Characterization of thrombin/factor Xa inhibitors in Rhizoma Chuanxiong through UPLC-MS-based multivariate statistical analysis

Background The dry root and rhizome of Ligusticum chuanxiong Hort., or Chuanxiong, has been used as a blood-activating and stasis-removing traditional Chinese medicine for 1000 years. Our previous studies have shown the inhibitory activity on platelet and thrombin (THR) of Chuanxiong. THR and factor Xa (FXa) play significant roles in the coagulation cascade and their inhibitors are of valuable in the treatment of thromboembolic diseases. The aim of the present study is to screen THR and FXa inhibitors from Chuanxiong. Methods Four extracts [ethyl acetate (EA), butanol (BA) and remained extract (RE) from 75% ethanol extract, and water extract (WE)] of Chuanxiong were prepared, and their THR/FXa inhibitory activities were assessed in vitro. Following silica-gel column chromatography (SC), the active EA extract and BA extract was further partitioned, respectively. Their active fractions (EA-SC1 to EA-SC5; BA-SC1 to BA-SC5) were obtained and analyzed by LC–MS. After modeling by the principal component analysis (PCA) and orthogonal partial least squares discriminate analysis (OPLS-DA), the specific marker compounds were predicted and identified. Their enzyme inhibitory was assessed in vitro and interactions with THR/FXa were investigated by molecular docking analysis. Results Chuanxiong EA extract showed strong activity against THR and BA extract was more effective in inhibiting FXa activity, and their fractions exhibited obvious difference in enzyme inhibitory activity. Furthermore, marker compounds a–h were predicted by PCA and OPLS-DA, and their chemical structures were identified. Among them, senkyunolide A, Z-ligustilide, ferulic acid and senkyunolide I (IC50 was determined as 0.77 mM) with potential THR inhibitory activity, as well as isochlorogenic acid A with FXa inhibitory activity were screened out. It was found that the four components could interact with the active site of THR, and the binding energy was lower than − 5 kcal/mol. Isochlorogenic acid A were bound to the active site of FXa, and the binding energy was − 9.39 kcal/mol. The IC50 was determined as 0.56 mM. Conclusions THR/FXa inhibitory components in different extracts of Chuanxiong were successfully characterized by the method of enzyme inhibition activity assays with ultra performance liquid chromatography-quadrupole time of flight mass spectrometry-based multivariate statistical analysis.


