Revealing Potential Bioactive Compounds and Mechanisms of Lithospermum erythrorhizon against COVID-19 via Network Pharmacology Study

Lithospermum erythrorhizon (LE) is known in Korean traditional medicine for its potent therapeutic effect and antiviral activity. Currently, coronavirus (COVID-19) disease is a developing global pandemic that can cause pneumonia. A precise study of the infection and molecular pathway of COVID-19 is therefore obviously important. The compounds of LE were identified from the Natural Product Activity and Species Source (NPASS) database and screened by SwissADME. The targets interacted with the compounds and were selected using the Similarity Ensemble Approach (SEA) and Swiss Target Prediction (STP) methods. PubChem was used to classify targets linked to COVID-19. The protein–protein interaction (PPI) networks and signaling pathways–targets–bioactive compounds (STB) networks were constructed by RPackage. Lastly, we performed the molecular docking test (MDT) to verify the binding affinity between significant complexes through AutoDock 1.5.6. The Natural Product Activity and Species Source (NPASS) revealed a total of 82 compounds from LE, which interacted with 1262 targets (SEA and STP), and 249 overlapping targets were identified. The 19 final overlapping targets from the 249 targets and 356 COVID-19 targets were ultimately selected. A bubble chart exhibited that inhibition of the MAPK signaling pathway could be a key mechanism of LE on COVID-19. The three key targets (RELA, TNF, and VEGFA) directly related to the MAPK signaling pathway, and methyl 4-prenyloxycinnamate, tormentic acid, and eugenol were related to each target and had the most stable binding affinity. The three bioactive effects on the three key targets might be synergistic effects to alleviate symptoms of COVID-19 infection. Overall, this study shows that LE can play a role in alleviating COVID-19 symptoms, revealing that the three components (bioactive compounds, targets, and mechanism) are the most significant elements of LE against COVID-19. However, the promising mechanism of LE on COVID-19 is only predicted on the basis of mining data; the efficacy of the chemical compounds and the affinity between compounds and the targets in experiment was ignored, which should be further substantiated through clinical trials.


Introduction
The outbreak of a new health crisis due to coronavirus disease 2019 (COVID- 19) occurred in Wuhan, Hubei Province, China, in December 2019 [1]. On 9 January 2020, the novel coronavirus SARS-CoV-2 was officially acknowledged as the sole contributor to this outbreak [2]. The World Health Organization (WHO) initially called this situation a Public Health Emergency of International Concern on 30 January [3], and then announced it was a global pandemic on 11 March [4].

Targets Associated with Selected Compounds or COVID-19
Based on SMILES, targets related to the selected compounds were identified via both SEA (http://sea.bkslab.org/) (accessed on 23 October 2021) [22] and STP (http://www.swisstargetprediction.ch/) (accessed on 24 October 2021) [23] using the "Homo Sapiens" mode. We selected the overlapping targets between SEA and STP databases based on the use of a Venn diagram. Moreover, the COVID-19 targets were adapted from PubChem (https://pubchem.ncbi.nlm.nih.gov/) (accessed on 24 October 2021). The final targets between the selected compounds-related targets and COVID-19 targets were visualized by a Venn diagram plotter.

The Analysis of the Protein-Protein Interaction (PPI) Networks
The final targets were utilized to construct a protein-protein interaction (PPI) network on STRING (https://string-db.org/) (accessed on 27 October 2021). In the PPI network, the size of circle represents the degree of value. In particular, a target indicated in red color in the most central position was considered the most significant target.

Targets Associated with Selected Compounds or COVID-19
Based on SMILES, targets related to the selected compounds were identified via both SEA (http://sea.bkslab.org/) (accessed on 23 October 2021) [22] and STP (http://www. swisstargetprediction.ch/) (accessed on 24 October 2021) [23] using the "Homo Sapiens" mode. We selected the overlapping targets between SEA and STP databases based on the use of a Venn diagram. Moreover, the COVID-19 targets were adapted from PubChem (https://pubchem.ncbi.nlm.nih.gov/) (accessed on 24 October 2021). The final targets between the selected compounds-related targets and COVID-19 targets were visualized by a Venn diagram plotter.

