Structure-Guided Design of C4-alkyl-1,4-dihydro-2H-pyrimido[4,5-d][1,3]oxazin-2-ones as Potent and Mutant-Selective Epidermal Growth Factor Receptor (EGFR) L858R/T790M Inhibitors

Epidermal growth factor receptor (EGFR) T790M acquired drug-resistance mutation has become a major clinical challenge for the therapy of non-small cell lung cancer. Here, we applied a structure-guided approach on the basis of the previous reported EGFR inhibitor (compound 9), and designed a series of C4-alkyl-1,4-dihydro-2H-pyrimido[4,5-d][1,3]oxazin-2-one derivatives as novel mutant-selective EGFR inhibitors. Finally, the most representative compound 20a was identified, which showed high selectivity at both enzymatic and cellular levels against EGFRL858R/T790M (H1975 cell lines) over EGFRWT (A431 cell lines). The representative compound 20a also showed promising antitumor efficiency in the in vivo antitumor efficacy study of H1975 xenograft mouse model driven by EGFRL858R/T790M. These results provide a new scaffold for the treatment of dual-mutant-driven non-small cell lung cancer.

Extensive efforts have been devoted to the development of novel covalent EGFR inhibitors to overcome gefitinib-and erlotinib-resistant mutant (T790M mutation). These irreversible inhibitors are designed with electrophilic Michael-acceptor systems to covalently react with the conserved Cys797 in the EGFR active site, so as to increase inhibition potency against T790M mutant relative to reversible agents. Unfortunately, because of the dose-limiting toxicities attributed to inhibition of the wild-type (WT) EGFR, these second-generation irreversible inhibitors ( Fig. 1) including 3 (afatinib) 13 , 4 (neratinib) 14 , 5 (dacomitinib) 15 did not improve clinical efficacy for NSCLC patients who have developed T790M acquired resistance. Recently, the third-generation (mutant-selective) irreversible EGFR-tyrosine kinase inhibitors (TKIs) based on an amino pyrimidine scaffold, such as compounds 6 (WZ4002) 16 , 7 (CO-1686) 17 and 8 (AZD9291) 18 have demonstrated promising selectivity for EGFR L858R/T790M mutant over WT EGFR, indicating that this strategy is feasible for overcoming EGFR T790M gatekeeper mutation in NSCLC treatment (Fig. 1). Based on their clinical significant benefits for NSCLC patients with EGFR T790M acquired drug-resistance mutation, United States Food and Drug Administration (FDA) has awarded compounds 7 and 8 "Breakthrough Therapy" designations in 2014 19 . Furthermore, 8 has been granted accelerated approval by FDA for the treatment of late-stage NSCLC patients with EGFR T790M mutation-positive who have progressed after other EGFR TKIs therapy 20 .

Results
Scaffold Hopping and Binding Mode Analysis. Using a structure-based approach, we reported compound 9 as a potent irreversible inhibitor with excellent selectivity against EGFR L858R/T790M kinase over EGFR WT kinase (4.7 nM vs 474 nM, 101-fold) 21 . As shown in Fig. 2B, the modeling binding mode suggests that compound 9 binds to the EGFR T790M (PDB code: 3IKA) via covalently modifying the conserved Cys797 residue at the lip of the ATP-binding site. The 6,7-dioxo-6,7-dihydropteridine core is predicted to interact with the residue Met793 at the hinge region by an expected bidentate hydrogen bonding and also form a hydrophobic sandwich with the ceiling residue Ala743 and the floor residue Leu844. The N-methylpiperazine ring faces to the solvent exposed region. The two phenyl rings substituted at C2 and N8 adapt a U-shape binding mode. These interactions suggested by   [4,5-d] [1,3]oxazin-2-ones