6-Amino-2,4,5-trimethylpyridin-3-ol and 2-amino-4,6-dimethylpyrimidin-5-ol derivatives as selective fibroblast growth factor receptor 4 inhibitors: design, synthesis, molecular docking, and anti-hepatocellular carcinoma efficacy evaluation

Abstract A novel series of aminotrimethylpyridinol and aminodimethylpyrimidinol derivatives were designed and synthesised for FGFR4 inhibitors. Structure-activity relationship on the FGFR4 inhibitory activity of the new compounds was clearly elucidated by an intensive molecular docking study. Anti-cancer activity of the compounds was evaluated using hepatocellular carcinoma (HCC) cell lines and a chick chorioallantoic membrane (CAM) tumour model. Compound 6O showed FGFR4 inhibitory activity over FGFR1 − 3. Compared to the positive control BLU9931, compound 6O exhibited at least 8 times higher FGFR4 selectivity. Strong anti-proliferative activity of compound 6O was observed against Hep3B, an HCC cell line which was a much more sensitive cell line to BLU9931. In vivo anti-tumour activity of compound 6O against Hep3B-xenografted CAM tumour model was almost similar to BLU9931. Overall, compound 6O, a novel derivative of aminodimethylpyrimidinol, was a selective FGFR4 kinase inhibitor blocking HCC tumour growth.


Introduction
Hepatocellular carcinoma (HCC) has been increasingly diagnosed over the past few decades and has become the third leading cause of cancer-related mortality worldwide 1 . After sorafenib (a multi-tyrosine kinase inhibitor) was authorised for the first-line treatment option of HCC treatment 2 , several multi-tyrosine kinase inhibitors including cabozantinib have been approved for patients with advanced-stage HCC 3 . In addition, PD-1 immune checkpoint inhibitors, such as nivolumab and pembrolizumab, have been added as therapeutic options in HCC treatment 3 . Despite these drugs being approved as the first-line or second-line therapy, tumour regression by immune checkpoint blockade only occurs in a limited proportion of patients 4 .
Fibroblast growth factor receptor 4 (FGFR4), one of the families of fibroblast growth factor receptors, is a tyrosine kinase receptor with a distinct 802 amino acid sequence. Normally, FGFR4 expression level is high during foetal development and drastically reduced thereafter 5 . Most of FGF family members, except FGF11 subfamily, are ligands of FGFR4 6 . FGF19, which is produced from the ileum as a postprandial hormone, has a more specific selective affinity to FGFR4 than the other FGF ligands. In contrast to the normal production and action of FGF19 as an endocrine hormone, FGF19 is overexpressed and co-expressed with FGFR4 in various cancers of liver, breast, lung, bladder, head, and neck 7-10 , indicating FGF19 as a driver oncogene. Similarly, amplification of FGFR4 gene is the predominant type and accounts for 78% of all FGFR4 gene alterations in various cancers [11][12][13][14] . In addition, a significant correlation between overexpression of FGFR4 in tumour tissues and patients' poor survival rate indicates FGFR4 is also acting as an oncogene 15,16 .
Numbers of research for the discovery of selective FGFR4 inhibitors have underlined the covalent interaction between Cys552 residue in the middle-hinge domain of the ATP-binding site of FGFR4 and acrylamide moiety of small molecule compounds, which was supported by X-ray co-crystal structures. BLU9931 (1) was discovered as a notable compound to possess potent FGFR4 inhibitory activity with good selectivity over FGFR1 $ 3. After that, a number of analogues of BLU9931 were designed and prepared, where fisogatinib (BLU554) (2) with better physicochemical properties has been evaluated in a clinical trial for the treatment of HCC. In addition, other Cys552 À acrylamide inhibitors such as H3B-6527 (3), aminopyrimidines (4), and aminopyridines (5) are undergoing either preclinical or clinical studies in recent years ( Figure 1) [17][18][19][20][21][22][23][24][25][26] . Multiple pan-FGFR inhibitors have shown weak efficacy against FGFR4 together with toxicity (hyperphosphatemia and soft tissue mineralization) due to activity against FGFR1 $ 3, which indicates a strong need for the discovery of FGFR4-selective inhibitors 27 .
We recently reported a string of anti-HCC compounds where bioisosterism was employed as the main strategy for the design of new compounds 28,29 . Bioisosterism is one of the efficient and practical strategies of new drug discovery. Replacement of a functional group of a lead compound may improve efficacy by changes in binding mode to molecular target, metabolic stability by modifying structure of labile sites, or physicochemical properties, etc 30,31 . In our previous studies, we designed and prepared a series of bioisosteres of sorafenib and cabozantinib for the discovery of anti-HCC agents, respectively 28,29 . As part of our on-going project of the discovery of new drug by bioisosteric replacement of a benzene, pyridine, or pyrimidine core ring of well-known drugs or promising compounds in the development stage with methyl-containing pyridine/pyrimidine rings, the present study was focussed on the synthesis and evaluation of a new series of aminotrimethylpyridinol-and aminodimethylpyrimidinol-derived FGFR4 inhibitors which are depicted as a general structure (6) in Figure 1. In addition, a molecular docking study was done indepth for an explanation of the difference in FGFR4 inhibitory activity with structural change.

