Analysis of Plant-derived Phytochemicals as Anti-cancer Agents Targeting Cyclin Dependent Kinase-2, Human Topoisomerase IIa and Vascular Endothelial Growth Factor Receptor-2

Cancer is caused by a variety of pathways, involving numerous types of enzymes, among them three enzymes: Cyclin dependent kinase-2 (CDK-2), Human topoisomerase IIα and Vascular Endothelial Growth Factor Receptor-2 (VEGFR-2) are three most common enzymes that are involved in the cancer development. Although many chemical drugs are available in the market, plant sources are known to contain a wide variety of agents that are known to possess anticancer activity. In this experiment, total thirty compounds were analysed against the mentioned enzymes using different tools of bioinformatics and in silico biology like molecular docking study, druglikeness property experiment, ADME/T test, PASS prediction and P450 site of metabolism prediction as well as DFT calculations to determine three best ligands that have the capability to inhibit the mentioned enzymes. Form the experiment, Epigallocatechin gallate was found to be the best ligand to inhibit CDK-2, Daidzein showed best inhibitory activities towards Human topoisomerase IIα and Quercetin was predicted to be the best agent against VEGFR-2. They were also predicted to be quite safe and effective agents to treat cancer. However, more in vivo and in vitro analysis are required to confirm their safety and efficacy in this regard.


Introduction
Cancer is defined as the uncontrolled proliferation and abnormal spread of the body's specific cells. According to WHO, cancer was responsible for 13% of world deaths accounted in 2005.
Moreover, projections have shown that cause-specific years of life lost (YLL) rate due to cancer would increase in 2005, 2015 and 2030. Millions of species of plants, animals, marine organisms and microorganisms act as attractive sources for new therapeutic candidate compounds. However, the development of novel agents from natural sources face many obstacles that are not usually met when one deals with synthetic compounds. Moreover, there may be difficulties with identification, isolation, assessing and obtaining the appropriate amounts of the active compound in the sample.
(1, 2) The search for anti-cancer compounds from plant sources started in earnest in the 1950s with the discovery and development of the various natural compounds like vinca alkaloids, vinblastine, vincristine and cytotoxic podophyllotoxins. In the recent years, new technologies have been developed by the scientists to enhance natural product drug discovery in an industrial manner.
Indeed, several new anticancer agents of natural origin have been introduced to the market recently and there is a promising pipeline of natural products in cancer-related clinical trials (3,4,5,6).
Future advances in the directed biosynthesis of small molecules will improve the ability of the scientists to control the shape and topology of various small molecules and thus creating new anticancer compounds that will interact specifically with biological targets. In the future, plants (300,000-500,000 such species) will continue to be a vital and valuable resource for anticancer drug discovery. More than 60 compounds from different plant sources are currently in the pipeline as potential anticancer agents (7,8,9,10). Many chemical and synthetic drugs are already available for treating cancers i.e. alvocidib, lenvatinib and daunorubicin etc. These chemical drugs have many adverse effects like sepsis, diarrhea, stomach and bladder pain, hair loss, paralysis, joint pain etc. However, plant phytochemicals are considered as safe in this regard since they generally don't possess any adverse effect to the human health in appropriate doses (11,12,13,14). Therefore, using alternatives from plants can have great potential for cancer treatment. Table 01 lists the potential phytochemicals used in the experiment. Table 01. List of the plant derived anti-cancer agents that work via CDK-2, human topoisomerase IIa and VEGFR-2 pathways. NA; Not available.  (68,69,70,71,72,73,74). The inhibitors are proteins from inhibitors of CDK-4 (INK4) and cyclin-dependent kinase inhibitor (CKI) families. CDK-4/6 is inhibited by p15/16

Involvement in Cancer
inhibitors and CDK-2 is inhibited by p21/p27 inhibitors. However, any type of mutation in the CDK genes causing hyperactivity or any type of mutation in the inhibitory genes, may lead to the uncontrolled proliferation of the cells, which can lead to different forms of cancers (75). For this reason, the targeting and inhibition of CDK-2 is a potential strategy for anticancer drug development (76) (Figure 01). Figure 01. The cyclin/CDK signalling pathway. Upon activated by mitogen signal, the cyclin D-CDK-4/6 complex is activated and cause the inactivation of Rb by phosphorylation and thus release the active E2F, which takes part in cell cycle progression. However, E2F activates cyclin E-CDK-2 complex, which phosphorylates the Rb protein and activates the E2F in a feedback loop. P15/p16 inhibitors repress cyclin D-CDK-4/6 complex and p21/p27 inhibit cyclin E-CDK-2. Anti-CDK-2 agents inhibit the CDK-2 protein, thus can help in cancer treatment.  (Figure 02). For this reason, DNA topoisomerase IIα is a potential target for anti-cancer drug development (77,78,79,80,81).

