Comprehensive functional evaluation of variants of fibroblast growth factor receptor genes in cancer

Abstract

Various genetic alterations of the fibroblast growth factor receptor (FGFR) family have been detected across a wide range of cancers. However, inhibition of FGFR signaling by kinase inhibitors demonstrated limited clinical effectiveness. Herein, we evaluated the transforming activity and sensitivity of 160 nonsynonymous FGFR mutations and ten fusion genes to seven FGFR tyrosine kinase inhibitors (TKI) using the mixed-all-nominated-in-one (MANO) method, a high-throughput functional assay. The oncogenicity of 71 mutants was newly discovered in this study. The FGFR TKIs showed anti-proliferative activities against the wild-type FGFRs and their fusions, while several hotspot mutants were relatively resistant to those TKIs. The drug sensitivities assessed with the MANO method were well concordant with those evaluated using in vitro and in vivo assays. Comprehensive analysis of published FGFR structures revealed a possible mechanism through which oncogenic FGFR mutations reduce sensitivity to TKIs. It was further revealed that recurrent compound mutations within FGFRs affect the transforming potential and TKI-sensitivity of corresponding kinases. In conclusion, our study suggests the importance of selecting suitable inhibitors against individual FGFR variants. Moreover, it reveals the necessity to develop next-generation FGFR inhibitors, which are effective against all oncogenic FGFR variants.

Fig. 1. The structure of FGFR protein and the distribution of FGFR mutations.

a FGFRs consist of an extracellular domain with three immunoglobulin-like domains (IgI, IgII, and IgIII), followed by a transmembrane domain and tyrosine kinase domain. An acid box (AB), which regulates FGFR interaction with partners except fibroblast growth factors (FGFs), is located between IgI and IgII. The structures of FGFR1/2/3/4 are highly conserved. b Mutational hotspots in FGFRs were observed across various types of cancer. The number of each cell indicates sample count reported in the COSMIC database. Clinical significance annotated in OncoKB and ClinVar databases were described at the bottom of the charts. The top ten major primary tissues and mutations reported in at least four samples are shown. c Structures of FGFR fusion proteins. FGFR fusions are classified into two types: type 1 fusion is found in hematological malignancies encoding non-transmembrane-type FGFR kinases with N-terminal substitution of fusion partner genes; type 2 fusion is common in solid tumors encoding transmembrane-type FGFRs with C-terminal substitution of fusion partner genes. Ig immunoglobulin-like domain, AB acid box, LB ligand-biding region, TM transmembrane domain, AA amino acid, CNS central nervous system, UAT upper autodigestive tract. d Brief overview of the MANO method. 3T3 mouse fibroblasts are infected with recombinant retrovirus expressing FGFR variants with individual 10-bp bar codes. Equal numbers of the stably transduced cells were mixed and cultured with different types of medium and/or treated with TKIs or vehicle. gDNA was harvested from the mixture of the remaining viable cells at the appropriate periods for each assay. Bar code sequences were PCR-amplified and then analyzed through deep sequencing using MiSeq sequencers to quantitate their relative abundance as a direct reflection of cell number. The read number for each bar code was normalized and compared with each other to assess transforming potential and drug sensitivity. FBS fetal bovine serum, BS bovine calf serum.

Fig. 2. Transforming activity of FGFR variants.

a Fold changes from day 3 to day 18 of 3T3 cells harboring each FGFR variant in the mixed cell population were calculated using the MANO method and shown as relative proliferation to wild-type on a base 2 logarithmic scale. Relative proliferation significantly different from those of wild-type (#) or GFP (*) are shown (paired t-test, p < 0.05). The color of each bar indicates the TAS, which are integrated assessments based on the results of the focus formation assay and low-serum cell proliferation assay. The bars are sorted according to the position of amino acids. b The results of the focus formation assay and growth competition assay by the MANO method were summarized to annotate the oncogenicity of variants according to the classification described in the method. The oncogenicity evaluated by the method is shown in colors and compared with the variant count number of COSMIC or AACR Project GENIE. c The oncogenicity evaluated by the method is shown in colors and compared with that of OncoKB. The OncoKB annotations are shown at the top of the pie chart. WT wild-type, EC extracellular domain, TM transmembrane domain, TK tyrosine kinase domain; error bars, SD.