Background
Thrombin (THR) and factor Xa (FXa) are the members of the serine protease family. As the pivotal enzyme in the blood coagulation processes, FXa act as a catalyst in the THR production by activating prothrombin without existing THR affected [1,2], and THR catalyze the conversion of fibrinogen into insoluble strands of fibrin, as well as stimulate and recruit platelets to the injured site [3]. Due to the importance during blood coagulation cascade, FXa and THR potentially have emerged as attractive targets for new anticoagulants to treat thrombotic diseases. On the other hand, there have been several literature focusing on the study of THR/FXa inhibitory activity of natural products, which include polypeptides [4][5][6][7], polyphenols [8,9], saponins [10] and other compounds [11][12][13]. Rhizoma Chuanxiong, the dried root of Ligusticum chuanxiong Hort. (Umbelliferae), namely Chuanxiong in Chinese, is a famous blood circulation promoting medicine and is one of the clinically used traditional Chinese medicine (TCM) in protecting cardiovascular system. As genuine medicinal material, Chuanxiong is mainly distributed in Sichuan province [14]. Its major chemical components include phthalide lactones, alkaloids, phenolic acids and other constituents [15]. A large body of studies has shown that Chuanxiong possesses multifarious pharmacological effects, including protective effects on neuron [16], heart [17], liver [18] and kidney [19], as well as antioxidation [20,21], anti-inflammation [22][23][24], etc. In our previous studies, Chuanxiong extracts had inhibitory effects on platelet aggregation [25] and THR activity [12], while there are few studies on its effects on THR and FXa so far. It is of reasonable to screen THR/FXa inhibitors from Chuanxiong.
Natural products, especially TCMs, are valuable sources of active components for the discovery of novel clinical drug candidates [26][27][28]. Various techniques to characterize bioactive components from natural products had been reported. The multivariate statistical analysis method can analyze huge amount data generated from liquid chromatography paired with mass spectrometry (LC-MS), and rapidly distinguish the chemical difference among different sample groups [29]. This method had been adopted several times in natural product research for screening of bioactive components [30] or quality control markers [31], studying mechanisms of TCM processing [32] and compatibility [33]. The pharmacological ingredients from the natural products can be efficiently determined by multivariate statistical analysis when combined with bioactivity analysis [34]. Recently, this method has been proved practical and effective in identifying antiplatelet components of edible Citrus limon [35], for the analysis of antidiabetic compounds from TCM formula Ge-Gen-Qin-Lian decoction [36] and for the screening of potential THR/FXa inhibitors from Danshen [37].
In this study, ultra performance liquid chromatography-quadrupole time of flight mass spectrometry (UPLC-QTOF-MS) combined with enzyme inhibition activity assays were carried out for analyzing different Chuanxiong fractions. Principal component analysis (PCA) was used to reduce the dimensionality of MS data. Then, orthogonal partial least squares discriminant analysis (OPLS-DA) models were fitted to find out the differential marker compounds based on activity assay results, and their structures were identified. Furthermore, the interaction behaviors between the selected compounds and the enzyme were elucidated by molecular docking analysis. Finally, the enzyme inhibition activities of the marker compounds were evaluated. The Tris-HCl buffer was prepared by adding 1 M HCl to 10 mM Tris solution (pH 8.0). All samples were prepared in Tris-HCl buffer (10 mM, pH 8.0) with cosolvent DMSO, and diluted to the required concentrations for THR/FXa inhibitory assay, which were stored at 4 °C and shielded from light before use. The stock solution of THR was prepared in Tris-HCl buffer (10 mM, pH 8.0) with the enzyme activity of 500 U/mL, and stored at − 20 °C. FXa was also dispensed in Tris-HCl buffer (10 mM, pH 8.0) with the enzyme activity of 0.5 IU/mL and stored at 4 °C.

Preparation of sample extracts
After comminution, 100 g of Chuanxiong powder was extracted with 800 mL 75% ethanol (1:8, w/v) in a 2 L glass-stoppered conical flask on water bath at 80 °C for 1 h; then the extract was filtered, and the residue was collected and extracted with the above process for twice. The three filtrates were combined and concentrated in a rotary evaporator (ZFQ 85 A, Shanghai Medical Instrument Special Factory, Shanghai, China) at 45 °C. After removing ethanol completely, the concentrate were degreased with petroleum ether (2:1, v/v), and further subjected to liquid-liquid partitioning to afford EA (2:1, v/v), BA (1:1, v/v) soluble extracts as well as the remained extract (RE). After removing the solvent from each solution, the extracts were obtained by reduced pressure distillation and vacuum dry method (DZF-6050, Shanghai Jing Hong Laboratory Instrument Co., Ltd., Shanghai, China). In addition, the residues was extracted twice with 600 mL water (1:6, w/v) on water bath at 80 °C. The two filtrates were combined and evaporated, and further vacuum-dried to obtain the water extract (WE). Total 200 g Chuanxiong powder was used for extraction.

