OUP user menu

★ editor's choice ★

Non-canonical fibroblast growth factor signalling in angiogenesis

Masahiro Murakami, Arye Elfenbein, Michael Simons
DOI: http://dx.doi.org/10.1093/cvr/cvm086 223-231 First published online: 1 January 2007


Whereas fibroblast growth factors (FGFs) classically transmit their signals via high-affinity tyrosine kinase receptors (FGFR1–4), recent evidence strongly implicates non-tyrosine kinase receptors (NTKR) or cell-surface FGFR-interacting proteins as important players in FGF signalling. Although NTKR have lower affinity for FGFs in comparison with cognate tyrosine kinase receptors, because of their high abundance they can effectively bind FGFs and produce unique biological effects independent of FGFRs. A prime example of such NTKR is the syndecan family of plasma membrane proteoglycans and, in particular, syndecan-4, which transmits FGF signalling via a protein kinase Cα pathway. Another NTKR, αvβ3 integrin, functions as an FGF signalling modulator by binding both FGF2 and FGFR1. Yet another NTKR, neural cell adhesion molecule (NCAM), can serve as an FGFR ligand and assemble an FGFR signalling complex in the absence of FGFs. Furthermore, N-cadherin, which has been reported to associate with FGFR, appears to activate FGFR in both ligand (FGF)-dependent and ligand-independent manners. Finally, gangliosides are implicated as a co-receptor system of FGFs. The biological consequence of non-canonical FGF signalling tends to be less discernable compared to the canonical FGFR activation because of the overlap between these two pathways; nevertheless, non-canonical signalling is important and sometimes essential for cellular functions. Given the diversity of FGF activities through embryonic development to adult physiology, the existence of the non-canonical signalling system may account for the different cellular response to the FGF input in different biological contexts. In this review, we will discuss recent findings related to non-canonical FGF signalling with emphasis on the endothelial biology and angiogenesis.

  • Angiogenesis
  • Cadherins
  • Growth factors
  • FGF
  • NCAM
  • Signal transduction
  • Syndecan-4

1. Introduction

Fibroblast growth factors (FGFs) and their tyrosine kinase receptors (FGFR) comprise one of the most diverse growth factor signalling families in vertebrates. In mice and humans, 22 FGF ligands and four tyrosine kinase high-affinity receptors have been identified. Their biological activities encompass various aspects of embryonic development and adult pathophysiology.1,2

FGFs are broad spectrum mitogens and stimulate a wide range of cellular functions including migration, proliferation, differentiation, and survival. Although it is reported that human umbilical vein endothelial cells express FGF1, 2, 5, 7, 8, 11, 12, 16, and 18 in vitro, 3 the functional difference among these FGFs remains largely unclear. FGFs are produced by many cell types, and endothelial cells are one of the main targets of ‘angiogenic FGFs’ (FGF1, 2, 4, and 5) that have been shown to have angiogenic activities in vivo.

One characteristic feature of FGF signalling is that the response to an FGF is not only cell type specific, but also variable, depending upon the cell cycle phase, cell differentiation stage, and the magnitude and duration of FGF stimulation even in the same cell type. For instance, FGFs’ role in anti-apoptosis has been accepted in the regular 2D endothelial culture system given their capability of activating phosphoinositide-3 kinase (PI3K)-Akt and Bcl-2. However, when cultured in the 3D-collagen gel system, FGF-induced p38 MAPK activation regulates tubulogenesis by possibly increasing apoptosis.4,5 Moreover, FGFs can induce cell-cycle arrest, proliferation or differentiation depending upon the biological context. The paradigm of FGF signalling being capable of producing different, sometimes opposite outcomes is one of the most intriguing areas of investigation. Underlying mechanisms of these differential FGF responses are extensively discussed in the recent review by Dailey et al.6 To address diverse cellular responses elicited by FGF signalling, we should also consider the role of signalling modulators that are differentially expressed in various tissues and cell types, and have the ability to modify FGF signalling in a variety of cellular contexts.

