This article is mentioned in:

Introduction

Non-small cell lung cancer (NSCLC) is the leading cause of cancer-associated mortality in both men and women worldwide (1). In China, there has been a drastic increase in lung cancer-associated mortality since 1970 (2). At present, treatments for NSCLC are marginally effective (3) and the primary target for treatment of NSCLCs is the epidermal growth factor receptor (EGFR) pathway (4). Other than genetic factors, patient characteristics including gender, race and lifestyle influence the effectiveness of the treatment for NSCLC (5). The most prominent genetic factor associated with the responsiveness to NSCLC treatment is a mutation in EGFR (6,7). Mutations in EGFR mediate the carcinogenic effect by manipulating the apoptotic mechanism (8). Tyrosine kinase inhibitors (TKIs) are the first line of treatment of cases featuring mutated EGFR (9). The initial response to TKIs is limited by acquired resistance and the T790M mutation in EGFR has previously been demonstrated to have this acquired resistance (10). A previous study revealed that the role of the T790M mutation is associated with an increase in the affinity of adenosine triphosphate (ATP) to mutated EGFR (11). Naturally derived compounds are being developed to treat NSCLC featuring a T790M mutation (12). Furthermore, computer-aided drug design has become a prominent tool in drug discovery (13). In the present study, an in-house Traditional Chinese Medicine (TCM) database was used to investigate novel drugs for EGFR treatment of NSCLC featuring this T790M mutation. In silico virtual screening tools, absorption, distribution, metabolism and excretion (ADME)/Tox analysis, and automated docking were all used to identify an effective single TCM compound.

Materials and methods

Protein preparation

The crystallographic structure of the kinase domain of the EGFR protein was retrieved from the Worldwide Protein Data Bank (PDB; wwpdb.org; ID:1XKK) (14). The structure was cleared of water and other ions and a T790M mutation was added using Discovery Studio Visualizer (Release 3.5.′; Accelrys, Inc., San Diego, CA, USA). The mutated structure was subjected to energy minimization using SPDB viewer version 4.1 (Swiss institute of Bioinformatics, Lausanne, Switzerland) following a previously documented protocol (15). GROMOS force field was used for the energy minimization.

Virtual screening

An initial dataset of 3,000 in-house selected TCM compounds, identified as exhibiting high activity, was used for the present study. This dataset was subjected to analysis with AutoDockVina version 1.1.2 (Scripps Research Institute, La Jolla, CA, USA) (16) platform using a Pymol interface (version 1.4.1; DeLano Scientific, Portland, OR, USA) (17). A grid of 40 Å was created around the ATP binding site. The number of compounds was limited on the basis of Gibbs free energy (ΔG). Only those compounds with an ΔG <-10 kcal/mol were selected for further ADME/Tox analysis.

PreADME/Tox

An online sever (preadmet.bmdrc.org/) was used to evaluate the ADME of 25TCM compounds. The 4 TCM compounds, nardosinon, artesunate, daidzin and emetine were selected based on their ADME properties. The toxicity testing provided the mutagenicity and carcinogenicity properties of the selected compounds and only those compounds with no mutagenic and carcinogenic properties were selected for ADME evaluation. The predictive in silico ADME values and drug-likeliness values were used to further shortlist the compounds. Properties including molecular weight, the octanol/water partition coefficient (logP), number of hydrogen bond donors (HBD), number of hydrogen bond acceptors (HBA) and total polar surface area (tpsa) were taken into account.

Automated docking

The 4 most appropriate compounds, as determined by their ADME properties, were subjected to automated docking by AutoDock version 4.2 (Scripps Research Institute) (18). The Lamarckian algorithm of the tool was used to evaluate the optimal TCM compound for binding to the ATP binding pocket of EGFR. A total of 4 independent docking experiments were performed for the limited TCM compounds with maximum evaluation criteria based on number of runs for generating the suitable pose. The optimum generated pose was studied using LIGPLOT+ software version 4.5.3 (European bioinformatics institute, Hinxton, UK).

