Identification of blood-activating components from Xueshuan Xinmaining Tablet based on the spectrum–effect relationship and network pharmacology analysis

With the aim of identifying the active components of Xueshuan Xinmaining Tablet (XXT) and discussing the potential mechanism involved, the relationship between HPLC fingerprints and its blood-activating effect were established by multivariate statistical analysis, including gray relational analysis (GRA) and partial least squares regression analysis (PLSR). Network pharmacology was used to predict the potential mechanism based on the identified active components. GRA and PLSR analysis showed close correlation between the HPLC fingerprints and blood-activating activity, and peaks P1, P3, P11, P15, P22, P34, P36, P38 and P39 might be potential anti-blood stasis components of XXT. The pharmacological verification showed that salvianic acid A (P1), rutin (P3), ginsenoside Rg1 (P11) and Rb1 (P22), cinobufagin (P36), and tanshinone I (P38) and IIA (P39) had significant blood-activating effects. Based on these seven active compounds, network pharmacology analysis indicated that the anti-blood stasis effect of XXT might be closely related to TNF, PI3K-Akt and NF-κB signaling pathways. The spectrum–effect relationship of XXT was successfully established in this study. The blood-activating components and the anti-blood stasis mechanism were revealed and predicted. These findings could also be beneficial for an exploration of the active components of TCM.


