Lung cancer is the most common malignancy worldwide (33), which is histologically classified into adenocarcinoma, large cell carcinoma, small cell carcinoma and squamous cell carcinoma. FGFR1 amplification preferentially occurs in squamous cell lung cancer (34); 9.3% of stage I cases, 22% of stage II cases (35) and 19% of stage IV cases with brain metastasis (36). FGFR2-CCAR2, FGFR2-CIT and FGFR3-TACC3 fusions and gain-of-function mutations of FGFR2 (A266_S267insSTVVGGD and 290_291WI>C) are also observed in lung cancer (37–39). Aberrant FGFR signaling may promote the proliferation and survival of tumor cells in the early stage, and invasion and metastasis in the later stage. As smoking is a risk factor of squamous cell lung cancer, the smoking-induced DNA damage of bronchial epithelial cells may cause lung carcinogenesis in part through FGFR1 gene amplification.

Breast cancer is the second most common malignancy worldwide (33). Breast cancers have been classified based on immunohistochemical analyses of estrogen receptor (ER), progesterone receptor (PR) and ERBB2 (HER2) receptor. The prognosis for luminal breast cancer (ER⁺) is better than that for non-luminal HER2⁺ breast cancer (ER⁻/PR⁻/HER2⁺) and triple-negative breast cancer (ER⁻/PR⁻/HER2⁻) (40). Endocrine therapy using tamoxifen or aromatase inhibitors is the standard therapy for patients with luminal breast cancer without lymph node metastasis; however, recurrence as a result of resistance to endocrine therapy is a serious issue in clinical practice (41). FGFR1 gene amplifications and FGFR2 gain-of-function missense mutations are detected in circulating tumor cells of two cases and one case of breast cancer patients, respectively (42,43). The FGFR1 gene is more frequently amplified in metastatic breast cancer than the FGFR2 and FGFR3 genes (44), which leads to FGFR1 overexpression and resistance to endocrine therapy (45). The preclinical studies mentioned above indicate that FGFR1-targeted therapy is applicable for a subset of breast cancer patients.

Gastric cancer is the fifth most common malignancy worldwide, although its incidence and mortality have been decreasing (33,46). The amplifications of genes encoding RTKs, such as EGFR, ERBB2, FGFR2 and MET, occur in gastric cancer (47,48). Gastric cancer with FGFR2 amplification is significantly associated with lymphatic invasion and a poor prognosis (49,50); however, the molecular mechanisms through which FGFR2 amplification promotes lymph node metastasis remain unclear. Preclinical studies using small-molecule FGFR2 inhibitor and patient-derived cancer xenografts revealed that FGFR2 amplification in human gastric cancer is a promising therapeutic target (51,52).

Somatic FGFR1 mutation (N546K) in Ewing sarcoma (53), FGFR2 mutations (R203C, N549K and K659N) in breast cancer (43,54), FGFR2 mutations (S252W and N549K) in endometrial cancer (55), FGFR2 coding in/del mutations in lung cancer (39) and FGFR3 mutations (R248C and S249C) in bladder cancer (56) are gain-of-function mutations (Fig. 3), whereas somatic FGFR2 mutations (D530N, I642V and A648T) in melanoma are loss-of-function mutations (57).

Missense mutations of FGFRs involved in congenital disorders rather than cancer have been well characterized. Gremlin FGFR1 mutations (P252R and Y372C) and FGFR2 mutations (S252W, P253R, K526E, N549K, K641R, etc.) in patients with craniosynostosis and FGFR3 mutations (R248C, S249C, G380R, N540K, K650E, etc.) in patients with skeletal dysplasia are gain-of-function mutations (17–20); however, germline FGFR1 mutations (R254W and V429E) in patients with congenital hypogonadotropic hypogonadism are loss-of-function mutations (58,59).

These facts clearly indicate that there are gain-of-function, as well as loss-of-function mutations in FGFRs. Therefore, validation of gain-of-function based on kinase or cell-based assay is mandatory before prescribing FGFR inhibitors to cancer patients with FGFR coding mutations.

