Designing of Disulfide Cyclic Peptide for Inhibiting Polymerase A and B1 (PAC-PB1N) in H1N1 Virus using Molecular Simulation Approach

Email: usman@ui.ac.id Abstract: The drug resistance A/H1N1 flu virus is emerging rapidly. Therefore, looking for potential therapy is very important. PB2, PB and PA are subunits of viral RNA-dependent RNA polymerase (RdRp). They play an important role in viral replication. The PA and PB1 binding sites can be considered as potential targets for the development of new influenza drugs. The peptide inhibitors can be designed specifically due to their highpreferred activity. In this study, the cyclic peptide ligands were designed based on the crystal structure of PAC-PB1N in the surface of the molecule, resulting 1728 cyclopentadienyl compounds. The MOE 2008.10 software was utilized for molecular docking and dynamics simulation approach, while Lipinski’s Rules of Five were utilized to evaluate the feasibility of drug candidates. Thus, molecular dynamics simulation was applied, in order to facilitate the interaction between the ligand and enzyme. The simulations have successfully produced two cyclopentyl peptides, namely CKKTC and CKTTC, which results in both ligands providing a potent inhibitor of polymerase PAC-PB1N of Influenza A/2009 (H1N1).


Introduction
Since March 2009, a new strain of Influenza A/H1N1 flu has spread rapidly and become a pandemic. The H1N1 virus infection has been officially reported in many countries (WHO, 2014). This could be a big burden to the health care system in the world due to the high mortality and morbidity associated with H1N1 pandemics (Donaldson et al., 2009). The antigenic seasonal flu vaccine can give protection against the new strain A/2009 (H1N1) and render it ineffective. Until now, there are two classes of antiinfluenza drugs that exist in the market; neuraminidase inhibitors (e.g., zanamivir and oseltamivir) and M2 inhibitors (e.g., amantadine, rimantadine and adamantine). Neuraminidase (NA) inhibitors prevent offspring distribution and release of viral particles (Ushirogawa and Ohuchi, 2011). The Centers for Disease Control and Prevention (CDC) and the World Health Organization (WHO) have recommended oseltamivir as the most effective anti-influenza therapy (Rungrotmongkol et al., 2009). However, the rapid mutation of influenza virus showed several strains of drug resistance (Regoes and Bonhoeffer, 2006). Therefore, identifying new potential drug candidates are necessary. Hence, pharmaceutical biotechnology has determined the core components of the polymerase complex, namely Polymerase B1 (PB1), Polymerase B2 (PB2) and Polymerase A (PA) (Toyoda et al., 1996). Cyclopentyl peptides were chosen due to their durability in utilizing vesicles delivery system (Pfaff et al., 1994). The hydrophobic pockets of PA C -PB1 N crystal structure were located on the molecular surface. Hydrophobic pockets would facilitate an array of hydrogen bonds in the PA C -PB1 N (Moen et al., 2014). Hydrophobic interactions could be a significant contribution to the binding energy (Obayashi et al., 2008). The PB1 Nterminal (PB1 N ) of 48 amino acids is known as the αbinding domain. It is required for the PA that contains residues (Perez and Donis, 2001).
In this study, our approach is was based on bioinformatics computation (Fischer and Serapião, 2010). Bioinformatics tools have been widely applied for drug development and molecular medicine (Bartoli, 2011). Moreover we constructed the cyclopentyl peptide ligands from two cysteine and three amino acids which can be found in the PB1 sequences, such as asparagine, glutamine, methionine, threonine, aspartic acid, alanine, valine, leucine, glycine, lysine, proline and phenylalanine. Thus, we obtained 1728 cyclopentyl peptide ligands by this approach. Those ligands would be piped into the Absorption, Distribution, Metabolism, Excretion and Toxicology (ADMET) screening. Further processing of the screening approach was based on the molecular docking and dynamics methods (Sudi et al., 2012). The purpose of this study is to inhibit the activity of PA C -PB1 N domain. We expected that the lead compounds from this study could be developed further as drug candidates.

Sequence Retrieval
The Database resources of the National Center for Biotechnology Information (NCBI) were employed to retrieve the H1N1 virus polymerase B1 (PB1) (http://www.ncbi.nlm.nih.gov/) (Parikesit, 2009). About 1644 sequences were downloaded. The virus was isolated from human and the data archive was not later than December 2009. ClustalW program was used to align the PB1 sequences (http://www.ebi.ac.uk/Tools/clustalw2/index.html) (Thompson et al., 2002). According to the results of ClustalW2 computation, all sequences (716 amino acids) have about 99% similarity.