Recent evidence documents the existence of FGF-induced signalling events that bypass the classic FGF–FGFR interaction. This raises the possibility that the non-canonical pathway, besides modulating FGF signalling, may serve as a backup signalling system for the canonical pathway. In fact, studies using either FGF2 or FGF1/FGF2 null mice did not demonstrate clear vascular phenotypes.7,8 Although at present, the redundancy of the FGF ligand system that shares overlapping receptor binding is thought to contribute to nearly negligible vascular phenotypes observed in these studies, compensation by the non-canonical signalling system has not been carefully examined.

In this review, we define canonical FGF signalling as a signalling event produced by an FGF upon binding to an FGFR, whereas a non-canonical pathway refers to FGF- or FGFR-induced signalling that does not necessarily require FGF–FGFR interaction. Both canonical and non-canonical signalling pathways are not mutually exclusive and likely intersect at the second messenger level.

2. Canonical fibroblast growth factor signalling

In the canonical pathway, FGFs produce their biological effects in target cells by signalling via cell-surface tyrosine kinase receptors. There are four known high-affinity receptors for FGFs: FGFR1 through FGFR4, which are transmembrane proteins containing two or three extracellular immunoglobulin (Ig)-like domains, an acidic region, a heparin-binding motif, a transmembrane domain, and an intracellular tyrosine kinase domain.9 Alternative splicing generates multiple isoforms of FGFRs. Of particular importance is the splicing of the carboxy-terminal half of the Ig domain III, which produces IIIb and IIIc isoforms and determines the binding specificity for different FGFs.10,11 This alternative splicing event is regulated in a tissue-specific manner. In endothelial cells, the predominant FGFR is FGFR1IIIc. However, the expression of the FGFR2IIIc, R3IIIc, and R4 has also been reported3,12 with an in vitro observation that FGFR2 signalling in endothelial cells only contributes to cell migration with sustained Erk1/2 activation.13

FGF binding to an FGFR leads to its dimerization. The receptors’ intrinsic tyrosine kinases are activated, causing phosphorylation of seven tyrosine residues on the receptor. In the case of FGFR1, Tyr653 and Tyr654 are important for the regulation of the receptor’s catalytic activity and are essential for signalling. Phosphorylated Tyr766 has been shown to bind the SH2 domain of phospholipase C-gamma (PLC-γ). A phosphotyrosine-binding (PTB) domain-containing protein FRS2α is a key FGFR substrate that mediates activation of MAKP and PI3K pathways. Tyrosine phosphorylated FRS2α serves as a docking protein by producing binding sites for the assembly of a signalling complex that includes the SH2 domain-containing adaptors Grb2 and Shp2. This complex recruits the guanine nucleotide exchange factor SOS and activates Ras, resulting in the activation of MAPK pathways (Erk1/2, p38, JNK).9 Although the Erk1/2 pathway is involved in cell proliferation and migration, p38 and JNK MAPKs are generally associated with inflammatory or stress responses.14 In endothelial cells, p38 MAPK is involved in FGF2-induced tubular morphogenesis by regulating endothelial cell survival, proliferation, and differentiation.5 FGFR can also activate PLC-γ, thereby stimulating production of diacylglycerol (DAG) and inositol 1,4,5-trisphosphate (IP3). This, in turn, releases intracellular calcium and activates Ca2+-dependent PKCs. However, the functional significance of this pathway in endothelial cells is not yet established.15,16 The activation of the PI3K-Akt cell survival pathway is one of the important biological responses induced by FGF2 in endothelial cells.17 Studies using other cell types demonstrated that Grb2 recruits the docking protein Gab1, which is tyrosine phosphorylated by FGFR, leading to the recruitment and activation of PI3K.18 Other signalling molecules involved in FGF-driven angiogenesis, especially cell migration and tubulogenesis, are members of the Src kinase family.19