In vitro THR/FXa inhibitory activity assays
THR/FXa inhibitory activity assays were performed on an Agilent 7100 3 D capillary electrophoresis (CE) system (Agilent Technologies, Palo Alto, CA, USA), which was equipped with a diode array detector (DAD) and Agilent ChemStation software. All of the experimental procedures were implemented according to literature with minor modifications [38]. The preparation process of immobilized enzyme microreactor (IMER) is described as follows. A new bare fused-silica capillary (i.d. 75 μm, purchased from Yongnian Ruifeng Chromatographic Device Co., Ltd., Hebei, China) were pretreated with NaOH (1 M) and deionized water for 15 min and 10 min, respectively. An automated program was set to prepare the IMER: The dopamine solution (2 mg/mL) was injected into the capillary with a voltage of + 10 kV for 10 s, stayed for 30 min, and then washed out the free dopamine using running buffer (10 mM Tris-HCl buffer solution, pH 8.0) with a pressure of − 100 mbar for 90 s. Then, the 125 U/mL THR solution (or 0.5 IU/ mL FXa solution) was introduced into the capillary with a voltage of + 10 kV for 10 s, kept for 30 min; and then flushed by running buffer with a pressure of − 100 mbar for 90 s to wash out free enzyme. The prepared IMER could be immediately used for enzyme inhibitory activity assay. The temperature of the capillary cartridge was maintained at 25 °C during the CE analysis. The enzyme inhibitory activity assays were carried out by a reaction that the 2.5 mg/mL substrate solution (S-2238 for THR assays, S-2765 for FXa assays) with/without inhibitors were injected into the IMER inlet by applying a voltage of + 10 kV for 10 s and incubated for 60 s to trigger amidolytic reaction. In order to detect the product p-nitroaniline, the voltage of + 25 kV was applied to separate the reaction mixtures and the detection wavelength was set at 405 nm. The inhibition percentage was calculated by the formula: where A blank and A sample represent the peak area of product yielded by enzymatic reaction of the blank and sample group, respectively. All assays were performed in triplicate and the % of inhibition was the mean of three observations.

HPLC-DAD analysis
HPLC analysis was performed on an Agilent 1260 Series liquid chromatography system (Agilent Technologies, Palo Alto, California, USA) which was equipped with a vacuum degasser, a binary pump, an autosampler, and (1)

UPLC-QTOF-MS analysis
A Waters ACQUITY ™ UPLC system coupled with a QTOF SYNAPT G2-Si high-definition mass spectrometer (Waters, Manchester, UK) equipped with an electrospray ionization (ESI) source, was used for the LC-MS analysis and identification. The LC conditions were implemented according to literatures with minor modifications [39]. Chromatographic separations were achieved on an ACQUITY ™ BEH C18 column (100 mm × 2.

Data processing and multivariate analysis
The raw LC-MS data of Chuanxiong samples were extracted and processed by Progenesis QI software (Waters Corporation, Milford, MA, USA). After peak recognition, alignment and integration, the intensity of each ion was normalized across samples according to total intensity of each chromatogram. A resultant threedimensional dataset composing of the sample code, peak name (t R -m/z pair) and ion intensity, was generated. After data pre-treatment by mean-centered and paretoscale methods, multivariate statistical analysis, including PCA and OPLS-DA, were conducted by SIMCA-P + 13.0 Software (Umetrics, Umeå, Sweden).

In silico molecular docking of THR/FXa and identified potential active compounds
The purpose of in silico molecular docking study is to validate the binding energy between enzymes and small molecular compounds, which were carried out by Auto Dock 4.2 program (The Scripps Research Institute, La Jolla, CA, USA). The docking operation was performed according to the following steps. Firstly, prepare the file of receptor protein. The X-ray co-crystal structure file of THR-argatroban complex (PDB code = 1DWC [40]) and FXa-rivaroxaban complex (PDB code = 2W26 [41]) were acquired from Protein Data Bank database. Next, the co-crystallized ligand, water were removed, and polar hydrogen atoms were added. Then, the 3D chemical structures of small molecular compounds were drawn by Chem Office and minimized energy with outputting in PDB format. Finally, Lamarckian genetic algorithm (LGA) was employed and the number of GA runs was equal to 50 for finding the most favorable ligand binding orientations. The 2D interaction diagrams of optimum conformation after docking was generated by Discovery Studio 4.5 (Dassault Systèmes BIOVIA, San Diego, CA, USA) to observe the interaction between molecular compounds and the residues of enzyme.