Protein Structure
Templates were used to identify PB1 by using Swiss model (http://swissmodel.expasy.org) (Biasini et al., 2014). Polymerase PA C -PB1 N complex was found with Protein Data Bank (PDB) entry code of 3CM8. The identification results score based on the template is 844. Thus, PA C -PB1 N 3D structure was extracted from Research Collaboration Structural Bioinformatics-Protein Data Bank (RCSB-PDB) database at http://www.rcsb.org/pdb/static.do?p=search/index.html (Rose et al., 2013). The 3D structure was saved in .pdb format.

Construction of the Ligand Structure
In this study, 1728 cyclopentadienyl peptide inhibitors were developed. At present, the ligand is composed of five disulfide bridges in the amino acid residues (cyclo pentapeptide) (Tambunan et al., 2014). The ACD/Phys Chem Suite 12.0 was used for drawing the ligands (Spessard, 1998).

Preparation of the Ligand
The ligands were converted into .mol format using VegaZZ software and then opened on database viewer of Molecular Operating Environment (MOE, version 2008.10). The ligands went through several steps of preparation, including "Wash", "Partial Charge" and "Energy Minimize" step. The selected force field in these process was Merck Molecular Force Field 1994 (MMFF94).

Protein-Ligand Docking
The docking simulation was conducted using MOE 2008.10 software with the default settings (Vilar et al., 2008). Docking simulations determine the optimal ligand-protein interaction (Shoichet et al., 2002). Thus, our results would be screened by Lipinski's rules using SCFBio-IT (Shaikh et al., 2007). The protein structure of 3CM8 was used for the preparation of the docking. The missing hydrogen atom was added. The protein-ligand complex was obtained for docking constitution. In order to estimate the free energy of binding of the ligands, the score composed is selected as molecular docking score function (London dG) (Ben et al., 2013).

Molecular Dynamics
Molecular dynamics simulations were conducted with a fixed Number of particles, Volume and Temperature (NVT) and born solvation stages. Thus, Nose'-Poincaré-Andersen (NPA) method was utilized (Sturgeon and Laird, 2000). It leads to the stability and durability of the simulation of 300 K and 312 K. The heating stage requires 20 picoseconds (ps), while to observe the enzyme-ligand enzyme inhibitor interaction stability needs 5 nanoseconds (ns) simulation calculations for all ligands (Durrant and McCammon, 2011).

Sequence Analysis
In molecular systems, the crystal structure PA-PB1 with 2,3 and 2,9Ǻ resolutions are 2ZNL and 3CM8, respectively. The existence of PA-PB1 3D structure provides a starting point for further investigation and dissemination of the structure of viral polymerase. In this study, the polymerase sequences were aligned to obtain the conserved regions.

Molecular Docking Simulation
In this study, the docking process was performed on 1728 cyclopentadienyl peptide inhibitors into PA-PB1 protein. There are three regions binding site between the Polymerase A and Polymerase B1, which is (1) Trp706 and Phe411, (2) Phe710 and Leu666 and (3) Leu640, Val636, Met595 and Trp619. Unlike the other two binding sites, The Trp706 and Phe411 binding sites have a great contribution to the bonding interactions of polymerase PA-PB1 (Liu and Yao, 2010). Therefore, the docking simulations were carried out at this binding site. Based on docking simulation results on the Trp706 and Phe411 binding sites, we obtained two ligands, namely CKTTC and CKKTC, as the best ligands which have the lowest ∆G binding value compared to other ligands. The ligand CKTTC itself has a ∆G binding value of -20.9946 kcal/mol and ligand CKKTC has ∆G binding value of -18.0293 kcal/mol. The docking results showed that both ligands have better ∆G binding value than the standard ligand PTLLFL and ribavirin, both of which each have ∆G binding value of -14.4066 kcal/mol and -9.7573 kcal/mol, respectively.

Molecular Dynamics Simulation
The Conformation of Ligand CKKTC with the Enzyme The conformational changes in the molecular dynamics are shown in Fig. 1. It illustrates Root Mean Square Deviation (RMSD) as a function of elapsed time of MD simulations for the main chain alpha carbon atom. At 300 K and 312 K, RMSD value at 0 ps is the same with the value of RMSD in complex enzyme without ligand. Increased RMSD values also occur in molecular dynamics simulation. Ligand CKTTC and CKKTC are more stable than ligand PTLLFL. Upon disulfide formation, the cross-linked vesicles remained stable under conditions that disrupted the noncrosslinked peptide ones.

Enzyme-Ligand PTLLFL and Ribavirin Three-Dimensional Structure
The RMSD as functions of elapsed time of MD simulations for the main chain alpha carbon atom can be seen in Fig. 2. In molecular dynamics simulations at 300 K, RMSD at 0 ps has the same value with the RMSD in enzyme complex without ligand. The increase in RMSD values also occurs at this temperature. The increase occurred as a result of increasing RMSD values from the influence of solvent molecules.

