Synthesis, Antitumor Evaluation, Molecular Modeling and Quantitative Structure-Activity Relationship (QSAR) of Novel 2-[(4-Amino-6-N-substituted-1,3,5-triazin-2-yl)methylthio]-4-chloro-5-methyl-N-(1H-benzo[d]imidazol-2(3H)-ylidene)Benzenesulfonamides.

A series of novel 2-[(4-amino-6-R2-1,3,5-triazin-2-yl)methylthio]-4-chloro-5-methyl-N-(5-R1-1H-benzo[d]imidazol-2(3H)-ylidene)benzenesulfonamides 6–49 was synthesized by the reaction of 5-substituted ethyl 2-{5-R1-2-[N-(5-chloro-1H-benzo[d]imidazol-2(3H)-ylidene)sulfamoyl]-4-methylphenylthio}acetate with appropriate biguanide hydrochlorides. The most active compounds, 22 and 46, showed significant cytotoxic activity and selectivity against colon (HCT-116), breast (MCF-7) and cervical cancer (HeLa) cell lines (IC50: 7–11 µM; 15–24 µM and 11–18 µM), respectively. Further QSAR (Quantitative Structure–Activity Relationships) studies on the cytotoxic activity of investigated compounds toward HCT-116, MCF-7 and HeLa were performed by using different topological (2D) and conformational (3D) molecular descriptors based on the stepwise multiple linear regression technique (MLR). The QSAR studies allowed us to make three statistically significant and predictive models for them. Moreover, the molecular docking studies were carried out to evaluate the possible binding mode of the most active compounds, 22 and 46, within the active site of the MDM2 protein.


Introduction
Cancer is a major public health problem worldwide and is the second leading cause of death in developed nations. The greatest number of deaths are from cancers of the lung, prostate, colon and rectum in men and the lung, breast, colon and rectum in women [1]. One of the basic methods of cancer treatment is chemotherapy, which uses cytotoxic drugs with systemic effects. Despite years of effort in the field of designing different molecules, there are still few selective drugs against cancer cells as compared with normal cells [2].
The mouse/murine protein, MDM2, a promising target for developing anti-cancer therapies, is an important negative regulator of the p53 tumor suppressor protein [3,4]. Under normal conditions, the MDM2 protein binds to the transactivation domain of p53, preventing its binding to DNA and labelling Our previous works on a search for antitumor agents among benzenesulfonamide derivatives, carried out by Sławinski's group, indicate the importance of the 2-methylthiobenzenesulfonamide fragment for cytotoxic activity of compounds against cervical, breast and colon cancer. We have proved that our compounds showed an apoptotic effect in cancer cells. Continuing the search for more active compounds, we designed and developed a method for the synthesis of new molecules with potential inhibitory activity against the MDM2 protein. We carried out molecular docking for various targets associated with tumors that showed the affinity of designed compounds for MDM2. In this work, we report on a series of 2-[(4-amino-6-R 2 -1,3,5-triazin-2-yl)methylthio]--4-chloro-5methyl-N- (5-R 1 -1H-benzo[d]imidazol-2(3H)-ylidene)benzenesulfonamides designed as molecular hybrids combining the 2-mercaptobenzenosulfonamide fragment with the imidazoline ring ( Figure  1). In the structure of our compounds, we also incorporated different substituents R 2 to investigate their impact on anticancer activity and establish structure-activity relationships. All compounds were tested for their cytotoxic activity against HCT-116, MCF-7 and HeLa cell lines.
The final compounds were characterized by IR and NMR (see representative spectra in Supplementary Materials) spectroscopy as shown in the Experimental Section. Elemental analyses (C, H, N) and HRMS were in accordance with the proposed structures.
For example, in the 1 H NMR spectra the presence of a S-CH 2 -1,3,5-triazine ring and NH groups in the benzimidazole rings of 6-49 were identified from the singlet signals at 3.81-4.05 and two signals at 10.5-12.15 ppm, respectively. Meanwhile the appearance of H-3 and H-6 of the benzene ring at 7.61-7.94 and 7.90-8.00 ppm, respectively, confirmed the proposed structure of final compounds 6-49.

