EGFR Inhibition Overcomes Resistance to FGFR4 Inhibition and Potentiates FGFR4 Inhibitor Therapy in Hepatocellular Carcinoma

Hepatocellular carcinoma (HCC), accounting for up to 90% of all primary liver cancers, is one of the leading causes of cancer mortality worldwide. To date, several therapies have been approved as first-line treatments for HCC, namely sorafenib, lenvatinib, and a combination of bevacizumab with atezolizuma. Agents that have been approved as second-line therapies include nivolumab, regorafenib, and cabozantinib (1). Unfortunately, these therapies have only been shown to provide limited therapeutic efficacy to patients with HCC, with relatively short progression-free survival and overall survival. Therefore, novel and more effective therapies are still in urgent need for this markedly lethal and prevalent disease.

Genomic and functional studies have identified FGF19 as a potential driver oncogene in HCC. FGF19 is a gut secreted endocrine hormone that functions in the liver via its receptor, FGFR4, and co-receptor, klotho-beta, in the regulation of bile acid synthesis (2, 3). FGF19 also activates MAPK signaling to induce hepatocyte proliferation (4). It has been reported that 20% to 30% of HCC cases exhibit overexpression of FGF19, with ∼7% exhibiting FGF19 amplification (5–7). In a previously conducted study, overexpression of FGF19 led to the development of liver tumors in transgenic mice (8), which were abolished in an FGFR4-null background (9). These studies all provided a strong rationale for the development of FGFR4-targetting therapies as novel treatment options for patients with HCC harboring FGF19 alterations.

Several FGFR4 inhibitors have advanced into clinical studies for the treatment of HCC. Among them, BLU-554 (fisogatinib) is a particularly potent, selective, and irreversible FGFR4 inhibitor that has demonstrated preliminary efficacy with an overall response rate (ORR) of 17% as a second-line treatment for patients with HCC harboring FGF19 overexpression (6). The observed responses, however, only lasted a few months in average before relapses, suggesting the likely occurrence of acquired resistance. Indeed, acquired resistance has been frequently found in the utilization of tyrosine kinase inhibitors (TKI) for the treatment of cancers, posing a major challenge to treatment. Several mechanisms often underlie acquired resistance. In some cases, gatekeeper mutations of receptor tyrosine kinases (RTK) have been identified as the causes of resistance, notably T790M mutations of EGFR following treatment with first or second generation EGFR-TKIs (10, 11). On the other hand, numerous studies have also suggested that the activation of bypass RTKs signaling is often associated with acquired resistance. For instance, previous studies have indicated that mesenchymal–epithelial transition factor (MET) and FGFR activation confer resistance to EGFR inhibition while MET and erb-b2 receptor tyrosine kinase 2/3 (ERBB2/3) activation confer resistance to FGFR inhibition (12–15). In a recent phase I study of BLU-554 in HCC, FGFR4 mutations were detected in only a few cases (16). In the majority of patients, it still remains unknown what resistance mechanisms led to quick relapses.

To further understand the molecular mechanisms involved in acquired resistance to FGFR4 inhibition, we used a resistant cellular system established in-house to study the signaling cascades that may contribute to the resistance. Multiple resistant clones were established via the long-term exposure of Huh7, an FGF19 amplified and overexpressed HCC cell line known to be sensitive to FGFR4 inhibition, to a high concentration of BLU-554. Through phospho-RTK screening of the resistant cells, we identified EGFR as the only activated RTK, which under activation leads to the consequential activation of downstream PI3K and MAPK signaling. Inhibition of EGFR or its downstream components effectively restored BLU-554 sensitivity in these cells. In addition, EGF stimulation of Huh7 and two other FGFR4-dependent HCC cell lines, Hep3B and JHH7, resulted in resistance towards BLU-554. The above findings were confirmed in vivo in relevant xenograft tumor models. Furthermore, we found in patient samples from The Cancer Genome Atlas (TCGA) that EGFR activation did indeed occur in a significant percentage of patients with HCC harboring FGF19 overexpression. Our studies altogether have, for the first time, demonstrated that EGFR activation functions as a critical resistance factor in FGFR4 inhibition resistance, suggesting that concurrent inhibition of both EGFR and FGFR4 pathways will provide a clinically effective therapeutic strategy for the treatment of patients with HCC harboring FGF19 overexpression.

Human HCC cell lines, Huh7 and Hep3B were purchased from National Collection of Authenticated Cell Cultures and JHH7 was purchased from Cobioer Co., Ltd. We cultured Huh7 in DMEM medium (Gibco, catalog no. 11965188) supplemented with 10% FBS (Gibco, catalog no. 10099–14) and 100 U/mL penicillin/streptomycin (Gibco, catalog no. 15140122), Hep3B in MEM medium (Gibco, catalog no. 11095098) with 10% FBS, 100 U/mL penicillin/streptomycin and 100 μmol/L Nonessential Amino Acid (Gibco, catalog no. 11140050), and JHH7 in William's E medium (Gibco, catalog no. 12551032) with 10% FBS, 1% penicillin/streptomycin and 2 mmol/L l-Glutamine (Gibco, catalog no. 25030081). All cell lines were maintained in humidified incubator with 5% CO₂ at 37°C. Mycoplasma contamination was excluded via Myco-Lumi Luminescent Mycoplasma Detection Kit (Beyotime, catalog no. C0298M). The cell lines were authenticated by short tandem repeat DNA profiling. Cells were used within 10 passages after thawing.