Fig. 3. The sensitivity of FGFR variants to FGFR-targeted drugs.

a 3T3 cells expressing FGFR variants, GFP, EGFR L858R, and KRAS G12V were treated with DMSO or FGFR-targeted drugs (AZD4547, infigratinib, E7090, erdafitinib, futibatinib, pemigatinib, and dovitinib) at the indicated concentrations. The relative viability of the treated cells with each drug versus DMSO-treated cells was measured, and the results are illustrated using the color-coded scale. b 3T3 cells with FGFR2 variants or KRAS G12V were incubated with the indicated concentrations of inhibitors for 5 days. Cell viability was measured using the PrestoBlue cell viability assay and plotted relative to the untreated control. Data are presented as mean ± SD (n = 6). The IC₅₀ values of inhibitors are shown in the table at the bottom. The p-value in the comparison of IC₅₀ between each two variants is shown in Supplementary Data 2. c Inhibition of phosphorylation of FGFR and downstream pathway by E7090 and erdafitinib. 3T3 cells with FGFR variants were treated with E7090 or erdafitinib for 2 h, and whole-cell lysate were analyzed by western blotting using antibodies against FGFR2, FGFR3, phospho-FGFR (p-FGFR, Tyr653/654), p44/42 MAPK, phospho-p44/42 MAPK (p-MAPK, Thr202/204), and β-Actin. WT wild-type, EC extracellular domain, TM transmembrane domain, TK tyrosine kinase domain.

Fig. 4. The IC₅₀ values of FGFR-targeted drugs estimated using the MANO method.

The IC₅₀ values of FGFR-targeted drugs against FGFR variants were evaluated using the drug sensitivity assay of the MANO method in 3T3 cells.

Fig. 5. Inhibition of tumor growth in vivo by FGFR-targeted drugs.

3T3 mouse fibroblasts expressing FGFR variants were subcutaneously injected into 6-week-old female nude mice. The mice were treated with erdafitinib (12.5 mg/kg body weight), E7090 (25 mg/kg body weight), or vehicle control once daily by oral gavage (n = 5 mice for each group). ^#p < 0.05 vs. vehicle; ^†p < 0.05 vs. erdafitinib; error bars, SD.

Fig. 6. FGFR structural analysis.

a Comparison of FGFR2 and FGFR3 kinase domain structures bearing N549H/T, K659E/N/M/Q/T, K650E, or K656E activating mutations (shades of green/teal) with wild-type FGFR1 in the active conformation (magenta). Autophosphorylated Tyr residues in the activation loop are shown in spheres. All mutant crystal structures were determined in the apo state or in the presence of an ATP analog (ball-and-stick model shown in center). Superimposed structures are from Protein Data Bank (PDB) entries 3GQI, 4K33, 2PWL, 5UHN, 2PZ5, 4J97, 5EG3, 4J98, 2PVY, 4J95, 4J99, 4J96, and 5UI0. b Comparison of four inhibitor-bound FGFR1 structures (shades of blue) with the structure of FGFR1 in the active conformation (magenta). In the inhibitor-bound structures, a ~10° rotation in the N-lobe was observed compared with its orientation in the active state, consistent with the inhibitor-bound structures being in the inactive, “open” conformation. The common site of mutation K656 in the activation loop also changes conformation in the inhibitor-bound structures. All structures are aligned to the C-lobe of the kinase. PDBs 3GQI, 4V05, 3TT0, 4TYI, and 5VND.

Fig. 7. Functional analysis of FGFR compound mutations.

a The frequency and patterns of FGFR compound mutations were investigated in the hotspot variants of FGFRs in the COSMIC database. b The drug sensitivity of FGFR compound mutations was evaluated with the PrestoBlue cell viability assay. 3T3 cells expressing a FGFR single mutation and compound mutations were treated with the indicated concentrations of E7090 or erdafitinib for 5 days. Cell viability was measured using the PrestoBlue cell viability assay and plotted relative to the untreated controls. Data are presented as the mean ± SD (n = 6). Estimated IC₅₀ values are shown in the tables under the dose response curves. error bars, SD.