Fibroblast Growth Factor Receptor Inhibitors as a Cancer Treatment: From a Biologic Rationale to Medical Perspectives

FGFs and FGFRs

The FGF ligand family contains 18 known components that can be divided into the following 2 categories: hormone-like FGFs (i.e., FGF19, 21, and 23) and canonical FGFs (i.e., FGF1–10, 16–18, and 20). The FGF–FGFR interaction is stabilized by the formation of a ternary complex that involves either cell surface heparan sulfate proteoglycans (HPSG) or Klotho proteins that further increase the specificity of the interaction in the case of hormonal FGFs (1).

Four of 5 identified FGFRs (i.e., FGFR1–4) are single-pass, transmembrane, tyrosine kinase receptors (Fig. 1). The extracellular domain contains 3 immunoglobulin (Ig)–like fragments (Ig-I, -II, and -III). Ig-I and the acid box are supposed to have a role in receptor autoinhibition, whereas Ig-II and Ig-III bind ligands and HPSGs (1). Alternative splicing of the second half of the Ig-III extracellular fragment of FGFR1, 2, or 3 may generate isoforms, that is, Ig-IIIb and Ig-IIIc, which are specifically expressed in the epithelium and mesenchyme, respectively, and differ in terms of ligand-binding specificity (2). The cytoplasmic portion of FGFR1–4 contains a tyrosine kinase domain and a COOH tail. A fifth receptor, FGFRL1, also binds FGFs, but it lacks the tyrosine kinase domain; moreover, FGFRL1 reportedly reduces cell growth and accelerates cell differentiation (3).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

The FGFR structure and signaling network. FGFRs are single-pass transmembrane receptors with an extracellular domain comprising 3 Ig-like domains (Ig-I–III) and an intracellular tyrosine kinase domain. The ligand-receptor binding is stabilized by the interaction with HPSG, thus inducing receptor dimerization and transphosphorylation at several tyrosine residues in the intracellular portion of FGFR. Two main downstream pathways are shown. The first induces Ras-dependent MAPK signaling, and the second leads to PI3K/Akt activation, independently and in parallel with Ras. PLC-γ may also be activated, thus triggering PKC, which converges with the MAPK pathway. Other pathways may be switched on, such as the STAT-dependent pathway. Many negative regulators are able to attenuate signaling at different levels, including FGFRL1, SEF, SPRY, and MAPK phosphatase 1 and 3 (MKP1 and MKP3) MEK, MAP-ERK kinase.

The specificity of the FGF–FGFR interaction is established through the different ligand-binding capacities of FGFRs as well as the alternative splicing of FGFR mRNA, and the tissue-specific expression of ligands, receptors, and cell-surface proteins.

FGF/FGFR Signaling

The binding of an FGF to an FGFR induces receptor dimerization, thereby enabling the intermolecular transphosphorylation of several tyrosine residues in the COOH terminal tail, kinase domains, and juxtamembrane region. The activated FGFR phosphorylates FGFR substrate 2 (FRS2) and recruits growth factor receptor-bound 2 (GRB2), finally resulting in the activation of RAS and the downstream RAS/mitogen-activated protein kinase (MAPK) pathway. Likewise, GRB2 activates phosphoinositide 3-kinase (PI3K)/AKT-dependent signaling. Operating independently from FRS2, phospholipase C-γ (PLC-γ) binds to a phosphotyrosine at the COOH tail and hydrolyzes phosphatidylinositol 4,5-bisphosphate (PIP2) to phosphatidylinositol 3,4,5-triphosphate (PIP3) and diacylglycerol (DAG), thus activating protein kinase C (PKC), which converges with the MAPK pathway (Fig. 1; ref. 4).

Negative feedback loops involve the induction of MAPK phosphatase 3 (MAPK3), Sprouty (Spry) proteins, and “similar expression to FGF” (SEF) family members, which can attenuate the signal cascade at different levels. Moreover, following activation, the receptor is internalized and then degraded or recycled, partly depending on the extent of ubiquitination (4).