Canonical FGF signalling is regulated by feedback systems that fine tune the response via a negative or positive feedback loop.20,21 Members of the Sprouty (Spry) and Sef families, Cbl and MAP kinase phosphatase 3 (MKP3) are negative modulators of FGF signalling, whereas positive regulators include low-molecular-weight protein tyrosine phosphatase (LMW-PTP) and a leucine-rich-repeat transmembrane protein XFLRT. Although Spry is thought to be a general inhibitor of the Ras-MAPK pathway, it behaves differently depending on the activated receptor tyrosine kinase. Overexpression of Spry-1 or Spry-2 in human endothelial cells inhibits MAPK activation induced by FGF2 or vascular endothelial growth factor (VEGF), but not by epidermal growth factor (EGF).22 Spry family members have different binding partners that are effectors of the MAPK pathway such as Grb2, SOS, and Raf1, suggesting that different Spry isoforms coordinately inhibit multiple steps of the signalling pathway. Another negative feedback loop involves MAPK phosphorylation of FRS2α. In addition to tyrosine phosphorylation induced by FGFR, FRS2α is phosphorylated by MAPK on multiple threonine residues in response to FGF stimulation, thereby reducing its binding to FGFR and shutting down FGFR signalling. Prevention of FRS2α threonine phosphorylation potentiates MAPK activation, cell migration, and proliferation, indicating that FGFR signalling is subjected to MAPK-mediated negative feedback regulation.23

3. Non-canonical fibroblast growth factor signalling

3.1 Syndecans

The syndecans, a four-member family of heparan sulfate proteoglycans (HSPGs), comprise a distinct class of non-tyrosine kinase receptors for FGFs. Although the affinity of FGFs for HSPGs is ∼100-fold less than that for FGFRs, syndecans strongly influence the ability of FGFs to induce signalling responses within the cell, in part because of their abundant cell surface expression that compensates for lower affinity.

Syndecans bind FGFs via their extracellular heparan sulfate chains, where particular requisite patterns of sulfation underlie the formation of FGF binding sites. Both 2-O and 6-O sulfations are required for heparan sulfates to bind FGF2, for the syndecan-FGF complex’s association with tyrosine kinase FGFRs, and for FGF-mediated signalling in development and angiogenesis.24,25 Each of the four syndecans contains these sulfation sites of their heparan sulfate chains and bind FGFs, yet there are substantial differences in FGF binding affinities among the syndecans. This is presumably a result of variable post-translational sulfation and the many differences in extracellular amino acid sequences among the syndecan isoforms.

Given that the set of syndecan-binding and activating ligands is far greater than that of the tyrosine kinase receptors, and that the FGFs’ affinities for syndecans are significantly lower than those for FGFR1–4, the precise physiological functions of syndecans have largely eluded classical models of receptor-mediated signalling. Although early models depicted a strictly cooperative relationship between syndecans and tyrosine kinase receptors in FGF signalling, more recent studies have defined independent and distinct FGF-induced signalling events, specific to the syndecan family.

The syndecans were originally identified as co-receptors for FGF tyrosine kinase receptors, and a wealth of evidence supports the notion that syndecans facilitate the formation of a signalling FGF–FGFR complex. To begin, syndecan-4 significantly increases the affinity of FGF2 for FGFR1 in both living cells and in cell-free binding assays; this effect requires close physical proximity between the FGFR and syndecan.26 A heparin-binding domain in the extracellular domain of FGFR1 similarly suggests that syndecan can bind directly to FGFRs.27 Heparan sulfate chains are capable of enhancing the activity of FGF2, further implicating syndecans in the facilitation of complex formation.28

The conditions of ternary complex formation between syndecans, FGFs, and FGFRs remain incompletely characterized, although it is clear that heparan sulfates do not induce a change in FGF conformation.29 Heparan sulfates promote the homodimerization of FGF1 and FGF2,30,31 a likely prerequisite of receptor-dependent FGF signalling. More recently, studies using synthesized oligosaccharides have revealed that overall sulfation in heparan chains correlates strongly with increased ternary complex affinities; surprisingly, these protein associations are largely mediated by nonspecific charge forces, rather than the unique structural details of the heparan chains.32 Whether such forces similarly define FGF interactions in living cells has yet to be determined.

