Crosstalk between Fibroblast Growth Factor (FGF) Receptor and Integrin through Direct Integrin Binding to FGF and Resulting Integrin-FGF-FGFR Ternary Complex Formation

Fibroblast growth factors (FGFs) play a critical role in diverse physiological processes and the pathogenesis of diseases. Integrins are involved in FGF signaling, since integrin antagonists suppress FGF signaling. This is called integrin-FGF crosstalk, while the specifics of the crosstalk are unclear. This review highlights recent findings that FGF1 directly interacts with integrin αvβ3, and the resulting integrin-FGF-fibroblast growth factor receptor (FGFR) ternary complex formation is essential for FGF1-induced cell proliferation, migration and angiogenesis. An integrin-binding defective FGF1 mutant (Arg-50 to Glu, R50E) is defective in ternary complex formation and in inducing cell proliferation, migration and angiogenesis, while R50E still binds to the FGF receptor and heparin. In addition, R50E suppressed tumorigenesis in vivo, while wild-type (WT) FGF1 enhanced it. Thus, the direct interaction between FGF1 and integrin αvβ3 is a potential therapeutic target, and R50E is a potential therapeutic agent.


Introduction
It has been proposed that biological processes of all multicellular organisms require signals from growth factor receptors and those from adhesion receptor integrins [1].A family of FGFs and their receptors regulate numerous physiological, as well as pathological processes, including cell proliferation, survival, motility, differentiation and angiogenesis [2][3][4].These biological activities induced by FGFs require integrins [5,6].The integrin family of cell adhesion receptors mediates cell-extra cellular matrix and cell-cell interaction and also transduces signals [7].It has been believed that integrins and growth factor receptors can independently propagate intracellular signals, but they interact with one another on multiple levels on the pathway [8].In current models of integrin-growth factor crosstalk, integrins bind to extracellular matrix proteins, and growth factor receptors bind to growth factors.Additionally, two separate signals merge with one another in multiple levels inside the cells [1,8,9].In contrast to current models, we discovered that FGF1 directly binds to integrin αvβ3 and induces the FGFR1-FGF1-integrin αvβ3 ternary complex [10].In this review, we will summarize recent findings on the fibroblast growth factor receptor (FGFR)-integrin crosstalk, in which FGF directly interacts with integrin αvβ3.
Among the FGF family members, FGF1 and FGF2 are prototypic FGF and have been extensively characterized.They were originally isolated from the brain and pituitary as mitogens for cultured fibroblasts [23,24].FGF1 has been implicated in a range of physiological processes, including development, wound healing, and cardiovascular changes [3].Surprisingly, FGF1 knockout mice are viable and normal under standard laboratory conditions, the physiological role of FGF1 is not clear [25].FGF2 knockout mice are also viable, but have decreased vascular smooth muscle contractility and displayed cerebral cortex defects [26,27].

Signal Transduction of FGF and FGFR
FGFs mediate their biological responses as extracellular proteins by binding to and activating the cell surface FGF receptors (FGFRs).The FGFRs consist of four structurally related receptor tyrosine kinases (FGFR1-4); in addition to these genes, alternative splicing variants of FGFR1-3 provide variations for ligand specificity [28,29].FGFRs are composed of three extracellular immunoglobulin domains (D1-D3), the transmembrane domain and the cytoplasmic tyrosine kinase domain [2,29,30].The FGF ligands bind the D2-D3 fragment of the FGFR, which results in the receptor dimerization and subsequent autophosphorylation of specific tyrosine residues within the intracellular domain.This leads to the activation of intracellular signaling cascades.The four main signaling pathways downstream of receptor activation are (1) the Janus kinase/signal transducer and activator of transcription (Jak/Stat), (2) phosphoinositide phospholipase C (PLCγ), (3) phosphatidylinositol 3-kinase (PI3K) and (4) mitogen-activated protein kinase/extracellular signal-regulated kinase (MAPK/Erk) [3,4,11].
Binding specificity between FGFs and FGFRs is determined by amino acid sequence differences of both 18 FGFs and seven cardinal FGFRs (FGFR1b, 1c, 2b, 2c, 3b, 3c and 4), by spatial and temporal expression patterns of these proteins and, also, is regulated by co-factors, such as heparan sulfate glycosaminoglycan (HSGAG) and cell adhesion molecules.Intriguingly, FGF1 binds to all known cell-surface FGFR isoforms [3,31].