The Analysis of the Protein-Protein Interaction (PPI) Networks
The final targets were utilized to construct a protein-protein interaction (PPI) network on STRING (https://string-db.org/) (accessed on 27 October 2021). In the PPI network, the size of circle represents the degree of value. In particular, a target indicated in red color in the most central position was considered the most significant target.

The Construction of a Bubble Chart
A bubble chart was constructed according to the rich factor, which is defined as the proportion of the number of genes expressed differentially in a signaling pathway [24].
Thus, we identified a hub signaling pathway related to a key target in PPI networks. The bubble chart was plotted in RPackage based on STRING (https://string-db.org/) (accessed on 27 October 2021).

The Assembly of Signaling Pathways-Targets-Bioactive Compounds (STB) Networks
The STC networks were illustrated the relationships of the three components (signaling pathways-targets-bioactive compounds) and obtained the most critical target in the correlation. In the network, yellow rectangles (nodes) stood for the signaling pathways; blue triangles (nodes) represented targets; and pink circles (nodes) denoted bioactive compounds. The size of the blue triangles marked the number of correlations with the signaling pathways; the size of the pink circles depicted the number of connections with the targets. The merged networks were completed in RPackage.

The Preparation of the Bioactive Compounds and Targets for Molecular Docking Test (MDT)
The bioactive compounds were associated with the key signaling pathway were extracted .sdf format from PubChem, which were converted into .pdb format using Pymol, and then they were changed into .pdbqt format via AutoDock. The number of three proteins on the MAPK signaling pathway, i.e., TNF (PDB ID: 5YOY), RELA (PDB ID: 2O61), and VEGFA (PDB ID: 3P9W) were obtained using STRING via RCSB PDB (https: //www.rcsb.org/) (accessed on 28 October 2021). The targets were converted .pdb format into .pdbqt format via AutoDock (http://autodock.scripps.edu) (accessed on 30 October 2021).

The MDT on a Key Signaling Pathway
The MDT were performed to verify the affinity between bioactive compounds and targets on a key signaling pathway. The set-up condition consisted of a value of 4 for the energy range and a value of 8 for the exhaustiveness as the default settings in order to  [25]. The lower the binding energy (the higher the negative value), the greater the stable binding is between the bioactive and the target.

Potential Bioactive Compounds from LE
We obtained the 82 bioactive compounds in LE through NPASS database and physicochemical properties of these compounds are listed in Table 1. All of them were confirmed by Lipinski's rule [26] and TPSA (<140 Å 2 ) [27]. Thus, we considered that these compounds might be potential therapeutic agents against COVID-19.

Targets Associated with the 82 Compounds or COVID-19
As shown in Supplementary Table S1, the total number of 1262 targets (SEA + STP) related to 82 compounds was identified in DisGeNET and OMIM. Then, the overlapping 249 targets between SEA and STP were obtained ( Figure 2). The 249 targets (Supplementary  Table S2) were analyzed with 356 COVID-19 related targets (Supplementary Table S3). Finally, the Venn diagram ( Figure 3)

Targets Associated with the 82 Compounds or COVID-19
As shown in Supplementary Table S1, the total number of 1262 target related to 82 compounds was identified in DisGeNET and OMIM. Then, th 249 targets between SEA and STP were obtained ( Figure 2). The 249 targets tary Table S2) were analyzed with 356 COVID-19 related targets (Supplem S3). Finally, the Venn diagram ( Figure 3) showed that 19 overlapping targets tary Table S4) were directly associated with response to COVID-19 infection

The Protein-Protein Interaction (PPI) Networks from 19 Targets
In the PPI networks, the TNF target was considered to be the most significant targ with the highest degree of value (15) ( Table 2). Moreover, the 19 targets were closely terconnected with each other (19 nodes and 69 edges) ( Figure 4). The TNF in the m central position was the most significant target in the PPI networks.

The Protein-Protein Interaction (PPI) Networks from 19 Targets
In the PPI networks, the TNF target was considered to be the most significant target with the highest degree of value (15) ( Table 2). Moreover, the 19 targets were closely interconnected with each other (19 nodes and 69 edges) ( Figure 4). The TNF in the most central position was the most significant target in the PPI networks.