scaffold. Compound 16a (orange) was aligned to the modeled conformation of compound 9 (marine) shown in panel A by SHAFTS. The EGFR WT crystal structure (shown in yellow, PDB code: 4G5J) was aligned to the EGFR T790M (shown in cyan, PDB code: 3IKA). The hinge regions of both EGFR WT and EGFR T790M were presented as stick. The carbon atoms of 9 are colored in marine and the nitrogen atoms of 9 are colored in dark blue. (D) The 2D diagram interaction of compound 16a. (E) The proposed binding mode of compound 20a bound to EGFR T790M (cyan line, PDB code: 3IKA). (F) The proposed binding mode of compound 20a bound to EGFR WT (yellow line, PDB code: 4G5J). The conformations of compound 9, 16a and 20a in the panel C,E and F are depicted as stick and highlighted in grey surface presentation. the docking study stabilize the basic recognition of inhibitor-kinase complex. Moreover, the methyl group at N5-position points to the mutant gatekeeper Met790 residue and strengthens nonpolar contacts with T790M mutants, which is a vital contribution proposed by the docking study to the selectivity against EGFR L858R/T790M over EGFR WT . Besides, the L858R mutation is far from the ATP-binding pocket of EGFRs and would contribute little to the interaction at the inhibitor-binding site. Given the former successful design strategy targeting EGFR L858R/T790M and aimed at improving selectivity over EGFR WT , a scaffold hopping of the 6,7-dioxo-6,7-dihydropteridine core was conducted. The piperazinediones part of compound 9 was replaced with 1,3-oxazinan-2-one in order to introduce one or two hydrophobic substituents at the C4-position, and thus it could not only occupy the lipophilic subpocket formed by the gatekeeper residue but also strengthen nonpolar contacts with Met790 residue (Fig. 2C and D). This computer-aided approach resulted in compound 16a. As expected, the superposition of compound 9 and compound 16a demonstrates that the C4-position methyl group is in van der Waals contacts (shown in little green disc, distance of 3.3 Å) with the side chain of Met790 residue in the case of the EGFR T790M while for the EGFR WT the corresponding distance is much closer (shown in large red disc, distance of 2.5 Å), resulting in a steric clash between the N5-methyl group and side chain of Thr790 residue. These different interactions between N5-position moiety and the gatekeeper residue would affect the potency and selectivity.
Chemistry. The synthesis of compounds 16a-e is outlined in Fig. 3. In summary, treatment of the ethyl 2,4-dichloropyrimidine-5-carboxylate with Boc-protected amine yielded the ethyl4-((3-((tert-butoxycarbonyl) amino)phenyl)amino)-2-chloropyrimidine-5-carboxylate (10), followed by reduction of the ethyl ester group and oxidization of the hydroxymethyl to obtain the key intermediate (12). A reaction of 12 with Grignard reagents generated compounds 13a-d. Then compounds 11, 13a-d were cyclized with 1,1′-carbonyldiimidazole to obtain the ring-closure compounds (14a-e), which were reacted with arylamine by displacement of 2-chloro in the pyrimidine to yield compounds 15a-e. Subsequently, removal of the protecting group with trifluoroacetic acid in dichloromethane and reacted with acryloyl chloride to provide the desired compounds 16a-e.
Synthesis of compounds 20a-c is shown in Fig. 4. Treatment of compound 10 with appropriate Grignard reagents afforded the corresponding intermediates 17a-c, which were performed by a procedure analogous to that for synthesis of compound 16a to give the desired compounds 20a-c.
