Study on Structure Activity Relationship of Natural Flavonoids against Thrombin by Molecular Docking Virtual Screening Combined with Activity Evaluation In Vitro

Thrombin, a key enzyme of the serine protease superfamily, plays an integral role in the blood coagulation cascade and thrombotic diseases. In view of this, it is worthwhile to establish a method to screen thrombin inhibitors (such as natural flavonoid-type inhibitors) as well as investigate their structure activity relationships. Virtual screening using molecular docking technique was used to screen 103 flavonoids. Out of this number, 42 target compounds were selected, and their inhibitory effects on thrombin assayed by chromogenic substrate method. The results indicated that the carbon-carbon double bond group at the C2, C3 sites and the carbonyl group at the C4 sites of flavones were essential for thrombin inhibition, whereas the methoxy and O-glycosyl groups reduced thrombin inhibition. Noteworthy, introduction of OH groups at different positions on flavonoids either decreased or increased anti-thrombin potential. Myricetin exhibited the highest inhibitory potential against thrombin with an IC50 value of 56 μM. Purposively, the established molecular docking virtual screening method is not limited to exploring flavonoid structure activity relationships to anti-thrombin activity but also usefully discovering other natural active constituents.


Introduction
In recent years, thrombotic disease has been a major threat to human health worldwide. As mentioned earlier, thrombin, a principal serine protease, plays a critical role in blood coagulation cascade response by activating blood platelet and catalyzing the conversion of soluble plasma-protein fibrinogen to insoluble fibrin [1,2]. This implies that thrombin metabolic disorder is a major triggering factor for thrombotic diseases. In clinical practice, there are some effective anti-thrombin inhibitors such as bivalirudin, argatroban, heparin among others. However, the cost of treatment with these drugs is expensive and also associated with serious side effects [2][3][4]. These reasons underscore the need to seek and discover safer, affordable and effective natural thrombin inhibitors for the treatment of thrombotic disease.
Structurally, 2-phenylchromanone and C6-C3-C6 form the backbone of flavonoids. Some well-known flavonoids include but are not limited to flavonols, isoflavones, flavones, flavanonols, anthocyanidins, chalcones, dihydrochalcone and aurones. [11,12]. These compounds elicit diverse pharmacological activity such as antioxidant, anti-inflammation, antibacterial, antithrombus and antineoplastic due to the difference in structural diversities in basic parent structure or substituent group [13]. Structure activity relationship (SAR) of flavonoids based on the above biological activities has gained lots of attention in recent years [14][15][16]. To the best of our knowledge, there is only one literature report on the SAR of flavonoids inhibiting thrombin [17]. Noticeably, it did not investigate anti-thrombin activity for all the substituent groups on flavonoid. In view of this, it was imperative to study the structural characteristics of more natural flavonoids for unraveling the mechanism of thrombin inhibition.
Molecular docking technology, an algorithm that predicts the putative geometry of a protein-ligand complex has been successfully applied in the studies of SAR of compounds and TCM as well as predicting binding affinities of ligands [18][19][20][21][22]. Irrespective of these gains, there also exist disadvantages such as low accuracy and high false positive rate. A chromogenic substrate assay (CSA), which is a rapid and very effective technique is another in vitro method for examining thrombin inhibition activity [4,19,23]. This assay determines the activity of enzymes in samples by measuring the absorbance of colored p-nitroaniline produced by the reaction of blood coagulation factor (such as FX, thrombin and prothrombin) and chromogenic substrate [24]. Clinically, CSA is widely used in diagnosing and monitoring of hemophilia A/B, investigating anticoagulant effects and screening enzyme inhibitors [5,25,26]. Therefore, the combination of the chromogenic substrate method and molecular docking could greatly improve the screening accuracy. In this study, the SAR of 103 flavonoids was studied using molecular docking combined with chromogenic substrate assay. The findings of this research aim to provide a basis for the study of mechanism of thrombin inhibition and also serve as a model for studying new flavonoid-type thrombin inhibitors.