Role of FGF and FGFR in Tumorigenesis, Tumor Angiogenesis, and Inflammation
The underlying mechanisms of FGF signaling in tumors are classified into four types: (1) activating mutations of FGFR; (2) FGF or FGFR overexpression due to gene amplification; (3) chromosomal translocation-derived FGFR fusion proteins and (4) FGF ligand-dependent autocrine or paracrine signaling [32,33].Through these mechanisms, deregulated FGF signaling facilitates tumor development by affecting downstream biological processes, cell proliferation, migration, invasiveness, and tumor angiogenesis [32].The FGFR aberrations mentioned above can be observed in many cancer tissues, such as mammary gland, bladder, prostate, endometrial, lung and brain [33].
FGFs are potent mitogens for cancer cells.In fact, high expression of FGFs has been reported in several tumors.More than 80% of prostate cancer cells express FGF8, and the levels of FGF8 expression correlate with the levels of invasiveness [34].FGF19 is also expressed in prostate cancer and acts in an autocrine fashion, although it is categorized into endocrine FGF [35].In breast cancer cells, cells that overexpress FGF1 or FGF4 grow faster than cells with low FGF expression in vivo [36].Amplification of FGF3 is detected over 20% of the primary breast carcinomas and shows a significant relationship to lymph node metastases [37,38].In ovarian adenocarcinoma, the levels of amplification of FGF1 at chromosome 5q31 correlate with those of FGF1 mRNA and protein expression and the number of CD31-positive blood vessels [39].The levels of FGFR expression also correlate with the invasiveness of cancer [40].FGF1/FGFR-1 signaling (both autocrine and paracrine loops) thus potentially plays a critical role in cancer progression.Because FGF signaling enhances multiple biological processes that promote tumor progression, it is an attractive therapeutic target, particularly since therapies targeting FGF receptors and/or FGF signaling can affect both the tumor cells directly and tumor angiogenesis.
It should be noted, however, that the role of FGF signaling in tumor progression is still controversial.In some tumors, FGFR expression is downregulated, suggesting that FGFRs can act as tumor suppressors.FGFR2 is downregulated or lost in thyroid tumors [41], salivary gland tumor [42] and most hepatocellular carcinoma cell lines and tissues [43].FGFR4 is also downregulated in fibroblasts in pulmonary adenocarcinoma [44].These FGFRs might play a role in the maintenance of homeostasis rather than tumorigenicity; however, it remains unknown how FGFRs negatively affects tumorigenesis and the invasiveness of tumor cells.
FGF plays a role in pathological angiogenesis in inflammatory diseases, as well.Transient exposure to FGF1 upregulates the expression of the cell adhesion molecules, ICAM (intercellular adhesion molecule)-1 and VCAM (vascular cell adhesion molecule)-1, in endothelial cells and increases polymorphonuclear leukocyte adhesion and transendothelial migration [45].