Biologic and Context-Dependent Responses

FGFRs/FGFs are key molecules involved in embryogenesis, tissue homeostasis, tissue repair, wound healing, and inflammation (5). The main effects of the FGFR pathway include proliferation, migration, and antiapoptotic signals. Proliferation is mainly achieved through the MAPK cascade, whereas antiapoptotic signals are mediated by PI3K/AKT with cross-talk between both pathways (4). However, in specific cellular contexts, FGF signaling may induce growth arrest and cell differentiation according to the cell type–specific expression of FGFs, FGFRs, and adaptor molecules.

FGF and Angiogenesis

Embryogenesis, angiogenesis, and wound healing were the first fields in which FGF/FGFR proliferation and migration signals were shown to play a fundamental role. Both FGF1 and FGF2 drive potent angiogenic signals according to the results of various preclinical assays. FGF1 and FGF2 directly promote endothelial cell proliferation, migration, vessel formation, and maturation (6). Indirectly, FGFs can synergize with the VEGF and platelet-derived growth factor (PDGF) pathways. For instance, in one preclinical microvascular endothelial cell model, the angiogenic response occurred faster when both FGF2 and VEGF were added, as compared with the addition of either factor alone (7).

Activating FGFR Genomic Alterations

Table 1 summarizes the main activating FGFR genomic alterations found in human cancers. These major oncogenic aberrations have features that make them potentially good therapeutic targets for specific cancers, and the details of these targets are discussed subsequently.

Autocrine/Paracrine Signaling

An increased availability of FGFs can promote cancer, as shown in several mouse models (4). High FGF levels may be derived from the upregulation of FGF expression in cancer or stromal cells and the enhanced release of FGFs from the extracellular matrix (46).

Much evidence supports the carcinogenetic and metastatic role of stromal–epithelial interplay involving FGFR signaling in prostate cancer. Indeed, according to previous studies, multiple FGFs, including FGF1 and 2 and FGF6–9, are upregulated in this disease, whereas FGFR1 is expressed in 40% of poorly differentiated localized prostate cancers (46). The stromal expression and release of various FGFs are related to angiogenic and tumor-promoting effects. More specifically, studies in transgenic mice have linked FGFR1 activation and FGF8 epithelial overexpression to epithelial–mesenchymal transition and the induction of prostatic adenocarcinomas (47). A humanized monoclonal anti-FGF8 antibody and selective FGFR TKIs, such as AZ8010, displayed in vivo antitumor activity against prostate cancer cell lines (46, 48).

FGF2 has also been shown to induce epithelial–mesenchymal transition in NSCLC cell lines (49). The inhibition of FGFR signaling through antisense RNA, naturalizing FGF2 antibodies, or TKIs led to the in vitro inhibition of cellular proliferation and tumor growth (50).

Finally, autocrine loops of FGF2 have also been reported in triple-negative breast cancers. This autocrine loop occurred mainly in basal-like B cells (showing mesenchymal-like features) and was associated with high sensitivity to FGFR inhibitors in preclinical models (51).

Overall, these data suggest that autocrine/paracrine loops of FGF could be involved in epithelial–mesenchymal transition and cancer progression in at least 3 common cancers.

FGF and Tumor Neoangiogenesis

FGFs have also been implicated in tumor neoangiogenesis. FGFs exert their angiogenic functions on endothelial cells in both a paracrine fashion and via autocrine release from capillary endothelial cells (4). Overexpression of a soluble FGFR led to an impairment in the maintenance of tumor angiogenesis and a decrease in tumor vessel density. Conversely, the inhibition of the VEGF pathway affected the initiation phases of tumor angiogenesis (52). Other compelling evidence comes from a preclinical study in which antisense-oriented FGF2 or FGFR1 cDNAs were delivered in preclinical models of melanomas. Intratumoral angiogenesis blockage caused the cessation of tumor growth (53). Studies on ovarian cancer revealed that FGF1 overexpression is correlated with both microvessel tumor density and poorer overall survival (OS). Moreover, both ovarian cancer cells and endothelial cells showed increased survival and motility when treated with exogenous FGF1 (54).

