Fibroblast Growth Factor Signalling: From Development to Cancer

Nicholas Turner; Richard Grose

In This Article

Deregulation of FGF Signalling in Cancer

In this section we discuss the substantial evidence that supports the existence of aberrant FGF signalling in the pathogenesis of multiple types of cancer. The underlying mechanism driving FGF signalling is largely tumour specific, and can be split into genomic FGFR alterations that drive ligand-independent receptor signalling compared with alterations that support ligand-dependent activation (Fig. 3; Table 1).

Figure 3.

Mechanisms of pathogenic cancer cell FGF signalling. The ways in which fibroblast growth factors (FGFs) and FGF receptors (FGFRs) can be altered in cancer fall into four main groups. a | Genomic alteration of FGFR can occur through three mechanisms, leading to ligand-independent signalling. First, activating mutations can result in ligand-independent dimerization or constitutive activation of the kinase (shown by yellow lightning). Second, chromosomal translocations can also lead to ligand-independent signalling. Intragenic translocations generate fusion proteins, usually with the amino terminus of a transcription factor fused to the carboxy-terminal FGFR kinase domain, resulting in dimerization of the fusion protein and constitutive signalling; for example, FGFR3 is brought under the control of an unrelated promoter, resulting in FGFR3 overexpression. Third, receptor gene amplification, which results in supraphysiological receptor overexpression. FGFR2 overexpression can also be accompanied by altered C-terminal splicing that might contribute to receptor accumulation. b | Establishment of a paracrine loop. Altered FGFR expression on a cancer cell can potentially occur by splicing, which alters FGFR specificity, or by amplification of an FGFR gene to express FGFR out of context, which is activated by FGF (green) expressed by a stromal component. Tumour cells can stimulate stromal cells to release FGF ligands and increase the release of ligands from the extracellular matrix. c | Establishment of an autocrine loop. FGF ligands are produced in an autocrine fashion by a cancer cell (brown). The autocrine loop can be established by FGFR expression out of context or by the increased expression of FGF ligands. d | FGF stromal effects, including angiogenesis. FGF released from stromal cells or cancer cells can act on endothelial cells to promote angiogenesis.

Activating Mutations

The importance of FGF signalling in tumour pathogenesis was highlighted by a screen of more than 1,000 somatic mutations found in the coding exons of 518 protein kinase genes from 210 different human cancers.[41] Of the non-synonymous mutations, components of the FGF signalling pathways were the most commonly mutated.[41]

Bladder cancer has the most established link to FGFR mutations. Overall ~50% of bladder cancers have somatic mutations in the FGFR3-coding sequence,[42] and most of the mutations precisely match the activating germline mutations of thanatophoric dysplasia, a lethal form of dwarfism.[43] Mutations in bladder cancer are strongly associated with non-muscle invasive disease, with 50–60% of non-muscle invasive cancers possessing FGFR3 mutations, and mutations occurring less commonly in muscle-invasive bladder cancers (10–15% of these cancers only). In contrast to the epidermal growth factor receptor ( EGFR ) gene, in which activating mutations occur almost exclusively in the kinase domain, more than half of the mutations in FGFR3 occur at a single position in the extracellular domain (S249C). This mutation leads to the formation of an aberrant intermolecular cysteine disulphide bridge, which results in constitutive dimerization and activation of the receptor.[44,45] Mutations are also commonly found in the transmembrane domain (such as Y373C), as well as less common kinase domain mutations (such as K652E),[46] with Y373C and to a lesser extent K652E constitutively activating the receptor.[44,45]

FGFR3 mutations have also been identified in many other cancer types, including cervical cancers,[47] multiple myeloma, prostate cancer[48] and spermatocytic seminomas[49] (Table 1). A single report identified FGFR3 mutations in oral squamous carcinomas,[50] but this was not confirmed by a follow-up study.[51]FGFR3-activating mutations are also found at a high frequency in the benign skin conditions epidermal nevi[52] and seborrhoeic keratosis,[53] which do not progress to malignancy.[54]

Mutations of FGFR2, which are also frequently extracellular and identical to the activating germline mutations found in craniosynostosis syndromes,[28] have been described in 12% of endometrial carcinomas.[55]FGFR2-mutant endometrial cancer cell lines are highly sensitive to FGFR tyrosine kinase inhibitors,[28] which reflects oncogenic addiction to the mutant-activated FGFR.

Interestingly, FGFR3 and HRAS mutations are mutually exclusive in bladder cancer,[56] but PIK3CA mutations are more commonly found in bladder cancers with FGFR3 mutations.[57,58] Similarly, FGFR2 and KRAS mutations are mutually exclusive in endometrial cancer.[59] This suggests that in these cancers only a single mechanism of activation is required to fully activate MAPK signalling, although oncogenic PI3K signalling can be enhanced by multiple 'hits'.

