Selectivity of Guanine Scaffold in Anticancer Drug Development: A Computational Repurposing Approach

Drug repurposing is one of the modern techniques used in the drug discovery to nd out the new targets for existing drugs. Insilico methods have a major role in this approach. We used 60 FDA approved antiviral drugs reported in the last 50 years to screen against different cancer cell receptors. The thirteen compounds selected after virtual screening are analyzed for their druggability based on ADMET parameters and found the selectivity of guanine derivatives- didanosine, entecavir, acyclovir, valganciclovir, penciclovir, ganciclovir and valacyclovir as suitable candidates. The pharmacophore model, AARR, suggested based on the common feature alignment, shows that the two fused rings as in guanine and two acceptors-one from keto-oxygen(A5) and other from the substituent attached to nitrogen of imidazole ring(A4) give the druggability to the guanine derivatives. The NBO analysis on N9 is indicative of charge distribution from the ring to substituents, which results in delocalization of negative character in most of the ligands. The molecular dynamics simulations also pointed out the importance of guanine scaffold, which stabilizes the ligands inside the binding pocket of the receptor. All these results are indicative of the selectivity of guanine scaffold in anticancer drug development, especially as PARP1 inhibitors in breast, ovarian and prostate cancer. As these seven molecules are already approved by FDA, we can safely go for further preclinical trials. on the top of them. Entecavir shows better binding anity with 1M17(-7.6kcal/mol), 3CS9(-7.1kcal/mol), 5LNZ(-7.8kcal/mol), 3NUP(-7.9kcal/mol), 4U80(-7.9kcal/mol) and 3OG7(-8.6kcal/mol). Didanosine and ganciclovir shows better binding with 1M17(-7.1 & -8.0kcal/mol), 3CS9(-6.9 & -6.9 kcal/mol), 5LNZ(-6.8 & -7.0 kcal/mol), 3NUP(-7.6 & -6.5 kcal/mol), 4U80(-6.9 &-7.3kcal/mol) and 3L3M(-7.5 & -108kcal mol) respectively while acyclovir shows maximum anity with 1M17(-6.2kcal/mol), 3CS9(-7.1kcal/mol), 5LNZ(-6.0kcal/mol), 5VAM(-7.7kcal/mol) and 3L3M(-10.3kcal/mol). Penciclovir and valganciclovir shows healthier anity with 3CS9(-8.5 5LNZ(-7.4 kcal/mol), 3NUP -7.7 kcal/mol), 4U80(-7.5 3L3M(-10.3 and 3OG7(-7.7 -8.2kcal/mol) Bank (PDB). The D-score analysis shows that 13 compounds- entecavir, didanosine, saquinavir, ritonavir, atazanavir, asunaprevir, paritaprevir, acyclovir, ganciclovir, valacyclovir, penciclovir, valganciclovir and laninamivir octanoate have better binding anity with more than 5 receptors. The stability of these complexes were conrmed by MD simulations, which focuses the importance of guanine moiety in all the screened cases. The QikProp analysis of these 13 compounds indicate that, only the 7 guanine derivatives-didanosine, entecavir, acyclovir, valganciclovir, penciclovir, ganciclovir and valacyclovir are best drugs based on zero violation from Ro5 and zero #stars. The results suggest that compounds can be developed as chemotherapeutic agents, specically as PARP1 inhibitors in breast, ovarian and prostate cancer. The pharmacophore model suggested based on the common feature alignment, AARR shows that the two fused rings as in guanine and two acceptors-one from keto-oxygen(A5) and other from the substituent attached to nitrogen of imidazole ring(A4) give the druggability to the guanine derivatives. The NBO analysis on N9 is indicative of charge distribution from the ring to substituents, which results in delocalization of negative character in most of the ligands. As these molecules are already approved by FDA as antiviral drugs, we can directly go for preclinical and clinical studies of them against different cancer cell proliferations.


