Herb pair Danggui-Honghua: mechanisms underlying blood stasis syndrome by system pharmacology approach

Herb pair Danggui-Honghua has been frequently used for treatment of blood stasis syndrome (BSS) in China, one of the most common clinical pathological syndromes in traditional Chinese medicine (TCM). However, its therapeutic mechanism has not been clearly elucidated. In the present study, a feasible system pharmacology model based on chemical, pharmacokinetic and pharmacological data was developed via network construction approach to clarify the mechanisms of this herb pair. Thirty-one active ingredients of Danggui-Honghua possessing favorable pharmacokinetic profiles and biological activities were selected, interacting with 42 BSS-related targets to provide potential synergistic therapeutic actions. Systematic analysis of the constructed networks revealed that these targets such as HMOX1, NOS2, NOS3, HIF1A and PTGS2 were mainly involved in TNF signaling pathway, HIF-1 signaling pathway, estrogen signaling pathway and neurotrophin signaling pathway. The contribution index of every active ingredient also indicated six compounds, including hydroxysafflor yellow A, safflor yellow A, safflor yellow B, Z-ligustilide, ferulic acid, and Z-butylidenephthalide, as the principal components of this herb pair. These results successfully explained the polypharmcological mechanisms underlying the efficiency of Danggui-Honghua for BSS treatment, and also probed into the potential novel therapeutic strategies for BSS in TCM.


Results
The molecular mechanisms of the herb pair Danggui-Honghua against BSS were investigated by a network construction approach with system pharmacology model based on chemical, pharmacokinetic and pharmacological data ( Fig. 1). All of the ingredients from this herb pair were explored to build a compound library. Next, the oral bioavailability (OB), Caco-2 cell permeability and drug-likeness (DL) of the ingredients were collected and calculated to screen the potential active compounds. Then the potential targets and their corresponding diseases and pathways of the active compounds were data-mined from literature and public database/software sources. Subsequently, the acquired pharmacological data were integrated into the compound-target (C-T), target-disease (T-D), and target-pathway (T-P) networks, respectively. Finally, a contribution index of every active ingredient based on network based efficacy weighted by literature was calculated. Based on the above analyses, the underlying mechanisms of Danggui-Honghua for treating BSS were explored.

Ingredient comparisons in Danggui and Honghua. The ingredients in Danggui and Honghua were
retrieved from Traditional Chinese Medicine Systems Pharmacology Database and Analysis Platform (TcmSP ™ , http://sm.nwsuaf.edu.cn/lsp/tcmsp.php) 12 and were manually supplemented. The main components of Danggui are phthalides and organic acids 13 , whereas the major ingredients in Honghua are quinochalcone C-glycosides and flavonoid glycosides possessing one or more glucose units 4 . Since those glycosides in this herb pair might be deglycosylated by the glycosidase in the intestinal tract, 11 aglycones were also incorporated into the compound library labeled by_qt. Thus, a total of 347 ingredients were retrieved for Danggui (143) and Honghua (204). The detailed information about these molecules was provided in Supplementary Table S1.
The above analysis suggested that the constituents of Danggui and Honghua were diverse but the majority of them satisfied the Lipinski's rule of five. Apart from the hydrotropic property, Danggui was significantly different from Honghua in other properties of the chemical components. These differences are not surprising due to the distinct chemo-physical properties of the ingredients from these two herbs. The above results also showed that the constituents from Danggui have better pharmacokinetic properties (OB and Caco-2), whereas the ingredients from Honghua possess better drug-likeness (DL). From a chemotaxonomic point of view, phthalides from Danggui and quinochalcone C-glycosides from Honghua have only been found in the families Apiaceae and Asteraceae, respectively. Although these major ingredients between Danggui and Honghua are obviously different, the two herbs exhibit the identical blood-activating and stasis-dissolving effects and the warm nature of TCM, which may also clarify why Danggui-Honghua could produce synergistic and complementary effects.