Beyond the extracellular heparan chains’ contribution to FGF signalling, the syndecans owe much of their signalling capabilities to their ability to bind and activate proteins in the cytoplasm. The intracellular tail of all four syndecans contain two conserved regions and one intervening variable domain, which collectively enable binding to a multitude of proteins and phosholipids, (reviewed by Zimmerman33). The importance of this cytoplasmic tail was highlighted in studies of syndecan-4 chimera and deletion mutant expression. These studies demonstrated that absence of the syndecan-4 cytoplasmic tail abrogates FGF signalling, despite an increase in extracellular heparan sulfate mass.34 Furthermore, introduction of syndecan-4 mutants with either a disrupted phosphatidylinositol 4,5-bisphosphate (PIP2) binding domain or a disrupted postsynaptic density 95, disc large, zona occludens-1 (PDZ)-binding domain rendered endothelial cells unresponsive to FGF2, while preserving serum-induced signalling.35

This PDZ-binding domain, found at the C-terminus of all syndecans, binds cytoplasmic proteins including synectin, syntenin, CASK/LIN-2, and synbindin.36 These interactions have not been proven across all syndecans, and their differential affinities likely contribute to isoform-specific signalling. Among these syndecan-binding partners, synectin has particular importance in the context of angiogenesis because of its role in regulating the pro-migratory GTPase, Rac1.

During angiogenesis, active Rac1 is found at the leading edge of migrating endothelial cells, where it orchestrates actin rearrangements required for lamellipodia formation and directional movement. Upon syndecan-4 clustering by antibodies, Rac1 is activated, endothelial cells form membrane protrusions, and subsequently undergo directional migration.37 Syndecan-4 binds to the extracellular matrix component fibronectin, and this protein has also been used to isolate the effects of syndecan-4 from those of the tyrosine kinase receptors. Studies using fibronectin to cluster syndecan-4 corroborated the antibody-derived reports and similarly describe syndecan-4’s ability to independently activate Rac1 and induce cell migration.38 In endothelial cells lacking synectin, syndecan-4, or a functional syndecan-4 PDZ-binding domain, baseline levels of active Rac1 are high and this GTPase is not properly localized to the leading edge.37,39,40 Such findings suggest that in an unclustered state, syndecan-4 suppresses Rac1 activation and cell migration, likely via synectin. In mice, a knockout of synectin is associated with a smaller heart size and reduced volume and complexity of arterial vasculature that may reflect abnormal Rac function.40

Another key signalling event controlled by syndecan-4 is the activation and localization of protein kinase Cα (PKCα).41 This presumably occurs upon direct PKCα binding to the syndecan-4 cytoplasmic tail, and requires the binding of PIP2.42 The consequences of this in endothelial cells include the increased angiogenic activity and activation of endothelial nitric oxide synthase (eNOS).43,44 The implications of PKCα activation by syndecan-4 are likely to be further reaching, given the protein’s broad signalling capabilities; the downstream signalling processes remain a field of active investigation.

Finally, syndecan-4 and other HSPGs are also intricately involved in the endocytosis of FGFs.45,46 Syndecan-4 expression is upregulated in response to FGF2 treatment,47,48 a potential mechanism of signalling complex removal from the membrane. Depending upon the endocytic route, this may result in prolonged FGF signalling or complex degradation. To this effect, early studies characterized FGF’s endocytic fate and signalling potency as processes dependent upon whether it is bound to HSPGs or tyrosine kinase FGFRs.49 At present, the relation between FGF signalling and endocytosis is further complicated by an emerging body of evidence demonstrating the eventual translocation of both the ligand and tyrosine kinase receptor to the nucleus (reviewed by Stachowiak50). This discovery is particularly intriguing because transport to the nucleus appears to be critical for FGF2’s mitogenic effects.51 How this transport occurs, under what conditions, and which mechanisms are employed by nuclear FGF for its mitogenic effects remain undefined.

