Computational screening of chalcones acting against topoisomerase IIα and their cytotoxicity towards cancer cell lines

Abstract Targeted cancer therapy has become one of the high potential cancer treatments. Human topoisomerase II (hTopoII), which catalyzes the cleavage and rejoining of double-stranded DNA, is an important molecular target for the development of novel cancer therapeutics. In order to diversify the pharmacological activity of chalcones and to extend the scaffold of topoisomerase inhibitors, a series of chalcones was screened against hTopoIIα by computational techniques, and subsequently tested for their in vitro cytotoxicity. From the experimental IC50 values, chalcone 3d showed a high cytotoxicity with IC50 values of 10.8, 3.2 and 21.1 µM against the HT-1376, HeLa and MCF-7 cancer-derived cell lines, respectively, and also exhibited an inhibitory activity against hTopoIIα-ATPase that was better than the known inhibitor, salvicine. The observed ligand–protein interactions from a molecular dynamics study affirmed that 3d strongly interacts with the ATP-binding pocket residues. Altogether, the newly synthesised chalcone 3d has a high potential to serve as a lead compound for topoisomerase inhibitors.


Introduction
Nowadays, cancer is one of the most serious groups of diseases in the world, with the number of deaths attributed to cancers being about 8.8 million in 2015 (World Health Organization, 2017) 1 . Chemotherapy is currently a highly effective cancer treatment. However, most chemotherapeutic agents give severe side effects with limited selectivity against various cancer cells. Hence, development of anti-cancer drugs with no or less side effects and a high selectivity is of prime concern 2 . Targeted cancer therapy, in which the drugs are used to specifically block the growth of cancer by interfering with molecular targets and consequently causing less damage to normal cells, has become one of the high potential cancer treatments. Several kinds of molecular targets have been focused on in recent years, including human topoisomerase II (hTopoII). This enzyme catalyzes the cleavage and rejoining of double-stranded DNA and so it is essential in several vital cell processes, such as replication, transcription, chromosome separation and segregation 3 . Generally, hTopoII exists in two homologous structures but in different isoforms, hTopoIIa and hTopoIIb. The hTopoIIa isoform shows a low expression level in the G cell cycle phase but an increased concentration in the S and G 2 /M phases compared to normal cells, whilst hTopoIIb does not change its concentration during the cell cycle 4 . Since hTopoIIa is highly overexpressed in proliferating cancer cells 5 , it has gained attention from many researchers who are developing new anti-cancer drugs.
There are two important motifs for drugs targeting hTopoIIa, namely the ATPase domain ( Figure 1(A)) and the DNA-binding core (Figure 1(B)) 6 . The hTopoIIa inhibitors can be divided into two categories, hTopoIIa poisons and hTopoIIa catalytic inhibitors 3 . For hTopoIIa poisons (etoposide, doxorubicin, anthracyclines and mitoxantrone), they are clinically active agents that generate a high level of hTopoII-DNA covalent complexes by stimulating cleavage of the G-segment and blocking relegation of DNA 7 . On the other hand, hTopoIIa catalytic agents (ICRF-187, novobiocin, merbarone and salvicine) affect the catalytic cycle of hTopoIIa by elimination of the enzymatic activity [8][9][10] . Although these different catalytic agents share the same effect, they interact with hTopoIIa at different binding sites. For example, the ICRF-187 binding pocket is located in the middle of the primary dimer interface 8 , while merbarone acts by blocking the DNA cleavage reaction of hTopoIIa. The merbarone-binding site possesses an interaction domain overlapping with that of etoposide [11][12][13] . Salvicine, a derivative of diterpenoid quinones isolated from the traditional Chinese medicinal plant Salvia prionitis 10,14 , targets the ATPase domain [15][16][17] .
Chalcones have attracted attention because of their promising therapeutic effects, since they are able to target multiple cellular molecules, such as MDM2/p53, tubulin, proteasome, NF-jB, TRIAL/death receptors and mitochondria-mediated apoptotic pathways, cell cycle, STAT3, AP-1, NRF2, AR, ER, PPAR-c, b-catenin/ Wnt 34 and especially hTopoIIa 24,[35][36][37] . Moreover, epipodophyllotoxin-chalcone hybrids exhibited an enhanced in vitro cytotoxicity and higher topoisomerase II inhibitory efficiency than etopoiside 38 . A series of chalcone-triazole derivatives presented a promising anticancer activity against the A-549 cell line and showed high binding affinities towards DNA topoisomerase IIa and a-glucosidase targets 39 . Moreover, the novel bis-fluoroquinolone chalcone-like derivatives were found to inhibit both hTopoIIa and tyrosine kinase 40 . Recently, a series of 2 0 -and 4 0 -aminochalcones were found to inhibit the growth of a canine malignant histiocytic cell line (DH82) and the transcription of the hTopoIIa and TP53 genes 41 . In the present study, in order to find new potential anti-cancer agents against hTopoIIa, the new 47 chalcone derivatives were designed ( Figure 2) and then screened in silico using a molecular docking approach. The potent chalcones with a more favorable interaction energy than that of the known hTopoIIa inhibitors were then synthesised and tested for their in vitro cytotoxicity towards three cell lines derived from urinary bladder (HT-1376), cervical (HeLa) and breast (MCF-7) cancers. Then, all-atom molecular dynamics (MD) simulations were performed to investigate the structure and dynamics properties as well as the ligand-target interactions between the most potent chalcone and hTopoIIa.