Introduction
Cancer, one of the foremost causes of mortality and morbidity worldwide, is a collective form of diseases with loss of control on growth and division of cells. 1,2 The existing treatments for this condition involves surgical removal of the growth with radiation and chemotherapies. 3 The key challenges of chemotherapy are the recurrence of the disease and severe side effects, which spoil the quality of life of the patient. In spite of its demerits, chemotherapy is still one of the broadly used method in treating all classes and stages of cancer progression. 1 Even though there are many heterocyclic compounds commercially available as anticancer agents, it still faces many challenges such as lack of speci c targeting. 4 From drug discovery through FDA approval, developing a new medicine takes at least 10 years on average and costs an average of $2.6 billion. 5 Repurposing existing drugs that may have unanticipated effects as potential candidates is one way to resolve this barrier. 6 Revaluating the prevailing drugs through drug repurposing holds the potential to counterpart traditional drug discovery by modifying the high economic and time related costs and risks as many compounds have demonstrated safety in humans it often negates the need for phase I clinical trials. 7 Also it has other advantages, such as implicit knowledge of its toxicity pro le, drug metabolism, pharmacokinetics, and drug interactions. 8 Repurposing approaches can be basically of two types-experimental screening approaches and in silico approaches. In silico methods apply sophisticated systematic approaches to existing data to identify new potential relations between drug and infection. 9 There are few recent reports on successful drug repurposing. Recently, Lorenzo et al. reported the repurposing effect of salicylanilide, anthelmintic drugs, in adenovirus infections. 10 Recognition of prostaglandin E2 as an activator of blood stem cell production and shows long-term safety in preclinical non-human primate transplant models. 11 White et al. reports the melanoma inhibiting capacity of le unomide, an oral anti-lymphocyte agent that has been approved by the Food and Drug Administration (FDA) since 1998 for treatment of rheumatoid arthritis. 12 One of the initial reports toward the computer-aided drug repositioning for DNMT1 inhibitors is the work of Méndez-Lucio et al. that identi ed olsalazine as a novel hypomethylating agent. 13 The identi cation of the antiviral drug ribavirin as inhibitor of histone methyltransferase zeste homolog 2 (EZH2) is another example of computer aided drug repurposing. 14 Shaimerdenova et al. reported the effect of antiviral treatment in breast cancer cell lines. They reveal that acyclovir inhibits colony formation ability, diminishes the proliferation rate of cells and cell invasion capacity of the cancer cells. 15 Molecular docking, dynamics and other computational tools are used to verify the drug and target interactions to envisage the potentiality of a drug or a ligand using mathematical calculations. 16 We applied inverse docking approach to identify the anticancer leads from antiviral drug database. As part of this, we selected 60 FDA approved small molecules over the past 50 years and screened them against 12 selected anticancer targets-EGFR, ABL, HSP90, CDK4/6, BRAF-wild and mutant, MEK1, BCL2, PARP, CMET and VEGF. [17][18][19][20][21][22][23][24][25][26][27][28][29][30][31][32][33][34] The best 7 compounds screened based on druggability pro le predicted by QikProp. The general features of the compounds were analyzed by pharmacophore modeling and electrostatic potential analysis. The results pointed out the selectivity of guanine moiety in anticancer drug development.

Methodology
The protein preparation, receptor grid generation, ligand conformation generation, ADMET screening, pharmacophore modeling, molecular docking and dynamics were done by Schrodinger suite (2018-2). 35 For this protein preparation wizard, LigPrep, QikProp, Phase-Pharm, Glide XP docking and Desmond tools were used in Maestro 11.2 interface in OPLS-2005 force eld. All the guanine derivatives were optimized and their HOMO-LUMO calculations were done by Density Functional Theory (DFT) using Gaussian 09 software packages and Chemcraft software was used for visualization purpose. 36 Poisson-Boltzmann electrostatic potential (ESP)was generated by Schrodinger. 37

Ligand preparation
The three-dimensional structures of 60 FDA approved antiviral drugs reported in the last 50 years were collected from PubChem. 39 The structures were imported to maestro workspace to generate the optimized geometry and conformers using Ligprep tool. The best conformers are selected for docking and further analysis.

QikProp analysis and ADMET prediction
The output of the Ligprep was used to analyze ADME/T(Absorption, Distribution, Metabolism, Excretion and Toxicity) properties using QikProp tool. The druggability of the compounds were predicted based on the parameters already de ned by the software.

Molecular Docking
The optimized conformers of the 60 ligands were docked against the grid generated using extra precision mode (XP). The exible docking of the ligands with the active binding sites were generated and the best poses corresponding to the interaction are de ned based on Glide and Dock score. These scores are used to arrange the ligands on the basis of interactions. Here we compare the D-score and − 6kcal/mol is used as the standard maximum for comparison of inhibiting power.