Figure 02.
The DNA topoisomerase IIA pathway in cancer development. Upon the cleavage of the target DNA, the topoisomerase can remain bound to the cleaved ends of the DNA fragments and form cleavage complexes. If the concentration of cleavage complexes falls too much, then this may lead to cell death due to the mitotic failure. Moreover, if the concentration rises too much, abnormal translocations and mutagenesis may occur, which lead to cancer development. Antitopoisomerase agents aid in cancer treatment by inhibiting the activity of DNA topoisomerase IIα.

Angiogenesis Pathway and Its Involvement in Cancer
Angiogenesis is the process of generating new capillary blood vessels (82). It plays important functions in organ development and differentiation during embryogenesis as well as wound healing and reproductive functions. However, angiogenesis is also responsible for a number of disorders including tumor formation, cancers, rheumatoid arthritis etc. Vascular Endothelial Growth Factor (VEGF) plays key role in angiogenesis process. VEGF protein has many isoforms and all of the isoforms mediate their effects by specific receptors known as VEGF receptors (VEGFRs). VEGFRs are receptor tyrosine kinases (RTKs) and there are three main isoforms: VEGFR-1, VEGFR-2, VEGFR-3. The expression of VEGF proteins are found to be dramatically increased in cancers like lung, thyroid, breast, ovary, kidney, uterine cancers etc. (83,84). Since VEGF mediates its effects by binding to specific receptors (like VEGFR-2), inhibiting the actions of the receptors is thought to be a therapeutic target for cancer treatment (85). When VEGF protein binds with VEGFR-2, the VEGFR-2 becomes activated which then activates phosphatidylinositol 3-kinase (PI3K). PI3K further activates phosphoinositide-3-kinase (PIP3), which in turn activates the Akt/PKB (protein kinase B) signaling pathway. This pathway contributes to endothelial cell survival by activating proteins, like BAD (Bcl-2 associated death promoter) and caspase proteins.
Moreover, the Akt/PKB signaling pathway can activate the endothelial nitric-oxide synthase (eNOS), which is responsible for vascular permeability. Both the endothelial cell survival and vascular permeability mechanisms contribute to the angiogenesis process. The binding of VEGF to VEGFR-2 can sometimes activate MAP kinase (mitogen activated protein kinase) pathway which is responsible for the proliferation of endothelial cells. In this pathway, activated VEGFR-2 activates phospholipase c-γ (PLC-γ). The PLC-γ then activates the protein kinase C (PKC). PKC further activates the proteins of MAP kinase pathway: RAF1, MEK, ERK, sequentially. This MAP kinase pathway causes the endothelial cell proliferation, which also contributes to the angiogenesis process ( Figure 03) (86,87,88). Since VEGFR-2 is involved in angiogenesis process in cancer development, inhibition of VEGFR-2 is considered as therapeutic approach to treat cancer.
Three approved drugs were used as positive controls in this study: alvocidib (inhibits CDK-2), daunorubicin (inhibits human topoisomerase Iiα) and lenvatinib (inhibits VEGFR-2) (32,89,90). Figure 03. The angiogenesis pathway. The VEGF protein binds with VEGFR-2 and activates the receptor. The VEGFR-2 activates PI3K, which activates PIP3 and thus activating the Akt/PKB signaling pathway. This pathway contributes to endothelial cell survival by activating BAD and caspase proteins. Moreover, the Akt/PKB signaling pathway can activate eNOS, which is responsible for vascular permeability. Both the endothelial cell survival and vascular permeability mechanisms contribute to the angiogenesis process. Binding of VEGF to VEGFR-2 can sometimes activate MAP kinase pathway which is responsible for the proliferation of endothelial cells. The activated VEGFR-2 activates PLC-γ. The PLC-γ further activates PKC. PKC further activates RAF1, MEK, ERK, sequentially. This MAP kinase pathway causes the endothelial cell proliferation, which also contributes to the angiogenesis process. VEGFR-2 inhibitors inhibits VEGFR-2, thus aid in cancer treatment.