Material
Human urinary bladder, cervical and breast cancer-derived cell lines (HT-1376, HeLa and MCF-7, respectively) were obtained from the American Type Cell Culture Collection (ATCC), Manassas, VA. Dulbecco's modified eagle's medium (DMEM), fetal bovine serum (FBS), penicillin-streptomycin (Pen-Strep) and trypsin were purchased from Life Technologies (Carlsbad, CA). Thiazolyl blue (MTT), dimethyl sulphoxide (DMSO), sodium dodecyl sulphate (SDS) and phosphate-buffered saline (PBS) were purchased from Sigma-Aldrich (Darmstadt, Germany). Salvicine was purchased from Chemfaces (Wuhan, P.R. China). The purity of the compound was more than 98.0%. All other chemicals and solvents used were of analytical grade. Plasmid pET28b-hTopoIIa-ATPase was gifted from Dr. Nonlawat Boonyalai. ADP-Glo TM Kinase Assay kit was purchased from Promega (Madison, WI). All solvents used for the synthesis were purified prior to use by standard methodologies. The reagents used for synthesis were purchased from Sigma-Aldrich, Merck or TCI chemical companies and were used without further purification.

Molecular docking
Due to the possibility of the inhibition of two motifs of the hTopoIIa (ATP-binding site in the ATPase domain and the etoposide-binding pocket in the hTopoIIa/DNA complex), the predicting mode of the inhibitory activity of chalcones on both sites was studied by molecular docking using the CDOCKER module of Accelrys Discovery Studio 3.0 (Accelrys Inc, San Diego, CA, USA) as previously reported 42 . The starting structures of the 47 designed chalcone derivatives were built by the GaussView program 43 , while those of salvicine and etoposide were taken from the ZINC database 44 . To validate the docking method, the co-crystallised ligands were initially docked into the binding pocket with 100 independent runs, i.e. docking of AMP-PNP into the ATP-binding site of the hTopoIIa ATPase domain (1ZXM.pdb), and etoposide into its binding pocket of the hTopoIIa/DNA complex (3QX3.pdb). The position of docked ligands did not differ significantly from the crystallised conformation ligands (RMSD ¼ 0.80 Å for AMP-PNP and 0.44 Å for etoposide) and so the 47 chalcones were then separately docked into both sites, while salvicine (used as the reference compound at the ATPase domain) was only docked into the ATP-binding site. The chalcones with predicted interaction energies towards hTopoIIa that were more favorable than those of the known inhibitors were synthesised and their in vitro cytotoxicity against the three cancer cell lines was tested (see Section 2.3.3).

