [98].

One of the main regulators of FGF signalling, SEF, was found first in zebrafish [57, 99], and fundamental understanding of the roles for FGF-8 signalling in neural development has been identified through the study of mutant zebrafish strains [100–102].

6.5.4 Xenopus

The first studies of FGF signalling in early development focused on its role as a competence factor, using Xenopus as a model system and showing that cell fate in the developing embryo was regulated by FGFs [103, 104]. FGF signalling was shown to be essential for cells to respond to mesoderm inducing TGF-β superfamily members [105, 106], and components of the entire pathway, from FGFRs to HSPGs to signal transduction proteins, have all been studied in detail in the frog [76]. Defects caused when FGFR signalling is inhibited, by small molecule inhibition [107], morpholino knockdown [108, 109] or expression of dominant negative receptor [105, 110, 111], confirm its fundamental importance in mesoderm induction, morphogenetic movements, neural induction, neuronal determination and anterior–posterior patterning.

6.5.5 Chick

The ability to manipulate and culture chick embryos has helped reveal several key roles for FGFR signalling, including elegant grafting studies showing the importance of FGFR activity in specifying and driving limb development [112, 113] and dynamic studies of presomitic mesoderm determination and subsequent somitogenesis [114–116]. Furthermore, FGFR signalling has been shown to act as a competence factor for neural induction [117, 118].

6.5.6 Mouse

All of the FGFs and FGFRs have been targeted using genetically modified mouse models, with approaches including germline deletion, conditional knockout and constitutive/inducible expression of either dominant negative or activating mutation constructs. The phenotypes of FGF ligand knockout mice are summarised elsewhere [119]. Extensive studies have revealed key roles for FGFRs in development, homeostasis and disease, and these are detailed in Table ‘FGFR1–4 at a glance’.

6.6 FGFR1

FGFR1 is used as the model receptor in the majority of studies and many of the findings are relevant to all FGFRs.

Syndecan 4 (S4), a transmembrane proteoglycan with extracellular HS chains, can regulate FGFR1 signalling, as well as signal independently as a growth factor receptor, to initiate cell adhesion and migration [120]. Recent work has hypothesised S4 could also be involved in FGFR1 trafficking [121]. S4 has a PDZ-binding domain, which is capable of activating the small GTPase, RhoG. RhoG is kept in a complex with S4 in its inactive form. Upon FGF binding to FGFR1, aided by the HS chains of S4, a ligand–receptor–S4 complex is formed. Signalling pathways of the individual receptors are initiated, for example MAPK from FGFR1. Upon this complex formation, RhoG is released from S4 and is activated by guanine exchange factors (GEFs). This induces membrane ruffling, leading to macropinocytosis of the complete FGF–FGFR–S4 complex. Trafficking of the internalised complex is dependent on another small GTPase, Rab5. When Rab5 function is absent, the vesicles cannot mature and become functional signalling endosomes. In this scenario, the MAPK pathway is not activated. However, when Rab5 activity is restored and localises to the macropinosome containing the FGF-FGFR-S4 complex, maturation of the vesicle is facilitated and MAPK signalling is activated. When S4 is absent, RhoG activity is high, leading to increased macropinocytosis and therefore receptor internalisation. S4 controls the rate of FGF–FGFR–S4 complex macropinocytosis; overactive S4 and Rab5 can lead to inadequate attenuation of the MAPK signal leading to continuous downstream signalling effects, for example cell migration. Hence, a novel method of FGFR1 MAPK signalling regulation via S4-medated trafficking is proposed.

Nuclear localisation of both FGFs and FGFRs has been reported in a number of cell lines and tissues [122]. The mechanism of nuclear translocation of FGFR1 has recently been elucidated by Chioni & Grose.

Studies have also shown Importin β is involved in FGFR1 nuclear translocation [123]. It is proposed that this occurs via the Integrative Nuclear FGFR1 Signalling (INFS) pathway [124]. FGFR1 is released from the cytoplasmic membrane into the cytosol. As it does not contain a nuclear localisation signal (NLS), FGFR1 associates with Importin β, a carrier protein that does. FGFR1 can then be transported into the nucleus where it is able to influence expression of, for example, C-JUN.