Bioactivity-guided fractionation
The activity evaluation tests against THR/FXa of Chuanxiong different polar extracts (EA, BA, RE and WE extracts, 1.5 mg/mL) and each positive control, argatroban and rivaroxaban (0.5 mg/mL) were employed, and the results were expressed in Fig. 1. The EA extract was more effective in inhibiting THR activity, and the BA extract showed the strongest inhibitory activity toward FXa, which were prioritized for further fractionation. The EA extract (4.16 g) was applied to normal silica gel column chromatography (SC), and eluted with PE-EA (40:1 to 1:20, v/v) followed by 100% EA. HPLC analysis was applied to recombine the obtained fractions to give five fractions (EA-SC1 to EA-SC5). The BA extract (4.35 g) was eluted with 100% EA, followed by EA-ME (50:1 to 1:3, v/v) and 100% ME in sequence. HPLC analysis was also applied and five fractions were yielded (BA-SC1 to BA-SC5). THR inhibitory activity of the five EA fractions was measured (Fig. 2a). Fraction EA-SC3 exhibited the strongest inhibitory effect, and fractions EA-SC2, EA-SC4 and EA-SC5 were shown moderate activity, while the inhibition percentage (%) of THR by EA-SC1 was consistently low compared to the other fractions. As shown in Fig. 2b, the results of FXa inhibitory activity assays among BA fractions indicated that these five fractions could be grouped into two classifications: the most active (BA-SC1 and BA-SC2) and moderate active (BA-SC3 to BA-SC5) groups.

Multivariate statistical analysis of active compounds from different fractions
The Chuanxiong fractions, separated by silica column, was subjected to LC-MS analysis in order to conduct an   generated by Progenesis QI and then was separately imported into SIMCA for unsupervised PCA analysis and supervised OPLS-DA analysis.
To provide visualization of the differences among the Chuanxiong EA fractions, unsupervised PCA analysis was conducted. The score plot (Fig. 5Aa) showed preferably discriminative distribution, and the PCA map could be mainly divided into two clusters: EA-SC1, EA-SC2 to EA-SC5. A correlation was found that this cluster was similar to total ion chromatography of each EA fraction (Fig. 3), which is also consistent with the THR inhibitor activity (Fig. 2). The values of the PCA model fit parameters were 0.782 of R 2 X (cum) and 0.622 of Q 2 (cum) and all the samples fell well inside the 95% confidence interval, indicating a good PCA model. Subsequently, to explore the potential active marker compounds, supervised OPLS-DA analysis was employed to group the Chuanxiong EA fractions in a binary classification as the active and less active groups. The OPLS-DA score plot was presented in Fig. 5Ab, and the five fractions are clearly distinguished and could be classified as active (EA-SC2 to EA-SC5) and less active (EA-SC1). The model fit parameters R 2 Y (cum) and Q 2 (cum) were 0.996 and 0.984, respectively, and all the observations fell within the Hotelling T2 (0.95) ellipse, which suggested that the OPLS-DA model (M THR ) exhibited good fitting and predictability [42]. In the S-plot (Fig. 5Ac), the ion points far away from the centre (the corner of "S"-shaped curve)  indicated a larger contribution to the classification of the samples. Marker ions a-g were selected with high variable importance in the projection (VIP) scores (VIP > 1). The detailed information was listed in Table 1.
The same procedure was employed to discover potential marker compounds from Chuanxiong BA fractions. PCA analysis were employed for investigating the similarity of the constituent profiles of Chuanxiong BA fractions. The values of the PCA model fit parameters were 0.995 for R 2 X (cum) and 0.985 for Q 2 (cum) and all the samples fell well inside the 95% confidence interval. As shown in Fig. 5Ba, in the PCA scores plot of Chuanxiong BA fractions, BA-SC1 and BA-SC2 were separated into a cluster distinct from other fractions, which exhibited a trend similar to total ion chromatography of each BA fraction and were observed corresponding with the results of FXa inhibitory activity assays. Then, the OPLS-DA model (M FXa ) was fitted and showed good fitness and predictability with Q 2 (cum) = 0.963, R 2 Y (cum) = 0.985. All the observations fell within the Hotelling T2 (0.95) ellipse. As presented in Fig. 5Bb, the OPLS-DA score plot illustrated that the five fractions could be clearly distinguished and classified as active (BA-SC1, BA-SC2) and less active (BA-SC3 to BA-SC5). In the S-plot (Fig. 5Bc), marker ion h in the extreme lower left quadrant was selected.