Evidence for synergism between FGFs and other angiogenic pathways in tumor angiogenesis has also been reported. For example, the simultaneous expression of VEGF and FGF2 rapidly increased tumor growth, blood vessel density, and permeability in xenograft models. The inhibition of FGF2 expression resulted in a significant decrease in tumor volume concomitant with a decrease in vessel density, whereas VEGF inhibition led to impaired pericyte organization and permeability (55).

As FGFR and VEGF receptor (VEGFR) pathways are integrated in tumor neoangiogenesis through complementary and, in part, overlapping functions, the upregulation of FGFs/FGFRs may also serve as a mechanism of resistance to anti-VEGF therapy. In the preclinical setting, murine pancreatic neuroendocrine tumor models initially responding to anti-VEGF treatment showed a higher expression of FGF2 at the time of progression, as compared with that of stable tumors (56). A TKI that inhibited both VEGFR and FGFR presented antitumor activity in a model of anti-VEGF failure (57). In agreement with these observations, reports have indicated that FGF2 is higher in patients with colorectal cancer after the failure of bevacizumab-containing regimens and in glioblastoma patients after treatment with a VEGFR TKI (6).

FGF and Drug Resistance

A significant amount of cross-talk between FGF and oncogenic pathways could explain emerging data that indicate a role for FGF/FGFR in the mediation of resistance to targeted and endocrine therapies. As recently shown, the sensitivity of cancer cells to TKIs can be impaired by the increased availability of growth factor ligands. More specifically, when testing different growth factors on a series of kinase-addicted cancer cell lines that are sensitive to specific TKIs, hepatocyte growth factor, FGFs, and neuregulin seem to confer drug resistance to most of the cells (58). Interestingly, the secretion of these ligands may come from tumor stroma (59). These data suggest that FGF released by stromal cells can be involved in the resistance to targeted therapy. This assumption constitutes the rationale for evaluating dovitinib in patients with imatinib-resistant gastrointestinal stromal tumors (GIST; clinicaltrials.gov; NCT01478373, NCT01440959).

Interestingly, in a mouse model of cervical carcinogenesis, using imatinib to target stromal PDGF receptor (PDGFR) resulted in reduced proliferation and angiogenesis via suppression of the expression of FGF2 and FGF7 by cancer-associated fibroblasts. Thus, PDGF ligands expressed by cancer epithelial cells stimulated PDGFR-expressing stroma to upregulate FGFs, thereby promoting angiogenesis and epithelial proliferation. Given this cross-talk, one mechanism of resistance to imatinib might involve upregulation of the FGF pathway (60).

In addition to FGF expression, FGFRs can also be involved in resistance to conventional therapies. First of all, a previous study showed that FGFR1 overexpression/amplification mediates resistance to 4-hydroxytamoxifen, and the effect was reversed by the inhibition of FGFR1 via siRNA (40).

More recently, EGF receptor (EGFR) inhibitors have been shown to induce transcriptional de-repression of FGFR2 and FGFR3 expression in NSCLC cell lines, thereby suggesting a novel mechanism for acquired resistance to anti-EGFR agents (61). Other reports have also suggested that FGFR3 activation mediates resistance to cetuximab in wild-type squamous tumor cells with KRAS mutations and to vemurafenib in BRAF V600E melanoma cells (62, 63).