Active ingredients in Danggui-Honghua. Although a single herb or TCM formula usually contains a considerable number of bioactive components, maybe only a few with desirable pharmacodynamic and pharmacokinetic properties are responsible for its therapeutic effects. In the present work, three crucial ADME (absorption, distribution, metabolism, and excretion) parameters, including OB, Caco-2, and DL, were employed to screen most of the active compounds from Danggui-Honghua. A few active compounds that do not meet all of these three criteria were also selected for their high amounts and high bioactivities. Consequently, a total of 31 active compounds were selected from the 347 compounds of this herb pair (Table 1).
Active ingredients from Danggui. By ADME screening, 15 out of 143 ingredients with excellent pharmacological effects were extracted from Danggui, and the majority of them possess satisfactory pharmacokinetic profiles. For instance, caffeic acid (DG-14, OB = 54.97%, Caco-2 = 0.21 and DL = 0.19) has antioxidant, anti-inflammatory, antimutagenic, antibacterial and anti-carcinogenic effects 14 . Likewise, β-sitosterol (DG-15, OB = 36.91%, Caco-2 = 1.33 and DL = 0.75) exhibits potent anti-inflammatory and antipyretic activities 15 . It should be point out that phthalides, the representative ingredients in Danggui, show low DL values, but they exhibit potent antifungal, antibacterial, anti-inflammatory, and antioxidant activities 16 . Similarly, although phenolic constituents hold low DL values, they also exhibit remarkable pharmacological effects 17,18 . Specifically, Z-ligustilide (DG-1) and ferulic acid (DG-12) have been chosen as the marker components for quality control of Danggui in Chinese Pharmacopoeia 19 . In view of the facts mentioned above, phthalides and phenolic constituents were also deemed as the active ingredients for further analysis. Furthermore, the contents of the selected constituents in Danggui were also considered. Through a wide-scale text mining of Google Scholar, the total content of ferulic acid, Z-butylidenephthalide (DG-5), senkyunolide A (DG-2), senkyunolide I (DG-3), and Z-ligustilide was found to be nearly 43 mg/g 20 , and nicotinic acid (DG-8) was up to 0.198 mg/g 21 . Noteworthy, coniferyl ferulate (DG-9) is also abundant in Danggui and exhibits multiple biological activities such as antioxidant, vasodilating and antibacterial effects 18 . Nodakenin (DG-11) has been reported to possess neuroprotective, anti-inflammatory, antibacterial, and memory-improving effects 22 . Based on the above considerations, it was reasonable to believe that fifteen compounds could be listed as potential active ingredients for Danggui (Table 1) 26 . Thus, quinochalcone C-glycosides were also selected for targeting. Notably, scutellarin (HH-12, OB = 2.64%, Caco-2 = − 1.08 and DL = 0.79) has been used in tablet and injection forms in China since 1984 for treatment of acute cerebral infarction and paralysis induced by hypertension, cerebral thrombosis, and cerebral hemorrhage 27 . Scutellarein (HH-13), the aglycone of scutellarin which is a candidate drug for cardio-cerebrovascular diseases was also selected for our further analysis. In addition, it is necessary to incorporate nicotiflorin (HH-8, OB = 3.64%, Caco-2 = − 1.77 and DL = 0.73) into further investigation for that this compound has potent neuroprotective, analgesic, anti-hypertensive and anti-anaphylactic effects 28 .
