A 3D structure model of AGPS (Fig.1A) was downloaded from the Protein Data Bank (www.rcsb.org; PDB ID: 2UUV) and the protein modules were processed using Discovery Studio 3.5 software (BIOVIA, San Diego, CA, USA) as follows: A single chain (the A chain) in the tetramer was retained; missing atoms to residues were added; the loop area was repaired; combinations and free water molecules were removed; unsaturated atoms were hydrogenated; valence was repaired; protein and original ligands were retained; and six basic pharmacophore elements were selected, namely hydrogen bond acceptor, hydrogen bond donor, hydrophobic, positive electricity center, negative electricity center and aromatic center. This obtained a final 3D AGPS structure for docking. The structure of TMZ used is depicted in Fig. 1B (15).

The structures of TMZ derivatives were combined with benzyl isothiocyanate (BITC) and the ITC-TMZ derivatives were drawn using ChemBioOffice 2010 (PerkinElmer, Waltham, MA, USA), and each structure was saved in mol2 format. The structures were imported into Discovery Studio 3.5 for virtual screening. The CDOCKER protocol in Discovery Studio 3.5 was used for molecular docking; ‘CDOCKER energy’ was used as a scoring index, with higher scores indicating greater docking affinity.

Firstly, it used the pharmacophore model generated by the HipHop algorithm to perform fit value sorting of the compounds based on the characteristics of the pharmacophore model, with small molecular compounds with the highest scores retained (16). The pharmacophore model generated by the Counting Belief Propagation algorithm was then used on reserved compounds to obtain preferable compounds that conformed to the characteristics of the two pharmacophore types (17). BITC was used as a control for the virtual screening.

Protein structure processing was completed with the QikProp module of Schrodinger Suite 2009 software (Schrodinger, LLC, Cambridge, MA, USA). The downloaded AGPS target crystal structure (PDB ID: 2UUV) was imported into the Schrodinger software and processed as follows: Metal ions were removed; unsaturated atoms were hydrogenated; amino acid residues at the N and C termini were made neutral; and protein and original ligands were retained to remove atom and water molecule impurities. The OPLS_2005 position was selected to optimize protein structure and utilize sample water orientations as parameters to optimize hydrogen bonding.

Ligand preparation utilized ChemBioDraw Ultra 11.0 in the ChemBioOffice 2010 to draw planar structures of the micromolecules, which were imported into ChemBio3D Ultra 11.0 for the generation of 3D structures, which were saved in mol2 format. The mol2fileswere then imported into Schrodinger Suite 2009 software and processed with the LigPrep module, complete desalting, add charge and produce ionization consistent with human body pH. The OPLS_2005 position was also selected for the optimized molecules to match that of the optimized acceptor. ADME parameters, namely oil-water partition coefficient (logPo/w), polarization surface area (PSA), water solubility (logS) and apparent Maden Darby Canine Kidney cell permeability (PMDCK), and toxicity factors, namely rodent carcinogenicity, mutagenicity, skin irritancy, ocular irritation and aerobic biodegradability, were subsequently simulated in the Schrodinger Suite 2009 software.

The human glioma cell lines U87MG and U251 were obtained from the Cell Bank of Type Culture Collection of Chinese Academy of Sciences (Shanghai, China). Cells were cultured in Dulbecco's modified Eagle's medium (Thermo Fisher Scientific, Inc., Waltham, MA, USA) with 10% fetal bovine serum (Thermo Fisher Scientific, Inc.) at 37°C with 5% CO2. It is important to note that U87MG has been identified to differ from the original cell line, and is probably an authentic human glioblastoma cell line of unknown origin according to DNA profiling (18).

A total of 5×103 cells/well were cultured in 96-well plates at 37°C with 5% CO2 for 24 h. ITC-TMZ derivatives were designed and synthesized via CADD (14) and added to the medium at final concentrations of 0 (negative control), 1, 2, 5, 10, 20, 50 and 100 µM, respectively. Subsequently, the cells were cultured for 72 h at 37°C, after which the medium was replenished with 200 µl plasma-free medium and 40 µl MTS, followed by incubation at 37°C for another 4 h. The optical density (OD) value was measured using an enzyme-labeled meter at 490 nm. Inhibition rate (%) was calculated as follows: (1-OD value of the medicine-treated group/OD value of the negative control group) × 100%. Half maximal inhibitory concentrations (IC50s) were subsequently determined using GraphPad Prism 6 software (GraphPad Software, Inc., La Jolla, CA, USA).

Visual screening of 50 small molecular compounds was performed using Schrodinger Suite 2009. A total of 10 ITC-TMZ derivatives and the compound BITC exhibited micromolecular structures with improved combination with the target protein AGPS compared with TMZ, determined by referring to the combination mode of the micromolecules and surrounding amino acids, and evaluation of scoring function from CDOCKER energy. The scoring of the docking between these micromolecules [structures (S)1-10 and BITC] and the AGPS active site is presented in Fig. 2. The 11 candidate compounds exhibited marked interactions with the protein cavity, with key amino acid residues forming hydrogen bonds and the hydrophobic effect occurring with nitrogen and sulfur atoms and carbon-sulfur covalent bonds of the ITC-TMZ derivative molecules. The other substituent residues in the micromolecules formed certain hydrogen bonding interactions with surrounding residues to an extent. As a result, as evident in the pattern diagrams (Fig. 2), the main ITC active group and small regions of substitute residues in the TMZ derivatives with some impurity atoms were respectively connected to both sides of the hydrophobic and rigid ring planes. These structural data suggest improved interactions and combination of the ITC-TMZ derivatives with the AGPS target compared with TMZ. Indeed, the docking scores of S1-9 and BITC were increased compared with that of TMZ, suggesting enhanced AGPS targeting.

ADME properties, including logPo/w, PSA, logS and PMDCK, and toxicity properties, including rodent carcinogenicity, mutagenicity, skin irritancy, ocular irritation and aerobic biodegradability, aid to determine whether therapeutic candidates are of clinical standard (16). The present results on ADME (Table I) and toxicity (Table II) parameters indicated the 10 ITC-TMZ derivatives to have improved interaction with AGPS. Firstly, the results suggested adequate ADME parameters for BITC regarding physico-chemical properties. In turn, the ADME parameters of S3, −6, −7, −8, −9 and −10 were optimized compared with TMZ, particularly regarding logPo/w, logS and PMDCK, suggesting the structure of BITC may improve TMZ physico-chemical properties. Meanwhile, the general toxicity parameters of the above compounds were not increased according to the predictions.

MTS assay was performed to observe the effect of the ITC-TMZ derivatives on the viability of human glioma U87MG and U251 cells in vitro. On determination of IC50 values, S1, −3, −8 and −10 were identified to exert greater inhibitory effect on glioma cell viability compared with TMZ (Table III).