BLU-554 was either internally synthesized following previous publication (6) or obtained from Shanghai Caerulum Pharma Discovery Co., Ltd. H3B-6527, FGF-401, and JNJ-42756493 were internally synthesized following previous publications (5, 17, 18). Gefitinib was obtained from Accela ChemBio Co., Ltd. Sorafenib was obtained from J&K Scientific Ltd. and Bide Pharmatech Ltd. Other chemicals, including erlotinib (catalog no. S7786), TNO155 (ref. 19; catalog no. S8987), BI3406 (ref. 20; catalog no. S8916), MK2206 (ref. 21; catalog no. S1078), trametinib (catalog no. S2673), and staurosporine (ref. 22; catalog no. S1421) were obtained from Selleck Chemicals. ET070 (23) was obtained from ETERN Therapeutics. For in vitro studies, compounds were dissolved in DMSO. For in vivo efficacy studies, BLU-554 was formulated in a mixture of 0.5% CMC, 20% PEG300, and 10% solutol HS 15; gefitinib was formulated in 0.5% HPMC and 0.1% Tween 80; and sorafenib was formulated in a mixture of 10%ETOH, 10% Cremophor EL, and 80% Saline.

N-ethyl-N-nitrosourea (ENU, Sigma, catalog no. N3385), a mutagenic reagent, was dissolved in DMSO at 50 mg/mL and stored in aliquots at −80°C. We attempted to generate BLU-554–resistant clones in three HCC cell lines (Huh7, Hep3B, and JHH7) with FGF19 overexpression and sensitivity to BLU-554. These cell lines were pre-treated with ENU at a working concentration of 50 μg/mL for 24 hours in 10-cm dish, then washed 3 times with DMEM and reseeded in multiple 96-well plates at a cell density of 0.5 cell per well. Subsequently, the cells were allowed to recover overnight before long-term BLU-554 treatment at 1 μmol/L. Ultimately, we successfully established Huh7 resistant clones, which designated as Huh7-A1, A2, and A3. Such long-term treatment of Hep3B and JHH7 cell lines did not lead to viable clones. This suggests that titration of additional doses might be needed in future studies in attempt to establish other resistant clones. Our established BLU-554–resistant cells are available from the corresponding author upon reasonable request.

Huh7-A1, A2, and A3 cell lines were seeded at 3,500 cells per well in 96-well plates. After incubation overnight, triplicates of cells were exposed to BLU-554 plus gefitinib, erlotinib, ET070, or TNO155 at the indicated concentrations, respectively, in matrix for 3 days. Cell viability was assessed using CellTiter-Glo (Promega, catalog no. G7570) according to the manufacturer's instruction. Measured anti-proliferation effects for each well were normalized to DMSO control and values were tabulated in percentage format based on instruction of the software. Loewe synergy and antagonism index were calculated with output using Combenefit_v2.02 software (24).

CellTiter-Glo assay kit (Promega) was used for the measurement of cell viability according to the manufacturer's instruction. Briefly, cells were seeded at the density of 3,500 (Huh7 parental and resistant clones) or 2,000 (Hep3B, JHH7) cells per well in 96-well black plates (Corning, #3603), incubated overnight and treated with drugs for 3 days (Huh7, resistant clones) or 5 days (Hep3B, JHH7). After treatment, 50 μL CellTiter-Glo reagent was added to each well and incubated on a shaker for 20 minutes at room temperature, protected from light. Cell viability was determined by measuring luminescent signal with Envision Multilabel Reader (PerkinElmer). The day-0 luminescent signal was subtracted from terminal signal then normalized to DMSO control. Data was further analyzed by fitting to standard dose–response curve using GraphPad Prism 9.0 (RRID:SCR_002798).