Results and discussion
2.1. Synthesis Figure 2 shows the structures of twenty-eight new compounds for this study. They were designed based on common aminopyridine/ aminopyrimidine backbones of about hundred compounds such as compounds 4 and 5 in Figure 1 introduced in the literature [21][22][23] . The pyridine/pyrimidine centred rings were replaced by trimethylpyridine/dimethylpyrimidine, which was the major structural variation. Fluorine and chlorine were employed at the halogen position in the dimethoxy ring part and seven kinds of acryl amide-containing aromatic rings were combined.
Preparation of the required substrates 7 $ 10 is shown in Scheme 2. First, the fluoro-containing benzyl bromide 7 was readily prepared by a simple bromination from commercially available alcohol 22 21 . For the synthesis of the chloro-version substrate 8, we put the ringchlorination step at the final stage. Reduction of methyl 3,5-dimethoxybenzoate (23) to primary alcohol 24 followed by bromination afforded the corresponding bromide 25 32 . Lastly, chlorination of the ring with sulphuryl chloride gave the chloro-substrate 8. The centred ring, 2,4,5-trimethylpyridin-3-ol (9), was prepared from pyridoxine (26) by the known procedure developed by us 33 . The two hydroxymethyl groups of 26 were chlorinated to give 27 which was then reductively cleaved to afford methyl compound 28. Ring bromination of 28 finally gave the substrate 9. Another centred ring, 2-bromo-4,6dimethylpyrimidin-5-ol (10), was obtained by a three-step sequence. 3-Chloropentane-2,4-dione (29) was reacted with formamide in formic acid to give an oxazole intermediate 30 which was then treated with ammonia water to afford 4,6-dimethylpyrimidin-5-ol (31) 34 . The pyrimidine substrate 10 was finally obtained by bromination of 31.

Inhibitory activity of new compounds against FGFR kinases
The new compounds (6A$6BB) were subjected to cell-free kinase assay to examine their inhibitory activities against FGFR4 kinase at a fixed concentration (1 mM) along with BLU9931 (1), the positive control. Most compounds generally showed low inhibitory activity against FGFR4 ( Figure 3). Roughly, the inhibitory activity of the difluoro-containing compounds was better than the corresponding dichloro analgoues. The inhibitory activity of compounds 6A, 6O, and 6S was notable, and, in particular, compound 6O showed an excellent inhibition comparable to that of BLU9931 at the 1 mM concentration tested.
A follow-up kinase assay of selected compounds, 6A and 6O, with 7 different concentrations, showed inhibition of the target kinase FGFR4 with IC 50 values of 190 nM and 75.3 nM, respectively (

