Three-dimensional quantitative structure activity relationship (QSAR) of cytotoxic active 3,5-diaryl-4,5-dihydropyrazole analogs: a comparative molecular field analysis (CoMFA) revisited study

In vitro antitumor evaluation of the synthesized 46 compounds of 3,5-diaryl-4,5-dihydropyrazoles against EAC cell lines and 3D QSAR study using pharmacophore and Comparative Molecular Field Analysis (CoMFA) methods were described. CoMFA derived QSAR model shows a good conventional squared correlation coefficient r2 and cross validated correlation coefficient r2cv 0.896 and 0.568 respectively. In this analysis steric and electrostatic field contribute to the QSAR equation by 70% and 30% respectively, suggesting that variation in biological activity of the compounds is dominated by differences in steric (van der Waals) interactions. To visualize the CoMFA steric and electrostatic field from partial least squares (PLS) analysis, contour maps are plotted as percentage contribution to the QSAR equation and are associated with the differences in biological activity. Background Pyrazole derivatives exhibit a wide range of biological properties including promising antitumor activity. Furthermore, Aldol condensation assisted organic synthesis has delivered rapid routes to N-containing heterocycles, including pyrazoles. Combining these features, the use of chalconisation-assisted processes will provide rapid access to a targeted dihydropyrazoles library bearing a hydrazino 3D QSAR study using pharmacophore and Comparative Molecular Field Analysis (CoMFA) methods were described for evaluation of antioxidant properties. Results Chalcones promoted 1 of the 2 steps in a rapid, convergent synthesis of a small library of hydrazinyl pyrazole derivatives, all of which exhibited significant antitumor activity against Ehrlich Ascites Carcinoma (EAC) human tumor cell line comparable to that of the natural anticancer doxorubicin, as a reference standard during this study. In order to understand the observed pharmacological properties, quantitative structure-activity relationship (3D QSAR) study was initiated. Conclusions Chalcones heating provides a rapid and expedient route to a series of pyrazoles to investigate their chracterization scavenging properties. Given their favorable properties, in comparison with known anticancer, these pyrazole derivatives are promising leads for further development and optimization.


Introduction
The cancer chemotherapy has entered a new era of molecularly targeted therapeutics, that highly selective and not associated with the serious toxicities of conventional cytotoxic drugs [1]. The urea and thiourea derivatives played an important role in anticancer agents because of their good inhibitory activity against receptor tyrosine kinases (RTKs), protein tyrosine kinases (PTKs), and NADH oxidase which played also critical roles in many aspects of tumorigenesis [2][3][4]. Many pyrazole derivatives were found to possess a wide range of bioactivities, such as anti-viral and anti-tumor [5][6][7].
In continuation to anticancer compounds with EAC cells [9], the biological evaluation indicated that some of synthesized compounds are potent inhibitors of cytotoxic activity. The compound D5 displayed the most potent inhibitory activity with IC 50 of 0.09 μM, which was comparable to the positive control doxorubicin (IC 50 = 7.36 μM) as reference drug. In the present work, we report the synthesis of pyrazole derivatives containing thiourea and urea skeletons D1-30 and F1-16 with CoMFA study, not only for validating the observed pharmacological properties but also, for investigating the most important parameters controlling these properties as anticancer agents.

Chemistry
The synthetic Pyrazoles are provided to be useful synthetic intermediates for synthesis of biologically active deazafolates ring system and pyrimidine nucleosides, which are reported to be significantly active, both in vitro and in vivo [10,11] as dihydrofolate reductase inhibitors [12]. Their cytotoxicity against various experimental tumors were found to be as potentially methotrexate [13,14], one of the most effective antimetabolites currently used in treatment of various solid tumors [15,16]. Owing to our plane to develop an efficient and simple procedure for the synthesis of new antimetabolites [17,18], in this part of new synthesis and due to the development of an efficient and simple method for synthesis of new antitumor agents for some of 3, 5-diaryl-4, 5-dihydropyrazole-1-carbothioamide D1-30 and 1-(3, 5-diaryl-4, 5-dihydropyrazol-1-yl)ethanone F1-16 analogs we presented here this research article. Firstly, Chalcones can be prepared by an aldol condensation between a substituted benzaldehyde and substituted acetophenone in the presence of sodium hydroxide as a catalyst.
This reaction has been found to work without any solvent at all a solid-state reaction [19]. The reaction between substituted benzaldehydes and acetophenones has been used to demonstrate green chemistry in undergraduate chemistry education [20]. In a study investigating green chemistry synthesis, chalcones were also synthesized from the same starting materials in high temperature (200 to 350°C) [21].