6.7 FGFR2

Developmental disorders are commonly associated with FGFR mutations, including Kallmann and Lacrimo-Auriculo-Dento-Digital (LADD) syndromes [125]. In skeletal disorders, for example, Crouzon, Pfiffer and Jackson–Weiss syndromes, receptor mutations tend to cluster in the linker region, connecting IgII and IgIII, and in both IgIII and the IgIII-transmembrane domain linker, functioning by either promoting receptor dimerisation or altering ligand–receptor specificity. Mutations in two conserved cysteine residues in IgIII of FGFR2 are commonly found in these skeletal disorders [33]. These cysteine residues usually function by linking to another cysteine in IgIII of the receptor via intramolecular bonds. Substitution of this amino acid with another creates an unpaired cysteine residue able to form an intermolecular disulphide bridge, leading to receptor dimerisation and therefore activation.

The craniosynostosis syndrome, Apert syndrome, depends on FGFR2 mutations. Gain-of-function changes in the highly conserved residues S252 and P253 of the IgII and IgIII linker of FGFR2 result in a change in ligand binding specificity [126, 127]. These are the cause of the majority of Apert syndrome cases [128, 129]. This has been further shown in mouse models; S252W FGFR2c mutants showed activation of the c isoform of the receptor by mesenchymally expressed FGF7, while FGFR2b was activated by FGFs associated with epithelial expression [130]. It is also possible that S252W and P253R mutations lead to the modified receptor remaining on the cell membrane for an extended period of time, rather than undergoing rapid endocytosis into the lysosomes like its wild-type counterpart. Downstream signalling pathways are affected, leading to increased ERK phosphorylation and therefore increased cell proliferation and migration capabilities, as well as premature differentiation [131].

Mouse modelling of Apert syndrome has shown that a soluble, truncated FGFR2 isoform is upregulated and influences FGF1-FGFR2 binding. This glycosylated IIIa-TM isoform is generated by direct splicing of exon 7 (IIIa) to exon 10 (TM), generating a premature stop codon three amino acids into the TM exon [8]. This loss-of-function mutation can thus negatively regulate FGF signalling.

A number of cancers have been found to contain somatic mutations identical to germ line mutations in FGFRs associated with developmental disorders. For example, FGFR2 mutations commonly seen in Apert syndrome and Pfeiffer are frequently identified in endometrial cancer [132], for example S252W and N550K, both of which result in receptor activation. The S252W mutation resides in the linker region between IgII and IgIII, the area responsible for providing key contacts with the ligand. This increases the binding affinity of the receptor for a range of FGFs while also leading to violation of ligand specificity of the receptor isoforms [133].

Other FGFR2 mutations in endometrial cancer include S373C and Y376C, which result in gain of a cysteine residue, allowing formation of intermolecular disulphide bonds [134]. This leads to constitutive receptor dimerisation and therefore downstream signalling. Although these findings were established using FGFR2c functional studies, it is known FGFR3 contains paralogous mutations.

6.8 FGFR3

FGFR3 is mutated in a range of developmental and skeletal disorders and is the most frequently mutated FGFR in cancer, as noted by the extensive list in the Catalogue of Somatic Mutations in Cancer (COSMIC) database. Gain-of-function FGFR3 mutations are involved in the most severe form of dwarfism in humans, thanatrophoric dysplasia types I and II [135, 136] and achondroplasia [137, 138]. Similar somatic mutations have been found in bladder and cervical cancer, amongst others, where they are believed to have a positive effect on proliferation and inhibit apoptosis [139].

The FGFR3 germline mutation, A391E, is known to cause abnormal cranium growth and is responsible for Crouzon syndrome [140]. This mutation is also found in bladder cancer [141]. A391E leads to stabilisation of the transmembrane domain of the dimerised receptor independent of ligand binding [142] and is therefore responsible for ligand-independent receptor activation [143].