The Feasibility of the Established Docking Protocol
To verify the feasibility of the established docking protocol, the original ligand molecule was extracted from the co-crystallized thrombin complex (2GDE and the original ligand SN3) and re-docked into 2GDE using CDOCKER. The conformation of the re-docked co-crystallized thrombin complex (2GDE and SN3) is shown in Figure 1. The value of low root mean-square deviation (RMSD) between re-docked conformation and the crystal conformation was 0.7457 Å. In general, if RMSD ≤ 2.0 Å, the molecular docking model should be considered as feasible [27].

In Silico Screening for Thrombin Inhibitors
One hundred and three flavonoids compounds were docked into the active site of thrombin model and identified by CDOCKER module in discovery studio 4.5 software (Accelrys, Biovia, Boston, MA, USA). The results showed 62 hit compounds with -CDOCKER_energy above 20.0 kcal/mol and 54 hit compounds with -CDOCKER_INTERACTION energy above 40.0 kcal/mol (see Table S1 in Supporting Material). The docking effect was good since the -CDOCKER_energy was higher than 15 kcal/mol [24].
Based on the above results, the anti-thrombin activities of 42 target flavonoids were selected for further research using a chromogenic substrate in vitro method. Selection of flavonoids was done according to three principles. The first rule was the value of -CDOCKER_energy should exceed 25 kcal/mol or -CDOCKER_INTERACTION energy should exceed 45 kcal/mol in order to improve the accuracy of measurement results. Secondly, flavonoids including flavonols, flavones, flavanonols and catechins with comparable structural features were given preference. Lastly, naturally derived and easily obtained compounds were chosen. The chemical structures of the 42 flavonoids are shown in Figure S1 (in Supporting Material).

In Silico Screening for Thrombin Inhibitors
One hundred and three flavonoids compounds were docked into the active site of thrombin model and identified by CDOCKER module in discovery studio 4.5 software (Accelrys, Biovia, Boston, MA, USA). The results showed 62 hit compounds with -CDOCKER_energy above 20.0 kcal/mol and 54 hit compounds with -CDOCKER_INTERACTION energy above 40.0 kcal/mol (see Table S1 in Supporting Material). The docking effect was good since the -CDOCKER_energy was higher than 15 kcal/mol [24].
Based on the above results, the anti-thrombin activities of 42 target flavonoids were selected for further research using a chromogenic substrate in vitro method. Selection of flavonoids was done according to three principles. The first rule was the value of -CDOCKER_energy should exceed 25 kcal/mol or -CDOCKER_INTERACTION energy should exceed 45 kcal/mol in order to improve the accuracy of measurement results. Secondly, flavonoids including flavonols, flavones, flavanonols and catechins with comparable structural features were given preference. Lastly, naturally derived and easily obtained compounds were chosen. The chemical structures of the 42 flavonoids are shown in Figure S1 (in Supporting Material).   Pearson correlation analysis was established using SPSS 21.0 (SPSS, Chicago, IL, USA) based on the anti-thrombin IC 50 , -CDOCKER_energy and -CDOCKER_INTERACTION energy of the 22 flavonoids with IC 50 values less than 500µM. Results showed that IC 50 was negatively correlated with -CDOCKER_energy (p(corr) = −0.55, p = 0.008) and positively correlated with -CDOCKER_INTERACTION energy (p(corr) = 0.45, p = 0.035). It was demonstrated that the target compound shows high anti-thrombin activity when the -CDOCKER_energy was higher and -CDOCKER_INTERACTION energy lower.

