CoMFA Model of Anti-tumor Activity for Fluoroquinolon-3-yl s-Triazole Sulfide-ketone Derivatives and Implications for Molecular Design①

CHEN Kui-Qing FENG Chang-Jun

(School of Chemistry & Chemical Engineering, Xuzhou University of Technology, Xuzhou 221018, China)

ABSTRACT Comparative molecular field analysis (CoMFA) techniques were used to perform three-dimensional quantitative structure-activity relationship (3D-QSAR) studies on the anti-tumor activity (pIH and pIC) of 28 fluoroquinolon-3-yl s-triazole sulfide-ketone derivatives (FQTSDs) against two cancer cell lines, including human hepatoma Hep-3B cells and human pancreatic cancer Capan-1 cells. 23 compounds were randomly selected as the training set to establish the prediction models, which were verified by the test set of 6 compounds containing template molecule. The obtained cross-validation (Rcv2) and non-cross-validation correlation coefficients (R2) of the CoMFA models were 0.477 and 0.850 for pIH, and 0.421 and 0.836 for pIC, respectively. The contributions of steric and electrostatic fields to pIH were determined to be 48.1% and 51.9%, and those to pIC were 49.4% and 50.6%,respectively. The CoMFA models were then used to predict the activities of the compounds in the training and testing sets, and the models had a strong stability and good predictability. Based on the 3D contour maps, four novel FQTSDs with a higher anti-tumor activity were designed. However, the effectiveness of these novel FQTSDs is still needed to be verified by experimental results.

Keywords: fluoroquinolon-3-yl s-triazole sulfide-ketone derivative, anti-tumor activity, 3D-QSAR,comparative molecular field analysis, molecular design; DOI: 10.14102/j.cnki.0254–5861. 2011–3000

1 INTRODUCTION

Fluoroquinolone antibiotics are a class of drugs commonly used by people and animals. Because of the broad-spectrum antibacterial property, strong antibacterial activity, no cross-resistance with other antibacterial drugs and little toxic and side effects, they are widely used in animal husbandry,aquatic products and other aquaculture industries. Since topoisomerase (Topo), the antibacterial target of fluoroquinolones, is also an important target of anti-tumor drugs,the antibacterial activity of fluoroquinolones can be regarded as anti-tumor activity[1,2].

It is the most economical and reliable strategy in the research and development of new drugs to modify the structure of existing drugs to obtain lead compounds with the potential to be developed into powerful drugs[3]. Among various strategies, rational drug molecular design based on the substitution of bioelectronic isoelectric isomer of pharmacophore and the combination of pharmacophore is a common method in the construction of new drug molecules[4].The structural modification of fluoroquinolones showed that C-3 carboxyl group was a pharmacophore necessary for antibacterial activity but not for anti-tumor activity, and the antibacterial activity can be transformed into anti-tumor activity by the substitution of heterocycles or condensed heterocycles[5].

In this context, much research work has been conducted on the design of anti-tumor fluoroquinolone molecules. Ni et al.[6]designed and synthesized 14 fluoroquinolone C-3 s-triazole thioether ketone derivatives by splicing effective pharmacophore or shifting dominant drug skeleton.Meanwhile, Xie et al.[7]designed and synthesized 14 fluoroquinolone C-3 s-triazole thioether ketone derivatives by using s-triazole as C3 carboxyl iso-isomer of antibacterial fluoroquinolones, based on the bioelectronic iso-isomer design principle of pharmacophore. These 28 compounds are collectively called fluoroquinolon-3-yl s-triazole sulfideketone derivatives (FQTSDs). The authors[6,7]further evaluated the anti-tumor activities of these FQTSDs to human hepatoma Hep-3B cells and human pancreatic cancer Capan-1 cells (expressed as “IC50”, unit: ?mol·dm-3) by the methyl thiazolyl tetrazolium (MTT) assay, with doxorubicin as the positive control drug.

