Synthesis, biological evaluation and molecular modelling of 2,4-disubstituted-5-(6-alkylpyridin-2-yl)-1H-imidazoles as ALK5 inhibitors

Abstract A series of 2,4-disubstituted-5-(6-alkylpyridin-2-yl)-1H-imidazoles, 7a–c, 11a–h, and 16a–h has been synthesised and evaluated for their ALK5 inhibitory activity in an enzyme assay and in a cell-based luciferase reporter assay. Incorporation of a quinoxalin-6-yl moiety and a methylene linker at the 4- and 2-position of the imidazole ring, respectively, and a m-CONH2 substituent in the phenyl ring generated a highly potent and selective ALK5 inhibitor 11e. Docking model of ALK5 in complex with 11e showed that it fitted well in the ATP-binding pocket with favourable interactions.


Introduction
Transforming growth factor-b (TGF-b) is one of the most potent immunosuppressive cytokines in the tumour microenvironment 1 . Elevated serum levels of TGF-b commonly observed in patients with advanced colorectal cancer 2 , breast cancer 3,4 , bladder carcinoma 5,6 , prostate cancer 7,8 , malignant melanoma 9 , pancreatic ductal adenocarcinoma 10 , and hepatocellular carcinoma 11 have been strongly associated with tumour progression and poor clinical outcome. Overexpression of TGF-b receptors has been implicated in cancer 12 . In patients with advanced hepatocellular carcinoma treated with the first clinically available ALK5 inhibitor, galunisertib (1) 13 , an approximately two-fold longer overall survival was observed in patients having a TGF-b1 response compared to patients who did not have a TGF-b1 response 14 . Therefore, TGF-b signalling pathway is an attractive target for development of cancer immunotherapeutic agents.
Vactosertib exhibited subnanomolar ALK5 inhibitory activity in a kinase assay and in a cell-based luciferase reporter assay 15 , high selectivity against a panel of 320 protein kinases including p38a 15 , moderate oral bioavailability in rats 15 , and high efficacy in animal models of cancer [19][20][21] and fibrosis [22][23][24][25][26] . In this report, we examined whether structural modification of vactosertib could increase its subnanomolar ALK5 inhibitory activity, thus further increasing its selectivity. For this purpose, we replaced a [1,2,4]triazolo [1,5a]pyridin-6-yl moiety of vactosertib with either a benzo [1,3]dioxol-5-yl or a quinoxalin-6-yl moiety and inserted either a methylene, an ethylene, or a propylene linker instead of a methyleneamino linker to optimise the distance between a central imidazole ring and a phenyl ring.

Chemistry
1 H NMR spectra were recorded on a Varian Unity-Inova 400 MHz instrument. The chemical shifts are reported in parts per million (ppm). For 1 H NMR spectra, CDCl 3 was used as solvent, and it served as the internal standard at d 7.26. Infra-red spectra were recorded on a FT-infra-red spectrometer (Bio-Rad). Electrospray ionisation mass spectra (ESIMS) were obtained on a Q-Tof2 mass spectrometer (Micromass). Elemental analyses (C, H, and N) were used to determine the purity of all tested compounds, and the results were within ±0.4% of the calculated values (Carlo Erba 1106 elemental analyzer). Analytical thin-layer chromatography (TLC) was performed on Merck silica gel 60 F-254 glass plates. Medium-pressure liquid chromatography (MPLC) was performed using Merck silica gel 60 (230-400 mesh) with a YFLC-540 ceramic pump (Yamagen).
2.1.6. General procedure for the preparation of the 5-(6-alkylpyridin-2-yl)-4-(quinoxalin-6-yl)-1H-imidazoles 16a-h To a stirred solution of 15a-d (3.79 mmol) in a mixture of t-BuOMe (35 mL) and MeOH (25 mL) were added either 14a or 14b (5.69 mmol) and NH 4 OAc (18.95 mmol), and the mixture was stirred at 30 C overnight under argon atmosphere. The pH of the reaction mixture was adjusted to pH$8 at 0 C with saturated NaHCO 3 solution. After removal of solvent, the reaction mixture was extracted with CH 2 Cl 2 (25 mL Â 3), and the organic solution was dried over anhydrous Na 2 SO 4 , filtered and evaporated to dryness under reduced pressure. The residue was purified by MPLC on silica gel with CH 2 Cl 2 /MeOH as eluent to afford the titled compounds 16a-h as a solid.

Luciferase reporter assay
To establish HaCaT (3TP-luc) stable cells, cells were seeded on sixwell plates. Cells were allowed to adhere overnight and then transfected with the p3TP-luc (neo) expression plasmid using PEI reagent (Sigma Aldrich). Transfected cells were cultured for four weeks in the presence of G418 (500 mg/mL). Several single clones were isolated and measured luciferase activity. The clone showing response to TGF-b1 treatment was used for reporter assay. HaCaT (3TP-luc) stable cells were seeded at 2.5 Â 10 4 cells/well in 96-well plate and were allowed to adhere overnight. Cells were concomitantly treated with TGF-b1 (2 ng/mL) and indicated concentrations of ALK5 inhibitors in 0.2% FBS medium and incubated for 24 h at 37 C in 5% CO 2 . Cell lysates were prepared using Luciferase Assay System (Promega) according to the manufacturer's instruction, and luminescence was measured by a luminometer, Micro Lumat Plus (Berthold, Germany).

