The Inhibition of Folylpolyglutamate Synthetase (folC) in the Prevention of Drug Resistance in Mycobacterium tuberculosis by Traditional Chinese Medicine

Tuberculosis (TB) is an infectious disease caused by many strains of mycobacteria, but commonly Mycobacterium tuberculosis. As a possible method of reducing the drug resistance of M. tuberculosis, this research investigates the inhibition of Folylpolyglutamate synthetase, a protein transcript from the resistance association gene folC. After molecular docking to screen the traditional Chinese medicine (TCM) database, the candidate TCM compounds, with Folylpolyglutamate synthetase, were selected by molecular dynamics. The 10,000 ps simulation in association with RMSD analysis and total energy and structural variation defined the protein-ligand interaction. The selected TCM compounds Saussureamine C, methyl 3-O-feruloylquinate, and Labiatic acid have been found to inhibit the activity of bacteria and viruses and to regulate immunity. We also suggest the possible pathway in protein for each ligand. Compared with the control, similar interactions and structural variations indicate that these compounds might have an effect on Folylpolyglutamate synthetase. Finally, we suggest Saussureamine C is the best candidate compound as the complex has a high score, maintains its structural composition, and has a larger variation value than the control, thus inhibiting the drug resistance ability of Mycobacterium tuberculosis.


Introduction
Mycobacterium tuberculosis (M. tuberculosis) is the principle causative agent in the development of tuberculosis (TB). TB typically attacks the lungs and can be spread from person to person through the air [1] when patients with an active TB infection cough, sneeze, spit, or otherwise transmit respiratory fluids. The mycobacteria may remain latent, causing the host to become weak and maybe develop anorexia. When M. tuberculosis becomes active the patients will develop a chronic cough with blood-tinged sputum, fever, night sweats, and weight loss. At this time the disease is infective. In 2012, there were estimated 8.6 million cases of TB worldwide and 1.3 million dead people from the infection (World Health Organization).
Due to the development of drug resistance by M. tuberculosis, coupled with the necessity for long-term treatment, it has become difficult to cure this disease by drugs [1,2]. A recent report in Nature Genetics indicates that the drug resistance genes gyrA, rpoB, rpoC, rpsL, katG, folC, thyA, embB, Rv3806c, and rrs are essential for M. tuberculosis. From the above genes, folC and Rv3806c remain uninvestigated [3].
Computer-aided drug design (CADD) is a popular in silico simulation technique due to its speed and low cost. The main investigations for CADD are structure-based and ligand-based. In this investigation we use molecular docking and molecular dynamics (MD), two aspects of structurebased drug design, to analyze protein structural variations during the complex interactions [4][5][6][7][8][9]. Personalized medicine and biomedicine have recently been attracting much attention [10], especially in areas such as the analysis of regional disease [11], clinical diagnoses, disease associated mutations [12]. And, as it is well known throughout the Asian region, traditional Chinese medicine (TCM) is the main personalized medicine resource.
Based on the above research, this study uses the CADD techniques of molecular docking and molecular dynamics to define the protein-ligand interactions and thus reports putative compounds for the inhibition of folC.

Disorder Protein Detection.
A disordered region of a protein plays an important role in drug design due to the character of the docking site structure affecting the suitability of the complex and the drug efficiency. The floC disorder region could be predicted from the database of protein disorder (DisProt, http://www.disprot.org/) [27], and comparisons between the docking site and the disorder region could help to define the drug effect on the protein [7,28].

Molecular
Docking. Accelrys Discovery Studio 2.5 (DS2.5) software was used to process the molecular docking produced in the CHARMm force field [29] by LigandFit, a receptor-rigid docking algorithm program [30]. The protein transcript from folC has shown that Folylpolyglutamate synthetase, dihydropteroate, and tetrahydrofolate could all dock with the protein. Based on the calculation of Ligplot [31,32], the complexes formed from the control with the protein product of folC and the top three TCM compounds with the protein product of folC contained hydrophobic interactions.