Binding Site of Myricetin, Scutellarein, Isorhamnetin and Myricitrin in Thrombin Model
Binding mechanisms of flavonoids and thrombin of strong inhibitors and moderate inhibitors were investigated using Discovery Studio (DS). The theoretical binding sites of these compounds interaction with thrombin model (2GDE) are shown in detail ( Table 2). For a detailed observation, the top four flavonoids with the IC50 values less than 100 µM (myricetin, scutellarein, isorhamnetin and myricitrin) were selected. The binding conformations of the 4 target flavonoids at the active site of thrombin are shown in Figure 2. The OHs on sites, C5 and C7 of the A ring of myricetin interacted with related residues of 2GDE, which displayed the two OHs on these sites were beneficial for anti-thrombin activity. Furthermore, there were interactions between the OH groups at C3 , C4 and 5 sites at the C ring of myricetin and SER195, CYS191 and GLY216, respectively. It was worth noting that more OH groups on C ring could increase thrombin inhibition. With respect to scutellarein, the oxygen atom on its C1 site and hydroxyl on the B ring interacted with 2GDE. The steric hindrance effect was weak due to lack of substituent group and this provided an excellent opportunity for the whole flavonoid molecule to enter the interior of the protein and interact with the protein residues. For isorhamnetin, hydroxyls on the C5, C7 sites and O-methylation on the B ring interacted with 2GDE. The results showed that the presence of OCH 3 on B ring improved bioactivity. Lastly, 2GDE residues of thrombin were bound by hydroxyl groups at C7 and C3 sites and O-glycosyl groups at the C3 site increased the interaction between the glycoside and protein; however, this hindered some interactions between other sites and the target.  The presence of C2=C3 and C4=O bonds of flavones was essential for inhibition of thrombin. Lack of either led to the loss of thrombin inhibition, which was consistent with the published report [17]. Thrombin inhibition was higher with trihydric phenol hydroxyl groups at the C5, C6 and C7 sites compared to meta-phenolic hydroxyl at the C5, C7 sites. In addition, more OH groups in the A-ring are beneficial for inhibitory effect [17]. However, the phenomenon was broken by the O-methylation at the C6 or C7 or C8 site. For example; scutellarein (C5, 6, 7-OH, IC 50 = 70.8 µM) vs. apigenin (C5, 7-OH, without C6-OH, IC 50 = 96.2 µM) vs. hispidulin (C5, 7-OH, C6-OCH 3 , IC 50 = 125 µM) vs. genkwanin (C5-OH, C7-OCH 3 , IC 50 = 212 µM). Also, baicalein (C5, 6, 7-OH, IC 50 = 249 µM,) vs. chrysin (C5, 7-OH, IC 50 > 500 µM) vs. wogonin (C5, 7-OH, C8-OCH 3 , IC 50 > 500 µM). In another instance, luteolin (C5, 7-OH, IC 50 = 146 µM) vs. 6-methoxyluteolin (C5, 7-OH, C6-OCH 3 , IC 50 > 500 µM). There was an exception with hydroxygenkwanin (C5-OH, C7-OCH 3 , IC 50 = 99.7 µM) with an OCH 3 at the C7 site that had a higher activity than luteolin (C5, 7-OH, IC 50 = 146 µM). Moreover, a hydroxyl at the C5 site weakened the anti-thrombin activity in fisetin (C7-OH, IC 50 = 175 µM) vs. quercetin (C5, 7-OH, IC 50 = 205 µM). This could be attributed to hydrogen-bond interaction between C5-OH and C4 = O on the quercetin. Moreover, an ortho-hydroxyl substitution on the C7 site near the C8-OH site had a moderate influence on inhibitory activity; for instance, kaempferol (C5, 7-OH, without C8-OH) vs. herbacetin (C5, 7, 8-OH) with the IC 50 values 107 µM and 133 µM, respectively.
The thrombin inhibitory activity of flavones and flavanones increased when OH group at the C7 site was replaced by O-glycosyl group. For example, baicalin vs. baicalein with IC 50 values of 88.6 µM and >500 µM, respectively. In the case of flavonol, there was an exception. Kaempferol-7-O-β-d-glucopyranoside with an O-glycosyl groups at the C7 site was inactive to thrombin with IC 50 > 500 µM compared with kaempferol (IC 50 = 107 µM). Glycosylation at the 7 position on the ring system might have affected activity in a different way to glycosylation at C3.

Materials and Chemicals
Chromogenic substrate S-2238 was purchased from Chromogenix (Milan, Italy); bovine thrombin and argatroban were purchased from Sigma-Aldrich (

The Pretreatment of Receptor and Ligands
The crystal structure of thrombin (2GDE, PBD ID) and target protein were downloaded from the PBD database and pre-processed by Discovery Studio (DS) V 4.5 (Accelrys, Biovia, Boston, MA, USA) including addition of hydrogen atoms, protonation treatment, and deletion of water molecules, original ligand molecule (SN3) and other unrelated protein conformations. Then the crystal structure was energy optimized by adding CHARMm force field. The active pockets (sphere) were defined and radius modified to 9.0 Å based on its original ligand position.

