Discovery of boron-conjugated 4-anilinoquinazoline as a prolonged inhibitor of EGFR tyrosine kinase†
Boron-conjugated 4-anilinoquinazolines were designed and synthesized as inhibitors of EGFR tyrosine kinase with possible covalent bond interactions between the boron atom and the nucleophilic groups of the EGFR kinase domain. Among the compounds synthesized, compounds 6c, 7b, and 7d reduced the EGF-mediated phosphorylation of EGFR tyrosine kinase and its downstream kinases including ERK and Akt in A431 cells. The cell growth was inhibited by these compounds through arrest of G1 cell cycle, which induced apoptosis. A time-dependent in vitro preincubation assay demonstrated the irreversible inhibition of compound 7d against EGFR tyrosine kinase. Quantum mechanical docking simulation revealed that the boronic acid moiety of compound 7d formed a covalent B–O bond with Asp800 in addition to hydrogen bonds with Asp800 and Cys797, which may cause the prolonged inhibition of compound 7d toward EGFR tyrosine kinase.
Introduction
The signal transduction pathway through growth factor receptor tyrosine kinases is important for cell proliferation. Deregulation of signaling pathway has been observed in many human tumors, and thus these kinases have been investigated as potential targets for cancer therapy.1 Epidermal growth factor receptor (EGFR) (also known as erb-B1 or Her-1) is a member of ErbB family of receptor tyrosine kinases, and the uncontrolled activation of EGFR-mediated signaling may be due to overexpression of the receptors in numerous tumors including head and neck, lung, breast, bladder, prostate, and kidney cancers, mutations resulting in constitutive activation in brain tumors, or overexpression of the ligands including EGF or TGF-a in pancreatic, prostate, lung, ovary, and colon cancers.2,3
In 1994, Fry and co-workers discovered 4-anilinoquinazoline (PD 153035) as a specific inhibitor toward EGFR tyrosine kinase.4 Since their discovery, various 4-anilinoquinazoline derivatives have been synthesized, and ZD-1839 (IressaTM),5,6 and OSI-774 (TarcevaTM)7,8 have been developed as reversible binding inhibitors of EGFR kinase activity and recently approved for non-small-cell lung cancer (NSCLC) therapy. Furthermore, Lapatinib has been investigated as a dual reversible ErbB inhibitor with potent activity against EGFR and Her-2 kinases, and approved for breast cancer therapy.9,10
However, despite the benefits of Iressa and Tarceva for treatment of NSCLC, most recurrent patients ultimately develop acquired resistance to these agents. Recently, a mutation, T790M, within the EGFR kinase domain has been found in ~50% of these patients, and therefore much attention has been paid for potency of irreversible EGFR inhibitors, which have proven to be effective for the EGFR tyrosine kinase with T790M mutation.11,12 Among the various 4-anilinoquinazoline derivatives synthesized,13–15 EKB- 569 (phase II),16,17 HKI-272 (phase I),18 and CI-1033 (phase I)19,20 have been investigated as irreversible inhibitors. These compounds have a Michael acceptor function, as a common structure, substituted at the 6-position of the quinazoline ring to trap the Cys-797 (Cys773 in an alternative numbering scheme) of EGFR, which may cause the irreversible inhibition. According to X-ray structural analysis of a complex with Tarceva, the thiol of Cys-797 in fact is located near the methoxyethoxy group substituted at the 6-position of Tarceva,21,22 as shown in Fig. 1.