FGFR fusions in human cancers are classified into type 1 fusions caused by chromosomal translocations in hematological malignancies, and type 2 fusions caused by chromosomal rearrangements in solid tumors (Fig. 3). Chromosomal translocations involving the FGFR1 gene occur in patients with myeloproliferative neoplasms initially presenting with eosinophilia and lymphadenopathy and then developing acute myeloid or mixed-lineage leukemia (60,61); chromosomal translocations involving the FGFR3 gene have been shown to occur in patients with peripheral T cell lymphoma (62). Type 1 FGFR fusion proteins, including BCR-FGFR1, CNTRL-FGFR1, CUX1-FGFR1, FGFR1OP-FGFR1, FGFR1OP2-FGFR1, LRRFIP1-FGFR1, MYO18A-FGFR1, RANBP2-FGFR1, TPR-FGFR1, TRIM24-FGFR1, ZMYM2-FGFR1 and ETV6-FGFR3 are non-receptor-type FGFR kinases (Fig. 3). CNTRL-FGFR1 fusion kinase promotes the expansion of the stem cell population and causes myeloid and lymphoid malignancies (63). By contrast, type 2 FGFR fusion proteins in solid tumors, including FGFR1-TACC1, FGFR2-AFF3, FGFR2-BICC1, FGFR2-CASP7, FGFR2-CCAR2, FGFR2-CCDC6, FGFR2-CIT, FGFR2-OFD1, FGFR2-PPHLN1, FGFR3-BAIAP2L1, FGFR3-JAKMIP1 and FGFR3-TACC3, are transmembrane-type FGFRs with C-terminal substitution to the region of fusion partners (37,38,56,64,65) (Fig. 3).

Both types of FGFR fusion proteins are endowed with oncogenic potential through the acquisition of protein-protein-interaction modules from fusion partners for ligand-independent dimerization and/or recruitment of aberrant substrates. Type 1 FGFR fusion proteins acquire oncogenic potential through altered subcellular localization as a result of the loss of the extracellular and transmembrane domains of wild-type FGFRs. Type 2 FGFR fusion proteins lose the PLC-γ-binding tyrosine (Tyr or Y) residue (Y766 in FGFR1, Y769 in FGFR2 or Y760 in FGFR3) owing to the C-terminal alterations. To understand the mechanisms of carcinogenesis caused by the FGFR fusions, substrates and downstream signaling cascades of FGFR fusion proteins need be elucidated.

AZD4547 (66,67), BGJ398 (infigratinib) (68), BLU9931 (69), Debio-1347 (70), dovitinib (TKI258) (27,71), FIIN-2 (72), JNJ-42756493 (73), LY2874455 (74) and ponatinib (9) are small molecule compounds that inhibit FGFRs at low nM levels (IC₅₀ value <20 nM) (Fig. 4A). AZD4547, BGJ398, Debio-1347, dovitinib, JNJ-42756493 and LY2874455 are reversible FGFR inhibitors that occupy the ATP-binding pocket in the kinase domain, whereas BLU9931 and FIIN-2 are irreversible FGFR inhibitors that covalently bind to each specific cysteine (Cys or C) residue in the kinase domain. These FGFR inhibitors reduce phosphorylation of FGFRs themselves and their direct targets, FRS2 and PLC-γ, and inactivate downstream signaling, such as the RAS-ERK, PI3K-AKT, IP₃-Ca²⁺ and DAG-PKC signaling cascades.

AZD4547, BGJ398, Debio-1347 and dovitinib are FGFR1/2/3 inhibitors that are less effective on FGFR4; BLU9931 is a selective FGFR4 inhibitor; FIIN-2, JNJ-42756493, LY2874455 and ponatinib are pan-FGFR inhibitors (Fig. 5A). Phylogenetic analysis on 54 RTKs revealed diversification of FGFR4 from other FGFRs (Fig. 2), and amino-acid alignment of the tyrosine kinase domains in the FGFR family members revealed relatively frequent amino-acid substitutions specifically in FGFR4 (Fig. 4B). Phospho-tyrosine residues involved in catalytic activation (Y653 and Y654 in FGFR1), STAT recruitment (Y677 in FGFR1) and PLC-γ recruitment (Y766 in FGFR1) are conserved in all members of the FGFR family. By contrast, one tyrosine residue in the hinge region (Y563 in FGFR1, Y566 in FGFR2 and Y557 in FGFR3) is changed to C552 in FGFR4, and phospho-tyrosine residue in the kinase insert region (Y583 in FGFR1, Y586 in FGFR2 and Y577 in FGFR3) is changed to L572 in FGFR4 (Fig. 4B). Y563 in FGFR1 is necessary for the interaction with Debio-1347 (70), whereas C552 in FGFR4 is necessary for the covalent binding with BLU9931 (69). As the diversification of FGFR4 significantly affects the biding affinities of TKIs, it is reasonable to functionally classify FGFR-targeting TKIs into FGFR1/2/3 inhibitors, FGFR4 inhibitor and pan-FGFR inhibitors.