Molecular Dynamics
The best scored receptor-ligand complexes were selected for molecular dynamics (MD) simulations, which predict the suitability and stability of the binding mode by integrating Newton's equation of motion. The cubic box is de ned around the selected complexes, solvate them with TIP4P water molecules and counter ions were added to neutralize the system. The energy minimized and equilibrated systems were further used to perform MD simulations for a period of 100ns by using the Desmond module of the Schrödinger with an OPLS-2005 force eld by setting NPT ensemble at 300K and 1atm pressure. Root mean square deviation (RMSD) plots for the backbone atoms for both proteins and the ligands were analyzed to predict the stability of binding. The protein-ligand (P-L) interaction diagram and histogram help us to de ne the interacting residues and their selectivity.

Pharmacophore Modeling
A pharmacophore model of guanine derivatives based on atleast 50% match seven-point hypothesis was generated to de ne the pharmacophore features of the ligands in common. The seven compounds were aligned based on common featuresacceptor (A), donor(D), hydrophobic (H), negative ionic (N), positive ionic (P) and aromatic ring (R) and generated a model based on Phase-Hypo score.

Optimization and Electrostatic Potential
All the guanine derivatives were optimized using density fuctional theory (DFT) by applying Becke, 3-parameter, Lee-Yang-Parr (B3LYP) functional 6-311 + + G** basis set. The electrostatic potential (ESP) is generated by solving the Poisson-Boltzmann equations, with the partial charges of all the atoms in the structure in the workspace.

Molecular Docking and Binding Pose Analysis
The 60 selected FDA approved antiviral drugs reported on the last 50 years were docked against the 12 receptors existing in different cancer cell proliferations. The result analyzed on the basis of D-score ( Fig. 1 and Fig. 2) shows that 13 compoundsentecavir, didanosine, saquinavir, ritonavir, atazanavir, asunaprevir, paritaprevir, acyclovir, ganciclovir, valacyclovir, penciclovir, valganciclovir and laninamivir octanoate have better binding a nity with more than 5 receptors. Among them the 7 guanine derivatives-entecavir, didanosine, acyclovir, ganciclovir, valacyclovir, penciclovir and valganciclovir are on the top of them.

ADMET Screening
The QikProp analysis (

Molecular Dynamics
To  (Fig. 4), it is very clear that the proteins are stabilized under 3.5Å and the ligands are stable under 1.5 Å in all the six cases without notable uctuations. The protein-ligand (P-L) histogram (Fig. 5) and the interaction diagrams revealed that, in the case of a, strong H-bonded interaction between -NH from pyrimidine and Gly202 lasts for 97% of the simulation time while the other two H-bonded interactions between -CO of guanine and Ser243 lasts for 62% and -CH-NH 2 with Glu327 lasts for 69% of the simulation time. In addition, π-π stacking of imidazole and Tyr235 and Tyr246 lasts for 79%, also stabilizes the complex. When we analyze the data for the complex b, it is very interesting that H-bonding of -NH from pyrimidine ring with Gly202 (98%), -OH with Tyr235(40%) and π-π stacking of imidazole with Tyr246 (88%) are the major stabilizing factors. The complex c also forms H-bonded interaction of -NH from pyrimidine ring with Gly202 (100%) and π-π stacking interaction of imidazole with Tyr235 and Tyr246 (76%). In all these cases, Gly202, Tyr235 and Tyr246 are acting as the major points of interaction of ligands with 3L3M. The ligand entecavir inside 4U80 (d) also shows strong H-bonded interaction with Met146 (100%) through the pyrimidine ring and Ser194 due to -OH (52%). In addition, hydrophobic interaction with Glu144(86%) also stabilizes the complex. The H-bonded interactions of -NH from pyrimidine with Cys532(95%) and -OH with Asn580 (37%) and ππ stacking interaction of the pyrimidine ring with Trp531 (42%) stabilizes the ligand entecavir inside the binding pocket of 3OG7 (e). In the case of f also H-bonding formed between -CO and Asp594 (94%), -NH 2 and Glu501 (100%), and π-π stacking of imidazole with Phe595 (57%) hold the ligand valganciclovir inside the binding site of 5VAM. From these analyses, we can concluded that the guanine moiety, speci cally the -NH from pyrimidine and the -CO of guanine are act as the source for strong H-bonded interaction and the π-π stacking interaction.