In total, the data reported to date highlight a critical role for the FGFR pathway in human cancer. Altered FGF/FGFR function may not only promote cancer progression by directly stimulating cancer cell proliferation and survival but also play a major role in tumor neoangiogenesis. This proangiogenic effect is relevant per se to tumor growth and progression because it represents the key mechanism that sustains cancer cells, and it has also implications for the resistance to available antiangiogenic treatments. Finally, interesting recent data indicate that the FGFR pathway is a potential driver of resistance to various targeted drugs. Therefore, anti-FGF/FGFR agents can be used to affect tumor progression in cancers driven by FGF/FGFR-activatingalterations, target angiogenesis as a fundamental process in cancer cell survival, and revert acquired resistance to antiangiogenic or other targeted agents. The fact that this strong rationale comprises at least 3 major issues represents a challenge for drug development and has implications for current and future clinical research. Each of these 3 goals requires specifically designed clinical trials that differ in terms of patient selection, clinical setting, and treatment to be tested. For example, trials targeting FGFR mutations as oncogenic drivers should preferably be focused on patients harboring the specific mutations. Trials addressing the angiogenic issue may include patients who acquired resistance to other antiangiogenic treatments and should plan the evaluation of potential markers of resistance involving the FGF/FGFR pathway. The neoadjuvant setting could allow researchers to directly test the antiangiogenic properties of an FGF/FGFR blockade using an in vivo study. Finally, trials that investigate the ability of anti-FGFR drugs to overcome resistance to other targeted treatments should test drug combinations or sequential schedules and should select patients on the basis of their previous treatments.

Small-Molecule TKIs

Several TKIs targeting the ATP-binding site of the intracellular tyrosine kinase domain of FGFRs are in clinical development (Table 2). Overall, FGFR inhibitors can be classified into 2 different families. The first class is represented by “multitarget” TKIs that include FGFR in their panel of targets (i.e., nonselective FGFR TKIs). These compounds usually also target VEGFRs and usually present modest but significant bioactivity against the FGFR family. The second class includes highly selective and highly bioactive FGFR inhibitors (i.e., selective FGFR TKIs). For the purpose of this review, we will focus on those compounds exhibiting an IC₅₀ less than 200 nmol/L against at least one of the FGFRs. Those compounds that have entered the clinical phases of development are discussed subsequently.

Drugs and Phase I Data

Phase II Data

Monoclonal Antibodies

Several monoclonal antibodies against FGFRs have been assessed in preclinical studies, and these studies have shown an antiproliferative effect on bladder cancer cells, t(4;14) myelomas, and mouse xenograft models of FGFR2-amplified gastric cancer and breast cancer (4). However, the evaluation of a single-chain, anti–FGFR1-IIIc was stopped at the preclinical phase after it was shown to induce severe anorexia in rodents and monkey models (90). In another example, MGFR1877S, an anti-FGFR3 antibody, is currently being evaluated in 2 phase I trials, that is, one including patients with advanced solid tumor and the other recruiting patients with t(4;14) myeloma (clinicaltrials.gov; NCT01363024, NCT01122875).

FGF-Ligand Traps

FP-1039 consists of the extracellular domain of an FGFR1-IIIc splice isoform fused to the crystallizable fragment region of human IgG. This compound is supposed to hamper the activity of multiple FGFs. Indeed, it has shown antiangiogenic and antitumoral efficacy in preclinical in vivo studies (75). A phase I study in advanced cancers is under way (clinicaltrials.gov, NCT00687505), but a phase II study that would have assessed FP-1039 efficacy in patients with FGFR2-mutant endometrial cancer has been withdrawn before the enrollment phase. The study was deemed infeasible because the original assumption was that at least 5% of patients who were screened would qualify, but after 70 patients had been screened, none qualified (clinicaltrials.gov; NCT01244438).

Running Therapeutic Trials in Patients Presenting a Low-Frequency Genomic Alteration

As mentioned in the previous sections, FGF/FGFR genomic alterations are rare for each single cancer type. Preclinical studies suggest that patients presenting these genomic alterations are more likely to be sensitive to FGFR inhibitors. Therefore, early clinical trials should include patients presenting these genomic alterations to determine the efficacy of these agents in this rare subset of patients. This approach has been already pursued in the TKI258A2202 trial, which evaluated TKI258 in patients with breast cancer; indeed, in this trial, 50% of the patients who were enrolled showed FGFR1 amplification. Many other ongoing phase I/II trials with anti-FGFR agents (e.g., TKI258, BGJ398, AZD4547, and E3810) are now including patients with specific FGF/FGFR alterations. In addition to the change in the population of early trials, validation randomized trials will likely need to include only patients with the genomic alteration under study.