Introduction
Xueshuan Xinmaining Tablet (XXT), a Chinese traditional compound medicine composed of ten herbs, possesses the activities of promoting blood circulation and removing blood stasis. 1 It is recorded in the Chinese Pharmacopoeia (2015 edition) and has been clinically applied for the treatment of cerebral thrombosis and coronary heart disease for at least een years. 2,3 In our previous studies, we have reported that XXT could treat blood stasis through regulating related genes and proteins or endogenous metabolite pathways. 1,4,5 The multitarget mechanism is closely related to the variation in the chemical components contained. We have reported the comprehensive component screening of XXT based on UPLC/Q-TOF-MS. The results showed that XXT was rich in chemical constituents of different structural types. 6 Additionally, the contents of salvianolic acid B, rutin, ferulic acid, and ginsenoside Re and Rg 1 were determined in the national standard of XXT. However, whether these ingredients are the bloodactivating substances is not quite certain.
Fingerprints are widely recognized as a quality control method worldwide. 7,8 The World Health Organization (WHO, 2007) has approved the use of ngerprints to evaluate the quality of natural drugs or traditional Chinese medicine (TCM). Fingerprints, with the advantages of convenience and a large amount of information, could systematically mark the chemical composition of TCM. 9 The UPLC-PDA ngerprint of XXT, marked with 28 common peaks in 280 nm, has been established to assess the quality consistency of XXT. 10 However, the ngerprints only focused on the chemical characteristics, but gave little information about the components' therapeutic effects. The components in XXT related to the antiblood stasis effect have not been identied.
The spectrum-effect relationship was established by linking the peaks in the ngerprints of TCM with specic bioactivity to screen the active components in TCM. 11 Multivariate statistical methods, including partial least squares regression analysis (PLSR) and gray relational analysis (GRA), have been commonly applied to analyze the regression coefficients, variable importance in projection (VIP) contribution, and their correlation. [12][13][14] As a result, the active compounds closely related to a specic effect could be effectively screened for further pharmacological activity evaluation in vitro or in vivo.
With the development of bioinformatics, network pharmacology, as a more comprehensive approach for integrating compound-target-pathway interactions from a molecular to a systematic level, is evolving as a Frontier research eld in drug discovery and development. The boom in network pharmacology has prompted more researchers to elucidate the possible mechanisms of natural products in recent years. 15,16 In this study, the blood-activating components in XXT were predicted by analysis of the spectrum-effect relationship for the chemical characteristic peaks and the anti-blood stasis effect. In view of the uniqueness of the manufacturer, and the good similarity and stability of chemical components in the XXT ngerprint, 10 ten different polar solvent extracts of XXT were used to establish HPLC ngerprints. Then, a pharmacological test in vitro was performed to verify the predicted results. Finally, the possible targets and signaling pathways of the blood-activating components were predicted by network pharmacology. This study identied the anti-blood stasis components in XXT, which could be benecial for the quality control of XXT and the exploration of blood-activating components.
(Certicate No. SCXK-(Ji) 2018-0003) and used for experiments aer a week of adaptive feeding. The rats were divided randomly into 13 groups and assigned to PC polycarbonate mouse cages (53.5 cm Â 39.0 cm Â 20.0 cm) (n ¼ 10 for each group per cage), then allowed standard diets and water ad libitum. The rats were kept in a controlled environment (temperature: 20 AE 2 C; relative humidity: 50 AE 10%) with a 12 h dark/light cycle. All animal procedures were performed in accordance with the Guidelines for Care and Use of Laboratory Animals of Jilin University and approved by the Animal Ethics Committee of Jilin University. The approval number was 201904108. , -Re, -Rf, -Rb 1 , -Rg 3 , -Rh 1 , -Rc, -Rg 2 , -F 1 , -Rb 2 , -Rb 3 , -Rd, -F 2 , cholic acid, cinobufagin, resibufogenin, tanshinone I and tanshinone IIA were dissolved in methanol to obtain the mixed reference standards solutions (the concentration of each reference standard was 0.1 mg mL À1 ).
2.2.3 Validation of method. The optimized method was validated by evaluating the precision, repeatability and stability. Precision was evaluated by successively analyzing the six test solutions from the S 4 sample. Six replicates of the S 4 sample were successively assessed to evaluate the repeatability. The stability was determined by analysis at different time intervals (0, 2, 4, 8, 16, and 24 h) in a day. Furthermore, the precision, repeatability and stability were expressed as the relative standard deviation (RSD) of each peak area (PA) of each sample.
2.2.4 Evaluation of HPLC ngerprints 2.2.4.1 SA. Each sample solution (S 1 -S 10 ) and the reference standard solutions were injected into the HPLC. Chromatograms showing the retention time (t R ) and the peak area (PA) were obtained. According to the Similarity Evaluation System for Chromatographic Fingerprint of TCM (version 2012A; Beijing, China), the ngerprints were automatically matched and established. 17 Chromatograms of S 1 -S 10 samples, including PA and t R , were put into the Analytical Instrument Association (AIA) form (*.cdf). Subsequently, a reference ngerprint was automatically generated by comparing S 1 -S 10 samples based on the median method, and the similarity (SA) between the chromatogram of the reference ngerprint and each sample chromatogram was calculated with the soware.
2.2.4.2 HCA. Hierarchical cluster analysis (HCA), a multivariate statistical analysis technique to measure dissimilarity or similarity, is commonly used to sort samples into clusters. 18 The common PAs of the samples were used as features, and the HCA of the S 1 -S 10 samples was established from the squared Euclidean distance (metric) and the between-groups linkage method (the amalgamation rule) with SPSS statistics soware (SPSS 19.0, SPSS Inc., Chicago, IL, USA). 19 2.3 Blood-activating experiments 2.3.1 Sample preparation. XXT extract samples were prepared according to the method described in Section 2.2.2. Signicantly, the desiccated extracts were suspended in distilled water instead of being dissolved in methanol. The concentrations of XXT extracts were calculated from the amount of XXT, and the nal concentrations were equivalent to 70.0 mg mL À1 of XXT. A Buchang Naoxintong Capsule (BCN) was also suspended in distilled water to produce a solution containing 70.0 mg per mL BCN. So 10 kinds of XXT extract and BCN test solutions were prepared according to a dosage of 0.7 g kg À1 . 1 2.3.2 Grouping, modeling, and administration. Aer one week of adaptive feeding, 130 male Wistar rats were randomly divided into 13 groups (10 rats per group): (i) normal (N) group, (ii) model (M) group, (iii) BCN positive control group, and (ivxiii) the XXT extract (S 1 -S 10 ) groups. Both i and ii groups were administered with distilled water. All groups were gastrointestinally administered at 10 mL kg À1 once a day for 8 days. On day 1 to day 6, half an hour aer each administration, the iixiii groups were all placed in ice water (0-1 C) for 5 min. On day 7, 30 minutes aer intragastric administration, the ii-xiii groups were injected subcutaneously with 0.1% adrenaline (1 mg kg À1 ) twice to establish an acute blood stasis model. 20 The rst dose was 1 mg kg À1 , and the second dose was reduced to 0.8 mg kg À1 at an interval of 4 hours. Two hours aer the rst dose of adrenaline, the rats in the ii-xiii groups were also immersed in an ice bath for 5 min.
2.3.3 Determination of hemorheology and coagulation indicators. In order to assess the model in rats and the effect of XXT extracts, the hemorheology indexes (whole blood viscosity: WBV; plasma viscosity: PV) and coagulation indicators (thrombin time: TT; prothrombin time: PT; activated partial thromboplastin time: APTT; brinogen: FIB) were evaluated. [21][22][23] On the night of day 7, all the rats were fasted for 12 h, but water was allowed. On day 8, 30 min aer administration, the rats were anesthetized by an intraperitoneal injection of 10% chloral hydrate (3 mL kg À1 ). Blood was drawn from the abdominal aorta and was kept in tubes containing heparin sodium (20 U mL À1 ). Some of the whole blood was used to measure the whole blood viscosity (WBV). The WBV was determined (at shear rates of 10, 60, and 120/s) at 37 C. Plasma, used to determine PV, TT, PT, APTT, and FIB, was obtained from the remaining whole blood by centrifugation at 3000 rpm for 10 min. PV was measured at a shear rate of 120/s. TT, PT, APTT, and FIB were determined with an SC 40 (LG-PABER-I) coagulation analyzer with commercial kits, in accordance with the manufacturers' instructions. The calibration curves of PT (x) and solidication time (y), FIB (x) and solidication time (y) were established. TT was tested by incubating 100 mL of plasma at 37 C for 3 min, followed by the addition of 100 mL of thrombin agent. PT was examined by incubating 50 mL of plasma for 3 min at 37 C, followed by the addition of 100 mL of thromboplastin agent. APTT was evaluated by incubating 50 mL of plasma with 50 mL pf APTT-activating agents for 3 min at 37 C, followed by the addition of 50 mL of CaCl 2 . FIB was assessed by incubating 10 mL of plasma with 90 mL of imidazole buffer at 37 C for 3 min, followed by the addition of 50 mL of FIB agent. All hemorheological indexes and coagulation function indexes were determined within 3 h aer blood collection.