Anti-tumor properties
The synthesized compounds (D1-30 and F1-16) were screened for their anti-tumor properties against Ehrlich  Tables 1 and Table 2 it has been noticed that most of the synthesized compounds reveal mild to potent anti-tumor properties against (EAC) human tumor cell lines. Meanwhile, compound D5 exhibit high potent anti-tumor activity with potency (IC 50 , concentration required to produce 50% inhibition of cell growth compared to control experimental) = 0.09 μM, compared with Doxorubicin (IC 50 = 7.36 μM), which was used as a reference standard during this study. In order to understand the observed pharmacological properties, quantitative structure-activity relationship (3D QSAR) study was initiated.

QSAR study
Pharmacophore modeling All compounds were built using the Discovery Studio 2.5 software (Accelrys Inc., San Diego, CA, USA), and optimized by using a CHARMm-like force field [22]. The Hip Hop model treats molecular structures as templates consisting of strategically positioned chemical functions that will bind effectively with complementary functions on receptor ( Table 1). The biologically most important binding functions are deduced from a small set of compounds that cover a broad range of activity. In this paper, considering the relatively high flexibility of ligands under investigation, the conformational analyses are carried out with an energy window high enough to include the bioactive conformation. DS 2.5 automatically generated conformational models for each compound. Diverse conformational models for each compound were generated such that the conformers covered accessible conformational space defined within 20 kcal/mol of the estimated global minimum. The estimation of conformational energy was performed based on the CHARMm force field. CATALYST provides two types of conformational analysis: fast and best quality. The best option was used, specifying 250 as the maximum number of conformers. Hypotheses approximating the 3D QSAR pharmacophore were described as a set of features distributed within 3D space with the correlation to the biological results. Five kinds of surface-accessible functions were considered, including two hydrogen bond acceptors (HBA) and 3 hydrophobic centers, as shown in Figure 1.
Furthermore, to emphasize the importance of an aromatic group of the studied molecules, aromatic ring (AR), which consists of directionality, was also included in the subsequent run. Considering the complexity of the studied molecules, the hypothesis generator was restricted to select only five or less features. During a hypothesis generation run, it attempts to minimize a score function consisting of two terms. One term penalizes the deviation between the estimated activities of the molecules in data set and their experimental values; the other term penalizes the complexity of the hypothesis. After assessing all generated hypotheses, the most plausible one is considered the best, as presented in Table 3 and displayed in Figure 2.

Molecular alignment
However, for most studies, the reliability and efficiency of a CoMFA analysis depend on the correct molecular alignment of the ligands. So we performed a pharmacophore mapping study to select both a proposed bioactive conformer and a superposition rule [23]. It is naturally deduced that the pharmacophore model thus obtained provides a good starting point for CoMFA studies. In  CATALYST, the studied molecules were automatically superimposed on the best pharmacophore model in generation hypothesis workbench using the generated conformers. For each ligand, one aligned conformer based on the lowest RMS deviation of feature atom coordinates from those of the corresponding reference features was superimposed on the hypothesis. In the studying process, we found that the best alignment based on the lowest RMS sometimes was not the most appropriate for a subsequent CoMFA analysis, and artificial adjustment was usually needed to get the reasonable alignment, as shown in Figure 3. In most cases of the CoMFA calculations, the alignment from the simple atom-by-atom fits may be the most widely used procedure. So a rigid alignment was applied to superimpose D-5 compounds onto an unsubstituented template shown in Figure 2 using an atom-by-atom least-squares fit as implemented in the SYBYL FIT option, and compound D-5 with the best biological activity was treated as the reference molecule. A conventional CoMFA was performed with the usually used steric and electrostatic fields implemented in SYBYL. The atomic Gasteiger charges were applied in the determination of the electrostatic field. All CoMFA calculations were performed with SYBYL standard setup (steric and electrostatic fields with Lennard-Jones and Coulomb-type potentials, dielectric constant 1/r, cutoff 30 kcal/mol) using an sp3 carbon atom with a charge of +1.0. The extent and the orientation of the grids surrounding the tested molecules were the same as the grid spacing was set to 2 Å.