Materials and methods
10 ligands (total) for each of the target molecule i.e., CDK-2, human topoisomerase IIa and VEGFR-2, were selected from literature that have already been proven to have inhibitory effects on the respective target molecule. Their IC50 values were collected by reviewing literatures discussing their anticancer potentiality. On sequential docking experiment one best ligand molecule was selected as the best inhibitor of respective target. Then their different drug like parameters were analysed in different experiments.

Protein Preparation and Ramachandran plot generation
Three dimensional structures of Cyclin-dependent kinase-2 (3EZV), Human topoisomerase II (1ZXM) and Vascular Endothelial Growth Factor Receptor-2 (2OH4) were downloaded (sequentially) in PDB format from protein data bank (www.rcsb.org). The proteins were then prepared and refined using the Protein Preparation Wizard in Maestro Schrödinger Suite 2018-4 (91). Bond orders were assigned and hydrogen molecules were added to heavy atoms as well as all the waters were deleted and the side chains were adjusted using Prime (92). Finally, the structure was optimized and then minimized using force field OPLS_2005. Minimization was done setting the maximum heavy atom RMSD (root-mean-square-deviation) to 30 Å and any remaining water less than 3 H-bonds to non-water was again deleted during the minimization step. After successful minimization, the proteins were used to generate Ramachandran plots for each of the protein by Maestro Schrödinger Suite 2018-4, keeping all the parameters as default.

Ligand Preparation
Three dimensional structures of 30 selected ligand molecules as well as controls were downloaded (sequentially) from PubChem database (www.pubchem.ncbi.nlm.nih.gov). These structures were then prepared using the LigPrep function of Maestro Schrödinger Suite (93). Minimized 3D structures of ligands were generated using Epik2.2 and within pH 7.0 +/-2.0 (94). Minimization was again carried out using OPLS_2005 force field which generated 32 possible stereoisomers.

Receptor Grid Generation
Grid usually confines the active site to shortened specific area of the receptor protein for the ligand to dock specifically. In Glide, a grid was generated using default Van der Waals radius scaling factor 1.0 and charge cutoff 0.25 which was then subjected to OPLS_2005 force field. A cubic box was generated around the active site (reference ligand active site). Then the grid box volume was adjusted to 15×15×15 for docking test.

Glide Standard Precision (SP) Ligand Docking, Prime MM-GBSA Calculation and
Induced Fit Docking for total free energy (ΔGbind) calculation. The total free energy of binding was calculated by the following equation: (96).
At this stage the docking parameters of our compounds under investigation was compared with 3 controls name with respective receptors.
To carry out the IFD of the nine selected ligand molecules, again OPLS_2005 force field was applied after generating grid around the co-crystallized ligand of the receptor and this time the best five ligands were docked rigidly. Receptor and Ligand Van Der Waals screening was set at 0.70 and 0.50 respectively, residues within 2 Å were refined to generate 2 best possible posses with extra precision. Best performing ligand was from each enzyme category was selected according to the IFD score and XPGscore. The 3D representations of the best pose interactions between the ligands and their respective receptors were obtained using Discovery Studio Visualizer (97).

Ligand Based Drug Likeness Property and ADME/Toxicity Prediction
The drug likeness properties of the three selected ligand molecules were analyzed using SWISSADME server (http://www.swissadme.ch/) (98

PASS (Prediction of Activity Spectra for Substances) and P450 Site of Metabolism (SOM) prediction
The PASS (Prediction of Activity Spectra for Substances) prediction of the three best selected ligands were conducted by using PASS-Way2Drug server (http://www.pharmaexpert.ru/passonline/) by using canonical SMILES from PubChem server (https://pubchem.ncbi.nlm.nih.gov/) (101). To carry out PASS prediction, Pa (probability "to be active") was kept greater than 70%, since the Pa > 70% threshold gives highly reliable prediction (102). In the PASS prediction study, both the possible biological activities and the possible adverse effects of the selected ligands were predicted. The P450 Site of Metabolism (SOM) of the three best selected ligand molecules were determined by online tool, RS-WebPredictor 1.0 (http://reccr.chem.rpi.edu/Software/RS-WebPredictor/) (103). The LD50 and Toxicity class was predicted using ProTox-II server (http://tox.charite.de/protox_II/) (104).