Data analysis.
All quantitative data were presented as mean AE standard derivation (SD) calculated with GraphPad Prism 6.0 soware (GraphPad Soware, CA). Multiple comparisons were analyzed by one-way analyses of variance (ANOVA). A Student's t-test was carried out by comparing two groups. A pvalue < 0.05 was considered statistically signicant.

Spectrum-effect relationship analysis
2.4.1 GRA. Gray relational analysis (GRA), expressing the strengths of relationships between factors, has been widely applied for TCM studies of spectrum-effect relationships. 24,25 Compared with other statistical methods, the gray correlation analysis, with lower requirements on sample data, possesses the advantages of requiring a small sample size and amount of computation, and has been applied widely. The principle of GRA is to utilize the degree of similarity between two series of curve geometries to determine a correlation based on assessing the inuence of factors on the index. 26,27 Its basic method is to use the pharmacodynamic index of traditional Chinese medicine as the reference sequence and the characteristic peak area shared by the ngerprints as the comparison sequence. According to the gray correlation between each factor of the comparison sequence and the reference sequence, the "contribution" of each common characteristic peak to the efficacy is determined. 28,29 The higher the gray relational grade is, the greater the correlation. In this study, GRA was performed using DPS 7.05 soware (Data Processing system, China).
2.4.2 PLSR. Partial least squares regression analysis (PLSR), a multivariate statistical regression model, 30 could effectively solve the multicollinearity between a set of dependent variables (y) from a large set of independent variables (x). 31 In this study, the x matrix was composed of PA, and the y vector was constructed with APTT and the levels of WBV (10/s). The PLSR model was established with SIMCA-P soware (SIMCA-P13.0, Umetrics, Umeå, Sweden). 32 The relative inuence of the independent variables on the dependent variables was reected by the regression coefficient.

