Network Pharmacology and Component Analysis Integrated Study to Unveil Molecular Mechanisms of A Traditional Chinese Medicine Decoction (Polygonum Multiorum, Rehmannia Glutinosa, Senna Obtusifolia and Crataegus) in Hypertension Treatment

Background: Traditional Chinese Medicines (TCM) are known for their curative effects on hypertension through a holistic approach. The molecular mechanisms of the formulation comprising Polygonum multiorum, Rehmannia glutinosa, Senna obtusifolia and Crataegus, used by Chinese practitioners in ameliorating hypertension, however remain a mystery. This initial study is thus aimed at unveiling the molecular mechanisms of this TCM formulation in treating hypertension. Methods: The methanolic extract compounds of the decoction were identied through Liquid chromatography mass spectrometry-mass spectrometry (LC-MS/MS). Oral bioavailability and drug likeness were then measured to lter out identied compounds. Several databases, such as the SwissTargetPrediction, STRING, OMIM and KEGG, were used to retrieve information on the predicted targets for the purpose of developing a network using Cytoscape Version 3.8. Enrichment analysis was then performed to elucidate the mechanisms of the decoction in hypertension mitigation. Results: A total of 11 compounds identied were revealed to possess bioavailable and drug like characteristics, based on the Veber and Quantitative Estimation of Drug-likeness (QED) parameters. Pathway analysis showed enrichment of pathways such as cardiac muscle contraction, uid shear stress and atherosclerosis, dilated cardiomyopathy, renin-angiotensin system and hypertrophic cardiomyopathy (HCM), which are all strongly associated with hypertension. Conclusion: The network pharmacology analysis clearly shows that this TCM decoction ameliorates hypertension through several indirect pathways where most of the targets are involved in HCM, which is caused by hypertension.


Introduction
In the category of noncommunicable diseases, cardiovascular disease is the foremost cause of mortality worldwide, accounting for approximately 17.7 million deaths in 2017, more than three quarters of which were from low-income and middle-income countries [1]. The severity of cardiovascular disease also causes economic burdens resulting in it being categorized as the most costly disease among noncommunicable diseases, with costs predicted to increase to $368 billion by 2035 [2]. One of the leading causative factors of cardiovascular disease, with its strong association established by numerous studies, is hypertension [3,4]. One previous report projected that 1.56 billion adults may develop hypertension by 2025 [5]. Due to substantial evidence of its role as a silent killer, signi cant efforts have expended in the identi cation of critical hypertension levels and ideas for treatments to prevent or halt atherosclerosis and further development of cardiovascular disease [6]. Western medicine is currently the most common treatment for hypertensive conditions but it is limited by availability, cost and adverse effects [7]. The e cacy and limitation of western drugs in treating such conditions remain unsatisfactory thus it is essential that alternative approaches be considered and explored.
Due to the bottlenecks in drug discovery focus has been shifted from one target to network targets, or more commonly known as multitargets, to derive solutions [8]. In tackling the complexity of the multitarget mechanism, the rapid growth in bioinformatics provides a powerful platform for studying network-based drug discoveries, which is a more cost-effective drug development approach [9]. Bioinformatics can be used to mine and comment on the biological information of disease networks [10]. Further, through utilization of one of its powerful tools, network pharmacology has been widely used to reveal the relationships between compound-targets and correlate it to disease gene targets and pathways, via a sea of data in databases [11]. Data from network pharmacological analysis, including underlying pathways, genes and protein-protein interaction and integration of these data, can ultimately lead to the formulation of mathematical models to decipher the system biology of humans, in its response to drugs' e cacies [9]. The rst concept or pioneer of network pharmacological analysis was developed for unraveling the effects of traditional Chinese medicine on system biology [8].
Traditional Chinese Medicine (TCM) is well known for the use of formulations comprising multiple herbs to treat numerous diseases through a holistic view and multitarget approach [12], concomitant with the pivotal concept of new modes of drug discovery. A wide variety of cases have been studied and randomized trials performed using TCM as treatment for hypertension [13,14] where some have shown curative effects [12]. However, the establishment of evidence for TCM's effects on hypertension is still in progress as the unravelling of the molecular mechanisms of TCM remains limited and vague. There are still a lot of uncertainties about the curative effects of TCM due to its complex multitarget properties in the human system biology. This hampers the development of novel TCM [9] cures, particularly those involving prescriptions which have been passed down from generation to generation verbally by Chinese practitioners.
One such prescription recommended by Chinese practitioners in East Malaysia involves the use of a formulation comprising the combination of Polygonum multi orum, Rehmannia glutinosa, Senna obtusifolia and Crataegus, as medicine for reducing or preventing hypertension. These plants are known for their anti-hyperlipidaemia, hypodynamia and anti-ageing properties [15,16]. They are also high in antioxidants and are antigenotoxic [17], besides having cardiotonic and cardioprotective characteristics [18]. However, identi cation of the bioactive compounds and mechanisms of action of this decoction remains undetermined. To ascertain the e cacy of this decoction from a cost-effective perspective, we deciphered its molecular mechanisms in the human system biology through liquid chromatography mass spectrometry-mass spectrometry (LC/MS-MS) screening and the bioinformatic approach, via network pharmacological analysis. The extracts were dissolved in 2 ml of methanol and ltered using 0.22 µm hydrophobic PTFE prior to LC/MS-MS analysis using the Agilent 6520 Accurate Mass Q-TOF LC/MS system. Reverse-phase liquid chromatography was used, with a multistep gradient consisting of Solvent A (water HPLC grade + 0.1% formic acid HPLC grade) and Solvent B (acetonitrile HPLC grade + 0.1% formic acid HPLC grade) at a ow rate of 0.2 mL/min. Data acquisition was performed using positive and negative ionization modes (separate runs) for both pro ling (MS1) and fragmentation (MS2), with a detection range of between 100 to 1500 m/z.