Drug treatment and cell proliferation assay

The human lung adenocarcinoma cell lines PC9GR4 and H2347 were purchased from the American type culture collection (ATCC, Manassas, VA, USA) and used in the present study. Cells were seeded at a concentration of 1.5×104 cells/ml in a 24-well plate and cultured at 37°C in an incubator containing 5% CO2 for 24 h in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum (Gibco; Thermo Fisher Scientific, Inc., Waltham, MA, USA). Concentrations of nardosinon and gefitinib (Sigma-Aldrich; Merck Millipore, Darmstadt, Germany) ranging from 0.001 to 10 µM dissolved in dimethylsulfoxide (DMSO), were subsequently added. The control group was treated with 0.1% DMSO alone (vehicle control). Each experiment was performed five times independently at each concentration. Cells were incubated at 37°C in a humidified atmosphere containing 5% CO2 for 24 h. The medium was subsequently removed and 0.1 mg/ml MTT solution (Sigma-Aldrich; Merck Millipore) was added to the cells followed by incubation for 4 h at 37.8°C in the dark. For control, 0.1 MTT solution was added to a plate containing no cells. The supernatant was subsequently removed and an equal volume of DMSO was added to dissolve the formazan crystals. The absorbance was measured at 565 nm against background absorption at 650 nm using an EPOCH Microplate Reader (BioTek Instruments, Inc., Winooski, VT, USA).

Statistical analysis

One-way analysis of variance was used to calculate significance. P-values between 0.001 and 0.01 were considered to indicate a significant difference. All statistical tests were performed using SPSS 18.0 for Windows (SPSS, Inc., Chicago, IL, USA).

Results and Discussion

Protein preparation and virtual screening

The T790M mutant form of EGFR kinase domain was prepared from crystallographic structure (PDB ID: 1XKK). The structure was subjected to energy minimization using GROMOS force field with SPDB Viewer, and PsI-PhI distributions prior to and following the energy minimization are depicted in Fig. 1. The purpose of virtual screening was to limit the number of compounds without affecting the success rate of compounds reaching further stages of drug development. A total of 25 TCM compounds met the cut-off criteria (∆G <-10 kcal/mol) set in virtual screening.

Figure 1. The PsI-PhI distribution plot prior to and following energy minimization using GROMOS force field.

Predictive in silico ADME

The subset of 25 TCM compounds was further screened on the basis of their drug-likeliness according to Lipinski's rule of five (19). The online ADME/Tox tool was used to check Lipinski's parameters and provided an insight into the mutagenic and oncogenic potential of the compounds tested, parameters that are an important basis for modern drug design. A total of 4 compounds were shortlisted and used for automated docking analysis. The following drug-likeliness parameters were used for the shortlisted compounds: Total polar surface area, molecular weight, calculated octanol/water partition coefficient, number of hydrogen bond donors and number of hydrogen bond acceptors (Table I).

Table I. Druglikeness and toxicity prediction.

Table I.

Druglikeness and toxicity prediction.

TCM No.	MW	logP	HBD	HBA	tpsa	Mutagenicity	Carcinogenicity
00812	384.178	3.10	1	8	101	NO	NO
01273	416.111	0.46	5	9	146	NO	NO
01578	480.229	4.49	1	6	52.2	NO	NO
02108	250.333	2.96	1	3	35.5	NO	NO

[i] Data generated using Pre ADME/Tox online tool. TCM, Traditional Chinese Medicine; MW, molecular weight; logP, octanol/water partition coefficient; HBD, number of hydrogen bond donors; HBA, number of hydrogen bond acceptors; tpsa, total polar surface area.