Cytotoxic Activity
Compounds 6-49 were evaluated in vitro for their effects on the viability of the three human cancer cell lines: HCT-116 (colon cancer), MCF-7 (breast cancer) and HeLa (cervical cancer) as well as the non-cancerous keratinocyte cell line (HaCaT). The concentration required for 50% inhibition of cell viability IC 50 was calculated and compared with the reference drug cisplatin, the results are shown in Table 1. Cell lines: colon cancer (HCT-116), breast cancer (MCF-7), cervical cancer (HeLa), the human keratinocyte cell line (HaCaT); NT-not tested; IC 50 was measured at concentrations 1,10,25,50,and 100 µM. IC 50 values are expressed as the mean ± SD of at last three independent experiments.

Quantitative Structure-Activity Relationships (QSARs) of Cytotoxic Activity
Correlation between structure and activity was performed according to QSAR methodology. Three-dimensional structures of all compounds were obtained by applying a conformational search with the LowModeMD method (MOE software) using MMFF94X forcefield (MOE software) followed by geometry optimization with the semi empirical PM6 method (MOPAC2016 software). The energy of final structures were calculated using GAMESS software and the STO3G HF method. Molecular descriptors were calculated using MOE software. In order to obtain QSAR models, stepwise linear progressive regression (a type of MLR) was applied. Compounds with activity above 100 µM were removed from the model development (29 and 32 for the HCT-116 cell line, 7, 18, 27, 29, 32, 39 for the MCF-7 cell line and 7, 27, 29, 15, 32 for the HeLa cell line).
The obtained models correlated cytotoxic activity (IC 50 ) toward cancer cells with different topological (2D) and conformational (3D) molecular descriptors. The equations were statistically significant, they explained 75-86% of the variability of the IC 50 coefficient and were characterized by usefulness of model to predict the antitumor activity of new sulfonamides, as indicated by the values of Q 2 from 68% to 75% (Table 2 and Figure 3). -test set. Molecular descriptors used in the models: a_nO-number of oxygen atoms (The atom count and bond count descriptors); pmi3-third diagonal element of diagonalized moment of inertia tensor (Surface Area, Volume and Shape Descriptors); E_oop-Out-of-plane potential energy (The energy descriptors); GCUT_SLOGP_1 and GCUT_SLOGP_2-descriptors using atomic contribution to logP (using the Wildman and Crippen SlogP method) instead of partial charge (Adjacency and Distance Matrix Descriptors); b_max1len-length of the longest single bond chain (The atom count and bond count descriptors); vsurf_IW6-hydrophilic integy moment (Surface Area, Volume and Shape Descriptors); a_IC-atom information content (total) (The atom count and bond count descriptors); PEOE_VSA+1-sum of vi where qi is in the range [0.05,0.10] (Partial Charge Descriptors); SMR_VSA0-adjacency and distance matrix descriptor (The Subdivided Surface Areas); std_dim3-standard dimension 3: the square root of the third largest eigenvalue of the covariance matrix of the atomic coordinates. A standard dimension is equivalent to the standard deviation along a principal component axis. (Surface Area, Volume and Shape Descriptors); ast_violation-number of Astex fragment-likeness violations (The atom count and bond count descriptors); a_nF-number of fluorine atoms (The atom count and bond count descriptors); b_1rotN-number of rotatable single bonds (The atom count and bond count descriptors); h_pstrain-the strain energy (kcal/mol) needed to convert all protonation states into the input protonation state: (kT ln 10) ( pC1 + log sum {10-pCi} ) (The Hueckel Theory descriptors); pmi-principal moment of inertia (Surface Area, Volume and Shape Descriptors); SlogP_VSA5-represent different aspects of van der Waals surface area's contribution to compound lipophilicity (The Subdivided Surface Areas).