We focused on the substituent at the 6-position and designed boron-conjugated 4-anilinoquinazolines. A boron atom has a vacant orbital and interconverts with ease between the neutral sp2 and the anionic sp3 hybridization states, which generates a new stable interaction between a boron atom and a donor molecule through a covalent bond. Recently the X-ray crystal structure of the 20S proteasome in complex with bortezomib was reported.23 In this structure, boronic acid of bortezomib covalently interacts with the Thr-1 hydroxyl in the active site of the 20S proteasome, forming the hybridized borate. Bond energies of C–S, B–S, and B–O s-bonds have been estimated as 65, 24, and 18 kcal/mol, respectively,24,25 and that of hydrogen bonding is 3–10 kcal/mol.24,26 Therefore, if we can introduce a boron atom into the 6-position of the quinazoline framework via a linker of suitable length, as illustrated in Fig. 1A, a covalent bond interaction would be expected between the boron atom and the thiol of Cys797 or the carboxyl group of Asp800. This interaction, which is stronger than the hydrogen bond but not as strong as the C–S or C–O covalent bond,27 may affect the prolonged inhibition of EGF-stimulated signal transduction.
In this paper, we have introduced a boron moiety into the anilinoquinazoline framework via various lengths of linkers, such as methylene (1.68 A˚ ), propenyl (3.90 A˚ ), and benzyl (5.81 A˚ ) groups (Fig. 2). We synthesized various boron-conjugated 4-anilinoquinazolines and investigated their selective inhibition of EGFR kinase, inhibition of the EGFR-stimulated signaling pathway, and effects on cancer cell growth and cell cycle, as well as their irreversible interaction properties.
Furthermore, in order to analyze the EGFR-inhibitor binding forms and their interactions in detail, molecular simulations were carried out based on quantum mechanical (QM) methods. Classical molecular mechanical (MM) approaches based on an empirical force field have been widely used for molecular simula- tions of protein properties such as enzyme–inhibitor interactions. However, such an approach is not sufficiently accurate in the description of the nature of the interactions between molecules. Also, the force field of sp3-hybridized boron atom has still not been established. On the other hand, a QM approach based on ab initio molecular orbital (MO) calculations has been successfully used to understand the structure and chemical properties of molecules including coordinate linkages of sp3-hybridized boron atoms, and therefore the QM approach has potential for accurate representation of molecular interactions that could aid drug design.28 Here, we have investigated docking geometries between the EGFR and inhibitor by hybrid QM/MM calculations in the ONIOM scheme,29,30 employing a Hartree–Fock (HF) method for the QM part and a UFF force field for the MM part. Binding energies between the EGFR and inhibitors were also evaluated by full QM treatment using ab initio fragment molecular orbital (FMO) methods31–34 at the HF level. This FMO approach has been successfully used to evaluate molecular interactions of bio- macromolecules such as a binding affinity of ligands to the human estrogen receptor.35,36
Results and discussion
Chemistry
The synthesis of the boron-conjugated anilinoquinazolines 6a–d and 7b–d is shown in Fig. 2. Briefly, 5-hydroxyanthranilic acid 1 was converted to the quinazolinone 2 in 70% yield in three steps according to the reported procedure.37 Protection of the hydroxy group of 2 with acetic anhydride gave the 6-acetoxyquinazoline 3, which underwent chlorination with POCl3 to give the 4-chloroquinazoline 4 in 66% yield. Displacement of 4 with 3-chloroaniline followed by deprotection of the acetyl group afforded the 6-hydroxy-4-anilinoquinazoline 5, quantitatively.38 Various boronic ester groups, such as (RO)2BCH=CHCH2Cl (8 and 9), (RO)2BCH2Br (10),39 and (RO)2BC6H4CH2Br (13),
were synthesized and introduced into the 6-position of the preferable to the corresponding boron esters due to their steric and hydrophilic properties. In contrast, boronic acid 7c showed less potent inhibition because the linker between the boronic acid and the quinazoline ring may be too short to occupy the hydrophilic region.
We next examined the effects of the boron-conjugated 4-anilinoquinazolines on the EGF-induced tyrosine phosphory- lation of EGFR and the signaling cascades in A431 cells by immunoblot analysis. Treatment of A431 with EGF (10 ng/ml) rapidly induced autophosphorylation of EGFR, and the level of the phosphorylation reached a maximum at 10 min after EGF stimulation (data not shown). Under these conditions, the boron- conjugated 4-anilinoquinazolines potently suppressed the EGF- induced phosphorylation of EGFR at 1 mM concentration of compounds (Fig. 3A).
