Discovery of Orally Bioavailable FGFR2/FGFR3 Dual Inhibitors via Structure-Guided Scaffold Repurposing Approach

Abstract

Fibroblast growth factor receptors (FGFRs) are transmembrane receptor tyrosine kinases that regulate multiple physiological processes. Aberrant activation of FGFR2 and FGFR3 has been linked to the pathogenesis of many tumor types, including cholangiocarcinoma and bladder cancer. Current therapies targeting the FGFR2/3 pathway exploiting small-molecule kinase inhibitors are associated with adverse events due to undesirable inhibition of FGFR1 and FGFR4. Isoform-specific FGFR2 and FGFR3 inhibitors that spare FGFR1 and FGFR4 could offer a favorable toxicity profile and improved therapeutic window to current treatments. Herein we disclose the discovery of dual FGFR2/FGFR3 inhibitors exploiting scaffold repurposing of a previously reported ALK2 tool compound. Structure-based drug design and structure–activity relationship studies were employed to identify selective and orally bioavailable inhibitors with equipotent activity toward wild-type kinases and a clinically observed gatekeeper mutant.

The mammalian fibroblast growth factor receptor (FGFR) family comprises four tyrosine kinases designated as FGFR1–4. Each member of the FGFR family consists of an extracellular ligand-dependent domain, a single transmembrane domain, and a cytoplasmic domain containing the catalytic protein tyrosine kinase.¹ The binding of fibroblast growth factor (FGF) or other ligands to the extracellular domain of FGFR induces receptor dimerization, followed by activation of the kinase domain, and transphosphorylation of the receptor, which in turn triggers multiple downstream signals, including the phospholipase Cγ, PI3K-AKT, and RAS-MAPK pathways. FGFR signaling plays a critical role in a variety of physiological processes, including embryonic development, fetal organogenesis, metabolic homeostasis, and tissue repair and regeneration.^2,3 Genetic alterations of FGFR2 and FGFR3, such as mutation, fusion/rearrangement, and amplification, drive activation of the receptors and the associated pathways. These genetic aberrations are associated with the development and/or progression of cancer. Alterations of the FGFR3 gene are frequently found in bladder cancer and multiple myeloma. FGFR2 genetic alterations are mainly found in cholangiocarcinoma, gastric cancer, and uterine cancer.⁴ Besides indications where FGFR2/3 dysregulation is the primary driver of tumorigenesis, FGFR2/3 alteration has been reported as a mechanism of resistance to cancer treatments that target other signaling pathways.^5,6

To date, four small-molecule FGFR inhibitors have been approved by the FDA to treat cancers associated with FGFR2 and/or FGFR3 aberrations (Figure 1). The pan-FGFR inhibitor erdafitinib⁷ was granted approval for treatment of metastatic or locally advanced bladder cancer. Pemigatinib⁸ and infigratinib,⁹ both of which are reversible FGFR1/2/3 inhibitors, were approved to treat locally advanced or metastatic cholangiocarcinoma. Futibatinib,¹⁰ an irreversible pan-FGFR inhibitor, was approved for metastatic intrahepatic cholangiocarcinoma. In addition, several FGFR kinase inhibitors are currently in clinical development.¹¹ Although current therapies targeting the FGFR2/3 pathway exploiting small-molecule kinase inhibitors have achieved significant progress, the lack of subtype specificity in these first-generation inhibitors invariably results in undesirable side effects and/or a narrow therapeutic index. The most frequently reported adverse events in patients treated with current FGFR inhibitors include hyperphosphatemia and diarrhea, which are attributed to inhibition of FGFR1 and FGFR4, respectively.¹²⁻¹⁴ Thus, isoform-specific FGFR2/3 inhibitors¹⁵ that spare FGFR1 and FGFR4 could be employed to optimize tolerability and efficacy in FGFR2/3-altered tumors. Herein we report the discovery and development of potent, selective, and orally bioavailable dual inhibitors of FGFR2 and FGFR3 exploiting scaffold repurposing of a previously disclosed activin receptor-like kinase 2 (ALK2) tool compound.

Figure 1.

FDA-approved FGFR inhibitors and their biological targets. First-generation FGFR kinase inhibitors share the 3,5-dimethoxyphenyl motif (highlighted) that interacts with the gatekeeper region of the ATP binding pocket.

