Molecular Interaction of Cytotoxic Anticancer Analogues as Inhibitors of β-tubulin Protein Against UACC-62 Melanoma Cell

In previous research, a series of cytotoxic anticancer analogues related to 2-acylamino-1,4-naphthoquinone derivatives has been demonstrated. As microtubule plays an important role in many essential cellular processes such as mitosis, tubulin is an important target of anticancer drug. This study performed molecular docking simulation, pharmacophore model, comparative force eld analysis model, and comparative similarity indices analysis model to investigate the relationship between inhibitory activities and the properties of compounds, in order to further progress the development of cytotoxic anticancer analogues. employed to construct the CoMFA (19) and CoMSIA (20) models. CoMFA was performed with distance-dependent dielectric method to evaluate the steric eld descriptors using Lennard–Jones potential energies and electrostatic eld descriptors using Coulombic potential energies. CoMSIA was performed with a Gaussian function based on distance to evaluate ve physico-chemical properties, which are steric, electrostatic, hydrophobic, H-bond donor, and H-bond acceptor. The partial least-squares regression was performed to obtain the linear correlation between cytotoxicity pGI 50 values and descriptors obtained by CoMFA and CoMSIA, respectively.


Introduction
In previous research, it demonstrates a series of cytotoxic anticancer analogues related to 2-acylamino-1,4-naphthoquinone derivatives with their inhibitory concentration (1). Microtubules, which are polymerized by α-tubulins and β-tubulins, are a major component of the cellular cytoskeleton (2,3). It is a major cellular target of anticancer drug as it plays an important role in many essential cellular processes such as mitosis (4)(5)(6)(7). Tubulin-binding drugs kill tumor cells by inhibiting microtubule dynamics required in cell division (8)(9)(10). In this study, we performed molecular docking simulation to identify binding conformations and interactions between compounds and β-tubulin protein, in order to investigate the protein-ligand interaction network responsible for their anticancer activities. For quantitative structureactivity relationship (QSAR) models study, we also performed pharmacophore model to investigate the common pharmacophore features, and we performed comparative force eld analysis (CoMFA) and comparative similarity indices analysis (CoMSIA) models to investigate relationship between inhibitory activities and their ve physico-chemical properties in order to further progress the development of designing of lead compounds with improved bioactivity.

Data Collection
The X-ray crystal structure of tubulin protein was obtained from RCSB Protein Data Bank with PDB ID: 5M7E (11). We performed the Prepare Protein protocol in Discovery Studio 2.5 (DS2.5) to remove water atoms in crystal structure, repair and optimize side-chain conformation of incomplete amino acids, and protonate the structure of β-tubulin protein using Chemistry at HARvard Macromolecular Mechanics (CHARMM) force eld (12). The co-crystallized compound, BKM120, in X-ray crystal structure of tubulin protein was employed to de ne the volume and position of binding site (Fig. 1).
The compounds displayed in Fig. 2 and Table 1 with their cytotoxicity pGI 50 values against UACC-62 melanoma cell growth were obtained from previous research in our laboratory (1). All 47 compounds drawn by ChemBioO ce 2010 were prepared by Prepare Ligand protocol in DS2.5 to modify their ionization state to physiological ionization setting.

Molecular Docking Simulation
We de ned the structure of β-tubulin protein obtained from RCSB Protein Data Bank with PDB ID: 5M7E as receptor and de ned the volume of co-crystallized compound, BKM120, as binding site (Fig. 1). To validate the accuracy of docking simulation using LigandFit protocol, we performed a docking simulation using LigandFit protocol to redock the co-crystallized compound into the binding site of β-tubulin protein.
The docking pose of BKM120 is displayed in Fig. 1 with the root-mean-square deviation value of 0.3475 We performed LigandFit protocol (13) in DS 2.5 to simulate the docking poses of each compound. LigandFit protocol employed Monte-Carlo ligand conformation generation and a shape-based docking, and then it optionally minimized the docking poses with CHARMM force eld (12) and ltered out similar docking poses using the clustering algorithm. We consider four different scoring functions, which are -PLP1 (14), -PLP2 (15), -PMF (16), Dock Score, and the interactions between compounds and β-tubulin protein to determine the suitable docking pose of each compound. -PLP1, -PLP2, and -PMF are the scoring functions evaluated by summing two types of pairwise interaction, namely hydrogen bonds (Hbonds) and steric interaction, between protein and compound. Dock Score is the scoring function evaluated based on a force eld approximation as following equation, (17) was performed to generate the 2D ligand-protein interaction diagrams with the Hbond and hydrophobic contacts between compound and β-tubulin protein.