Molecular docking analysis

The shortlisted TCM compounds were docked using the AutoDock 4.2 tool into the optimized adenosine triphosphate (ATP) binding site of the mutated EGFR protein and the results were analyzed using Pymol and LIGPLOT+ software. The results generated by the in silico analysis are presented in Table II. All 4 shortlisted TCM compounds (Fig. 2) interacted with the ATP binding site via non-covalent interactions (hydrogen bonding). The two dimensional representation of the TCM-EGFR interactions (hydrophobic or non covalent) are presented in Fig. 3. The compounds were ranked based on the spontaneity of their interaction with EGFR. Nardosinon was determined to have a ΔG of −6.38 kcal/mol and form 3 hydrogen bonds with the LYS852, ARG776 and LYS846 of the EGFR ATP binding pocket (Fig. 4A). The hydrophobic interaction pocket EGFR with nardosinon is comprised of the following amino acids: ASN771, GLN791, VAL774, MET790, LEU778, ALA1013, ASP1014, LEU1017 and PRO772. Oxygen at the second and third positions of the ligand interacts with LYS 852, LYS 846 and ARG776. The strongest interaction was between N on LYS852 and O2 on nardosinon, with a distance of 2.82 Å.

Figure 2. A total of 4 compounds were shortlisted from the Traditional Chinese Medicine database following virtual screening and absorption, distribution, metabolism and excretion/Tox analysis. (A) Nardosinon, (B) artesunate, (C) daidzin and (D) emetine.

Figure 3. Two-dimensional contact plot depicting the noncovalent and hydrophobic interactions of (A) nardosinon, (B) artesunate, (C) daidzin and (D) emetine with the kinase domain of T790M mutated epidermal growth factor receptor.

Figure 4. (A-D) Three-dimensional representation of atomic interactions. Red lines depict hydrogen bonds. (A) Nardosinon and (C) daidzin form hydrogen bonds, whereas (B) artesunate and (D) emetine exhibit contacts with hydrophobic pockets.

Table II. Auto Dock analysis of four lead natural products.

Table II.

Auto Dock analysis of four lead natural products.

Name	TCM No.	ΔG, kcal/mol	Ligand binding pocket	Hydrogen bonds
Lead 1	02108	−6.38	ASN771, GLN791, VAL774, LYS852a,	LYS852:N-: TCM:O2 (2.82 Å)
			ARG776a, MET790, LEU778, ALA1013,	RG776:N-: TCM:O3 (3.18 Å)
			ASP1014, LEU1017, LYS846a, PRO772	LYS846:N-: TCM:O2 (3.15 Å)
Lead 2	00812	−5.82	ALA743, LEU792, VAL726, LEU844,	NIL
			LEU718, MET766, PHE856, CYS775, LEU777,
			LEU858, THR854, ASP855, ARG841
Lead 3	01273	−1.32	ALA743, VAL726, LEU792, LEU718,	THR854:OG1-:TCM:O7 (3.13 Å)
			MET793, LEU777, LEU844, THR854a,	CYS775:O-:TCM:O6 (2.24 Å)
			CYS755a, MET766, PHE856a, LEU858,	PHE856:N-:TCM:O4 (2.59 Å)
			ASP855a, MET790, ILE789, LEU788	PHE856:O-:TCM:O4 (2.00 Å)
				PHE856:O-:TCM:O3 (2.83 Å)
				ASP855:OD1-:TCM:O3 (3.21 Å)
Lead 4	01578	−0.32	VAL726, ALA743, ILE744, LYS745,	NIL
			LEU788, LEU777, MET790, MET766,
			ASP855, CYS755, THR854, PHE856,
			GLN791, LEU792, MET793, LEU844

{ label (or @symbol) needed for fn[@id='tfn2-etm-0-0-4168'] } The ligand binding pocket and the hydrogen bond formation was calculated using discovery studio 3.5 software.

a Represents the amino acids involved in forming hydrogen bond with the ligand. TCM, Traditional Chinese Medicine.

The second ranked TCM was artesunate, which interacted with the ATP binding pocket of mutated EGFR only. This binding pocket comprises ALA743, LEU792, VAL726, LEU844, LEU718, MET766, PHE856, CYS775, LEU777, LEU858, THR854, ASP855 and ARG841, and has a ΔG of −5.82 kcal/mol (Fig. 4B).