3.2 Integrins

Perhaps the most critical link between extracellular matrix proteins and intracellular signalling is the family of integrins, a diverse group of proteins that are receptors for extracellular matrix molecules. Named for their ability to integrate signals from outside the cell, integrins form distinct non-covalent pairings of one alpha and one beta component. At least 18 alpha subunits and eight beta subunits have been identified to date, and each alpha–beta combination is able to bind to unique matrix components.52

The fibronectin-binding α5β1 integrin dimer upregulates FGF2 expression, and secreted FGF2 binds directly to αvβ3 integrin, stimulating endothelial cell spreading and adhesion.53,54 αvβ3 integrin is expressed on the surface of endothelial cells during angiogenesis, and its blockade by antibodies causes vessel regression in several models.55

Classically defined integrin ligands include extracellular matrix components such as fibronectin, laminin, and collagen. However, integrins also influence the signalling capabilities of extracellular growth factors such as platelet-derived growth factor-BB (PDGF-BB) via direct association with their receptors.56 Furthermore, FGFR1 is associated with αvβ3 integrin, and the formation of an integrin–FGFR signalling complex appears to potentiate FGF2 signalling, particularly in the activation of Erk1/2.57 Given the similarly observed cross-talk between integrins and syndecans,58,59 it appears that signalling specificity among these receptor types and the FGFRs is determined not exclusively by each receptor’s downstream signalling repertoire, but instead by the receptor composition and spatial dynamics of individual signalling complexes.

3.3 Neural cell adhesion molecule

Neural cell adhesion molecule (NCAM) belongs to the Ig superfamily of CAMs, which contains a combination of Ig-like and fibronectin type III (FNIII) domains. This family of CAMs has traditionally been viewed as mediators of both homophilic and heterophilic adhesion; however, recent research has extended this view and demonstrated that CAMs can function as signalling receptors.60,61 Abundantly expressed in the nervous system, NCAM is known to play an important role in neurite outgrowth due to its ability to initiate intracellular signalling via homophilic (NCAM–NCAM) and heterophilic (NCAM–FGFR) interactions. Notably, cross-linking NCAM by specific antibodies triggers signalling, providing evidence that NCAM cis-interactions are sufficient for carrying out at least part of NCAMs biological function.62,63

NCAM has a number of heterophilic binding partners that can affect the signalling process promoting neurite growth, among which the most characterized structurally and functionally is FGFR.64,65 A peptide derived from the NCAM FNIII region can bind FGFR1 directly and stimulate FGFR1 phosphorylation, leading to neurite outgrowth and neuronal survival in primary rat neurons.66 In PC12 cells, NCAM requires both FGFR and Fyn, a member of Src family kinases, for promoting neurite growth. Specifically, the NCAM–FGFR interaction activates PLC-γ and DAG lipase to generate arachidonic acid, elevating intracellular calcium levels and activating Ca2+-dependent PKCs. On the other hand, NCAM-mediated activation of receptor protein tyrosine phosphatase α (RPTPα) dephosphorylates Fyn C-terminal regulatory tyrosine in lipid rafts, leading to Fyn and Fak activation. These two signalling pathways seem to converge on the Erk1/2 MAPK pathway to promote neurite outgrowth.61