It has been reported that the autophosphorylation of EGFR tyrosine kinase activated various downstream kinases including ERK and Akt, which play an important role in the regulation of cell proliferation and apoptosis, respectively.42 Therefore, we examined the effects on the EGFR-dependent activation of downstream signaling pathways. As shown in Fig. 3B, the boron- conjugated 4-anilinoquinazolines, except compounds 6d and 7c, also significantly suppressed the EGF-induced phosphorylation of ERK and Akt in parallel with the inhibition of EGFR autophosphorylation. These results indicate that the boron- conjugated 4-anilinoquinazolines also induce the inhibitory effect on autophosphorylation of EGFR tyrosine kinase in cells as well as EGFR kinase in vitro, and arrest the downstream signaling pathway.
To demonstrate whether the inhibitory activity of the boron- conjugated 4-anilinoquinazolines toward EGFR tyrosine kinase is implicated in cell growth, we examined the effects of the compounds on proliferation of various human carcinoma cell lines (A431, HepG2, KatoIII, and SKBR3). The cells were incubated with compounds at various concentrations for 72 h and their cell growth inhibition was determined by MTT assay. As shown in Fig. 3C, the boron compounds reduced the proliferation of A431 cells, which overexpress EGFR, and their GI50 values were in the range 0.54–2.59 mM. Lower cell growth inhibitions were observed in other cell lines, which express a low level of EGFR. These results indicate that the selective inhibition of A431 cell growth by the boron-conjugated 4-anilinoquinazolines is a consequence of the inhibition of EGFR tyrosine kinase.
Cell cycle G1 arrest and induction of apoptosis
To characterize the cell growth inhibition induced by the boron- conjugated 4-anilinoquinazolines, effects of compounds on cell cycle were investigated by flow cytometry analysis. Treatment of A431 cells with compounds 6c, 7b, and 7d at 1 mM for 24 h led to profound changes in cell-cycle profiles as shown in Fig. 4A. The boron-conjugated 4-anilinoquinazolines induced a massive accumulation of cells in the G1 phase and decreased the proportion of cells in G2/M phase. Tarceva also induced G1-arrest with a similar potency. To further characterize the cell-death mech- anism induced by the boron-conjugated 4-anilinoquinazolines, a biparametric cytofluorimetric analysis was performed using FITC-labeled annexin V and propidium iodide (PI), which stain phosphatidylserine residue and DNA, respectively.43 A431 cells were treated with the compounds for 24 h and then stained with PI and FITC-labeled annexin V. The resulting green (FITC) and red (PI) fluorescence was monitored by flow cytometry. As indicated in Fig. 4B, the number of annexin V-positive and PI-negative cells increased by the treatment with compounds 6c, 7b, 7d, and Tarceva (the area labelled D). These results indicate that the boron-conjugated 4-anilinoquinazolines possess not only antiproliferative but also apoptotic activities in the A431 cell growth inhibition through the suppression of EGFR tyrosine kinase activity.