Cells were washed twice with PBS, lysed with RIPA buffer (Sigma, catalog no. R0278) supplemented with protease inhibitor cocktail (Sigma, catalog no. P8340) and phosphatase inhibitor (Sigma, catalog no. P5726 and catalog no. P0044) then incubated 10 minutes on ice. Supernatants were collected by centrifuging at 20,000 × g for 10 minutes at 4°C. Protein concentrations were measured using the BCA Protein Assay Kit (Thermo Fisher Scientific, catalog no. 23225). Equal amounts of total protein were loaded for electrophoresis with 4% to 15% Criterion TGX Precast Gels (Bio-Rad, catalog no. 4561086) and transferred (Bio-Rad, catalog no. 1704150) to PVDF membrane. Then membranes were blocked in QuickBlock (Beyotime, catalog no. P0252–500 mL) reagent for 1 hour at room temperature. Subsequently, membranes were probed with primary antibodies and horseradish peroxidase (HRP)-conjugated secondary antibodies. Finally, Clarity Max Western ECL Substrate (Bio-Rad, catalog no. 1705062) was added to the membranes and the blots were resolved with Bio-Rad Chemi-Doc imaging system. Primary antibodies were purchased from Cell Signaling Technology: phospho-FRS2 (Cell Signaling Technology, catalog no. 3861, RRID:AB_2231950), phospho-ERK1/2 (Cell Signaling Technology, catalog no. 9101, RRID:AB_331646), ERK1/2 (Cell Signaling Technology, catalog no. 4695, RRID:AB_390779), Phospho-AKT (Cell Signaling Technology, catalog no. 4060, RRID:AB_2315049), AKT (Cell Signaling Technology, catalog no. 9272, RRID:AB_329827), Phospho-EGFR (Cell Signaling Technology, catalog no. 2234, RRID:AB_331701), EGFR (Cell Signaling Technology, catalog no. 2232, RRID:AB_331707), GAPDH (Cell Signaling Technology, catalog no. 2118, RRID:AB_561053), and SantaCruz: FRS2 (Santa Cruz Biotechnology, catalog no. sc-17841, RRID:AB_2106230). Goat anti-Rabbit (Bio-Rad, catalog no. 170–6515, RRID:AB_11125142), and anti-mouse (Bio-Rad, catalog no. 170–6516, RRID:AB_11125547) IgG-HRP–conjugated secondary antibody were purchased from Bio-Rad.

Cells were seeded in 96-well plates in triplicate at density of 10,000 cells per well. After overnight starvation, cells were treated with BLU-554 and gefitinib or erlotinib for additional 24 hours. Cells were fixed in cold 70% ethanol for 2 hours at 4°C. Fixed cells were subsequently centrifuged at 400 × g for 5 minutes at 4°C, resuspended in PBS with RNase A for 1 hour at 37°C, and stained with propidium iodide (PI) for 30 minutes at room temperature. The stained cells were measured with flow cytometer and analyzed with NovoExpress software (ACEA, NovoCyte).

Cells were seeded in 96-well plates in triplicate at density of 10,000 cells per well. After incubation overnight, cells were treated with BLU-554 and gefitinib or erlotinib for 4 days. Cells were then labeled with annexin V and PI (Beyotime, Inc. C1062S) at room temperature for 10 minutes according to the manufacturer's instruction. Apoptosis was determined by flow cytometer and NovoExpress software. Early and late apoptosis were defined as annexin V⁺ PI⁻ and annexin V⁺ PI⁺, respectively.

Cells were seeded in triplicate in 96-well black/clear plate at density of 10,000 cells per well. After incubation overnight, cells were treated with BLU-554 and gefitinib or erlotinib for another 3 days. Caspase 3/7 activity of cells was measured using Caspase-Glo 3/7 Assay kit (Promega, G8091) according to the manufacturer's instruction. Briefly, 50 μL of Caspase-Glo reagent was added to the cells, incubated and protected from light for 1 hour at room temperature. Luminescence signal was measured by Envision Multilabel Reader (PerkinElmer). The fold change of caspase 3/7 activity was normalized to DMSO.

Screening of activated RTKs was performed using the Proteome Profiler Human Phospho-RTK array kit (R&D Systems, ARY001B) according to the manufacturer's instruction. Briefly, Huh7 parental and resistant cell lines were maintained in media at 37°C. In collecting cell extracts, cells were washed with excess pre-cold PBS and lysed with lysis buffer plus protease and phosphatase inhibitors on ice for 30 minutes. Lysates were centrifuged at 12,000 rpm for 15 minutes and supernatants were collected. Total protein at 300 μg was incubated with blocked array membranes overnight at 4°C. Array membranes were washed 3 times and incubated with HRP conjugated anti-phospho-tyrosine for 2 hours at room temperature. The membranes were then washed 3 times again before development with ECL Western blotting detection reagent (R&D Systems, VL001–200). The signals were densitometrically quantified by Image Lab software. The amount of phosphorylated RTKs from resistant cells was compared with parental cells.

Tumors were collected from subcutaneously xenografted mice 2 hours post the final oral administration, and fixed in 10% formalin solution at room temperature for 48 hours. The fixed sample were then washed in water for 20 minutes and transferred to 75% alcohol solution. Each tumor was embedded in paraffin and sliced for subsequent Immunohistochemistry (IHC) analysis. Formalin-fixed, paraffin-embedded slices were stained using the following antibodies: Ki67 antibody (1:500, Biolynx, catalog no. BX50040-C3), anti-cleaved caspase 3 (1:200, CST, catalog no. 9661S) as well as terminal deoxynucleotidyl transferase-mediated dUTP nick-end labeling (TUNEL) assay. The stained slides were digitalized using tissue slide scanner (3DHISTECH).