AZD4547, BGJ398, Debio-1347, dovitinib, JNJ-42756493 and ponatinib are currently being investigated in clinical trials (https://clinical-trials.gov): phase II studies of AZD4547 in patients with breast, gastric and squamous-cell lung cancer (FGFR1 or FGFR2 amplification) and metastatic breast or non-small-cell lung cancer (FGFR genetic alterations, umbrella trial); phase II studies of BGJ398 in patients with solid tumors or hematological malignancies (FGFR genetic alterations); phase II study of dovitinib in patients with gastric cancer (FGFR2 amplification); phase II study of JNJ-42756493 in patients with urothelial cancer (FGFR genetic alterations); phase II study of ponatinib in patients with advanced biliary cancer (FGFR2 fusion) or refractory metastatic solid tumors (genetic alterations in FGFRs and other targets).

TKIs have been approved for cancer therapy by regulatory authorities in expectation of an improved risk/benefit ratio; however, adverse effects on viral organs, such as the cardiovascular system and liver, are serious issues that may occur in the clinic (75). Hypertension, bleeding and thrombosis are adverse effects of anti-angiogenic therapy targeting the VEGF signaling pathway (76), while cardiovascular events are serious adverse effects of ponatinib for the treatment of chronic myeloid leukemia (77). AZD4547, dovitinib and ponatinib are representative multi-kinase inhibitors targeting FGFRs and other tyrosine kinases (Fig. 5A). Selective FGFR targeting is expected to reduce adverse effects, whereas the dual targeting of FGFR and VEGFR/CSF1R is expected to enhance the anti-tumor effects indirectly through the normalization of tumor microenvironment.

The tumor microenvironment consists of cancer cells and stromal/immune cells, such as fibroblasts, endothelial cells, lymphocytes, macrophages, monocytes and neutrophils (Fig. 6). The interactions of tumor cells and stromal cells are involved in almost all stages of tumor development, including neoplastic transformation, proliferation, invasion and metastasis, through the regulation of various cellular processes in a context-dependent manner (78–80). FGFs derived from cancer cells, as well as stromal cells play a key role in the tumor microenvironment.

Cancer-associated fibroblasts (CAFs) are activated stromal fibroblasts that support tumorigenesis (Fig. 6). FGF2 activates human dermal fibroblasts through transcriptional downregulation of the TP53 gene, whereas BGJ398 or ponatinib treatment induces their senescence through the upregulation and activation of TP53 (81). By contrast, FGF2 signaling through FGFR1 causes resistance to EGFR inhibitor in lung cancer cells, and combination therapy using EGFR inhibitor and AD4547 is effective to overcome drug resistance (82). Multiple myeloma cells induce FGF23 secretion from osteocytes, and then FGF23 signaling through FGFR3 to multiple myeloma cells promotes proliferation and induces heparanase upregulation, which explains the pathogenesis of osteolytic 'punched-out lesion' in patients with multiple myeloma. BGJ398 treatment inhibits FGF23-dependent growth and heparanase expression of multiple myeloma cells (83). These results indicate the rational for the application of FGFR inhibitors to target paracrine FGF signaling in the tumor microenvironment.

Tumor angiogenesis is largely classified into sprouting angiogenesis and vasculogenesis (84). Sprouting angiogenesis is the formation of new blood vessels as a result of endothelial sprouting from preexisting blood vessels, whereas vasculogenesis is the de novo formation of blood vessels owing to endothelial differentiation of progenitor cells and endothelial-like differentiation of cancer cells. Tumor angiogenesis is involved in the supply of oxygen and nutrient (mature blood vessels), as well as the formation of hypoxic environment (immature blood vessels).