Integrin-Growth Factor Receptor Crosstalk
Integrins are a family of cell adhesion receptors that recognize extracellular matrix ligands and cell surface ligands [46].Integrins are transmembrane α-β heterodimers, and at least 18 α and 8 β subunits are known [7].Integrins serve as transmembrane mechanical links between the extracellular contacts and the cytoskeleton by the actin-based microfilament system.Multiple submembrane linker proteins connect the cytoplasmic domains of integrins to the cytoskeleton, resulting in complex interactions [46].In addition to the function in cell adhesion, integrins are involved in signal transduction upon ligand binding, and their functions are, in turn, regulated by signals from within the cell [7].Ligation of integrins triggers a large variety of signal transduction events that contribute to modulate cell proliferation, survival/apoptosis, shape, polarity, motility, gene expression and differentiation.
Integrins and growth factor receptors interact with each other at several levels in the pathway: (1) integrins control growth factor receptor activity by recruiting of adaptor proteins proximal to growth factor receptors; (2) integrins modulate growth factor receptor activity by changing their localization to focal contacts, where signaling molecules are accumulated; (3) growth factor receptors control integrin-dependent cell adhesion and motility by regulating binding affinity to ligands; (4) growth factor receptor regulates the integrin-associated signaling proteins, such as FAK, Src and PI3K, and regulates the downstream regulatory proteins, such as ERK, Akt, JNK and the Rho GTPases [8,9].These collaborative signaling events include direct or indirect physical interaction between integrins and growth factor receptors [47,48,58,59].
Several growth factors have been reported to act as direct ligands for integrins (Table 1).For instance, FGF1 directly binds integrin αvβ3, and the interaction is required for fibroblast proliferation and migration [10].IGF1 directly binds to integrin αvβ3 and α6β4 [60] and serves to regulate IGF signaling in anchorage-independent conditions [61].Furthermore, neuregulin-1, an EGF family member, binds to integrin αvβ3 and α6β4 and, thereby, induces integrin-neuregulin-1-ErbB3 (a member of the EGF receptor family, a specific neuregulin-1 receptor) complex [62].Vascular endothelial growth factor-C (VEGF-C) and VEGF-D binds α9β1 and induced α9β1-dependent Erk and paxillin phosphorylation [63].Additionally, VEGF-A binds integrin α3β1 and α9β1 and contributes to angiogenesis [64,65].Activation of TGF-β depends on the binding of latent TGF-β to integrin αvβ6 or integrin αvβ8 [66,67].Other examples include nerve growth factor (NGF), neurotropin-3 (NT3) and fractalkine, a transmembrane chemokine (Table 1).Ligand-binding function of integrins plays a role in integrin-FGF receptor crosstalk.Blocking antibodies for integrin αvβ3 diminishes FGF2-induced vascular growth and endothelial cell migration, suggesting that the FGF signaling cooperates with integrin αvβ3 in angiogenesis [53,55].How do FGFs and integrins crosstalk?It has been reported that immobilized FGF2 facilitates endothelial cell adhesion through direct interaction with integrin αvβ3 [73] and induces endothelial cell proliferation, motility and recruitment of FGFR1 at the cell-substratum contact [78].However, because heat-denatured FGF2 still supports integrin binding [78], it is unclear whether this interaction is biologically relevant or how integrins interact with FGF2.It has also been reported that FGF2 binds to fibrinogen, which induces endothelial cell proliferation and subsequent co-localization of αvβ3 and FGFR1 [79,80].However, because FGF1 does not bind to fibrinogen [81], fibrinogen mediated FGF-integrin crosstalk cannot be applied to FGF1 and, also, perhaps, not to other members of the FGF family.Thus, there may be an alternative model by which FGF1 crosstalks with the integrins.