GRID size
Once the molecules are aligned a grid or lattice is established which surrounds the set of analogs in potential receptor space. Current CoMFA studies seldom use grid resolution less than 1 Å and, most often, 2 Å. The choice of grid resolution represents a compromise between computational practicality and detailing of the fields. If the grid resolution is too small, the number of field-points (cells) becomes too large to perform a timely analysis. Moreover spatial information on field preference can be lost, through a 'smearing out' effect, if the cells become too small. The grid resolution in the 1 Å to 2 Å range corresponds to, at best, differentiating single carbon-carbon (1.54 Å) from one another.

CoMFA interaction energy
The steric and electrostatic (potential fields) energies were calculated at each lattice intersection of a regularly spaced grid box. The lattice spacing was set a value of 2.0 Å. CoMFA region was defined automatically which extends the lattice walls beyond the dimensions of each structure by 4.0 Å in all directions. The Lennard-Jones potential and coulomb potential energy which represent, respectively steric and electrostatic fields, were calculated using the TRIPOS force fields. A sp3 carbon atom with a van der Waals radius of 1.52 Å and a +1.0 charge served as the probe atom to calculate steric and electrostatic fields. The default value 30.0 kcal/mole was used as the maximum electrostatic and steric energy cutoff.

Partial least squares (PLS) and cross-validation in CoMFA
The last step in a CoMFA is a partial least square analysis to determine the minimal set of grid points which is necessary to explain the biological activities of the compounds. Partial least-square is an iterative procedure that applies two criteria to produce its solution. First, to extract a new component, the criterion is to maximize the degree of commonality between all of the structural parameter columns (independent variable) collectively and the experimental data (dependent variable). Second, in the evaluation phase of a PLS iteration, the criterion for acceptance of the principal component just generated is an improvement in the ability to predict, not to reproduce, the dependent variable, as outlined in Figure 4. The technique used in PLS to assess the predictive ability of a QSAR is cross-validation [24]. Cross-validation is based on the idea that the best way to assess predictive performance is to predict. When cross-validating, one pretends that one or more of the unknown experimental value is, infect, unknown. The analysis being cross-validated is repeated, excluding the temporarily 'unknown' compounds and then using the resulting equation to predict the experimental measurement of the omitted compound(s). The cross-validation cycle is repeated until each compound has been excluded and predicted exactly once. The results of cross-validation are the sum of the squared prediction errors, sometimes called the predicted residual sum of squares (PRESS). For evaluation of the overall analysis, the PRESS is commonly expressed as a cross-validated correlation coefficient r 2 , or xv -r 2 value, as outlined in Figure 5.
The results of the CoMFA studies are summarized in Table 4. From this table it is evident that the CoMFA derived 3D QSAR shows a good cross validated r 2 , (0.568)   Table 5. In this analysis both steric and electrostatic field contribute to the 3D QSAR equation by 73.8% and 26.2%, respectively, suggesting that variation in biological activity of compounds is dominated by differences in steric (van der Waals) interactions.
The real test for model predictiveness is to predict the activity of ligands, which were not used in the model generation. Our test set has 16 compounds, which were randomly kept with bias given to both chemical and biological diversity in both the training set and the test set molecules, as shown in Tables 6.
The CoMFA models exhibited a good predictiveness on these ligands with conventional r 2 0.469 and standard error of estimation 0.372 as the data described in Table 7. The contour maps are plotted as percentage contribution to the QSAR equation and are associated with the differences in biological activity.
In Figures 6(a) the regions of high and low steric tolerance are shown in green and yellow polyhedral, respectively. The areas of high bulk tolerance (80% contribution) are observed near 4-methoxy phenyl of the ligands . The active analogue (5) outlined in Figure 6(b), shows that 4methoxy phenyl ring embedded in the green region. The antitumor activity shown by the different training compounds was due to the presence of bulky groups at position 5 and 3, 4-disbstituted phenyl at position 3 surrounded by green contours. In the present sterically unflavored yellow regions were observed near the 3, 6-substituted phenyl at position 3. The steric bulk in this region has a negative effect on the antitumor activity, suggesting that there is a

CoMFA applications in drug design
There are now a few hundred practical applications of CoMFA in drug design. Most applications are in the field of ligand protein interactions, describing affinity or inhibition constants. In addition, CoMFA has been used to correlate steric and electronic parameters [25]. Less appropriate seems the application of CoMFA to in vivo data, even if lipophilicity is considered as an additional parameter. As most CoMFA applications in drug design have been comprehensively reviewed in three books [26,27] and in some reviews [28,29].