For each model, a residue analysis was carried out to confirm the correctness of used linear regression and to confirm its assumptions (such as demonstration of an absence of deviations from linearity, and normality of residue distribution to confirm homoscedasticity). The predictors that corresponded most with antitumor activity were estimated: a_nO (number of oxygen atoms in the molecule) for the HCT-116 model with a correlation coefficient of 0.50, SMR_VSA0 (adjacency and distance matrix descriptor) for the MCF-7 model with a correlation coefficient of 0.75 and for the HeLa model there is SlogP_VSA5, which represents different aspects of the van der Waals surface area's contribution to lipophilicity with correlation coefficient of 0.55. standard deviation along a principal component axis. (Surface Area, Volume and Shape Descriptors); ast_violation -number of Astex fragment-likeness violations (The atom count and bond count descriptors); a_nF -number of fluorine atoms (The atom count and bond count descriptors); b_1rotN -number of rotatable single bonds (The atom count and bond count descriptors); h_pstrain -the strain energy (kcal/mol) needed to convert all protonation states into the input protonation state: (kT ln 10) ( pC1 + log sum {10-pCi} ) (The Hueckel Theory descriptors); pmiprincipal moment of inertia (Surface Area, Volume and Shape Descriptors); SlogP_VSA5 -represent different aspects of van der Waals surface area's contribution to compound lipophilicity (The Subdivided Surface Areas). From the HCT-116 model, it is clear that higher activity correlates with lower values of the number of oxygen atoms (a_nO), third diagonal element of diagonalized moment of inertia tensor (pmi3), out-of-plane potential energy (E_oop), length of the longest single bond chain (b_max1len), hydrophilic integy moment (vsurf_IW6). On the other hand, the negative coefficient of GCUT_SLOGP_1 shows that the high value of this descriptor is valuable for anticancer activity. The From the HCT-116 model, it is clear that higher activity correlates with lower values of the number of oxygen atoms (a_nO), third diagonal element of diagonalized moment of inertia tensor (pmi3), out-of-plane potential energy (E_oop), length of the longest single bond chain (b_max1len), hydrophilic integy moment (vsurf_IW6). On the other hand, the negative coefficient of GCUT_SLOGP_1 shows that the high value of this descriptor is valuable for anticancer activity. The cytotoxic activity of the compounds against MCF-7 has correlation with six descriptors. Two beneficial impacts were shown: atom information content (a_IC) and shape (std_dim3) descriptors, which prefer high values and b_max1len, GCUT_SLOGP_2, PEOE_VSA+1, SMR_VSA0 descriptors favoring low values. In the HeLa model, it can be noticed that the increase of biological activity relates to higher values of both parameters: the number of fluorine atoms (a_nF) and number of rotatable single bonds (b_1rotN). Increased values of descriptors related to atom counts and bond counts (ast_violation), Huckel theory (h_pstrain), and the structure connectivity and conformation (pmi, SlogP_VSA5) decrease the anticancer activity of molecules.

Molecular Modeling and Docking Results
In order to better understanding the anticancer activity of synthesized compounds, molecular docking was carried out for various therapeutic targets of cancer. It was found that the proper fitting with good energy scores was shown for the MDM2 protein, while the majority of the compounds had a moderate score with other targets, e.g., serine-threonine protein kinases Akt-1 [21], RAF [22] and B-RAF [22] or epidermal growth factor receptor EGFR [23] among others.
Molecular docking of some of the newly synthesized compounds within the active site of the MDM2 protein was performed and the amino acid interactions and docking patterns were investigated using the protein data bank file (PDB ID:5C5A). This file contains the MDM2 protein co-crystalized with Nutlin-3a. The docking procedures were performed by Molecular Operating Environment (MOE, 2018) software. The docking setup was first validated by self-docking of the co-crystallized ligand (Nutlin-3a) in the binding site of the protein, with energy score S = −10.8029 kcal/mol and root mean standard deviation (RMSD) = 0.2534. The ligand interacts with Met62, His96, Gly58, Gln59, Leu54 and Val93 in the active site of MDM2 ( Figure 4). cytotoxic activity of the compounds against MCF-7 has correlation with six descriptors. Two beneficial impacts were shown: atom information content (a_IC) and shape (std_dim3) descriptors, which prefer high values and b_max1len, GCUT_SLOGP_2, PEOE_VSA+1, SMR_VSA0 descriptors favoring low values. In the HeLa model, it can be noticed that the increase of biological activity relates to higher values of both parameters: the number of fluorine atoms (a_nF) and number of rotatable single bonds (b_1rotN). Increased values of descriptors related to atom counts and bond counts (ast_violation), Huckel theory (h_pstrain), and the structure connectivity and conformation (pmi, SlogP_VSA5) decrease the anticancer activity of molecules.

