Once FGFs bind to the extracellular domains of FGFRs, the dimerization of transmembrane and intracellular domains takes place along with a series of conformational changes that lead to trans-phosphorylation of dimerized kinase domains for activation.

The activation of kinase domain is a fined-tuned process (Furdui et al., 2006). Several critical tyrosine residues including Y463, Y583, Y585, Y653, Y654, Y730, and Y766 are autophosphorylated by precisely ordered reactions while ATP binds to the highly conserved pocket located in the hinge region during the activation of FGFR1 (Mohammadi, Dikic, et al., 1996). The autophosphorylations of Y653 and Y654 in the A-loop, which appear to induce the binding of substrate but not ATP, have increased by 50–100 fold and 500–1,000 fold in the rate of substrate phosphorylation, respectively. This suggested that these autophosphorylations have an indispensable role in kinase activation. The function of other tyrosine autophosphorylation sites contributes to the activation of FGFRs and downstream signal transduction through diverse biochemical reactions yet to be discovered.

In addition to the phosphorylation of critical tyrosine residues, the DFG motif of kinase domain toggles between two different conformations in line with the state of FGFR (active or inactive). When the motif adopts a DFG-in conformation, its Asp coordinates with phosphate groups of ATP and/or magnesium ion and causes the A-loop to display an open conformation, rendering the kinase an active state. Conversely, a DFG-out conformation, where the Asp and Phe point away from and toward the ATP binding pocket, respectively, is an indicator of inactive state of kinase. Noticeably, the flipped conformation of the DFG motif results in the formation of a large adjacent hydrophobic pocket (Hu et al., 2015; Vijayan et al., 2015). A valine in the ATP binding pocket (V561/564/555/550 in FGFR1/2/3/4), which is highly conserved in a variety of kinases and known as a “gatekeeper” residue, is the switch of the large hydrophobic pocket. The gatekeeper mutations give rise to many drug resistances due to hampered drug binding (Azam et al., 2008; Roskoski, 2010; Vijayan et al., 2015).

The autophosphorylated kinase domain can recruit and phosphorylate multiple downstream effector molecules to initiate several signaling pathways (Figure 3).

The Fibroblast Growth Factor Receptor Substrate 2 (FRS2), a major FGFR substrate, binds to the juxtamembrane region of FGFR via its phospho-tyrosine binding domain (PTB) in a constitutive manner, regardless of the activation and phosphorylation state of the kinase domain. Following the activation of FGFR, multiple tyrosine residues of FRS2 are subject to phosphorylation and serve as docking sites for subsequent molecules.

The RAS-MAPK-ERK signaling pathway is activated by a serial docking of FRS2 with multiple proteins, including SH2-containing protein tyrosine phosphatase (SHP2), growth factor receptor-bound protein 2 (GRB2), Son of Sevenless (SOS) and noted RAS. Both SHP2 and GRB2 contain a Src homology domain (SH2 domain), which can recognize and bind the phosphorylated tyrosine residues of FRS2 and GRB2. Therefore, the GRB2-SOS complex is recruited to FRS2 directly or through the formation of the SHP2-GRB2-SOS complex (Hadari et al., 1998; Ong et al., 2000). The complex, in turn, initiates a phosphorylation cascade in the RAS-MAPK-ERK signaling pathway. Upon activation, ERK1/2 is translocated from cytoplasm into nucleus and regulates the activity of diverse transcription factors to influence cell proliferation, differentiation and signal transduction, which makes it the most persuasive signaling molecules in this pathway for the evaluation of FGFR inhibitors (Guo et al., 2020).

The docking protein GRB2 associated binding protein 1 (GAB1) is recruited to the complex via binding to the SH3 domain of GRB2, which enables tyrosine phosphorylation on itself. Similarly, the phosphorylated tyrosine residues of GRB2 are captured by the phosphoinositide 3 kinase (PI3K) containing a SH2 domain, thus initiating the activation of PI3K-AKT signaling pathway. The downstream effector molecules of AKT vary, including the well-known mTOR, which is closely related to cell metabolism, transcription and so forth (Quan et al., 2020).