Conclusion
This work demonstrated that the 3D QSAR pharmacophore modeling and MFA technique (CoMFA) could be effectively combined. The best model identified by 3d QSAR pharmacophore search served as suitable modes of superimposition for subsequent CoMFA utilizing compounds (D1-30), as training set. The better results were obtained by superimposing the molecules onto the 3D QSAR pharmacophore model than onto the common structure used in the energy-lowest conformers. Moreover, from our calculations, it can be found that the active conformers and the 3D QSAR pharmacophore mapping automatically afforded by CATALYST are usually not the optimal ones. The manual adjustments are usually needed to obtain the best results. As illustrated by the contour maps, our 3D-QSAR model, which is based on a homogeneous set of ligands, is expected to correctly predict affinities of structurally related compounds. These results will provide useful information in understanding the structural and chemical features required for EAC cells cytotoxicity and in designing new potential compounds. General synthetic procedure of chalcones Equimolar portions of the appropriately substituted aromatic aldehydes (10 mmol, 1 equiv) and ketones (10 mmol, 1 equiv) were dissolved in approximately 15 mL of ethanol. The mixture was allowed to stir for several minutes at 5-10°C. A 10 mL aliquot of a 40% aqueous sodium hydroxide solution was then slowly added dropwise to the reaction flask via a self-equalizing addition funnel. The reaction solution was allowed to stir at room temperature for approximately 4 h. Most  commonly, a precipitate formed and was then collected by suction filtration.

General synthetic procedure of pyrazole derivatives C1-C30
A mixture of chalcone (0.01 mol), thiosemicarbazide (0.01 mol), and NaOH (0.025 mol) was refluxed in ethanol (25 mL) for 8 h. The solution was poured into icewater. The precipitate was filtered and crystallized from methanol to give white powder compounds in good to excellent yields.

General synthetic procedure of pyrazole derivatives F1-F16
A solution of chalcone (5 mmol) in 30 mL of acetic acid was added dropwise to 0.6 mL of hydrazine hydrate (12.5 mmol) and kept under stirring at 120°C for 24 h. The mixture was then poured onto ice-water, the crude pyrazole derivatives D1-D16 were obtained by vacuum filtration, which were crystallized from ethanol to give white powder compounds in good to excellent yields.

Cell growth inhibition assay
The in-vitro growth inhibitory activity of the test compounds against EAC cell line was evaluated in NCI. The evaluation depends on using the standard 48 h exposure assay [30]. Ehrlich Ascites Carcinoma (EAC) cells were obtained by needle aspiration of ascetic fluid from the pre-inoculated mice under aseptic conditions. Tumor cells suspension (2.5 × 10 6 per ml) was prepared in saline. The parent line was kindly supplied by the National Cancer institute (NCI), Cairo, Egypt, Diagnostics Lab. Cairo University, Egypt. The tumor cells were maintained by weekly intra-peritoneal transplantation of cells.

Bioassay in vitro
EAC (Ehrlich Ascites Carcinoma cell line) was obtained from the Pharmacology Unit, Cancer Biology Department, National Cancer Institute, Cairo University, Egypt. Cells were maintained in DMEM medium with 10% foetal calf serum, sodium pyruvate, 100 U/ml penicillin and 100 mg/ml streptomycin at 37 ο C and 5% CO 2 . Potential cytotoxicity of D1-30 and F1-16 were tested using the method of Skeha et al. [31,32] briefly, 104 cells/well were plated onto 96-well dishes overnight before the treatment with the tested compounds to allow the attachment of cells to the wall of the plate. Different concentrations of each tested compound (0.1, 2.5, 5, 10 mg/ml) were added to the cell monolayer; triplicate wells were used for each individual dose. Monolayer cells were incubated with the tested agent(s) for 48 h at 37°C and 5% CO 2 . At the end of the incubation period, the cells were fixed and stained with sulforhodamine B dissolved in acetic acid. Unbound stain was removed by washing four times with 1% acetic acid and the protein bound dye was extracted with tris-EDTA buffer. Absorbance was measured in an ELISA reader. The relation between surviving fraction and compound concentration was plotted to get the survival curve of each tumor cell line and IC 50 , the concentration of an agent that causes a 50% growth inhibition for each tested agent using each cell line was obtained from the survival curve.