Obtaining small-molecule inhibitors targeting FGFR2 and FGFR3 isoforms selectively is challenging due to the high homology of the FGFR family.¹⁶ At the time we initiated our discovery campaign, none of the reported FGFR inhibitors displayed selectivity toward the FGFR2 and/or FGFR3 isoform. Furthermore, currently approved FGFR inhibitors share a common structural motif (i.e., the 3,5-dimethoxyphenyl group) that fills a hydrophobic pocket in the region near the gatekeeper residue of the FGFR kinase domain (Figure 1). As a result, clinical resistance to current FGFR-targeted therapy via gatekeeper mutation has been observed and suggests the need to identify a new chemotype that does not target the gatekeeper region.^17,18 During our search for a suitable starting point for next-generation FGFR kinase inhibitors, we became interested in compound 1 (Figure 2), which was reported previously as a potent and selective in vivo tool for ALK2 inhibition (IC₅₀ = 0.98 nM).¹⁹ Broad kinase panel screening data obtained from the off-target characterization of 1 demonstrated a clean profile with only modest activity toward FGFR3 (IC₅₀ = 178 nM). Given the unique chemical structure of 1 in comparison to all other reported FGFR inhibitors, we further profiled this molecule against the remaining FGFR isoforms in enzymatic biochemical assays. Interestingly, 1 possesses similar activity for FGFR2 (IC₅₀ = 213 nM), while displaying a 3-fold selectivity against FGFR1 (IC₅₀ = 577 nM), which further substantiated our interest in initiating a structure–activity relationship (SAR) campaign to develop dual FGFR2/3-selective inhibitors based on this scaffold.

Figure 2.

Lead identification via scaffold repurposing.

Our initial focus was to demonstrate the possibility of abolishing ALK2 activity while maintaining FGFR2/3 potency for this chemical series. To that end, the pyrazolo[4,3-d]pyrimidine bicyclic core was kept intact while the SAR study was performed on the remaining areas of the molecule (i.e., the R₁ and R₂ vectors of 2; Figure 2). Analysis of the binding mode of compound 1 to ALK2 suggested that in addition to the hydrogen bonds between the bicyclic core and the ALK2 hinge region (His284 and His286), there are two key interactions that drive potency toward ALK2. First, the methyl carbamate group forms a strong hydrogen bond with the catalytic lysine Lys235. Second, the methylpiperazine group, protonated at physiological pH, forms a hydrogen bond with Asp293 in front of the ATP binding pocket (Figure 3). Thus, removal of these structural features was anticipated to reduce ALK2 activity. In addition, SAR data from a previous ALK2 discovery campaign showed that replacement of the central phenyl ring in 1 with a five-membered heteroaryl group dramatically reduced the ALK2 potency. Preliminary SAR exploration on the basis of this logic identified pyrazole 3 (Figure 2) as an early lead, displaying moderate FGFR3 enzyme inhibition (FGFR3 IC₅₀ = 266 nM) with no significant ALK2 activity (ALK2 IC₅₀ > 500 nM). Encouraged by the promising result, further SAR study was performed on the piperazinone ring in 3, and key results are highlighted in Table 1. At this stage of the program, compounds were triaged based on FGFR3 potency and selectivity over FGFR1. Importantly, characterization of the previously disclosed ALK2 chemical series revealed that compounds possessing a cis-2,6-dimethylpiperazine group (e.g., 1) displayed improved FGFR3 potency in comparison to the unsubstituted piperazine analogues, suggesting an opportunity to enhance the potency of the new lead (i.e., 3) by substitution at C-2 and/or C-6 of the piperazinone ring. Following-up on the above observation, incorporation of a methyl group at C-6 furnished compound 4 with no improvement in FGFR3 potency or selectivity. On the other hand, methyl group substitution at C-2 yielded compound 5 displaying potent FGFR3 activity (FGFR3 IC₅₀ = 19 nM) with good selectivity against FGFR1 (18-fold). Interestingly, gem-dimethyl incorporation at C-2 significantly diminished the FGFR3 potency and selectivity (6). Pleasingly, compound 7, featuring an ethylene bridge between C-2 and C-6 of the piperazinone, displayed single-digit-nanomolar potency against FGFR3 (FGFR3 IC₅₀ = 5.6 nM) while maintaining good selectivity over FGFR1 (16-fold). This bridged piperazinone group was selected for further investigation. An additional SAR vector was explored by substitution at N-4 of the piperazinone ring (e.g., 8–10). However, no improvement in either FGFR3 potency or selectivity against FGFR1 was observed. Subtle modifications such as replacement of the N-4-methyl with the slightly larger ethyl group (8) led to a 3-fold decrease in FGFR3 potency. Overall, the above SAR studies demonstrated the pyrazolo[4,3-d]pyrimidine core to be a suitable scaffold to develop a new class of highly potent FGFR binders not previously reported in the literature.

Figure 3.