Pharmacophore Modeling
The pharmacophore model (Fig. 6) generated based on the common feature alignment, AARR shows that the two fused rings as in guanine and two acceptors-one from keto-oxygen(A5) and other from the substituent attached to nitrogen of imidazole ring(A4) give the druggability to the guanine derivatives. In valganciclovir and valacyclovir, A4 is from keto-oxygen while in others such as entecavir, didanosine, acyclovir, ganciclovir and penciclovir, the oxygen from -CH 2 -O-H act as the acceptor. The distance between imidazolyl-N and A4 (oxygen) varies from 4.42 to 5.02 Å. The dihedral angle between the ring plane and the plane containing oxygen(A4) is 15.0 o

Electronic structure analysis (Optimization/MESP/HOMO-LUMO )
Guanine and its seven derivatives-acyclovir, didanosine, entecavir, ganciclovir, penciclovir, valacyclovir and valganciclovir were optimized using density functional theory (Fig. 7.; Supplementary Information, S1-S7) and their energies are tabulated below ( Table 2). Even though guanine itself is planar and aromatic, the substitution changes its planarity and this bent structure help these molecules to entrap into the binding pocket of the receptors. The exibility of the vanganciclovir by its linear and bent conformers makes it most suitable inhibitor in almost all cases. The ESP analysis shows that all these compounds with common guanine scaffold share precise electronic properties (Fig. 8). Poisson-Boltzmann tool help to generate isosurfaces and a mapping to the molecular surface by using partial charges of all the atoms in the input structure. The ESP pro les shows that guanine scaffold is highly negatively charged (red colour), while the substituent tail is somewhat positive (blue colour) in valacyclovir and valganciclovir. In total the negative charge is delocalized within these molecules and which helps in binding. The HOMO-LUMO energy gap (E LUMO -E HOMO ) varies from 4.81 to 5.58 eV, which displays the relative activity of these compounds. The natural charge on N9 calculated by natural bond orbital (NBO) analysis indicate that the charge from the ring is distributed to substituents, which results in delocalization of negative character in most of the ligands. The natural charge on N9 in guanine is -0.581 while on the derivatives, it ranges from − 0.432 to -0.452. Table 2 The optimized energy (B3LYP/6-311 + + G**, gas phase), natural charge on N 9 , E LUMO -E HOMO (eV) and PARP interacting residues of guanine and its derivatives.

Conclusions
We effectively used virtual screening for predicting the activity of antiviral drugs as inhibitors of selected cancer targets. The sixty FDA approved drugs from the last 50 years were selected and screened against twelve receptors-epidermal growth factor receptor, EGFR (1M17); human proto-oncogene tyrosine-protein kinase ABL1, ABL kinase (3CS9); heat shock protein HSP 90alpha, HSP90(5LNZ); cell division protein kinase 4, CDK4(3G33); cell division protein kinase 6 CDK6(3NUP); serine/threonineprotein kinase B-raf, BRAF wild(5VAM); mitogen-activated protein kinase kinase 1, MEK1(4U80); apoptosis regulator, BCL2(4AQ3); poly [ADP-ribose] polymerase 1, PARP(3L3M); B-Raf Kinase V600E oncogenic mutant, BRAF(3OG7); hepatocyte growth factor receptor, CMET(4MXC) and vascular endothelial growth factor, VEGF(1FLT) retrieved from RCSB Protein Data Bank (PDB). The D-score analysis shows that 13 compounds-entecavir, didanosine, saquinavir, ritonavir, atazanavir, asunaprevir, paritaprevir, acyclovir, ganciclovir, valacyclovir, penciclovir, valganciclovir and laninamivir octanoate have better binding a nity with more than 5 receptors. The stability of these complexes were con rmed by MD simulations, which focuses the importance of guanine moiety in all the screened cases. The QikProp analysis of these 13 compounds indicate that, only the 7 guanine derivatives-didanosine, entecavir, acyclovir, valganciclovir, penciclovir, ganciclovir and valacyclovir are best drugs based on zero violation from Ro5 and zero #stars. The results suggest that compounds can be developed as chemotherapeutic agents, speci cally as PARP1 inhibitors in breast, ovarian and prostate cancer. The pharmacophore model suggested based on the common feature alignment, AARR shows that the two fused rings as in guanine and two acceptors-one from ketooxygen(A5) and other from the substituent attached to nitrogen of imidazole ring(A4) give the druggability to the guanine derivatives. The NBO analysis on N9 is indicative of charge distribution from the ring to substituents, which results in delocalization of negative character in most of the ligands. As these molecules are already approved by FDA as antiviral drugs, we can directly go for preclinical and clinical studies of them against different cancer cell proliferations.
Declarations D-score analysis of sixty FDA approved drugs against selected twelve anticancer targets.   Optimized geometries