Fibroblast growth factor receptor signaling in hereditary and neoplastic disease: biologic and clinical implications

In humans, there are five known FGFRs, called FGFR1–FGFR4 and FGFRL1 (also known as FGFR5). FGFR1–FGFR4 are typical growth factor receptor tyrosine kinases, with extracellular immunoglobulin (Ig)-like domains and intracellular tyrosine kinase domains, while FGFRL1 lacks the intracellular kinase domain and has less clear function(s)[22, 23]. Upon binding of their ligands, the typical receptors homo- or hetero-dimerize, leading to sequential phosphorylation of specific intracellular tyrosine residues and activation of an intracellular signaling cascade and gene transcription [24] (Fig. 1). The FGFR signaling pathway interacts with several other important intracellular pathways, including PI3K/Akt, Wnt, hedgehog, and bone morphogenic protein (BMP) [24, 25]. FGFR1–4 have different ligand specificities based on developmental aspects, tissue distribution, and RNA splicing variation [26]. For example, the FGFR2b isoform is predominantly expressed in epithelial cells, while the FGFR2c isoform is expressed predominantly in mesenchymal cells, and switching from FGFR2b to FGFR2c occurs during progression and invasion of prostate and bladder cancers [27].

There are 18 human ligands for FGFRs (Fig. 1). They are FGF1 (acidic FGF), FGF2 (basic FGF), FGF3, FGF4, FGF5, FGF6, FGF7 (KGF), FGF8, FGF9, FGF10, FGF16, FGF17, FGF18, FGF19, FGF20, FGF21, FGF22, and FGF23. FGF11–14 are not ligands for FGFRs and are known as FGF homologous factors (FHF1–4) [28, 29]. There is no human FGF15 (FGF15 is the mouse equivalent of human FGF19) [30]. The FGFR ligands are secreted proteins that bind to the extracellular matrix, thereby restricting their influence to the tissue where they are produced (autocrine or paracrine function). However, three of the FGFs—FGF19, FGF21, and FGF23—bind less tightly to extracellular matrix heparin sulfates, so they are able to act systemically in an endocrine fashion (hormone-like), allowing them to spread from their production site into the circulation [30, 31]. Most FGFs are secreted proteins with cleavable amino terminal portions, but FGF9, FGF16, and FGF20 have non-cleavable secretion sequences, and FGF1 and FGF2 have no secretion sequences, although they are found in the extracellular compartment, suggesting an alternate process of release [29, 32]. Most FGFs are released from binding to the extracellular matrix via action of heparin sulfate proteoglycans and/or fibroblast growth factor–binding protein (FGFBP1)[33].

In general, FGFs are promiscuous, and most can bind to any of the four main FGFRs, although some of the FGF/FGFR ligand/receptor pairs differ in their affinities (Fig. 1c). FGFs interact with cell surface heparan sulfate proteoglycans (HSPGs) and with the transmembrane protein Klotho (for the hormone-like FGF19, 20, 23) to stabilize binding to FGFRs. Ornitz et al. [34, 35] systemically investigated affinities of FGFs for each FGFR isoform and showed, for example, that FGFR2b is a high-affinity receptor for FGF1, FGF7, FGF10, and FGF22, while FGFR2c has high affinity for FGF1, FGF2, FGF4, FGF6, FGF8, FGF9, and FGF20. But, differential expression of either FGFs or FGFRs in time or tissue may also contribute to tissue-specific effects of FGF/FGFR signaling. For example, FGF1 and FGF2 are expressed in both embryonic and adult tissues, while FGF7–FGF9 are predominantly expressed in developing or in restricted adult tissues [36].