Pharmacophore model
For each compounds, we performed FAST generation protocol in DS2.5 to generate their low-energy conformations, and performed 3D-QSAR Pharmacophore Generation protocol in DS2.5 to generate a pharmacophore model using Catalyst HypoGen algorithm (18). Four different pharmacophore features, H-bond donor, H-bond acceptor, hydrophobic, aromatic ring, were considered to construct the common pharmacophore models.
CoMFA and CoMSIA models SYBYL-X was employed to construct the CoMFA (19) and CoMSIA (20) models. CoMFA was performed with distance-dependent dielectric method to evaluate the steric eld descriptors using Lennard-Jones potential energies and electrostatic eld descriptors using Coulombic potential energies. CoMSIA was performed with a Gaussian function based on distance to evaluate ve physico-chemical properties, which are steric, electrostatic, hydrophobic, H-bond donor, and H-bond acceptor. The partial least-squares regression was performed to obtain the linear correlation between cytotoxicity pGI 50 values and descriptors obtained by CoMFA and CoMSIA, respectively.
between crystallized structure and docking pose. It represents a suitable docking pose of BKM120 in the docking simulation using LigandFit protocol.
The chemical scaffolds of 47 compounds were displayed in Fig. 2 and Table 1 with their cytotoxicity pGI 50 values against UACC-62 melanoma cell growth. The results of docking simulation for 47 compounds listed in Table 2 were determined due to their scoring functions and interactions between each compound and β-tubulin protein. The docking poses of compound 06 and 18 with their interactions illustrated in Fig. 3.

Pharmacophore model
In this study, we performed 3D-QSAR Pharmacophore model using 47 compounds to investigate the common pharmacophore features. Figure 4a illustrated the result of the best pharmacophore hypothesis with the distances between each pharmacophore feature. It indicates two H-bond acceptor features, one aromatic ring feature, and one hydrophobic feature. Figure 4b-c displayed the compounds 06 and 18 mapping in the pharmacophore feature, respectively.

CoMFA and CoMSIA models
The 21 compounds of training set were selected due to their docking poses (Table 2). For alignment of compounds, 21 compounds were superimposed based on their docking poses, and then we constructed the CoMFA and CoMSIA models to determine the correlation between the e cacy of cytotoxicity pGI 50 values and the functional groups of compounds. After partial least-squares analysis, the predicted pGI 50 values of compounds of training set listed in Table 3 were evaluated by CoMFA model using four components with signi cantly steric elds (100% contribution) and by CoMSIA model using four components with signi cantly steric (17.9%), hydrophobic (42.5%), and H-bond donor (39.6%) elds, respectively. The correlations between predicted pGI 50 values and experiment pGI 50 values for CoMFA and CoMSIA models were displayed in Fig. 5 with the square correlation coe cients (R 2 ) of 0.857 and 0.817, respectively. The results of CoMFA and CoMSIA models were then illustrated by eld contribution maps in Fig. 6 with the high a nity compounds 18. The favorable and unfavorable regions (85% and 15%, respectively) were evaluated using StDev*coe cients for each eld.

Molecular Docking Simulation
In docking simulation, these compounds have common H-bond interactions with key residues Lys254 and Lys352. However, the docking poses of compounds 01-06 with small R 2 substituent have different docking poses than others. The docking pose of compound 06 illustrated in Fig. 3a-b indicates that the oxygen atoms in 1,4-naphthoquinone moiety and amino fragment of acylamino moiety have H-bonds with key residues Lys254 and Lys352. The docking pose of compound 18 illustrated in Fig. 3c-d indicates that the H-bonds with key residues Lys254 and Lys352 are formed with the oxygen atoms in 1,3dimethoxybenzene moiety and amino fragment of acylamino moiety. It indicates that compounds with the large alkyl moiety in R 2 substituent have different docking poses. In addition, Fig. 3 indicates that compound 06 also has H-bonds with residues Asn249 using its amine group and has hydrophobic contacts with residues Cys241, Leu248, Asn249, Ala250, Lys254, Leu255, Ala316, and Lys352. For compound 18, it also has two π-cation interactions with residues Asn258 and Lys352, and it has hydrophobic contacts with residues Gln247, Leu248, Ala250, Lys254, Leu255, Asn258, Met259, Val315, and Lys352. These interactions and hydrophobic contacts supported each compound to bind stabilized in the binding site of β-tubulin protein.
Pharmacophore model In the CoMSIA model, the favorable and unfavorable regions of steric eld displayed in Fig. 6b are similar to CoMFA model, which has favorable regions around the R 1 substituent of compound 18 and unfavorable regions closed to the terminal of R 2 substituent of compound 18. Figure 6c indicates that there are a favored hydrophobic eld (blue) closed to the terminal of R 2 substituent toward the direction of residues Ala316, Lys352 and a disfavored hydrophobic eld (white) closed to the terminal of R 1 substituent. Figure 6d introduced the favored (cyan) and disfavored (purple) areas of H-bond donor may further improve the activity.