Discovery of 5-methyl-N-(2-arylquinazolin-7-yl)isoxazole-4-carboxamide analogues as highly selective FLT3 inhibitors.

A series of 4-arylamido 5-methylisoxazole derivatives with quinazoline core was designed and synthesised based on conformational rigidification of a previous type II FMS inhibitor. Most of quinazoline analogues displayed activity against FLT3 and FLT3-ITD. Compound 7d, 5-methyl-N-(2-(3-(4-methylpiperazin-1-yl)-5-(trifluoromethyl)phenyl)quinazolin-7-yl)isoxazole-4-carboxamide, exhibited the most potent inhibitory activity against FLT3 (IC50= 106 nM) with excellent selectivity profiles over 36 other protein kinases including cKit and FMS kinase. Compound 7d was also active in FLT-ITD, with an IC50 value of 301 nM, and other FLT3 mutants showing potential as an AML therapeutics.


Introduction
FLT3, a type of trans-membrane receptor tyrosine kinase expressed on lympho-hematopoietic cells, is an attractive target in acute myeloid leukaemia (AML) 1 . When FLT3 ligand binds to FLT3 kinase, it is activated by auto-phosphorylation, which then activates its multiple downstream signalling pathways, including signal transducer and activator of transcription 5 (STAT5), ras/ mitogen-activated protein kinase, and phosphatidylinositol 3-kinase/Akt pathways, playing a key role in cell proliferation, survival, and immune response [2][3][4][5] .
However, mutation of FLT3 causes its ligand-independent activation and can be classified by location and type, including internal tandem duplication (ITD) and point mutation in the tyrosine kinase domain (TKD). FLT3-ITD mutations are present in 20-30% of AML patients and are especially related to leucocytosis and poor prognosis and point mutation in TKD is found in 5% of AML patients 6 . Despite the significance of seeking AML therapeutics, only a few molecules have been approved, such as midostaurin and gilteritinib (ASP2215), which are depicted in Figure 1 7,8 .
Protein kinase inhibitors can be classified as type I, II, III, IV and V 9 based on their binding mode. Among these inhibitors, type II kinase inhibitor can take advantage of selectivity acquisition and exhibits promising potency due to additional interaction with the DFG pocket adjacent to the ATP-binding pocket on top of the hinge hydrogen bond in the ATP site 10 . Previously, we reported very potent type II FMS inhibitors 11 and selective benzimidazoleincorporated FLT3 inhibitors 12 . Here, we demonstrate that the amide or urea between the middle phenyl ring and the secondary hydrophobic aromatic ring can be rigidified further, introducing a quinazoline structure as a bioisostere ( Figure 2). The quinazoline structures are well-known privileged structures in medicinal chemistry, exhibiting diverse biological activities 13 . However, this is the first attempt to incorporate one in type II PKI modification.
Introducing quinazoline into type II FMS kinase inhibitor, we could discover a novel FLT3 inhibitor with excellent selective profile.
All quinazoline compounds, 7a-q, were evaluated for their activity against FLT3 kinase and FLT3-ITD mutation and the results are shown in Table 1. Most of the synthesised compounds exhibited selective activity against FLT3, particularly those incorporating the piperazine moiety. Among the compounds evaluated, 7d showed the most potent activity against FLT3 with an IC 50 value of 106 nM, and FLT3-ITD with an IC 50 value of 301 nM. Structure activity relationships (SARs) were inferred from the data.
In our previous work, benzimidazole compounds retained their activity against FLT3 regardless of presence of 1,3,5-substituted or 1,3,4-substituted benzoic acid, and the activity was determined as piperazine > imidazole > morpholine substituents 12 . We optimised quinazoline derivatives based on the observation of previous benzimidazole derivatives. Those with methyl piperazine or morpholine as the phenyl substitution group (7d and 7b) were more potent about 2-to 5-fold (IC 50 values of 0.106 and 3.98 mM, respectively) compared to corresponding benzimidazole series (IC 50 values of 0.495 and 7.94 mM), and 7c displayed better potency (IC 50 value of 1.58 mM) than that of benzimidazole (IC 50 value of 2.33 mM). Introduction of 3,5-disubstituted benzoic acid through quinazoline connection maintained the activity (7b, 7c, 7d, 7e, 7n), but quinazoline compound with 1,3,4-substituted benzoic acid (7a) and one with pyrazole (7h) caused loss of activity towards FLT3.
With the result of 7d, we synthesised compound 7e to optimise the linkage between the phenyl group and the piperazine moiety. Although inhibitory activity towards FLT3 was retained, 7e exhibited decreased activity, about 10-fold less than that of 7d. On the predicted binding mode of 7d, strong ionic interaction between the protonated nitrogen of the piperazine and Asp829 might enhance its binding affinity. Almost the same ionic interaction seems possible in case 7e, but the ionic interaction might push the whole compound slightly out of the active site due to its length, resulting in loss of multiple interactions such as hydrogen bonding with Asp829, p-p interaction with Phe691, and p-cation interaction with Lys644 ( Figure 3). We also replaced piperazine with a piperidine moiety (7n) to investigate the role of nitrogen in the piperazine structure. The IC 50 value of 7n was 3.59 mM, similar but weaker than 7e despite their similar structures. Our docking study showed that one conformer of 7n with equatorial O linkage bound with FLT3 similarly to compound 7e, but the other conformer with axial O linkage was not suitable to bind tightly to the active site because of its non-linear piperidine moiety ( Figure 4).