(methyl, ethyl, propyl and isopropyl) at the C4-position of the scaffold were designed and synthesized (Fig. 3). Their in vitro enzymatic inhibitory activities against EGFR L858R/T790M and EGFR WT were evaluated by using the well-established ELISA-based assay. As shown in Fig. 5, compounds 16a and 16b indeed demonstrated different inhibitory activities for dual-mutant (DM) and WT EGFR kinases. They displayed single nanomolar inhibitory activities for EGFR L858R/T790M with IC 50 values of 5.4 and 6.1 nM, respectively, while their inhibition for EGFR WT were ~4-7-fold less potent. Introduction of propyl and isopropyl groups in the 4-position of the core led to compounds 16c and 16d, which showed decreased potency for EGFR kinases and significant loss in selectivity profiles between EGFR L858R/T790M and EGFR WT (Fig. 5). The bioactivities of 16c and 16d indicated that this hydrophobic subpocket is unable to accommodate these two longer and bulkier alkyl groups, thus resulting in detrimental influence on potency and selectivity. To validate the key contribution of the introduced alkyl groups for EGFR kinases selectivity, compound 16e, a C4-unsubstituted analogue, was also prepared (Fig. 3). Compared to compound 16a, compound 16e displayed not only less potent inhibition effect for EGFR L858R/T790M (IC 50 = 7.3 nM), but also slight improvement in inhibitory activity for EGFR WT (IC 50 = 24 nM), leading to a 3.3-fold decrease in its selectivity between the DM and WT EGFRs. These results therefore showed that the small hydrophobic  C4-substitutent, such as methyl group, occupying the lipophilic subpocket formed by the mutant gatekeeper residue was favorable to inhibitor of DM EGFR with selectivity over WT EGFR.
On the basis of above single-substituent structural modification, we hypothesized that introducing one more suitable hydrophobic moiety at C4-position to further strengthen the nonpolar contacts with mutant gatekeeper Met790 may be beneficial for improving the mutant selectivity profile. So we applied this rationale to compound 16a through attaching a dual-alkyl group at the C4-position, and resulted in compounds 20a-c (Fig. 4). As expected, compound 20a, dual-methyl groups substituted derivative at C4-position of the 1,4-dihydro-2H -pyrimido [4,5-d] [1,3]oxazin-2-one, inhibited EGFR L858R/T790M with an IC 50 value of 4.5 nM, which was similar to that of compound 16a. Whereas its inhibition of WT EGFR was less potent (IC 50 = 156 nM). Thus, it showed approximate 35-fold selectivity between EGFR L858R/T790M and EGFR WT , and was highly comparable to the positive control compound 7. The structure-based insights from molecular modeling are consistent with the experimental observations. As shown in the Fig. 2E, compound 20a was covalently docked into the ATP-binding site of EGFR T790M and the hinge binder part is predicted to tightly interacts with the backbone of Met793 (two H-bonds distance: 3.1 and 3.1 Å). Its upper methyl group fits the methionine gatekeeper pocket well and its lower methyl group is involved in van der Waals contacts with Met790. In the case of the proposed binding mode of EGFR WT (Fig. 2F), the branched residue Thr790 leads a steric clash to the dual methyl moiety ( Fig. 2C and D) and pushed the core of compound 20a to the right side of the ATP-binding site, which attenuates the hydrogen-bonding with the hinge region (two H-bonds distance: 3.3 and 3.9 Å). These comparable docking studies also validated this selectivity profile. However, when introducing bulkier substitutions, such as dual-ethyl and dual-propyl groups at the C4-position of the core, the compounds (20b and 20c) led to a complete loss of inhibitory activities (>1000 nM) for both DM and WT EGFR kinases. These results implied that longer and bulkier substituents than dual-methyl groups are not tolerated to the hydrophobic subpocket.
In Vitro Cellular Anti-proliferation Assay. Having explored potent inhibition and obvious selectivity of EGFR kinases by introduction of alkyl groups to C4-position to the 1,4-dihydro-2H-pyrimido [4,5-d] [1,3]oxazin-2-one ring, we subsequently evaluated their anti-proliferative activities in H1975 cells (expressing EGFR L858R/ T790M ) and A431 cells (expressing EGFR WT ) in vitro by using the sulforhodamine B (SRB) colourimetric assay. The anti-proliferative activities of our designed compounds against H1975 and A431 cell lines range from 135 nM to more than 9 μM and from 554 nM to more than 7 μM, respectively (Fig. 5). Additionally, the selectivity at the cellular level is from 0.7 to 9.4-fold. Importantly, compound 20a exhibited potent growth-suppressing activities against the H1975 cell lines with IC 50 values of 135 nM, and achieved high selectivity over A431 cell lines (9.4-fold, IC 50 = 1274 nM), which is comparable with the positive control compound 7 (15-fold). Therefore, compound 20a, which shows best potency at both enzymatic and cellular levels, was selected as the representative compound evaluated in the following pharmacology assays.