The differences in activity/function of the different FGF/FGFR pairs are also highlighted by murine gene knockout experiments. Mice heterozygous for FGFR knockout mutations develop normally, so haplo-insufficiency is not likely to be a factor [37]. However, mice homozygous for FGFR1 or FGFR2 null mutations die in utero, and FGFR3-null mice develop normally other than overgrowth of cancellous bones and deafness [37]. Most of the FGFs have also been knocked out in mouse models, with varying phenotypic effects, including lethality (FGF4, FGF8, FGF9, FGF10, FGF19, FGF18, and FGF23), defects in neuromusculoskeletal development or function (FGF2, FGF3, FGF6, FGF7, FGF12, FGF14, FGF17, FGF18), cardiac defects (FGF2, FGF9, FGF16, FGF19), and no identifiable abnormality (FGF1) [29].

To signal, FGFs have to be released from the extracellular matrix by heparinases, proteases, or specific fibroblast growth factor–binding protein (FGFBP1). FGFBP1 are secreted heparin proteins that reversibly bind FGF1 and FGF2, releasing them from the extracellular matrix and increasing the local levels of free ligand available for receptor binding [38]. FGFBP1 is secreted by keratinocytes and human epidermal carcinomas, particularly squamous cell carcinomas [33, 38]. Its expression is seen in development, wound healing, cancer, and angiogenesis, and it is downregulated by pharmacologic agents in vitro [33]. The fibronectin-leucine-rich transmembrane protein 3 (FLRT3) is co-expressed with FGF8 during development, co-immunoprecipitates with FGF8/FGFRs, physically interacts with FGFRs via its fibronectin type III domain, and increases activity through the MAPK pathway, facilitating FGFR signaling [39, 40].

FGF/FGFR signaling is negatively regulated by several mechanisms. FGFR stimulation activates sprouty proteins (SPRY1–4), which in turn negatively feedbacks on FGF/FGFR/MAPK signaling by interacting with growth factor receptor bound protein 2 (GRB2), son of sevenless homolog 1 (SoS1), and/or RAF proto-oncogene serine/threonine-protein kinase (RAF1) [41, 42]. Intriguingly, sprouty proteins are thought to be general inhibitors of receptor tyrosine kinase growth factor signaling via inhibition of Ras-MAPK signaling, but their behavior differs among specific growth factor receptors. For example, sprouty protein inhibits FGFR signaling but potentiates epidermal growth factor receptor (EGFR) signaling [43]. The cytoplasmic domain of similar expression to FGF (SEF, also known as IL-17 receptor D) interacts with the cytoplasmic domain of FGFR and inhibits downstream signaling of the pathway [43]. Finally, MAP kinase phosphatases (MKPs) dephosphorylate ERKs and thereby downregulate the signaling pathway. MKP3 functions within the cytoplasm, whereas MKP1 is localized in the nucleus [43].

FGFRL1 is the atypical receptor that lacks an intracellular kinase domain. Like the other FGFRs, it is found in all vertebrates, is expressed in a wide variety of tissues, and binds to some FGF ligands. In vitro experiments show that FGFRL1 binds FGF3, FGF4, FGF8, FGF10, and FGF22, but not the other FGFs tested [44]. Unlike FGFR1–4, it is shed from cell membranes [44], suggesting that it may serve as a ligand trap to negatively regulate signaling [45]. Other regulatory mechanisms might include heterodimerization with other FGFRs to prevent trans-autophophorylation and thereby negatively regulating intracellular signaling or by increasing membrane turnover rates of the other FGFRs, but there is currently no evidence that these mechanisms exist [45]. The exact functions of FGFRL1 are not known. FGFRL1-null mice die in infancy with diaphragmatic defects and renal agenesis [45], and there are case reports of FGFRL1 mutations in human disease: craniosynostosis [46] and ovarian cancer [47].

Several types of inherited syndromes are due to germline aberrations in FGFR. These include craniosynostosis syndromes and achondroplasia [48], mainly due to gain-of-function mutations, as well as loss-of-function anomalies associated with congenital hypogonadotropic hypogonadism (Table 1). Interestingly, although some of these aberrations are identical to those that, in somatic form, are associated with cancer, for the most part, the individuals with these hereditary syndromes have not been reported to have an increased incidence of cancer. Apart from case reports [49‐53], there is no epidemiologic evidence that people with craniosynostosis, dwarfism, or congenital hypogonadotropic hypogonadism are at increased risk for malignancy. This is possibly related to differential effects of activating FGFR mutations in cell or tissue type in the context of stage of development. It is possible, for instance, that there are as yet unknown anti-neoplastic compensatory effects in individuals with these disorders, perhaps stemming from the presence of germline FGFR aberrations during development as opposed to the appearance of somatic FGFR mutations in human cancers.