Besides FRS2, the phospholipase Cγ (PLCγ) binds to a phosphorylated tyrosine in the C-terminal of phosphorylated kinase domain, and hydrolyzes phosphatidylinositol 4,5-bisphosphate (PIP2) into two secondary messengers, inositol triphosphate (IP3) and diacyl glycerol (DAG). The binding between IP3 and its receptor on the endoplasmic reticulum leads to the release of Ca²⁺ from intracellular stores and thus increases Ca²⁺ concentration (Mikoshiba, 2007). When coordinated with Ca²⁺, DAG activates PKC signaling pathway, which causes crosstalk with RAS-MAPK pathway due to the competition between GRB2 and PLCγ to bind with FGFR (Fearon et al., 2014).

In addition, FGFR can activate the signal transducer and activator of transcription (STAT) proteins to partially mediate cell transformation (Hart et al., 2000).

The negative regulation of FGF/FGFR signaling includes dephosphorylation, ubiquitination and obstruction in a serial of docking. In response to FGF stimulation, an ubiquitin ligase called Casitas B-lineage lymphoma (Cbl) is recruited to the FRS2 (-SHP2)-GRB2-SOS complex and induces ubiquitination and subsequent degradation of FGFR and FRS2α (Wong et al., 2002). In addition, the binding of Sprouty to GRB2 can block the interaction between GRB2 and FRS2 or SHP2 so as to exert an inhibitory effect on downstream RAS-MAPK signaling (Hanafusa et al., 2002). The dual-specificity MAPK phosphatases 3 (MKP3) also inhibits RAS-MAPK signaling by dephosphorylating activated MAPK (Farooq et al., 2004).

The activations of these FGFR-dependent or related signaling pathways converge into the regulation of diverse cellular events and physical functions.

As the most commonly altered FGFR subtype, FGFR1 aberrations account for 49% of all cases with FGFR aberrations according to the sequencing analysis (Helsten et al., 2016). The most frequent type of FGFR1 aberrations is gene amplification, which is reported in 8.7–20.0% of non-small cell lung carcinoma (NSCLC) cases (Miao et al., 2016; Miao et al., 2020) and involved in several acquired resistances against NSCLC therapeutics (Gammelgaard et al., 2019; Zhang et al., 2019). FGFR1 amplification is common in breast cancer (10%), predominantly in estrogen receptor-positive breast cancer, and harmful to the survival of patients (Gelsi-Boyer et al., 2005; Wang et al., 2017). FGFR1 amplification is also seen in prostate cancer (15%) (Edwards et al., 2003), bladder cancer (9%) (Ross et al., 2014), and other cancers (ovarian cancer, colorectal carcinoma, and squamous non-lung tumors). FGFR amplification (mainly in FGFR1 and 2) causes overexpression of proteins and increases the FGFR-dependency of cancer cells. Therefore, it is regarded as a biomarker for the efficacy of some FGFR inhibitors (Weiss et al., 2010). FGFR1 mutation has been detected in several tpyes of cancers, including midline gliomas (18%), glioblastoma and melanoma, whereas FGFR1 fusion is rare.

Amplification (predominantly observed in triple-negative breast cancer, 4%) and mutation (e.g., K660N) of FGFR2 occur frequently in breast cancer. Besides, FGFR2 also amplifies in gastric cancer (4.5–9%) and is associated with its venous and lymphatic invasion (Jung et al., 2012; Cancer Genome Atlas Research Network, 2014). Apart from breast cancer, 10–12% of endometrial carcinoma and 4–5% of NSCLCs bear FGFR2 mutations (Mohammadi et al., 1998; Dutt et al., 2008; Kandoth et al., 2013). FGFR2 mutants are infrequently reported in urothelial cancers (1.2%). Several FGFR2 fusions have been reported including FGFR2-AFF3, FGFR2-CASP7 and FGFR2-CCDC6 (Turner et al., 2010; Reintjes et al., 2013; Wu et al., 2013). In addition, FGFR2 fusions occur in cholangiocarcinoma, lung squamous cell carcinoma (LSCC), thyroid cancer, prostate cancer, according to a study of FGFR targetable gene fusions (Wu et al., 2013). Notably, a FGFR2-PPHLN1 fusion generated by the chromosomal translocation t (10; 12) (q26; q12) is identified to possess oncogenic and transforming activity in 16% of intrahepatic cholangiocarcinoma (iCCA).