The Bonding between Enzyme and Ligand
PA C interacts with PB1 N through an array of hydrogen bonds and hydrophobic contact. Most intersubunit hydrogen bonds were formed through main-chain atoms of PB1 N . Residues Asp 2 to Asn 4 formed antiparallel βsheet-like interaction with Ile621 to Glu623 of PA C . Based on the consistent information of interaction pair analysis and virtual alanine scanning, several residues of PA C , such as Trp706, Phe710, Asn412, Gln408, Phe411, Leu666, Trp619, Glu623, Gln670, Leu667, Met595, Thr639 and Leu640, were found to contribute largely to the binding process. If we compare to all of the ligands, based on the ∆G binding value, CKTTC and CKKTC ligands are the best. Hopefully, these ligands can block the interaction between PA C and PB1 N . We presented the three-dimensional ligand-receptor interaction diagrams ( Fig. 3 and 4). In Fig.  3, CKTTC ligand interacted with several residues of PA C , such as Glu623, Glu630, Lys635, Ala448, Glu449, His452, Ile621, Gly631 and Ser632. CKKTC ligand interacted with the residue of PA C Glu410, Glu449, His452, Glu410, Phe446, His452, Cys453 and Val636. In Fig. 4, CKTTC ligand interacted with the residue of PA C such as Asp272, Gly273, Gly622, Glu623, Lys716, Pro271, Pro620, Ile621 and Phe710. CKKTC ligand interacted with Pro267, Leu270, Asp272, Gly273, Glu327, Glu691, Arg266, Leu268, Lys539, Asp682, Tyr687, Trp706 and Phe710. This depiction is relatively strong connections or hydrogen bonds as well as the electrostatic or charge-transfer interaction between a ligand and the residue protein.

Ligand's Pharmacological Properties
The pharmacological properties of the ligands were computed by using the Lipinski's rule of five (RO5) (Lipinski et al., 1997). The classification of these rules was derived from the physical and chemical properties of the drug compounds. In Table 1, we presented properties of our ligand compound. Lipinski's RO5 helps in distinguishing between a drug like and not a drug like molecules. It computes whether the ligand was breaching with at least 2 rules to determine its feasibility as a drug candidate. The further descriptor analysis helped in the identification of the better inhibitor (Table 2). Based on the RO5 and descriptor analysis, the whole ligands could become candidate drugs for blocking of interaction between polymerase A and B1.

Discussion
On one hand, the current biomedical research has greatly utilized cyclic peptides. The bioengineering methods have been applied to generate robust cyclic peptides (Thapa et al., 2014). Thus, structure-based docking method was useful to generate cyclic peptide inhibitor to cope with dengue virus (Idrees and Ashfaq, 2014). Assessment of small cyclic peptides was conducted for generating a library of VEGF inhibitors (Feng et al., 2014). Potential application in generating peptidomimetic compounds was in place as well (Razavi et al., 2014). On the other hand, the new generation of cyclic peptides databases was already in place, such as antiparasitic and antimicrobial peptides libraries (Gogoladze et al., 2014). These strong developments in cyclic peptide design were in line with the interest of our group to work in this field. These are shown in the previous works within the subjects of H1N1 virus (Tambunan et al., 2012). The resulted best cyclic peptides, namely the threoninedominated CKTTC and lysine-dominated CKKTC could be affected by the posttranslational modification, if no cyclization is applied (Tsiboli et al., 1997). Threonine can undergo glycosylation while lysine can be hampered with methylation (Zhang et al., 2012). Thus, it is already proven that cyclization could prevent the modification of peptides by digestive or biotransforming enzymes (Oliyai, 1996). Hence, the next feasible step is to evaluate the synthetic accessibility of these lead compounds (Ertl and Schuffenhauer, 2009). The synthesized works, bioassay and drug delivery construction will also follow suit (Jois et al., 1996).

Conclusion
There are 1728 cyclo-pentapeptide inhibitors already designed in this study. The docking of these inhibitors to determine their interaction activities with polymerase A and B1 was carried out. The small value of free energy binding and estimated Ki values for all the inhibitors indicate that the molecules bound considerably well with the binding site. We also analyzed our ligand properties based on Lipinski's rules. To ensure the absorption, distribution, metabolism, excretion and toxicity, ADMET calculations on the ligands are recommended. Thus, from this study, cyclo-pentapeptide CKTTC and CKKTC could be proposed as potential inhibitors of the interaction of PA C -PB1 N in H1N1 virus.