MD simulation
All-atom MD simulations under a periodic boundary condition were performed on the most potent chalcone selected from the in vitro cytotoxicity study (Section 2.3.3) in complex with hTopoIIa in aqueous solution, following the previously reported MD study on the binding of mansonone G to hTopoIIa 42 . The partial charges of the ligand were prepared according to standard procedures [45][46][47] . The ligand was optimised with ab initio calculation using the HF/6-31G Ã method in the Gaussian09 program 48 . The electrostatic potential (ESP) charges of the ligand were calculated using the same level of theory, and then the restrained ESP (RESP) charges were obtained by the charge fitting procedure using the antechamber module in the AMBER 14 package program 49 . The general AMBER force field (GAFF) 50 and AMBER ff03 force field 51 were applied for the ligand and protein, respectively. The protonation states of all ionizable amino acids were determined using PROPKA 3.1 52 . The complex was solvated by TIP3P water molecules 53 within 12 Å around the system surface. Chloride ions were introduced to neutralise the total positive charge of the chalcone/ hTopoIIa complex.
To remove the bad contacts and steric hindrances, the added hydrogen atoms were minimised with 1000 steps of steepest descents (SD) followed by 2000 steps of conjugated gradients (CG) using the Sander module in AMBER 14. The water molecules and ions were then minimised with 500 steps of SD followed by 500 steps of CG, while a 500 kcal/mol Å 2 force constant was used to restrain hTopoIIa. The whole system was then fully minimised with 1000 steps of SD and CG. All covalent bonds involving hydrogen atoms were constrained by the SHAKE algorithm (Amber, San Francisco, CA) 54 . The long-range electrostatic interactions were calculated according to the Particle Mesh Ewald (PME) approach 55 with a cutoff distance of 12 Å for non-bonded interactions.  The system was heated to 310 K for 100 ps and then simulated at the same temperature for 80 ns in the NPT ensemble using a time step of 2 fs. The MD trajectories in the production phase were taken for analysis in terms of the per-residue decomposition free energy and intermolecular hydrogen bonds (H-bonds) between the ligand and hTopoIIa using the MM/PBSA.py and cpptraj modules, respectively. The percentage of H-bond occupation was calculated using the two criteria of: (i) the distance between proton donor (HD) and acceptor (HA) atoms 3.5 Å and (ii) the angle of HD-H … HA >120 .

Synthesis of chalcone derivatives
The three selected chalcones (3c, 3d and 3f) were synthesised by Claisen-Schmidt condensation with some modifications between selected acetophenones and benzaldehydes under a basic condition, according to the procedures described by Cabrera 56 . The target products were purified by column chromatography and their structures were elucidated by NMR spectroscopy.

Cytotoxicity assay
The cytotoxicity of the chalcones and salvicine was measured according to a published method 57 with some modifications. The cell viabilities of three cancer cell lines (HT-1376, HeLa and MCF-7) exposed to the screened chalcone derivatives were evaluated by the MTT assay. The cell suspension (100 lL) was seeded into 96-well plates at a density of 2 Â 10 6 cells/well and then incubated for 24 h under normal culture conditions before the addition of the respective test compound at various concentrations [100, 50, 25, 12.5 and 0 (control) lM] and incubated for another 24 h. Then, 10 mL of fresh MTT solution (5 mg/mL) was added to each well and incubated at 37 C for 2 h, before the reaction was stopped by adding 100 lL of DMSO. The absorbance was measured at 570 nm with correction for background at 690 nm using a microplate spectrophotometer system (Infinite M200 micro-plate reader, Tecan, M€ annedorf, Switzerland). The amount of the colored product is assumed to be directly proportional to the number of viable cells. Each experiment was performed in triplicate and repeated three times. The percentage cell viability in each compound was calculated relative to the control, and the IC 50 values were determined in comparison with untreated controls using the Table  Curve 2D program version 5.01 (Systat, San Jose, CA).

