Discovery of a covalent FGFR2-selective inhibitor overcoming clinically-acquired resistance mutations

Abstract

Abnormal activation of fibroblast growth factor receptor 2 (FGFR2) drives tumorigenesis in various cancers. Clinical use of pan-FGFR inhibitors is limited due to emerging acquired resistance mutations within the FGFR2 kinase domain and adverse effects associated with FGFR1/4 off-target inhibition. Herein, we describe the structure-based discovery of LC-F2-1, an FGFR2-selective inhibitor that demonstrates irreversible covalent binding to the P-loop. Cellular assays confirm the high selectivity of LC-F2-1 for FGFR2 over FGFR1 and FGFR4, along with potent inhibition of FGFR2 signaling. LC-F2-1 maintains strong activity against clinically observed FGFR2 resistance variants, including gatekeeper, molecular brake, and activation loop variants. X-ray crystallography reveals conformational rearrangement of the kinase domain by LC-F2-1, which overcomes the recalcitrant V565F gatekeeper mutation. In vivo, LC-F2-1 induces tumor regression in xenograft models harboring FGFR2 resistance mutations without affecting serum phosphate levels. In this work, we identify LC-F2-1 as a therapeutic candidate for FGFR2-driven cancers.

This study reports the structure-based discovery of LC-F2-1, a selective FGFR2 inhibitor that overcomes clinically observed resistance mutations in cancers. Its co-crystal structure with FGFR2V565F reveals how it may restore drug sensitivity.

Introduction

Fibroblast growth factor receptors 1 to 4 (FGFR1–4) constitute a family of single-pass transmembrane receptor tyrosine kinases. They are characterized by three conserved domains: an extracellular ligand-binding region (D1–D3), a short transmembrane segment, and an intracellular tyrosine kinase catalytic domain (Fig. 1a)^1–3. FGFRs regulate diverse biological functions, such as cellular proliferation, differentiation, and maintaining metabolic homeostasis through paracrine/endocrine FGF ligand recognition⁴. Furthermore, FGF-induced receptor dimerization triggers sequential trans-autophosphorylation of activation loop (A-loop) tyrosine residues within the kinase domain^5,6. This facilitates the recruitment of two principal adaptor proteins, FGFR substrate 2α (FRS2α) and phospholipase C gamma 1 (PLCγ), initiating the activation of three major downstream signaling cascades^2,7–10.

Fig. 1. Discovery of LC-F2-1 overcomes FGFR2 V565F gatekeeper mutation.

a Architecture of FGFR domains. b Chemical structure of pan-FGFR inhibitors approved by the Food and Drug Administration. The shared dimethoxyphenyl ring is highlighted. c Analysis of co-crystal structures of futibatinib (shown in yellow sticks) bound to WT FGFR2 (PDB ID: 8W2X). The V565 gatekeeper residue is shown as sticks superimposed on a surface representation. d Structure-based drug design of FGFR2 inhibitors to overcome gatekeeper mutation. e Optimization process of FGFR2-selective inhibitors. f Biochemical in vitro inhibitory activities of synthesized target compounds against FGFR2^WT and the FGFR2^V565F mutant. g Biochemical activities of LC-F2-1 on FGFR1^WT, FGFR2^WT, FGFR2^V565F, and FGFR4^WT. Data in (f, g) are shown as mean values from two independent experiments (n = 2). FGFR fibroblast growth factor receptor, WT wild-type, PDB Protein Data Bank.

Oncogenic activation of FGFR2 occurs via three principal mechanisms: gene amplification, activating mutations, and chromosomal fusion events^2,8–10. Aberrant FGFR2 signaling has been clinically associated with multiple malignancies, including 10–16% cases of intrahepatic cholangiocarcinoma (iCCA) and 3.7–7.9% of gastric cancer^10,11. Among FGFR2 alterations, chromosomal rearrangements, such as FGFR2-BICC1 fusion, have emerged as the primary mechanism, driving constitutive kinase activation through ligand-independent dimerization¹⁰. These findings indicate that FGFR2-targeted therapy offers notable clinical benefits to a broad spectrum of patients harboring FGFR2 alterations¹².

FGFR2-targeted therapy has demonstrated efficacy in clinical trials evaluating pan-FGFR inhibitors, such as pemigatinib and futibatinib, for patients with iCCA^13,14. Despite these advances, FGFR1/4-related toxicity remains a considerable challenge. The primary toxic effect is hyperphosphatemia, caused by the unintended inhibition of FGF23-FGFR1c-mediated endocrine signaling. Hyperphosphatemia is characterized by elevated serum phosphate levels, documented in approximately 55–90% of patients with iCCA enrolled in phase II clinical trials^13,14. This dose-limiting adverse event necessitates proactive management strategies, including phosphate binder administration and therapeutic dose modifications¹⁵. In addition, off-target inhibition of FGFR4 has been implicated in gastrointestinal toxicities, manifesting as treatment-related diarrhea (15–35% incidence) in clinical cohorts¹⁶.

On-target resistance is another major factor that limits the long-term efficacy of FGFR inhibitors^17–20. A clinical study previously demonstrated the high frequency of secondary FGFR2 mutations in over 60% of patients who initially benefited from therapy²¹. Similarly, a comprehensive analysis of resistance mechanisms in patients with cholangiocarcinoma progressing on FGFR inhibitors has revealed detectable secondary mutations in 60% (49/82) cases²². Notably, the FGFR2 V565 gatekeeper mutations (V565F/L/I) and molecular brake region alterations (N550K, E566A, and K642N) accounted for 47% and 63% of these resistance mutations, respectively²². These landscape analyses establish secondary FGFR2 kinase domain alterations as the primary drivers of drug resistance. Lirafugratinib (RLY-4008), an FGFR2-selective inhibitor evaluated in phase II clinical trials, shows clinical activity in FGFR inhibitor-naïve cholangiocarcinoma and has an isoform selectivity profile aimed at limiting FGFR1/4-associated toxicities^23–25. However, reported clinical data indicate that acquired resistance to lirafugratinib is associated with secondary FGFR2 mutations at M538 and/or L618 in 69% (11 of 16) of analysed cases^26,27, whereas resistance to futibatinib is mainly linked to mutations at FGFR2 gatekeeper (V565) and molecular brake (N550) residues¹².

In this work, we describe LC-F2-1, a covalent FGFR2-selective inhibitor designed to address resistance-associated FGFR2 kinase-domain mutations. LC-F2-1 forms an irreversible covalent bond with the P-loop residue Cys492 and retains activity against multiple clinically relevant FGFR2 resistance mutations. Structural analyses define its binding mode and provide a basis for its activity against resistance-associated FGFR2 variants. Together, these results support Cys492-directed covalent inhibition as an approach for addressing heterogeneous FGFR2 resistance mechanisms.

