Screening of Potential Thrombin and Factor Xa Inhibitors from the Danshen–Chuanxiong Herbal Pair through a Spectrum–Effect Relationship Analysis

In this study; a spectrum–effect relationship analysis combined with a high-performance liquid chromatography–mass spectrometry (LC–MS) analysis was established to screen and identify active components that can inhibit thrombin and factor Xa (THR and FXa) in Salviae Miltiorrhizae Radix et Rhizoma–Chuanxiong Rhizoma (Danshen–Chuanxiong) herbal pair. Ten potential active compounds were predicted through a canonical correlation analysis (CCA), and eight of them were tentatively identified through an LC–MS analysis. Furthermore; the enzyme inhibitory activity of six available compounds; chlorogenic acid; Z-ligustilide; caffeic acid; ferulic acid; tanshinone I and tanshinone IIA; were tested to verify the feasibility of the method. Among them; chlorogenic acid was validated to possess a good THR inhibitory activity with IC50 of 185.08 µM. Tanshinone I and tanshinone IIA are potential FXa inhibitors with IC50 of 112.59 µM and 138.19 µM; respectively. Meanwhile; molecular docking results show that tanshinone I and tanshinone IIA; which both have binding energies of less than −7.0 kcal·mol−1; can interact with FXa by forming H-bonds with residues of SER214; GLY219 and GLN192. In short; the THR and FXa inhibitors in the Danshen–Chuanxiong herbal pair have been successfully characterized through a spectrum–effect relationship analysis and an LC–MS analysis.


Introduction
Thrombin (THR) and factor Xa (FXa), which are members of the Serine Protease family, play critical roles in the coagulation cascade [1]. In thrombotic diseases, THR catalyzes the transformation of soluble fibrinogen into insoluble fibrin in response to a vascular injury, leading to the formation of a clot [2]. FXa is located at the center of endogenous and exogenous coagulation pathways. It is worth mentioning that the formation of THR can also be reduced by inhibiting its activity to prevent the expansion of a thrombus [3]. In reality, thrombotic disease, a common type of cardiovascular disease, has gradually become a threat to public health [4]. Due to the importance of THR and FXa in the formation of thrombosis, inhibiting their activity is a feasible strategy for the treatment of thrombotic diseases [5]. At present, anticoagulant drugs including argatroban (THR inhibitor) and rivaroxaban (FXa inhibitor) are used clinically [6,7], but risks such as drug resistance and gastrointestinal bleeding may be caused by these inhibitors [8]. Therefore, it has become a trend to search for THR and FXa inhibitors in natural sources to provide a reference for the development of anticoagulant drugs [9]. Danshen (Salviae Miltiorrhizae Radix et Rhizoma, DS) and Chuanxiong (Chuanxiong Rhizoma, CX) are both effective for invigorating blood circulation and eliminating blood stasis [10]. DS and CX have attracted widespread attention as primary materials for the treatment of cardiovascular diseases (CVDs) [11], tumor disorders [12] and inflammatory diseases [13]. DS is often employed clinically combined with other traditional Chinese medicines (TCMs) for promoting blood circulation, of which Danshen-Chuanxiong (DC) is one of the most frequently used herbal pairs (mixture of plant materials) among best-selling herbal formulae [14]. In our previous studies, potential THR and FXa inhibitors in DS and CX were screened, respectively [1,5]. However, the THR and FXa inhibitory activity of the DC herbal pair has not been evaluated, and the potential active components are still unclear. Therefore, a spectrum-effect relationship analysis was performed to screen the active ingredients in the DC herbal pair that inhibit THR and FXa. A spectrum-effect relationship analysis [15], correlating the peak area in the fingerprint of TCMs with pharmacodynamic data, is employed to reveal the correlation between the components and their pharmacological activities. Due to the advantages of being a simple operation and saving time [16], spectrum-effect relationship analyses have been reliably used in the screening of natural active compounds. For instance, protocatechuic aldehyde, hydroxysafflor yellow A and tanshinone IIA were screened and identified as active tyrosinase inhibitors in the Salvia miltiorrhiza-Carthamus tinctorius herbal pair through a spectrum-effect relationship analysis [17].
In this study, the inhibition effects of the DC herbal pair with different ratios on THR and FXa were compared. Then, the components in the DC herbal pair were analyzed through HPLC and the chemical fingerprint was constructed. A spectrum-effect relationship analysis was employed to predict the potential active compounds in the DC herbal pair, and their chemical structures were identified through an LC-MS analysis. Furthermore, the in vitro THR and FXa inhibitory activities of the predicted compounds were tested. Finally, molecular docking was used to explore the possible binding sites and modes of action between the active compounds and THR and FXa, respectively.