Crosstalk between Integrin αvβ3 and FGFR through Direct Binding to FGF
Computational simulation of FGF1 and integrin αvβ3 interaction predicts direct interaction of FGF1 and αvβ3 (Figure 1a).FGF1 directly binds to soluble and cell-surface integrin αvβ3 (Kd about 1 μM).Antagonists to αvβ3 (mAb 7E3 and cyclic RGDfV) block this interaction [10].The CYDMKTTC sequence (the specificity loop) within the ligand-binding site of β3 plays a role in FGF1 binding (Figure 1b), suggesting that FGF1 binds to a binding site common to other αvβ3 ligands (classical RGD-binding site) [10].The integrin binding site in FGF1 is distinct from the FGFR-binding site (Figure 1c).The structure of the FGF1-FGFR complex (Protein data base (PDB) code 1E0O) was superimposed on the FGF1-integrin αvβ3 complex model, resulting in the putative integrin-FGF-FGFR ternary complex structure (Figure 2a).Thus, this model predicts that the integrin-FGF-FGFR ternary complex can be generated with little or no steric hindrance.In fact, wild-type (WT) FGF1 induces ternary complex formation, which is detected by co-precipitation of FGFR and integrin αvβ3 [82].We propose a model in which integrin and FGFR bind to FGF1 simultaneously and make a ternary complex on the cell surface upon FGF1 stimulation (Figure 2b).
We generated a point mutant (Arg at position 50 in FGF1 to Glu, R50E) within the predicted integrin-binding interface based on the docking model.R50E is defective in integrin binding, but normal in heparin and FGFR binding [10].Interestingly, R50E is defective in DNA synthesis, cell proliferation, cell migration and chemotaxis, suggesting that the direct integrin binding to FGF1 is critical for FGF signaling, while it binds to FGFR1 and heparin [10].Interestingly, R50E is defective in ternary complex formation, as predicted from the model [82], suggesting that the process is critical for FGF signaling.
It has been reported that the extracellular domain of integrin αvβ3 can form both bent and extended conformations based on electron microscopy and the crystal structures of integrins [83,84].It has been proposed that a high affinity state is related to an extended conformation with an open headpiece [85].It has also been reported that most of the ligand (fibronectin) bound integrin complexes are in a bent conformation [86].In extended integrins, the ligand-binding site reaches up to approximately 200 Å from the plasma membrane [83].It is unclear if integrin αvβ3 in the ternary complex has an extended conformation, since the FGF binding site in FGFR is estimated to be located about 50 Å from the membrane.
ERK1/2 activation can be divided into a transient phase (within three hours of stimulation) and a sustained phase (after three hours) in NIH3T3 cells [87].Sustained ERK1/2 activation is integrin-dependent and crucial to cell cycle entry upon FGF stimulation [55,87].The time-course of ERK1/2 activation shows that WT FGF1 induces both transient and sustained activation of ERK1/2 in NIH3T3 cells [82].In contrast, R50E is defective in inducing sustained ERK1/2 activation, while it induces transient ERK1/2 activation.R50E induces transient activation, but is defective in sustained activation of FGFR1 and FRS2α, as well [82].It is, thus, likely that the inability of R50E to induce sustained ERK1/2 activation is related to its defect in inducing cell cycle entry.It is interesting to study how the ternary complex formation is related to sustained ERK1/2 activation.Docking simulation of the interaction between FGF1 (PDB code 1AXM) and integrin αvβ3 (PDB code 1L5G) was performed as described in [10].1AXM was used for the docking, and 1L5G (headpiece of integrin αvβ3) was used as a receptor.This pose indicates the most stable state of FGF1 when FGF1 interacts with integrin αvβ3.(b) Mutagenesis analysis reveals that indicated basic amino acids within the integrin-binding interface in FGF1 are responsible for the interaction between FGF1 and integrin αvβ3.The specificity loop in the I-like domain of β3 plays a critical role in binding FGF1 to integrin αvβ3.(c) It is noteworthy that the positions of the amino acid residues (E102, Y109, and N110) at the fibroblast growth factor receptor (FGFR)-binding site are distinct from the predicted integrin-binding site.
Figure 2. A model of an FGF-FGFR-integrin ternary complex.(a) FGF1-FGFR2 crystal structure (PDB code 1E0O) was superimposed on the simulated FGF1-integrin αvβ3 complex; (b) a schematic model of the FGF1-FGFR-integrin ternary complex was depicted based on (a).FGF1 and FGFR complexes juxtapose in a symmetric dimer (two-end model).Two receptors contact each other through D2 domains [88].Integrins interact with FGFs that are located outside the FGFR dimer (integrin is shown on only one-side).According to the predicted structure of (a) and the differences in the molecular size between FGFR and integrin αvβ3, integrin αvβ3 may form a bent conformation to bind FGF1.The common signaling pathways of FGFR and integrin are the Ras/Raf/Mek pathway (red), the PI3K/AKT pathway (purple) and the PLCɤ pathway (blue).Complex formation is required for mediating signals to elicit cell proliferation, migration and angiogenesis [82].D1-D3, Ig-like domains, 1 to 3; AB, acid box; TM, transmembrane domain; TK, tyrosine kinase domain; KI, kinase insert; α, integrin αv; β, integrin β3.