FGFR Gene Amplifications

In contrast to the activation of FGFR3 by mutation, amplifications of FGFR3 have been described only rarely in cancers.[60] Conversely, amplifications of both FGFR1 and FGFR2 have more commonly been described. Approximately 10% of gastric cancers show FGFR2 amplification, which is associated with poor prognosis diffuse-type cancers.[61,62] Gastric cancer cell lines with FGFR2 amplifications show evidence of ligand-independent signalling and are highly sensitive to FGFR inhibitors,[61,62] although paracrine secretion of FGF7 by fibroblasts may also contribute to cellular proliferation in vivo.[63] Interestingly, in some gastric cancer cell lines amplification of FGFR2 is accompanied by deletion of the most C-terminal coding exon.[64] This results in the expression of a C-terminally truncated receptor, which can also be generated by aberrant splicing in cell lines that lack the C-terminal deletion. This C-terminal FGFR2 truncation interferes with receptor internalization,[65] therefore preventing a potential mechanism of signalling attenuation and contributing to constitutive activation of the receptor.

Amplification of the chromosomal region 8p11–12, the genomic location of FGFR1, is one of the most common focal amplifications in breast cancer,[66–68] and occurs inapproximately 10% of breast cancers, predominantly in oestrogen receptor (ER)-positive cancers.[66]FGFR1 amplifications have also been reported in oral squamous carcinoma[69] and are found at a low incidence in ovarian cancer,[70] bladder cancer[71] and rhabodomyosarcoma[72] (Table 1). In contrast to FGFR2 amplifications, overexpression of wild-type FGFR1 occurs in cancer; it is unclear whether the higher levels of FGFR1 lead to tumours that aberrantly respond to paracrine FGF ligands, such as FGF2, or whether at higher levels of FGFR1 expression ligand-independent signalling occurs. It is also important to note that the 8p11–12 region is gene dense and it is not universally accepted that FGFR1 is the causative oncogene in this amplified region in breast cancer.[73,74] FGFR1 might also be important in breast cancers that lack FGFR1 amplifications, and one study suggested that an FGFR inhibitor blocked the proliferation of non-amplified cancer cell lines by downregulating D-type cyclins.[75]

Chromosomal Translocations in Haematological Malignancies

Some of the strongest evidence linking FGF signalling to oncogenesis has come from the study of haematological malignancies, in which translocations involving the FGFRs have been identified. Several FGFR intragenic translocations have been identified, which typically result in a fusion protein comprising the N terminus of a transcription factor fused to an FGFR kinase domain. This leads to constitutive FGFR dimerization and activation[76–78] (Table 1). A different translocation is found in multiple myeloma: 15% of multiple myelomas harbour a t(4;14) translocation that links FGFR3 at 4p16.3 to the immunoglobulin heavy chain IGH locus at 14q32 (Refs[79,80]). These translocations are intergenic, with the breakpoints occurring ~70 kb upstream of FGFR3, and bring FGFR3 under the control of the highly active IGH promoter. It is important to note that the translocations involving FGFR3 in multiple myeloma also involve the adjacent multiple myeloma SET domain-containing ( MMSET ) gene, and the relative contributions of FGFR3 and MMSET to oncogenesis are subject to ongoing debate.[81] However, the importance of FGFR3 overexpression and mutation in haematological malignancy has been modelled using transgenic mice[82] (Box 3), and t(4;14) myeloma cell lines are highly sensitive to FGFR3 targeting.[83,84]

FGFR3 translocation in multiple myeloma is associated with a poor prognosis and is rarely found in monoclonal gammopathy of uncertain significance, a precursor condition of multiple myeloma, which suggests that FGFR3 translocations promote a rapid conversion to full multiple myeloma.[85] The ultimate effect of the translocation is to overexpress FGFR3 out of context, which might result in aberrant ligand-dependent signalling[86] (with hypersensitivity to ligands by swamping negative feedback and receptor internalization and degradation pathways) or ligand-independent signalling. In a small proportion of t(4;14) multiple myeloma, FGFR3 is also mutated (~5% translocated cases),[87] presumably further reinforcing FGFR3 signalling.[82] Interestingly, FGFR3 translocations also occur mutually exclusively of NRAS and KRAS mutations.[88]

Autocrine and Paracrine Signalling

Most of the genomic aberrations discussed above lead to constitutive receptor activation and ligand-independent signalling. Ligand-dependent signalling is likely to have a similarly important role in the pathogenesis of cancer, through either autocrine production of ligand in cancer cells or paracrine production of ligand from stromal cells that may be expressed physiologically or in response to cancer cells in a 'paracrine loop'. Several mouse models have shown that ectopic expression of FGF can promote cancer. This has been achieved by expressing FGF in either epithelial cells or stromal fibroblasts, which results in the autocrine and paracrine stimulation of cancer cells, respectively (Box 3).