Although the role of NCAM-induced FGF signalling is well established in the nervous system, information regarding biological effects of NCAM in other systems is largely limited at this point. NCAM expression, however, is reported to occur elsewhere such as in the muscle and endocrine system.67,68 Moreover, NCAM is upregulated in a variety of physiologic and pathologic conditions including tissue regeneration and malignancies.6971 In comparison to neurons which abundantly express NCAM, its expression in endothelial cells is not uniform and demonstrates a dramatic variability in different vascular beds. NCAM is detected in human dermal microvascular endothelial cells (HDMEC) and corneal endothelial cells (HCEC); however, this is not the case for endothelial cells from larger vessels such as aorta, umbilical vein, pulmonary artery, and pulmonary vein.72,73 Renal tumour-derived endothelial cells express NCAM and can induce capillary-like structures in an in vitro angiogenesis assay upon stimulation with C3d, an NCAM mimetic peptide. This angiogenic process is in part mediated by the NCAM–FGFR1 interaction and concomitant FGFR1 phosphorylation. Interestingly, normal endothelial cells transiently express NCAM during VEGF-induced capillary formation in vitro, suggesting a possible role of NCAM in the vascular formation.74 Moreover, NCAM is suggested to play a regulatory role in neovascularization processes that occur in conjunction with tissue regeneration. In fact, NCAM is upregulated in muscle wound repair which is normally accompanied by angiogenesis, muscle cell regeneration, and re-innervation. NCAM expression occurs on the surface of muscle cells in response to the formation of neuro-muscular junctions.69 Given the similarity of axon guidance and angiogenesis that often occur in parallel during embryogenesis and wound healing, it is also possible that axonal and vascular growth are coordinately regulated by molecules like NCAM. Adenovirus-mediated gene delivery of either FGF2 or FGF6 in rat muscle wound generates enhanced angiogenic and myogenic responses that are characterized by high NCAM expression in myotubes, implying possible FGF–NCAM interplay during the repair process.75

In contrast, loss of NCAM in tumours has been implicated in development of a metastatic phenotype due to reduced adhesive function mediated by NCAM. Interestingly, NCAM has been reported to form a signalling complex that contains FGFR4 and N-cadherin. This complex stimulates FGFR4 kinase activity in pancreatic tumour cells, leading to β1-integrin-mediated cell-matrix adhesion and decreased tumour cell mobility.76

3.4 N-cadherin

Cadherins are a class of transmembrane adhesion proteins that engage in homophilic binding in a calcium-dependent manner, and play an important role in the formation of adherens junctions. They also modulate the actin cytoskeleton dynamics through coupling with catenins, and, importantly, regulate cell functions by adhesion-dependent signalling.77 This involves the activation of Rho family GTPases, regulation of β-catenin availability for participation in Wnt signalling, and modification of receptor tyrosine kinase functions.78 Different members of the cadherin family are expressed in a cell type-specific manner, and endothelial cells, like other cell types, have multiple cadherins including VE-, N-, and T-cadherin. Among them, VE-cadherin is expressed exclusively in endothelial cells and is essential for a variety of endothelial functions including cell survival, contact inhibition of cell growth, shear stress sensing, and angiogenesis.79,80

VE-cadherin is one of the earliest markers expressed in endothelial progenitors, and mouse embryos that are null for Cdh5, the gene encoding VE-cadherin, die in utero within 9.5 days after fertilization, showing significant defects in vascular development.81 In these mice, the early phase of vascular formation such as assembly of endothelial cells in vascular plexi appears to occur normally, although subsequent remodeling and maturation are severely affected, manifested by vessel collapse, regression, and hemorrhage. Furthermore, VE-cadherin is reported to associate with VEGFR, thereby controlling VEGFR internalization and signalling.82

In contrast to the established role of VE-cadherin, the function of N-cadherin in endothelial cells is still elusive despite its abundant expression. Cdh2 (N-cadherin) null mice die by Day 10.5 of gestation. The most notable defect seen in these mice is dissociation of cardiomyocytes, which is followed by impaired heart tube formation. Intriguingly, endothelial specific deletion of N-cadherin in mice results in embryonic lethality due to severe vascular defects at mid-gestation before mural cell investment, raising the possibility that N-cadherin also plays an important role in the vascular development.83

In endothelial cells, N-cadherin is diffusely distributed on the cell surface while VE-cadherin assumes a junctional localization, suggesting a less significant involvement of N-cadherin in homophilic endothelial cells binding.84 Instead, N-cadherin appears to mediate mural cell (pericyte or smooth muscle cells) adhesion to endothelial cells during angiogenesis, thereby contributing to vessel maturation and stabilization.85,86 Underlying mechanisms to promote this endothelial–mural cell interaction are not well studied; however, a recent report strongly implicates involvement of sphingosine 1-phosphate (S1P), a sphingolipid metabolite found in high concentrations in platelets and blood. S1P stimulation of endothelial cells results in S1P receptor-mediated activation of Rac1, leading to N-cadherin trafficking to plasma membrane and establishment of the N-cadherin–catenin complex formation.87 Importantly, inhibition of N-cadherin expression with small interfering RNA profoundly attenuates the process of vascular stabilization in vitro and in vivo.

