Cancer is the second leading cause of death, globally, and was responsible for 8.8 million deaths in 2015. Globally, nearly 1 in 6 deaths are due to cancer [1
]. Chemotherapy is generally the main treatment for various cancers. Chemotherapeutic agents (anti-cancer drugs) have a range of side-effects such as immunosuppression, myelosuppression, anemia, teratogenicity, infertility, and even secondary neoplasm [2
]. The major goal of oncology scientists is to design a selective and effective anticancer agent that is only sensitive in normal cancer cells, as well as the ability to predict, alter, or block the hallmark of cancer cells and is likely to improve the therapeutic index [7
]. Therefore, the search for a targeted, effective drug with minimum toxicity is urgently necessary [3
Previous studies have shown that the phosphatidylinositol-3-kinase (PI3K) signaling pathway is a crucial one for many aspects of cell growth and survival. Abnormalities in the PI3K pathway are common in cancer and have a role to play in neoplastic transformation [8
]. The most frequent genetic aberrations in cancer are linked to somatic missense mutations in the gene encoding PIK3CA (p110α) [9
]. Given the important role of the PI3K signaling pathway, some selective inhibitors—PX-866 and PEG Wortmannin—have entered into preclinical status [10
Imidazolium salts serve as the nuclear skeleton in many compounds with anticancer activity [12
], and some of them showed an inhibited effect of PI3K [15
]. A series of imidazolium salt derivatives were designed and synthesized by molecular hybridization tools in the prior research, with the hybrid compound demonstrating potent cytotoxic activity against HL-60, A549 and MCF-7 tumor cell lines (the 77 hybrid compounds with the mean IC50
values of 2.84 μM) [18
]. There was no further structure-function relationship, target or mechanism with respect to these novel imidazolium salt derivatives.
Structural modification of a familiar natural product, active compound or clinical drug is an efficient method for designing a novel drug. The main purpose of structural modification is to reduce the toxicity of target compound, while enhancing the utility of the drug [20
]. This is generally done by altering the key substituent group in the nuclear skeleton of target compounds to increase the binding affinity and specificity to the active site of receptor protein, and improve ADME (absorption, distribution, metabolic and excretion), and changing the lipid-aqueous partition [20
]. The most important step in drug design is to predict the target of a given compound and investigate the binding affinity for and specificity to the active target, which is achievable through the application of Computer-Aided Drug Design (CADD) techniques, which can improve the efficiency of this process [22
Target identification is a fundamental step in the drug design pipeline and process, and makes use of PharmMapper. PharmMapper is a freely accessible web-based tool that is utilized for predicting the potential drug targets via a “reverse pharmacophore” (also known as “target fishing”) mapping method [23
]. Benefiting from a highly efficient and robust mapping method, PharmMapper, with its high-throughput ability, is able to identify the potential target candidates from the database with a runtime of a few hours [23
Protein–protein interactions (PPIs) can illustrate the interaction between two or more protein molecules that share a substrate in a metabolic pathway, regulate each other transcriptionally, or participate in larger multi-protein assemblies, under the PPI network [24
]. Cancer-related proteins obtained by reverse docking techniques using the PharmMapper platform and the STRING database will be combined together to construct the PPI network. The weight of a node in the PPI network was determined by its own properties and its associated edges [25
]; three centrality measures calculated by CytoNCA plugins Subgraph Centrality, Betweenness Centrality and Closeness Centrality [26
] were utilized to generate the sub-network and to screen the potential target of the imidazolium salt derivatives.
The 3D-QSAR (three-dimension-quantitative structure-activity relationship) and docking techniques were regarded as effective and useful tools for drug discovery, a combined method consisting of ligand-based 3D-QSAR and receptor-based docking was utilized to identify the structural requirements of the imidazolium salt derivatives. A molecular dynamics (MD) simulation was utilized to estimate the strength of the intermolecular interaction between the imidazolium salt derivatives and their putative target. This integrated in silicon study not only limits these imidazolium salt derivatives, but could also serve as a guideline for identifying the target of other derivatives with potent activity and to modify the structure of these compounds based on the structure-activity relationship information obtained from the 3D-QSAR and docking study.