Inhibitory Effects of DC Extracts on THR and FXa
As shown in Figure 1A, different ratios of DC herbal pair extracts show a weaker inhibitory effect on THR than the CX extract when the final concentration is 2.5 mg/mL. However, compared with the DS extract, all ratios of DC herbal pair extracts display stronger inhibitory activity on THR. The DC herbal pair extracts (1:1 and 5:1) display a weaker effect than the other DC herbal pair extracts (2:1, 3:1, 1:5, 1:3 and 1:2). As shown in Figure 1B, different ratios of DC herbal pair extracts show a stronger inhibitory effect on FXa than the CX extract when the final concentration is 2.5 mg/mL. Among them, the DC herbal pair (1:1) exhibits the strongest effect, even stronger than the DS extract. However, the inhibitory effects of other DC herbal pair extracts (2:1, 3:1, 5:1, 1:5, 1:3 and 1:2) on FXa are weaker than the DS extract. In reality, studies show that the combined use of DS and CX display a better antithrombotic effect at a ratio of 1:1 than their separate application [14], indicating the mechanism for the DC herbal pair to exhibit an antithrombotic effect may be by inhibiting the activity of FXa. The results may provide scientific evidence for the experience-based application of the DC herbal pair [18].

Spectrum-Effect Relationship Analysis
The HPLC fingerprints of DC samples with different ratios are shown in Figure 2. A canonical correlation analysis (CCA) was used to assess the spectrum-effect relationship between the areas of 71 chromatographic peaks and the inhibition rate of different extracts on THR and FXa. As shown in Table 1, the results are expressed through a Pearson correlation coefficient; "r" indicates the correlation strength, 0.6 ≤ |r| ≤ 0.8 indicates a strong correlation and 0.8 ≤ |r| ≤ 1.0 indicates a significant correlation [19]. Therefore, compounds of peaks 4, 17, 24, 55, 61 and 62 with "r" higher than 0.8 may be the active components of herbal pairs for inhibiting THR. On the other hand, compounds of peaks 5, 9, 61, 67 and 71 with "r" higher than 0.6 are potential active FXa inhibitory compounds. A further LC-MS analysis was performed to identify the structures of these compounds, and the inhibitory activity of the pure compounds was tested in vitro.

Spectrum-Effect Relationship Analysis
The HPLC fingerprints of DC samples with different ratios are shown in Figure 2. A canonical correlation analysis (CCA) was used to assess the spectrum-effect relationship between the areas of 71 chromatographic peaks and the inhibition rate of different extracts on THR and FXa. As shown in Table 1, the results are expressed through a Pearson correlation coefficient; "r" indicates the correlation strength, 0.6 ≤ |r| ≤ 0.8 indicates a strong correlation and 0.8 ≤ |r| ≤ 1.0 indicates a significant correlation [19]. Therefore, compounds of peaks 4, 17, 24, 55, 61 and 62 with "r" higher than 0.8 may be the active components of herbal pairs for inhibiting THR. On the other hand, compounds of peaks 5, 9, 61, 67 and 71 with "r" higher than 0.6 are potential active FXa inhibitory compounds. A further LC-MS analysis was performed to identify the structures of these compounds, and the inhibitory activity of the pure compounds was tested in vitro.

Identifications of the Potential Active Compounds through an LC-MS Analysis
An LC-MS analysis was used to study the chemical structures of potential active compounds; the total ion chromatogram was shown in Figure 3. Eight compounds, chlorogenic acid, caffeic acid, ferulic acid, Z-ligustilide, tanshindiol B, methyl dihydronortanshinonate, tanshinone I and tanshinone IIA, were tentatively identified by comparing their MS data with those reported in the literature [20][21][22]. The detailed MS data of ten predicted compounds are shown in Table 2, and their chemical structures are presented in Figure 4.