Pharmacological test
The predicted components related to the blood-activating activity of XXT were identied with reference standards. The anticoagulant activity of the predicted components of different concentrations (10, 20, and 40 mM) on normal rat plasma was determined by the method described in Section 2.3.3. 33,34 The bioactive components were nally screened out with this in vitro test.

Network pharmacology
Network pharmacology, an application of systems biology, is applied to predict the complex mechanism of action of TCM formulas. 14,35 It could be used to construct an entire drug-target interaction network and to predict the core targets and pathways involved. It is more efficient to clarify how multiple components interact with core targets by intersecting with the disease target database. In this study, the aforementioned screened active components were used as "candidate compounds" for network pharmacological analysis. The interaction network between active components and a blood stasis related target was then presented, and the potential involved pathways were also predicted.
2.6.2 GO function enrichment and KEGG pathway enrichment analysis 2.6.2.1 GO. We enriched the overexpressed GO (gene ontology) terms of the obtained target networks. The gene symbol IDs of the core targets were transformed to gene ensemble IDs by using the Ensembl database (https:// asia.ensembl.org/). Gene ensemble ID les were then uploaded to the OmicShare database (https:// www.omicshare.com/) for GO enrichment analysis.

Results of HPLC ngerprints
3.1.1 Method validation. The t R and PA of common peaks were both calculated to assess the precision, repeatability, and stability. The results demonstrated that the relative standard deviation (RSD) of the precision (t R : 0.26-1.54%; PA: 0.32-1.85%), repeatability (t R : 0.25-1.36%; PA: 0.12-1.76%), and sample stability (t R : 0.07-1.27%; PA: 0.18-1.42%) were all lower than 3%. All results illustrated that the established HPLC analysis method was valid and suitable for sample analysis.
3.1.2 Results of HPLC ngerprints. HPLC ngerprints of 10 batches of S 1 -S 10 samples are shown in Fig. 1. The automatically generated reference ngerprint (R) is shown in Fig. 2.
A total of 41 peaks with good segregation and PA greater than 1 000 000 were selected as the "common peaks" from consecutive peaks. Rutin (Peak 3), identied by comparison with reference standards with t R at 44.4 min, was selected as the reference peak to calculate the relative t R and relative PA of the other 40 common peaks. In accordance with the mixed reference standard HPLC spectrum shown in Fig. 3, 23 characteristic peaks, P1, P3, P4, P11, P12, P14, P18, P20, P22, P24, P25, P26, P27, P29, P32, P33, P15, P34, P35, P36, P37, P38, and P39, were identied as salvianic acid A, rutin, quercetin, ginsenoside Rg 1 , -Re, -Rf, -Rh 1 , -Rg 2 , -Rb 1 , -Rc, -Rb 2 , -Rb 3 , -Rd, -Rg 3 , -F 2 , -Rh 2 , salvianolic acid B, cholic acid, hyodeoxycholic acid, cinobufagin, resibufogenin, tanshinone I, and tanshinone IIA, respectively. The variance coefficients for almost all common peaks were greater than 35%. This is due to the diversity in the contents of constituents contained in different samples extracted with different polar solvents. 3.1.2.2 HCA. To further evaluate the results of the similarity analysis, HCA was performed. As shown in Fig. 4, the S 1 -S 10 samples were obviously separated into two main clusters: S 5 , S 7 , S 13 , S 9 , and S 8 in cluster I, and the other samples in cluster II. Cluster II was further separated into two subgroups: subgroup A (S 6 ) and subgroup B (S 2 , S 4 , S 1 ). The results demonstrated that the chemical components varied from S 1 to S 10 , which was conrmed by the similarity analysis.