DFT Calculations
Minimized ligand structures obtained from LigPrep were used for DFT calculation using the Jaguar panel of Maestro Schrödinger Suite using Becke's three-parameter exchange potential and Lee-Yang-Parr correlation functional (B3LYP) theory with 6-31G* basis set (105,106,107).

Ramachandran Plot and Molecular Docking Analysis
After preparing the proteins, the Ramachandran plot for each of the receptor proteins was generated. In the plot, the orange regions represent "favored" regions, the yellow regions represent "allowed" regions and the white regions represent "disallowed" regions (110). CDK-2 protein generated Ramachandran plot with almost all of the amino acids in the "favored" region and no amino acids in the "disallowed region". Human topoisomerase II generated Ramachandran plot with 15 amino acids in the "disallowed region". It also had majority of the amino acids in the "favored" region. VEGFR-2 generated Ramachandran plot with only 4 amino acids in the "disallowed" region and most of the amino acids in the "favored" region ( Figure 04). All the selected ligand molecules were docked successfully with their respective receptor proteins.
The ligand molecules that had the lowest binding energy were considered the best ligand molecules in inhibiting their respective receptors since lower binding energy (docking score) corresponds to higher binding affinity (111). In the MM-GBSA study, the most negative ΔGBind score (the lowest score) is considered as the best ΔGBind score (112). IFD study is carried out to understand the accurate binding mode and to ensure the accuracy of active site geometry. The lowest values of IFD score and XP GScore are considered as the best values (113,114,115,116). Nine ligands: Geraniol, Epigallocatechin gallate and Indirubin (inhibit CDK-2), Daidzein, Camptothecin and Salvicine (inhibit human topoisomerase II) and Quercetin, Decursinol and Plumbagin (inhibit VEGFR-2), were initially selected based on the lower free binding energy and MM-GBSA study since they were reported to show comparable binding energy with respective controls (Table 02).
Then these molecules were subjected to IFD study. Epigallocatechin gallate, Daidzein and Quercetin were considered as the three best ligand molecules from the IFD study among the nine initially selected ligands. The 3D representations as well as interaction of different amino acids with Epigallocatechin gallate, Daidzein and Quercetin are illustrated in Figure 05. Now, these 3 best ligands (one from each of the receptor category) were used to in next phases of this experiment to analyze their drug potentials. Valine 914 and many other amino acid residues within the binding pocket of VEGFR-2 (Table   03).

ADME/T Tests
The results of ADME/T test with probability scores are summarized in

PASS and P450 Site of Metabolism Prediction
The predicted LD50 value for epigallocatechin gallate, daidzein and quercetin were 1000 mg/kg, 2430 mg/kg and 159 mg/kg, respectively. The prediction of activity spectra for substances (PASS prediction) was for the three selected ligands to identify 20 intended biological activities and 5 adverse & toxic effects. The PASS prediction results of all the three selected ligands are listed in

Analysis of Frontier's Orbitals
In the analysis of Frontier's orbitals, the DFT calculations and HOMO-LUMO studies were conducted. The results of the DFT calculations are listed in Table 09. In these studies, Epigallocatechin gallate showed the lowest gap energy of 0.070 eV as well as the lowest dipole moment of 1.840 debye. On the other hand, quercetin generated the highest gap energy of 0.167 eV and the highest dipole moment of 5.289 debye. The order of gap energies and dipole moments of these three compounds were, epigallocatechin gallate < daidzein < quercetin (Figure 06).  Quercetin.