Expression and enrichment of the recombinant (r)hTopoIIa ATPase domain
Expression and enrichment of the rhTopoIIa ATPase domain was modified from that reported 58 . The expression plasmid pET28b-hTopolla-ATPase was transformed into Escherichia coli BL21 (DE3) cells and a transformant colony was selected for large-scale protein expression and grown at 37 C to an optical density at 600 nm of $0.6 in LB broth (2 L) containing 50 mg/mL kanamycin. Protein expression was then induced by adding 0.5 mM IPTG at 30 C for 5 h. The cells were harvested by centrifugation at 6000Âg, 4 C and resuspended in lysis buffer [50 mM Tris-Cl pH.8, 0.5 M NaCl, 5 mM imidazole, 0.5% (v/v) Triton X-100, 1 mM PMSF] and lysed by sonication. After clarification by centrifugation (as above) the supernatant was harvested, and the rhTopoIIa-ATPase enriched for using HisTrap Chelating HP and Resource S column chromatography, eluting in exchange buffer [50 mM Tris pH.7.5, 50 mM NaCl, 5% (v/v) glycerol, 50 mM KCI, 5 mM MgCl 2 ] from a PD-10 desalting column. The enriched protein was analyzed by 12% sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE) and stained by Coomassie blue.

ATPase assay
The inhibitory activities of salvicine and chalcone 3d were determined by measuring the ATPase activity of rhTopoIIa-ATPase using the ADP-Glo TM Kinase Assay. Briefly, 8 mL of buffer (40 mM Tris-HCI pH 7.5, 20 mM MgCl 2 , 0.1 mg/mL BSA) was added to each well of a 384-well plate (Promega, solid white) with 5 mL of enzymes (10 ng/ mL) and 2 mL of the test compound at different concentrations. The reaction was initiated by the addition of 5 mL of 12.5 mM ATP, allowed to proceed for 1 h at room temperature and then stopped by the addition of 5 mL of ADP-Glo TM Reagent and incubating at room temperature for 40 min. Next, 10 lL of Detection Reagent was added and incubated for 30 min prior to the addition of luciferase and luciferin to detect the ATP by measuring the luminescence of each well with a microplate spectrophotometer system (Synergy HTX Multi-Mode reader, BioTek, Winooski, VT). All assays were performed in triplicate. The percentage relative inhibition of salvicine and 3d was calculated as shown in Equation (1); (1) where negative and positive are the luminescence without and with the enzyme activity, respectively, and sample is luminescence with the sample. Finally, the IC 50 curve was determined by GraphPad Prism version 6 (GraphPad Sofware, La Jolla, CA).

Molecular docking studies
To investigate the most favorable binding site of the 47 designed chalcones, each compound was separately docked into the ATPbinding site in the ATPase domain of hTopoIIa and the etoposidebinding pocket in the hTopoIIa/DNA complex. The predicted interaction energies of all chalcones at both sites were plotted and compared with those of salvicine and etoposide (Figure 3). The interaction energies of the chalcones ranged from À45.6 to À32.4 kcal/mol in the etoposide-binding pocket and from -60.0 to À37.5 kcal/mol in the ATP-binding site. This suggested that all the chalcone derivatives specifically interacted with the ATPase domain rather than with the hTopoIIa/DNA complex. Among all 47 chalcones, the group-3 compounds (3c, 3d and 3f) showed high interaction energies with the hTopoIIa ATPase domain (À61.1, À60.7 and À59.3 kcal/mol), which were better than that of salvicine (À58.7 kcal/mol) at the hTopoIIa ATPase domain. However, none of the tested chalcones were stronger than etoposide binding in the hTopoIIa/DNA complex (À72.4 kcal/mol). Additionally, the mode of action of these three compounds was likely comparable with salvicine in the ATP-binding pocket ( Figure 4). The important residues that contributed to ligand stabilisation via van der Waals (vdW) and H-bond interactions are summarised in Table 1. There are at least four conserved residues between each chalcone and salvicine. The obtained results were similar to the docking study of 4-ethoxycarbonylmethyl-1-(piperidin-4-ylcarbonyl)-thiosemicarbazidehydrochloride, and napthoquinone-containing compounds, which specifically targeted the ATPase domain 42,59 . Since 3c, 3d and 3f may be effective as ATP competitors at the ATP-binding site of the hTopoIIa ATPase domain, these three compounds were synthesised and their in vitro cytotoxicity towards three cancer cell lines was then tested.