Western Blot Analysis. In order to further validate the observed selectivity of enzymatic and cellular inhibition, we selected the representative compound 20a to examine its suppressive functions on the activation of EGFR and the downstream signaling proteins in H1975 and A431 cell lines. As illustrated in Fig. 6A, compound 20a inhibited the phosphorylation of EGFR L858R/T790M and AKT in H1975 cells at the concentration as low as 10 nM, and inhibited downstream molecule ERK phosphorylation at the concentration up to 100 nM, demonstrating similar activity with compound 8 (positive control). Meanwhile, its inhibitory activity against A431 cell lines harboring EGFR WT was obviously less potent (Fig. 6B). For example, compound 20a significantly suppressed the activation of EGFR L858R/T790M in H1975 cells at 10 nM, but it did not show obvious effect until reaching high concentration at 1 μM in A431 cells. The results demonstrated that 20a strongly inhibited EGFR L858R/T790M while EGFR WT sparing at the cellular level, further suggesting that 20a was a potent and selective EGFR L858R/T790M inhibitor. In Vivo Antitumor Efficacy Study of Compound 20a. In view of its encouraging ability of inhibition for EGFR kinases and cells, we also evaluated the preliminary in vivo antitumor activity of compound 20a in H1975 xenograft mouse model driven by EGFR L858R/T790M (Fig. 7). Compound 20a and positive control compound 7 were administered orally once per day at a dosage of 50 mg/kg for continuous two weeks (14 days). As shown in Fig. 7A, compound 20a (tumor growth inhibition (TGI) = 71%, P < 0.05) potently suppressed the proliferation and growth of the tumor, which is comparable with that of compound 7 (TGI = 72%, P < 0.05). Furthermore, the dosage of compound 20a was well tolerated, with no obvious effect on the body weight observed during the treatment period (Fig. 7B).

Discussion
Herein, we combined the ligand-based scaffold hopping with the structure-based analysis of differences of the binding pockets between EGFR L858R/T790M and EGFR WT to design a new 1,4-dihydro-2H-pyrimido [4,5-d] [1,3] oxazin-2-one scaffold targeting EGFR L858R/T790M and improve selectivity over EGFR WT . The scaffold hopping method in silico helps us modify more than one hydrophobic substituent at C4-position. Compared with the compound 16e (no substitution at C4), the single methyl group substituent on the compound 16a occupies hydrophobic subpocket, leading to a slightly 1.4-fold increase of the enzymatic inhibition against EGFR L858R/T790M (7.3 nM vs 5.4 nM). With the slight potency decline against EGFR WT (16e vs 16a, 24 nM vs 38 nM), the selectivity against EGFR L858R/T790M over EGFR WT boost 2.1-fold (16e vs 16a, 3.3-fold vs 7-fold). In the second round dual-substituent optimization, the one more methyl group at C4-position stabilized the nonpolar interaction with gatekeeper residue Met790 (Fig. 2E), keeping the enzymatic inhibitory activities against EGFR L858R/T790M at the same level (16a vs 20a, 5.4 nM vs 4.5 nM). Besides, dual-methyl group introduces more steric clashes into the interaction between the inhibitor and the EGFR WT (Fig. 2F), leading to a pronouncing decrease of enzymatic inhibition against EGFR WT (16a vs 20a, 38 nM vs 156 nM). Thus, the representative compound 20a achieved the best selectivity (35-fold) in this series, which is comparable with the compound 7 (29-fold). The selectivity improvements achieved along the whole structural optimization verified the rationality of our design strategy.
According to the results from enzymatic and cellular inhibition assays and western blot analysis, this series of compounds present good targeting properties. In particular, the antiproliferation potencies against H1975 and A431 cell lines are highly correlated with the enzymatic inhibitory activities against EGFR L858R/T790M and EGFR WT , respectively (Fig. 8). The western blot analysis of representative compound 20a validated the selectivity profile observed both in the enzymatic and cellular assays (Fig. 6). Moreover, the in vivo efficacy for double-mutant tumor xenografts of 20a is highly comparable with that of the drug at clinical trials (compound 7). These good targeting and selective properties of representative compound may imply the potential clinical use for the treatment of the drug-resistance NSCLC and less side-effects caused by the EGFR inhibitor without mutant selectivity.