Prolonged inhibition by irreversible binding to EGFR tyrosine kinase in cells
The prolonged inhibition ability of the boron-conjugated 4-anilinoquinazolines toward EGFR tyrosine kinase was inves- tigated by experiment using A431 exposed to the compounds and washed prior to the assay.15 After continuous incubation of A431 cells with the boron-conjugated 4-anilinoquinazolines at 1 mM or Tarceva at 0.3 mM for 1 h followed by EGF stimulation, each compound inhibited autophosphorylation of EGFR in A431 cells, as shown by immunoblot analysis (Fig. 5A). When the cells were washed to remove unbound compounds from the cell-medium and incubated for an additional 5 h prior to EGF stimulation, the activity of EGFR autophosphorylation was restored, while for cells treated with compound 7d, the autophosphorylation was still inhibited. These results suggest that compound 7d irreversibly inhibited to EGFR tyrosine kinase, whereas other compounds, such as 6c, 7b, and Tarceva, reversibly interacted to EGFR tyrosine kinase. Furthermore, the preincubation time- dependent inhibition of compound 7d against the EGF-induced phosphorylation of EGFR was observed in A431 cells (Fig. 5B). These results also support the irreversible inhibitory property of compound 7d. We next investigated the prolonged inhibition implicated in continuous reduction of cell growth. As shown in Fig. 5C, MTT assay revealed that compound 7d showed retentive reduction of cell growth after removal of compounds from the cell- medium by washing even though its inhibitory activity was not potentially high in comparison with other compounds, whereas the inhibitory effects of other compounds including Tarceva were essentially lost upon washing. These results indicate that continued reduction of cell growth by compound 7d may be due to the irreversible inhibition of EGFR tyrosine kinase. Fig. 5D shows dose-dependent growth inhibition curves of cells treated with compound 7d with and without washing, and similar values for GI50 were observed in both cases. It was reported that the lack of autophosphorylation sites resulted in reduction of internalization and degradation rates of EGFR.44 Therefore, the irreversible in- hibition of EGFR autophosphorylation by compound 7d induces prolonged reduction effects of internalization and degradation rates of EGFR, which may lead to the retentive reduction of cell growth.
In vitro irreversible inhibition of EGFR tyrosine kinase activity
To clarify the irreversible inhibition of EGFR tyrosine kinase by the boron-conjugated 4-anilinoquinazoline in detail, in vitro time-dependent inhibition assay was performed.45 As shown in Fig. 6, EGFR irreversible inhibitor PD168393 at 10 nM inhibited EGFR tyrosine kinase activity in a time-dependent manner; however, the inhibitory activity of reversible inhibitor Tarceva at 10 nM plateaued after incubation for 5 to 15 min. Under these conditions, compound 7d inhibited EGFR tyrosine kinase activity in a concentration- and time-dependent manner (Fig. 6). These results suggest that the boron-conjugated 4-anilinoquinazoline 7d irreversibly suppresses EGFR tyrosine kinase.
Recently, the X-ray cocrystal structure of EGFR and PD168393 was reported and a covalent bond has been observed be- tween the acrylamide Michael acceptor on PD168393 and the targeted Cys797 in the complex (PDB code: 2JF5).21 We tried to detect the Cys797-containing peptide fragment VQLITQLMPFGCLLDYVREHKD, which was generated by treated with tryps¯in digestion of EGFR kinase domain, by ESI-TOF and MALDI-TOF mass spectroscopies in order to clarify the covalent bond interaction of compound 7d and EGFR. However, this peptide fragment was not ionized enough to be detected by the mass spectroscopy under our tested conditions. In fact, no report has been found for detection of this peptide fragment by mass spectroscopic analysis. Fry and coworkers reported that the covalent conjugation of Cys773 in the fragment EILDEAYVMASVDNPHVCR with PD168393 was detected by ESI mass spectroscopy.14 How¯ever, Cys773 is located near the aniline ring of both inhibitors Tarceva and PD168393 in the reported cocrystal structures.21,22 Therefore, the report for detection of the covalent conjugation of Cys773 and PD168393 has exhibited the possibility of alternative binding mode of PD168393 in EGFR kinase domain. Since mass-spectroscopic analysis was not been suitable for clarification of the covalent bonding interaction of Cys797 and compound 7d, we next demonstrated molecular docking and quantum mechanical simulation to predict the plausible binding mode of compound 7d in the EGFR kinase domain.