This process of selecting patients emphasizes the need for development of optimal molecular diagnosis procedures for FGFR alterations. The challenge here is to be able to select the right patients for the FGFR compound while not denying another potentially effective compound to FGFR-negative patients. One possible strategy is to “screen” patients using multiplex approaches. This screening allows the detection of several alterations and thus increases the likelihood of identifying at least one actionable genomic alteration in each patient. The “screened” alteration must be centrally confirmed in the context of a Clinical Laboratory Improvement Amendments (CLIA)–certified laboratory. Several tests are being used for the diagnosis of genomic alterations in the FGF pathway, including FISH, chromogenic in situ hybridization (CISH), quantitative real-time PCR, and Sanger sequencing (91).

In addition, autocrine and paracrine FGF/FGFR overexpression promotes carcinogenesis and progression. Therefore, developing anti-FGFR drugs only for the treatment of genomic aberrant tumors is short sighted and may miss other potentially therapeutic uses. Future research efforts must strive to define the subset of cancers driven by FGFR pathway component overexpression by means other than amplification and may benefit from an anti-FGFR approach. In the case of tumors driven by the overexpression of an FGF, research should define which of the FGFRs really mediates the effect of the upregulated ligand and then specifically target that FGFR.

Selecting the Right Drug Family: Nonselective TKI or Selective FGFR Inhibitors?

This dilemma of developing highly specific versus multitarget inhibitors is shared with other drug classes. Although the time is not right for direct comparisons between nonselective and selective anti-FGFR TKIs, some considerations can be contemplated. As previously reported, multi-TKIs can cause adverse effects related to VEGF, in addition to potential FGF-related problems. In addition, these compounds present less bioactivity in terms of their actions against FGFR. On the basis of these considerations, the development of strongly bioactive, highly selective FGFR inhibitors is of greater need than multitarget inhibitors. Nevertheless, the most impressive antitumor activity to date has been observed with multitarget kinase inhibitors in patients presenting genomic alterations in the FGF pathway (78, 85). In addition, multitarget inhibitors are more advanced in the clinical development process, as dovitinib is part of a phase III registration trial for kidney cancer treatment and a phase II randomized trial for breast cancer treatment.

Overall, these considerations suggest that multitarget kinase inhibitors will be the first-in-class to gain approval. Highly bioactive FGFR inhibitors (e.g., BGJ398 and AZD4547) will come later and could prove to be interesting choices when combination therapies are needed. Indeed, they present FGFR-specific toxicity profiles that could favor their use in drug combinations. In contrast, the broad nonspecific toxicity profile of multitarget TKIs makes them difficult to include in drug combinations.

Managing Long-Term Toxicities of FGFR Inhibition

Given the multiple recognized physiologic functions of FGF/FGFR signaling and the suggested oncosuppressive role of FGF/FGFR activation in some contexts, the feasibility of a long-term FGF pathway inhibition is questionable. This issue will be investigated in ongoing studies; however, some efforts are currently under development. Anti-FGFR antibodies that target a single FGF or FGFR could offer an alternative and may display a more narrow range of toxicity as compared with pan-FGFR inhibitors. An additional suggestion aimed at avoiding continuous pan-FGFR inhibition could be the provision of therapeutic windows by continuing other treatments that target additional downstream kinases, such as MAPK.