6.9 FGFR4

FGFR4 has a diverse range of roles, from involvement in the vascular system to regulation of hepatic bile acid and lipid metabolism [144, 145]. Recently, a single nucleotide polymorphism (SNP) in FGFR4 has been identified which is thought to have both positive and negative prognostic value in different diseases. This SNP (rs351855) results in a glycine-arginine change (G388R) in the transmembrane domain, leading to increased receptor stability and sustained receptor activation [146].

FGFR4 is expressed at high levels in coronary artery disease (CAD). Investigation of the SNP status of CAD patients in a Chinese population study showed this SNP is low in CAD patients [145]. It is therefore thought that having this SNP may be beneficial, acting as a protective factor against CAD development in Asian populations. This SNP is also associated with poor prognosis in prostate and breast cancer [146, 147].

6.10 FGFRL1: The Fifth FGFR

A fifth member of the FGFR family has been discovered, Fibroblast Growth Factor Receptor Like 1 (FGFRL1). This protein, which exists as a homodimer consisting of the three characteristic extracellular Ig-like domains, acid box between IgI and IgII and a transmembrane helix, differs from the classic receptors in that it has no intracellular tyrosine kinase domain [148–150]. Instead, the intracellular portion of FGFRL1 consists of only 100 residues including a histidine-rich sequence and a tandem tyrosine-based motif [148, 151, 152]. These two sequences function as signals for FGFRL1 trafficking from the plasma membrane to endosomes and lysosomes. Deletion of these sequences resulted in inefficient FGFRL1 internalisation and prolonged time at the plasma membrane [151].

Interactions with both FGFs and heparin have also been confirmed through dissociation studies [148]. FGFRL1 binds strongly to FGF3, 4, 8, 10 and 22 [153] and the affinity of FGF3 for FGFRL1 is at least one order of magnitude higher than the majority of FGFs for their receptors [151]. Affinity of this magnitude between FGFs and their receptors is only seen in mutant receptors in, for example, Pfeiffer, Apert and Muenke craniosynostosis syndromes [154]. The gain-of-function P253R mutation in Apert syndrome exactly matches an arginine residue at position 243 in FGFRL1; this residue could be responsible for the high affinity of FGF3 for FGFRL1 [155]. Its interaction with HS is also stronger than that of classic FGFRs and heparin [156, 157]. FGFs bind FGFRL1 between IgII and IgIII domains while heparin binds at the basic region at the beginning of the IgII loop [149, 158]. Autoinhibition of FGFRL1 via the IgI loop also occurs and the protein can be post-transcriptionally modified on one of its four glycosylation sites.

As FGFRL1 does not contain a tyrosine kinase domain it is not able to signal in the classical FGFR fashion. Its signalling function is yet to be fully determined, but a number of theories have been postulated. Firstly, the receptor could have an inhibitory effect on FGF signalling by sequestering ligands and therefore preventing them binding to FGFR1–4 [148, 149, 153]. Secondly, FGFRL1 could aid in internalisation and degradation of the classic receptors by binding to the same HS chain as the signalling receptor and effectively dragging it into endosomes/lysosomes. Thirdly, the tandem tyrosine-based motif and histidine-rich sequence could act as a docking site for tyrosine phosphatases, which could act on the signalling receptors and therefore attenuate signalling.

Although signalling mechanisms are yet to be elucidated, FGFRL1 can affect multiple cellular behaviours, inhibiting cell proliferation, increasing cell differentiation, regulating cell–cell contact and inducing cell–cell and cell–matrix adhesion. FGFRL1 is often found at the site of cell–cell contact and it is thought it may mediate cell adhesion by interacting with HS expressed on other cells [157]. FGFRL1 is also thought to play a role in craniosynostosis diseases and mutations in the protein have been found in ovarian cancer [151, 159].

Although not itself a receptor tyrosine kinase, it is clear that FGFRL1 plays an important role in FGF/FGFR signalling. Though full understanding of this role is yet to be determined, it is important to consider this fifth member of the FGFR family when understanding the complexity of FGF signalling.