Molecular docking and quantum mechanical simulation
We performed a molecular docking simulation of compound 7d with the ATP-binding site of EGFR kinase domain (PDB code 1M17). The initial structure of protein backbone was taken from the X-ray structure of EGFR kinase domain in complex with Tarceva.22 A conformation search for three rotatable bonds of compound 7d (Fig. 7A) was performed to examine the boron docking mode toward surrounding residues. As shown in Fig. 7B, 65 conformers were observed and classified into three clusters. The boronic acid of compound 7d was located around Cys797 or Asp800 in the clusters 1 and 2, respectively, suggesting interactions of the boronic acid with such residues. Based on these two conformations, three docking models (models 1–3) were constructed for further calculations. The initial structure of model 1 was selected from the cluster 2, and those of model 2 and model 3 were selected from the cluster 1. The thiol group of Cys797 was in neutral form (–SH) for model 1 and model 2, and negatively charged form (–S-) for model 3.
The docking structure of Tarceva (X-ray structure) and the three models of compound 7d were further optimized by using the ONIOM-QM/MM simulation at the HF/6-31G(d):UFF level of theory. As shown in Fig. 8, the QM region includes both interaction sites of quinazoline ring and boronic acid moi- eties: binding site residues of EGFR, a hydrogen-bonding water molecule, and inhibitor. Fig. 9 shows the three docking modes of the compound 7d (models 1–3) into the active site of EGFR kinase domain overlaid with the binding conformation of Tarceva (schematic figures and geometrical values for each docking mode are available in the ESI†). According to the simulation, the 4-anilinoquinazoline moieties of both compound 7d and Tarceva bind to the kinase domain in a similar manner. In all binding structures, three hydrogen bonds were observed as well as the X-ray structure: one direct hydrogen bond between the nitrogen (N1) of the quinazoline ring and the amide nitrogen of Met793 (N–H lengths 2.28–2.44 A˚ ); two hydrogen bonds mediated by single water molecule among the nitrogen (N3) of the quinazoline ring, the thiol hydrogen of Cys775, and hydroxyl oxygen of Thr790. The corresponding calculated values are 2.28–2.40 A˚ between N3 and the water hydrogen; 2.32–2.65 A˚ between the thiol hydrogen of Cys775 and the water oxygen; 2.02–2.05 A˚ between the hydroxyl oxygen of Thr790 and the water hydrogen. Also, the chloro group on the aniline ring of compound 7d is located in a very similar position to the ethynyl group of Tarceva. In the binding structures of compound 7d (models 1–3), additional bonds were observed at the boronic acid site. Model 1 indicated a covalent B–O bond between the sp3-hybridized boron atom and the carboxyl oxygen atom of Asp800. The calculated B–O distance is 1.58 A˚ , which is longer than other two B–O lengths in the boronic acid (1.44 and 1.42 A˚ ). In contrast, the hybridization state of boron atom was sp2 for models 2 and 3. A strong hydrogen bond between hydroxyl hydrogen of boronic acid and carboxyl oxygen atom of Asp800 was observed, with lengths of 1.60 and 1.73 A˚ for models 2 and 3, respectively. A second hydrogen bond was also observed in model 3 between the other hydroxyl hydrogen of boronic acid and thiol ion S- of Cys797, with a length of 2.43 A˚ . Therefore, the interaction between the EGFR and inhibitor at the boronic acid site is even stronger than that of the quinazoline ring site, and this covalent interaction may induce the prolonged inhibition of compound 7d toward EGFR tyrosine kinase.