Somatic or acquired mutations in FGFR3 have been observed in benign skin conditions like seborrheic keratosis and epidermal nevi [54, 55]. These aberrations are activating and, when evaluated, are not seen in adjacent normal skin [55]. Many of these are identical to mutations that are also seen in bladder and/or cervical cancers.

Aberrations in FGFR and its ligands are common in malignancy (Tables 2, 3, and 4 and Figs. 2 and 3). Across malignancies, FGF anomalies are found in ≈14 % and FGFR in ≈7 % of malignancies [56] (FGF/FGFR in about 20 %).

Only four drugs (ponatinib, regorafenib, pazopanib, and more recently lenvatinib) are FDA-approved for use in human cancers (Table 5). Ponatinib is a multi-tyrosine kinase inhibitor that was approved for imatinib-resistant chronic myelogenous leukemia (CML) and Philadelphia chromosome-positive (CP) acute lymphoblastic leukemia (ALL). The efficacy results demonstrated a 54 % major cytogenetic response (MCyR) rate in patients with CP-CML, and seventy percent of patients with CP-CML with the T315I mutation in BCR-ABL achieved MCyR. In addition of targeting BCR-ABL, ponatinib can also inhibit members of the VEGFR, PDGFR, FGFR (IC₅₀ = 2 nM for FGFR1), and SRC families of kinases, KIT, or RET, with IC₅₀ between 0.1 and 20 nM. Of note, ponatinib was briefly taken off the market by the FDA because of concerns about cardiovascular side effects (clotting), but soon after, it was returned to the market with updated safety monitoring recommendations. Regorafenib was approved for imatinib-resistant gastrointestinal stromal tumor (GIST) and metastatic colorectal cancer, based on a statistically significant survival prolongation observed in patients randomized to receive regorafenib (6.4 vs 5.0 months in the placebo arm, P = 0.01). Regorafenib, and its active metabolites inhibit multiple membrane-bound and intracellular kinases including those in the RET, VEGFR1, VEGFR2, VEGFR3, KIT, PDGFR-α and -β, FGFR1-2, and Abl pathways. Pazopanib was approved for advanced renal cell carcinoma (based on a progression-free survival of 9.2 months compared to 4.2 months in the placebo arm), as well as soft tissue sarcomas (improved progression-free survival: 4.6 months versus 1.6 for patients who received placebo). Pazopanib is a multi-tyrosine kinase inhibitor whose targets include VEGFR1-3, PDGFR-α and -β, FGFR-1 and -3, and KIT. Lastly, lenvatinib is a multi-kinase inhibitor (targets including VEGFR1–3, FGFR1–4, PDGFR-α, KIT, and RET) indicated for the treatment of patients with locally recurrent or metastatic, progressive, radioactive iodine-refractory differentiated thyroid cancer. The recent approval in February 2015 was based on an improved progression-free survival (18.3 vs 3.6 months in the placebo group, P < 0.001). The approvals of these four agents (ponatinib, regorafenib, pazopanib, and lenvatinib), all three of which are multi-kinase inhibitors, were not based on activity against FGFR. Of interest, nintedanib is an FGFR inhibitor that is FDA-approved for a non-cancer indication—idiopathic pulmonary fibrosis (Table 5).