In summary, through structure-guided design approach based on our previous work on mutant-selective EGFR inhibitor (compound 9), we have designed and generated a novel 1,4-dihydro-2H-pyrimido [4,5-d] [1,3] oxazin-2-one scaffold, followed by introduction of alkyl groups contacting the Met790 gatekeeper residue into the C4-position to keep the selectivity for the EGFR L858R/T790M over EGFR WT , leading to discovery of one of the most potent compound 20a. 20a not only showed remarkable selectivity of enzyme inhibition for EGFR L858R/ T790M over EGFR WT , but also strongly inhibited the proliferation of H1975 cells, while significantly less effective on A431 cell lines. A further preliminary in vivo antitumor efficacy evaluation on 20a suggested that it significantly inhibited tumor growth in H1975 NSCLC xenograft mouse model via po dosing at 50 mg/kg/day for 14 days. Considering the encouraging enzymatic and cellular selectivity and in vivo antitumor efficiency of compound 20a, this agent might be a promising lead compound for further optimization as a selective EGFR L858R/T790M inhibitor to overcome T790M resistance mutation. The results also provide more insights for designing new classes of mutant-selective EGFR inhibitors. Further pre-clinical evaluations for the candidate compound are in progress and will be performed in due course.

Materials and Methods
Molecular Modeling. The X-ray structures (PDB code: 3IKA and 4G5J) were downloaded from the Protein Data Bank (PDB, http://www.pdb.org). The comparative docking study was conducted via three steps: First, compounds 9 and 20a were docked into the EGFR T790M (PDB code: 3IKA) and EGFR WT (PDB code: 4G5J) as reversible inhibitors using Glide 23 (version 5.5) in extra precision (XP) mode with default settings; Then, the covalent bond between the Cys797 and the electrophilic group was formed by the build panel in Maestro 24 ; Finally, to fix the bond lengths and eliminate steric clashed, the covalently modified inhibitor-kinase complexes were minimized in water solvent using MacroModel application with flexible ligand and constrained protein.
The scaffold hopping was conducted by SHAFTS 25,26 . The query molecule is the docked pose of compound 9 obtained from the above docking study. The aligned pose of compound 16a (shown in Fig. 2C) was generated by fast 3D similarity calculation of SHAFTS. The score for the similarity calculation between compound 9 and 16a is 1.494. The range of the SHATFS' similarity score is from 0 (totally dissimilar) to 2 (completely similar). The Fig. 2D was generated by LigPlot+ 27 .
Reagents and General Methods. All chemical reagents and solvents were purchased from commercial sources, and used without further purification. The positive control compound 7 was purchased from Shanghai Future Chemical Technology Co., Ltd. Thin-layer chromatography (TLC) was used to monitor the progress of reactions. Column chromatography was performed using 300−400 mesh silica gel (Qingdao Haiyang Chemical Co.,Ltd). 1 H and 13 C NMR spectra were recorded on a Bruker AV-400 spectrometer at 400 MHz and 100 MHz, respectively. LC-MS and high-resolution mass spectra (HRMS) were measured at the Institute of Fine Chemistry of East China University of Science and Technology. The purity of all final compounds were analyzed by using HPLC (Hewlett-Packard 1100), with the purity of all compounds to be over than 95% (>95%). HPLC instrument: (column: Zorbax RP-18, 5 μm, 4.6 mm × 250 mm, reverse phase; detector: photodiode array detector). The mobile phases A and B were acetonitrile and 10 mM NH 4 OAc in water (pH 6.0), respectively. A gradient of 5−100% the mobile phase A was run at a flow rate of 1.0 mL/min.