Conclusions
In the present study, we designed boron-conjugated 4-anilino- quinazolines as inhibitors of EGFR tyrosine kinase with possible covalent bond interactions between the boron atom and the nucleophilic thiol of the EGFR kinase domain. Selective inhibition of boron-conjugated 4-anilinoquinazolines was observed at 1 mM; compound 6c exhibited significant inhibition of EGFR tyrosine kinase with an IC50 value of 0.19 mM. Boron compounds also reduced the proliferation of A431 cells, selectively, with GI50 values of 0.53–2.42 mM. Among the compounds synthesized, compounds 6c, 7b, and 7d significantly suppressed the EGF- induced phosphorylation of ERK and Akt in parallel with the inhibition of EGFR autophosphorylation in A431 cells, and compound 7d was found to possess prolonged inhibitory activity toward autophosphorylation of EGFR tyrosine kinase at 5 h after cell wash, although inhibition of the cell growth was not particularly high in comparison with other compounds. Flow cytometry analysis of compounds 6c, 7b, and 7d revealed that these compounds induced G1 arrest in cell cycle and apoptosis. Docking studies suggest that the quinazoline ring of compound 7d is bound to the EGFR kinase domain in a very similar manner to that of Tarceva. The boronic acid moiety of compound 7d is shown to be able to form a covalent B–O bond with Asp800 and hydrogen bonds with Cys797 and Asp800, which may cause the prolonged inhibition of EGFR tyrosine kinase by compound 7d. Although the acidity of each boronic acid and physico-chemical nature of aromatic and aliphatic boronic acids should affect the binding to the protein, these results suggest that boron has potential in pharmaceutical drug design for prolonged interaction to target proteins through covalent bonds.
Cell growth assays
The human carcinoma cell lines, A431 (epidermoid carcinoma), HepG2 (liver carcinoma), KatoIII (gastric carcinoma), and SKBR3 (breast carcinoma) were used for the cell growth assay. Each cell line (5 ¥ 103 cells/well of 96-well plate) was incubated at 37 ◦C for 72 h in 100 ml of RPMI-1640 medium containing various concentrations of the boronic acid- conjugated 4-anilinoquinazolines. After the incubation, 10 ml of 3¢-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT, Sigma) in PBS (5 mg/ml) was added into the each well, and the cells were further incubated at 37 ◦C for 4 h. After the removal of the medium, 100 ml of DMSO was added, and the absorbance at 570 nm was determined. For cell-based wash experiment,15 A431 cells were incubated for 18 h with compounds and then washed with fresh medium. After further incubation for 54 h, the rates of viable cells were determined by MTT assay. The drug concentrations required to inhibit cell growth by 50% (GI50) were determined from semilogarithmic dose–response plots.
Immunoblot analysis
After stimulation for the specified period, the cells were washed three times with PBS, dipped in 100 ml of ice-cold lysis buffer (20 mM HEPES, pH 7.4, 1% Triton-X 100, 10% glycerol, 1 mM sodium vanadate, 5 mg/ml of leupeptin, and 1 mM EDTA) for 15 min, and disrupted with a Handy Sonic Disrupter, and the lysate was boiled for 5 min in a sample buffer (50 mM Tris, pH 7.4, 4% sodium dodecylsufate (SDS), 10% glycerol, 4% 2-mercaptoethanol, and 0.05 mg/ml of bromophenol blue) at a ratio of 4:1. The proteins were separated by 10% SDS- polyacrylamide gel electrophoresis and transferred onto a PVDF membrane (Millipore). For the immunoblotting, antibodies used were phosphor-EGFR(Tyr1173) antibody (Santa Cruz), EGFR- antibody (Santa Cruz), phospho-Akt (Ser473) antibody (Cell Signaling, Beverly, MA), and phospho-p44/42 MAP kinase (Tyr204) antibody (Santa Cruz). After incubation with HRP- conjugated secondary antibody (Santa Cruz), the blot was reacted with ECL kit (GE healthcare), and the levels of each protein were visualized by a ChemiDoc XRS image analyzer (Bio-Rad).
Cell cycle analysis
After incubation of A431 cells with the drugs, the cells were washed with PBS, and fixed with 70% ethanol for 2 h at 4 ◦C. The cells were incubated for 30 min at 37 ◦C in 1 ml of RNase solution (0.25 mg/ml in PBS), and further incubated for 30 min at 4 ◦C in propidium iodide staining solution (PI, 0.05 mg/ml in PBS). The suspension was then passed through a 40 mm nylon mesh filter and analyzed by a Cytomics FC500 flow cytometer (Beckman Coulter).