Developing Third-Generation Angiogenesis Inhibitors

The identification of the most effective timing for adding FGF/FGFR inhibitors to VEGF inhibitors is an emerging area of interest. Current research has not elucidated whether targeting FGF signaling is more effective in the front-line setting or after the failure of a regimen of sunitinib or bevacizumab. Although inhibition of the VEGF and FGF pathways has been shown to be synergistic, preclinical and clinical evidence suggests that FGF levels increase at the time of anti-VEGF resistance, thereby suggesting that the addition of FGFR inhibitors at the time of the failure of anti-VEGF treatment, rather than upfront, makes the most clinical sense. From this perspective, it would be unlikely to expect great differences in terms of efficacy through a head-to-head, first-line drug comparison. Hence, the initial development of FGF inhibitors at the time of PD during VEGF exposure has the strongest basis. Optimally, eligible patients could be selected on the basis of FGF levels. In contrast, a current, ongoing debate rages over whether treatment with an antiangiogenic drug may alter disease biology by making it more aggressive and finally hampering survival despite initial responses. Although the intrinsic mechanisms of this suggested “rebound” process are not completely clear, we can postulate that an upfront dual targeting of VEGF and FGF might help to prevent it. To better address these issues, researchers need to develop more representative preclinical models that have the increased capability to provide results that can be translated into the clinical setting. Furthermore, biomarker studies must be prioritized, and assays must be developed to better characterize the role of FGF-dependent angiogenesis during the different phases of diseases and its relationship with other angiogenic pathways. This practice will allow researchers to better define the clinical settings and study populations for further anti-FGF drug development.

Combining FGFR and Other Kinase Inhibitors

FGFR alterations in cancer can be associated with the activation of other oncogenic drivers. For example, FGFR1 is frequently coamplified with CCND1 on 11q mainly in luminal breast cancers. Evidence from breast cancer cell line assays shows a substantial functional interaction between the genes on 8p11-12q and their cooperation with major oncogenic pathways. These findings may represent the basis for the development of novel drug combinations, including an FGFR inhibitor and a CDK4 inhibitor (92). In addition, ER-positive breast cancers present a high frequency of PIK3CA mutations, and studies to evaluate whether these cases are more numerous in FGFR1-amplified breast cancer samples are ongoing. Such findings could provide a rationale to combine PI3K inhibitors and FGFR inhibitors in a treatment regimen. Such a combination may deserve evaluation even in the triple-negative breast cancer group and, more specifically, in the subtype showing mesenchymal features. Indeed, mesenchymal-like triple-negative cells may contain more FGF/FGFR pathway components and show sensitivity to PI3K inhibition (42, 51). One of the most evident mechanisms of endocrine resistance is the activation of the PI3K/AKT/mTOR pathway; thus, mTOR inhibitors (everolimus) will soon become a standard of care in ER-positive breast cancer. Further studies will evaluate to what extent FGFR1 could contribute to resistance to everolimus and whether evidence suggests the need to combine everolimus with FGFR inhibitors.

Additional data support the combination of FGFR and kinase inhibitors. Cancer cell line assay results suggest that cooperation exists between FGFR and receptors of the EGFR family in oncogenesis (20). Overexpression and activation of FGFR1 and HER2 have been observed in breast cancer, and the combined inhibition of both receptors led to a stronger antitumor activity compared with each treatment alone in a mouse model (93). In addition, we described different mechanisms by which FGFR signaling could contribute to resistance to available targeted treatments, such as anti-EGFRs (e.g., vemurafenib and imatinib), and future research must determine the optimal selection of the most appropriate combinations from the pool of new molecularly targeted agents. Finally, targeted drugs may not be the only potential candidates for association with anti-FGFRs, as preclinical assays found increased levels of FGFR4 in chemotherapy-resistant breast cancer cell lines as a result of ectopic endogenous expression. Indeed, treatment with an anti-FGFR4 antibody restored sensitivity to chemotherapy-induced apoptosis (94).

Identifying Resistance to FGFR Inhibitors

Deeper insight into tumor biology from the performance of biomarker analysis is required not only to identify markers for sensitivity to FGFR inhibition but also to define possible markers of anti-FGFR resistance at earlier stages. As previously stated, cross-talk with other oncogenic or angiogenic pathways has been proposed; therefore, adaptation mechanisms that involve the activation of other downstream or parallel signals may occur, and those parallel signals may be targeted by different agents. Studies of tumor tissue evaluations before and after anti-FGFR treatments are preferred. In this context, the neoadjuvant setting can help in in vivo evaluating drug effects.