Systems Pharmacological Approach of Pulsatillae Radix on Treating Crohn's Disease

In East Asian traditional medicine, Pulsatillae Radix (PR) is widely used to treat amoebic dysentery and renowned for its anti-inflammatory effects. This study aimed to confirm evidence regarding the potential therapeutic effect of PR on Crohn's disease using a system network level based in silico approach. Study results showed that the compounds in PR are highly connected to Crohn's disease related pathways, biological processes, and organs, and these findings were confirmed by compound-target network, target-pathway network, and gene ontology analysis. Most compounds in PR have been reported to possess anti-inflammatory, anticancer, and antioxidant effects, and we found that these compounds interact with multiple targets in a synergetic way. Furthermore, the mRNA expressions of genes targeted by PR are elevated significantly in immunity-related organ tissues, small intestine, and colon. Our results suggest that the anti-inflammatory and repair and immune system enhancing effects of PR might have therapeutic impact on Crohn's disease.


Introduction
Inflammatory bowel disease (IBD) may be categorized clinically as Crohn's disease or ulcerative colitis [1]. Crohn's disease usually causes a variety of systemic symptoms, which include chronic inflammation of the bowel [2]. Although any part of the digestive tract from mouth to anus may be affected, usually the small intestine (ileum) and the large intestine (colon) are involved; ileocolic Crohn's accounts for 50% of cases, ileal Crohn's for 30%, and colic Crohn's for 20% of cases [3]. Symptoms vary though its common manifestations include persistent diarrhea, rectal bleeding, abdominal cramps, and pain, though fever, extreme fatigue, and weight loss are also common [1,2,4]. Constipation is also a frequent symptom and can lead to bowel obstruction and, thus, increase the risk of bowel cancer. Complications involving other than the gastrointestinal (GI) tract include anemia, arthritis, liver disease, eye inflammation, and skin rashes [2,4].
According to a statistical report issued by the Health Insurance Review and Assessment Service in Korea in 2016, the number of Crohn's disease patients increased from 13,920 in 2011 to 18,332 in 2015, an average annual increase of 7.1%. Furthermore, over the same period, total medical cost has increased by 19.4% annually, and more than half of patients are in 20s or 30s. A systematic review about the economic and quality-of-life burden of Crohn's disease reported that, in the USA, Germany, France, UK, Italy, and Spain, in all countries combined, Crohn's medical costs totaled €30 billion annually and that patient quality of life was substantially diminished by the physical, emotional, and social effects of the disease [5].
Although Crohn's disease is a global health problem [5], its pathology remains poorly understood [1,6]. Nevertheless, it has been established that its etiology is associated with complex interactions between environmental, immune, microbial, and genetic factors [4,7], though a number of authors have suggested that the primary defect in Crohn's disease is one of relative immunodeficiency [6,8,9].
A systematic review of publications from 1947 to 2013 involving controlled studies of herbal therapies in IBD indicated that at least 50 percent of IBD patients used some form of complementary and alternative medicine (CAM), since most herbal therapies had been reported to have antiinflammatory effects and plausible action mechanisms in IBD with minimal adverse effects [10]. Moreover, herbal medicines are being increasingly used and requested by IBD patients not only in Asia, but also in Western countries [11].
Of the anti-inflammatory herbal medicines, Pulsatillae Radix (Baekduong, PR) is worth considering as a potential treatment for Crohn's disease as it was widely used in traditional medicine to treat amoebic dysentery and has also been demonstrated scientifically to have anti-inflammatory effects experimentally [12,13]. PR is traditional Korean herbal medicine prepared from the roots of Pulsatilla koreana Nakai and contains several phytochemicals, including anemonin, hederagenin, oleanolic acid, and deoxypodophyllotoxin [14,15]. Some experimental study results have shown that PR has various biological activities. For instance, aqueous and ethanol extracts of PR have been reported to demonstrate anticancer effects in anaplastic thyroid cancer [16], methanol extract of PR was found to have anti-inflammatory effects in lipopolysaccharide (LPS) exposed rats [17], and PR has also been reported to inhibit adipocyte differentiation and to suppress adipogenesis [13].
According to the meridian tropism theory of traditional medicine, the effects of PR reach to the stomach meridian and the large intestine meridian, which is in accord with the organs commonly affected by Crohn's disease. In terms of the selection of traditional medication, meridian tropism theory plays an important role, as it is one of the core principles of traditional medicine [18]. Based on meridian tropism theory, each herbal medicine possesses different affinities for certain organs and meridians of the body [19].
In the present study, we sought to confirm the therapeutic effects of PR in Crohn's disease using system level analysis and a network based in silico approach. A schematic of the network pharmacological study is summarized in Figure 1.