Structure-activity relationship (SAR) based on molecular docking
As shown in Figure 3, most of the trimethylpyridine-or dimethylpyrimidine-containing new compounds generally have low inhibitory activity against FGFR4 except for compounds 6A, 6O, and 6S, which have fluorine in common at X position in Figure 2. We thought that a combination of fluorine substitution at the dimethoxyphenyl ring and the introduction of methyl groups to the pyridine or pyrimidine ring may contribute to the higher FGFR4 activity and selectivity of 6A and 6O as shown in Table 1. To clarify these substituent effects, we focussed on compounds 6E and 6G. These compounds are methylated analogs of known FGFR4 inhibitors 41 and 42, which have bare pyridine and pyrimidine at the centre 21 . The FGFR4 percent inhibition data at 0.01 mM of these compounds are shown in Table 2. It is clear that the substitution of methyl groups at the pyridine and pyrimidine rings is detrimental to the activity. In order to further investigate the substitution effects of fluorines and methyl groups on the FGFR4 activity and selectivity, we performed covalent docking studies on relevant compounds.

Effect of methyl groups and halogen substituents on SAR
Covalent docking studies were carried out using Flare from Cresset according to the methods described in detail in the experimental section. The LF Rank score and LF VSscore were used to identify correct binding pose and rank compounds in a virtual screening context, respectively. Introducing three methyl groups into the pyridine core (compound 41 ! 6E) cost LF VSscore of 0.908. As shown in Figure 4(a,b), this is a direct result of losing hydrogen bonds from Ala553 to the pyridine N and anilyl NH. Due to steric bulkiness of three methyl groups in compound 6E, the pyridine ring plane is twisted by 90 and thus the hydrogen bonds are destroyed. On the other hand, introducing only two methyl groups into the pyrimidine ring is generally tolerated (compound 6O). Unlike compound 6E, it does not cause a twist of the pyrimidine ring plane (Figure 4(c)). However, replacing the fluorine atoms of the dimethoxyphenyl ring with chlorine (6O ! 6P) abolished the FGFR4 inhibitory activity (Figure 4(d)). To find out the origin of the activity change, we replaced hydrogens of the pyrimidine ring with methyl in the crystal structure of FGFR4 À ligand complex (PDB 7DTZ) and examined the binding interactions ( Figure 4(e)). It turned out that the methyl group would clash with the peptide carbonyl oxygen of Glu551 and thus the ideal binding pose cannot be maintained as shown in Figure  4(d). LF VSscore also changed from À12.23 to À10.81 in agreement with the observed FGFR4 activities of 6O and 6P. Interestingly, removing methyl groups from the pyrimidine ring of 6P restores the FGFR4 activity, i.e., it is known that compound 6P-1 exhibits an FGFR4 activity of IC 50 2.6 nM 37 . Figure 4(f) shows the X-ray co-crystal structure of FGFR4À6P-1 complex (PDB 6NVI), in which, unlike 6P (Figure 4(d)), one of the pyrimidine nitrogen atoms forms a hydrogen bond to Ala553. Therefore it can be concluded that both the small size of fluorine at the dimethoxyphenyl ring and the bulkiness of methyl groups at the pyrimidine ring amount to the observed FGFR4 activity and selectivity of compound 6O.