Quantitative structure-activity relationship (QSAR) plays an important role in predicting the biological activity of compounds and is becoming increasingly important in the field of chemistry, medicine, environment and other disciplines[8-13]. The research methods of QSAR mainly include two-dimensional QSAR (2D-QSAR) and threedimensional QSAR (3D-QSAR) methods[14-17]. The major difference is that 3D-QSAR methods incorporate the three-dimensional conformational properties of bioactive molecules. Therefore, they can reflect the real images of interactions between bioactive molecules and receptors more accurately, and reveal the mechanisms of drug-receptor interactions more deeply. Moreover, the fitting results of 3D-QSAR are generally better than 2D-QSAR. In recent years, 3D-QSAR methods have been applied extensively in correlating chemical structure features with biological activities to obtain some basis information for further compound design and synthesis. Especially, the comparative molecular field analysis (CoMFA)[14-17]is a widely used 3D-QSAR method that establishes relationships between the biological activity of drug molecules and their steric and electrostatic fields. The Van Der Waals potentials and Coulomb potentials, which represent steric and electrostatic fields, respectively, can be calculated by the standard Tripos force field with the Gasteiger-Hückel charges.

In this study, we aim to establish 3D-QSAR models of the anti-tumor activity[6,7]of FQTSDs against two tumor cell lines (Hep-3B and Capan-1) by CoMFA method. According to the established model, the key structural features affecting anti-tumor activity were revealed, and novel molecules with a higher anti-tumor activity were further designed for future potential applications.

2MATERIALS AND METHODS

2. 1 Studied compounds and their anti-tumor activity data

28 FQTSDs[6,7]were tested for cytotoxic activity (IC50,Half maximal inhibitory concentration) against two tumor cell lines (Hep-3B and Capan-1) by invitroMTT assay. TheIC50for Hep-3B and Capan-1 are denoted by "IH" and "IC",respectively. The basic structures of these compounds are shown in Fig. 1, with the corresponding substituents Ar and theIHandICdata listed in Table 1.

Table 1. Experimental and Predicted pIH and pIC Values of FQTSDs Using the CoMFA Models

23 4-CH3O-Ph 27.5 21.6 4.561 4.634 4.666 4.772 24 4-CH3-Ph 28.3 25.4 4.548 4.462 4.595 4.509 25 4-Cl-Ph 15.2 10.3 4.818 4.916 4.987 5.143 26 4-Br-Ph 17.6 12.7 4.755 4.908 4.896 5.132 27* 4-F-Ph 4.7 3.2 5.328 4.971 5.495 5.218 28 4-NO2-Ph 4.3 1.5 5.367 5.117 5.824 5.435 29 4-SO3H-Ph 5.185 5.531 30 4-SO2NH2-Ph 5.094 5.421 31 4-CO2H-Ph 5.369 5.803 32 4-CO2CH3-Ph 5.138 5.448

Fig. 1. Basic structures of FQTSDs

According to the equilibrium principle of physical chemistry, there is a logarithmic relationship between free energy change (ΔGr) and concentration (c)[18]. Therefore, by taking the negative logarithm ofIC50(pIC= –logIC50), all initial anti-tumor activity (IHandIC) values[6,7]are converted into pIHand pICto be used as dependent variables in CoMFA research. The pIHand pICvalues of the 28 FQTSDs are also shown in Table 1. The data set was randomly divided into a training set of 23 compounds for model generation and a test set of 6 compounds (including template molecule, marked with “*” in Table 1) for model validation.

2. 2 Molecular modeling and alignment

All molecular modeling and calculations for CoMFA[19-22]were carried out by Sybyl package (Latest version Sybyl-x2.1.1, Tripos Inc., St. Louis, MO, USA) on a LINUX RH6.4/WIN7 64BIT/WIN864BIT/MAC operating system.The energy minimization was performed using the Tripos force field with a distance-dependent dielectric function and Powell conjugate gradient algorithm with a convergence criterion of 0.005 kcal/(mol??), and the number of iterations was 1000. Partial atomic charges were calculated using the Gastieger-Huckel method. The remaining parameters used in CoMFA were all default unless otherwise stated. Structural alignment is considered as one of the most sensitive parameters in CoMFA analysis. The accuracy of the prediction of CoMFA model and the reliability of contour maps are directly dependent on the structural alignment rule.The most active compound,i.e., No. 28 for Hep-3B and Capan-1, was selected as the template for the molecular alignment.

The common substructures of these molecules, which are composed of the non-hydrogen atoms, are shown in Fig. 2.Each analog was aligned to the template by rotation and translation using the DATABASE ALIGNMENT command in Sybyl so as to minimize root mean square deviation(RMSD) between atoms in the template and the corresponding atoms in the analog. These alignments were subsequently used in calculations of CoMFA probe interaction energy. The alignment results based on the above strategies are shown in Fig. 3.