All animal studies were approved by Institutional Animal Care and Use Committee of Shanghai BK/KY Biotechnology Co., Ltd. and Shanghai Medsyin Biotechnology Co., Ltd. The BALB/c-nude female mice at approximately 8-week-old, weighing 18–20 grams were obtained from Sippr-BK Laboratory and housed under specific pathogen–free conditions. Resistant Huh7-A3 and parental Huh7 cells were cultured and kept in exponential growth phase. Cells were collected after trypsinization and washed with PBS for 3 times, then suspended in a 1:1 mixture of PBS and Matrigel (Corning, 354234) at a final concentration of 100 million cells per mL. To generate xenografts, 0.1 mL of the inoculum was injected subcutaneously into the right flank region of mice, giving a final concentration of 10 million cells per mouse. When the mean tumor volume (TV) reached approximately 180 mm³, mice were randomized into four groups with 6 animals per group. BLU-554 (30 mpk, orally, twice a day), gefitinib (150 mpk, orally, once daily), or combination was administered orally. Body weight and tumor measurements were performed twice a week. Dosing was suspended when the body weight loss exceeded 10%. Mice with more than 20% body weight loss or TV exceeded 2,000 mm³ were immediately euthanized. The TV in mm³ was calculated according to the following formula: TV = length × width² × 0.5, while length was the largest diameter of tumor (mm) and width was the diameter perpendicular to length (mm). Tumor growth inhibition% (TGI) was calculated according to the following formula: TGI = [(Average control TV day X – control TV day 0) – (Treatment TV day X – treatment TV day 0)] / (Average control TV day X – control TV day 0) × 100%, where day X is any day of the treatment.

Total RNA was extracted using TRIzol Reagent (Invitrogen, catalog no. 15596026) and assessed for quality with the Agilent 2100/2200 Bioanalyzer (Agilent Technologies) and NanoDrop (Thermo Fisher Scientific). The sequencing was performed on the NovaSeq 6000 platform (Illumina) using 1 μmol/L of the total RNA. Sequencing reads were aligned to the GRCh38 genome by Hisat2 (v2.0.1). Gene expression levels were calculated using HTSeq (v0.6.1) and normalized using the fragment per kilobase of transcript per million fragments mapped (FPKM) method. Differential expression analysis was conducted by the negative binomial model implemented in the Bioconductor package DESeq2 (V1.6.3), with P values adjusted using the FDR procedure. The EGFR pathway enrichment profiles were generated using the Preranked gene set enrichment analysis (GSEA) method (25, 26), in which genes were pre-ranked by P values of the resistant cells over parental cells. The MSigDB Hallmark and C2 pathway databases were used for enrichment analysis (25, 27).

Samples in the HCC cohort with FGF19 mRNA expression in the top 25% were classified as the high FGF19 expression (FGF19^hi) HCC group. To categorize the FGF19^hi HCC subgroups, we selected 40 EGFR activation signatures from the MSigDB C2 pathway database. For consensus clustering, we used the ConsensusClusterPlus R package (28) with the following parameters: maxK = 6, reps = 50, pItem = 0.8, pFeature = 1, distance = euclidean, and clusterAlg = km. The optimal number of subgroups was determined by evaluating the consensus matrix, delta area curves representing changes in the area under the CDF curve, and the average silhouette distance for consensus clusters. In the TCGA-Liver Hepatocellular Carcinoma (LIHC) cohort, the consensus matrix with k = 2 was identified as the optimal choice based on the CDF curve of the consensus score (29) and showed a distinct separation from the other clusters. Silhouette plots with k = 2 did not exhibit and significant negative values. In the International Cancer Genome Consortium Liver Cancer-RIKEN (ICGC-LIRI) cohort, the cleanest consensus matrix was obtained for k = 3. The CDF curve indicated that k = 3 was the optimal cluster number, and the 3-cluster did not contain silhouette widths with significant negative values. PCA of the ade4 R package was used to distinguish different subtype information in a two-dimensional space (30).

The results were analyzed with Graphpad Prism v9.0 using appropriate statistical tests. Data were presented as the mean ± SEM of at least three technical replicates (except for in vivo studies). P values of <0.05 were considered to be statistically significant.

The raw RNA sequencing (RNA-seq) data generated in this study had been deposited at the National Center for Biotechnology Information under the BioProject accession number PRJNA987200. Publicly available RNA-seq data comprising 371 samples were collected from the TCGA-LIHC dataset in the Genomic Data Commons (GDC) database (https://portal.gdc.cancer.gov/). For external validation, a cohort of 232 samples was collected from the ICGC-LIRI dataset (https://dcc.icgc.org/).

Huh7 is an HCC cell line harboring FGF19 amplification/overexpression and sensitive to FGFR4 inhibitors in proliferation. To explore the potential mechanisms of acquired resistance to FGFR4 inhibitors, we generated resistant cell lines via the long-term culture of Huh7 cells with exposure to BLU-554 at high concentrations. Three independent clones, Huh7-A1, A2, and A3, were selected and established as resistant cell lines. Results from proliferation experiments confirmed that all three resistant cell lines were indeed insensitive to the treatment of BLU-554, as well as with other FGFR4 inhibitors namely FGF-401, H3B-6527, and JNJ-42756493 (erdafitinib), while parental Huh7 cells were sensitive to all mentioned FGFR4 inhibitors. All of cells, including the parental and resistant clones, displayed sensitivity to staurosporine, a nonselective multi-kinase inhibitor. This indicates that the resistant clones specifically developed resistance to FGFR4 inhibitors (Fig. 1A). The treatment with BLU-554 effectively decreased the phosphorylation of downstream FRS2 and ERK1/2 in parental cells, these effects were completely absent in the resistant cell lines (Fig. 1B).