Molecular Dynamics Simulation.
After preparation based on the reference force field [33] of GROMACS 4.5.5 [34] by using SwissParam (http://www.swissparam.ch/) [35], the ligands were subjected to molecular dynamics simulation. The Folylpolyglutamate synthetase with ligands was placed into a simulation box with appropriate buffer, or other solutions, at a minimum distance of 1.2Å from the complex. The solution for simulation was based on the TIP3P water model in which sodium and chloride ions were added to neutralize  complex charges. The MD of GROMACS 4.5.5 had three steps: minimization, equilibration, and production. After minimization with the steepest descent method for 5,000 steps, the structures were transferred for MD simulation. The electrostatic interactions were based on the particle-mesh Ewald (PME) method [36] which calculates each time step at 2 fs and the numbers of steps were repeated 5,000,000 times. Under the 100 ps constant temperature (PER ensemble), the simulation was equilibrated by the Berendsen weak thermal coupling method.

BioMed Research International
Docking site After a MD simulation time of 10,000 ps, the protocols in Gromacs used the MD data to analyze the MD trajectories, RMSD, energy variations, and pathway analysis.

Results and Discussion
3.1. The Detection of Disorder Protein. The disordered protein is intrinsically an unstructured protein, and therefore the docking site will consist of a disordered region that will create challenges for drug docking, and the complex will stabilize only with difficultly. In recent references [7,28], the disordered protein cannot be established as a common domain; thus a drug docking to a disordered region might have lower side effects. On the other hand, a common domain for a similar structure will allow the drug to dock to the protein easily but may have an effect on other tissues and thus create side effects. This disorder for drug design is not a bad choice and should not be identified as difficult work. The important amino acids around the docking site of the synthetase protein are Asn75, Gly76, Lys77, Thr78, Ser79, His299, Asn303, Arg340, Ala354, Ala355, and His356 and are defined as a nondisordered region (Figure 1). From this result, and the understanding of disorder, the compounds and Folylpolyglutamate synthetase could combine as a stable complex.

Molecular
Docking. The top three TCM compounds based on the ranking of docking by Discovery Studio 2.5 were selected as candidate compounds for molecular dynamics investigation. These compounds and their botanical sources are listed in Table 1.
The structures of the control drug and the selected compounds Saussureamine C, methyl 3-O-feruloylquinate, and Labiatic acid are presented in Figure 2. The compound with the highest docking score, Saussureamine C, which is extracted from Saussurea lappa Clark, is also known as an antiulcer medication [37] and has been used to prevent breast cancer cell migration [38], represses inflammatory responses [39], has antihepatotoxic activity [40], and regulates immunity [41]. The compound ranked second, methyl 3-O-feruloylquinate, derived from Phellodendron amurense Rupr., has been assessed for antiviral treatment of H5N1 infections [42], the regulation of fatty acids [43], its role in the protection of human osteoarthritic cartilage [44], the treatment of Alzheimer's disease [45], as an antiinflammatory [46] and as an antimicrobial, activity against herpes simplex virus type 1 [47], and its effect on the human immune response [48,49]. The compound ranked third, Labiatic acid, which is derived from Rosmarinus officinalis L., has been shown to improve memory impairment [50], as well as having anti-inflammatory activity [51], being able to attenuate oxidative stress and reduce blood cholesterol [52], and having hypoglycemic and hepatoprotective activity [53]. The preceding references indicate these compounds could regulate immunity, be antimicrobial and antiviral and thus may be successful candidate compounds for the inhibition of the activity of bacteria and viruses, and may have the ability to modify drug resistance.
The docking poses ( Figure 3) and hydrophobic interactions ( Figure 4) could help with the identification of important amino acids. The results in Figure 3 show the docking poses and the amino acids around docking site that interact with the ligands. The amino acids Asn75, Gly76, Lys 77, Thr78, Ser79, Asn303, Arg340, and Asp353 have been defined in Uniprot as important binding sites. These amino acids are always present during interactions with ligands, not only in docking possess but also in hydrophobic interactions. This result confirms that the docking site is correctly defined as the functional domain of the protein.