Target proteins of Danggui-Honghua. Searching for the targets of candidate drugs solely by the experimental approaches is overspending, labor-intensive, and time-consuming. In the present work, an integrated in silico approach was introduced to identify the target proteins for the active ingredients of Danggui-Honghua. Predictive models were used including Similarity Ensemble Approach (SEA, http://sea.bkslab.org/) 29 34 ; PTGS1 and PTGS2 contribute to atherosclerosis and thrombosis by regulating the production of eicosanoids that modulate physiological processes in the vessel wall 35 ; JUN modulates smooth muscle cell proliferation in response to vascular angioplasty 36 ; additionally, ICAM1 mediates the adhesion of neutrophils and monocytes to vascular endothelium 37 . Specifically, Z-ligustilide, senkyunolide I, ferulic acid and coniferyl ferulate may mediate NOS2 or NOS3 to increase the nitric oxide biosynthesis, thereby possibly exerting blood-vessel dilation, neuronal signal transmission, coordination of heart rhythm and regulation of cellular respiration activities 38 . Vanillin may interact with four potential targets including PTGS2, JUN, MMP9 and MAPK1 which are also relevant to nervous system and vascular diseases 39 . Except for vanillin, nicotinic acid and β-sitosterol, other active ingredients from Danggui also interacted with the targets related to inflammation, abdominal pain, dysmenorrhea, embolism, thrombosis and ischemia. Six potential targets including F2, F7, F10, F11, TBXA2R and MAPK14 are relevant to disorders of thrombosis, which could clarify why Danggui exhibits strong anticoagulant function 16 . Additionally, there are also several targets such as PPARG, SOD1, RELA and GSK3B are involved in the ischemic and inflammation processes. It should  be pointed out that some major targets, such as PTGS2, NOS3 and CHRM2, are also closely concerned with the pain-related diseases, which may contribute to the blood-activating and stasis-dissolving effects of Danggui.
Target proteins of Honghua. For Honghua, by target fishing, 16 active ingredients were validated to bind with 36 target proteins related to BSS. There are also plenty targets mediated by the active ingredients from Honghua that are involved in vascular and central neural systems. For example, HSYA may have the potential to act on 18 targets including HIF1A, VEGFA, HMOX1, PTGS2, CASP3, CASP9, AGTR1, PTAFR, and GSK3B. Actually, HSYA has been identified as an inhibitor of HIF1A, which might contribute to its therapeutic application in vascular diseases 40 . Beyond that, it also has a strong antagonistic effect on the PTAFR protein, which may explain its function in both inflammatory and neuropathic pain responses 41 . Analogously, scutellarin exhibits strong neuroprotective effect because it may interact with STAT1, NOS2, NOS3, VEGFA and FGFR1 simultaneously 42 . Inflammatory response is of enormous significance in BSS. Six active compounds of Honghua, including 6-hydroxykaempferol (HH-6), eriodictyol (HH-11), rosmarinic acid (HH-16), kaempferol, quercetin, and luteolin, may have interactions with PPARG which is expected to control inflammation associated with gut, myocardial, lung and cerebral ischemia 43 . Other five active ingredients, HSYA, rutin (HH-10), safflor yellow A (HH-2), safflor yellow B and eriodictyol, were identified to interact with PTGS1, which is a potential target for the next generation of anti-inflammatory drugs 44 .
Of note, we have implemented molecular docking combined with the in vitro experiments to explore the structure-activity relationships of the active ingredients among Si-Wu- Tang 46 . It could be concluded that the predicted targets were in agreement with our previous experiments. Therefore, the integrated in silico approach should be feasible and convincible to explore the compound-target interactions of Danggui-Honghua.

Target and contribution index analysis to decipher the combination rule of Danggui-Honghua.
To facilitate the visualization and interpretation of the complex relationships between all active ingredients of Danggui-Honghua and their targets, a bipartite graph of C-T network was constructed (Fig. 3). Two characteristics of Danggui-Honghua were observed from network analysis: (1) the promiscuous properties of its active ingredients and (2) the existence of highly interconnected compounds. The average number of potential targets per active ingredient was 6.5. And all active ingredients in this herb pair were potential multiple-kinase inhibitors or activators. Amongst them, those ones with high interconnection degrees were responsible for the high interconnectedness of the C-T network, especially HSYA (degree = 18), Z-ligustilide (degree = 15), quercetin (degree = 12), luteolin (degree = 11), ferulic acid (degree = 11) and Z-butylidenephthalide (degree = 10). From the topological features of this network and the functional properties of the proteins (Supplementary Table S2 and Fig. S1A), there were different modes of actions between compounds and targets in Danggui-Honghua. As shown in the C-T network (Fig. 3), the efficacy of this herb pair not only concentrated on modulating the crucial targets involving in the vascular and central neural systems (ESR1, ICAM1, HMOX1 and NOS3), but also, more essentially, focused on the regulation of the other proteins mediating inflammation, thrombosis, ischemia, dysmenorrhea and abdominal pain (F7, TBXA2R, PTGS2, CHRM2, and NOS2) to relieve the pathological changes and prolong the efficient curing process. For example, in animal models, increased expression of HMOX1 has been shown to protect tissues and cells against ischemia-reperfusion injury, oxidative stress, inflammation and hypoxia-induced vascular stasis 47 . The protein HMOX1 was found to have interactions with 9 ingredients, and senkyunolide A, kaempferol and eriodictyol synergistically increase HMOX1 expression [48][49][50] . In addition, the neurotoxic proinflammatory mediator PTGS2 was also connected with several active ingredients. Especially, HSYA, Z-ligustilide and ferulic acid were confirmed to enhance anti-inflammatory effects via significantly attenuating the expression level of PTGS2 4,51,52 .