Identi cation of compounds from databases
The raw data obtained were processed using MZmine 2.53, and only the MS1 features (associated with their respective MS2 spectra) that appeared in at least 2 of the 3 biological replicates were retained and converted into MSP formatted les, using a custom PERL script. MSFinder 3.26 was used for both molecular formula prediction and compound annotation of the MSP (MSFinder) les. 4.0 Oral Bioavalability (OB) and Drug Likeness (DL) screening and establishment of a compound-target network The identi ed compounds' oral bioavailability (OB) and drug likeness (DL) were analyzed and calculated using ADMETlab [19] and DruLiTo, based on Veber rule [20] and Quantitative Estimate of Drug likeness (QED). The bioactive compounds were ltered based on OB≥50% and DL=pass criteria. Predicted target interaction with compounds were obtained through the SwissTargetPrediction database (http://www.swisstargetprediction.ch/) based on chemical similarities [21] and the relationships of a compound target's (CT) visual interaction network was produced using Cytoscape program Version 3.8 [22]. All of the genes' names were standardized using the UniProt database (http://www.uniprot.org/).

Establishment of a hypertension gene network
Genes related to hypertension or cardiovascular disease were downloaded and identi ed via the Online Mendelian Inheritance in Man, OMIM (https://omim.org/) database and the human gene database, GeneCards (http://www.genecards.org/), using the keywords 'hypertension' and 'cardiovascular disease related genes of Homo sapiens' , while other species were excluded [23,24]. The identi ed hypertension or cardiovascular disease genes were then used to construct a protein-protein interaction (PPI) network with the compound target genes (from section 4.0). 6.0 Construction of protein-protein interaction (PPI) network PPI analysis was performed using STRING 11.0 (https://string-db.org/) database and generated in accordance with systematic co-expression analysis, detection of shared selective signals across genomes and automated textmining of the scienti c literature [25]. It was imported to Cytoscape Version 3.8 for network analysis and data visualization through STRINGApp [26], using the con dence score of 0.7 for higher con dence in the interaction.

Gene Ontology (GO) functional annotation and KEGG pathway analysis
The mechanisms of action of the decoction in hypertension were deciphered by analyzing the primary pharmacological units through Gene Ontology (GO) (http://geneontology.org/) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway (https://www.genome.jp/kegg/pathway.html) [27,28] analysis. Only biological processes and pathways with p-values <0.05 were considered as signi cant in this enrichment analysis. The data were analyzed using R version 3.6.3 [29].
Results ADME pro le of the screened components from Polygonum multi orum, Rehmannia glutinosa, Senna obtusifolia and Crataegus A total of 51 compounds, were identi ed from MSFinder after performing LC-MS/MS (Additional File 1). The absorption, distribution, metabolism and excretion (ADME) properties of the metabolites of the selected plants' compounds were analyzed as ADME properties are pivotal keys in bioactivities. OB limits F (20%) and F (30%) indicators were used to lter the chemical constituents of the plant samples that displayed favourable pharmacokinetic properties. The ADME-related parameters of OB and DL were selected to screen for potential bioactive compounds in the TCM decoction. This is because only bioactive compounds can contribute to therapeutic effects [30]. The values of OB ≥50% and DL=pass were considered as being indicative of good absorption after oral administration and thus, demonstrative of chemical suitability for drug development. The results indicated that only 11 compounds met the OB ≥50% and DL=pass criteria, as shown in Table 1.