Results

Discovery of LC-F2-1 for FGFR2 V565F mutation

All FGFR inhibitors approved by the FDA occupy the kinase back pocket region through a functionalized dimethoxyphenyl ring to optimize binding specificity and potency (Fig. 1b, highlighted in red). While clinically effective, these inhibitors are inherently predisposed to resistance mutations. The V565F substitutions were the most frequent mutations across all patient samples, significantly lowering the affinity for futibatinib and pemigatinib. Co-crystal structures of FGFR2-futibatinib indicated that phenylalanine substitution at V565 may cause steric hindrance, which disrupts drug-receptor binding (Fig. 1c). Recently, studies revealed that FGFR2 gatekeeper mutations further increase in the population of the active state kinase²⁸.

To discover compounds with broad-spectrum inhibitory activity across clinically relevant FGFR2 alterations, we developed a structure-based drug design strategy. This strategy optimizes key binding interactions within the hydrophobic back pocket of FGFR2 based on the binding mode of futibatinib (Fig. 1d). Accordingly, we replaced the pyrazolo[3,4-d]pyrimidine scaffold of futibatinib with a pyrrolo[2,3-d]pyrimidine core. Furthermore, the pyrrolidine ring linker connecting the acrylamide warhead of futibatinib was structurally modified to a phenyl linker. This enabled covalent targeting of the conserved Cys492 residue in the FGFR2 phosphate-binding loop (P-loop; a cysteine residue conserved across all FGFR isoforms). While Cys492 is conserved across all four FGFR isoforms, previous long-timescale molecular dynamics simulations by Schönherr et al. have identified distinct, isoform-specific conformational dynamics in the P-loop of FGFR2 compared to FGFR1²⁴. Consequently, we hypothesized that strategic modification of the warhead linker could potentially enhance FGFR2 isoform selectivity.

Guided by design principles, we prioritized structural diversification by replacing the conserved dimethoxyphenyl moiety of futibatinib with an indole bioisostere, which yielded the lead compound F1 (Fig. 1e). Notably, F1 inhibited FGFR2^WT (IC₅₀ = 61.1 nM) while retaining potent activity against the FGFR2^V565F mutant (IC₅₀ = 61.6 nM). These biochemical data reveal that the indole pharmacophore has superior steric compatibility with the FGFR2^V565F gatekeeper mutation compared to the dimethoxyphenyl moiety (Fig. 1e and Supplementary Table 1). In contrast, futibatinib and pemigatinib exhibited significantly reduced efficacy against the FGFR2^V565F mutant compared to that against FGFR2^WT, with a 67-fold (IC₅₀ = 114.2 nM) and 1192-fold (IC₅₀ = 5712 nM) decrease, respectively (Fig. 1f and Supplementary Table 1).

Building upon these findings, we conducted a systematic structure-activity relationship analysis by modifying the back pocket binding ring systems of F1, yielding compounds F2–F10 (Fig. 1e). As expected, variations in the back pocket site pharmacophore significantly affected FGFR2^V565F binding affinity. Notably, compound F9 (LC-F2-1), featuring a methylimidazo [4,5-b]pyridine pharmacophore, exhibited a significant boost in inhibitory potency. Compared to the parent F1 compound, F9 showed 6.1-fold (IC₅₀ = 10.0 nM) and 10.6-fold (IC₅₀ = 5.8 nM) activity against both the WT and V565F gatekeeper mutant, respectively (Fig. 1f and Supplementary Table 1). Further, we carried out medicinal chemistry campaigns including N-methyl modifications targeting the kinase solvent region (compounds F11–F15) and exploration of electrophile moieties targeting the kinase P-loop (compound F16), but these changes did not enhance activity (Fig. 1e, f). F9 (LC-F2-1) emerged as a highly potent FGFR2^V565F inhibitor, with 21-fold selectivity for FGFR1 and >1000-fold for FGFR4 in biochemical assays (Fig. 1g). Therefore, LC-F2-1 was selected as the optimal compound for subsequent pharmacological evaluations.

LC-F2-1 selectively inhibits FGFR2 isoform

Given the high conservation of catalytic domains across kinase families (Fig. 2a), the development of isoform-selective kinase inhibitors remains challenging. To comprehensively establish its selectivity profile, we screened LC-F2-1 against an extensive panel of 416 human kinases at a concentration of 1000 nM (Fig. 2b). The profiling results confirmed that LC-F2-1 achieves complete inhibition of FGFR2—including the wild-type (WT) and the N550H, R612T, K526E, and K660N mutants. Additionally, LC-F2-1 exhibited robust inhibitory activity (>90% inhibition) against several other FGFR2 mutants, including E566G, K642R, V565I, and V565F. Importantly, regarding off-target effects, the compound showed >80% inhibition against only five non-FGFR targets: RET, TNK1, EPHA6, PDGFRα, and KDR (Fig. 2b and Supplementary Fig. 2).

Fig. 2. LC-F2-1 selectively inhibits FGFR2 isoform activity.

a Sequence alignment of the kinase regions of FGFR1–4. Sites of drug-resistant mutations are indicated using red circles. C492 in the P-loop is marked using a blue triangle (note: the alignment spans the β1 through the A-loop, representing a partial kinase domain sequence). b Kinase selectivity screening of LC-F2-1 (1000 nM) against 416 kinases. Cell proliferation assay of LC-F2-1, pemigatinib, futibatinib and lirafugratinib in Ba/F3 cells expressing FGFR1 (c), FGFR2 (d), FGFR3 (e), and FGFR4 (f). Data in (b–f) are shown as mean values from two independent experiments (n = 2). FGFR fibroblast growth factor receptor, P-loop phosphate-binding loop, A-loop activation loop.

Next, we employed four Ba/F3 cell models expressing FGFR1–4 to assess the isoform specificity of LC-F2-1 within the FGFR family. Cellular sensitivity analysis (Fig. 2c–f and Supplementary Table 2) confirmed the strong selectivity for FGFR2, where LC-F2-1 demonstrated potent inhibition with an IC₅₀ of 4.0 nM. In contrast, FGFR1- and FGFR3-expressing cells exhibited markedly reduced responsiveness, with 44-fold (IC₅₀ = 175.9 nM) and 23-fold (IC₅₀ = 93.4 nM) lower selectivity, respectively. Notably, no anti-proliferative effect of LC-F2-1 was observed in FGFR4-expressing cells. This selectivity profile contrasts with that of the pan-FGFR inhibitors futibatinib and pemigatinib, which showed comparable activity across all FGFR isoforms. The superior isoform selectivity of LC-F2-1 may effectively mitigate hyperphosphatemia induced by FGFR1 inhibition, while circumventing FGFR4 suppression-related side effects.

LC-F2-1 potently inhibits FGFR2 downstream signaling

To demonstrate the effects of LC-F2-1 on FGFR2-mediated signaling, we initially used the gastric cancer cell line KATOIII with distinct FGFR2 alterations²⁹. The inhibitory effect on FGFR2 signaling was evaluated by assessing the phosphorylation status of Tyr653/654 residues on FGFR2 and Thr202/Tyr204 residues on its downstream effector ERK1/2 after 8 h of drug exposure (Fig. 3a). Consistent with the observed biochemical and cell-based antiproliferation potency, LC-F2-1 inhibited pFGFR and pERK in KATOIII with a potency of 4.7 and 12.0 nM, respectively. Next, we performed a pharmacological validation to determine the effects of LC-F2-1 on the FGFR2-mediated downstream cascade. In addition to the KATOIII model, we used the SNU-16 cell line—an FGFR2-dependent model characterized by FGFR2 amplification and low-frequency fusion variants, for comprehensive pathway analysis. Clinically approved pan-FGFR inhibitors, pemigatinib and futibatinib, were used as reference compounds in parallel assays.