Next, we sought out to identify the potential mechanisms mediating the acquired resistance to FGFR4 inhibition in the resistant Huh7-A1, A2, and A3 cells. Whole-genome sequencing was performed but the results did not indicate any particular mutation on FGFR4 itself or on any other key oncogenes. Following this, a phospho-RTK array was used to examine potential feedback signaling activation by comparing the parental and resistant cell lines. Among the 49 RTKs evaluated by this array, only phospho-EGFR was found to be upregulated in all three resistant cell lines (Fig. 2A). Consistently, subsequent Western blot results confirmed the strong increase of phospho-EGFR, phoshpo-ERK1/2, and phosphor-AKT, in all three resistant cell lines (Fig. 2B). We also performed RNA-seq and GSEA on both parental and resistant cell lines. Results from these analyses revealed that multiple gene sets related to EGF or EGFR activation were significantly enriched in the resistant cell lines, further confirming that activation of EGFR signaling occurs in these cells and provide resistance to FGFR4 inhibition (Fig. 2C–F).

To study the functional roles of EGFR activation and activation of its downstream signaling components in the resistant cell lines, we evaluated whether or not direct inhibition of EGFR could restore the sensitivity of BLU-554 in these cells. Two selective EGFR-TKIs, gefitinib and erlotinib, were used. Both inhibitors reverted the sensitivity of all three resistant cell lines to the treatment of BLU-554 in cellular proliferation experiments (Fig. 3A; Supplementary Fig. S1A). To assess the underlying mechanisms involved, we first examined cell apoptosis upon treatment with EGFR-TKIs and BLU-554 via caspase 3/7 and annexin V/PI analyses. When applied alone, FGFR4 and EGFR inhibition did not produce significant cell apoptosis in any of the three resistant cell lines, whereas concomitant FGFR4 and EGFR inhibition induced pronounced apoptosis signals in both assays. Similarly, we found via parallel cell-cycle analysis that only FGFR4 and EGFR combinatory inhibition induced cell-cycle arrest at S phase (Fig. 3B; Supplementary S1B), while treatment with either inhibitor alone did not produce such effect. Furthermore, combined inhibition of FGFR4 and EGFR decreased the phosphorylation of EGFR and downstream phosphorylation of ERK1/2 and AKT (Fig. 3D; Supplementary S1F). Gene expression analysis was also conducted to compare the RNA-seq data from resistant cell lines treated with BLU-554 alone versus those treated with the combination of BLU-554 and gefitinib. The GSEA results revealed significant suppression of EGFR activation, KRAS, and cell proliferation signaling in the group treated with the combination of BLU-554 and gefitinib (Supplementary Fig. S1C). In addition, inhibition of PI3K/AKT/mTOR signaling was observed with combinatory treatment in certain resistant cell lines (Supplementary Fig. S1D). These results indicate that EGFR activation in combination with FGFR4 activation play important roles in the cause of FGFR4 inhibition resistance, thus suggesting that concomitant inhibition of both pathways may be required to sufficiently block downstream signaling, inhibit proliferation, and induce apoptosis.

On the basis of the above findings, we then proceeded to test if inhibition of EGFR downstream components could produce similar effects. Indeed, SHP2 inhibition via the application of two different inhibitors, TNO155 and ET070, potently restored BLU-554 sensitivity in all three resistant cell lines (Fig. 3C; Supplementary Fig. S1E). Combined inhibition of FGFR4 and SHP2 effectively suppressed downstream phosphorylation of ERK1/2 and AKT (Fig. 3D; Supplementary Fig. S1F). Application of SOS1 inhibitor BI3406 also produced a similar effect in restoring BLU-554 sensitivity (Supplementary Fig. S1G). Interestingly, standalone application of AKT inhibition by MK2206 or MEK inhibition by trametinib was not effective in restoring sensitivity. However, combination of AKT and MEK inhibition was able to potently restore sensitivity towards FGFR4 inhibition (Supplementary Fig. S1H). These results altogether suggest that concomitant and complete inhibition of EGFR, FGFR4, and their common downstream signaling pathways may be required to completely inhibit the growth of the resistant cells.