Cytotoxicity towards cancer cell lines
After screening the potent chalcones for inhibition of hTopoIIa by molecular docking, the three compounds that exhibited better interaction energies than salvicine (3c, 3d and 3f) were selected for synthesis to test their cytotoxicity on the HT-1376, HeLa and MCF-7 cancer-derived cell lines using the MTT assay. The derived IC 50 values of the three chalcone derivatives and salvicine on the three cancer cell lines are summarised in Table 2. All three screened chalcones showed a higher cytotoxicity to all three cell lines than salvicine, with 3d being the most cytotoxic with an IC 50 value of 10.8 ± 1.1, 3.2 ± 2.2 and 21.1 ± 6.3 mM against the HT-1376, HeLa and MCF-7 cell lines, respectively. The IC 50 of salvicine in a lung cancer cell line (A549) was previously reported to be 18.66 mM 60 . The diversity of the cytotoxicity of these three chalcones could suggest that the position of the methoxy group on the B ring of the chalcones affected the cytotoxicity. The methoxy groups substituted at the R 2 , R 3 and R 4 positions were found to be most important in terms of anti-cancer activities. Moreover, the different IC 50 values of the chalcone derivatives in each cancer cell line may reflect the different expression levels of hTopoIIa and proliferation rates between those cell lines [61][62][63][64] . Cells containing a high concentration of hTopoIIa are more sensitive to hTopoIIa-inhibiting drugs than cells containing a lower concentration of hTopoIIa 61,65 . Thus, these chalcones might    , 3d and 3f). The residues in bold format stabilise the ligand binding via H-bond interaction, while the conserved residues between each chalcone and salvicine are shown in underlined format. inhibit HeLa cells better than MCF-7 and HT-1376 cells because of the higher hTopoIIa levels typically expressed in cervical cancer cells than in breast and urinary bladder cancer cells 66 . Considering the data from the in silico molecular docking and the in vitro cytotoxicity against cancer cell lines, it is possible that 3d tends to inhibit the hTopoIIa ATPase domain in a somewhat similar manner as salvicine.
However, to gain additional information about the inhibition of hTopoIIa at the ATPase domain by salvicine and 3d, their in vitro inhibitory activity against the ATPase activity of rhTopoIIa was evaluated.

Inhibition of the hTopoIIa ATPase domain
In order to assess the inhibition of ATPase activity by salvicine and 3d, the rhTopoIIa ATPase domain was expressed from the pET28bexpression vector and enriched by following a previously reported protocol 58 for use in the ATPase enzymatic activity assay. The rhTopoIIa ATPase domain was enriched to apparent homogeneity, with the 45 kDa ATPase domain evident as a single band following SDS-PAGE resolution and Coomassie blue staining ( Figure 5(A)). The ATPase inhibitory activity of different concentrations of salvicine and 3d was then comparatively studied using a commercial kit (ADP-Glo TM Kinase Assay, see also in material). The obtained IC 50 curves of salvicine and 3d are shown in Figure 5(B,C), respectively, and listed in Table 2. The chalcone 3d showed an ATPase inhibitory activity with an IC 50 value (7.5 nM) that was some 43.5-fold lower than that for salvicine (326.5 nM). To investigate the binding and interaction of 3d against hTopoIIa at the ATPase domain a detailed investigation of the 3d/hTopoIIa complex in aqueous solution was then performed in silico using MD simulations.