Detection of apoptosis
Phosphatidylserine externalization was measured using an An- nexin V-FITC kit (Beckman Coulter) according to the manufac- turer’s instructions. After incubation of A431 cells with the drugs, the cells were suspended in 100 ml of 1X binding buffer, and added 5 ml of annexin V-FITC (5 mg/ml) and 2.5 ml of PI (250 mg/ml in PBS). The cell suspensions were incubated for 10 min at 4 ◦C and analyzed by a flow cytometer.
Molecular docking and quantum mechanical simulations:
Docking and binding energy calculations were carried out based on the X-ray structure of EGFR kinase domain in complex with Tarceva (PDB code: 1M17).22 Based on this structure, the entire kinase domain of the receptor protein (residues 672–964), inhibitor, and a water molecule were selected for simulations. All water molecules in the crystal structure were removed except one molecule which mediate hydrogen bond network between the protein and inhibitor (Fig. 8). Hydrogen atoms were added to the protein, considering the terminal amino acid and all charged amino acid residues the charged state. The positions of added hydrogen atoms were minimized using MMFF94x force field using the Molecular Operating Environment (MOE) software (Chemical Computing Group Inc.). The binding site of the boron-conjugated 4-anilinoquinazoline 7d in EGFR kinase domain was defined based on that of Tarceva in the 1M17 X- ray structure. As the initial structure, the quinazoline ring of compound 7d was superimposed on that of Tarceva retaining hydrogen bond network with the EGFR. Docking structures were calculated in two steps. First, systematic conformation search of compound 7d in the EGFR kinase domain was performed for three rotatable bonds (Fig. 7) using the MOE software. Then two conformations were selected, in which boronic acid moiety was located at the Asp800 (model 1) or Cys797 (model 2 and model 3) interaction sites, followed by the MMFF94x minimization of the inhibitor structure at the EGFR binding site. Second, the detailed docking geometries for both Tarceva and compound 7d were further optimized by the hybrid QM/MM simulation of quantum mechanics (QM) and molecular mechanics (MM) methods. The QM/MM calculations were carried out employing the ONIOM scheme29,30 as implemented in GAUSSIAN0346 using the universal force field (UFF)47 for the MM region. The Hartree– Fock (HF) method using the 6-31G(d) basis set was chosen as the high level of theory for the QM region. The QM region consisted of side chains of Cys775, Thr790, Cys797, and Asp800, backbone of Met793, single water molecule, and the inhibitor (Fig. 8). Both side chains and backbones were capped at the Ca positions by using hydrogen link atoms, and the thiol group of Cys797 was either the –SH or the –S- form. Other residues in EGFR were treated with the UFF. With these optimized docking geometries, the binding energies between EGFR and inhibitors were calculated using the ab initio fragment molecular orbital (FMO) method at the HF/6-31G(d) level of theory. The FMO method is a fully QM approach and details of this method have been described elsewhere.31–34 The approximations of electrostatic potentials considered as the Mulliken orbital charge (esp-aoc) and the fractional point charge (esp-ptc) were carried out when the distance between the closest contact atoms in the two fragments exceed 0.0 and 2.0 in the units of van der Waals radii (vdW), respectively. In addition, the Coulomb interaction approximation (dimmer-es) was applied for the fragment pair calculations in two fragments with separations exceeding 2.0 in the same vdW unit. The fragmentation of the system was as follows: each amino acid residue of EGFR, the inhibitor molecule, and the water molecule were treated as a single fragment. All the FMO calculations were performed BI-4020 with the ABINIT-MP program (available at http://www.ciss.iis.u-tokyo.ac.jp/rss21/).