Drug-Likeness
Evaluation. DL helps filter out "druglike" compounds in oriental herbs, as DL represents a qualitative concept for valuations based on how "druglike" prospective compound is [23]. Accordingly, a high DL may lead to a greater possibility of therapeutic success, and compounds with a higher DL value are more likely to possess certain biological properties [24]. Calculations of DL in the TCMSP database are based on the Tanimoto coefficient formula [25] as follows: where represents the molecular parameters of herbal compounds and is the average molecular parameter of all compounds in the DrugBank database (http://www.drugbank.ca/) Evidence-Based Complementary and Alternative Medicine 3 [26]. In the present study, we excluded compounds with a DL value of < 0.18.

Oral Bioavailability (OB) Prediction.
Oral bioavailability (OB) is defined as the absorption ratio of an active compound into the systemic circulation, which represents convergence of the ADME process [21]. OB values are dependent on drug dissolution in the gastrointestinal (GI) tract, intestinal and hepatic first-pass metabolisms, and intestinal membrane permeation, and, thus, OB is considered a major pharmacokinetic parameter for drug evaluations [24]. In this study, the OB threshold was set as ≥ 15%.

Caco-2 Permeability
Screening. Caco-2 permeability is used to predict the absorption of an orally administered drug [22]. Surface absorptivity of the small intestine is maximized by villi and microvilli; for this reason, orally administered drugs are mostly absorbed in the small intestine [27]. Moreover, the movement of orally administered drugs across the intestinal epithelial barrier determines the rate and extent of human absorption and ultimately affects drug bioavailability [28]. In the present study, compounds with OB, DL, and Caco-2 values of >15%, >0.18, and >−0.4, respectively, were regarded as active and subjected to analysis. In addition, we included some compounds with lower ADME profile than above thresholds, for the reason that those were reported to possess anti-inflammatory, antioxidant, anticancer, and antibacterial effects. This study was about only one single herb, and for this reason we did not use a high threshold of ADME profile to filter potential active compounds. Instead, we lowered the standard of OB in order to analyze the most potential targets of PR.

Gene Ontology (GO) Analysis.
Biological process (BP) of gene ontology (GO) analysis was employed to determine the biological properties of target genes [31]. GO annotation provides statistical analyses on gene function information. In this research, GO BP terms with values < 0.01 were employed and the data was collected using the DAVID 6.8 Gene Functional Classification Tool (http://david.abcc.ncifcrf.gov/).

Network Construction and Analysis.
In order to understand the multiscale interactions between the active compounds of PR and targets, two types of networks were built: (1) a compound-target network (C-T network), in which nodes represented either compounds or target proteins, and edges indicated compound-target connections; and (2) a target-pathway network (T-P network), which was used to extract pathways from the KEGG database (http://www.genome.jp/kegg/) and to filter out terms highly associated with Crohn's disease and values of < 0.01. Related targets were mapped onto relevant pathways to produce the T-P network. Both networks were generated in Cytoscape 3.4.0, an open-source biological network visualization and data integration software package [32].