Proposed binding mode of 1 in the ATP pocket of ALK2. Key interactions highlighted include hydrogen bonds in the hinge region with the backbone CO and NH of His284 and His286 as well as hydrogen bonds with Lys235 and Asp293.

Table 1. SAR Exploration of Piperazinone Analogues.

^aCompounds 4, 5, and 7–10 are racemates.

^bFold selectivities over FGFR1 are indicated in parentheses.

With the exciting improvement in FGFR3 potency and selectivity, advanced profiling of 7 was warranted. The biochemical potency of this lead compound was examined across four isoforms of the FGFR family (Table 2). Compound 7 potently inhibited FGFR3 and FGFR2 kinases (IC₅₀ = 5.6 and 5.2 nM, respectively) while displaying favorable selectivity against FGFR1 (IC₅₀ = 89 nM) and FGFR4 (IC₅₀ = 351 nM). In addition, compound 7 also exhibited excellent selectivity against ALK2 (IC₅₀ = 930 nM) in the biochemical enzymatic assay, validating our scaffold repurposing approach. Pleasingly, 7 showed equipotent activity when tested against the clinically observed FGFR3 gatekeeper mutant V555L²⁰ (IC₅₀ = 3.8 nM), demonstrating the ability of this chemical series to overcome resistance due to gatekeeper mutation. Importantly, 7 was identified as a potent inhibitor of phosphorylated FGFR3 kinase activity in cells (IC₅₀ = 70 nM) by measuring the inhibitory effect of phosphorylated FGFR3 in Ba/F3 cells stably expressing the TEL–FGFR3 fusion proteins. Further analysis of physicochemical properties and ADME profiles revealed that 7 exhibited high metabolic stability in human liver microsomes (h-IntCL = 0.6 L h^–1 kg^–1) and high free fraction in human plasma (%free = 76%). However, 7 displayed low Caco2 permeability (P_m = 0.5 × 10^–6 cm/s) which translated to low absorption following oral administration in a rat pharmacokinetics (PK) study.

Table 2. In Vitro Properties of Lead Compound 7.

FGFR2 IC50	5.2 nM	FGFR3-V555L IC50	3.8 nM
FGFR3 IC50	5.6 nM	FGFR3-BAF3 IC50	70 nM
FGFR1 IC50	89 nM	CACO2 Pm	0.5 × 10–6 cm/s
FGFR4 IC50	351 nM	h-IntCL	0.6 L h–1 kg–1
ALK2 IC50	930 nM	h-PB (%free)	76%

Having identified a suitable lead compound, we continued our SAR campaign to enhance the FGFR3 potency as well as the pharmacokinetic profile via fine-tuning of the physicochemical properties. We reasoned that the low Caco2 permeability could be addressed by increasing compound lipophilicity. Thus, attention was next focused on designing inhibitors with increased cLogP compared to 7 (cLogP = 1.2). Molecular modeling of 7 with the FGFR3 kinase structure suggested that introducing lipophilicity at the methyl pyrazole fragment of the molecule would be tolerated. To this end, a variety of substituted pyrazoles were explored, and key examples are highlighted in Table 3. Replacing the N-Me with N-Et afforded compound 11 with similar potency and selectivity compared with 7. A further increase in size of the alkyl substituent such as isopropyl (12) or methoxyethyl (13) diminished the FGFR3 potency. Addition of a basic morpholinoethyl group (14) led to a 3-fold drop in FGFR3 potency, albeit with some improvement in selectivity against FGFR1 (19-fold). Installation of a cyclopropyl group furnished compound 15 with potent FGFR3 activity (IC₅₀ = 5.9 nM) and reasonable selectivity against FGFR1 (10-fold). It was particularly interesting that compound 16, bearing a cyclobutyl group on the pyrazole ring, displayed a noted improvement in FGFR3 potency (IC₅₀ = 3.2 nM) with favorable selectivity against FGFR1 (14-fold). Several larger saturated carbocyclic and heterocyclic substituents were incorporated, and no further beneficial effect on FGFR3 potency or selectivity against FGFR1 was observed (e.g., 17 and 18).

Table 3. SAR Exploration of Pyrazole Analogues.

^aRacemates.

^bFold selectivities over FGFR1 are indicated in parentheses.