VEGF, FGF, angiopoietin (ANGPT) and Notch signaling cascades are major players of tumor angiogenesis (85,86). VEGF signaling through VEGFR2 promotes endothelial cell proliferation via the DAG-PKC signaling cascade, endothelial cell survival via the PI3K-AKT signaling cascade, endothelial cell migration via the FAK-Paxillin signaling cascade as well as vascular permeability and vasodilatation via the IP₃-eNOS (NOS3) signaling cascade (76,87,88). Pro-angiogenic FGF2 also promotes the proliferation and migration of endothelial cells directly through FGFR1 (or FGFR2) signaling activation (89,90) and indirectly through the induction/secretion of VEGF and ANGPT2 from endothelial cells (91,92). ANGPT1 is secreted from pericytes and maintains endothelial quiescence or stabilization through TIE2 signaling activation. ANGPT2 is secreted from endothelial cells and promotes the endothelial activation or sprouting through TIE2 signaling inhibition (92,93). VEGF signaling in endothelial tip cells induces DLL4 expression, which subsequently activates Notch signaling in endothelial stalk cells for vascular quiescence through VEGFR downregulation (94–96). VEGF, FGF2 and ANGPT2 are involved in endothelial activation, whereas ANGPT1 and Notch are involved in endothelial quiescence. VEGFR2 and FGFR1/2 on endothelial cells are representative RTKs that promote tumor angiogenesis (Fig. 6).

VEGF signaling is targeted using anti-VEGF monoclonal antibody (mAb) or small-molecule VEGFR inhibitors in cancer patients; however, some tumors do not respond to the VEGF blockade therapy and other tumors recur after transient response [(Gacche and Meshram (76); Jain (78)]. As FGF signaling activation in endothelial cells is one of the mechanisms responsible for intrinsic and acquired resistance to the VEGF blockade therapy (97), FGFR inhibitors may be applicable to overcome the resistance to the VEGF blockade therapy. There are two options for the dual blockade of FGF and VEGF signaling cascades. Combination therapy using FGFR inhibitor and anti-VEGF mAb is a preferable choice to reduce adverse effects, whereas monotherapy using small-molecule FGFR/VEGFR2 dual inhibitors, such as AZD4547 and dovitinob, may be a preferable choice to reduce medical cost. FGF/VEGF dual blockade therapy should be optimized in consideration of safety issues and medical costs.

Cancer immunity and immune tolerance in the tumor microenvironment are regulated by the interaction between cancer cells and immune cells (98). CD8⁺ T cells, NK cells and NKT cells are immune effector cells involved in tumor elimination (98), whereas myeloid-derived suppressor cells (MDSCs) (99), tumor-associating macrophages of M2 type (M2-TAMs) (100) and regulatory T (Treg) cells (101) are immune modifier cells involved in immune evasion and tumor growth (Fig. 6).

MDSCs are heterogeneous populations of immature myeloid cells, including monocyte-like MDSCs (CD14⁺, CXCR4⁺, CSF1R⁺ and VEGFR2⁺), granulocyte-like MDSCs (CD15⁺, CXCR4⁺, KIT⁺ and VEGFR2⁺) and endothelial progenitor cells (CD31⁺, CXCR4⁺, KIT⁺ and VEGFR2⁺) (99). CSF1 (M-CSF), CSF2 (GM-CSF) and CSF3 (G-CSF) are secreted from the tumor microenvironment and stimulate the growth and survival of MDSCs and other myeloid-lineage cells, while CXCL12 (SFD-1α) and VEGF promote the recruitment of MDSCs to the tumor microenvironment (102–106). MDSCs activate M2-TAMs and Treg cells, but inhibit CD8⁺ T cells and NK cells, leading to immune evasion in the tumor microenvironment. In addition, endothelial progenitor cell-like MDSCs are involved in tumor angiogenesis (104). MDSC infiltration and tumor angiogenesis during mammary tumorigenesis in MMTV-Wnt1/iFGFR1 bi-genic mice are significantly enhanced in comparison with MMTV-Wnt1 transgenic mice, and BGJ398 treatment results in tumor regression and disappearance of MDSCs from the residual mammary gland (107). By contrast, AZD4547 treatment inhibits the proliferation and lung metastasis of 4T1 mouse mammary tumor cells, and reduces MDSCs in the tumor microenvironment and systemic circulation (108). FGFR inhibitors induce the reduction or disappearance of MDSCs from the tumor microenvironment, partly by targeting cytokine-producing CAFs.