Fig. 3. LC-F2-1 selectively inhibits proliferation of FGFR2-driven cancer.

a Inhibition of the phosphorylation of FGFR2 (at Y653/654) and ERK (at T202/Y204) in KATOIII cells. Cells were incubated with LC-F2-1 for 8 h prior to lysis and subsequent Western blotting. Data are shown as mean values from two independent experiments. b Representative immunoblots are shown. The experiment was independently repeated two times with similar results. Evaluation of LC-F2-1 inhibition potency against FGFR2-mediated signaling in KATOIII (left) and SNU-16 (right) cells. Immunoblots of cell lysates derived from treatment with DMSO, LC-F2-1, pemigatinib, or futibatinib for 8 h at concentrations of 20 and 100 nM. c Cell proliferation assay evaluating LC-F2-1, pemigatinib, and futibatinib activity against FGFR2-dependent KATOIII cells. Data are shown as mean ± SEM from three biologically independent experiments. d Immunoblotting analysis post-24 h LC-F2-1 treatment. e Quantification of LDH release—a biochemical marker of necroptotic cell death-following 24-h LC-F2-1 treatment of KATOIII cells. Data are shown as mean ± SEM from three biologically independent experiments. f Schematic showing the components participants in FGFR2 inhibition-induced necroptosis upon LC-F2-1 treatment. FGFR fibroblast growth factor receptor, ERK extracellular signal-regulated kinase, DMSO dimethyl sulfoxide, LDH lactate dehydrogenase.

In both KATOIII and SNU-16 cellular models, LC-F2-1 significantly attenuated the phosphorylation of critical FGFR2 signaling components following an 8-h treatment with inhibitors at concentrations of 20 and 100 nM (Fig. 3b). At a concentration of 20 nM, LC-F2-1 almost completely inhibited the phosphorylation of FRS2α (Tyr436), PLCγ (Tyr783), AKT (Ser473), SHP2 (Tyr524), ERK, and RSK1P90 (Thr573). To systematically assess LC-F2-1 efficacy in the inhibition of cellular FGFR2, we evaluated its anti-proliferative activity in FGFR2-dependent KATOIII cells. LC-F2-1 demonstrated robust growth inhibition with an IC₅₀ of 9.0 nM (Fig. 3c), exhibiting potency comparable to that of clinically approved FGFR-targeted agents. Collectively, these findings confirm that LC-F2-1 exerts anticancer effects by targeting FGFR2 kinase activity and disrupting downstream signaling cascades, ultimately suppressing the proliferation of FGFR2-driven malignancies.

Previous studies demonstrated that cell death following FGFR blockage is caspase-independent, and pan-FGFR inhibitors can induce necrosis in cholangiocarcinoma cells³⁰. A more recent study revealed that the pan-FGFR inhibitor AZD4547 exerts potent inhibitory effects on esophageal squamous cell carcinoma by triggering necroptosis via FGFR2 inhibition. These findings suggest a potential link between FGFR2 signaling and necroptosis³¹. To further investigate whether the FGFR2-selective inhibitor LC-F2-1 induces necroptosis in FGFR2-driven tumors, we conducted experiments in SNU-16 cells. Results showed that LC-F2-1 treatment increased protein levels of key necroptotic mediators (RIP1, RIP3, p-MLKL, and MLKL) in a dose-dependent manner (Fig. 3d). A concomitant dose-dependent increase in lactate dehydrogenase (LDH) release further indicated plasma membrane rupture (Fig. 3e). These observations together suggest that LC-F2-1 induces necroptosis in gastric cancer cells (Fig. 3f).

LC-F2-1 retains activity against resistant FGFR2 mutants

The development of resistance mutations within the FGFR2 kinase domain compromises the therapeutic efficacy of the existing pan-FGFR inhibitors. The three most frequently detected FGFR2 mutations in clinical samples were V565 gatekeeper, L618 DFG latch, and N550 mutations, which were observed in a large percentage of patients with acquired resistance (Fig. 4a, data from Goyal et al.²¹). The N550 mutations disrupt a network of hydrogen bonds that typically relieves the “molecular brake” and stabilizes the active conformation of the kinase (Fig. 4b). The L618V mutation impairs stabilizing interactions with the DFG motif. Mutations in these codons produce various substitutions (e.g., V565F/L/I) that exhibit differential prevalence patterns between reversible and irreversible inhibitors. Notably, FGFR2^L618V mutations were detected following treatment with the reversible inhibitor pemigatinib, but not with the irreversible inhibitor futibatinib (Fig. 4a).

Fig. 4. LC-F2-1 retains efficacy against clinically relevant drug-resistant mutations.

a Frequency of FGFR2 acquired resistance mutations detected at progression on reversible and irreversible pan-FGFR inhibitor. b Structural representation depicting the location of the mutation within the FGFR2 kinase domain (PDB ID: 2PVF). Residue numbering corresponds to the FGFR2b isoform (NM_022970.3). c Heat map showing the IC₅₀ fold change for LC-F2-1, pemigatinib, futibatinib, and lirafugratinib against 12 clinically relevant FGFR2 mutants, relative to wild-type FGFR2. Following 72 h of treatment with the indicated compound, IC₅₀ values against Ba/F3 cells expressing the TEL-FGFR2 fusions with a wild-type kinase domain or possessing different mutations were determined. Fold-change value of 1.0 indicates equivalent potency against wild-type and mutant FGFR2. Cell proliferation assay evaluating the activity of FGFR inhibitors against Ba/F3 cells expressing FGFR2^V565F(d) or FGFR2^M538I(e). Dose–response curves represent two biological replicates. f Representative immunoblots are shown. The experiment was independently repeated two times with similar results. Immunoblot of Ba/F3 cells with the indicated FGFR2 alleles following 5 h of treatment with the indicated inhibitors at 50 nM (left) and 100 nM (right). FGFR fibroblast growth factor receptor, PDB Protein Data Bank, IC₅₀ half-maximal inhibitory concentration, TEL-FGFR2 translocated ETS leukemia-FGFR2 fusion.