Target Organ Location Network.
Tissue-specific patterns of mRNA expression can indicate important associations with particular biological events or gene functions [33]. Thus, to explore the beneficial effects of PR on Crohn's disease, it was important to determine the tissue mRNA expression profiles of target proteins at the organ level [34]. The target organ location network was used with the Dataset: GeneAtlas U133A, gcrma (http://biogps.org). The BioGPS database provides expression data acquired by direct measurements of gene expressions by microarrays analysis [35]. First, the mRNA expression patterns of each target gene in 84 parts of organ tissues were obtained. Second, average values were calculated for each gene. Third, above average mRNA expressions in relevant organ tissues were extracted and arranged by frequency. Finally, a target organ location network was constructed using Cytoscape 3.4.0 and organspecific, Crohn's disease related, gene overexpression data.
2.6. GEO2R Analysis. Using Gene Expression Omnibus (GEO, https://www.ncbi.nlm.nih.gov/geo), we compared mRNA expression pattern of normal groups and Crohn's disease groups. GEOquery and limma R packages of GEO2R tool were employed to identify highly expressed genes.

Network Pathway.
In order to elucidate the action mechanisms of PR in Crohn's disease, filtered target proteins were input into the pathway map of inflammatory bowel disease acquired from the Kyoto Encyclopedia of Genes and Genomes (KEGG, http://www.kegg.jp/) database.  [36]. Pinoresinol (C23; DL = 0.52, OB = 4.25, Caco-2 = 0.52) was reported to have anti-inflammatory properties [37].

Target
Fishing. These 32 identified active compounds interact with 182 target proteins ( Table 2); that is, on average, they interact with 5.7 target genes, which does much to explain the polypharmacological and synergistic effects of PR on multiple targets [51].

Network Construction and Analysis.
To visualize more conveniently the multitargeted effects of PR, network analysis was used to investigate its actions within the context of the whole human genome [52,53]. As shown in Figure 3, constructed (A) C-T and (B) T-P network demonstrated multicompound and multitargeted effects and relations between various pathways and targets. Circular nodes represent compounds and targets in the C-T network and triangles and circular nodes show pathways and compounds in the T-P network. In both networks, node size was regulated by degree centrality and edges showed interactions between nodes.
The C-T network showed 415 interactions between the 182 targets and 32 active compounds of PR. Ursolic acid (C30, degree = 55) had the highest number of interactions with targets, followed by beta-sitosterol (C8, degree = 37) and isorhamnetin (C17, degree = 36), and, thus, these results demonstrated that single molecules can target multiple receptors [54]. Likewise, prostaglandin G/H synthase 2 (PTGS2, degree = 15) displayed the most affinitive connections with compounds, followed by prostaglandin G/H synthase 1 (PTGS1, degree = 13) and nuclear receptor coactivator 2 (NCOA2, degree = 11). According to betweenness centrality results, protein-tyrosine phosphatase 1B (PTPN1, betweenness = 0.11) was second highest followed by PTGS2 (betweenness = 0.12). 28 (88%) Of the 32 active compounds were connected with more than two targets and 86 (47%) of the 182 targets interact with more than one compound. This network analysis results clearly demonstrated the multitargeting natures of herbal compounds; besides it showed that ursolic acid (C30) is the most essential compound in PR.