Results of blood-activating activity
The effects of S 1 -S 10 on WBV, PV, TT, APTT, PT, and FIB are shown in Table 1.
The levels of WBV (10, 60, and 120/s) and PV of the M group were signicantly increased compared with the N group (p < 0.05, p < 0.01). While the levels of WBV and PV of the S 1 -S 10 groups were notably lower than those of the M group (p < 0.05, p < 0.01). The S 4 and S 7 groups exhibited the best effect on both indexes (p < 0.01). The BCN group could signicantly reduce PV and the low, medium and high share rates of WBV (p < 0.05, p < 0.01).
In the model rats, TT and APTT were signicantly shortened (p < 0.01), the content of PT was prominently decreased (p < 0.01), and the level of FIB was signicantly increased (p < 0.01). TT and APTT were remarkably prolonged, the content of PT was signicantly increased, and the contents of FIB were considerably reduced in the S 1 -S 10 groups as well as in the BCN group (p < 0.05, p < 0.01).
It was demonstrated that different batches of XXT samples exhibited a noticeable blood-activating effect on the acute blood stasis model rats. However, considerable differences in the levels of WBV (10/s) and APTT between different batches of XXT samples were found by the ANOVA analysis, which revealed that the contents of active components in each kind of XXT extract were different to some extent.
The normalized regression coefficient maps of bloodactivation were drawn with the regression equation coefficients (Fig. 5). Except for chromatographic peaks P2 and P6, the other 39 peaks were all negatively related to WBV, indicating that as the levels of compounds represented by these peaks increased, the WBV decreased. Except for peaks P6, P8, P10, P13, P16, and P23, all other peaks were positively related to APTT. In addition, the signicance of the x-variables for the   were considered to be closely related to prolonged APTT (Fig. 5).

Results of verication experiment
The results of the spectrum-effect relationship demonstrated that P1, P3, P11, P15, P22, P34, P36, P38 and P39 might be the blood-activating components in XXT. To further conrm the reliability of the results and determine their dedication to the blood-activating effect of XXT, the anticoagulant effects on  APTT, PT, TT, and FIB of normal rat plasma were then tested. As shown in Fig. 6, salvianic acid A (P1), rutin (P3), ginsenoside Rg 1 (11), ginsenoside Rb 1 (P22), cinobufagin (P36), tanshinone I (P38), and tanshinone IIA (P39) could increase PT, TT, and APTT and reduce FIB at 40 mM and 20 mM (p < 0.05, p < 0.01), respectively, but had no obvious effect on these indexes at 10 mM (p > 0.05). A certain dose dependence appeared. In addition, the positive control drug, aspirin, 19 showed a remarkable effect on PT, TT, APTT, and FIB at 40 mM and 20 mM (p < 0.05, p < 0.01). However, cholic acid (P34) and salvianolic acid B (P15) had no sizeable effect on PT, TT, APTT, or FIB (p > 0.05). The results indicated that salvianic acid A, rutin, ginsenoside Rg 1 , ginsenoside Rb 1 , cinobufagin, tanshinone I, and tanshinone IIA showed a signicant anticoagulant effect on the rat plasma. As a whole, these seven compounds might play vital roles in the blood-activating effect of XXT.