To evaluate the effect of LC-F2-1 on these drug resistance mutations, we determined the antiproliferative activity of LC-F2-1 in 12 Ba/F3 cells harboring common FGFR2 mutations or fusions. The experimental design incorporated three comparator agents, including pemigatinib, futibatinib, and lirafugratinib. Fold changes in potency (IC₅₀) against the specified Ba/F3 FGFR2 mutants compared with cells expressing FGFR2^WT were calculated (Fig. 4c and Supplementary Tables 3, 4). As expected, pemigatinib showed significantly reduced potency against gatekeeper mutations, resulting in IC₅₀ values of 938.5 nM for FGFR2^V565F (426.5-fold increase) and 152.7 nM for FGFR2^V565L (69.4-fold increase). Futibatinib demonstrated a 44.7-fold shift in potency against FGFR2^V565F relative to that against Ba/F3 cells expressing FGFR2^WT (Fig. 4d). This is consistent with previous findings that the replacement of phenylalanine at V565 induces steric hindrance in drug binding. Futibatinib retained its inhibitory efficacy against FGFR2 V565L/I mutants, presumably because of less severe hindrance compared to that with the V565F variant. Notably, lirafugratinib demonstrated stronger antiproliferative activity against FGFR2^V565F/L mutants than against FGFR2^WT, while showing a moderate decrease in potency against the FGFR2^V565I variant (Fig. 4c and Supplementary Tables 3, 4). In contrast, LC-F2-1 was more potent against FGFR2^V565F/L/I than against FGFR2^WT, with IC₅₀ ranging 2.1–3.0 nM (Fig. 4c, d). These results indicate that replacing the dimethoxyphenyl ring allows for better accommodation of gatekeeper resistance mutations. Indeed, the recently reported compound KIN-3248 maintained robust activity against FGFR gatekeeper mutations following the optimization of binding interactions within the back pocket³².

For molecular brake mutations, LC-F2-1 demonstrated a 9.4-fold potency shift against FGFR2^N550K-expressing Ba/F3 cells (IC₅₀ = 37.6 nM) compared to that against FGFR2^WT. This was markedly lower than the 141.1- and 19.2-fold shifts observed with pemigatinib (IC₅₀ = 310.5 nM) and futibatinib (IC₅₀ = 7.7 nM), respectively. For the E566A mutation, LC-F2-1 showed a 14.0-fold reduction in potency compared with that against FGFR2^WT, whereas pemigatinib exhibited a more substantial decrease of 32.9-fold. Pemigatinib lost its potency against the DFG latch mutation L618V, with a 22.7-fold increase in IC₅₀ compared to that against FGFR2^WT. Notably, the irreversible inhibitors LC-F2-1, futibatinib, and lirafugratinib maintained their efficacy, exhibiting only ~2-fold IC₅₀ differences between FGFR2^WT and FGFR2^L618V-expressing Ba/F3 cells. These mechanistic differences in mutations aligned with the clinical observations showing a differential prevalence of L618V resistance mutation between reversible and irreversible FGFR inhibitors (Fig. 4a).

Interestingly, a relatively rare M538I mutation located in the hydrophobic regulatory spine (R-spine) of the αC-helix compromised the activity of lirafugratinib (137.9 nM, 81.1-fold IC₅₀ increase, Fig. 4e). This diminished efficacy is likely due to the M538I mutation disrupting interactions between the 4-methylpyrimidine pharmacophore of lirafugratinib and M538 residing on the αC-helix of the FGFR2 kinase domain. The other three compounds, LC-F2-1, pemigatinib, and futibatinib, largely retained their activity against both FGFR2^M538I and FGFR2^M539I (Fig. 4c). Importantly, while this manuscript was under review, two pivotal clinical studies revealed that patients with acquired resistance to lirafugratinib manifest an unprecedented emergence of FGFR2 mutations (M538I and/or L618F) within the kinase domain^26,27. Corroborating these clinical insights, our subsequent validation assays demonstrated that the activity of lirafugratinib against the FGFR2-L618F mutation is drastically diminished (a 65.3-fold increase in IC₅₀).

We further investigated the effects of the compounds on the FGFR signaling pathway in engineered Ba/F3 cells at 50 or 100 nM concentrations (Fig. 4f). Western blot analysis revealed that, compared to V565F and M538I mutations, both FGFR2 N550K and L618V mutations activated the constitutive and robust FGFR2 signaling with elevated pFRSα and pSHP2 levels. Immunoblotting results revealed distinct efficacy profiles of the three agents (LC-F2-1, pemigatinib, and lirafugratinib) against four key FGFR2 mutations. Pemigatinib showed markedly reduced efficacy against V565F and L618V mutations, whereas LC-F2-1 and lirafugratinib were effective against V565F gatekeeper and L618V DFG latch mutations (Fig. 4f). LC-F2-1 and lirafugratinib remained at least partially active against FGFR2^N550K mutation-mediated signal activation at 50 nM and completely blocked the FGFR2^N550K pathway at 100 nM. Both LC-F2-1 and pemigatinib completely suppressed the downstream signaling of FGFR2^M538I, whereas lirafugratinib exhibited a significant loss of sensitivity to the M538I and L618F mutations. Overall, these results identify LC-F2-1 with activity against most clinically observed FGFR2 resistance variants.

Covalent binding of LC-F2-1 with FGFR2^V565F

Cell-based experimental data show that the covalent inhibitors futibatinib and lirafugratinib overcome drug resistance more effectively than reversible inhibitors (Fig. 4). We subsequently validated whether LC-F2-1 can covalently bind to the FGFR2 V565F mutant. The recombinant kinase domain of FGFR2K^V565F (molecular weight: 35,683.27 Da) was incubated with LC-F2-1 (molecular weight: 448.49 Da) or futibatinib (molecular weight: 418.45 Da) at molar ratios of 2:1 or 5:1 (ligand:protein). Following 4 h of incubation at room temperature, intact mass analysis revealed a mass shift of 448.64 Da, consistent with the formation of a 1:1 covalent LC-F2-1:FGFR2^V565F complex (Fig. 5a, b). Complete protein labeling was achieved under both stoichiometric conditions, with no detectable multiple adduct formation, even at a 5:1 ligand excess (Supplementary Fig. 3). In addition, a cellular washout experiment was conducted to verify the covalent target engagement between LC-F2-1 and FGFR2 in KATO III cells (Fig. 5c). LC-F2-1 treatment resulted in sustained inhibition of p-FGFR2 and p-ERK, which persisted for up to 12 h after washout. After 24 h, the LC-F2-1-treated group maintained a potent inhibitory effect on the phosphorylated proteins. In contrast, the reversible inhibitor, pemigatinib, recovered the original p-FGFR2 levels after 3–6 h of washout. This demonstrates the irreversible binding of LC-F2-1 to cellular FGFR2. Finally, in vitro enzymatic kinetic studies were performed to determine the efficiency of covalent modification (Fig. 5d). The k_inact/K_I ratio for the modification of FGFR2 by LC-F2-1 was calculated to be 1.57 × 10⁴ (M⁻¹ × s⁻¹), indicating that LC-F2-1 rapidly and covalently inactivates the FGFR2 kinase (Fig. 5d).