In Vitro Activity Tests for the Predicted Compounds
In order to confirm the feasibility of the spectrum-effect relationship method for screening of active compounds, in vitro enzymatic activity assays were performed. Argatroban, a well-known THR inhibitor [6], was selected as the positive control drug. The results are shown in Figure 5. Argatroban shows a strong inhibition effect with an IC50 value of 64.94 µM. Chlorogenic acid with an IC50 value of 185.08 µM exhibits a better inhibitory activity than Z-ligustilide with an IC50 value of 1460.65 µM on THR. As shown in Figure 6, the IC50 value of rivaroxaban, as the positive control drug [7], is 47.21 µM. Among four predicted compounds, tanshinone I and tanshinone IIA strongly inhibit the activity of FXa in a dose-dependent manner with IC50 values of 112.59 µM and 138.19 µM, respectively. In addition, caffeic acid and ferulic acid do not show a significant inhibitory effect with IC50 values of 3789.96 µM and 2488.90 µM, indicating that they may not act directly on the active center of FXa.

In Vitro Activity Tests for the Predicted Compounds
In order to confirm the feasibility of the spectrum-effect relationship method for screening of active compounds, in vitro enzymatic activity assays were performed. Argatroban, a well-known THR inhibitor [6], was selected as the positive control drug. The results are shown in Figure 5. Argatroban shows a strong inhibition effect with an IC 50 value of 64.94 µM. Chlorogenic acid with an IC 50 value of 185.08 µM exhibits a better inhibitory activity than Z-ligustilide with an IC 50 value of 1460.65 µM on THR. As shown in Figure 6, the IC 50 value of rivaroxaban, as the positive control drug [7], is 47.21 µM. Among four predicted compounds, tanshinone I and tanshinone IIA strongly inhibit the activity of FXa in a dose-dependent manner with IC 50 values of 112.59 µM and 138.19 µM, respectively. In addition, caffeic acid and ferulic acid do not show a significant inhibitory effect with IC 50 values of 3789.96 µM and 2488.90 µM, indicating that they may not act directly on the active center of FXa.

Molecular Docking of THR and FXa and Identified Active Compounds
Molecular docking can be used to reveal the possible modes and sites of compounds interacting with THR and FXa. As shown in Table 3, the binding energies of chlorogenic acid and Z-ligustilide are less than −5.0 kcal·mol −1 , indicating that they are potential THR inhibitors. Chlorogenic acid is combined with THR through Van der Waals forces and ionic bonds. Z-ligustilide can enter THR catalytic activity pockets with a strong affinity by forming H-bonds with residues of SER214 and GLY226. The 2D interaction diagrams of chlorogenic acid and Z-ligustilide with THR can be observed in Figure 7. The binding energies and residues of caffeic acid, ferulic acid, tanshinone I and tanshinone IIA with FXa are summarized in Table 4. Among them, tanshinone I and tanshinone IIA, which have binding energies of less than −7.0 kcal·mol −1 , can interact with FXa by forming H-bonds with residues of SER214, GLY219 and GLN192. In addition, caffeic acid and ferulic acid, which have binding energies of −5.82 kcal·mol −1 and −6.35 kcal·mol −1 , are combined with FXa through Van der Waals and other forces. It is worth mentioning that the molecular docking results of the four active components are consistent with their activity in in vitro tests. Tanshinone I and tanshinone IIA with stronger affinity exhibit better inhibition effects. The 2D interaction diagrams are shown in Figure 8.

In Vitro Activity Tests for the Predicted Compounds
In order to confirm the feasibility of the spectrum-effect relationship method for screening of active compounds, in vitro enzymatic activity assays were performed. Argatroban, a well-known THR inhibitor [6], was selected as the positive control drug. The results are shown in Figure 5. Argatroban shows a strong inhibition effect with an IC50 value of 64.94 µM. Chlorogenic acid with an IC50 value of 185.08 µM exhibits a better inhibitory activity than Z-ligustilide with an IC50 value of 1460.65 µM on THR. As shown in Figure 6, the IC50 value of rivaroxaban, as the positive control drug [7], is 47.21 µM. Among four predicted compounds, tanshinone I and tanshinone IIA strongly inhibit the activity of FXa in a dose-dependent manner with IC50 values of 112.59 µM and 138.19 µM, respectively. In addition, caffeic acid and ferulic acid do not show a significant inhibitory effect with IC50 values of 3789.96 µM and 2488.90 µM, indicating that they may not act directly on the active center of FXa.