R50E Is a Dominant-Negative Antagonist of FGFR
If ternary complex formation (integrin-FGF-FGFR) is required to induce signals, it is predicted that the R50E mutant should compete with WT FGF1 for binding to integrins.Thus, R50E should be antagonistic.We found that this is the case.Excess R50E suppresses DNA synthesis, cell viability, and angiogenesis induced by WT FGF1 [82].We thus conclude that: (1) ternary complex formation is required for FGF signaling; (2) the defect of R50E to bind to integrin and subsequent ternary complex formation is directly related to the functional defect in R50E; and (3) R50E is a dominant-negative mutant (Figure 3).These results suggest that R50E has therapeutic potential [82].Using cancer cells that stably express WT FGF1 or R50E, WT FGF1 markedly enhanced tumorigenesis, and R50E suppressed it [89].Furthermore, R50E suppressed angiogenesis in different assays (tube formation assays, Matrigel plug assays, rat aorta ring assays and chick embryo chorioallantoic membrane (CAM) assays [89]).These results suggest that R50E has translational potential as an anti-angiogenesis and anti-cancer therapeutic agent.Ternary complex formation (integrins-FGF-FGFR) is required for signaling.R50E competes with wild-type (WT) FGF1 for binding to integrins.Thereby, dominant-negative effects were observed when treated cells with an excess of R50E in WT FGF1 induced cell proliferation, migration and, also, angiogenesis.

Potential Advantage of R50E over Antibodies and Kinase Inhibitors
The potential advantage of the FGF1 mutant is that: (1) R50E is highly specific to FGFR1 compared to tyrosine kinase inhibitors, which are selective rather than specific; (2) R50E has high affinity to FGFR1 (Kd 10 −12 M); (3) R50E binds to all FGFR isoforms and, thus, is expected to suppress signaling by other members of the FGF family (indeed, R50E (an FGF1 mutant) suppressed angiogenesis induced by FGF2 [89]); (4) R50E (17 Kd) is much smaller than IgG (180 Kd) and is expected to have better penetrance (access to cancer tissues) than antibodies; and (5) currently used target therapeutics (antibodies and kinase inhibitors) almost always induce resistance after a while.This is partly due to point mutations in antibody epitopes or inhibitor-binding sites.Cancer cells obviously benefit from mutations that block the binding of antagonists.R50E may not induce such mutations in FGFR because R50E and wild-type (WT) FGF1 bind to FGFR exactly the same way, and blocking binding of WT FGF1 (and other members of the FGF family) to FGFR would not benefit cancer cells; (6) FGF may bind to multiple integrins, and therefore, antagonists to multiple integrins would be required for effectively suppressing FGF-integrin crosstalk.In contrast, the R50E mutation is expected to effectively suppress integrin binding to FGF1 regardless of the integrin species involved.

Can Mutant Human Protein Be Used As a Therapeutic Agent?
There is a precedent that a mutant of human protein be used for human diseases.A mutant of human growth hormone (hGH) has been used as an antagonist of the GH receptor in the treatment of acromegaly (Pegvisomant) [90].The Gly-120 of hGH was mutated to Arg (G120R), and this mutant was further modified by poly(ethylene glycol) (PEG)-5000 to elongate half-life.Pegvisomant prevents functional dimerization of the hGH receptor by sterically inhibiting conformational changes within the GH receptor dimers [90].Pegvisomant is generally well-tolerated, with a safety profile similar to that reported in clinical trials, and can effectively reduce IGF1 in patients with acromegaly refractory to conventional therapy [91].