Moreover, the most potent 7d was examined further for inhibitory activity against other FLT3 mutants and the result was impressive ( Table 2). The 7d was a potent inhibitor towards FLT3 (F594_R_R595) (IC 50 ¼ 0.524 mM), FLT3 (F594_EY_R595) (IC 50 ¼ 0.495 mM), FLT3 (Y591_ VDFREYEYD_V592) (IC 50 ¼ 0.728 mM), and FLT3 (D835Y), with an IC 50 value of 0.228 mM. Since FLT3-ITD mutation occur mainly in the juxtamembrane domain away from the ATP binding site, the FLT3 inhibitors binding in ATP site exhibit potency against FLT3-ITD mutant 14 . On the other hand, a mutation in active site such as FLT3 (D835Y) induces active conformation of kinase and type II inhibitors generally could not bind well enough. This is also known to be a mechanism of secondary resistance of type II FLT3 inhibitor. 13 Nevertheless, our compound 7d, which is a type II FLT3 inhibitor, exhibits a potent inhibitory activity (IC 50 ¼ 228 nM) in FLT3-TKD (D835Y) similar to that of FLT3 wild type also. FLT3-TKD, especially D835Y, D835H mutant was identified as a main drug resistance mechanism of FLT3 inhibitor in the clinical trial. [15][16][17][18] Like quizartinib, 7d displays potent inhibitory activity against FLT3-TKD (D835Y) with value of IC 50 228 nM, while decreased about 2-fold compared to that of FLT3 wild type.
Next, we investigated kinase panel screening of 7d and 7b over 36 different kinases at a single dose of 10 mM ( Figure 5). Both compounds 7d and 7b (data in Supplementary material) showed an excellent selectivity profile, having almost no activity on other kinases including FMS and cKit. High selectivity over cKit might be an opportunity to avoid myelosuppression toxicity, which was reported from a dual FLT3/cKit inhibitor [19][20][21][22] . It is a valuable result For better understanding of the interactions between our compounds and FLT3, molecular docking of compound 7d into the ATP binding pocket of FLT3 (PDB: 4RT7) was performed using Glide software (Schr€ odinger, Version 14.2). The docking result is depicted in Figure 6. In the binding model of 7d, 5-methyisoxazole forms a p-p interaction with Phe830 instead of hydrogen bonding with Cys694, contrast to the co-crystal structure of quirzatinib 23 . Instead, the quinazoline ring of 7d showed possibilities of multiple interactions: one hydrogen bond with Asp829, p-p interaction with Phe691, and p-cation interaction with Lys644. As it was designed as a hybrid of the middle phenyl ring and urea (or amide) linker, the quinazoline seems tightly bound with ATPbinding site and adjacent region like other type II inhibitors. However, instead of fitting into a hydrophobic pocket surrounded by hydrophobic residue, the strong ionic interaction between the protonated nitrogen of the piperazine ring and the Asp796 might enhance the binding affinity, resulting in potent inhibitory activity and considerable selectivity about FLT3. Due to its strong ionic interaction, the fused ring system of quinazoline might make additional interactions in the active site of FLT3 unlike benzimidazole derivative 12 . The quinazoline derivative 7d interacts with not only Lys644 (p-cation interaction between guanidine group of Lys644 and quinazoline ring) also gatekeeper residue, Phe691, (p-p interaction between phenyl group of side chain and quinazoline structure) properly, which could contribute to enhanced activity of compound 7d presumably compared with benzimidazole compound 12 .
We also tried to explain the selectivity of quinazoline compound for FMS and cKit. As shown in Figure 6, FMS kinase has   Thr663 instead of Phe691 in the active site. 7d could not bind well in the FMS kinase without multiple interaction which quinazoline forms. Although 7d forms hydrogen bonding with Cys666 and strong ionic interaction with Asp796, the quinazoline moiety does not form any interaction with Thr663 or other residue. The difference of only one residue in the highly conserved active site seems to play crucial role in selectivity profile and the interaction between quinazoline structure and Phe691 also seems to be  important to binding affinity. In the case of cKit kinase, threonine residue locates at the same position, and same explanation could be applied.

Conclusions
We designed and synthesised a series of 4-arylamido 5-methylisoxazole derivatives containing quinazoline using a conformational rigidification strategy. Various analogues were synthesised and tested for inhibitory activity against FLT3. Compound 7d displayed the most potent inhibitory activity against FLT3, with an IC 50 of 106 nM, and was also active against FLT3-ITD and FLT3-TKD (D835Y), with an IC 50 of 301 nM and 228 nM respectively. The novel type II FLT3 inhibitors showed an excellent selectivity profile having 20% or less activity towards other kinases including similar FMS, cKit. Specifically, the selectivity over FMS and cKit potentially relieve the myelosuppression toxicity observed in the dual FLT3/ cKit kinase inhibitor. Molecular docking of 7d into the ATP binding site of FLT3 kinase was performed and the result disclosed that the quinazoline structure could perform an adequate or even better bioisostere of benzimidazole in previous report.
Considering that FLT3 and FLT3 mutants have been significantly associated with FLT3 mutant-mediated AML, compound 7d could be a valuable molecule for FLT3-mutant positive physiology and pathology research and a potential therapeutic agent. Even if further optimisation would be necessary in terms of potency, our research towards a novel selective FLT3 inhibitor suggests the potential of quinazoline compounds as AML therapeutics.

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