Considering the favorable profile of 16, the individual enantiomers were separated for characterization. Only one enantiomer (compound 19; Figure 4A) displayed significant activity in the biochemical enzymatic assays, and the stereochemistry (1R,5S) of this active enantiomer was assigned based on chemical synthesis. The binding mode of 19 with the FGFR3 kinase structure (PDB code 4K33) was evaluated through in silico docking studies. As shown in Figure 4B, we propose that 19 binds in the ATP binding pocket of FGFR3 with the pyrazolopyrimidine fragment making hydrogen bonds with the backbone NH and CO of Ala558 and Glu556 in the hinge region. The bicyclic core also exhibits favorable van der Waals interactions with the side chains of Val486 and Leu478. The cyclobutylpyrazole group extends toward the solvent-exposed region, with the pyrazole making a van der Waals interaction with Gly561. The bridged piperazinone occupies the pocket under the P loop, wherein the ethylene bridge is oriented toward a hydrophobic groove formed by Leu624 and Ala634, placing the amide carbonyl group in the proper position to form a strong hydrogen bond with the catalytic lysine Lys508 (Figure 4C). This key hydrogen bond and the shape complementarity provided by the ethylene bridge contribute to the improved potency observed for 19 in comparison to the other piperazinone analogues. Importantly, analysis of the binding mode of 19 reveals no interaction with the gatekeeper region (Val555), explaining the ability of this chemical series to maintain potency against the gatekeeper mutant. Given the very high structural homology of the FGFR family, current molecular modeling is ineffective at rationalizing the observed FGFR isoform selectivity of 19. Presumably, the favorable FGFR2/3 selectivity is attributed to a better shape complementarity between 19 and the FGFR2/3 binding pockets.

Figure 4.

Proposed binding mode of compound 19 in the ATP binding pocket of FGFR3. (A) Structure of 19. (B) Interaction of 19 with FGFR3 in the ATP binding pocket. (C) Molecular surface showing the ethylene bridge filling the hydrophobic pocket under the P loop and the amide carbonyl forming a hydrogen bond to the catalytic Lys508.

Detailed characterization of compound 19 is provided in Table 4. Specifically, 19 potently inhibited FGFR2, FGFR3, and the gatekeeper mutant FGFR3-V555L (IC₅₀ = 1.8, 2.0, and 1.5 nM, respectively) with good selectivity against FGFR1 (IC₅₀ = 27 nM) and FGFR4 (IC₅₀ = 157 nM) in enzymatic assays. This molecule also demonstrated potent FGFR3 inhibition in Ba/F3 cells (IC₅₀ = 16 nM). Activity against FGFR2 was also measured in KATOIII cells, an FGFR2-amplified gastric cell line, wherein compound 19 potently blocked phosphorylation of FGFR2 with an IC₅₀ value of 24 nM. As discussed above, the corresponding enantiomer of 19 displayed weak activity in enzymatic biochemical assays, with FGFR3 and FGFR1 IC₅₀ values of 482 nM and 4.5 μM, respectively. To determine the ability of 19 to attenuate downstream signaling (pERK) in the presence of human serum proteins, a whole blood (WB) assay was developed employing the RT112/84 bladder carcinoma cell line. Pleasingly, compound 19 displayed potent inhibition of pERK signaling in the WB setting with a WB IC₅₀ value of 209 nM. Following in vitro ADME profiling, compound 19 demonstrated moderate metabolic stability in human liver microsomes (h-IntCL = 1.0 L h^–1 kg^–1) and a moderate free fraction in the human plasma protein binding study (%free = 11%). Pleasingly, 19 displayed significant improvement in permeability (P_m = 4.0 × 10^–6 cm/s) across Caco-2 cell monolayer compared to that of 7 (P_m = 0.5 × 10^–6 cm/s). This compound also exhibited good solubility in both fasted-state simulated intestinal fluid (FaSSIF) (557 μg/mL) and simulated gastric fluid (SGF) (>400 μg/mL) conditions. In addition, 19 was not a potent CYP inhibitor, with IC₅₀ values greater than 10 μM for all major CYP isoforms tested (1A2, 2B6, 2C8, 2C9, 2C19, 2D6, and 3A4), and showed a low hERG patch clamp activity (<10% inhibition at 10 μM). Compound 19 exhibited good selectivity over ALK2 (IC₅₀ = 477 nM) as well as our in-house panel of 52 additional kinases, where only KDR (IC₅₀ = 184 nM) was identified as having an IC₅₀ value less than 500 nM (Table S2). In pharmacokinetics experiments, 19 demonstrated an excellent rat PK profile with moderate systemic clearance, high oral exposure, and complete bioavailability (100%). In cynomolgus monkeys, 19 exhibited low systemic clearance, good oral exposure, and moderate oral bioavailability (20%). Importantly, the favorable PK data observed for 19 across species validated our strategy of addressing the suboptimal pharmacokinetic profile of lead compound 7 by improving the compound’s cellular permeability.