Fig. 5. Covalent binding of LC-F2-1 with FGFR2^V565F.

a Chemical structure and molecular weight of LC-F2-1. b Deconvoluted intact mass spectra of the FGFR2^V565F kinase domain (left), LC-F2-1-FGFR2^V565F mixture (middle), and futibatinib-FGFR2^V565F mixture (right) at a molar ratio of 2:1. The spectra shown were obtained with 28 µM of kinase. Mass spectrometry analysis revealed a +448.6 Da or +418.1 Da mass shift, indicating stoichiometric binding of one LC-F2-1 or futibatinib molecule. This schematic was created with BioRender. Chen, L. (2026). (www.biorender.com/lq0eri6). c Representative immunoblots are shown. The experiment was independently repeated two times with similar results. This schematic was created with BioRender. Chen, L. (2026). (www.biorender.com/k2rkl7k). KatoIII cells were treated with 100 nM of LC-F2-1 (covalent) or pemigatinib (reversible) for 2 h and then washed out. Both pFGFR and pERK levels were measured at the indicated time points. d LC-F2-1 concentration-dependent modification rate against FGFR2. e Cell proliferation assay evaluating the activity of FGFR inhibitors against Ba/F3 cells expressing FGFR2^C492F. Data in (d, e) are shown as mean values from two independent experiments (n = 2). FGFR fibroblast growth factor receptor, ERK extracellular signal-regulated kinase.

Collectively, these data confirm that covalent attachment to the C492 residue in the FGFR2 P-loop drives the efficacy of LC-F2-1. Indeed, in our 416-kinase panel screening, although LC-F2-1 maintained robust activity against most variants, its inhibitory potency was notably reduced against the FGFR2 C492S and C492F mutants (Fig. 2b and related Source data). Given that the C492F mutation, while rare, has been clinically detected and confers potent resistance to covalent agents²¹, we engineered Ba/F3-FGFR2-C492F cells for rigorous validation (Fig. 5e). As expected, all tested covalent inhibitors (futibatinib, lirafugratinib, and LC-F2-1) completely lost efficacy against this mutant (IC₅₀ > 800 nM). Interestingly, the reversible inhibitor pemigatinib retained excellent potency (IC₅₀ = 3.4 nM). These findings establish the C492F mutation as an acquired resistance vulnerability specifically for FGFR-targeted covalent inhibitors.

Crystal structure of the LC-F2-1 and FGFR2 V565F complex

Next, we determined the binding mode of LC-F2-1 to FGFR2 and the mechanism by which it overcomes drug resistance mutations. We successfully co-crystallized LC-F2-1 with FGFR2 carrying the V565F mutation (PDB ID: 9U3N). The crystal structure showed that the pyrrolo-pyrimidine core of LC-F2-1 is anchored to the ATP pocket through hydrogen bonds with the E566 and A568 residues (Fig. 6a). The acrylamide warhead was irreversibly captured by Cys492 from the P-loop, forming hydrogen bonds with the backbone of the P-loop E490 residue. Notably, the imidazo[4,5-b]pyridine moiety of LC-F2-1 deeply inserts into the back pocket of the FGFR2 kinase domain (Fig. 6b).

Fig. 6. Crystal structure of the LC-F2-1 and FGFR2 V565F complex.

a Co-crystal structure of LC-F2-1 (green) bound to FGFR2^V565F (blue). b Close-up view of LC-F2-1 in the ATP-binding pocket of FGFR2^V565F. The V565F gatekeeper residue is depicted in magenta. c Structural superimposition of futibatinib (yellow)-FGFR2^WT and LC-F2-1(green)-FGFR2^V565F. Note that the gatekeeper substitution (V565F) induces significant steric hindrance, creating spatial conflict with the dimethoxy phenyl moiety of futibatinib. d Close-up view highlights the differential spatial orientations of the Cys492 residue in the P-loop induced by LC-F2-1 versus futibatinib. e Superposition of the ATP analog-bound FGFR2 structure (PDB ID: 2PVF, gray) and the LC-F2-1-FGFR2^V565F complex (blue) reveals N-lobe structural rearrangements in the αC-helix and P-loop. FGFR fibroblast growth factor receptor, PDB Protein Data Bank, P-loop phosphate-binding loop, N-lobe N-terminal lobe.

Phenylalanine substitution at position V565 in the ATP-binding pocket of FGFR2 is known to sterically reduce the binding affinity of several approved pan-FGFR inhibitors. Therefore, we investigated the molecular mechanisms by which LC-F2-1 circumvents gatekeeper mutations. The LC-F2-1-bound structure was superimposed on a futibatinib-bound FGFR2 kinase domain structure (PDB ID: 8W2X²¹) for comparison. The heterocyclic fragment of LC-F2-1 optimally accommodated steric hindrance within the binding pocket and established additional hydrophobic interactions with the F565 residue. In contrast, the dimethoxyphenyl ring of futibatinib clashed sterically with F565 (Fig. 6c). Although both LC-F2-1 and futibatinib share overlapping binding sites (e.g., E566 and A568) in the FGFR2 hinge region, the core scaffold of LC-F2-1 adopts a distinct binding conformation characterized by a ~20° rotational shift relative to futibatinib. This conformational adaptation optimizes its steric complementarity with the mutated binding pocket, effectively mitigating the steric constraints imposed by the V565F mutation. These structural insights explain the enhanced inhibitory potency of LC-F2-1 against the FGFR2^V565F mutant compared with that against FGFR2^WT (Fig. 6c).

Additionally, the acrylamide moiety of LC-F2-1 induced a torsional strain on Cys492, shifting the P-loop into a downward-oriented conformational state associated with functional inhibition (Fig. 6d). Structural analysis revealed that the different warhead linker lengths between LC-F2-1 and futibatinib induce distinct conformations of Cys492 in the P-loop. Specifically, in the LC-F2-1-bound FGFR2 complex, Cys492 adopts an outward orientation relative to the ATP-binding pocket. However, in the futibatinib-FGFR2 complex, the sulfhydryl group of Cys492 undergoes an inward rotation to form a covalent adduct with futibatinib (Fig. 6d). Previous molecular dynamics simulations conducted by Schönherr et al. revealed distinct isoform-specific dynamics and conformational variations in the kinase P-loop between FGFR2 and FGFR1²⁴. These findings suggest that the isoform selectivity of LC-F2-1 may originate from an unfavored P-loop conformation, preventing the covalent modification of Cys488 within the FGFR1 P-loop.

LC-F2-1 binding was also found to induce conformational changes in the FGFR2 kinase structure. To analyze these changes, we superimposed the LC-F2-1-bound FGFR2 kinase structure on an ATP analog-bound FGFR2 kinase domain structure (PDB ID: 2PVF)³³. Notably, LC-F2-1 triggered distinct conformational changes. As illustrated in Fig. 6e, the N-lobe of LC-F2-1-bound FGFR2 adopted a distinct rotated conformation. Comparative structural analysis revealed that the N-lobe of the LC-F2-1-bound kinase was rotated by approximately 18° relative to the ATP analog-bound structure. These structural changes induced positional shifts in the hydrophobic regulatory spine of the αC-helix, better accommodating LC-F2-1 within the ATP-binding pocket.