FGFR3 aberrations are predominantly implicated in bladder cancer (Baldia et al., 2016; Helsten et al., 2016). The incidence of FGFR3 mutations in non-muscle-invasive bladder cancer is as high as 75%, as determined by the presence of mutations in the p53 suppressor gene (Zhang et al., 2015), whereas it is relatively low (20%) in muscle-invasive bladder cancer (Couffignal et al., 2015; Solomon et al., 2016; Siracusano et al., 2020). Suppression of FGFR3 activation is sufficient to reduce the survival and proliferation of carcinoma cells harboring FGFR3 mutations (Markham, 2019; Montazeri et al., 2020). Furthermore, FGFR3 mutations are found in uterine cervical cancer (16.3%) (Yoshimoto et al., 2020), including invasive cervical cancer (5%) (Rosty et al., 2005), myeloma (2.16%) (Walker et al., 2015), and spermatocytic seminoma (6.66%) (Goriely et al., 2009). FGFR3 amplification is not frequent in cancers, but is sporadically reported in bladder cancer and adenoid cystic carcinomas (Vékony et al., 2007). Translocations at the t (4; 14), in which FGFR3 is significantly mutated, occurs in multiple myeloma (15%) frequently (Walker et al., 2018). Fusions of BAIAP2L1 or TACC3 to 5’ terminal of FGFR3 can also cause aberrant activation of FGFR3 by inducing oligomerization of fusion proteins even in the absence of FGFs. These fusions are reported in a variety of cancers including bladder cancer, LSCC, NLSCC, glioblastoma and oral cancer.

Amplification or mutation of FGFR4 is rarely perceived as an oncogene except in rhabdomyosarcoma (7.5%). It is confirmed that kinase inhibitor treatment increased cell apoptosis in FGFR4-mutant rhabdomyosarcoma (RMS) cell lines, which is consistent with increased SubG1 fraction and high level of activated caspase-3, suggesting the strong dependency of RMS on FGFR4 (Taylor et al., 2009).

The three approved FGFR inhibitors and quite a few candidates are all non-covalent inhibitors with pan-FGFR inhibitory activity, although some of them showed reduced, yet still considerable, potency for FGFR4 because of its relatively notable difference from FGFR1-3.

Erdafitinib (JNJ-42756493) is the first approved FGFR inhibitor for treatment of adult patients with locally advanced or metastatic urothelial carcinoma. It is an orally active and selective pan-FGFR inhibitor (Perera et al., 2017) that inhibits the kinase activity of FGFR1-4 with similar potency (Markham, 2019). Erdafitinib features quinoxaline element, which was first identified through virtual screening based on the crystal structure with FGFR1. The compound 2 was next generated through fragment growing approach. Removal of the methylene group in compound 2 produced compound 3, which has shown much improvement in activity due to better shape complementarity with the hydrophobic pocket. An additional substitution on the secondary nitrogen occupied the ribose-binding region, leading to the discovery of erdafitinib with increased affinity, better physicochemical and pharmacokinetic properties (Figure 6 ) (Murray et al., 2019).

In order to replace the pyrido [2,3-d]pyrimidin-7-one core, which was a common feature in a well-established class of protein kinase inhibitors (e.g., PD166285), Guagnano et al. (2011) developed a pseudo six-membered ring structure stabilized by an intramolecular hydrogen bond (e.g., prototype compound) (Figure 7) (Furet et al., 2008). Using the same strategy, infigratinib was eventually discovered through a lead optimization of a known FGFR inhibitor PD173074. Structural studies confirmed that hydrogen bonds with hinge region were retained and both chlorines and methoxy groups would form favorable hydrophobic contacts with the deep pocket inside ATP binding site. It was also observed that the phenyl ring at the entrance of the pocket was hydrophobically sandwiched between Leu478 and Gly561. These hydrophobic effects contributed to the selectivity of infigratinib for FGFR, especially FGFR1-3, over other tyrosine kinases. Infigratinib was approved by FDA for the treatment of cholangiocarcinoma patients with FGFR2 fusion in 2021 (Botrus et al., 2021; Javle et al., 2021).