The Pretreatment of Receptor and Ligands
The crystal structure of thrombin (2GDE, PBD ID) and target protein were downloaded from the PBD database and pre-processed by Discovery Studio (DS) V 4.5 (Accelrys, Biovia, Boston, MA, USA) including addition of hydrogen atoms, protonation treatment, and deletion of water molecules, original ligand molecule (SN3) and other unrelated protein conformations. Then the crystal structure was energy optimized by adding CHARMm force field. The active pockets (sphere) were defined and radius modified to 9.0 Å based on its original ligand position.
All the collected compounds (103 flavonoids) from traditional Chinese herb medicines were sketched by ChemDraw V10.0 (Cambridge Soft, Cambridge, MA, USA). These compound structures were imported into DS, and then, CHARMm force field was added to minimize the energy.

Molecular Docking and Virtual Screening
According to the above conditions, the compound structures and 2GDE were imported into DS for screening the potential flavonoids components which could act on thrombin. Docking results of 2GDE and ligand molecule were graded by "-CDOCKER_energy" and "-CDOCKER_INTERACTION energy".

Thrombin Inhibitory Activity Assay
Activity of thrombin was measured on Multiskan™ FC microplate photometer (Thermo Scientific, Waltham, MA, USA) based on published method with modification to suit our requirements [4,19,23]. Briefly, 80 µL thrombin solution (14 µg/mL (46 U/mg) in phosphate buffer saline, pH 7.2-7.4, 0.01 M) was pre-incubated in 96-well plates with 10 µL of each flavonoid solution (1-10 mM in DMSO) and 90 µL PBS at 37 • C for 5 min and shaken for 1 min. Twenty microliter (20 µL) chromogenic substrate S-2238 (4 mg/mL in PBS) was then added to activate the reaction and incubated for 15 min at 37 • C after shaking for 1 min. All assays were paralleled three points and then measured at 405 nm in the micro-plate reader immediately. The blank group was 10 µL DMSO without flavonoids whereas Argatroban was used positive control. Background group was 10 µL flavonoid solutions (0.5-10 mM in DMSO) with 190 µL phosphate buffer solution. According to the pretest, 5% of DMSO in the total volume had no effect on thrombin. The thrombin inhibition rate was calculated using the following equation: where A blank was the absorbance of the blank group; A sample was the absorbance of the sample group, and A background was the absorbance of background group. IC 50 values for active compounds were determined on three replicates of six concentrations. Statistical analyses and IC50 values were calculated using GraphPad Prism version 5.01 (GraphPad Software Inc., La Jolla, CA, USA).

Conclusions
In this study, potential active flavonoids were screened from traditional Chinese medicine by molecular docking; the activity was studied by enzymatic bioassay, and SARs of natural flavonoids against thrombin were revealed. Molecular docking provided a method of screening target compounds with specific structure from a large molecular dataset quickly and easily. The enzymatic bioassay focused on functional verification of thrombin inhibitors from the perspective of molecular biology. The combination of the two methods provided a reference to accurately explore the structural and activity characteristics of flavonoids against thrombin. In this paper, the results showed that the C2, C3 and C4 sites on C ring and C5, C7 sites on A ring as well as C3 , C4 and C5 sites on B ring were the key sites of flavonoids for activity against thrombin. Moreover, C=C, C=O bonds and OH groups were the indispensable substituent group for flavonoids against thrombin. In conclusion, these are the key aspects revealed through the SAR analysis of natural flavonoids on anti-thrombin activity. These findings would promote the study on activity mechanism of flavonoid-type compounds as well as aid medicinal chemists to design and develop more potent flavonoid-type inhibitors against thrombin.
Supplementary Materials: The following are available online. Figure S1: the molecular structures of 42 flavonoids. Table S1: the docking score of 103 compounds.