The FGF system is involved in N-cadherin function through direct interaction of FGFR with N-cadherin. At least a significant portion of N-cadherin-mediated biological processes, such as axonal growth and guidance, tumour metastasis and angiogenesis, is perturbed by blocking of FGFR function.88 Besides heterophilic interaction supporting tumour cell invasion, N-cadherin in cancer cells modifies FGFR signalling by inhibiting the ligand-induced internalization of cell-surface FGFR. N-cadherin retains FGFR1 at the membrane, resulting in a sustained Erk1/2 activation and MMP-9 upregulation, which supports tumour invasion.89

The acid box between IgI and IgII domain of the FGFR extracellular part has been recently identified as a common binding motif of FGFR to N-cadherin and NCAM. However, neither deletion of this motif in the rat skeletal muscle cell line (L6 cells) nor overexpression of N-cadherin in NIH3T3 cells affects the ability of FGF2 to induce Erk1/2 activation.90 The FGFR-N-cadherin interaction is consistently observed, but co-localization of these molecules is not detected at the cell–cell contact of NIH-3T3 cells, although N-cadherin-based junctions are formed in these cells. This suggests that the role played by FGFR–N-cadherin interaction is different from that mediating homophilic interaction at adherens junctions.

In endothelial cells, attenuation of N-cadherin function by antagonistic peptides unrelated to VE-cadherin function results in the impairment of cadherin-mediated endothelial interaction, leading to induction of apoptosis. This effect appears to be mediated by FGF signalling since the N-cadherin inhibitory peptide reduces FRS2 phosphorylation, and exogenous addition of FGF2 completely rescues the phenotype.91

Several lines of evidence point to an N-cadherin role in neovascularization. In non-small-cell lung cancer patients, the frequency of hypervascular tumours is significantly higher for N-cadherin-positive carcinomas, suggesting that N-cadherin expression may drive tumour angiogenesis.92 Furthermore, soluble N-cadherin, which naturally occurs as a cleaved product, has been shown to promote angiogenesis in vivo.93 Soluble N-cadherin stimulates endothelial migration and Erk1/2 activation in vitro, which is likely to be FGFR-dependent, but cell–cell adhesion-independent. Finally, FGF2-induced angiogenesis, and especially vessel maturation, is severely affected by N-cadherin depletion in vivo, although VEGF-induced vascular sprouting is not perturbed.87 Taken together, it is reasonable to conclude that N-cadherin plays an important role in regulation of angiogenesis in conjunction with the activation of FGFR.

Next, the question arises of whether N-cadherin can induce angiogenesis or other biological processes independently of the classic FGF–FGFR axis. Studies from the nervous system suggest the possibility of ligand-independent FGFR activation. Plating neurons on an N-cadherin substrate promotes neurite outgrowth in the absence of an FGF ligand, but the process is indeed FGFR dependent.64 A line of evidence from cancer biology also suggests such a possibility; mammary tumours which overexpress N-cadherin demonstrate higher basal Erk1/2 and p38 MAPK activities compared with control tumours in mice.94 Moreover, in many experimental settings, soluble N-cadherin or N-cadherin agonistic peptide is sufficient to trigger FGFR-mediated signalling processes in a cell–cell adhesion independent manner, suggesting that a homophilic N-cadherin interaction can bypass FGF ligand-mediated FGFR activation. Thus, together with convincing evidence showing synergistic effects of N-cadherin and FGFR, it is suggested that N-cadherin is involved in both ligand (FGF)-dependent and ligand-independent activation of FGFR.78