CT visual interaction network
To visualize the relationships of the medicinal plants' compounds and targets, the plant-compound-target network of the decoction was constructed using Cytoscape Version 3.8, as shown in Fig. 1. Only 11 compounds from the four medicinal plants were identi ed to be suitable as bioactive compounds, as attested by their OB and DL criteria. There were also duplicate compounds detected in the different plants used for the decoction through network visualization, as illustrated in Fig. 2. For instance, D-glucose was found in Crataegus, Senna obtusifolia and Polygonum multi orum, while pyroglutamic acid was found in all the plants except for crataegus, as shown in Table 1. The 11 selected compounds were found to exhibit activity on 124 targets, as illustrated in Fig. 1 and Fig.   2. Some of the compounds were also found to have the same targets, as shown in Fig. 2. For example, ADORA3 is the target gene for purine, camazepam, L-isoleucine and L-valine. In the meantime, only 1-(dimethylamino) propan-2-ol and L-lactic acid targets were found to be unique, which do not share an edge with other compounds. On other hand, PLG, IGF1R, REN, F2, PDE3A, TSPO, CDK1, CDK5 and ADORA2B were targets that were shared by two different compounds.

PPI Network
The PPI analysis results from STRING, after input of all the targets, revealed diversi ed interactions in the PPI network, as depicted in Fig. 3. The illustrated gure shows that viable protein target nodes (n=164) were linked to edges (n=497) at an average node degree of 6.06 and average local clustering coe cient of 0.591. The con dence level of 0.7 was selected as the high evidence level from STRING to indicate strong interactions.
GO and KEGG pathway enrichment analysis GO and KEGG pathway enrichment analysis were carried out to determine the characteristics of the compoundrelated and disease-related targets. GO analysis revealed that most of the existing potential targets were enriched in the regulation of biological processes, regulation of biological quality, signal transduction, response to chemicals, regulation of cell communications and response to stress ( Fig. 4 (a)). According to the GO enrichment analysis, the top 30 enriched hypertension-related biological processes included regulation of heart contraction, regulation of blood circulation, heart process and cardiac muscle regulation. These ndings suggest that rather than hypertension-related biological processes, other numerous biological processes were instead affected by the synergies of this decoction. As shown in Fig. 4 (b), KEGG enrichment analysis showed that 131 pathways (p value

Target-disease network
Based on the enrichment analysis (shown in Fig. 4), the target-pathway network was constructed to visualize the mechanisms of this decoction in hypertension treatment, by associating the pathways and targets in the network (depicted in Fig. 5). Other pathways known to be associated with hypertension, such as Cushing's syndrome, AGE-RAGE signaling pathway in diabetic complications and complement and coagulation cascades, were included.
The network constructed depicts the synergy of the TCM in managing or preventing hypertension through pathways such as cardiac muscle contraction, uid shear stress and atherosclerosis, DCM, renin-angiotensin system, HCM and arrhythmogenic right ventricular cardiomyopathy (ARVC). A total of 36 compound targets and 42 disease targets identi ed were mostly involved in HCM and DCM.