Table 4. In Vitro Data and Pharmacokinetic Profile of 19.

In Vitro Profilea
FGFR2 IC50	1.8 nM	WB IC50	209 nM
FGFR3 IC50	2.0 nM	CACO2 Pm	4 × 10–6 cm/s
FGFR1 IC50	27 nM	h-IntCL	1.0 L h–1 kg–1
FGFR4 IC50	157 nM	h-PB (%free)	11%
FGFR3-V555L IC50	1.5 nM	FaSSIF	557 μg/mL
FGFR3-BAF3 IC50	16 nM	SGF	>400 μg/mL
FGFR2-KATOIII IC50	24 nM	hERG patch	<10% inh at 10 μM

Pharmacokinetic Profileb
	Cmax (p.o., nM)	AUCinf (p.o., nM h)	T1/2 (h)	Vss (L/kg)	CL (L h–1 kg–1)	F (%)
rat	5553	5819	6.6	2.6	1.4	100
cyno	825	2215	3.2	0.66	0.7	20

^aEach IC₅₀ value represents the mean of at least two independent determinations with standard deviation <50%.

^b1 mg/kg i.v. and 3 mg/kg p.o.

On the strength of these data, 19 was selected as an in vivo tool to measure inhibition of FGFR downstream signaling in mice. Following oral administration into NSG mice at 30 mg/kg, 19 demonstrated high absorption with a plasma concentration of 760 nM measured 2 h after dose administration (ca. 3.6-fold over WB IC₅₀). The inhibition of the pERK level in the RT112/84 human epithelial bladder cell carcinoma model was then investigated to observe the impact of 19 on the RAS–MAPK pathway. Tumor-bearing mice were dosed by oral gavage with 19 at 30 mg/kg, and tumors were analyzed 2 h after dose administration. Treatment with 19 showed a 57% reduction in pERK in the RT112/84 model (Figure 5). Significantly, compound 19 at 30 mg/kg (760 nM plasma concentration) displayed target inhibition of pERK comparable to a high oral dose (10 mg/kg, 4.9 μM plasma concentration) of pemigatinib, a potent FGFR1/2/3 inhibitor employed as a positive control in this pharmacodynamics (PD) study. The good PK profile and the observed pERK modulation in tumors support further development of 19 and related analogues as potential therapeutic agents for FGFR2/3 targeted therapies.

Figure 5.

PD Study of 19 in the RT112/84 mouse model. Data are presented as mean ± SEM of three mice per group.

In summary, we have disclosed here a series of pyrazolopyrimidines as dual inhibitors of FGFR2 and FGFR3 via scaffold repurposing of a previously reported ALK2 tool compound. Medicinal chemistry rationale and structure–activity relationship studies led to the discovery of 19 displaying equipotent activity toward wild-type kinases and the gatekeeper mutant, high selectivity against the FGFR1 and FGFR4 isoforms, and favorable pharmacokinetic profiles across multiple species. Importantly, the pyrazolopyrimidines reported herein represent a new chemotype that is structurally distinct from previously reported FGFR inhibitors and is suitable for further development to treat cancers associated with FGFR2 and/or FGFR3 aberrations. The selection of the lead-finding strategy to identify chemical starting points is often regarded as one of the most important decisions made by drug discovery teams. This report showcases scaffold repurposing, employed in combination with structure-based drug design, as an effective lead generation approach for the rapid identification of in vivo tool compounds or development candidates.

Glossary

Abbreviations

FGFR

fibroblast growth factor receptor

ALK

activin receptor-like kinase

FGF

fibroblast growth factor

SAR

structure–activity relationship

ADME

absorption, distribution, metabolism, and excretion

PK

pharmacokinetics

PD

pharmacodynamics

p.o.

oral administration

i.v.

intravenous

h-PB

in vitro protein binding of compound in human plasma

h-IntCL

intrinsic clearance in human liver microsomes

P_m

permeability

C_max

maximum concentration

AUC

area under the curve

T_1/2

half-life

V_ss

volume of distribution at steady state

CL

systemic clearance

F

oral bioavailability

CYP

cytochrome P450

hERG

human ether-a-go-go related gene

ClogP

calculated log P

WB

whole blood

FaSSIF

fasted-state simulated intestinal fluid

SGF

simulated gastric fluid

KRD

kinase insert domain receptor

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsmedchemlett.3c00003.

• Experimental details for in vitro and in vivo studies; synthetic procedures and characterization of all compounds; kinase profiling and PK data of compound 19 (PDF)

Author Contributions

The manuscript was prepared by M.H.N. with contributions from all authors.

The authors declare no competing financial interest.