Our above findings demonstrated that feedback activation of EGFR and its downstream signaling components leads to acquired resistance towards chronical FGFR4 inhibition in HCC. To further understand the functional connection between EGFR signaling and FGFR4 dependency, we also examined whether or not acute EGFR activation in parental HCC cells could lead to rapid resistance against FGFR4 inhibition. Upon supplementing exogenous EGF to culture media of the three parental FGFR4-dependent HCC cell lines (Huh7, Hep3B, and JHH7), the IC₅₀s of BLU-554 in proliferation experiments significantly increased, suggesting that acute EGFR activation can indeed lead to the rapid development of resistance to FGFR4 inhibitors. In the presence of EGF stimulation, the addition of EGFR inhibitors gefitinib and erlotinib, as well as SHP2 inhibitors TNO155 and ET070, effectively re-sensitized the EGFR activated HCC cells to BLU-554, similar to the condition without EGF stimulation. With BLU-554 at low concentration ranges, combination of SHP2 inhibitors exhibited a more pronounced inhibition compared with EGFR inhibitors (Fig. 4A). This difference may be attributed to the fact that SHP2 acts as the downstream convergence point of both FGFR4 and EGFR signaling. Consistently, EGF supplementation induced a substantial increase in EGFR phosphorylation, followed by enhanced phosphorylation of ERK1/2 and AKT, even in the presence of BLU-554. These EGF-induced effects were eliminated by the addition of EGFR inhibitors or SHP2 inhibitors (Fig. 4B). Collectively, these findings suggest that EGF-induced activation of EGFR and MAPK/PI3K-AKT may serve as an innate bypass resistance mechanism, allowing cells to grow in the presence of FGFR4 inhibition.

To confirm the above findings in vivo, we performed efficacy studies in xenograft tumor models inoculated with the resistant or parental Huh7 cells. Upon testing all three resistant cell lines in vivo in mice, only the Huh-A3 xenografts had reasonable tumor growth suitable for efficacy study. Therefore, this model was used to examine the effects of dual EGFR and FGFR4 inhibition in vivo. When the inoculated Huh7-A3 and Huh7 parental tumors grew to an average of ∼180 mm³, mice were randomized into four groups and orally dosed with vehicle control, BLU-554 (30 mg/kg twice daily), gefitinib (150 mg/kg once daily), or a combination of BLU-554 and gefitinib. While monotherapy of BLU-554 demonstrated strong antitumor effects in the Huh7 parental xenograft (Fig. 5B), it only had minimal efficacy in the Huh7-A3 xenograft (Fig. 5A), confirming the resistance of Huh7-A3 to FGFR4 inhibition in an in vivo setting. Monotherapy of gefitinib had no inhibition on tumor growth in either xenograft models. However, combinatory treatment of BLU-554 with gefitinib demonstrated strong synergy and led to persistent tumor regression in both models (Fig. 5A and B). Mice were tolerable with the combination treatment (Supplementary Fig. S2A and S2B). In addition, we performed IHC staining using paraffin-embedded tumor slices and found that the levels of Ki67 were greatly downregulated and the apoptotic cells (cleaved caspase-3 and/or TUNEL-positive) were increased by the combination treatment of BLU-554 and gefitinib, indicating that such combination not only inhibited proliferation but also induced apoptosis of the tumor cells in vivo (Fig. 5C and D). Consistent with TGI, single agent treatment with BLU-554 or gefitinib produced only marginal decreases in phospho-ERK1/2 in the collected Huh7-A3 tumor samples, while combinatory treatment almost completely abolished ERK1/2 phosphorylation (Fig. 5E).

Because sorafenib is the standard first-line treatment for patients with HCC, we also assessed its antitumor effect in HCC xenografts with FGF19 overexpression. Daily dosing of sorafenib at 30 mpk only resulted in partial TGI, with TGI% at 44% for Huh7 and 50% for Hep3B xenografts, respectively (Supplementary Fig. S2C and S2D).

Collectively, these results indicate that EGFR activation occurs in vivo and render resistance to FGFR4 inhibition. Dual targeting of FGFR4 and EGFR could provide synergistic and enhanced antitumor effect, superior to the efficacy of sorafenib monotherapy.