Target Organ Location Network.
The tissue mRNA expression profiles of target genes at the organ level were investigated to identify effects of PR on Crohn's disease. No mRNA expression information of three of genes, muscarinic acetylcholine receptor M1 (CHRM1), G-protein coupled receptor TGR-1 (NMUR2), and taste receptor type 2 member 31 (TAS2R31), was found in the BioGPS. In total, the mRNA expression profiles of 179 of 182 genes were analyzed (Supplementary Table S2). 158 Of the genes displayed above average mRNA expressions in 17 relevant organ tissues, namely, in BDCA4+ dendritic cells, bone marrow, CD14+ monocytes, CD19+ B cells, CD33+ myeloid cells, CD34+ hematopoietic stem cells, CD4+ T cells, CD56+ NK cells, CD8+ T cells, colon, colorectal adenocarcinoma, liver, lymph nodes, lymphoblasts, small intestine, smooth muscle, and thymus. In addition, these 17 organ tissues, retina, prefrontal cortex, pineal, amygdala, cardiac myocyte, heart, whole blood, and other tissues, were also associated with relatively high mRNA expressions. Networks of the tissue mRNA expressions of 158 target genes and PR compounds are shown in Figure 4, nodes represent organs and genes, and node sizes indicate the number of interactions between nodes. In detail, among 84 organ tissues, CD33+ myeloid showed the most overexpressed mRNA expression, 104 genes of 158 target genes were overexpressed in CD33+ myeloid, followed by 97 genes in lymphoblasts, 95 in each of smooth muscle and CD34+ hematopoietic stem cells, 91 genes in liver and CD56+ NK cell, 84 in bone marrow and colorectal adenocarcinoma, 82 in BDCA4+ dendritic cells, 75 in CD14+ monocytes, 73 in small intestine, 72 in CD4+ T cells, 70 in CD19+ B cells, 67 in colon, 38 in CD8+ T cells, 23 in thymus, and 22 in lymph nodes. It is evident that most genes were overexpressed in several organs at the same time.
Additionally, intestinal alkaline sphingomyelinase (ENPP7), DNA polymerase catalytic subunit (POLG), and carbonic anhydrase 13 (CA13) recorded beyond average mRNA expressions in all 17 organs. Furthermore, more than 146 (92%) of 158 target genes were overexpressed in two or more organ tissues, suggesting that these organs and compounds in PR are closely related. Furthermore, because the above 17 organs are highly related to immunity, our study results indicate that the therapeutic effects of PR on Crohn's disease are due to its targeting and activating the immune system. The other 21 target genes, such as acetylcholinesterase (ACHE), pancreatic alpha-amylase (AMY2A), and muscarinic acetylcholine receptor M3 (CHRM3), did not show above average mRNA expression in the 17 organs.
3.6. GEO2R Analysis. Comparison data between normal tissue and Crohn's disease patients' mRNA expression pattern from Gene Expression Omnibus (GEO) was collected. We employed GEOquery and limma R packages of GEO2R tool to identify highly expressed genes in 6 datasets. Dataset accession numbers are as follows: GSE24287, GSE60083, GSE6731, GSE36807, GSE68570, and GSE72780. To sum up, 86 normal samples and 149 Crohn's disease samples were analyzed in each dataset. GEO2R presented the top 250 highly expressed genes in Crohn's disease group compared to the control group and we deleted overlaps, so, in the end, 1182 genes were sorted out.
We found out that there were 23 common genes (Table 3) between target genes of PR and highly expressed genes of Crohn's disease dataset from GEO.

Network Pathway.
In order to investigate further the effect of PR in Crohn's disease, we performed pathway enrichment analysis ( Figure 5). Using the IBD pathway provided by the KEGG pathway database, we confirmed the pathway mapping effect of PR in Crohn's disease by inputting the filtered human target genes into the pathway. The KEGG pathway was constructed according to the current knowledge of the pathogenesis IBD.