Pemigatinib is another FGFR inhibitor featuring a tricyclic urea scaffold for the treatment of adults with previously treated, unresectable locally advanced or metastatic cholangiocarcinoma with a FGFR2 fusion or other rearrangement. Like infigratinib, pemigatinib contains a 3,5-dimethoxyphenyl for the high affinity and selectivity by filling the hydrophobic pocket, and two fluorine atoms, which led to further improvements in potency (Figure 8) (Wu et al., 2021). Many rounds of optimization also demonstrated that the ethyl group on the N-1 position of the cyclic urea and the pyrrole ring were of great importance to the potency and PK profile.

Starting from erdafitinib and its quinazolinone analogue, a series of pyrido [1,2-a] pyrimidinone derivatives were designed as novel selective FGFR inhibitors through scaffold hopping (Figure 9A) (Ran et al., 2021). Molecular docking suggested an overall similar binding mode with erdafitinib, while the rotatable pyrazole ring could lead to increased potency. The rotation also disrupted the planarity, which might enhance the aqueous solubility owing to reduced crystal-stacking.

AZD4547 is another pan-FGFR inhibitor bearing the pyrazole scaffold and just completed phase 2 clinical trial (NCT04439240) in multiple cancers with FGFR alterations. Based on the structure of AZD4547, Zhao et al. developed 1H-Pyrazolo [3,4-b]pyridine derivatives using scaffold hopping strategy, and incorporated two chlorines at the dimethoxyphenyl ring (Figure 9B) (Zhao et al., 2016). Notably, AZD4547 maintains efficacy to FGFR1 harboring gatekeeper mutation V561M as the flexible linker between dimethoxyphenyl and pyrazole allows conformational adaption within the hydrophobic region (Sohl et al., 2015). However, it remains unknown whether 1H-Pyrazolo [3,4-b]pyridine derivative (a) has compromised its activity for mutated FGFRs.

CH5183284 (Debio1374) is a potent pan-inhibitor of FGFR1-3 with IC₅₀ values of 9.3, 7.6 and 22 nM, respectively. It was discovered through a conventional high throughput screening and its interactions with the hinge region, hydrogen bonding pattern and an additional π-π interaction were also identified (Ebiike et al., 2016). CH5183284 is under clinical investigation for the treatment of cancer patients with FGFR genetic alterations (Nakanishi et al., 2014). Turner et al. (2017) described the application of SPROUT, a de novo computation program, to develop the active indazole-based pharmacophore for the inhibition of FGFR kinases ( Figure 9C ). Beginning with the co-crystal structure of CH5183284-FGFR1, they modified the indole moiety and obtained a small library of 23 indazole derivatives. Subsequent biological evaluation indicated that these indazole-containing fragments inhibited FGFR1-3 with IC₅₀ values of 0.8–90 µM, suggesting that structure-based drug discovery (SBDD) is becoming an essential tool for identifying potent and selective FGFR inhibitors.

There are numerous other non-covalent pan-FGFR inhibitors in clinical trial or development. For example, ASP5878 inhibited cell proliferation of urothelial cancer cell lines harboring FGFR3 point mutation or fusion and has completed phase 1 clinical trial in 2017 (NCT02038673) (Kikuchi et al., 2017). LY2874455 is a phase 1 orally available inhibitor with IC₅₀ values of 2.8, 2.6, 6.4, and 6.0 nM against FGFR1-4, respectively (Michael et al., 2017). Rogaratinib (BAY 1163877) is another potent and selective FGFR1-4 inhibitor (Collin et al., 2018). Rogaratinib alone or in combination with other agents have been in a few clinical trials. 3D185, a highly selective FGFR1-3 inhibitor, was approved for investigational new drug by the NMPA in 2018 and followed by a phase 1 study in patients with the advanced solid tumors (NCT04221204). ICP-192 is a pan-inhibitor against FGFR1-4 and entered the phase 1/2 clinical trial for the treatment of advanced solid tumors, urothelial carcinoma, and cholangiocarcinoma (NCT04565275). E7090 has favorable pharmacokinetic profiles and sub-nanomolar inhibitory activity against FGFR1-3 with IC₅₀ values of 0.71, 0.50, and 1.2 nM, respectively (Watanabe Miyano et al., 2016). The phase 2 study of E7090 in participants with unresectable advanced or metastatic cholangiocarcinoma with FGFR2 fusion is recruiting (NCT04238715).