The first strong evidence for autocrine FGF signalling driving human tumorigenesis comes from studies of melanoma, which expresses high levels of FGFR1 and FGF2. The growth of human melanoma xenografts regressed after antisense-mediated inhibition of FGFR1 or FGF2 (Ref.[89]), suggesting that an FGF2–FGFR1 autocrine loop promotes the development of some melanomas. Frequent amplification of FGF1 , resulting in increased FGF1 expression, has also been reported in ovarian cancer and is associated with poor survival.[90] The FGF1 expression levels correlated with microvessel density, suggesting that aberrantly expressed FGF1 functions in a paracrine fashion to promote angiogenesis.[90] Whether FGF1 also functions in an autocrine manner in ovarian cancer to promote tumour cell proliferation or survival is unclear. An autocrine FGF2–FGFR1-IIIc feedback loop has also been reported in non-small-cell lung cancer cell lines that show resistance to the EGFR antagonist gefitinib.[91]

Unequivocal evidence that paracrine FGF released from the stroma functions on human cancer cells to promote tumorigenesis is lacking, principally because it is difficult to model such an interaction in vitro. Increased plasma levels of FGF2 and other FGFs are found in multiple cancer types.[92] This partly reflects the increased release of FGFs: as tumours invade and degrade the extracellular matrix FGF is liberated,[4] freeing it to function as a paracrine growth factor. Tumour cells may also induce FGF2 release from stromal inflammatory infiltrate (reviewed in Ref.[93]), which may promote tumour survival in a classic paracrine loop or promote angiogenesis. The angiogenic response can be further augmented by the establishment of an autocrine FGF2 signal in endothelial cells.[93]

In prostate cancer, several FGFs, including FGF2 (Ref.[94]) and FGF6 (Ref.[95]), are upregulated. Similarly, FGFR1-IIIc is upregulated in poorly differentiated prostate cancers,[94] which suggests the potential existence of a paracrine loop, and FGFR2-IIIb is downregulated. In astrocytomas, a similar increase in FGFR1 expression has been reported in high-grade tumours.[96] However, it is unclear to what extent the expression of FGFR1-IIIc in prostate cancer drives tumour progression, and to what extent the expression of this splice form is a consequence of tumour progression, particularly the epithelial–mesenchymal transition (EMT), which could lead to secondary changes in FGFR splicing and expression.[97] Loss of the expression of negative regulators, including SPRY1 and SPRY2 (Ref.[98]) and SEF[99] can also increase FGF signalling in prostate cancer. It has been proposed that these changes in prostate cancer may result in androgen independence.[100] As in prostate cancer, the expression of FGF1, FGF2 and FGF7 is higher in breast cancer stroma than in normal breast stroma.[101,102]

As mentioned above, increased levels of several FGFs are detected in the serum of cancer patients, but some hormonal FGFs, such as FGF19, signal physiologically through the bloodstream. Transgenic mice expressing FGF19 in skeletal muscle developed hepatocellular carcinomas,[103] which presumably reflects an endocrine action of the increased levels of circulating FGF19 (Box 3). FGF19 was overexpressed in a subgroup of liver, colonic and lung squamous carcinomas.[103] A monoclonal antibody that sequesters FGF19 blocked the growth of colonic cancer cell lines and xenografts that overexpress FGF19 (Ref.[104]), suggesting that the growth of these cells may be driven by an autocrine loop functioning through FGFR4 (Ref.[105]). FGF19 requires β-klotho expression to interact with FGFR4, at least in physiological signalling, however, the expression of β-klotho is still to be established in these tumour types.

Germline Single Nucleotide Polymorphisms

A further link between FGFR signalling and breast cancer has been provided by recent genome-wide association studies that identified FGFR2 as a breast cancer susceptibility gene.[106,107] In pioneering studies, single nucleotide polymorphisms (SNPs) located in the second intron of FGFR2 were found to correlate with an increased risk of developing breast cancer. The 'at risk' minor allele is prevalent, with 40% of the population carrying at least one copy, but the increase in risk is relatively small, increasing risk by 1.26-fold in a heterozygote and 1.63-fold in a homozygote.[106] Interestingly, the SNP seems to appreciably increase the risk of developing ER-positive breast cancer[108] only, with little or no effect on ER-negative breast cancer.

There remains substantial uncertainty regarding how the minor FGFR2 allele increases breast cancer risk and exactly which of the multiple SNPs in the second intron — which are in strong linkage disequilibrium — are mechanistically important. One study suggested that the SNPs result in modest increases in FGFR2 mRNA expression through the modification of a binding site for the transcription factors OCT1 and runt-related transcription factor 2 (RUNX2).[109] Binding sites for ER and OCT1 often cluster together,[110] suggesting a model in which ER and OCT1 cooperate in regulating FGFR2 expression and potentially also cooperate with C/EBPβ to drive transcription.[109] The potential role of ER in determining the functional effects of this SNP may explain the apparent restriction of the risk attributed for this SNP to ER-positive breast cancer.[108]

A second SNP, FGFR4 G388R, does not seem to increase the incidence of cancer, but has been reported to associate with poor prognosis in multiple cancer types, including breast cancer,[111] colon cancer[111] and lung adenocarcinoma.[112] The FGFR4 G388R allele is common, with at least one copy present in approximately half the population. How this SNP influences cancer prognosis is less clear, and it potentially promotes cancer cell motility and invasion[111] or resistance to chemotherapy.[113]


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.
Post as: