Optimization of Substituted 6-Salicyl-4-Anilinoquinazoline Derivatives as Dual EGFR/HER2 Tyrosine Kinase Inhibitors

4-Anilinoquinazolines as an important class of protein kinase inhibitor are widely investigated for epidermal growth factor receptor (EGFR) tyrosine kinase or epidermal growth factor receptor 2 (HER2) inhibition. A series of novel 6-salicyl-4-anilinoquinazoline derivatives 9–27 were prepared and evaluated for their EGFR/HER2 tyrosine kinase inhibitory activity as well as their antiproliferative properties on three variant cancer cell lines (A431, MCF-7, and A549). The bioassay results showed most of the designed compounds exhibited moderate to potent in vitro inhibitory activity in the enzymatic and cellular assays, of which compound 21 revealed the most potent dual EGFR/HER2 inhibitory activity, with IC50 values of 0.12 µM and 0.096 µM, respectively, comparable to the control compounds Erlotinib and Lapatinib. Furthermore, the kinase selectivity profile of 21 was accessed and demonstrated its good selectivity over the majority of the close kinase targets. Docking simulation was performed to position compound 21 into the EGFR/HER2 active site to determine the probable binding pose. These new findings along with molecular docking observations could provide an important basis for further development of compound 21 as a potent EGFR/HER2 dual kinase inhibitor.


Introduction
The epidermal growth factor receptor (EGFR) family of receptor tyrosine kinase consists of human epidermal growth factor receptor (EGFR/ErbB-1), human epidermal growth factor receptor 2 (HER2/ErbB-2), human epidermal growth factor receptor 3 (Her3/Erb-3), and human epidermal growth factor receptor 4 (Her4/Erb-4), each of which was closely implicated in cell proliferation, survival, adhesion, migration and differentiation [1]. A number of researches demonstrated that overexpressions of EGFR and ErbB-2 has been observed in many cancer patients and correlate with a poor prognosis for several tumors including nonsmall-cell lung cancer (NSCLC), prostate, breast, stomach, colon, and ovarian cancers [2][3][4][5][6]. The blockade of EGFR and ErbB-2 has been clinically validated as an attractive approach for cancer therapy. As shown in Figure 1, the first-generation EGFR inhibitors Gefitinib (Iressa TM ) [4,7] and Erlotinib (Tarceva TM ) [8] were approved for non-small-cell lung cancer (NSCLC) therapy, and Lapatinib as a dual reversible EGFR/HER2 inhibitor was approved for breast cancer therapy [9]. Two candidate drugs Afatinib [10,11] and Dacomitinib [10,12], which could bond with a unique cysteine797 residue located at the lip of the ATP binding cleft [13], as the most promising second generation EGFR kinase inhibitors, have been being evaluated in respective phase III study. Moreover, the pyrrolo [3,2-d]pyrimidine derivative TAK-285 [14], of which hinge binder was similar to the 4-anilinoquinazoline core, has also been developed as dual EGFR/HER2 inhibitor.
During our research for potent EGFR tyrosine kinase inhibitors, we previously reported a series of novel 4-anilinoquinazoline derivatives, mainly focusing on the modifications of the substituent at the 6-and 7-positions of the quinazoline ring [15,16]. It is noteworthy that several known EGFR inhibitors (Figure 1), no matter ATP-competitive agents or ATP-noncompetitive/irreversible agents, need to share the basic structural motif of 4anilinoquinazoline, which has been identified as a most versatile template for kinase inhibition [17]. Much of structural alterations of these compounds presented mostly exist at the 4-, 6-, and 7positons of the basic quinazoline scaffold. Accordingly, a huge volume of researches on the synthesis, structure-activity relationships, and antiproliferative activities of the quinazoline-containing derivatives have been reported over the past ten years [18][19][20][21]. To our knowledge, the nitrogen at position 3 is important to EGFR inhibitory potency, but the nitrogen at position 1 is an even more important contributor [18]. Electron-donating substituents on the 6-and 7 positions could enhance the binding of N1 and N3, as do specifically substituted C4-anilines. In addition, structure activity relationships (SAR) in the quinazoline series have already demonstrated tolerance of bulky substituents at the 6-and 7positions of the quinazoline ring [19].
Among thousands of these 4-anilinoquinazolines, compound 1 and 2 empolying Lavendustin A subunit at the 6-position (Figure 2), reported by Albuschat et al [22] would be selected as the starting point for the development of dual EGFR/HER2 inhibitors in our study. Molecular docking studies based on EGFR and HER2 co-crystal structures (EGFR pdb code: 1M17, and HER2 pdb code: 3RCD) suggested that compound 1 would fit into the ATP-binding site of the HER2/EGFR protein ( Figure 3A, 3B, 3D, and 3E). Our analysis of Figure 3A and Figure 3B suggested that in addition to the optimal binding pose similar to Erlotinib, introduction of o-hydroxyl group of Lavendustin A subunit could form two hydrogen bond interactions with CYS773 of the EGFR protein, which implied that this moiety was of importance for the binding potency. However, comparing Figure 3D with Figure 3E, it was seen that the binding mode of compound 1 to the HER2 protein was not similar with that of TAK-285, although the two hinge binders (pyrrolo [3,2-d]pyrimidine of TAK-285, and 4-anilinoquinazoline of compound 1) were anchored at the same space of the HER2 protein. In order to develop novel 4-anilinoquinazoline derivatives as dual EGFR/ HER2 inhibitors, we envisioned that four kinds of modifications (A, B, C, D) presented in Figure 2 based on the structure of compound 1 would generate novel potent inhibitors.
In this manuscript, we described the design and structure activity relationships (SAR) of a series of 6-salicyl-4-anilinoquinazoline analogs as dual EGFR/HER2 inhibitors and as antitumor agents. Molecular docking studies demonstrated key molecular interactions of the 6-salicyl-4-anilinoquinazoline scaffold with EGFR and HER2, which would provide additional clues for the development of novel EGFR/HER2-targeting therapeutic agents.

Molecular docking evaluation
According to Figure 2 as well as considering chemical reagents available in our laboratory, 4-anilinoquinazoline compounds 7-25 ( Figure 4) were designed, and virtually evaluated by the means of the CDOCKER protocol [23] in the Accelrys Discovery Studio 3.5 suite. Two groups of molecular dockings between these novel 4-anilinoquinazoline analogs and two protein-ligand complexes (as for EGFR, PDB code: 1M17; as for HER2, PDB code: 3RCD) retrieved from the RCSB Protein Data Bank were performed, and the obtained results were plotted as a line-scatter graph and presented in Figure 5, describing the CDOCKER_INTERAC-TION_ENERGY of two sets of molecular dockings. Compared with two positive drugs (Erlotinib and TAK-285), respectively, most of the designed molecules below the blue dash lines would possess lower interaction energy, demonstrating that they are likely to exhibit more potent against both of the two protein targets. The docking study performed on the 1M17 protein crystal structure was represented by a black solid line and as clearly seen in the picture, compounds 9-12, 17-19, and 22-26 showed lower interaction energy than Erlotinib that reached up to 248.1441 kcal/mol. On the other hand, the docking study based on the 3RCD protein crystal structure also indicated compounds 10, 12, 14, 16, 21, 22 could exhibit more affinity for HER2 than TAK-285, of which the interaction energy reached up to 265.8378 kcal/mol. Therefore, the preliminary analysis would serve as a modest spur to induce us to probe these novel 4anilinoquinazoline compounds. Chemistry (For experiment details please see Chemistry experimental section) Here we described our efforts to optimize the 6-salicyl-4anilinoquinazoline series, mainly focusing on two aspects of modification: one is the replacement of substituents (include hydroxyl group) at the salicylic ring, and the other is the replacement of the -CH 2 -NH-linker with the carbonyl group between the quinazoline ring and the salicylic ring. A general synthetic approach was developed to prepare the 6-amino-4anilinoquinazolines (compounds 7 and 8) that can be easily generated and the corresponding process was displayed in Figure 4A [24]. Substitutions of 6-NH2 of the intermediates 7 and 8 by a series of salicylides with halogen groups led to 2-((4-(phenylamino)quinazolin-6-ylamino)methyl)phenol analogs 9-16 ( Figure 4B). Furthermore, in order to explore how much the coplanar effect of the quinazoline core ring and salicylic ring contributed to compounds' activities, these corresponding salicylic amides 17-25 described also by Figure 4B were synthesized via a simple amidation pathway. Obviously seen that the flexible moiety (-CH 2 -NH-) was replaced by the amide structure (-CONH 2 -) in these compounds, and along with compounds 9-16, all these compounds would be considered as probes used for the structure activity relationship (SAR) discussion.
Besides, we also noted the hydroxyl group of the salicylic ring and attempted to investigate what role it should play in the SAR study. We synthesized two new reference molecules 26 and 27: one's hydroxyl group was replaced by an ethyoxyl, and the other possessed no hydroxyl group. As presented in Figure 4C, the salicylic acid dissolved in DMF, followed by reacting with C 2 H 5 I at 80uC, was converted to the intermediate 2-ethoxybenzoic acid and finally treated with thionyl chloride to yield compound 26, and compound 27 from the starting material para-chlorobenzoic acid, was easily obtained by a simple amidation.

EGFR/HER2 enzyme assay
The synthesized target compounds 9-27 were tested for their inhibitory activities toward the EGFR and HER2 tyrosine kinases [14] and their activity data were summarized in Table 1. As expected, most of all the synthesized compounds showed moderate to potent inhibitory activities for these two protein kinases. Of these 4-anilinoquinazoline compounds, compound 21 was screened and obtained as the most potent dual EGFR/HER2 inhibitors. In order to better displaying and comparing each other, these results were plotted as a bar graph presented in Figure 6A. It composed of two color bars in which the black bar represented EGFR inhibitory activities of all the compounds, while the white one represented HER2 inhibitory activities of all the compounds. Comparing the white one to the black one, the amplitude of activity variation depicted by the later of which IC 50 values were in the range of 0.01527 mM, exhibited higher than that of the former of which the IC 50 values were in the range of 0.03-4.6 mM. Generally, compounds 9-20, 22-25 displayed less potency against HER2 inhibitory activities than EGFR inhibitory activities while the remaining three compounds 21, 26, 27 performed on the contrary. These results above demonstrated that 6-salicyl-4anilinoquinazolines derived from compound 1 indeed preferred to bind to EGFR than HER2, with the exception of compound 21 or disruption of the salicyl ring (compounds 26 and 27).
Although the black bar could roughly correlate with the white one, there would be some attentions when comparing the variation tendency of EGFR to that of HER2. In part of compounds 9-16 with -CH 2 -NH-linker, the IC 50 values of compounds against EGFR lowered than the level of 1 mM while that against HER2 mostly exceeded to 1 mM. Particularly, compound 14 that possessed 3,5-dichloro substituents performed better than the others, indicating that for inhibitory potency of EGFR and HER2 kinases, the chlorine atom as one substituent was likely to be superior to the bromine atom and disubstituted inhibitors were better than monosubstitutions. In addition, comparing these inhibitors 9-16 with -CH 2 -NH-linkers to those 17-25 containing -CO-NH-linkers, as was seen that that the flexible C-N bond was replaced by the rigid C = N bond can not radically influenced the corresponding activity, demonstrating that the coplanar effect between the quinazoline core and the salicylic ring was not important for the SAR of 6-salicyl 24anilinoquinazolines.
Among compounds 17-27, compound 21 showed the most potent inhibition toward EGFR and HER2, of which IC 50 values can reach up to 0.12 and 0.096 mM, respectively. Figure 6A showed that 21 could be considered as a turning point among these eleven analogs, before which compounds 17-20 with the Br atom on the aniline ring exhibited moderate activities against both of the two target proteins. Besides, it can be noted that the OH on the salicylic skeleton actually increased the inhibitory activity, comparing 17 to 27. The deduction that the pseudo sixmembered ring formed [25] through the intramolecular hydrogen bond between OH and O = C in the salicylic acid probably account for this issue above. The last one worth mentioning was that 22 with the methyl substituent is superior to 23 with the Cl atom and compounds with substitutions at the meta (18-20, 25) position showed less potent activities than those with substitutions at the para position (17,21,23,24). Thus, the primary factor for differences exhibited in the level of inhibitory activity of these

Antiproliferation assay
The target compounds were also evaluated in in vitro antiproliferation assays against three human cancer cells shown in Table 2: A549 (carcinomic human alveolar basal epithelial cell), MCF-7 (breast cancer, with Her2/neu protein overexpression)  and A431 (overexpression of EGFR). As expected, due to different types of cancer cells correlated with EGFR/HER2 overexpression, the data revealed most compounds could perform better against MCF-7 cells and A431 cells than A549 cells. Among these compounds, compounds 7, 9, 12, 13, 19-24 could meanwhile inhibit MCF-7 and A431 at the level of IC 50 values less than 1 mM. In accord with the kinase inhibition, all the compounds, particularly from 17 to 25, possessed the same trend with enzyme inhibitory activity. Strikingly, the most potent small molecule screened via these antiproliferation assays was still 21 (MCF-7 and A431, IC 50 = 0.49 mM and IC 50 = 0.67 mM, respectively). On the basis of the results above, in order to get the cellular activity profile of salicyl derivatives of 4-anilinoquinazoline against several known lapatinib-resistant cell lines, we select the most potent compound 21 and lapatinib as the positive control for futher biological evaluation as a potential dual EGFR/HER2 inhibitor. The screening results of 21 were obtained in Table 3 by means of three drug-resistant cancer cell lines that expressed or harbored genetic information with resistance to lapatinib. Comparing with the positive control, 21 showed a better profile of inhibiting the growth of these corresponding cancer cell lines. Particuarly it showed excellent growth inhibitory activity with IC 50 of 0.77 mM in lapatinib-resistant H1781 cell line (IC 50 (lapatinib) = 2.1mM). However, compound 21 as similar with lapatinib, exhibited weak inhibitory activity against H1975 cell line.

Kinase selectivity
Since 4-anilinoquinazoline analogs exhibited widely kinase inhibitions [26], the selectivity profile of compound 21 was assessed against several key kinases close to EGFR/HER2 ( Figure 6B, for detailed data please see Table S1). The data confirmed the generally good selectivity profile of 21 but also identified some activity against CDK2, with an IC 50 value of 3.5 mM. In general, compound 21 bearing a tri the 4-position of the salicylic ring, which was proposed to fit the ATP binding pocket very well, showed more than 100-fold selectivity against major of other kinases screened.

Molecular docking
Docking study was performed to fit compound 21 into the active center of the two protein-ligand complexes of the epidermal growth factor family (PDB code: 1M17 and 3RCD, respectively). The obtained results were presented in Figure 3C and 6F, which showed the optimal binding mode of compound 21 interacting with the 1M17 protein and the 3RCD protein, respectively. Compared with Figure 3A and 3B, Figure 3C displayed that although the binding mode of compound 21 to EGFR kinase protein seemed to differ from that of the control Erlotinib and compound 1, the CDOCKER_INTERACTION_ENERGY of 21 was up to 249.3307 kcal/mol, lower than 247.3559 kcal/ mol, that of compound 1. When paying attention to Figure 3F, in addition to a similar manner to compound 1, the binding mode of 21 to the HER2 protein was further optimized, in which multiple interactions that three key hydrogen bonds, one Pi-cation bond, and one Pi-Sigma bond could be provided clearly. Of three hydrogen bond interactions, two adjacent hydrogen bonds were formed between one fluorine atom of the trifluoromethyl moiety of 21 and two residues PHE731 and ALA730. The other hydrogen bond was formed between the guanidyl of LYS753 and the oxygen atom of the hydroxy group of 21. One Pi-cation bond was located between the guanidyl of LYS753 and the aniline ring while the Pi-Sigma bond was formed by the quinazoline ring and VAL734. Insight into this picture, two amino acid residues SER783 and MET801 located in the hinge binding pocket seemed very important for the active conformation of compound 21. Besides, the trifluoromethyl moiety, the hydroxy group, and the ketonic oxygen of 21 could contribute to stabilizing the binding complex. These results could provide a molecular level foundation for 21 as the most potent dual inhibitor.

Conclusions
In summary, a series of 6-salicyl-4-anilinoquinazoline analogues 9-27 have been synthesized and showed modest to potent inhibitory potency against EGFR/HER2, with IC 50 values ranging from 7.7 mM to 0.096 mM. Among these small molecules, compound 21 that possessed a para substituent of trifluoromethyl at the salicylic ring exhibited potent EGFR and HER2 kinase inhibitory activity with an IC 50 of 0.12 mM and 0.096 mM, respectively, comparable to the control compounds Erlotinib and Lapatinib. The corresponding cellular activities of 21 against three lapatinib-resistant cancer cell lines were superior to the positive control lapatinib. Furthermore, molecular modeling confirmed this kind of novel binding modes of 21 to the EGFR/HER2 proteins, mainly due to the contribution of trifluoromethyl moiety. These wet results, along with molecular docking observations, could provide an important basis for further development of compound 21 as a potent tyrosine kinase inhibitor.

Materials and Methods
All of the synthesized compounds were chemically characterized by thin layer chromatography (TLC), proton nuclear magnetic resonance (1H-NMR) and elemental microanalyses (CHN). 1 H NMR spectra were measured on a Bruker AV-300 or AV-500 spectrometer at 25uC and referenced to Me 4 Si. Chemical shifts are reported in ppm (d) using the residual solvent line as internal standard. Splitting patterns are designed as s, singlet; d, doublet; t, triplet; m, multiplet. ESI-MS spectra were recorded on a Mariner System 5304 Mass spectrometer. Elemental analyses were performed on a CHN-O-Rapid instrument and were within 60.4% of the theoretical values.  Melting points were determined on a XT4 MP apparatus (Taike Corp., Beijing, China) and are as read. Analytical thin-layer chromatography (TLC) was performed on the glass-backed silica gel sheets (silica gel 60Å GF254). All compounds were detected using UV light (254 nm or 365 nm).
6-Nitroquinazoline 6 (2.0 g, 6.65 mmol, 1.0 equiv), iron (2.5 g, 45 mmol, 6.7 equiv) and acetic acid (6 ml, 90 mmol, 13.5 equiv) were suspended in aqueous ethanol (180 ml, 77.8% v/v) and heated at reflux about 70-80uC for 5-6 h. Meantime this mixture would be stirred for one minute every half hour with a glass rod and the yellow solution would become reddish-brown slowly. The reaction mixture was cooled to room temperature and alkalinized by addition of concentrated ammonia (40 ml). Insoluble material was removed by filtration through Celite, and the filtrate was evaporated under reduced pressure. The resulting solid was extracted with ethyl acetate for column chromatography.  (7) was also prepared as described for 8.
General procedure for the synthesis of 9-16.
Different substituent salicylal (100 mg) was added to a solution of the corresponding (7 or 8, 60 mg) in EtOH (20 ml). The mixture was stirred at 40-60uC for 4-6 h. In the next step, sodium borohydride (100 mg) was added into the liquid reactant, and the reaction mixture was stirred at room temperature for 1 d. The reaction mixture was evaporated under reduced pressure and the solid precipitated was dissolved with water (40 ml), extracted with ethyl acetate about three times. The extraction liquid was purified by a v) to give the target compounds as follows, which yields were between 60-80%.
General procedure for the synthesis of 17-25 The starting materials (salicylic acids) for the synthesis of amides should be activated in the first procedure: acid (100 mg) and  flash chromatography with EtOAc/petroleum ether (2:1, v/ SOCl 2 (6-10 ml) was mixed and stirred at reflux 80uC for 4 hours. The reaction mixture was cooled and evaporated to give reactive acyl chloride obtained as a kind of oil, which would be dissolved in ethyl acetate (5-6 ml) in the next step. A solution of acyl chloride dissolved with ethyl acetate was added dropwise to compound 7 or 8 (0.5 mM) in ethyl acetate containing potassium carbonate (300 mg) at 0uC with constant stirring overnight. The reaction mixture was then poured in excess of diluted NaOH and extracted with EtOAc. The extraction liquid was purified by a ash chromatography with EtOAc/petroleum ether (3:1, v/v) to give these amides as follows: the yields were between 30-60%.
General procedure for the synthesis of 26 and 27 Ethyl iodide (C 2 H 5 I, 2.4 ml) was added into the mixture containing salicylic acid (2 g) and potassium carbonate (4 g) in DMF (40-60 ml), and then the reaction mixture was stirred at 80uC for 6 h. The reaction mixture was poured in ice water and extracted with methylene dichloride. The solvent was removed in vacuo to obtain the corresponding salicylate as oil.
The total oil was added into the miscible liquid (NaOH, 4 g; EtOH, 30 ml; H 2 O, 50 ml) and was refluxed gently with stirring overnight. Adjusting pH value to 7 with hydrochloride, the 2ethoxybenzoic acid was precipitated in the solution, filtered off to obtain a white solid (1.2 g). The next two steps of the preparation for 24 and 25 were the same as Figure 4B 13