Molecular Modeling and Docking Results
In order to better understanding the anticancer activity of synthesized compounds, molecular docking was carried out for various therapeutic targets of cancer. It was found that the proper fitting with good energy scores was shown for the MDM2 protein, while the majority of the compounds had a moderate score with other targets, e.g., serine-threonine protein kinases Akt-1 [21], RAF [22] and B-RAF [22] or epidermal growth factor receptor EGFR [23] among others.
Molecular docking of some of the newly synthesized compounds within the active site of the MDM2 protein was performed and the amino acid interactions and docking patterns were investigated using the protein data bank file (PDB ID:5C5A). This file contains the MDM2 protein cocrystalized with Nutlin-3a. The docking procedures were performed by Molecular Operating Environment (MOE, 2018) software. The docking setup was first validated by self-docking of the cocrystallized ligand (Nutlin-3a) in the binding site of the protein, with energy score S = −10.8029 kcal/mol and root mean standard deviation (RMSD) = 0.2534. The ligand interacts with Met62, His96, Gly58, Gln59, Leu54 and Val93 in the active site of MDM2 ( Figure 4). Docking of the most active compounds 22, 46 was performed and showed proper fitting in the active site of MDM2 with positive energy scores (S), which supports the observed activity of these compounds as MDM2 inhibitors. The energy score (S) and amino acid interaction of the most potent MDM2 inhibitors are listed in Table 3. The docking results revealed that the amino acids Leu54 and Met62 located in the binding pocket of the protein played an important role. Thus, the most active compounds (22,46) showed interaction with Leu54 and Met62 formed π-H interaction with Leu54 and/or H-bond donor with Met62, which mimics the pattern of interaction of Nutlin-3a with the MDM2 protein ( Figure 4). Interestingly, the 2-(4-phenylpiperazin-1-yl)-1,3,5-triazine fragment is located in a hydrophobic binding pocket near amino acid residues Met62, Leu54, Val93, Gly58, and Gln59, interacting directly with the most important residues Leu54 and Met62 ( Figure 5). Apparently, Docking of the most active compounds 22, 46 was performed and showed proper fitting in the active site of MDM2 with positive energy scores (S), which supports the observed activity of these compounds as MDM2 inhibitors. The energy score (S) and amino acid interaction of the most potent MDM2 inhibitors are listed in Table 3. The docking results revealed that the amino acids Leu54 and Met62 located in the binding pocket of the protein played an important role. Thus, the most active compounds (22,46) showed interaction with Leu54 and Met62 formed π-H interaction with Leu54 and/or H-bond donor with Met62, which mimics the pattern of interaction of Nutlin-3a with the MDM2 protein ( Figure 4). Interestingly, the 2-(4-phenylpiperazin-1-yl)-1,3,5-triazine fragment is located in a hydrophobic binding pocket near amino acid residues Met62, Leu54, Val93, Gly58, and Gln59, interacting directly with the most important residues Leu54 and Met62 ( Figure 5). Apparently, compounds 22 and 46 showed the same orientation within the active site of MDM2, suggesting the binding pattern of these derivatives within the MDM2 protein ( Figure 6). The remaining parts of the molecules, i.e., the benzenesulfonamide fragment and the benzimidazole ring occupy other regions of the protein by interacting with the amino acids Lys51 or Met50, as well Tyr100, respectively. Based on the characterization of the protein-ligand interactions, the 4-phenylpiperazin-1-yl moiety played a key role in forming a H-bond interaction, while both 1,3,5-triazine and benzimidazole rings were responsible for π-H (polar) and aromatic π-π stacking noncovalent interactions ( Table 3). compounds 22 and 46 showed the same orientation within the active site of MDM2, suggesting the binding pattern of these derivatives within the MDM2 protein ( Figure 6). The remaining parts of the molecules, i.e., the benzenesulfonamide fragment and the benzimidazole ring occupy other regions of the protein by interacting with the amino acids Lys51 or Met50, as well Tyr100, respectively. Based on the characterization of the protein-ligand interactions, the 4-phenylpiperazin-1-yl moiety played a key role in forming a H-bond interaction, while both 1,3,5-triazine and benzimidazole rings were responsible for π-H (polar) and aromatic π-π stacking noncovalent interactions ( Table 3).     Ph (1,2,4-trisubstituted) π-H 3.75