MD simulations
All-atom MD simulations were performed on the docked 3d/ hTopoIIa complex with three different velocities for 80 ns to understand the structure and dynamics of 3d at the ATP-binding site of the hTopoIIa ATPase domain. Since the 3d binding patterns and interactions with hTopoIIa obtained from three different simulations were similar, the results presented here are taken from one representative simulation. The root mean square displacement (RMSD) plot in Supplementary Figure S1 showed that the 3d/hTopoIIa complexes had reached equilibrium by 50 ns. Herein, the snapshots taken from the last 10-ns were extracted for analysis in terms of the binding pattern and ligand-protein interactions as follows.
In order to elucidate the hTopoIIa ATPase residues important for 3d inhibition at the ATP-binding site, the per-residue decomposition free energy (DG residue ) was evaluated by the MM/PBSA approach using the 100 snapshots over the last 10-ns simulation. The results are given in Figure 6(A), where the binding orientation of 3d inside the ATP-binding pocket with the contour energy of residue contribution is drawn in Figure 6(B). The fingerprint in Figure 6(A) showed only residues 50-250 in chain A, while the rest of protein (chain A residues 29-49 and 251-405 plus all chain B residues) had no interaction with the ligand. The negative and positive DG residue values represented the degrees of stabilisation and destabilisation for ligand binding, respectively.
From Figure 6(A), 10 residues preferentially stabilised 3d with an energy contribution lower than À1.0 kcal/mol: E87, D94, R98, I125, I141, S148, S149, G164, Y165 and K168. This implies that these residues probably play a crucial role in 3d binding to the ATPase domain. The free energy contributions of each key residue, decomposed to backbone and side chain as well as electrostatic (E ele þG polar ) and vdW (E vdW þG nonpolar ) terms, are plotted in Figure 7. Most of the important residues support the 3d binding via the vdW energy contribution, while E87, D94 and K168 residues likely presented the electrostatic contribution. The strongest energy stabilisation for 3d (À3.8 kcal/mol) came from the K168 residue. In contrast, it has been reported that the K168 was not interfere with salvicine binding and even destabilised some mansonone G compounds in the ATP-binding pocket 42 . However, the observed binding patterns of 3d in this work are somewhat similar with salvicine (E87, I125 and I141) and mansonone G (D94, I125, I141 and G164) in our previous work. 42 The results also demonstrated that the 3d binding energy is mainly contributed from the side chains of the key residues (E87,  D94, R98, I125, I141, S148, S149, G164, Y165 and K168), except for the S148, S149, G164 and Y165 residues where the ligand-protein interactions substantially come from their backbone contributions. The information was well supported by the formation of two strong H-bonds between the carbonyl group of 3d and the backbone nitrogen of S149 (92%) as well as the 3-methoxy group on its A-ring and the backbone nitrogen of G164 (80%), (see intermolecular H-bonds between 3d and hTopoIIa residues in Figure 8).

Conclusions
A series of 47 designed chalcones were screened in silico as potent anti-cancer agents by computational methods. Molecular docking of the chalcone derivatives relative to salvicine, a known inhibitor of hTopoII at the ATPase domain, suggested that the ATP-binding site of hTopoIIa ATPase domain serves as the target site for the considered chalcones. The three most active chalcones (3c, 3d and 3f) had interaction energies towards the ATPase domain that were stronger than that of salvicine. Compound 3d, containing 2,4-dimethoxy and 6-hydroxy groups on A ring and 3 0 ,4 0 ,5 0 -trimethoxy on the B ring, showed the highest in vitro cytotoxicity against the HT-1376, HeLa and MCF-7 cancer cell lines. Moreover, 3d inhibited the rhTopoIIa ATPase activity in vitro with an IC 50 value some 43.5-fold lower than that for salvicine. From 80-ns MD simulations of the 3d/hTopoIIa complex, the key residues responsible for 3d binding via vdW and electrostatic interactions were E87, D94, R98, I125, I141, S148, S149, G164, Y165 and K168. The residue K168 exhibited the strongest energy stabilisation for 3d, while residues S149 and G164 formed  two strong H-bond interactions with the carbonyl and 3-methoxy groups of 3d. In summary, the in silico and in vitro results suggested that 3d can serve as a lead compound for further anti-cancer drug development. Figure 8. Hydrogen bond formation between chalcone 3d and the two residues in the ATP-binding pocket of hTopoIIa ATPase domain, where the percentage of H-bond occupation is also given.