Fig. 2. Common substructure of the molecules employed for molecule alignment

Fig. 3. 3D view of all the aligned molecules in the training set (a) and test set (b)

2. 3 Model generation and validation

CoMFA fields, which were the interaction energies between a probe atom or a molecule and a set of aligned molecules, are used to establish the 3D-QSAR model. The steric and electrostatic contributions were truncated at 30 kcal/mol, and electrostatic contributions were dropped at lattice intersections with maximum steric interaction.

Partial least-squares (PLS) analysis was used to construct a linear correlation between the 3D-fields (independent variables) and the anti-tumor activity values (pIHand pIC, as dependent variables) of FQTSDs. First of all, leave-one-out cross-validation method was used to determine the cross-validation correlation coefficients(Rcv2) and the optimal number (N) of components. Generally, whenRcv2is greater than > 0.3, the established model is statistically significant at the significance level of 5%,i.e., the reliability of the model is 95%[23]. Moreover, the model is required to have aSv (The ratio of compound number (m) to variable number (n)) of ≥ 5 has statistical significance and low probability[24]. In addition, non-cross-validation method was also used to further assess the robustness and statistical confidence of the derived models. According to the general statistical standard, a value ofR2≥ 0.80 indicates that the model fitting is good[25].

3 RESULTS AND DISCUSSION

3. 1 The established 3D-QSAR models

Based on the CoMFA method, the 3D-QSAR models of the anti-tumor activity of FQTSDs on two kinds of cells were constructed, and the results are shown in Table 2, whereNis the optimal component number,Rcv2the cross-validation correlation coefficient,R2the non-cross-validation correlation coefficient,SDthe estimated standard error,Fthe Fisher value,and St. and El. represent the contribution rates of stereo and electrostatic fields, respectively.

3. 2 Quality evaluation of the 3D-QSAR models

(1) Cross-validation: TheRcv2of the two models is both greater than 0.3 (Table 2), which indicates that they have good prediction ability and robustness. Furthermore, it meets the current preferred standard ofRcv2> 0.4[26]. TheSv equals to 5.75 (Sv = 23/4), larger than 5, indicating that the models are statistically significant.

Table 2. Fitting Results of the CoMFA Models for pIH and pIC

(2) Non-cross-validation: TheR2is both greater than 0.8(Table 2), suggesting the two models have good correlation and stability. Thus, over 83.6% of the change of anti-tumor activity can be explained by field energy of FQTSDs, and only less than 16.4% is attributed to a random factor. At the significant level of 95%, the critical value of statistical variance ratio (F) isF0.05(4,18) = 2.93[27]. TheFvalue of these two models is obviously greater than 2.93, suggesting that the relationship between dependent and independent variables is very significant.

Then, the calculated values (for the training set) and predicted values (for the test set) for pIHand pICof FQTSDs are obtained by using these two CoMFA models. These values are very close to the corresponding experimental data(the minimum deviation for pIHis 0.357, and the deviation for pICis less than 0.389, see Table 1 and Fig. 4 for specific data). All statistical parameters show that the models generated from CoMFA method are reasonable and have good predictability.

Fig. 4. Scatter plots of the experimental versus predicted pIH and pIC values of FQTSDs

(3) Field energy contribution: The steric and electrostatic contributions were found to be close to 48.8% and 50.8%,respectively. Electrostatic field has a slightly greater influence than the steric field, which indicates that the electrostatic interaction of ligands may be a slightly more important influencing factor of anti-tumor activity.

3. 3 Main factors affecting the anti-tumor activity from a combined CoMFA analysis

The results of CoMFA can be displayed as vivid 3Dcontour maps of steric and electrostatic fields, providing an effective way to explain the observed variance in the anticancer activity (expressed as pIHand pIC). CoMFA steric map is represented by green and yellow contours: green contours indicate the regions where the existence of bulky group would be favorable for anticancer activity, while yellow contours represent regions where bulky group would decrease the activity. In the electrostatic field map, blue and red contours denote the regions that the anticancer activity would be increased by the presence of electron-donating and electron-withdrawing groups, respectively.