Selective inhibition of FGFR4 by compound 6O over FGFR1 À 3
Compound 6O showed an excellent selectivity towards FGFR4 over other FGFR subtypes (FGFR1 $ 3), whereas compound 6A had a marginal selectivity (Table 1). Although detailed explanation for the selectivity profile of BLU554 for the hinge cysteine of FGFR4 has been reported recently 37 , more rationales are needed to address the exceptionally high selectivity profile of compound 6O. We further conducted a molecular docking study to provide an answer to such a high degree of selectivity by 6O. FGFR4 has a cysteine residue available for covalent binding of ligand in the hinge region, while other FGFR subtypes have one on the P-loop. X-ray crystal structures are found for FGFR À ligand complexes with a covalent bond at both the hinge and P-loop cysteine residues. It was proposed that the FGFR selectivity of BLU554 mainly comes from the rotational energy barriers of the ligand in solution and steric clash between the tetrahydropyran ring and the P-loop of FGFR1, specifically Leu484 37 . Although compound 6O does not have a tetrahydropyran-like fragment that would clash with the Ploop of FGFR1, the methyl groups in the central pyrimidine ring would be too close to Glu562 at the hinge, hindering the binding ( Figure 5). Therefore we propose that the outstanding FGFR4 selectivity of 6O originates from the steric clash of the pyrimidine methyl groups to the hinge loop.   compound 6A was fairly safe, and cytotoxicity of 6O and 1 was similar.
Next, we examined the in vivo antitumor efficacy of compound 6O using a chick chorioallantoic membrane (CAM) tumour model implanted with Hep3B cells. The implanted Hep3B cells in the vehicle-treated control group developed a tumour mass and tumour-induced angiogenesis (Figure 7). Compound 6O and positive control compound (BLU9931, 1) dose-dependently inhibited the tumour growth. The inhibitory effect of 6O on the tumour weight was slightly weaker than that of compound 1. The tumour-induced angiogenesis was also significantly blocked by compounds, 6O and 1, and the antiangiogenic effect of compound 6O was also slightly weaker than compound 1.

Conclusion
A series of novel compounds containing aminotrimethylpyri(mi)dinol were designed and synthesised with the aim of the discovery of selective FGFR4 inhibitors. Among them, compound 6O showed a potent and highly selective inhibitory activity against FGFR4 over FGFR1 $ 3. Structure-activity relationship study using intensive molecular docking calculations revealed that, by and large, introducing methyl groups at the central pyridine or pyrimidine ring can cause a steric clash with the hinge of FGFR4 kinase domain and thus weaken the binding affinity. However, exceptionally, compound 6O with dimethylpyrimidine core ring and difluoro substituent was observed to maintain its conformation suitable for strong binding interaction. Molecular docking study also suggests that a steric clash of dimethyl groups of compound 6O to a hinge of FGFR1 $ 3 interfere with adopting a proper pose for a strong binding between them. This explains the highly selective inhibition of compound 6O against FGFR4 over other families. The anti-proliferative activity of compound 6O on HCC cell lines was a little weaker than BLU9931 at low concentrations, however, the strength of the inhibitory effect was reversed at high concentrations. In addition, compound 6O effectively inhibited both the growth and angiogenesis of HCC, one of the hypervascular tumours. Overall, the antitumor effect of compound 6O was excellent and comparable to that of BLU9931.

General
Unless noted otherwise, materials were purchased from commercial suppliers and used without further purification. Air or moisture-sensitive reactions were carried out under argon atmosphere. The reaction progress was monitored by thin layer-chromatography (TLC) using silica gel F 254 plates. Products were purified by flash column chromatography using silica gel 60 (70-230 mesh) or by using the Biotage 'Isolera One' system with indicated solvents. Melting points were determined using a Fisher-Johns melting point apparatus and were not corrected. Low-resolution mass spectra (LRMS) were obtained using a JMS-700 (JEOL) and recorded either in molecular ion peak mode with an electron ionisation (EI) source or in positive ion mode with fast electron bombardment (FAB) source or using a Waters ZQ 2000 and recorded in a positive ion mode with an electrospray (ESI) source. High-resolution mass spectra (HRMS) were obtained using a JMS-700 (JEOL) and recorded either in molecular ion peak mode with an electron ionisation (EI) source or in positive ion mode with fast electron bombardment (FAB) source. NMR spectra were obtained using a Bruker-250 spectrometer (250 MHz for 1 H NMR and 62.5 MHz for 13 C NMR) and a Bruker Avance Neo 400 spectrometer (400 MHz for 1 H NMR). Chemical shifts (d) were expressed in ppm using a solvent as an internal standard and the coupling constant (J) in hertz.