The T-D network was further constructed based on all the targets and their corresponding diseases. As shown in the graphical network (Fig. 4), both Danggui and Honghua could regulate the proteins related to vascular and central neural system diseases as well as inflammation and pain. The target scope of Honghua was broader than that of Danggui. Several targets mediated by the active ingredients of Honghua including PTAFR, PTGS2, NOS3 and STAT1, may alleviate the accompanying symptoms of BSS such as inflammation and pain. As to Danggui, its active ingredients mediated several targets including ADRB1, CHRM2, GSK3B, BCL2 and ICAM1, which may be helpful for improving the immune system. Modern research has demonstrated that ischemic processes release mediators activating the innate immune system that may be induced by the formation of thrombi inside blood vessels 53 . Fortuitously, several targets associated with cancer were also discerned in the T-D network. HSYA from Honghua was validated to antagonize tumor angiogenesis by inhibiting the protein expression of VEGFA, MMP9 and HIF1A 4 . Additionally, Z-butylidenephthalide, senkyunolide A and Z-ligustilide from Danggui were confirmed to exhibit anti-proliferative potential and significant synergy on colon cancer cells 54 .
As mentioned earlier, a contribution index of every active ingredient was proposed based on network based efficacy weighted by literature. According to calculated results ( Fig. 5  ferulic acid, and Z-butylidenephthalide. They displayed the most contribution to the blood-activating and stasis-dissolving effects of Danggui-Honghua with a sum of CIs of 88.13%. Therefore, the above discussion may fully clarify why Danggui-Honghua could produce synergistic and complementary effects. Pathway analysis to explore the underlying mechanisms of Danggui-Honghua. Signaling pathways, as an important component of the system pharmacology, link receptor-ligand interactions to pharmacodynamics outputs 55 . The canonical pathways associated with BSS treatment and prophylaxis were extracted from Kyoto Encyclopedia of Genes and Genomes (KEGG, http://www.genome.jp/kegg/) database, which ends up with 20 KEGG pathways, including cAMP signaling pathway, calcium signaling pathway, NF-κ B signaling pathway and sphingolipid signaling pathway (Fig. 6). The NF-κ B signaling pathway is crucial for focal cerebral ischemia/reperfusion induced inflammatory injury 56 . Previous research has shown that HSYA, safflor yellow A, and caffeic acid could inhibit the NF-κ B signaling pathway contributing to the cross-talk of multiple targets in anti-inflammation 57 . The sphingolipid signaling pathway was proven to play a critical role in the ischemic preconditioning and the pathophysiology of stroke 58 .