Covalent inhibition is a re-emerging strategy especially in the development of kinase inhibitors, which can make a big difference in binding affinity and selectivity. A covalent inhibitor typically consists of a drug-like scaffold offering noncovalent interactions and an appropriate electrophilic warhead to react with nucleophilic residues of target proteins. Cysteine represents the most targeted amino acid in kinases, due to its non-catalytic roles, low abundance, high reactivity and chemical plasticity of the anionic thiolate (Abdeldayem et al., 2020; Galbiati et al., 2020). For FGFRs, the conserved cysteine in the P-loop (C488 in FGFR1, C491 in FGFR2, C482 in FGFR3 and C477 in FGFR4) and the unique C552 in FGFR4 in the hinge region are sites for covalent attachment (Dai et al., 2019). This section focuses on recent publications with regard to the discovery of covalent pan-FGFR covalent inhibitors.

Zhou et al. (2010) discovered FIIN-1 as the first irreversible inhibitor of FGFR1−4 in 2010. The acrylamide of FIIN-1 formed covalent bond with a conserved cysteine (Cys488 of FGFR1) located at the rim of the P-loop. Replacing the acrylamide with a propylamide led to the failure of covalent bond formation. In addition, its selectivity over some other kinases (e.g., c-Src, TNK1, and YES) bearing the P-loop cysteine at the same position as FGFRs was also confirmed. Tan et al. (2014) developed FIIN-2 and FIIN-3 as irreversible inhibitors with potent in vitro inhibitory activity against FGFR1 and FGFR2 gatekeeper mutants, which conferred resistance to first-generation FGFR inhibitors. The acrylamide moiety in both molecules was installed on the 4-poistion of their phenyl rings in contrast to 3-acrylamide as found in FIIN-1, which still maintained the bond formation with P-loop cysteine while changed the selectivity over other kinases (PDB 4QQC, 4R5S, 4R6V).

Brameld et al. (2017) developed another irreversible inhibitor, PRN-1371, which shared a common core with FIIN-1. PRN-1371 was proven to be highly selective for FGFR1-4 over other kinases, including KDR, FLT-4, etc., and showed high FGFR1 occupancy and ideal PK profile. Ding et al. focused on the modification of the acrylamide-containing side chain of FIIN-1 and obtained the promising lead compound showing inhibitory effect in FGFR-amplified cancer cell lines ( Figure 10A) (Li et al., 2017). The crystal structure of FGFR1 C488A in complex with lead compound revealed that the acrylamide side chain was located in the solvent accessible space and easily performed nucleophilic attack by the target cysteine.

Goyal et al. (2019) demonstrated that Futibatinib (TAS-120), a highly selective and irreversible pan-FGFR inhibitor, exhibited in vitro potency against wild-type FGFR1-4, as well as some FGFR2 kinase domain mutants. Hiroshi Sootome et al. (2020) reported the preclinical profile of futibatinib and suggested that it is an orally available, potent pan-FGFR inhibitor. Futibatinib is the most advanced candidate in the category of covalent pan-FGFR inhibitors because it is in phase 3 clinical trial to evaluate the treatment of metastatic and recurrent unresectable cholangiocarcinoma harboring FGFR2 gene rearrangements.

Wang et al. (2018) designed and discovered a set of irreversible inhibitors bearing novel pyrrolopyrimidine scaffold. By analyzing the ATP binding pocket, they introduced a substituted phenyl moiety at the C-3 position of pyrrolopyrimidine to interact with hydrophobic region I ( Figure 10B ). Then the electrophilic warhead attached to N-1 via a pyrrolidine linker led to the identification of lead compound that showed excellent potencies against FGFR1−4 and acceptable selectivity over VEGFR2.

Wei et al. (2021) recently reported a class of 2H-pyrazolo [3,4-d]pyrimidine derivatives as a potent irreversible pan-FGFR inhibitor (Figure 10C). The lead compound was derived from BTK inhibitor ibrutinib and also shares a similar core with the above-mentioned lead compound. Interestingly, an electron-withdrawing terminal-CF₃ substituted acrylamide group provided the most potent inhibition against FGFRs.