LC-F2-1 overcomes FGFR2 V565F/M538I mutations in vivo

Collective biochemical, biophysical, and cell-based evidence suggests that LC-F2-1 effectively overcomes acquired resistance to existing inhibitors, providing a rationale for in vivo studies. However, prior to conducting animal experiments, we tested the effects of the compound on hERG and its pharmacokinetic properties. LC-F2-1 showed a low inhibitory effect on the hERG potassium channel current (IC₅₀ > 40 μM, Supplementary Fig. 1) and demonstrated acceptable in vitro stability and oral pharmacokinetic profiles (Supplementary Tables 5–9). To evaluate the in vivo antitumor efficacy of LC-F2-1, we established a xenograft model using FGFR2^V565F-Ba/F3 cells (Fig. 7a). When the tumors reached approximately 100 mm³ in volume, the mice (n = 6 per group) were randomized into four treatment groups: (1) vehicle control, (2) LC-F2-1 (10 mg/kg), (3) pemigatinib (1 mg/kg), and (4) futibatinib (5 mg/kg). Treatment was administered via daily oral gavage, and the dosages for pemigatinib and futibatinib were determined based on clinical equivalents. After 14 days of treatment, LC-F2-1 demonstrated superior tumor growth suppression, with 87% tumor growth inhibition (TGI), significantly outperforming both futibatinib (32% TGI) and pemigatinib (22% TGI) at equivalent doses (Fig. 7b). Notably, the compound exhibited excellent tolerability, with stable body weight maintenance comparable to that of vehicle-treated controls throughout the study period (Fig. 7c).

Fig. 7. LC-F2-1 overcomes FGFR2 V565F- and M538I-mediated resistance in Ba/F3 xenograft models.

a Mice carrying Ba/F3-TEL-FGFR2^V565F were treated with vehicle, pemigatinib (1 mg/kg), futibatinib (5 mg/kg), or LC-F2-1 (10 mg/kg) daily for 14 days. Tumor volumes were recorded over time, and treatment was initiated once the tumors reached approximately 100 mm³. b Tumor weight. Ordinary one-way ANOVA with Dunnett’s multiple-comparison test was used. Note that LC-F2-1 reduces tumor weight relative to vehicle-treated mice (adjusted P < 0.0001), compared with the weaker effects of pemigatinib and futibatinib (adjusted P = 0.2203 and 0.0361). c Body weight change. d Terminal serum phosphate levels were determined. For a–d, n = 5 biologically independent mice for the vehicle group and n = 6 biologically independent mice for each treatment group. e Mice carrying Ba/F3-TEL- FGFR2^M538I were treated with vehicle, lirafugratinib, or LC-F2-1 (10 mg/kg) daily for 11 days. Tumor volumes were recorded over time. f Tumor weight. Data are presented as mean ± SEM from six mice per group. Ordinary one-way ANOVA with Dunnett’s multiple-comparison test was used. Note that LC-F2-1 reduces tumor weight relative to vehicle-treated mice (adjusted P < 0.0001), whereas lirafugratinib has limited activity in this model (adjusted P = 0.1610). g Proposed sequential FGFR inhibitor therapies tailored to emergent FGFR2 kinase domain mutations driving acquired resistance. FGFR fibroblast growth factor receptor, TEL-FGFR2 translocated ETS leukemia-FGFR2 fusion.

In our previous studies, we elucidated that endocrine FGF23 binds to FGFR1c, FGFR3c, or FGFR4 isoforms to form quaternary complexes of FGF23-FGFR-Klotho-heparan sulfate, thereby regulating FGF23-mediated phosphate metabolic signaling⁶. However, administration of pan-FGFR inhibitors has been shown to elevate systemic phosphate levels, leading to hyperphosphatemia as an adverse effect^34,35. Consequently, we evaluated whether LC-F2-1 modulates serum phosphate levels in animal models. As shown in Fig. 7d, serum phosphate levels in the LC-F2-1 treatment group remained within the normal range at the end of the therapeutic period. This observation aligns with biochemical and cellular studies showing that LC-F2-1 selectively spares FGFR1 and FGFR4. In contrast, mice treated with futibatinib or pemigatinib showed significantly elevated serum phosphate levels.

Recent clinical studies have reported that patients who acquire resistance to lirafugratinib show an unexpected emergence of FGFR2 kinase-domain mutations, including M538I in the αC-helix and L618F in the DFG latch^26,27. We therefore further evaluated the activity of LC-F2-1 in a Ba/F3 FGFR2 M538I xenograft model. The in vivo data were consistent with our in vitro findings (Fig. 4c) and with recent clinical reports describing FGFR2 M538I as a resistance mechanism to lirafugratinib. In contrast, LC-F2-1 displayed pronounced antitumour activity in this model (Fig. 7e, f). Structurally, the M538I substitution is likely to impair lirafugratinib binding by disrupting van der Waals interactions between Met538 and the methylpyrimidine moiety of lirafugratinib²⁶. By contrast, LC-F2-1 is positioned away from residue M538 (Fig. 6b), which may render its binding mode less vulnerable to perturbation by the M538I mutation.

Discussion

While pan-FGFR inhibitors demonstrate clinical benefits in FGFR2-driven malignancies, their therapeutic potential is substantially constrained by two major challenges: (1) the inevitable emergence of on-target resistance mutations (observed in >60% of treated patients²²) and (2) off-target toxicities mediated by FGFR1 (hyperphosphatemia) and FGFR4 (diarrhea) signaling inhibition. The acquired FGFR2 resistance mutations predominantly involve the “molecular brake” residues and the V565 gatekeeper position. Clinical data have revealed that FGFR2 V565F/L/I mutations emerge with high frequency in patients progressing on both reversible ATP-competitive and covalent inhibitors, such as futibatinib^17,22,36. Recently, FGFR2- or FGFR2/3-selective inhibitors/degraders have emerged as major research topics^37–41. Lirafugratinib was the first selective FGFR2 tyrosine kinase inhibitor to enter clinical development. Its broad therapeutic potential as a highly selective FGFR2 inhibitor has been highlighted in clinical studies, along with its promising initial efficacy in FGFR2-altered solid tumors. Moreover, its differentiated safety profile has been shown to effectively prevent FGFR1/4-related toxicity. However, clinical trial data have also revealed the emergence of acquired resistance mutations. These included M538I and L618F mutations in patients^21,26,27. Thus, development of novel inhibitors to address diverse resistance mechanisms is necessary.

This study presents the structure-guided rational design of LC-F2-1, a covalent FGFR2 inhibitor developed to overcome the therapeutic limitations of current pan-FGFR inhibitors. The co-crystal structure of LC-F2-1 carrying FGFR2^V565F revealed its distinct binding mode. Unlike the futibatinib-FGFR2 complex, the imidazo[4,5-b]pyridine moiety of LC-F2-1 accommodates the V565F mutation by establishing hydrophobic interactions with F565 and rotating its core scaffold to avoid steric clashes. This structural adaptation enhances the steric complementarity in the gatekeeper-mutated ATP pocket. Comparative analysis showed that the longer warhead linker of LC-F2-1 forces Cys492 into an outward orientation. This is in contrast with the inward Cys492 rotation of futibatinib, which potentially explains the FGFR2 selectivity of LC-F2-1 over FGFR1. These combined features underpin the enhanced potency of LC-F2-1 against resistance mutations while maintaining kinase selectivity.