Molecular Dynamics Simulation.
Variation in the complex RMSD, ligand RMSD, and total energy can help analyze the situation during MD simulation ( Figure 5). From Figure 5 it can be seen that the RMSD of the complex and ligand is around, or lower than, 0.2 nm. This result indicates that the protein, ligand, and their complex are stable and (1) (2) (3) that their position and structural variations are not too large. The total energy tends to the range between −254.5 and −255.5 10 3 kcal/mol. From these results we suggest this simulation will balance quickly according to the stability characteristics of the protein.
Clustering assists in grouping the data based on RMSD and thus defines similar structures as belonging to the same group ( Figure 6). These results demonstrate that there are some groups which are larger than the primary candidate. This implies that as the simulations tend to balance and thus the complexes have lower variation and similar structure, they become part of the same group. In our previous research we found there were commonly a lot of small groups under 5,000 ps. This interesting situation indicates this protein is stable enough during interactions.
The RMSF calculates the average RMSD focus of each amino acid in the complete MD simulation (Figure 7). In this result, we find the amino acid regions 24, 139-142, 203-208, and 455-460 have large variations. That the defined docking site is not in these regions means that the docking poses will not change significantly while the protein-ligand interactions are mobile. If the RMSF is similar, then the efficacy of compounds may be the same as the control.
Next we discuss the structural variations between protein-control interaction and protein-compound interaction (Figures 8-11). Figure 8(a) shows that Arg340 formed an Hbond with the ligand (distance <0.3 nm) at 200 ps. This suggests Arg340 may have function in protein-ligand interaction. From Figure 8(b) we can see that the variable region of the protein upper subunit will rotate counterclockwise, while the other subunit rotates clockwise. In the complete protein, there is only positional variation to transform the receptor site for ligand interaction.
In Figure 9(a), His299 produces H-bonds at an early time while Gly360 produces H-bonds at a later time. We suggest His299 might have an effect on the ligand target and Gly360 might have a protein function after the ligand interaction. Similar to the control, the primary candidate compound has the same variation in that the upper subunit rotates counterclockwise and the other subunit rotates clockwise. The variation value in the complex is larger than the control and thus our suggestion is that Saussureamine C might have a stronger effect on the protein.
In Figure 10(a), it is interesting that both Asn303 and Asp353 produce H-bonds during the MD simulation but one of the differences is that the H-bond of Asn303 does not change but the H-bond of Asn303 will exchange two atoms. We think the function of Asn303 may have an effect on the target ligand and that Asp353 may have an effect on the interaction. This complex also has a similar positional variation as the control, but in variation 1 the loop becomes a short helix that might make the protein different.
In Figure 11(a), the H-bond frequency is greater than in other compounds, indicating that Labiatic acid may have a higher activity in this protein. The variation of Arg340 in the MD simulation and Glu298 produce H-bonds from 1,500 ps, indicating that these two amino acids may have a great effect on the protein function. In Figure 10(b), besides the positional variation being similar, variation 1 is present as a short helix loss.
Pathway definition is based on the calculation of caver 3.0 to find out the path interprotein during MD [54]. These results indicate the different pathways deined form the ligand structure and protein variation caused by interaction ( Figures  12 to 14). In Figure 12, this result indicates the top 4 length pathways for dihydropteroate. But in these pathways, the third and fourth are in protein structure not in docking site. Actually, the ligand could not move through protein structure even the range of path could allow ligand pass; thus we suggest pathways 1 and 2 are the true pathways for dihydropteroate. In Figure 13, we also find out the top 4 length pathways in folC for Saussureamine C but we suggest only the first and the fourth are possible pathways. Finally, we can define the first and the third pathways as possible pathways for methyl 3-O-feruloylquinate ( Figure 14). In the pathway calculation, there is no pathway for Labiatic acid. We suggest the Labiatic acid makes protein variation; then the path is not larger or longer enough for ligand.

Conclusion
In the analysis of docking, this research indicates that the docking site and the ligand dock to protein are correct based on the amino acids interactions. The RMSD, energy, clustering, and RMSF show that Folylpolyglutamate synthetase is a stable protein according to low variation during interaction, with H-bonding providing appropriate assistance. We suggest Glu298, Asn303, Arg340, and Asp353 are important in the interaction based on the high frequency and stability during MD simulation. The structural variation shows that the conformation variation is focused on the protein character rather than the ligand affection. Finally, although the selected compounds are similar to the control in docking, hydrophobic interactions, and structural variations, we suggest that Saussureamine C is the best candidate for the complex as it has a high score, maintains its structural composition, and has a greater variation value than the control.