The third ranked TCM was daidzin, which was demonstrated to have a ΔG of −1.32 kcal/mol. The binding pocket daidzin with mutated EGFR comprises the following amino acids: ALA743, VAL726, LEU792, LEU718, MET793, LEU777, LEU844, THR854, CYS755, MET766, PHE856, LEU858, ASP855, MET790, ILE789 and LEU788. Of these, daidzin forms hydrogen bonds with 4 amino acids: THR854, CYS755, PHE856 and ASP855. The O atoms at the 3rd, 4th, 6th and 7th positions of daidzin form hydrogen bonds. Fig. 4C provides a 3D illustration of the interactions.

The fourth ranked TCM compound was emetine, which formed no non-covalent bonds with the mutated EGFR ATP binding domain. The hydrophobic binding pocket comprises the following amino acids: VAL726, ALA743, ILE744, LYS745, LEU788, LEU777, MET790, MET766, ASP855, CYS755, THR854, PHE856, GLN791, LEU792, MET793 and LEU844. The binding energy of the complex is −0.32 kcal/mol. Fig. 4D illustrates the hydrophobic interactions between emetine with the ATP binding domain.

Drug treatment and cell proliferation assay

To investigate the effect of the top ranked TCM compound, nardosinon, under in vitro conditions, the PC9GR4 cell line was used. These cells are derived from PC9 cells and are resistant to gefitinib treatment due to the T790M mutation. Cells were treated with nardosinon at 5 different concentrations (0.001, 0.01, 0.1, 1 and 10 µM). A total of five independent experiments were performed for the 72-h assay with the different concentrations of nardosinon. The procedure was repeated for the H2347 cell line, representing NSCLC with wild type EGFR. Fig. 5 illustrates the cell viability (%) vs. drug concentration of the two cell lines with and without the T790M mutation. The compound of interest, nardosinon (at a concentration >0.1 µg) had a significantly greater effect compared with geftinib on the viability of cells with the T790M mutation (P<0.01), indicating that it could overcome the drug resistance caused by the T790M mutation in NSCLC patients. Following 72 h treatment, an IC50 of 225 mg/ml was calculated via MTT assay.

Figure 5. Inhibition of the EGFR kinase domain in mutated (PC9GR4) and non-mutated (H2347) cell lines by nardosinon at various concentrations (0.1–10 µM). Cell viability was measured using crystal violet staining, following culture in 5% fetal bovine serum with 100 ng/ml EGFR at 72 h and exposure to indicated drug concentrations. Each data point represents the mean of 5 samples. Data represented as mean ± standard deviation of results obtained from five independent experiments. **P<0.01; ***P<0.001 compared with control. EGFR, epidermal growth factor receptor.

The T790M mutation is in the protein kinase domain of the EGFR and activates the ATP affinity of the EGFR kinase (20). Blocking this activity may have therapeutic potential. The difference in physical and chemical properties, including size, charge, and hydrophobicity value caused by the amino acid change from threonine to methionine at codon 790, increased the activity in the kinase domain. In the present study, computer-aided drug design of a pharmacological agent for the treatment of patients with NSCLC and the T790M mutation has revealed a potentially effective TCM compound, nardosinon. Nardosinon is a compound found in the root extract of Nardostachys jatamansi (21). This plant is abundant in China and has a long history of medicinal use (22–25). Nardosinon is able to bind to the ATP binding site of the EGFR (Fig. 6) and the results from the in vitro studies of the nardosinon compound in two different cell lines indicates that it is may be developed as a pharmacological agent capable of treating NSCLC in patients with T790M EGFR variation.

Figure 6. Physical interaction of nardosinon with the kinase domain of the EGFR. Inset shows the pocket where binding of the compound from the Traditional Chinese Medicine database with the mutated EGFR occurs. EGFR, epidermal growth factor receptor.

References