Molecular docking analysis of THR/FXa and identified potential active compounds
Molecular docking can be used to study the binding mechanism of compounds interacting with the active site of proteins. The docking energy and binding residues of four markers (from Chuanxiong EA fractions) with THR active site were gathered in Table 2. The active sites of THR have four binding pockets [47]: S1 pocket (specificity pocket), S2 pocket (proximal pocket), S3 pocket, and S4 pocket (aryl binding pocket). For the docking with THR (Fig. 7), it was observed that four marker compounds could insert into the catalytic active pocket of THR like original ligand argatroban via multifarious interactions such as hydrogen bond and van der Waals, etc. The main part of argatroban interacted with

In vitro activity assessment of the predicted compounds
The reference compounds, which were based on the four identified markers (senkyunolide A, Z-ligustilide, ferulic acid and senkyunolide I) screened from Chuanxiong EA fractions, were further examined for their inhibitory effects on the THR activity. Among them, senkyunolide I strongly inhibited THR activity with an IC 50 value of 0.77 mM (Fig. 9). Senkyunolide A and Z-ligustilide on the inhibiting of THR activity were weaker than that of senkyunolide I with % inhibition around 40% at a relatively high concentration (0.5 mM), while ferulic acid did not show inhibitory effect on the THR activity under such concentration. The inhibition results were summarized in Table 4. The marker compound screened from Chuanxiong BA fractions, isochlorogenic acid A, and its isomers (isochlorogenic acid B and C) were further examined for their potential inhibitory effects on the FXa activity. From the results shown in Fig. 9, these compounds possessed FXa inhibitory effects in a dose-dependent behavior, and IC 50 value were 0.56, 0.77 and 0.61 mM, respectively.

Discussion
Chuanxiong EA extract has the strongest inhibitory activity against THR among four different polar extracts, and its fractions (EA-SC1 to EA-SC5) exhibited activity difference. In order to compare the chemical profiles of five fractions, multivariate statistical analysis with two fraction classes were used. Based on OPLS-DA model, markers a-g were the main components that contribute to the difference of composition and enzyme inhibitory activity among fractions. Four of them were identified as senkyunolide A (a), Z-ligustilide (d), ferulic acid (f) and senkyunolide I (g). FXa inhibitory activity assessment results demonstrated a different tendency, in which Chuanxiong BA extracts showed the highest enzyme inhibitory activity. An FXa inhibitor isochlorogenic acid A (h) was screened using the same way. Chuanxiong extracts has the different effective position toward THR and FXa, which give a hint that THR inhibitor might mainly exist in low polar phthalides and FXa inhibitors could be found in high polar phenolic acids.