Targeting and Delivery of R50E to Cancer
It would be challenging to use R50E as a potential therapeutic, because exogenous growth factors are usually quickly cleared from circulation.It is expected that free FGF1 or R50E has a short half-life in vivo, and it is critical to stabilize and target R50E in vivo to the tumor tissues in order to effectively suppress FGF signaling.Most nanoparticles are expected to accumulate in tumors, due to the enhanced permeation and retention (EPR) effects, a tumor characteristic as a means to target anticancer agents to tumors [92].This one possibility is to deliver R50E as part of a nanoparticle.The good news is that WT FGF1 has already been tested in vivo and in clinical trials in angiogenic therapy.WT FGF1 has been delivered using two established methods: (1) sustained release [93,94]-WT FGF1 is encapsulated in heparin-alginate microparticles-and (2) plasmid-mediated gene transfer [95].Intramuscular administration of FGF1 plasmid induces localized expression and transient systemic distribution of the plasmid.The WT FGF1 expression plasmid can be repeatedly and safely injected at multiple sites to achieve a broad distribution of FGF1 transgene protein [96].We expect that R50E can be delivered using these methods.We will probably be able to exploit the delivery methods for WT FGF1 for angiogenic therapy to deliver R50E for anti-angiogenic therapy.

Conclusions
Although it has been well-established that integrins are involved in FGF signaling in biological and pathological conditions [6,97,98], the specifics in the crosstalk have not been established.We obtained evidence that FGF1 directly binds to integrin αvβ3, and this interaction is critically involved in FGF signaling.We propose a model in which FGF binding to FGFR induces integrin αvβ3 to the complex, resulting in the ternary complex formation.The integrin-binding defective R50E is defective in signaling and ternary complex formation, while it still binds to FGFR and heparin.We demonstrated that R50E is dominant-negative, as predicted by the model.R50E suppresses signaling induced by WT FGF1 and FGF2 and suppresses tumorigenesis and angiogenesis.R50E has potential as an anti-cancer and anti-angiogenesis therapeutic agent (-FGF1-decoy‖).R50E has several advantages over currently available targeting therapeutics: (1) R50E has high affinity and high specificity to FGFR (in contrast to kinase inhibitors that are selective, but not specific); (2) R50E binds to all FGFR isoforms and suppresses signaling by other members of the FGF family; (3) R50E is much smaller than IgG and is expected to have penetrance better than antibodies; and (4) the R50E mutation is expected to suppress integrin binding to FGF1, regardless of the integrin species involved.

Figure 1 .
Figure 1.Docking simulation of fibroblast growth factor 1 (FGF1)-integrin αvβ3 and identification of the integrin-binding site in FGF1.(a) Model of FGF1-integrin interaction.Docking simulation of the interaction between FGF1 (PDB code 1AXM) and integrin αvβ3 (PDB code 1L5G) was performed as described in[10].1AXM was used for the docking, and 1L5G (headpiece of integrin αvβ3) was used as a receptor.This pose indicates the most stable state of FGF1 when FGF1 interacts with integrin αvβ3.(b) Mutagenesis analysis reveals that indicated basic amino acids within the integrin-binding interface in FGF1 are responsible for the interaction between FGF1 and integrin αvβ3.The specificity loop in the I-like domain of β3 plays a critical role in binding FGF1 to integrin αvβ3.(c) It is noteworthy that the positions of the amino acid residues (E102, Y109, and N110) at the fibroblast growth factor receptor (FGFR)-binding site are distinct from the predicted integrin-binding site.

Figure 3 .
Figure 3.The integrin-binding defective mutant (R50E) is incapable of inducing signals.Ternary complex formation (integrins-FGF-FGFR) is required for signaling.R50E competes with wild-type (WT) FGF1 for binding to integrins.Thereby, dominant-negative effects were observed when treated cells with an excess of R50E in WT FGF1 induced cell proliferation, migration and, also, angiogenesis.

Table 1 .
Direct interactions between growth factors and integrins.