2. Results and Discussion
After executing the fragment method in Figure 1
a, the topomer CoMFA gave the q2
values of 0.648 and r2
values of 0.896, with 6 optimum components. The database results of CoMSIA alignment is shown in Figure 1
b; all 77 compounds were aligned based on the template and common moiety, which also provided reliable statistical values: q2
of 0.714, r2
of 0.925, with optimum components of 7. Other statistical values, such as SEE, MAE, F-test value and predictive r2
value are shown in Table 1
The correlation coefficient (R2
) between the Experimental pIC50
and predicted pIC50
of all 77 compounds is shown in Figure 2
; the correlation coefficients of the topomer CoMFA and CoMSIA models were found to be 0.9027 and 0.9204, respectively, showing that the topomer CoMFA and CoMSIA models were reliable and precise for the prediction of activity.
The steric and electrostatic contour maps of topomer CoMFA are shown in Figure 3
a–d, and the hydrophobic and hydrogen-bond acceptor contour maps of CoMSIA are shown in Figure 3
e,f. The structure of the most active compound—compound 72—was selected as the reference structure for the generation and visualization of the topomer CoMFA and CoMSIA contour maps.
The steric contour maps of the results of topomer CoMFA for fragment 1 and fragment 2 are shown in Figure 3
a,b, respectively. In Figure 3
a, the green contour map around the R1 substituent of imidazole/triazole ring indicates that this region was favorable for bulky groups. The results can be proved by the fact that compounds 72, 73 and 74 (with IC50
values of 0.45, 0.68, and 0.58 μM, respectively) with 5,6-Dimethyl-benzimidazole exhibit more potent cytotoxic activities than compound 47 and 45 (with IC50
values of 1.75 and 2.17 μM) without substituents in the imidazole ring. The green region was also near the R3 substituent; hence, the compounds with methyl in R2 shown (such as compound 14 with IC50
values of 4.18 μM and compound 07 with IC50
values of 5.94 μM) will increase their cytotoxic activities.
The configuration change of the substituent group in R2 occurred when the hydroxyl in this place (in the upper right corner of Figure 3
a) and the yellow region neared the substituent groups directly connected to the hydroxyl, which revealed that this place was sterically unfavorable for functional groups; for example, compounds 40 and 51, with 4-Bromobenzyl (with IC50
values of 1.73 and 1.09 μM, respectively) in the R2 substituent group, were more positively charged, resulting in more potent activity than compounds 36 and 53 with 4-Bromophenacyl (with IC50
values of 7.31 and 1.9 μM, respectively).
b shows the steric contour map of fragment 2, which is composed of two entirely different types of tricyclic substituent groups: flexible hexahydropyrrolo[2,3-b]indole substituents and rigid 9H-fluorene substituents. The green and yellow regions above the tricyclic substituent groups neared the 9H-fluorene and hexahydropyrrolo[2,3-b]indole, respectively; the combined region indicates that the compounds with 9H-fluorene substituents (35 compounds with mean IC50
values of 2.58 μM) demonstrated more potent activity than compounds with hexahydropyrrolo[2,3-b]indole substituents (42 compounds with mean IC50
values of 3.06 μM). In the case of hexahydropyrrolo[2,3-b]indole, the green region neared the N
-benzyl group, revealing that the N
-position was favorable for bulky groups, so compounds 23 and 24, possessing the N-benzyl moiety (with IC50
values of 1.27 and 1.29 μM, respectively), demonstrated higher activity than compounds 40 and 41 without substituents in the N-position (with IC50
values of 1.73 and 1.52 μM, respectively).