Molecular Docking of THR and FXa and Identified Active Compounds
Molecular docking can be used to reveal the possible modes and sites of compounds interacting with THR and FXa. As shown in Table 3, the binding energies of chlorogenic acid and Z-ligustilide are less than −5.0 kcal·mol −1 , indicating that they are potential THR inhibitors. Chlorogenic acid is combined with THR through Van der Waals forces and ionic bonds. Z-ligustilide can enter THR catalytic activity pockets with a strong affinity by forming H-bonds with residues of SER214 and GLY226. The 2D interaction diagrams of chlorogenic acid and Z-ligustilide with THR can be observed in Figure 7. The binding energies and residues of caffeic acid, ferulic acid, tanshinone I and tanshinone IIA with FXa are summarized in Table 4. Among them, tanshinone I and tanshinone IIA, which have binding energies of less than −7.0 kcal·mol −1 , can interact with FXa by forming Hbonds with residues of SER214, GLY219 and GLN192. In addition, caffeic acid and ferulic acid, which have binding energies of −5.82 kcal·mol −1 and −6.35 kcal·mol −1 , are combined with FXa through Van der Waals and other forces. It is worth mentioning that the molecular docking results of the four active components are consistent with their activity in in vitro tests. Tanshinone I and tanshinone IIA with stronger affinity exhibit better inhibition effects. The 2D interaction diagrams are shown in Figure 8.
Crude drugs (pieces of plant materials) of Danshen and Chuanxiong were purchased from Kangmei Pharmaceutical Co., Ltd. in May 2020. Their origins are from Shandong and Sichuan, respectively. The voucher specimens of Salvia miltiorrhiza Bge. (number SM2020050101) and Ligusticum chuanxiong Hort. (number CF2020050101) were deposited at the Pharmaceutical Engineering Laboratory in School of Chemistry and Chemical Engineering, Chongqing University, Chongqing, China.

Preparation of DC Extracts
Crude drugs of DS and CX were pulverized, then stored in a dry box after filtrating through a 50 mesh sieve (about 0.29 mm). Nine different ratios of DC herbal pairs (1:0, 1:1, 2:1, 3:1, 5:1, 1:5, 1:3, 1:2 and 0:1) were prepared for reflux extraction. First, 2.5 g of DS and CX mixed powder was extracted through ultrasonic extraction at room temperature for 15 min with 20 mL of 70% ethanol in a 250 mL round-bottom flask. Then, it was further extracted in a water bath at 80 • C for 30 min. The above process was repeated three times. After the extract solutions were filtered through filter paper and combined, they were evaporated in a rotary evaporator (ZFQ 85A, Shanghai Medical Instrument Special Factory, Shanghai, China) at 55 • C under reducing pressure to remove the ethanol solvent. Finally, the extracts were further dried by lyophilization with a freezing-drying system (DZF-6050, Shanghai Jing Hong Laboratory Instrument Co., Ltd., Shanghai, China) to obtain the DC extracts at a yield of about 20% (w/w, dried extract/crude herb). All solutions were stored in the dark at 4 • C. Before the HPLC analysis, the extract was dissolved in methanol (5.0 mg/mL) and filtered through a 0.22 µm membrane filter (Shanghai Titan Scientific Co., Ltd., Shanghai, China).
An Agilent 6545 equipped with electrospray ionization was used for LC/Q-TOF-MS analyses (Santa Clara, CA, USA). The LC conditions were the same as described previously. Nitrogen was used as the atomizing dry gas at a flow rate of 8.0 L/min. The Q-TOF-MS analysis was used in both negative and positive ion modes. The key parameters were set as follows: ion source temperature, 320 • C; sprayer pressure, 35 psi; spray voltage, 1000 V; and TOF acquisition speed, 0.5 s. For compound identification, MS 1 data and MS 2 data were recorded in the range of m/z 100 to 1000. An LC-MS software package (Agilent) was employed for data collection and processing.