Scientific RepoRts | 7: 3830 | DOI:10.1038/s41598-017-04184-9 To a solution of the above obtained compound (0.335 g, 0.68 mmol), DIPEA (0.131 g, 1.02 mmol) in CH 2 Cl 2 (5 mL) was added acryloyl chloride (72 μL, 0.88 mmol) dissolved in CH 2 Cl 2 (1 mL) at 0 °C. The reaction mixture was stirred overnight at room temperature. The mixture was partitioned between dichloromethane (100 mL) and saturated aqueous NaHCO 3 (10 mL). The organic layer was washed with saturated aqueous NaCl, dried over anhydrous Na 2 SO 4 and concentrated in vacuo. The residue was purified by silica gel chromatography (dichloromethane/methanol = 15:1, v/v) to give the product 20a (0.165 g, 45%). 1  The following compounds 20b and 20c were prepared by a method similar to that for the synthesis of compound 20a.    13  In Vitro Enzymatic Activity Assay. The enzymatic activity assay was evaluated by using the well-established ELISA-based assay. Kinases domain of EGFR L858R/T790M and EGFR WT were expressed by using the Bac-to-Bac ™ baculovirusexpression system (Invitrogen, US), followed by purified in Ni-NTA columns (QIAGEN Inc., US). The detailed experiment was performed according to the published literature 21 . In Vitro Cellular Antiproliferation Assay. In vitro cellular antiproliferation activities of the designed compounds against H1975 and A431 cell lines were examined by using the well-established sulforhodamine B (SRB) colorimetric assay 28 . The detailed experiment was performed according to the published literature 21 . Briefly, H1975 and A431 cells were seeded in 96-well plates, and allowed to adhere overnight followed by treated with gradient concentrations of the test compounds for 72 h. The inhibition rate for cell proliferation was calculated as [1 − (A 515 treated/A 515 control)] × 100%. The IC 50 values were obtained by using the Logit method and reported as the mean ± SD from three independent determinations. Western Blot Analysis. Western blot analysis was performed with standard procedures. Briefly, H1975 and A431 cell lines were seeded in six-well plates, starved in serum-free medium for 24 h, followed by incubation for 2 h in the presence of different concentrations of compounds 20a and 8 (as positive control). Cells were stimulated by the addition of EGF (50 ng/mL, R&D Systems, US) for 10 min, then washed with PBS and lysed in SDS lysis buffer. Antibodies against p-EGFR (Y1068; #3777 S), EGFR (#4267 S), p-ERK (T202/Y204; #4370 L), ERK (#4695 S), p-AKT (Ser473), AKT and Tubulin were used (Cell Signaling Technologies, Cambridge, MA).

N-(3-(7-((2-Methoxy-4-(4-methylpiperazin-1-yl)phenyl)amino)-2-oxo-4,4-dipropyl-
In Vivo Antitumor Efficacy Study. All experimental procedures involving animals were performed in accordance with the ethical guidelines from Shanghai Laboratory Animal Public Service Center and Animal Ethics Committee of the East China University of Science and Technology. The whole experimental protocol was also approved by the Animal Ethics Committee of the East China University of Science and Technology. All 4-6 weeks old nude mice were purchased from Shanghai Slac Laboratory Animal Co. Ltd (The license of experimental animal: SCXK (shanghai) 2012-0002). The mice were maintained under the 12 hour light/dark cycles condition in the specific pathogen free (SPF) cleanroom with fresh air (more than 50%), appropriate temperature (22-26 °C) and humidness (40-60%). All materials and containers were disinfected and sterilized before use. Nude mice, aged 4-6 weeks, were inoculated subcutaneously into the right flank with H1975 NSCLC cells (approximately 2 × 10 6 cells/mouse). When the tumor size reached a volume of ~200−400 mm 3 , animals were randomly assigned into control and treated groups (n = 3/group). For efficacy studies, mice in control and in treatment groups were treated with vehicle compound 7 and compound 20a (50 mg/kg/day) for continuous 14 days, respectively. The sizes of tumor volume (V) were measured every two or three days by using vernier caliper, and were calculated with the formula V = (L × W 2 )/2, where L and W stands for length and width, respectively.