The topomer CoMFA electrostatic contour map of fragment 1 and 2 are shown in Figure 3
c,d, respectively, in Figure 3
c, the blue regions are found around the R1 and R3 regions, demonstrating that these positions were favorable for the electropositive group. Compounds in which the electropositive 5,6-Dimethyl-benzimidazole and 2-Methyl-imidazole moieties were present (compounds 4 and 72, with IC50
values of 0.47 and 0.45 μM, respectively) will demonstrate higher cytotoxic activity than compounds with imidazole, triazole and benzimidazole (compounds 46, 57 and 30, with IC50
values 3.49, 2.80 and 8.29 μM, respectively). The blue region near and around the benzene ring revealed that, whether or not there was hydroxyl in R2, the electropositive group was favorable for the activity, hence the compounds with electropositive methoxy and naphthyl benzene ring groups substituent in the benzene ring demonstrated more potent cytotoxic activity than those with electronegative fluorine and bromine substituents. It can be proved that compounds 36 and 40, with 4-Bromophenacyl and 4-Bromobenzyl (IC50
= 7.31 and 1.73 μM) exhibited weaker cytotoxic activity than the compounds 37, 38 and 42, with 4-methoxyphenacyl, 2-naphthylacyl and 2-Naphthylmethyl (with IC50
values of 6.23, 1.6 and 1.35 μM, respectively). In addition, the more the electropositive group is replaced in the benzene ring, the more the activity will decrease. For example, compound 75 (with IC50
values of 1.78 μM) with fluorine substituent in the benzene ring demonstrated weaker activity than compound 74 (with IC50
values of IC50
= 0.58 μM) with bromine substituent. There is no distinct red region in the electrostatic contour map of fragment 1. In Figure 3
d, the blue region near the N
-benzyl group indicates that this place is favorable for electropositive groups, and compounds with benzyl groups demonstrated more potent activity than compounds without substituents in the N
-benzyl group, which is consistent with the steric contour map of fragment 2.
The CoMSIA hydrophobic contour is shown in Figure 3
e, the yellow region indicates that R1 was favorable for hydrophobic groups, which can be validated by the fact that the compounds with bicyclic Benzimidazole skeletons exhibited more potent activity than the monocyclic imidazole skeletons. The yellow region can also be seen in the benzyl at the R2 and R3 positions, so the more hydrophobic group naphthyl in R2 position and methyl in R3 position demonstrated higher cytotoxic activity (compound 71, with an IC50
value of 0.57 μM) than other compounds that did not have this moiety. The white region around hexahydropyrrolo[2,3-b]indole indicates that the introduction of this hydrophobic skeleton was unfavorable for the cytotoxic activity, which is in accordance with the steric contour map of fragment 2; substituent groups with weak hydrophobicity were more beneficial for the activity than hydrophobic groups.
The CoMSIA hydrogen bond acceptor contour map is shown in Figure 3
f, the red regions near the two sites of imidazole indicate that the introduction of the hydrogen bond acceptor carbonyl in this position was unfavorable for cytotoxic activity, which corresponds with the results of the topomer CoMFA steric contour map. There was no distinct magenta region around the compound, because the structural modifications of the hydrogen bond acceptors of compounds did not exhibit good biological activity.
We received 3000 protein targets for 10 active compounds from the PharmMapper result list. After removing duplicates, 722 targets were used for screening cancer-related proteins; only proteins with clear cancer drug ligands can be used for further PPI analysis. Finally, 27 cancer-related proteins were identified, as shown in Table 2
. Then all the proteins were uploaded to the STRING database to find their direct and functional partners and to obtain the primary PPI network of each protein. The software Cytoscape 3.5.0 (U.S. National Institute of General Medical Sciences, Bethesda, MD, USA) was utilized to merge the PPI network and analyze the merged network using its functional plugins. Finally, a network with 104 nodes and 496 edges was obtained, as shown in Figure 4
The CytoNCA plugin was utilized to calculate the Subgraph Centrality, Betweenness Centrality and Closeness Centrality of all 104 nodes. After calculating the eight centrality measures of all nodes (see Table S1
), all nodes were sorted by three centrality measures in descending order, the top 10% of the three centrality measures are colored black, blue and white respectively, and the overlapping nodes are colored with color mixtures, with the colored network being shown in the Figure. The top 10% of the three centrality measures were merged together to generate the sub-network with essential nodes, as shown in Figure 5
b; in the merged list (see Table 3
), the nodes are sorted by the three comprehensive centrality measures in descending order. The PIK3CA had the highest values for the three centrality measures, which indicates that PIK3CA produces some interactions with other proteins, and serves as a more essential node than the other proteins obtained from PharmMapper, hence the PIK3CA (with PDBID: 3ZIM in the PharmMapper results) was selected for the further docking study.