Network pharmacology
In order to discuss the potential pharmacological mechanisms of XXT, network pharmacology was used based on the seven active components.  To understand the complex interactions between the seven active compounds and their corresponding targets at a system level, we constructed a network (Fig. 7) based on the active compounds of XXT and their potential targets. This network contained 97 nodes and 111 edges in total. In the network, each node size was proportional to its degree number. The blue parallelograms represent the seven active components, while the circled dots are the targets. Each link represents the interaction between the compound and the target. The targets of tanshinone IIA are greater than 40. The targets of the other four compounds are all greater than 10. The degree values of a node display the number of routes connected to the node. The degree was ranked as tanshinone IIA, salvianic acid A, rutin, tanshinone I, ginsenoside Rb 1 , ginsenoside Rg 1 and cinobufagin. These targets are composed of eight cytokines, y-two enzymes, eleven Gprotein-coupled receptors (GPCRs), two integrins, one ion channel, three kinases, six nuclear receptors, and seven transporters. This highlights the critical roles of enzymes, kinase, and transporter.
3.5.1.2 The target-target network. The blood stasisassociated target proteins were obtained by searching databases. A Venn diagram of 48 core targets was then obtained by blood stasis-associated target intersecting with the above targets of active compounds (Fig. 8). The target-target network was constructed (Fig. 9) from the STRING database by uploading 48 core targets. In the target-target network, each node represents a protein target. Two proteins are linked if they are targeted by the shared component. In the 48 targets, 46 have at least one link to other targets: that is, they share compounds with other targets. A large highly interconnected network, with 46 nodes and 214 edges, was formed by most targets. The top three targets according to degree were TNF, AKT1, and VEGFA. Combined with the betweenness centrality score, AKT1 was selected as a potential key target protein for the treatment of blood stasis.
3.5.2 GO enrichment and KEGG path enrichment analysis. The gene symbol IDs of the 48 core targets were transformed into 53 gene ensemble IDs with the Ensembl database. The 53 genes were then uploaded to the OmicShare database for Go enrichment analysis. The GO function could be separated into three categories: biological process, molecular function and cellular component. The top three enrichments in the biological process were cellular processes, responses to stimuli, and biological regulation; the top three in the molecular function were binding, molecular function regulation, and catalytic activity; and the top three in the cellular components were cell part, cell, and organelle (Fig. 10).
The 48 core targets were uploaded to the DAVID database to obtain 14 pathways related to blood stasis (p < 0.05), which were then uploaded to the OmicShare database. A bubble map of the KEGG enrichment pathways of the core targets was obtained (Fig. 11). KEGG analysis demonstrated that multiple pathways were remarkably involved in the mechanism of XXT, including cAMP, TNF, HIF-1, NF-kB, PI3K-Akt, sphingolipid, MAPK, VEGF, platelet activation, apoptosis, arachidonic acid, hematopoietic cell lineage, and NOD-like receptor signaling pathway.
Information on pathways and corresponding targets is shown in Table 3. An active compound-target-pathway network map was constructed (Fig. 12). Fig. 7 The active compound-target network. Fig. 8 The Venn diagram of active ingredient targets (A) and blood stasis-associated targets (B). Fig. 9 The target-target network from the STRING database.