For the purpose of systematically dissecting the underlying mechanisms of Danggui-Honghua, all of the targets interacting with the active ingredients were mapped onto the 20 KEGG pathways and the T-P network was generated (Fig. 7). The TNF signaling pathway exhibited the highest number of target connections (degree = 11), followed by HIF-1 signaling pathway with 9 targets, estrogen signaling pathway and neurotrophin signaling pathway with 8 ones, respectively. These high-degree pathways were closely related to the vascular and central neural systems and inflammation. The TNF signaling pathway plays an important role in the ischemic stroke and the vascular injury involved in multiple targets including JUN, PTGS2 and ICAM1 59,60 . The HIF-1 signaling pathway was an underlying mechanism of neuroprotection and anti-ventricular cell apoptosis 61,62 . Fortunately, 22 out of 31 active compounds from Danggui-Honghua, especially HSYA, quercetin, Z-ligustilide and luteolin, were implicated in regulating the major targets of HIF-1 signaling pathway, such as HMOX1, NOS2, NOS3, RELA, PIK3CG and MAPK1. Recently, we successfully discovered the underlying blood-activating mechanisms of Danggui-Honghua by metabolomics analysis. Four potential metabolic pathways were speculated, including phenylalanine metabolism, sphingolipid metabolism, arachidonic acid metabolism, and arginine and proline metabolism 63 . In the present study, arginine and proline metabolism, arachidonic acid metabolism, and sphingolipid signaling pathway (including sphingolipid metabolism) were also deciphered, which were consistent with our previous metabolomics results 63 . Overall, it could be speculated that the herb pair Danggui-Honghua exert the blood-activating and stasis-dissolving effects mainly through the regulation of TNF, HIF-1, estrogen, and neurotrophin signaling pathways. As a holistic medicine, this herb pair may also be implicated in arginine and proline metabolism, arachidonic acid metabolism, VEGF signaling pathway, MAPK signaling pathway, calcium signaling pathway, and sphingolipid signaling pathway to regulate the vascular and nervous systems, as well as the inflammation and pain.

Discussion
Herb pairs, the smallest compatible units in TCM formulae, have become a prominent concern during the past decade 2 . Danggui and Honghua, combined as a classical herb pair, have been frequently used in TCM prescriptions 5 . Through data mining and pharmacological approaches, this herb pair was found to produce synergistic and complementary effects to treat BSS [64][65][66][67][68][69] . However, the complexity of the chemical components of this herb pair and their corresponding multiple targets in vivo led to extreme difficulty to elucidate the molecular mechanisms.
Nowadays, system pharmacology provides a powerful avenue for compatible and mechanistic exploration of TCM 10,11 . In our work, an integrated system pharmacology approach, combined a number of network-based computational methods and algorithm-based approaches, was used to select active compounds, predict targets, construct networks, and illuminate the molecular synergy of Danggui-Honghua on BSS. Thirty-one active ingredients with favorable bioactivities and contents were selected from the 347 compounds of Danggui-Honghua by ADME filtering, providing foundational clues for thorough investigation on this herb pair. It was found that some biological activities against BSS of these active ingredients have been reported previously [26][27][28][39][40][41][42][70][71][72] , highlighting the credibility of our ADME filtering criteria. Then, an integrated in silico approach was applied to decipher the 42 targets for these active ingredients related to BSS including 27 common targets, which distinctly explained the action modes and biological processes that active ingredients achieve their synergistic and complementary curative effects. It is worth mentioning that the results from several compound-target interaction experiments by us and other groups indicated the reasonability of our integrated in silico approach [45][46][47][48][49][50][51]73 . Subsequently, 20 signaling pathways associated with BSS treatment and prophylaxis by Danggui-Honghua were retrieved. A recent study on the anti-inflammatory effects of volatile oils from Danggui (78.61% Z-Ligustilide and 7.99% Z-Butylidenephthalide) using GC-MS-based metabolomics revealed that it may be involved in regulating the arachidonic acid metabolism 74 , which is comparable to our findings. Finally, the C-T, T-D and T-P networks clearly elucidated the molecular synergistic actions of Danggui-Honghua in a holistic context. By network systematic analysis and contribution index calculation, HSYA, safflor yellow A and safflor yellow B in Honghua, together with Z-ligustilide, ferulic acid and Z-butylidenephthalide in Danggui, displayed the most contribution to the blood-activating and stasis-dissolving effects of Danggui-Honghua. Meanwhile, this herb pair could regulate the proteins related to vascular and central neural system diseases as well as inflammation and pain implicated in TNF, HIF-1, estrogen, and neurotrophin signaling pathways, arginine and proline metabolism, arachidonic acid metabolism, and VEGF signaling pathway. Noteworthy, based on our research, several targets and signaling pathways of Danggui-Honghua acting on BSS have been found for the first time. All of these results are expected to help identify novel curative efficacy and take full clinical advantage of Danggui-Honghua.