We further investigated the potential clinical relevance of EGFR activation in patients with HCC harboring FGF19^hi. From the aforementioned preclinical findings, we observed elevated EGFR activation gene signature score and increased phosphorylation of EGFR and downstream proteins in BLU-554–resistant cell lines (Fig. 2B–E). But total EGFR protein or mRNA levels appeared not significantly increased in the resistant cell lines (Fig. 2B and F). Consequently, we employed consensus clustering on the enrichment score of published gene signatures for EGFR activation to assess EGFR activation status in FGF19^hi patients (see materials and methods). In TCGA datasets, the single-sample GSEA (ssGSEA) enrichment score of EGFR activation signatures clustered FGF19^hi patients with HCC (n = 93) into two subgroups based on EGFR activation levels: high (CL1, n = 51) and low (CL2, n = 42; Fig. 6A–C; Supplementary Fig. S3). The enrichment scores of EGFR-downstream pathways, namely KRAS, PI3K/AKT/mTOR, ERK1/2, and MAPK signaling pathways from the MsigDB HALLMARK and BioCarta databases, were significantly higher in CL1 than CL2, suggesting a positive correlation with EGFR pathway activation (Fig. 6D). Consistent with our in vitro results, total EGFR expression did not significantly differ between clusters with different EGFR activation status (t test P value = 0.14; Fig. 6E) and exhibited a weak correlation with enrichment scores of EGFR downstream pathways (Pearson R < 0.4; Fig. 6F). In addition, there was no significant difference in FGF19 expression between the subgroups (t test P value = 0.30; Fig. 6E), indicating that FGF19 expression does not drive the EGFR pathway activation. To validate the stability of these clusters based on the TCGA database, we used RNA-seq data from an independent dataset, ICGC-LIRI (n = 232), which included 58 FGF19^hi patients. CL1 (n = 21) and CL2 (n = 25) revealed that the majority of the FGF19^hi patients exhibited high EGFR activation scores, while only 21% of the patients displayed relatively low EGFR activation signals (CL3, n = 12; Supplementary Fig. S4A–S4F). Furthermore, the enrichment scores of EGFR downstream pathways are positively associated with EGFR activation status among the three clusters (Supplementary Fig. S4H). These findings suggest that a significant proportion of FGF19^hi patients with HCC also possess high EGFR activation (CL1 in TCGA cohort; CL1 and CL2 in ICGC cohort), potentially elucidating why the majority of patients with HCC in previous clinical trials did not respond to treatment with FGFR4 inhibitors. Consequently, combination therapy involving EGFR-TKIs and FGFR4 inhibitors may offer additional clinical benefits.

Frequent occurrence of acquired resistance to targeted therapies is a major challenge for the development of effective cancer medicines. While FGFR4 inhibitors have achieved preliminary clinical efficacy in patients with HCC harboring FGF19 overexpression, the ORRs and durations of response remain limited, likely due to resistance or other reasons. To better understand the resistance mechanisms of FGFR4 inhibition, we carried out our studies by developing a set of cellular models with resistance to BLU-554, one of the most advanced and selective FGFR4 inhibitors in clinical trials. Huh7 is a human HCC cell line harboring FGF19 overexpression/amplification and is highly sensitive to BLU-554 inhibition in proliferation. Long-term treatment of the Huh7 cells to high concentration of BLU-554 successfully generated multiple resistant cell lines. These cell lines were insensitive to the treatment of BLU-554 or other FGFR4 inhibitors in inhibition of proliferation or FGFR4 downstream signaling pathways.

Acquired resistance to TKIs could arise through various mechanisms, such as mutations in the target gene at gatekeeper or other residues, or feedback activation of bypass signaling pathways. Previous study showed that FGFR4 gatekeeper and hinge-1 mutations were found in patients with HCC, who initially responded to BLU-554 but relapsed. The study also found similar FGFR4 mutations in resistant Baf3 cells expressing constitutively active FGFR4 (16). We used ENU to facilitate mutagenesis in generating Huh7 resistant cells. However, we did not uncover any gatekeeper or hinge mutations in FGFR4, suggesting the resistance may be mediated by other factors. We then used phospho-RTK array to screen for activation of bypass pathways and found that EGFR phosphorylation was significantly increased in all resistant clones. We further demonstrated that activation of EGFR and its downstream signaling indeed mediated the resistance to FGFR4 inhibition. Dual EGFR/FGFR4 inhibition overcame the resistance both in vitro and in vivo. In a phase I clinical trial of BLU-554, the ORR was ∼17% in FGF19-positive HCC with a median PFS of merely ∼3.3 months (6). Among these patients, on-target FGFR4 mutations were found only in 2 of 7 cases (29%), suggesting that other mechanisms may contribute to the resistance and relapses (16). Our findings identified EGFR activation as a novel resistance mechanism towards FGFR4 inhibition upon not only chronical treatment of FGFR4 inhibitor but also acute EGF stimulation. On the basis of our findings, dual targeting of FGFR4 and EGFR through combination therapies might bring improved therapeutic benefit to patients with HCC harboring FGF19 overexpression.

Compensating activation of bypass pathways has been frequently reported to render resistance to TKIs. FGFR and EGFR share common downstream signaling in MAPK and PI3K-AKT pathways. Activation of FGFR1–3 and EGFR have been found to mediate resistance of each other in many types of cancer. For example, enhanced expression of FGFR1 and FGF2 were reported as an escape mechanism for survival of afatinib-resistant cancer cells (31). FGFR1 and FGFR3 activation were found to cause EMT-mediated resistance in EGFR mutated NSCLC (13). FGFR3 overexpression was identified to be responsible for trastuzumab resistance in gastric cancer models. A phase II clinical study was therefore initiated to assess safety and activity of a FGFR inhibitor, pemigatinib, as a second-line treatment strategy in metastatic esophagogastric junction gastric cancer patients progressed under trastuzumab-containing therapies (32, 33). Reciprocally, EGFR reactivation could also compensate FGFR inhibition and drive resistance. In one study, EGFR activation was identified by parallel RNA interference as an escape mechanism in FGFR3-mutant cancer cells (34). Another recent study in FGFR2 fusion-positive cholangiocarcinoma found that EGFR inhibition also potentiated FGFR inhibitor therapy and overcame its resistance (35). In addition, several studies reported that EGFR contributed to resistance to lenvatinib. In cellular systems, EGFR activation limited the response of liver cancer to lenvatinib. Lenvatinib synergized with gefitinib both in vitro and in vivo. On the basis of these preclinical results, combination of lenvatinib and gefitinib was treated in twelve patients who had shown progression under prior lenvatinib treatment, resulting in four confirmed partial response (36). Furthermore, NGS analysis of ctDNA revealed that EGFR/HER2 amplifications and alterations were associated with resistance to lenvatinib in clinical setting (37). Lenvatinib is a multi-kinase inhibitor targeting VEGFR, PDGFR, KIT, RET, and FGFRs. Although lenvatinib is reported to target FGFR4 (38), its potency on FGFR4 is relatively weak, particularly in comparison with FGFR4 specific inhibitors. Mechanistic studies have clearly shown that lenvatinib synergizes with EGFR inhibitor mainly by abrogating FGFR1–3 signaling (36). Together, these results demonstrated frequent crosstalk and reciprocal relationships between EGFR and FGFR1–3, but the relationship between FGFR4 and EGFR has not been clearly addressed.