LC-F2-1 exhibited broad inhibitory activity against clinically emerging FGFR2 resistance mutations, including variants that confer resistance to pan-FGFR inhibitors, such as pemigatinib and futibatinib, as well as to the FGFR2-selective inhibitor lirafugratinib. Systematic biochemical profiling and cell-based functional assays showed that LC-F2-1 retained high potency against 12 clinically validated FGFR2 resistance variants. These variants involve three major structural regions of the kinase domain: gatekeeper mutations, including V565F/L/I; molecular brake alterations, including N550K, E566A and K642N; and DFG-latch mutations, including L618V/F. These findings indicate that LC-F2-1 has the potential to overcome a broad spectrum of acquired FGFR2 resistance mechanisms.

Notably, lirafugratinib and futibatinib displayed largely non-overlapping resistance profiles (Fig. 4c), indicating a strong degree of complementarity between these two irreversible inhibitors. Accordingly, sequential treatment strategies using inhibitors with distinct resistance liabilities may be an effective approach to extend disease control (Fig. 7g). FGFR2-targeted covalent inhibitors, including LC-F2-1, futibatinib and lirafugratinib, share a common liability: a complete loss of activity against the FGFR2 C492F mutation. After the emergence of C492F-mediated resistance to irreversible covalent inhibitors, treatment with a reversible FGFR inhibitor, such as pemigatinib, may retain activity and provide prolonged clinical benefit (Fig. 7g). These findings support a precision-guided sequential treatment paradigm in which inhibitor selection is informed by the molecular identity of acquired FGFR2 resistance mutations. Furthermore, while our study focuses on on-target resistance, recent clinical analyses of selective FGFR2 inhibitors have identified RTK-MAPK bypass alterations as a major driver of acquired resistance²⁷. It is highly probable that such off-target bypass pathway activation will present a similar long-term resistance liability for LC-F2-1, remaining a broader challenge for selective FGFR2 kinase inhibition. Our subsequent investigations will focus on evaluating its translational potential in FGFR2-driven malignancies that exhibit resistance to first-line pan-FGFR inhibitors.

Methods

Ethics statement

All animal experiments in this study complied with relevant ethical regulations for the care and use of laboratory animals. The animal study protocol was reviewed and approved by the Institutional Animal Care and Use Committee (IACUC) of Precedo Pharmaceuticals Co., Ltd (approval no. IACUC-20260301). Mice were housed in a specific pathogen-free facility under a 12 h light/12 h dark cycle at 20–24 °C and 40–70% relative humidity, with ad libitum access to food and water. The in vivo efficacy experiments used 6-week-old female BALB/c mice. Sex was considered in the study design; only female mice were used to maintain consistency across the Ba/F3-FGFR2 xenograft experiments. Because only one sex was used, sex-disaggregated analysis was not applicable.

Chemicals

The drugs used in this study were futibatinib (BD764789; Bidepharm), pemigatinib (HY-109099; MCE), and lirafugratinib (E1431; Selleck). The synthesis of the target compounds F1–F16 is described in the Supplementary data.

LC-MS analysis of synthesized compounds

All synthesized target compounds were analyzed by mass spectrometry for structural confirmation and purity assessment, with 16 final compounds included. Each sample was analyzed in a single technical replicate, and sodium formate calibration standards were included as quality controls for each analytical batch. Samples were dissolved in HPLC-grade methanol to a concentration of ~10 μg/mL and filtered through 0.22 μm PTFE syringe filters prior to injection.

LC-MS was performed on a Waters Alliance e2695 HPLC system coupled to a QDa mass detector and a 2998 photodiode array (PDA) detector, operated in both positive and negative electrospray ionization (ESI) mode. Chromatographic separation was achieved on a Waters XBridge BEH C18 column (4.6 × 150 mm, 15 μm) with a mobile phase of 0.1% formic acid in water (phase A) and methanol (phase B), eluted at a flow rate of 1.0 mL/min with a linear gradient of 5–95% B over 15 min. MS data were acquired in full-scan mode (m/z 100–1000). PDA detection was monitored at 254 nm for purity assessment.

Cellular proliferation assay

Wild-type and mutant FGFR sequences (generated via site-directed mutagenesis) were cloned into the MSCV retroviral expression vector. Retroviral particles were produced by transfecting the MSCV-FGFR plasmids into HEK293T. Viral supernatants were harvested 48 h post-transfection and utilized to transduce wild-type Ba/F3 cells. Following transduction, the cells underwent a round of puromycin selection. The selected Ba/F3 cells were subsequently cultured and expanded in an interleukin-3 (IL-3)-depleted medium. LC-F2-1, futibatinib, pemigatinib, and lirafugratinib were evaluated using a proliferation assay in Ba/F3-TEL-FGFR1-4 or Ba/F3-TEL-FGFR2 cells harboring different mutations. A volume of 2 μL stock solution was added to 98 μL of growth medium, with a final dimethyl sulfoxide (DMSO) concentration of 0.1% in each well. A threefold, 9-point dilution was prepared, and the experiments were performed in duplicate. The cell lines were incubated at 37 °C for 3 days, following which the assay plate was equilibrated to room temperature and 50 μL of CellTiter-Glo Reagent (#G7573, Promega) was added to each well. The contents were mixed for 120 s on an orbital shaker to induce cell lysis and incubated at room temperature for 10 min to stabilize the luminescent signal. Luminescence was recorded using a SpectraMax Paradigm (Molecular Devices). Inhibition Rate (Inh%) was calculated relative to vehicle (DMSO)-treated control wells using the following formula:

$$\mathrm{Inhibition\; Rate}\left(\%\right)=\left(1-\frac{{RLU}_{\mathrm{Compound}}-{RLU}_{\mathrm{Blank}}}{{RLU}_{\mathrm{Control}}-{RLU}_{\mathrm{Blank}}}\right)\times 100\%$$ 1

Data were analyzed using Graphpad Prism 8.0, fitted to a four-parameter logistic model to generate the dose–response curves.

Immunoblotting

SNU-16 human gastric cancer cells, which harbor high-level FGFR2 gene amplification, were obtained from the Cell Bank of the Chinese Academy of Sciences and cultured in RPMI 1640 medium supplemented with 10% fetal bovine serum (FBS). KATO III cells, characterized by high-level FGFR2 gene amplification and constitutive FGFR2 activation, were purchased from the Cell Bank of the Chinese Academy of Sciences and maintained in RPMI 1640 medium supplemented with 10% FBS. SNU16 or KATOIII cells were treated with DMSO, LC-F2-1, futibatinib, and pemigatinib for 6 h at 37 °C with 5% CO₂. Cells were washed and lysed in RIPA buffer (#K1121, APExBIO) supplemented with phosphatase inhibitors (#P1046, Beyotime). The protein concentration was standardized using a bicinchoninic acid reagent kit (#A55860, Thermo Scientific), following the manufacturer’s instructions. Lysates were mixed with loading buffer for sodium dodecyl sulfate polyacrylamide gel electrophoresis and immunoblotting. Antibodies against pFGFR (Y653/654, #3471), pPLCγ (Y783; #14008), PLCγ (#2822), pAKT (S473, #4058), and AKT (#9272) were obtained from Cell Signaling Technology. The following antibodies were sourced from HUABIO: FGFR2 (#HA721434), FRS2 (#ER63474), pSHP2 (Y524, #ET1611-22), SHP2 (#ET1611-35), pERK (T202/Y204, #ET1610-13), ERK (#ET1601-29), pRSK1P90 (T573; #HA721793), and GAPDH (#ET1601-4). Antibody against pFRS2α (Y436; #31352-1-AP) was purchased from Proteintech. Blots were developed using enhanced chemiluminescence reagents (no. P10300, NCM Biotech Laboratories) by the FUSION FX EDGE (VILBER).