Hydrogen bond Van der Waals Electrostatic interaction and other
Isochlorogenic acid A − 9. 39  GLN192, GLY193, THR98, GLU97,  TYR99, SER214, ASP189   GLN61, MET180, ILE175, PHE174,  HIS57, SER195, GLY216, VAL213,  ILE227, TYR228, GLY226, TYR225,  GLY219, CYS191, ASP194 TRP215, ALA190, CYS220 Isochlorogenic acid B − 8.67  GLN192, GLN61, TRP215, SER214,  GLY216, GLU147, GLY226   HIS57, SER195, GLY193, ASP189,  ILE227, ARG143, LYS148, GLY219,  TYR99 TYR228, VAL213, ALA190, CYS191, CYS220 Molecular docking can be used to study the binding mechanism of compounds interact with protein such as characterizing the binding site and evaluating the strength of interaction [48]. It was demonstrated that the four screened marker compounds from Chuanxiong EA extracts were bound to the THR active sites and the binding energies were below − 5 kcal/mol. Generally, the region with binding energy lower than − 5.0 kcal/ mol could be considered the "Potential Targets" [49]. Therefore, senkyunolide A, Z-ligustilide, ferulic acid and senkyunolide I were potential THR inhibitors. Their inhibitory effects on THR activity were examined. Among them, senkyunolide I had the strongest activity, and senkyunolide A and Z-ligustilide showed moderate activity, while ferulic acid exhibited no effect which could explain the low binding energy of ferulic acid that is close to − 5 kcal/mol. In addition, the IC 50 of senkyunolide I differed from the report [12], which may be due to the different experimental methods and conditions (such as concentrations of substrate and enzyme activity). Likewise, the interaction between isochlorogenic acid A, B and C and FXa were investigated. The result of binding energy of isochlorogenic acid A and C (< − 9 kcal/mol) is better than isochlorogenic acid B (− 8.67 kcal/mol). Isochlorogenic acid A and C interacted with the S1 and S4 pocket of FXa, and isochlorogenic acid B could interact with S1 and S2 pocket. It has been reported that S1 and S4 pockets are commonly used to predict high-affinity FXa inhibitors [50]. Therefore, these three molecules are potential inhibitors of FXa. The results of inhibitory effects on the FXa activity indicated that isochlorogenic acid A and C had the lower IC 50 value. A possible explanation is that they have similar binding site and energy.
Senkyunolide I, which has been screened out as a THR inhibitor from Chuanxiong ethanol extract by ultrafiltration in our previous study [12], was also screened out in this study along with senkyunolide A and Z-ligustilide. Among the three compounds, senkyunolide I had the strongest inhibitory activity, probably because of its two phenolic hydroxyl groups have hydrogen bonding with SER195 that belongs to the THR catalytic triad. In addition, one of the FXa inhibitors found in this study is isochlorogenic acid C, which was also discovered as a THR inhibitor in our previous study [12]. The explanation may be that its larger molecular structure and more hydroxyl groups make it able to bind to the residues in the active pocket of the center of THR/FXa pocket. Therefore, it could be a dual-enzyme inhibitor.

Conclusion
Rhizoma Chuanxiong has been used for thousands of years in TCM, and is well-known for its properties of "HuoXueHuaYu" (activating blood circulation to remove blood stasis). It has various kinds of biological activities such as vasodilation, antiinflammatory, antioxidation, antiproliferation, etc. Through the combination method of LC-MS analysis, THR/FXa activity assessment and multivariate statistical analysis, this study predicted and identified four marker compounds (senkyunolide A, Z-ligustilide, ferulic acid and senkyunolide I) with potential THR inhibitory activity from Chuanxiong EA fractions, and one marker compound isochlorogenic acid A, with potential FXa inhibitory activity from Chuanxiong BA fractions. Docking results showed that five screened compounds could insert into the catalytic active site of enzyme, and the binding energy was lower than − 5 kcal/mol. The IC 50 of senkyunolide I and isochlorogenic acid A was 0.77 and 0.56 mM, respectively. In addition, two other FXa inhibitors, isochlorogenic acid B and isochlorogenic acid C, with similar structure to isochlorogenic acid A, were also found, with IC 50 value of 0.77 and 0.61 mM, respectively. These results proved that the proposed  method could effectively characterize the THR/FXa inhibitors in complex mixtures, which not only complemented the anticoagulant mechanism of Rhizoma Chuanxiong, but also provided a clue for the discovery of new active THR and/or FXa inhibitors.