In this study, we for the first time demonstrated that EGFR activation could lead to resistance to specific FGFR4 inhibition. Surprisingly, we did not uncover any other RTKs with similar functions in our system. Downstream of EGFR activation, we also observed strong increases in phosphorylation of ERK1/2 and AKT in the resistant cells. Hence, we tested whether inhibition of these downstream signaling components, such as SHP2, SOS1, PI3K, mTOR, AKT, and MEK, would produce similar effects as EGFR inhibition. Indeed, either of the two tested SHP2 inhibitors was able to effectively restore BLU-554 sensitivity, while also leading to decreased phosphorylation of ERK1/2 and AKT. SOS1 inhibitor showed similar but weaker effects than SHP2 inhibitors, possibly due to the compensation from SOS2. Interestingly, either AKT or MEK inhibitor alone did not alter BLU-554 sensitivity significantly. Combination of both restored the sensitivity. These results suggested that both MAPK and PI3K-AKT pathways contributed to the EGFR-driven resistance of FGFR4 inhibition. A complete inhibition of both pathways will likely be required to provide sufficient efficacy in overcoming FGFR4 resistance.

To investigate the potential clinical relationship between the EGFR and FGFR4 pathways, we examined whether EGFR activation occurs in patients with HCC with FGF19 overexpression. Through our analysis of data from the TCGA and ICGC database, we observed that the expression of EGF or EGFR does not appear to be associated with the activation status of EGFR signaling. These findings align with the previous reports on HCC and colon cancers (39, 40). Consistent with our observations, we also discovered that the levels of phospho-EGFR, rather than of the expression of total EGFR protein and mRNA, were elevated in the resistant cell lines. Given the identification of EGFR downstream signaling activation was also found in our preclinical studies, we employed a set of EGFR-activation signatures to evaluate the status of EGFR pathway activation in these patients with HCC. Notably, several subgroups (CL1 in Fig. 6; CL1 and CL2 in Supplementary Fig. S4) exhibited significant enrichment of these signatures, suggesting the activation of the EGFR signaling and likely resistance towards FGFR4 inhibition. Therefore, dual inhibition of FGFR4 and EGFR may potentially improve the clinical outcome of these specific patient groups.

In conclusion, our findings demonstrated that EGFR activation led to resistance to FGFR4 inhibition in HCC (Supplementary Fig. S5). Several FGFR4 inhibitors have advanced into phase I/II clinical trials and shown initial efficacy. However, the ORRs were less than 20% and the durations of response were relatively short. It is likely that other factors may compensate inhibition of FGF19-FGFR4 pathway and lead to resistance in the refractory or relapsed patients. Our results indicated that EGFR activation could be one of the main resistance mechanisms. Therefore, combination therapies of FGFR4 and EGFR inhibitors may provide synergistic and complete inhibition of both targets and their downstream signaling pathways, potentially bringing additional clinical benefit to patients with HCC harboring FGF19^hi.

No disclosures were reported.

B. Shen: Formal analysis, investigation, writing–original draft. J.P. Shi: Formal analysis, investigation. Z.X. Zhu: Data curation, formal analysis, writing–original draft. Z.D. He: Formal analysis, validation. S.Y. Liu: Formal analysis, validation. W. Shi: Data curation, formal analysis. Y.X. Zhang: Methodology. H.Y. Ying: Methodology. J. Wang: Methodology. R.F. Xu: Methodology. F. Fang: Methodology. H.X. Chang: Writing–review and editing. Z. Chen: Conceptualization, supervision, writing–review and editing. N.N. Zhang: Conceptualization, supervision, writing–original draft.

We appreciate the generosity of ETERN Therapeutics, Shanghai, China, and Dr. Jidong Zhu for kindly providing the SHP2 inhibitor ET070. We wish to thank Dr. Shuqun Yang and Dr. Mei Ning for advice on experimental design and manuscript revision.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.