Discussion
The present chemical study has established a valid and satisfactory HPLC method to analyze the ngerprints of 10 kinds of XXT extracts with different polarity solvents. The similarity of the ngerprints (correlation coefficients > 0.815) was acceptable for spectrum-effect relationship analysis, and the coefficients of variance of common peaks (>35%) clearly showed that there was diversity in the contents of the constituents of 10 kinds of XXT extracts. In the pharmacological experiment, hemorheology indexes (such as WBV and PV) and coagulation indicators (PT, TT, APTT, and FIB), the important evaluation indexes in estimating blood stasis, 38,39 were selected for pharmacodynamic evaluation. Blood stasis plays a vital role in the information process of multiple disorders and diseases. 40,41 While bloodactivation is the essential strategy to ameliorate blood stasis. The results indicated that different batches of XXT exhibited a considerable blood-activating effect on the acute blood stasis model rats. However, signicant differences in the levels of WBV (10/s) and APTT between different batches of XXT were found, which might be caused by the different contents of bioactive ingredients in each kind of XXT extract. Therefore, it is necessary to perform a spectrum-effect relationship analysis. Both GRA and PLSR, being widely applied for spectrum-effect relationship analysis of TCM, were used to predict the active components. As a result, nine components with potential strong blood-activating activity were analyzed. In the following pharmacological test, seven compounds, tanshinone I, tanshinone IIA, salvianic acid A, ginsenoside Rg 1 , ginsenoside Rb 1 , rutin, and cinobufagin, were then screened out by an in vitro test. Among them, tanshinone I, tanshinone IIA, and salvianic acid A come from Salvia miltiorrhiza; ginsenoside Rg 1 , and ginsenoside Rb 1 come from the total ginsenosides found in ginseng stems and leaves; rutin comes from sophorae os and cinobufagin comes from bufonis venenum. The pharmacological validation was almost the same as the expected results. In previous reports, the activities of tanshinone I, tanshinone IIA, and salvianic acid A mainly referred to anti-cardiac brosis, 42 anti-tumor, 43,44 anti-atherosclerosis, 45 and anti-heart damage effects, and ameliorating endothelial dysfunction, 46 suppressing excessive oxidation damage and cell apoptosis, anti-blood stasis, and reducing intracellular calcium ion overload, 47 possessing anticoagulant, 48 anti-inammatory, 49 reducing nerve cell apoptosis and protecting against ischemia/reperfusion injury 50 and anti-carotid atherosclerosis plaques. 51 The antiplatelet aggregation activity of ginsenoside Rg 1 and Rb 1 has been reported. 52 In addition, they have activities in treating cardiovascular diseases 53 and ischemic stroke, 54 and have anti-inammatory 55 and neuroprotective effects against cerebral ischemia. 56 Rutin had anti-oxidant, 57 anti-carcinogenic, 58 liver protection, 59 anti-inammatory, 60 neuroprotective, 61 and vasoprotective properties. 62 Cinobufagin inhibited PC3 cell growth 63 and had antibrosis properties. 64 The investigation in the present study showed that "spectrum-effect relationship analysis" is an effective method for looking for the active components in TCM.  To make a further primary prediction of the multiple mechanisms of the effects of XXT, network pharmacology was used to construct an entire drug-target-pathway interaction network in the last part of the present study. The seven screened active components were used as "candidate compounds" for the analysis. The results showed that the main kinds of targets involved in the blood-activating effect were enzymes, kinase, and transporter. The top three targets according to degree, TNF, AKT1, and VEGFA, were considered as potential key target proteins for the treatment of blood stasis in the present study. Moreover, the TNF, NF-kB, and PI3K-Akt signaling pathways were also predicted to be involved in the anti-blood stasis effect of XXT. In previous reports, AKT1, one of three closely related serine/threonineprotein kinases (AKT1, AKT2, and AKT3), is activated reliant on the P13K pathway and is a key node in the signal network, which controls many aspects regarding cell metabolism, proliferation, apoptosis, angiogenesis and anti-hematologic malignancies. [65][66][67] TNF (tumor necrosis factor), an inammatory cytokine (including TNF-a and TNF-b), mainly secreted by macrophages and T lymphocytes during acute inammation, acts as an inammatory mediator by preventing endotoxin shock, and has anti-tumor effects and anti-infective effects by inhibiting viral replication and killing virus-infected cells/tumor cells. 68 VEGFA, a key vascular endothelial growth factor to promote angiogenesis and axonal regeneration, 69 promotes endothelial cell proliferation, induces cell migration, induces permeabilization of blood vessels, and suppresses apoptosis. 70 Although further experimental validations of the network pharmacology prediction are required in future studies, utilizing network pharmacology in TCM research is an efficient method to elucidate how multiple components interact with multiple targets, which is consistent with the theory of TCM synergy.

Conclusions
In the present study, we quickly and accurately screened seven active components from Xueshuan Xinmaining Tablet using a comprehensive evaluation system combining chemical analysis, biological activity evaluation and mechanism study. Firstly, the spectrum-effect relationship was established by linking the peaks in the ngerprints of XXT with the blood-activating bioactivity to screen nine potential active components in XXT. Then, pharmacological tests veried seven of the nine compounds, salvianic acid A (P1), rutin (P3), ginsenoside Rg 1 (P11), ginsenoside Rb 1 (P22), cinobufagin (P36), tanshinone I (P38), and tanshinone IIA (P39), and had an anticoagulant effect in vitro. Finally, the putative mechanism of these seven components was further discussed using network pharmacology analysis, and three signal pathways (the TNF, NF-kB, and PI3K-Akt signaling pathways) that might be related to antiblood stasis were found. The study provided a new approach and ideas for the exploration of the active components of TCM.

Author contributions
Conceptualization: Jing Tan and Jinping Liu; data curation: Hongqiang Lin, Zhongyao Wang, Hanrui Si and Yutong Zhang; Table 3 The KEGG pathways and corresponding targets in XXT

Conflicts of interest
The authors declare that there are no conicts of interest.