3. 3. 1 Steric contour maps

Fig. 5 shows the steric contour maps for the CoMFA models of pIHand pICwith the most active molecule 28 as a reference. In the contour map, two models are very similar.The spatial distribution of the stereo field shows that the introduction of a larger group at the 4-position (green region)of the benzene ring and the presence of two smaller groups at the distant 3,5-position (yellow region) of the benzene ring are both conducive to improving the anti-tumor activity. For example, when the group at the 4-position of benzene ring is-NO2group, the pIHand pICare generally larger (Table 1:Nos. 7, 14, 21 and 28. Especially, the Nos. 14, 21 and 28 molecules have the largest activity value in the same group of compounds.).

3. 3. 2 Electrostatic contour maps

Fig. 6 shows the electrostatic contour maps for the CoMFA models of pIHand pICwith the most active compound 28 as a reference. The two figures are also very similar, which have two smaller blue regions near and on the 4-position of benzene ring, a larger red region far away from the 4-position,and a slender blue region at thiosemicarbazone. For the compounds containing thiosemicarbazone (No. 8～14 and 22～28), the H atoms with partial positive charge in thiosemicarbazone is in the blue region, so their anti-tumor activities are stronger than those of compounds No. 1～7 and 15～21. The electron-withdrawing ability of F, Cl and Br decreased in sequence, and thus the positive electricity of the neighboring 4-carbon atom in the halogenated FQTSDs was decreased correspondingly, leading to a decline in the anti-tumor activity, just as the case of No. 27 > 25 > 26. The anti-tumor activity of the compounds with 4-NO2group is almost the strongest in the same class, because the two negative oxygen atoms are located in the red region, while the positive nitrogen atoms in the blue region. This further verified the findings of Ni et al.[6]and Xie et al.[7]that the electron-withdrawing groups (such as F and N) in the benzene ring and the introduction of thiosemicarbazone can significantly enhance the anti-tumor activity.

Fig. 5. CoMFA steric contour maps with compound 28 as a reference

Fig. 6. CoMFA electrostatic field contour maps with compound 28 as a reference

3. 3 Theoretical design of new FQTSDs with higher anti-tumor activity

One of the purposes of QSAR research is to design molecules from the structural information implied in the established model and to predict the biological activity of the designed molecules by the model, providing theoretical basis for the synthesis of new highly active molecules. It can be seen from the above discussions on the spatial distribution of electrostatic and stereo fields that the introduction of a group similar to nitro group at the 4-position of benzene ring is conductive to improving anti-tumor activity. According to this rule, four new compounds were designed (Table 1,compounds 29～32), and their pIHand pICvalues were further predicted by the two 3D-QSAR models. For most of these compounds, they have even better anti-tumor activity than the most effective drugs reported in the literature. To be specific, the anti-tumor activity of molecules 29, 31 and 32 is stronger than that of template molecule (No. 28) due to the fact that the electronegativity of nitrogen atom is greater than that of sulfur and carbon atoms, and these atoms are located in the blue region. However, it should be pointed out that the anti-tumor activity of No. 30 molecule is weaker than that of template molecule because the connection of electron-donating NH2group reduces the positive charge on the sulfur atom. Despite of the theoretically high anti-tumor activity, these new FQTSDs need to be further verified by biomedical experiments.

4 CONCLUSION

(1) The CoMFA method was applied to systematically study the anti-tumor activities (pIHand pIC) of a series of FQTSDs against two cancer cell lines, including human hepatoma Hep-3B cells and human pancreatic cancer Capan-1 cells. The established CoMFA models showed good fitting and high predictive ability in terms of large correlation coefficient (pIH:R2= 0.850, pIC:R2= 0.836) and cross-validation correlation coefficient (pIH:Rcv2= 0.477, pIC:Rcv2= 0.421).

(2) The two CoMFA models give some insights into the mechanism of the anti-tumor effects of FQTSDs: (a) The action mechanism of FQTSDs on Hep-3B and Capan-1 tumor cells is very similar; (b) The contribution of electrostatic effect is close to that of stereo effect; (c) When the 4-position of benzene ring is introduced with a group similar to nitro structure, the anti-tumor activity will be greatly enhanced.

(3) According to the established 3D-QSAR models, four novel FQTSDs with excellent anti-tumor activity are designed, which needs to be further verified by biomedical experiments.

This work is the first attempt to successfully establish two CoMFA models of anti-tumor activity for FQTSDs, and the information obtained in this study would provide helpful theoretical information for predicting the anticancer activities of related novel FQTSDs derivatives and guiding the synthesis of new potent anticancer drugs through structural modifications.