Cell permeability assay
Caco-2 cells were seeded in Transwell V R polycarbonate filter at a density of 8 Â 10 4 cells/filter and cultured for 21 days. Culture medium was removed from both apical (AP) and basolateral (BL) chambers of transwell, and the wells were rinsed three times with PBS. AP buffer (HBSS, pH 6.5) containing 10 mM MES and BL buffer (HBSS, pH 7.4) containing 10 mM HEPES were loaded in AP (500 mL) and BL (1500 mL) chambers, respectively, followed by incubation for 30 min at 37 C. Then, test compound (100 mM) was added to the AP side and incubated for 2 h at 37 C. After the incubation, BL buffer was collected and analysed using an UV spectrophotometer at a maximum wavelength (wavelength 225-357 nm).

Docking study
All computational works were performed on the Sybyl-X 2.1.1 (Tripos Inc., St Louis, MO, USA) molecular modelling package with CentOS Linux 5.4. operating system1 27 .

Preparation of ligands and receptor
The 11e was prepared with sketch module embedded in Sybyl package and saved as mol2 format. After sketching the molecule, Gasteiger-H€ uckel charges were assigned to all atoms. To optimise the ligand, energy minimisation was conducted by the standard tripos force field with convergence to maximum derivatives of 0.001 kcal mol À1 ÁÅ À1 . The X-ray structure of ALK5 complexed with 5,6-dihydro-4H-pyrrolo[1,2-b]pyrazole inhibitor was used as a receptor for docking (PDB id: 1RW8) 28 . Receptor structure was retrieved from PDB (http://www.rcsb.org/) and optimised using structure preparation tool embedded in biopolymer module. Native ligand was extracted and all water molecules except key water molecule for water-mediated hydrogen bond network were deleted from the complex structure.

Molecular docking
To examine the binding poses of 11e, docking study was conducted using Surflex-Dock3 embedded in Tripos Sybyl X 2.1.1 software package. For docking performance, the active site was assigned as a protomol generated by using the native ligand in the X-ray structure. Flexible docking was carried out by using default parameter values (threshold ¼ 0.5 and bloat ¼ 0), producing 200 conformers as maximum number of poses per ligand. Binding affinity of each docking pose of ligand was calculated by Surflex-dock score and consensus scoring function (CScore). The total Surflex-Dock score was expressed as -logKd to represent binding affinities. To build the best docking model, key interactions between candidate compound and active site were investigated by visual inspection.

ALK5 inhibitory activity in an enzyme assay and in a cellbased luciferase reporter assay
To evaluate whether these potential inhibitors 7a-c, 11a-h, and 16a-h could inhibit ALK5, a kinase assay was performed using the purified human ALK5 kinase domain produced in Sf9 insect cells and casein as a substrate ( Table 1). The ALK5 inhibitory activity of 7b (IC 50 ¼ 0.093 mM) was 2.4-fold and 3.6-fold higher than those of 7a (IC 50 ¼ 0.224 mM) and 7c (IC 50 ¼ 0.338 mM), respectively, in a kinase assay (Table 1). In a cell-based luciferase reporter assay using HaCaT (3TP-luc) stable cells, 7c (40% inhibition) was much less inhibitory than 7a (72% inhibition) and 7b (74% inhibition) at a concentration of 0.1 mM.
Among this series of compounds, 11e possessing the most potent ALK5 inhibitory activity and the highest permeability in Caco-2 cells was selected, and its ALK5 inhibitory activity was compared with those of potential competitors, galunisertib and vactosertib in a kinase assay and in a luciferase reporter assay. In a kinase assay, 11e (IC 50 ¼ 0.013 mM) showed the same level of potency to that of vactosertib (IC 50 ¼ 0.013 mM) and 6.6-fold higher potency compared to that of galunisertib (IC 50 ¼ 0.086 mM) ( Table 2). Luciferase activity of HaCaT (3TP-luc) cells was increased by 65-fold after treatment of TGF-b1 (2 ng/mL), and 11e and vactosertib displayed the similar level of inhibition on the TGF-b1induced luciferase reporter activity with IC 50 values of 0.0196 mM and 0.0165 mM, respectively. Similar to a kinase assay, galunisertib displayed much lower inhibition (IC 50 >0.1 mM) compared to 11e and vactosertib ( Table 2).
The kinase domain of p38a is known to be one of the most homologous to that of ALK5 39 , therefore, this enzyme was chosen to compare selectivity of 11e, galunisertib, and vactosertib. In a p38a kinase assay, IC 50 values of 11e, galunisertib, and vactosertib were 0.288 mM, 0.320 mM, and 1.775 mM, respectively, thus, their selectivity indices (IC 50 for p38a/IC 50 for ALK5) were 22, 4, and 137, respectively. Although 11e was 5.5-fold more selective against p38a than galunisertib, it was 6.2-fold less selective than vactosertib.