Discussion
Substantial evidence from previous studies, either in-vitro or clinical trials, have revealed TCM's curative effects on hypertension and potential for drug discovery [12,14]. In this study, we performed LC-MS/MS for identi cation of the chemical constituents existing in the TCM formulation consisting of Polygonum multi orum, Rehmannia glutinosa, Senna obtusifolia and Crataegus, to determine the bioactive compounds responsible for mitigating hypertension. From the numerous compounds detected (Supplementary Table 1), only the bioactive compounds which can contribute therapeutic effects [30] were selected, based on their OB and DL values. The OB and DL values of constituents are pivotal keys for successful development of therapeutic drugs with good oral absorption and permeability across the blood-brain barrier [31,32], in line with their molecular physicochemical properties [33].
While most network pharmacological analysis selections are based on OB≥30%, in actual fact there are no universal criteria to de ne high and low bioavailable compounds [34]. Thus, in this study, instead of selecting bioactive compounds with OB≥30%, we chose OB≥50% (as shown in Table 1) as the arbitrary classi cation threshold, with the hope that the selected bioactive compounds with higher OB would result in higher absorption in the human system [31]. Our selection was based on the observation that a lot of drugs failed in clinical trials due to poor absorption in the human system.
The ltering process revealed that 11 compounds ful lled the selection criteria (Table 1), with some found in duplicate within the four medicinal plants. A total of 124 genes were predicted as the target genes for these 11 compounds, as illustrated in Fig. 1. While the SwissTargetPrediction Database estimates predictions of protein targets for druglike molecules using similarity principles e ciently, as has been experimentally observed previously, it does not however provide calculations for large peptides [35]. From our results (shown in Fig. 2), one prime example is 2-methoxy-N-[2-(methylthio)phenyl] benzamide, which was shown to possess most of the predicted targets, which might indicate that it has been experimentally observed to bind to or homologs thereof. It is an established fact that the targets for this bioactive compound, involving hypertension, is through regulation of blood circulation, such as ADORA1, ADORA2B, HMOX1, KCNJ5, NOS2 (illustrated in Fig. 1 and Fig. 4(a)) [36][37][38][39].
Apart from the target compounds, hypertension disease targets were also determined through the OMIM and GeneCards databases. In addition to the hypertension disease gene targets, we also searched the databases for hypertension-related diseases such as cardiovascular disease or categories of diseases possessing both keywords, which possess strong inter correlations, as shown in previous reports [3,4]. Subsequently, the targets for the bioactive compounds were overlapped with the targets from diseases to construct the PPI network and visualized using Cytoscape Version 3.8 for better illustration, as depicted in Fig. 3, with con dence level of 0.7. Our ndings revealed that two main clusters were formed through this network, which consisted only of the target compounds (red node) and target diseases (blue node), in the bottom left and bottom right of the network, respectively. It showed that the target compounds were involved in neuroactive ligand receptor interaction while the target disease clusters were involved in HCM pathways. This indicates that the TCM decoction might trigger the central neural pathways, one of the multiple converging pathways leading to hypertension [40,41], through neuroactive ligand receptor interaction, to ameliorate hypertension.
The enrichment analysis for GO in biological processes and KEGG pathways demonstrated the involvement of this TCM decoction in numerous biological processes and pathways in the human system biology, as summarized in Fig. 4. As shown in Fig. 4 (a), the biological processes of high signi cance involve the regulation of biological quality, glutamate receptor signaling pathway, response to stress, regulation of localization and trans-synaptic signaling, among others. Among these biological processes, we found that regulation of heart contraction and regulation of blood circulation, which are strongly associated with hypertension [42,43], are also biological processes with p-value of 6.58x10 -19 and 2.84x10 -21 , respectively. This is concomitant with our KEGG pathway enrichment analysis (Figure 3 (b)) which revealed the high signi cance of HCM, DCM and cardiac muscle contraction, with p-values of 2.12x10 -17 , 1.89x10 -14 and 5.93 x10 -08 , respectively. All of these pathways are interlinked, either directly or indirectly [42]. Additionally, other top pathways found are neuroactive ligand-receptor interaction, glutamatergic synapse, pathways in cancer and adrenergic signaling in cardiomyocytes, among others.
Our PPI network associated with enrichment analysis also suggests that this TCM decoction resolves hypertension conditions through neuroactive ligand receptor interaction [40]. We thus constructed a network to visualize all the hypertension-related pathways by overlapping the compound targets and disease targets, as shown in Fig. 5. No compound targets related to hypertension were however detected which indicates that the synergies of the decoction might instead be involved in pathways indirectly linked to hypertension, to improve hypertension conditions in the human system biology. For example, one of the compound targets, angiotensin converting enzyme (ACE), plays a role in elevating hypertension via constriction of the blood vessels in the pathway of the renin angiotensin system [44]. This indicates that this TCM decoction might be contributing to hypertension amelioration through alternative pathways instead of affecting hypertension itself directly, as most of the targets involved neuroactive ligand receptor interaction, one of the pathways leading to hypertension [40] and HCM, a disease caused by hypertension [45,46]. Availability of data and materials

Conclusion
The datasets analysed during the current study are not publicly available due to this study will be further studied for their puri ed compounds' curative effect in hypertension particularly compounds unidenti ed for their pharmacological effect but are available from the corresponding author on reasonable request.

Competing interests
The authors declare that they have no competing interests.

Funding
This work was supported by the Malaysia Toray Science Foundation (Ref: 18/G105).
Authors' contributions FFL and FTC were responsible for the conception and design of the study. FFL was responsible for the supervision, conducted the experimental work, data analysis and revision of the manuscript. FFL and FOS drafted the manuscript. WTLY was responsible for conducted the experimental work and FPL was responsible for review and editing of the manuscript. All the authors reviewed and approved the entire manuscript.

Figure 2
Compound-Target Network. Visualization of connection of 11 bioactive compounds (blue arrow) with its target genes (red arrow) from the decoction of four medicinal plants.

Figure 3
Visualization of PPI interaction network analyzed by STRING using Cytoscape. Compound targets (red node) and hypertension-related targets (blue node) were linked together through edges (green) for its PPI interaction.  Target-disease network constructed to visualize the interaction between targets to hypertension-related pathways.
Fifteen hypertension-related pathways (orange rectangle) were connected to compound targets (red arrow) and disease targets (blue arrow) to establish the correlation produced by TCM decoction from the four selected medicinal plants.