4,6-Dimethylpyrimidin-5-ol (31) [CAS RN: 70345-38-9]
To a solution of 3-chloro-2,4-pentadione (29) (20 ml, 177.17 mmol) in formic acid (30 ml) was added formamide (16 ml, 402.84 mmol) and refluxed for 12 h. The reaction mixture was cooled to room temperature then aq. NH 4 OH was added dropwise until the basic pH was maintained. After refluxing the mixture for 5 h, it was cooled to room temperature and then concentrated. Acetone was added to the residue to acquire product in acetone. The acetone fraction was concentrated, and the residue was purified by silica gel column chromatography with EtOAc to obtain nd 31 (7.4 g, 34%) as a pale yellow solid.

2-Bromo-4,6-dimethylpyrimidin-5-ol (10)
To a solution of 4,6-dimethylpyrimidin-5-ol (31) (150 mg, 1.21 mmol) in THF (4 ml) was added DBDMH (242 mg, 0.85 mmol). The resulting mixture was stirred at room temperature for 5 h and then concentrated. The residue was diluted with EtOAc and water, and the aqueous layer was extracted with EtOAc. The combined organic solution was dried over MgSO 4 , filtered, and concentrated. The residue was purified by silica gel column chromatography with 10% to 40% EtOAc/Hexanes to obtain 10 (156 mg, 63%   wise at 0 C. The mixture was stirred at 0 C for 3 h and then it was poured into ice water and stirred at room temperature for 12 h. The mixture was neutralised using a solution of 6 M NaOH followed by EtOAc extraction. The combined organic solution was dried over MgSO 4 , filtered, and concentrated. After addition of an excessive amount of 1 M HCl, the mixture was refluxed for 5 h and then neutralised using a solution of 6 M NaOH followed by EtOAc extraction. The combined organic solution was dried over MgSO 4 , filtered, and concentrated. The residue was purified by silica gel column chromatography with 1% to 5% EtOAc/Hexanes to obtain 18 (596 mg, 59%) as a yellow solid. addition of a 3% phosphoric acid solution. Then, the reaction was spotted onto a P30 filtermat, and unbound phosphate was removed by washing 3 times for 5 min in 75 mM phosphoric acid and once in methanol prior to drying and scintillation counting. The background counting derived from the control reactions containing the inactive enzyme was subtracted, and specific kinase activity data were expressed as the percentage of remaining kinase activity in the test compounds compared to the vehicle (dimethyl sulfoxide) reactions. IC 50 values and curve fits were obtained using GraphPad Prism 5 software. FGFR1 $ 3 kinase assays were performed using the adenosine diphosphate (ADP)-glow kinase assay kit of FGFR1, FGFR2, and FGFR3 kinase enzyme systems (Promega, WI, USA) in accordance with the manufacturer's instructions. Kinase activity was detected by the addition 50 mM adenosine triphosphate (ATP) to a mixture of 0.2 mg/mL poly (Glu4, Tyr1), test drug (1, 6A, 6O), and corresponding enzyme, FGFR1 (1.5 ng/mL), FGFR2 (3 ng/mL), or FGFR3 (6 ng/mL). The reaction was carried out at 25 C for 1 h in a total volume of 25 mL. Subsequently, 25 mL of ADP-Glow reagent was added to the mixture, and then incubated for 40 min at 25 C. After the addition of 50 mL of ADP detection reagents for 30 min at 25 C, luminescence was measured using a Fluostar Omega microplate reader (BMG LABTECH GmbH, Ortenberg, Germany).