6.11 Disease

As discussed for each receptor individually, both germ line and somatic FGFR mutations are known to play a role in a range of diseases, most notably craniosynostosis dysplasia and cancer (Table 6.1). Given the ability of the FGF signalling pathway to initiate cell survival and proliferation, amongst other cellular responses, it is not surprising this pathway is hijacked in cancer cells. Mutations in FGFRs in cancer are generally indicative of a more malignant phenotype. The majority of these mutations are activating, resulting in increased proliferation, migration and angiogenesis. However, recent data suggest that loss-of-function FGFR mutations may play a role in the development of some cancers [160, 161].

Table 6.1 Disease associated with FGFR1–4
Full size table

In cancer, chromosomal translocations lead to expression of constitutively active fusion proteins in which the FGFR tyrosine kinase domain is fused downstream of a constitutive dimerisation domain from a fusion protein. This has been seen in myeloproliferative syndromes (MPS), amongst other malignancies [162]. FGFR1 fusion proteins are known to cause 8p11 myeloproliferative syndrome (EMS), a form of MPS [163]. These fusion proteins are known to cause constitutive tyrosine kinase activation of FGFR1 while also leading to signalling independent of FRS2. Fusion proteins containing the FGFR3 kinase domain are also associated with multiple myeloma and peripheral T-cell lymphoma [33].

Other cancers associated with FGFR signalling deregulation include breast cancer, where FGFR1 and FGFR2 are amplified in approximately 10 and 2 % of breast cancers, respectively [164, 165]. Approximately 10 % of melanoma cases have FGFR2 mutations [160]. Interestingly, functional analysis has shown these mutations in melanomas result in loss of function of the receptor. The mutation spectrum, characteristic of those induced by UV radiation, includes 20 missense mutations occurring at conserved residues in FGFR2. Receptor loss of function due to this mutation is caused by loss of ligand binding affinity, impaired receptor dimerisation and decreased kinase activity.

FGFRs can also be involved in cellular transformation by interacting with other proteins. For example, in epithelial ovarian cancer (EOC), Neural Cell Adhesion Molecule (NCAM) is unregulated and promotes malignancy via interaction with FGFR [166]. However, as NCAM is known to inhibit FGF2–FGFR binding [167], NCAM/FGFR interactions increase malignancy via inhibition of normal FGF–FGFR interactions. FGF2/FGFR and NCAM/FGFR interactions therefore stimulate different receptor-mediated responses in EOC; NCAM/FGFR leads to increased cell migration, while FGF2/FGFR leads to increased proliferation. It is also possible the varying receptor interactions cause differential regulation of receptor trafficking to the endosome, resulting in different cellular responses and signalling kinetics [168]. Mouse models have shown that targeted abolition of NCAM/FGFR interaction with a monoclonal antibody results in elimination of metastatic dissemination of EOC. This has been shown via NCAM/FGFR1 interaction studies. However, as NCAM binds FGFR2 and FGFR4 [169], interaction of NCAM with multiple FGFRs may increase malignancy of EOC [166].

The high rate of FGFR mutation in a range of diseases makes this family of proteins a potential therapeutic target. Numerous studies have shown the benefits of FGFR knockdown and inhibition in cancer cell lines where the result is, for example, a decrease in cell proliferation [170]. However, translating this into a therapy for patients has proven difficult. Even specific FGFR inhibitors have off-target effects.

The most clinically advanced FGFR inhibitors to date are mixed kinase inhibitors, targeting the kinase domain of receptors to prevent downstream signalling. These include Dovitinib [171] and SU6668 [172]. However, their anti-FGFR activity is often weak, leading to investigation of more potent FGFR inhibitors. One such inhibitor currently in phase I clinical trials is AZD4547 [173]. This pyrazoloamide derivative targets FGFR1, 2 and 3 and resulted in cell growth inhibition versus cancer cell lines with known FGFR mutations and induces apoptosis. However, even this inhibitor has off-target effects, e.g. selectivity against VEGFR2 (also known as Kinase insert Domain Receptor, KDR), Insulin-like Growth Factor (IGF), PI3K and AKT, although this off-target inhibition is much lower than that of FGFRs. Such inhibitors still need more investigation, but the possibilities of potential FGFR inhibition are an exciting field of cancer therapeutics.