Caco-2 cell permeability assay
To estimate oral absorption of target compounds, their permeability in a Caco-2 monolayer was evaluated at a concentration of 100 mM ( Table 1). The 11e showed the highest permeability (56%), but 16f (6%), 16g (0%), and 16h (0%) showed very limited or no permeation in this assay, demonstrating that even simple structural modification markedly affected the permeability in this series of compounds.

Docking model of ALK5 in complex with 11e
To determine the binding pose of 11e in the active site of ALK5, we executed docking modelling with flexible molecular docking programme Surflex-dock 40 . We analysed the result of docking considering Surflex-dock docking score (-logKd) and consensus score (obtained from CScore module 27 ), and selected the poses of 11e with high scores (-logKd !7 and CScore !3). To select the best docked pose among them, we also identified key interactions between amino acid residues in the active site and 11e, in comparison with the X-ray pose of native ligand (3-(4-fluorophenyl)-2-(6-methylpyridin-2-yl)-5,6-dihydro-4H-pyrrolo[1,2-b]pyrazole) 28 . As shown in Figure 3(A,B), 11e fits well into cavity of active site, with quinoxaline ring and methyl group of 6-methylpyridine ring occupying the hydrophobic pockets. Hydrogen bond interactions between 11e and ALK5 are exhibited in Figure 3(C). The quinoxaline ring nitrogen acts as a hydrogen bond acceptor and interacts with NH of His283 in the backbone of hinge region of ALK5. The 6-methylpyridine ring nitrogen forms a water-mediated hydrogen bond network with the backbone NH of Asp351, and the side chain of both Tyr249 and Glu245. The imidazole ring NH of 11e is also involved in the water-mediated hydrogen bond network. The NH 2 group of carboxamide in the meta position of phenyl ring interacts with the side chain of Asn338 by forming hydrogen bond, which cannot be formed for the compounds, 11a-d having the carboxamide group in the para position. Our docking model for the most active compound 11e well supports the key interactions for ALK5 inhibition which was previously reported 28 . 11e is (magenta carbon atoms) superimposed over the X-ray pose of native ligand (grey carbon atoms). The active site of ALK5 is shown as MOLCAD lipophilic potential surface map (A), and lipophilicity increases from blue (hydrophilic) to brown (liphophilic). Grey capped sticks represent key amino acid residues within the binding site, and the backbone of ALK5 is shown as ribbon. The bound water molecule in the X-ray structure is represented by ball and stick. (C) Intermolecular interaction between 11e and ALK5. Grey capped sticks represent key amino acid residues in the active site. Red dashed lines are hydrogen bonding interactions (<3.0 Å).

Conclusions
In this report, a series of 2,4-disubstituted-5-(6-alkylpyridin-2-yl)-1H-imidazoles, 7a-c, 11a-h, and 16a-h has been synthesised and evaluated for their ALK5 inhibitory activity in an enzyme assay and in a cell-based luciferase reporter assay. The structure-activity relationships in this series of compounds revealed that an ethylene linker at the two-position of the imidazole ring was the most beneficial in ALK5 inhibitory activity. Replacement of a benzo[1,3]dioxol-5-yl moiety at the four-position of the imidazole ring with a quinoxalin-6-yl moiety markedly increased ALK5 inhibitory activity. Regarding the alkyl substituent at the six-position of the pyridine ring, the compounds having a methyl or an ethyl substituent displayed much higher inhibitory activity than the compounds having an i-propyl or a n-butyl substituent. The m-CONH 2 analogues were more inhibiting than the p-CONH 2 analogues in compounds having a methylene linker, whereas the p-CONH 2 analogues were more inhibiting than the m-CONH 2 analogues in compounds having an ethylene linker. In a cell permeability assay using Caco-2 monolayer, 11e showed the highest permeability in this series of compounds. The 11e was equipotent to vactosertib, but much more potent than galunisertib in an ALK5 kinase assay and in a cell-based luciferase reporter assay. Although 11e was 5.5-fold more selective against p38a than galunisertib, it was 6.2-fold less selective than vactosertib. Therefore, it can be concluded that combination of replacement of a [1,2,4]triazolo[1,5-a]pyridin-6-yl moiety with a quinoxalin-6-yl moiety, insertion of a methylene linker instead of a methyleneamino linker, and a m-CONH 2 substituent in the phenyl ring in vactosertib maintained its high ALK5 inhibitory activity, but decreased its selectivity against p38a. Selectivity profiling of 11e using a panel of 28 protein kinases showed that it is highly selective for ALK5. Our docking results demonstrate that 11e fits well in the ATP-binding pocket of ALK5 with favourable intermolecular interactions.

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