Cell lines and culture
Human HCC cell lines, Hep3B and Huh7 were obtained from American Type Culture Collection (ATCC, Manassas, VA, USA). An H6c7 human normal pancreatic duct epithelial cell line was purchased from Kerafast (Boston, MA, USA). Hep3B and Huh7 cells were cultured in Dulbecco 0 s Modified Eagle 0 s Medium (DMEM) (Hyclone, Logan, UT, USA). The media were supplemented with 10% foetal bovine serum (FBS) (Gibco/ThermoFisher Scientific) and 1% penicillin/streptomycin (Gibco/ThermoFisher Scientific). H6c7 cells were maintained in a keratinocyte serum-free medium supplemented with a recombinant endothelial growth factor (rEGF) and bovine pituitary extract (Gibco/ThermoFisher Scientific). All the cells were incubated at 37 C under a 5% CO 2 atmosphere.

Cell proliferation assay
Cells were seeded at a density of 5000 cells/well in a 96-well plate. After overnight incubation, the cells were serum-starved using 1% FBS for 24 h. The next day, the cells were pre-treated with the different concentrations (0.1, 1, 3, 10, 30, 100 mM) of drugs for 1 h prior to the treatment with serum (10% FBS). After 48 h of incubation, 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) dye solution was added and incubated for 4 h. Next, DMSO was added, and after 30 min, the colour intensities were measured using a microplate reader (Versamax, Molecular Devices, Inc., USA) at 490 nm.

Cytotoxicity assay
The cytotoxicity of the compounds was assessed by measuring the cell viability decrease using the MTT staining method. Briefly, H6c7 cells were seeded in a 96-well plate (Falcon, USA) at a density of 4 Â 10 4 cells per well in keratinocyte serum-free medium. After 24 h, Cells were incubated with different concentrations (0.1, 1, 3, 10, 30, and 100) mM of each compound for 48 h, and the cell viability was measured using MTT assay. Optical density was measured at 540 nm using a microplate reader (BMG LABTECH). 4.2.5. Cam (chick chorioallantoic membrane) tumour model Fertile chicken eggs were purchased from Byeolbichon Farm (Gyeongbuk, South Korea) and incubated at 37 C and under 55% relative humidity. On the 9th day of egg incubation (post-fertilization), false air sac was generated on the relatively flat side of eggs by a negative pressure technique. A small window (1 cm 2 ) was created on the false air sac surface of the eggs by separating the shell and membrane beneath (technique) using a grinding wheel (Dremel, Racine, WI, USA). Next, Hep3B cells were loaded at a density of 1.5 Â 10 6 cells/CAM with or without compound. After four days of drug treatment, the tumour weight, number of vessel branch points within the tumour region were analysed.
The chick embryo experiments were approved beforehand by the Institutional Animal Care and Use Committee of Yeungnam University and were performed accordingly the guidelines issued by the Institute of Laboratory Animal Resources (1996) and Yeungnam University (The care and use of animals 2009).

Statistical analysis
The results are presented as the mean ± SEM and were analysed using one-way ANOVA followed by the Newman-Keuls comparison method using the GraphPad Prism software (version 5.0) (San Diego, CA, USA). P values less than 0.05 were considered statistically significant.

Computational study
Flare version 5.0 from Cresset (http://www.cresset-group.com/ flare/) was used for covalent docking for this study. The crystal structure of the complex of FGFR4 and a 2-aminopyridine-based inhibitor (PDB ID 7DTZ) was chosen for docking owing to its high crystal resolution (2.01 Å) and the possession of strong hydrogen bonding interactions between the guanidyl group of Arg483 and the carbonyl oxygen of the acrylamide moiety of the ligand, which would help keep the ligands within the binding cavity in a proper orientation during the docking. Docking parameters were set up to use Cys552 as the covalent bonding residue and to use "Very Accurate but Slow" as the calculation method. For postdocking analysis, LF Rank Score was used to identify correct binding pose, while LF VSscore was used to rank compounds in a virtual screening context 41 .
Procedures for the synthesis of compound 6O are provided here. Synthetic procedures for all compounds are described in Supplementary Material.

Disclosure statement
No potential conflict of interest was reported by the author(s).