CSF1 signaling through CSF1R on monocyte/macrophage-lineage cells are involved in their proliferation, survival and differentiation (109–111). CSF1R inhibitors (GW2580 and PLX3397) and anti-CSF1R mAb (RG7155) have been developed as therapeutics for CSF1 signaling blockade in monocyte-like MDSCs and M2-TAM (112–115). Combination therapy of CSF1R inhibitor PLX3397 and paclitaxel inhibits tumor-infiltration of MDCSs and M2-TAM and suppresses mammary tumorigenesis (113,114). IC₅₀ value of PLX3397 to CSF1R is 20 nM (113), whereas IC₅₀ values of AZD4547 (67), ponatinib (9) and dovitinib (71) to CSF1R are 9.7, 8.7 and 36 nM, respectively. By contrast, PI3K is one of common signaling effectors CSF1R and FGFRs (Fig. 1A), and PI3K activation enhances immune suppressor and pro-angiogenic potentials of M2-TAMs (116). Therefore, as CSF1 and FGF signals are both involved in the accumulation of tumor-infiltrating/promoting MDSCs and M2-TAMs, the dual inhibition of CSF1R and FGFRs may be more effective for cancer therapy than selective CSF1R inhibition.

FGF19, FGF21 and FGF23 are endocrine FGFs that transduce signals to target organs through FGFRs and the Klotho family of co-receptors, α-Klotho (KL) and β-Klotho (KLB) (117). FGF19 is upregulated by bile acid in the intestine to transduce endocrine signaling through FGFR4 and β-Klotho in the liver. FGF21 is upregulated by fasting in the liver and adipose tissue to transduce paracrine signaling through FGFR1 and β-Klotho locally and endocrine signaling to the pancreas and brain. FGF23 is upregulated by serum phosphate, vitamin D and parathyroid hormone in bone to transduce endocrine signaling through FGFR1 and α-Klotho in the kidneys and negative feedback signaling through FGFR3 and α-Klotho in the parathyroid gland (118). As endocrine FGFs are involved in the maintenance of whole-body homeostasis, FGFR inhibitors elicit endocrine or metabolic abnormalities. This section will be focused on adverse effects of FGFR inhibitors on endocrine FGF signaling in cancer patients.

FGF19 signaling through FGFR4 in the liver stimulates hepatocyte proliferation and glycogen synthesis but reduces bile acid synthesis and triglyceride synthesis (69,119). FGF19-FGFR4 signaling blockade in cynomolgus monkeys using anti-FGF19 monoclonal antibody causes hepatotoxicity, increased bile acid secretion and severe diarrhea (120). Fgfr4 knockout in mice also causes increased bile acid secretion in the liver, which leads to induction of Fgf15 (mouse ortholog of human FGF19) in the intestine and subsequent improvement of insulin resistance and glucose metabolism (121). The selective FGFR4 inhibitor, BLU9931, may be applied for the treatment of patients with hepatocellular carcinoma depending on FGF19-FGFR4 signaling. By contrast, as FGFR4 blockade is associated with a risk of liver toxicity, FGFR1/2/3 inhibitors rather than pan-FGFR inhibitors are preferable for the treatment of cancer patients with genetic alterations in FGFR1, FGFR2 or FGFR3, especially those with liver dysfunction (Fig. 5B).

Serum FGF23 elevation is a biomarker indicating on-target effects of FGFR1/2/3 and pan-FGFR inhibitors in cancer patients, whereas serum FGF23 levels are also elevated in patients with non-cancerous diseases, such as hypophosphatemic rickets and chronic kidney diseases (118). Physiological FGF23 signaling through FGFR1 and α-Klotho in the kidneys decreases the serum phosphate level through the downregulation of phosphate reabsorption. FGFR inhibitors, hindering FGF23 signaling in the kidneys, promote hyperphosphatemia and subsequent FGF23 secretion from bone and soft-tissue mineralization. Pathological FGF23 upregulation is associated with endothelial dysfunction and arterial stiffness (122). Pathological FGF23 signaling through FGFR4 in cardiac myocytes then induces phosphorylation of PLC-γ and activation of the IP₃-Ca²⁺ signaling cascade, which results in cardiac remodeling, such as cardiac hypertrophy and cardiac fibrosis (123). As FGF23-FGFR4 signaling activation is associated with a risk of cardiac toxicity, pan-FGFR inhibitors rather than FGFR1/2/3 inhibitors may be selected for the treatment of cancer patients with FGFR genetic alterations, particularly those with heart dysfunction (Fig. 5B).