Preparation of Buffers and Solutions
A PBS buffer was prepared by dissolving 1.44 g of Na 2 HPO 4 and 0.24 g of KH 2 PO 4 in 800 mL of water for immediate use (pH was adjusted to 8.0 by 1.0 M NaOH). A stock solution of THR (125 U/mL) was prepared by dissolving 4.16 mg of THR in 1.0 mL of PBS. A stock solution of FXa (0.5 IU/mL) was prepared by dissolving 5.0 mg of FXa in 1.0 mL of PBS. Stock solutions of S-2238 (2 mg/mL) and S-2765 (2.5 mg/mL) were prepared in PBS.
The DC extract samples (10 mg/mL) were prepared in a PBS buffer and diluted to 5.0 and 2.5 mg/mL as necessary.

Calculation of the % Inhibition
THR and FXa inhibitory activity assays were performed with an iMark™ Microplate Absorbance Reader (Bio-Rad Laboratories, Inc., Hercules, CA, USA). A DC extract solution of 50 µL and a 50 µL THR solution were mixed and incubated for 10 min at 37 • C. Subsequently, a 100 µL S-2238 (2 mg/mL) solution as the chromogenic substrate was added to start the reaction in the microplate reader. The absorbance was monitored at 405 nm every 30 s for 10 min, and each group was measured three times. As a control group, the operation process was similar to the experimental group but without a DC extract. Except for the enzyme solution and chromogenic substrate, the procedure for assessing the FXa inhibitory activity was exactly the same as for THR. The inhibition rate (%) of THR and FXa was calculated using Formula (1).
where I(%) represents the % of inhibition, and dA dt blank and dA dt sample represent the reaction rate of the control and experimental group, respectively. The results were expressed as the mean ± standard deviation (SD) of three experiments.

Spectrum-Effect Relationship Analysis
A spectrum-effect relationship analysis was performed by transferring the DC fingerprint peak area and inhibition rate of THR and FXa into SPSS software (Ver. 24.0, Chicago, IL, USA) for a canonical correlation analysis (CCA). The chromatograms of nine ratios of DC samples were calculated and generated by "Similarity Evaluation System for Chromatographic Fingerprint of Traditional Chinese Medicine" composed by Chinese Pharmacopoeia Committee (Version 2004A). A CCA was used to establish the spectrumeffect relationships between the areas of 71 peaks in the fingerprint and inhibition rate of THR and FXa, respectively, to screen potential active components.

Molecular Docking of THR and FXa and Identified Active Compounds
AutoDock 4.2 (The Scrips Research Institute, La Jolla, CA, USA) was employed for molecular docking studies to predict the site and mode of action of active compounds from THR and FXa, respectively. The brief steps for molecular docking are as follows. First, the crystal structures of THR (No. 1DWC) and FXa (No. 2W26) from the PDB library were downloaded, which were output in PDBQT format after the following operations: the water molecules and the ligands of argatrobana and rivaroxaban were deleted while polar hydrogen was added and the point charge was calculated. Second, the 3D chemical structures of active compounds were drawn by ChemDraw 3D (Ver. 15.0) to minimize energy to prepare ligands. Semi-flexible docking was used to simulate the docking process. The grid size was set to (x, y, z) = (60, 60, 60), and the grid centers of THR and FXa were set to (x, y, z) = (35.887, 19.178, 18.856) and (x, y, z) = (7.951, 5.850, 21.876), respectively. The Lamarckian genetic algorithm (LGA) was employed and the number of GA runs was equal to 50 for finding the most favorable ligand binding orientations. Finally, Discovery Studio 4.5 (BIOVIA, San Diego, CA, USA) was used to observe the interaction of active compounds with THR and FXa.

Statistical Analysis
All data are presented as the mean ± standard deviations (SD) of at least three different experiments. The statistical analysis was performed by SPSS.

Conclusions
In this study, the potential THR and FXa inhibitory components in the DC herbal pair were screened by a spectrum-effect relationship analysis. Combined with a canonical correlation analysis, in vitro THR and FXa inhibitory activity test and LC-MS analysis, two potential THR inhibitory active compounds (chlorogenic acid and Z-ligustilide) and four potential THR inhibitory active compounds (caffeic acid, ferulic acid, tanshinone I and tanshinone IIA) were predicted. Furthermore, chlorogenic acid, tanshinone I and tanshinone IIA were verified to possess good THR and FXa inhibitory activity with IC 50 values of 185.08 µM, 112.59 µM and 138.19 µM, respectively. In addition, the molecular docking results, which are consistent with their in vitro activity verification, show that tanshinone I and tanshinone IIA can interact with FXa by forming H-bonds with residues of SER214, GLY219 and GLN192.