There are several other covalent pan-FGFR inhibitors in development. For example, Yamani et al. (2021) discovered a pyrazole-benzimidazole CPL304110 as a pan-FGFR inhibitor for the treatment of bladder, gastric and squamous cell lung cancer, which also showed favorable pharmacokinetic properties after oral administration. Dai et al. (2020) reported that DW14383 simultaneously inhibited tumor proliferation and angiogenesis via inhibition of FGFR1–4 with similar potency. In addition, they claimed that its pan-tumor spectrum potential might overcome compensatory activation among FGFR1–4. Wang et al. (2019) developed a series of compounds featuring pyrazolo [3,4-d] pyridazinone as covalent FGFR inhibitors. Their structural optimization resulted in more analogues that could remarkably inhibit proliferation of various FGFR-dysregulated cancer cells and display potent antitumor efficacy in xenograft model as well (Figure 10D) (Xie et al., 2020).

The kinase domains of FGFR1–3 share high structural similarity, whereas FGFR4 is relatively distinct from FGFR1-3 (Babina et al., 2017), which is consistent with the fact that many foregoing pan-FGFR inhibitors show strong inhibition of FGFR1-3 but reduced potency for FGFR4. Detailed comparison of the active sites of FGFR1-3 and FGFR4 revealed a key difference in the hinge region: Tyr563 in FGFR1-3 versus the Cys552 in FGFR4 ( Figure 11 ) (Tucker et al., 2014). This unique Cys552 provides great opportunity for the design of highly selective, covalent inhibitors of FGFR4. Although there is no approved FGFR4-specific drug, the past years have witnessed growing numbers of promising compounds as discussed below.

Hagel et al. (2015) discovered BLU9931 as the first selective FGFR4 inhibitor for the treatment of hepatocellular carcinomas (HCC) with aberrant FGFR4 signaling. The dimethoxyphenyl group of BLU9931 occupied the hydrophobic pocket located near the gatekeeper valine of all FGFRs. In addition, an acrylamide at the ortho-position of the aniline can form the covalent bond with Cys552 in the hinge region of FGFR4. To achieve a better selectivity, the rotation of the phenyl ring was also taken into consideration because it could cause steric clash with the corresponding tyrosine in FGFR1-3 hinge region. Moreover, the addition of 3-methyl group on the aniline ring rendered BLU9931 with high selectivity for FGFR4 over FGFR1-3 (Figure 12A). Unfortunately, BLU9931 failed to enter clinical stage, presumably due to its rapid metabolism in liver microsomes. BLU554 (fisogatinib), an orally available analog of BLU9931, is now in phase 1 clinical trial to treat patients with HCC (NCT02508467) and in phase 1b/2 clinical trial in combination with CS1001.

H3B-6527 (Joshi et al., 2017) is another highly selective covalent FGFR4 inhibitor and is currently undergoing phase 1 study for the treatment of advanced HCC, liver neoplasms, hepatic carcinoma and so on. Co-crystal structure of FGFR4 in complex with H3B-6527 illustrated the covalent bond formation between Cys552 and the acrylamide at the ortho-position of the N-aryl substituent.

Fairhurst et al. (2020) reported the discovery of FGF401 (roblitinib) as a potent, selective FGFR4 inhibitor. Through high throughput screening, 2-formylquinoline amide (2-FQA) derivatives were identified as the starting hits. After optimization of the 2-FQA and the substituent groups on the hinge-binding pyridyl ring, roblitinib was eventually obtained with outstanding selectivity for FGFR4. Being assessed in phase 2 clinical trial, roblitinib is the most advanced covalent FGFR4-specific inhibitor. This inhibitor features a covalent yet rapidly reversible mode of action that may reduce off-target related toxicity. Therefore, roblitinib is regarded as a promising next-generation drug to offer a new approach to target FGFR covalently.

Starting from BGJ-398 (infigratinib), Miranda et al. developed a novel covalent inhibitor of FGFR4 for the treatment of HCC (Rezende Miranda et al., 2020). According to the crystal structure of FGFR1-BGJ-398, the urea group and aminopyridine group should exhibit similar geometrical and electronic properties ( Figure 12B). An acrylamide group was also attached to the ortho position of a 2-methylaniline ring for targeting Cys552 of FGFR4. Interestingly, crystallographic study revealed that the introduction of the methyl group into aniline phenyl rings facilitates the covalent reaction from a conformational perspective (PDB: 6JPE). As expected, its exceptional selectivity among the FGFR family is due to the fact that the acrylamide group formed steric clash with the bulky tyrosine residue present in FGFR1-3 hinge region.