Cell Culture and Cell Viability Assay
All chemicals, if not stated otherwise, were obtained from Sigma-Aldrich (St. Louis, MO, USA). The HCT-116 cell lines was purchased from ATCC (ATCC-No: CCL-247), while the MCF-7, HeLa and HaCaT cell lines were purchased from Cell Lines Services (Eppelheim, Germany). Cells were cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum, 2 mM glutamine, 100 units/mL penicillin, and 100 µg/mL streptomycin. Cultures were maintained in a humidified atmosphere with 5% carbon dioxide at 37 • C in an incubator (Heraceus, HeraCell).
Cell viability was examined using the MTT (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide) assay. Cells were seeded in 96-well plates at a density of 5 × 10 3 cells/well and treated for 72 h with the tested compounds in the concentration range of 1-100 µM (1, 10, 25, 50 and 100 µM). Then, MTT (0.5 mg/mL) was added to the medium and cells were further incubated for 2 h at 37 • C. In the next stage, cells were lysed with DMSO and the absorbance of the formazan solution was measured at 550 nm with a plate reader (1420 multilabel counter, Victor, Jügesheim, Germany). The experiment was performed in triplicate. Values are expressed as the mean ± SD of at least three independent experiments. Cisplatin was used as a positive control.

Molecular Docking
All the molecular modeling studies were performed using Molecular Operating Environment (MOE, 2018) software. The partial charges were calculated automatically. All minimizations were performed with MOE until an RMSD gradient of 0.2534 kcal/molÅ with AMBER10 force field (a value below 2.0 kcal/molÅ indicates that the docking protocol was validated).
The X-ray crystallographic structure of MDM2 co-crystalized with Nutlin-3a (PDB ID:5C5A) was downloaded from the protein data bank available at the RCSB Protein Data Bank https://www.rcsb.org/. For each co-crystallized enzyme, water molecules and ligands that were not involved in the binding were removed. The Protonate 3D protocol in MOE with its default options was used to prepare the protein. The co-crystallized ligand (Nutlin-3a) was used to define the binding site for docking. The Triangle Matcher method was used, where 1000 poses were analyzed together with also redocking 1000 poses (to optimize docked structures) using the AMBER10 force field. From each obtained molecular docking result, five poses with the lowest energy were selected. Then one pose was selected that had the most interactions with amino acids in the MDM2 protein binding pocket. The choice of poses also took into account the number of interactions with the amino acids with which the known MDM2 (pdb: 5C5A) Nutlin-3a protein inhibitor binds. The docking scores, types of interactions and the bond lengths are shown in Table 3.
The obtained compounds were tested in vitro for their cytotoxic activity, with the use of the MTT assay, toward colon (HCT-116), breast (MCF-7) and cervical (HeLa) cancer cell lines (IC 50 : 7-11µM; 15-24 µM and 11-18 µM), vs. non-cancerous cells (HaCaT) (IC 50 : 34 µM and 28 µM), respectively. The multiple linear regression technique (MLR) was applied to build up the QSAR model for predicting the cytotoxic activity of novel compounds, based on different topological (2D) and conformational (3D) molecular descriptors. Developed models showed a good predictability and might be useful for further development of structurally similar derivatives with better cytotoxic properties. The molecular docking studies revealed the possible binding mode of the most active compounds 22 and 46 within the active site of the MDM2 protein suggesting that it may be a possible molecular target for the tested compounds.