Due to the above findings mainly relied on theoretical analyses, more experiments are anticipated to support these findings as well as potential clinical applications. It should be noted that the majority of the ingredients in both Danggui and Honghua were hydrotropic, and the OB values of a third of the selected 31 active ingredients were less than 30%. Therefore, the availability of these active constituents by gut microbiota especially under the disease state may be a critical step towards the emergence of their bioactivities in vivo 75 .

Methods
Chemical ingredients database building. All of the constituent data of Danggui (the radix of A. sinensis) and Honghua (the florets of C. tinctorius) were retrieved from TcmSP ™ , a unique system pharmacology platform designed for herbal medicines 12 , and then manually supplemented through a wide-scale text-mining method.
Meanwhile, four important pharmacology-related properties were also obtained from TcmSP ™ , including MW, MLogP, nHDon and nHAcc.  Active ingredients screening. The active ingredients from Danggui-Honghua were filtered by integrating three indexes including OB, Caco-2, and DL. A robust in silico model OBioavail 1.1 (Yangling, Shaanxi, China) that integrated the metabolism and transport information was employed to calculate the OB values of all herbal ingredients 76 . Those ingredients with OB ≥ 30% were selected. The VolSurf built-in Caco-2 permeability model (Tripos, St. Louis, America) 77 was implemented to screen active compounds. Given that molecule with Caco-2 value less than − 0.4 is not permeable, the threshold of Caco-2 permeability was set to − 0.4. Database-dependent DL evaluation approach based on Tanimoto coefficient 78 was applied and shown as follows.
In this eq. (1), A represents the molecular descriptors of herbal compounds, and B displays the average molecular properties of all compounds in Drugbank. Those ingredients with DL ≥ 0.18 were preserved. The ingredients were adopted as the candidate compounds for further analysis when they met all of these three criteria.
Targets fishing. To identify the corresponding targets of the active ingredients of Danggui-Honghua, several approaches combined with chemometric method, information integration and data-mining were implemented. First of all, the biological targets of active ingredients were obtained from SEA 29 , STITCH 30 , and PharmMapper sever 31 . Amongst them, the PharmMapper gives the best mapping poses by comparing with available targets in PharmTargetDB (a large, in-house repertoire of pharmacophore database belongs to PharmMapper) and the respective N-best fit poses are generated. All active compounds were also sent to HIT 32 , TTD 33 , DrugBank and Google Scholar to mine compound-target interactions supported by literature. Then, to better dissect the role of Danggui-Honghua in BSS treatment, all targets obtained from the previous two steps were sent to TTD, Comparative Toxicogenomics Database (CTD, http://ctdbase.org/) and PharmGKB (http://www.pharmgkb.org) 79 to mine target-related diseases. Finally, those targets which were implicated in the aforementioned pathophysiology and clinical manifestations of BSS were retained, and the others were eliminated. Networks construction. Three networks were constructed: (1) Compound-target network (C-T network).
Active ingredients of Danggui-Honghua and their corresponding targets were employed to generate the C-T network.
(2) Target-disease network (T-D network). All targets and their corresponding diseases were employed to build a bipartite graph of T-D network. (3) Target-pathway network (T-P network). The pathway information of targets were extracted from the database of KEGG, and then a bipartite T-P network composed of targets and their corresponding putative pathways was built. All visualized network graphs were constructed by Cytoscape 3.2.1 (http://www.cytoscape.org/), an open software package project for visualizing, integrating, modeling and analyzing the interaction networks 80 .

Contribution indexes calculation.
In order to estimate the contribution of each active ingredient to the blood-activating and stasis-dissolving effects of Danggui-Honghua, a contribution index (CI) based on network based efficacy (NE) weighted by literature was proposed and calculated by eqs (2) and (3): where n is the number of targets associated with ingredient j; d i is the degree of target i associated with ingredient j; c i is the number of BSS-related literature of ingredient i; m is the number of ingredients.
If the sum of CIs for the top N ingredients was more than 85%, these relevant N ingredients were considered to contribute the most to the blood-activating and stasis-dissolving effects.