Fairhurst et al. (2017) reported the discovery of dipyridylamine through high throughput screening (HTS). Dipyridylamine contains a novel warhead 3-nitro-6-chloropyridyl, which was designed to achieve covalent binding with Cys552 of FGFR4 (Figure 12C ). The 6-chloro substituent in this warhead is positioned for attack by the Cys552 thiolmethyl group through a nucleophilic aromatic substitution (S_NAr). Dipyridylamine demonstrated high potency against FGFR4 with IC₅₀ value of 53 nM while sparing the FGFR1-3 with IC₅₀ values higher than 10 μM. Besides, each nitrogen atom in pyridyl ring formed a hydrogen bond with hinge residue Ala553 (Lu et al., 2019). Dipyridylamine features a relatively low molecular weight and novel mechanism of covalent binding that may serve as a promising lead compound for future discovery of FGFR4-specific covalent inhibitors.

Several candidates have entered clinical stages without full disclosure of chemical structures. For example, INCB-62079 entered phase 1 trial in 2017 but was terminated for business strategic consideration. ICP-105, a selective FGFR4 inhibitor, is now in phase 1 clinical trial for the treatment of solid tumor (NCT03642834). Other clinical trials involving inhibitors including ZSP-1241 and ABSK-011 are actively recruiting patients.

Unlike FGFR4, few FGFR1-3 subtype-specific kinase inhibitors have been reported to date, mainly aiming at FGFR2. Casaletto et al. (2021) recently reported RLY-4008 as a highly selective inhibitor of FGFR2 WT/mutant, which exhibited >200-fold higher potency than FGFR1. Although the structure has not been disclosed, RLY-4008 showed no difference in binding modes with FGFR1 or FGFR2. Instead, a flexible loop in FGFR1/2 validated from MD simulation might be the cause of the selectivity. It is encouraging that a recent de novo design by Turner et al. (2021) has provided a paradigm for perhaps the next-generation member of FGFR-specific inhibitors. They started from a fragment with moderate potency and carried out iterative rounds of de novo design as well as a classical SAR study to generate compound 31. Interestingly, compound 31 specifically inhibited FGFR2 with an IC₅₀ of 29 nM, whereas 389 nM for FGFR1 and 758 nM for FGFR3, suggesting at least 10-fold selectivity for FGFR2 over FGFR1 (Figure 13).

An extracellular allosteric inhibitor of FGFR can bind to its extracellular domain and modulate the receptor conformation, thus blocking the signaling pathway. The extracellular domains of FGFR1-4 exhibit discernible structural differences compared to their kinase domains, therefore it is possible to achieve selective inhibition within FGFR members. Alofanib (RPT835) belongs to this type of inhibitors and has entered a phase 1 clinical trial in Russia. By specifically binding to IgIII of FGFR2, alofanib is able to inhibit the FGF2-induced phosphorylation of FRS2α with nanomolar activity in cancer cells expressing different FGFR2 isoforms (Tyulyandina et al., 2017). In addition, Tsimafeyeu et al. (2016) performed in vivo experiments to demonstrate that alofanib could ablate FGF-induced angiogenesis. Another example is SSR128129E, an orally-active, allosteric inhibitor of FGFR1 (Herbert et al., 2013). It interacts with extracellular part of FGFR without interfering with FGF binding or receptor dimerization. Critical conformational changes were observed in IgIII subdomain when treated with SSR128129E, resulting in defective internalization of FGFRs.

Efforts have been made to identify natural products that act on FGFRs. These include a few phenolic compounds (resveratrol, caffeic acid phenethyl ester, kaempferol, etc.), stilbene glycosides, and sesterterpenes compounds (leucosesterterpenone, leucosterlactone, etc.), and a detailed review has been completed by Yin et al. (2019). Recently, Pagano et al. (2021) reported rosmarinic acid (RA), a natural phenolic compound, could induce FGF2/FGFR complex dissociation as verified by experimental mechanistic study. In addition, gambogenic acid and ferulic acid, originated from natural plants, exhibited inhibitory effect of FGFR autophosphorylation. Taken together, natural products stand for a prime source of FGFR inhibitors, while more studies are needed to improve their activities and elucidate the molecular mechanisms. Chemical structures of these represent natural products are shown in Supplementary Figure S3 .