4. Other fibroblast growth factor signalling modulators

4.1 Ganglioside

Gangliosides are complex sialic acid-containing glycosphingolipids present on the outer leaflet of plasma cell membrane of eukaryotic cells. They form unique domains in the membrane, broadly termed as glycospingolipid-enriched microdomains (GEM) where many signalling molecules are known to localize. With their unique biological activities, gangliosides have been shown to modify the function of various receptor tyrosine kinases by regulating their ligand binding, receptor dimerization, and receptor subcellular localization.95 Early studies demonstrating direct interaction of FGF and GM1 (monosialotetrahexosylganglioside) have led to the notion that gangliosides are competitive inhibitors of FGF, inhibiting FGFR activation. This model has been countered by a recent study which shows that exogenously added GM1 can restore the ability of the heparan sulfate-deficient cells to respond to FGF2.96 In fact, cell-associated GM1 appears to interact with FGF2, suggesting that gangliosides can act as a functional FGF co-receptor, presenting FGF to its receptor much like HSPGs. In contrast to syndecans, gangliosides are not capable of transmitting a signal without FGF–FGFR interaction. They appear to simply act as a modulator of classic FGF signalling through their influence on the localization and concentration of ganglioside in the plasma membrane, thereby enabling the recruitment of FGFR to a specific membrane domain.

5. Concluding remarks and perspectives

Many different cellular and biochemical mechanisms that can modify classic FGF signalling have been discussed in this review (Figure 1). A growing body of evidence indicates important roles played by non-canonical FGF signalling systems; nonetheless, the relative significance of these to the canonical pathway and their indispensability to the relevant biological processes are still actively debated. Given the distinct distribution of components comprising non-canonical pathways, these pathways are not likely to function uniformly in different cell or tissue types, which may explain the differential FGF response in a various biological contexts. Further investigations are required to elucidate unique properties of these non-canonical signalling pathways.

Figure 1

FGF signal transduction pathways. The canonical FGF signalling pathway (highlighted in purple) is elicited by FGF binding to heparan sulfate and FGFR. The canonical pathway includes FRS2α-dependent (MAPK and PI3K–Akt) and FRS2α-independent (PLCγ–PKC) pathways. Non-canonical pathways are divided into two groups: ligand (FGF)-dependent (right side) and ligand-independent (left side) pathways. Syndecans are able to trigger FGF-induced signalling in an FGFR-independent manner, whereas FGF-induced αvβ3 integrin activation seems to involve FGFR. NCAM and N-cadherin can initiate signalling by heterophilic interaction with FGFR. NCAM is expressed in a subset of endothelial cells and can be upregulated by angiogenic stimuli. NCAM-induced PLCγ activation is FGFR dependent, but Fyn activation is not. N-cadherin mediates endothelial-mural cell adhesion, but involvement of FGFR in this process is not clear. Although N-cadherin interaction with FGFR is sufficient to trigger FGFR signalling, this interaction also results in the FGFR retention on the cell surface, thus increasing the FGFR responsiveness to FGF.

One major challenge inherent in the studies dealing with FGFR interacting proteins is that in most cases cell-based biochemical data relies on an overexpression system in which kinetics of the interaction may not be relevant to normal physiology. As a result, significance of these interactions can be overestimated, or in the worst case scenario, does not reflect the original role. Therefore, despite technical challenges, it is important to conduct experiments in more physiological settings. Another related point is that it will be exceedingly important to extend our understanding of canonical and non-canonical signalling to in vivo models. Cell culture systems provide great advantages in investigating molecular- and protein-based mechanisms; however, they involve the risk of studying an oversimplified and out-of-context situation. Considering the FGF functions that encompass multiple cell types and tissues, findings derived from in vitro studies need to be verified and integrated in studies using in vivo models.

Elucidating how all these canonical and non-canonical FGF signalling pathways operate to control differential cellular response remains a formidable challenge for the future. It will, however, lead to an understanding of the precise mechanism of angiogenesis and, ultimately, the development of new therapeutic applications.


National Institutes of Health (HL62289 and HL53793 to M.S, Funding Agency: National Heart, Lung and Blood Institute), American Heart Association (0615689T to AE).

Conflict of interest: none declared.


View Abstract