Discussion
Molecular docking is an effective strategy in computer aided drug designing which works on specific algorithm and assigns affinity score depending on the poses od ligand inside the binding pocket of a target. Lowest docking score reflects highest affinity meaning that the complex remains more time in contact (118,119).
In this study a total of 30 ligands targeting three macromolecules involved in cancer development were screened with the aid of molecular docking which generated comparable docking score as with positive controls (Table 02). At the initial step their quality was exemplified with the help of Ramachandran plot where they were predicted to perform well. Primarily, three ligands were selected for each receptor which were then subjected to IFD. Finally, Epigallocatechin gallate, Daidzein and Quercetin were selected as the best inhibitors of CDK-2, Human topoisomerase Iiα and VEGFR-2, respectively. Hydrogen and hydrophobic interactions are important for strengthening the receptor-ligand interactions (120). Selected best three ligands along with total ligands were predicted to form multiple hydrogen and hydrophobic interactions with the target molecules (Table 02 and Table 03).
Estimation of the drug likeness properties facilitates the drug discovery and development process.
Drug permeability through the biological barrier is influenced by the molecular weight and TPSA. whereas, Epigallocatechin gallate was reported to violate the rule which might subject it to further modification ( Table 04).
The main purpose of conducting ADME/T tests is to determine the pharmacological and pharmacodynamic properties of a candidate drug molecule within a biological system. Therefore, it is a crucial determinant of the success of a drug discovery expenditure. BBB is the most important factor for those drugs that target primarily the brain cells. P-glycoprotein in the cell membrane aids in transporting many drugs, therefore, its inhibition affects the drug transport. The permeability of Caco-2 cell line indicates that the drug is easily absorbed in the intestine. Orally absorbed drugs travel through the blood circulation and deposit back to liver and are degraded by group of enzymes of Cytochrome P450 family and excreted as bile or urine. Therefore, inhibition of any of these enzymes of these enzymes affects the biodegradation of the drug molecule (123,124). Moreover, if a compound is found to be substrate for one or more CYP450 enzyme or enzymes, then that compound is metabolized by the respective CYP450 enzyme or enzymes (125).
A drug's proficiency and pharmacodynamics are depended on the degree of its binding with the plasma protein. A drug can cross the cell layers or diffuse easily if it binds to the plasma proteins less efficiently and vice versa (126). Human intestinal absorption (HIA) is a crucial process for the orally administered drugs (127,128,129). Moreover, the half-life of a drug describes that the greater the half-life, the longer it would stay in the body and the greater its potentiality (130,131,132). HERG is a K + channel found in the heart muscle and blocking the hERG signaling can lead to the cardiac arrhythmia (133,134). Human hepatotoxicity (H-HT) involves any type of injury to the liver that may lead to organ failure and death (135,136). Ames test is a mutagenicity assay that is used to detect the potential mutagenic chemicals (137). Drug induced liver injury (DILI) is the injury to the liver that are caused by administration of drugs. DILI is one of the causes that causes the acute liver failure (138). The results of ADME/T test are listed in Table 05. All of the three ligands were predicted to perform similar and sound in the ADME/T test.
Prediction of Activity Spectra for Substances or PASS prediction is a process that is used to estimate the possible profile of biological activities associated with drug-like molecules. Two parameters are used for the PASS prediction: Pa and Pi. The Pa is the probability of a compound "to be active" and Pi is the probability of a compound "to be inactive" and their values can range from zero to one (101). If the value of Pa is greater than 0.7, then the probability of exhibiting the activity of a substance in an experiment is higher (139). PASS was predicted for Epigallocatechin gallate, Daidzein and Quercetin. Both Epigallocatechin gallate and Quercetin showed similar performances with 12 biological activities and 2 toxic effects (Table 06 and Table 07).
ProTox-II server estimates the toxicity of a chemical compound and classifies the compound into a toxicity class ranging from 1 to 6. The server classifies the compound according to the Globally  (107,143,144,145,146,148). All of the ligands were reported to have significant energy gap indicating their possibility to undergo a chemical reaction (Table 09 and Figure 06).
Finally, all the best performed ligands were analyzed in different post-screening study and they're predicted to perform sound. Overall, this study recommends Epigallocatechin gallate, Daidzein and Quercetin as the best inhibitors of CDK-2, Human topoisomerase Iiα and VEGFR-2, respectively among all selected ligands which could be potential natural plant-derived compounds to treat cancer. However, other compounds could also be investigated as they were also showed convincing docking scores (Table 02). Further in vivo and in vitro experiments might be required to strengthen the findings of this study.

Conclusion
In the experiment, 30 anti-cancer agents were selected to analyse against three enzymes, CDK-2, human topoisomerase IIa and VEGFR-2, of three different pathways that lead to cancer development. 10 ligands were studied for each of the enzyme group using different approaches used in computer-aided drug designing. Upon continuous computational experimentation, Epigallocatechin gallate, Daidzein and Quercetin were predicted to the best inhibitors of CDK-2, Human topoisomerase IIα and VEGFR-2 respectively. Then their drug potentiality was checked in different post-screening studies where they were also predicted to perform quite similar and sound. However, the authors suggest more in vivo and in vitro researches to be performed on these agents as well as the other remaining agents to finally confirm their potentiality, safety and efficacy.

Acknowledgements
Authors are thankful to Swift Integrity Computational Lab, Dhaka, Bangladesh, a virtual platform of young researchers, for providing the tools.