Discussion
In this study, a network pharmacology method with active compounds filtration, multiple drug target prediction, gene ontology, network analysis, relevant organ location network, and pathway enrichment analysis were employed to determine the targets of PR in relation to Crohn  study shows that PR is highly connected to the pathways, biological processes, and organs of Crohn's disease. A pharmacological approach was used to identify the actions of PR at the systems network level.
In this study, pathway mapping result showed that the target genes of PR overlap more with Crohn's disease than with ulcerative colitis. Experimental study also suggested that the markers of both diseases are different from each other [59,60]. The clinical symptoms of these diseases also differ; for instance, Crohn's disease affect any region of the entire gastrointestinal (GI) tract and all layers of the bowel wall, whereas ulcerative colitis affects only the mucosa and submucosa of colon [61]. Furthermore, ulcerative colitis can be cured by surgery, but Crohn's disease of any part of GI tract tends to relapse after surgery [62]. For this reason, long term management using herbal medicines might be highly recommendable treatment option for Crohn's disease, since herbal medicines have advantages for managing chronic diseases [63,64].
In addition, IBD is usually referred to as an autoimmune disorder [65], whereas Crohn's disease does not meet the criteria of an autoimmune disorder; rather it is associated with immune deficiency or a secondary immune response to altered intestinal microbiota [65]. Furthermore, ulcerative colitis is a mucosal disease where autoimmune autoantibodies are commonly detected [66], whereas Crohn's disease is a transmural disease, in which pathological changes in gut wall are thought to result from submucosal inflammatory changes [67]. Accordingly, the areas targeted for treatment in these two diseases should be differentiated.
In the present study, we focused on the use of PR as a potential therapy for Crohn's disease. However, herb pairs and combinations are more commonly prescribed and are regarded to be more effective and safer [68]. In terms of the Gunsinjwasa theory of traditional medicine combinations, there are four different roles for each herb in the formula. First, the major component targets the main symptom. Second, the supportive component assists the effect of the major component or targets the secondary symptoms. Third, the neutralizing component allays the side effects or toxins of the major and the supportive component. Fourth, the deliver/retaining component guides the medicine to the targeting part of the body [69,70]. This combination principle enables not only the enhancement of synergistic medicinal effects but also potentially reduces toxicities [69]. In order to induce better effects and reduce toxicities, an extended analysis of the Pulsatillae Radix (Baekduong, PR), Phellodendri Cortex (Hwangbaek), Coptidis Rhizoma (Hwangryeon), and Citrus reticulata (Jinpi) herb combination (a widely prescribed formula, known as Baekduong decoction) should be explored in the future.
Through GO analysis, we found out that targets of PR are associated with liver diseases such as hepatitis B, hepatitis C, and nonalcoholic fatty liver. In addition, the mRNA expression of 91 of 179 genes in liver was overexpressed according to the target organ location network result. Liver inflammation is as common extraintestinal symptom of Crohn's disease [71], and the number of liver abscesses in Crohn's disease patients is 15 to 20 times higher than that found in the general population [72]. Besides biochemical liver dysfunction [73] and hepatic fibrosis [74] are also frequently found in Crohn's disease, and a number of drugs used to treat IBD have been reported to be associated with liver injury [75]. Furthermore, in an experimental study using a mice model of Crohn's disease-like ileitis, it was found that TLR9 plays an important role in hepatic involvement in IBD [71]. More detailed pathways and relations between liver and Crohn's disease should be discussed in the future.
This study demonstrated that 73 and 67 of 158 targets of PR were highly expressed in small intestine and colon, respectively, results which were accordance with not only the organs commonly affected by Crohn's disease but also the properties of PR in terms of meridian tropism theory. However stronger evidence with another research design is required to support this result.
We confirmed a multicompound and multitarget interaction through the C-T network, which showed that 88% of the active compounds were connected with more than two targets and 47% of the targets interacted with more than one compound. Although it demonstrated the synergetic network of multitarget actions, we should explore more differentiated drug action based on degree centrality and how different the drug actions are when more compounds target one gene in simultaneous way compared to the case of one compound targeting one gene.

Conclusion
Previous studies indicate that most compounds in PR have anti-inflammatory, anticancer, and antioxidant effects. In the present study, we found that these compounds interact with multiple targets in a synergetic manner and that PR is highly connected to Crohn's disease related pathways, biological processes, and organs. C-T and T-P network results demonstrated complex multicompound and multitarget drug actions and the relations between targets and various pathways. Furthermore, target genes were found to be overexpressed in organs highly related to immunity. These findings suggest the anti-inflammatory effects of PR, and its enhancements of repair processes and immune system might be of therapeutic benefit in Crohn's disease.

Conflicts of Interest
The authors have no conflicts of interest to declare.