The 10 compounds had fit scores of 9.14 (compound 04), 7.77 (compound 54), 7.96 (compound 64), 8.25 (compound 66), 7.70 (compound 67) 7.99 (compound 70), 8.32 (compound 71), 8.52 (compound 72), 8.91 (compound 74) and 8.46 (compound 76) (see Table S2
). The ligand interaction of compound 04, 72 and 74 is exhibited in Figure 5
b–d, analyzing the interaction between the substituents in the compound skeleton and the key amino acids in the active pocket.
a exhibits the combinations of selective targeted covalent inhibitor CNX-1351 and PIK3CA, this ligand-receptor interaction was selected as the reference for analyzing the docking results of the three compounds, because the ligand-binding affinity to the receptor has been verified by in vitro biological experiment [29
In Figure 5
a, the inhibitor CNX-1351 formed several important interactions with amino acids Asp810, Val851, and Ile932. Similarly, in the three docking compounds, the Val851 formed arene–H interactions with the benzyl in N
-benzyl group; this docking result was consistent with the steric, electrostatic and hydrophobic contour maps of topomer CoMFA and CoMSIA, which revealed that the benzyl in this position was beneficial for the improvement of the activity. In compounds 72 and 74, the Val 851 and Ile932 formed arene–H interactions with the 9H-fluorene moiety of the compound skeletons, while the hexahydropyrrolo[2,3-b]indole moiety of compound 04 did not form any interaction with the amino acid, indicating that compounds with rigid 9H-fluorene substituents were more similar to inhibitor CNX-1351 than compounds with flexible hexahydropyrrolo[2,3-b]indole rings; this also confirms the topomer CoMFA steric contour map and the CoMSIA hydrophobic contour map. In the docking results for compounds 04 and 72, the Val851 and Lys702 formed arene–H interactions with phenyl in the R2 and R1 positions, respectively, and as the result of CoMFA and CoMSIA, bulky, electropositive and hydrophobic groups were favorable for the cytotoxic activity. The imidazolium moieties of compounds 04, 72 and 74 formed ion contacts with acidic amino acids Asp933 and Glu849; the imidazolium served as the core moiety in a series of imidazolium salt derivatives. While performing structural modification of the imidazolium moieties of compounds 45, 46, 47, 49 and 50 (with IC50
values of 2.17, 3.49, 1.75, 1.10 and 1.01 μM, respectively), we found that the remodeled compounds 56, 57, 58, 59 and 60, which possess the triazolium moiety (with IC50
values of 2.05, 8.29, 2.07, 2.55 and 1.70 μM, respectively) did not exhibit any improvement in cytotoxic activity. Additionally, some greasy amino acids, basic amino acids and acidic amino acids around compounds 04, 72 and 74 were also similar to the inhibitor CNX-1351.
The molecular dynamics simulation results for the 04-3zim complex and the 72-3zim complex at different temperatures is shown in Figure 6
. The values of root mean square deviations (RMSD) indicate that the transformation of the ligand-protein complex backbone forms the initial structure. In results of the compound 04-3zim, the RMSD values increased to 2.2 Å in 2.3 ns, and then maintained at between 2.5–3 Å until 5 ns; similarly, in the 72-3zim complex, the RMSD increased to 2.5 Å in 2.3 ns and then maintained at between 2.6–3.2 Å until 5 ns. The mean RMSD values of the 04-3zim complex and the 72-3zim complex were 2.115 and 2.253, respectively. The MD results revealed that these two active compounds had potent binding capacity and stability to the putative receptor.
Finally, the structural requirement of a series of imidazolium salt derivatives was identified by a comprehensive method that combined ligand-based QSAR study and receptor-based docking-MD simulation (see Figure 7