There are several pharmacologic agents that have been or are being developed for inhibition of FGFR via targeting of the ATP binding site of the intracellular tyrosine kinase domain(s) (Table 5). The inhibition varies by their affinities for FGFR signaling, as most of these molecules are promiscuous kinase inhibitors (inhibiting VEGF, PDGF, and many other TKIs in addition to FGFR). Figure 1b gives examples of selective versus non-selective FGFR inhibitors. Of note, the dual kinase inhibitor (VEGFR/FGFR) lucitanib has shown activity in FGFR1-amplified breast cancer, with an overall response rate of 50 % [69]. There are also selective tyrosine kinase inhibitors (TKIs) available (Table 5 and Fig. 1b). While several of these agents are currently in clinical trials, none of the more highly selective FGFR inhibitors (e.g., BGJ398 or AZD4547) are currently FDA-approved. A recent study demonstrated that Debio 1347 (a selective orally available FGFR1–3 inhibitor) displayed preferential anti-tumor activity against cells with FGFR genetic alterations in a panel of 327 cancer cell lines and xenograft models [70]. Debio 1347 is currently under investigation for the treatment of patients harboring FGFR genetic alterations.

Of interest, other types of agents have been developed. As an example, FP-1039 is a soluble fusion protein, consisting of the extracellular domains of human FGFR1 linked to the Fc region of the human immunoglobulin G1; it is designed to bind multiple FGF ligands [71] (TRAP molecule).

A large number of other drugs and indications are being pursued. Some FGFR inhibitors have failed to meet their phase III endpoints (Table 5). The majority of the failed trials, however, have been performed in patient populations that were not biomarker-selected.

In a recent phase 1 study reporting on patients with FGFR1-amplified (identified by fluorescent/chromogenic in situ hybridization) advanced or metastatic lung squamous cell carcinoma (SCC) treated with the selective pan-FGFR inhibitor BGJ398, only a limited number of patients achieved relatively short-lived partial responses (2 of 17 patients, 11.7 %); responses lasted 3 and 8 months. These observations suggest the existence of resistance mechanisms. Sohl et al. [72] reported that resistance mutations at the “gatekeeper” residue may arise (FGFR1 V561M mutation confers a 38-fold increase in autophosphorylation and significant resistance to lucitanib), leading to tumor progression and explaining the non-durable responses. For instance, it has also been shown that the heterozygous gatekeeper mutation FGFR3 V555M appeared as a mechanism of acquired resistance to selective FGFR inhibitors [73]. Several other activating mutations were identified in FGFR2-expressing cells treated with high concentrations of dovitinib, and the multi-kinase inhibitor ponatinib inhibitory activity was affected by the V565I gatekeeper mutation [74]. In addition, a previously undescribed FGFR3 variant was identified as a key contributor to resistance in the MGH156-1A cell line derived from a patient with acquired resistance to EGFR TKIs [75], and follow-up studies clearly indicated that FGFR inhibitors re-sensitized these cells to EGFR inhibitors.

Besides secondary mutations in the kinase domain, resistance to FGFR kinase inhibitors may also occur through activation of alternative signaling pathways, as demonstrated by Harbinski et al. [76] who showed a broad and versatile potential for tyrosine kinase receptor from the FGFR, HER, and MET family to compensate for each other. Javidi-Sharifi et al. [77] suggest that some patients with gastrointestinal stromal tumor (GIST) treated with imatinib can develop a functional dependence on FGFR3, illustrated by the fact that the addition of the FGFR3 ligand FGF2 to GIST cells restored KIT phosphorylation during imatinib treatment. Furthermore, signaling crosstalk between KIT and FGFR3 activated the MAPK pathway to promote resistance to imatinib. FGFR amplification and overexpression have also been related to poor prognosis and endocrine resistance in breast cancer [78]. Of note, relationships between cyclins and FGF/FGFRs have also been reported at the protein level. For instance, a study showed that FGFR4 contributed to the maintenance of CCND1 via the mTOR translation pathway, and several other studies demonstrated cooperation between FGFR and CCND1 [79]. Finally, most patients with advanced cancer have complex molecular portfolios, and hence there may be multiple genomic drivers that are active and supplant the role of FGFR [80‐82]. In that context, it appears evident that identification of resistance mechanisms is crucial to crafting effective drug combinations.