Purification of FGFR2 V565F kinase domains

The FGFR2 kinase domain (P458-E768, with a V565F mutation) was cloned into pET28a with a cleavable N-terminal 6×His tag. FGFR2 was co-expressed with YopH in the Escherichia coli T7 Express strain. The protein was purified using His FF, HiTrap Q, and Superdex 75 columns by Biortus Discovery Co., Ltd. After the first purification round, the N-terminal 6×His tag was cleaved using TEV protease, and the protein was purified a second time using affinity chromatography (High affinity Ni Resin FF). Purified FGFR2 was stored in 25 mM Tris-HCl (pH 7.5), 50 mM NaCl, and 2 mM TCEP at 14.41 mg/ml. The protein molecular weight was determined using liquid chromatography/mass spectrometry (LC/MS; FGFR2 V565F = 35683.27 Da). The WT FGFR2 kinase domain (P458-E768) was cloned into the pET28a expression vector with a cleavable N-terminal 8×His tag.

Intact mass analysis

The purified FGFR2 V565F was diluted to 28 μM, then incubated with LC-F2-1 (molar ratio 1:2 or 1:5) or futibatinib (molar ratio 1:2) at room temperature for 4 h. LC-MS analysis showed that all FGFR2 was covalently bound to the compound after the incubation. Next, intact mass analysis was performed via Dual AJS ESI TOF MS using an Agilent 6230 time-of-flight LC/MS system. FGFR2 protein samples were desalted prior to MS using an Agilent 1290 Infinity II instrument. BioConfirm software was employed to deconvolute the protein spectra.

Protein crystallization and structure determination

FGFR2 V565F (14.41 mg/ml) was incubated with LC-F2-1 (molar ratio 1:2, 1.6% DMSO) at RT for 4 h, then centrifuged at approximately 11,000 × g for 5 min prior to crystallization experiments. Nanoliter-scale vapor diffusion crystallization experiments were set up at 20 °C using a NT8 protein-crystallization system (Formulatrix) and commercially available sparse-matrix screens (Molecular Dimensions, Hampton Research). Following optimization, good-quality crystals were obtained via self-seeding in an Index-H6 + Silver bullets-G6 setup containing 0.2 M sodium formate, 20% PEG 3,500, 16% glutaric acid, 0.16% oxalic acid, 0.16% sebacic acid, 0.16% mellitic acid, 0.16% trans-cinnamic acid, 0.16% pimelic acid, and 0.02 M HEPES-Na (pH 6.8). The co-crystals of FGFR2-LC-F2-1 appeared in about 9 days at 20 °C. After cryoprotection with 20% glycerol in the crystallization solution, the crystals were rapidly vitrified in liquid nitrogen. X-ray diffraction measurements were performed at the BL45XU beamline (synchrotron SPring-8, Japan).

The crystals grew in the space group P212121, with the following unit cell dimensions: a = 63.07 Å; b = 88.71 Å; c = 127.42 Å. Data were reduced and phased using the XDS program package and the ccp4 program suite, respectively. The electron density corresponding to the compound was resolved unambiguously. Following iterative cycles of model building and refinement of the protein components, the ligand was successfully fitted to the composite omit maps. Structural refinement was performed using REFMAC5, with ligand geometry restraints generated by JLigand. The regions lacking interpretable electron densities were excluded from the final atomic model. Finally, the structure was refined to a resolution limit of 3.25 Å (detailed data collection and refinement statistics are provided in Supplementary Table 10). The PDB ID for the FGFR2- LC-F2-1 structure is 9U3N.

Kinome selectivity profiling

The kinase selectivity profile of LC-F2-1 across a 416-kinase panel was determined using homogeneous time-resolved fluorescence-based assays performed by ICE Bioscience (Beijing). The experimental procedure involved sequential addition using acoustic dispensing (ECHO) technology: initial loading of 100× compound concentration into 384-well plates, followed by the addition of a 2× kinase/metal cofactor mixture. After brief centrifugation (approximately 200 × g for 1 min), the reaction system underwent 10-min equilibration at 25 °C. Subsequently, a 2× substrate/ATP mixture in kinase buffer was introduced, followed by another round of centrifugation and 60-min of enzymatic reaction at 25 °C. Detection was performed using homogeneous time-resolved fluorescence-compatible reagents containing a cryptate-conjugated antibody and XL665 fluorophore. Fluorescence resonance energy transfer signals were quantified via dual-wavelength detection (620 and 665 nm) on a microplate reader.

Animal study

Ba/F3-FGFR2 M538I or Ba/F3-FGFR2 V565F cells formulated in Matrigel were subcutaneously injected into 6-week-old BALB/c female mice. Once the tumors reached an average volume of approximately 100 mm³, the mice were randomized into designated groups. For the Ba/F3-FGFR2^V565F xenograft study, n = 5 biologically independent mice for the vehicle group and n = 6 biologically independent mice for each treatment group. For the Ba/F3-FGFR2^M538I xenograft study, each group included n = 6 biologically independent mice. The mice received daily oral administration (p.o.) of the vehicle control, LC-F2-1, or the reference compounds (pemigatinib, futibatinib, or lirafugratinib). Tumor sizes and body weights were measured three times a week.

Supporting information

Supporting Figures, supporting Tables, chemistry and compound characterization, as well as copies of the ¹H NMR, ¹³C NMR, and HPLC spectra, and uncropped blots.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Supplementary information

Supplementary Information ^{(9MB, pdf)}

Reporting Summary ^{(102.8KB, pdf)}

Transparent Peer Review File ^{(2.5MB, pdf)}

Source data

Source Data ^{(118.6KB, xlsx)}

Author contributions

X.H., L.Z. (Lulu Zheng), R.L., and Z.F. performed the Western blotting experiments and prepared the related figures. X.C., L.Z. (Lei Zheng), and Z.X. synthesized the compounds and characterized their chemical structures. Y.L., Y.Z., and S.C. contributed to the cell-based and animal experiments. X.L. and G.L. contributed to manuscript writing and scientific discussion. L.C. conceived and supervised the project, analysed the structural models and wrote the manuscript. All authors discussed the results and contributed to revision of the manuscript.

Peer review

Peer review information

Nature Communications thanks Michael Cameron and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

The crystal structure generated in this study and the raw mass spectrometry data have been deposited in public repositories. Accession codes for newly generated and previously published datasets used in this study are listed below: 9U3N (FGFR2 V565F kinase domain in complex with LC-F2-1) PXD073416 (raw mass spectrometry data characterizing covalent binding between FGFR2 and LC-F2-1). Raw uncropped Western blot images are provided in Supplementary data. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains Supplementary material available at https://doi.org/10.1038/s41467-026-76339-0.