A Bubble Plot and Signaling Pathways-Targets-Bioactive Compounds (STB) Networks
The Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis indicated that the 19 targets were associated directly with 18 signaling pathways

A Bubble Plot and Signaling Pathways-Targets-Bioactive Compounds (STB) Networks
The Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis indicated that the 19 targets were associated directly with 18 signaling pathways (False Discovery Rate < 0.05) ( Figure 5). The identified 18 signaling pathways were involved in the response to a COVID-19 infection. Detailed information on the 18 signaling pathways is presented in Table 3. Additionally, a bubble plot suggested that the RAS signaling pathway might be a key signaling pathway due to the lowest rich factor (0.008). Moreover, we performed an STB networks analysis to identify the most important target based on the degree of value ( Figure 6). Thus, the highest degree of value in STB networks was the RELA target with 17 degrees, which was considered as a notable target against COVID-19 (Table 4). Comprehensively, a signaling pathway that combined both TNF (a key target in PPI networks) and RELA (a key target in STB networks) was the MAPK signaling pathway, which had an antagonistic propensity on the relatively lower rich factor found in the 18 signaling pathways. We observed that the uppermost signaling pathway was not the RAS signaling pathway but the MAPK signaling pathway, which consisted of the two core targets (TNF and RELA).

Discussion
The compounds-targets network indicated that LE compounds might be significant ligands that could be used to alleviate COVID-19 symptoms. Of these, RELA has a more significant effect than any other kinds of targets. Furthermore, based on the pathway enrichment, the MAPK signaling pathway is the uppermost mechanism of LE against

Discussion
The compounds-targets network indicated that LE compounds might be significant ligands that could be used to alleviate COVID-19 symptoms. Of these, RELA has a more significant effect than any other kinds of targets. Furthermore, based on the pathway enrichment, the MAPK signaling pathway is the uppermost mechanism of LE against COVID-19. Thus, three targets (RELA, TNF, and VEGFA) linked to the MAPK signaling pathway might be promising targets for use against COVID-19. Accordingly, the docking score on the three targets suggested that three bioactive compounds (methyl 4-prenylcinnamate, tormentic acid, and eugenol) were considered as the most notable compounds of LE against COVID-19. Meanwhile, the results of the KEGG pathway enrichment analysis showed that nine targets might play important roles against COVID-19. The relationships of the 18 signaling pathways with anti-virus were discussed as follows.
• AGE-RAGE signaling pathway in diabetic complications: Activation of the binding of AGE to its receptor RAGE can stimulate cytokine production, can cause tissue damage, and the suppression of AGE-RAGE can effectively reduce inflammation [28]. • Adipocytokine signaling pathway: Most adipocytokines are pro-inflammatory factors and they are closely linked to chronic inflammation [29,30]. • RIG-I-like receptor (RLR) signaling pathway: The RIG-I-like receptors (RLRs), RIG-I, MDA5, and LGP2, play an important role in pathogen recognition of RNA virus infections that instigate and regulate antiviral immunity [31]. • IL-17 signaling pathway: IL-17 can play vital roles in responding to pathogenicity in diverse tissues, as well as being important for inflammation balance and tissue cohesion during viral attacks [32]. • Toll-like receptor (TLR) signaling pathway: Nucleic acids originating from bacteria and viruses can be recognized by the intracellular Toll-like receptor (TLR) and they are also sensitive to self-nucleic acids in disease conditions such as autoimmunity [33].  [44].
Based on the pathway enrichment analysis, RELA was considered as a hub target in LE against COVID-19. The RELA was directly enriched in 17 out of 18 signaling pathways by the MAPK signaling pathway, indicating that the MAPK signaling pathway might be a hub signaling pathway in LE against COVID-19. The other two targets (TNF and VEGFA) that are directly related to the MAPK signaling pathway might be important targets for creating synergistic effects against COVID-19. A report demonstrated that inhibition of RELA of the NF-κB component reduce cytokine production and thus could alleviate inflammation severity [45]. Most recently, it has been reported that anti-TNF treatment for COVID-19 patients with rheumatoid arthritis diseases showed preventive effects against the high levels of cytokines involved in the immune response of infection, and the therapeutic application of anti-TNFs can lessen the incidence of severe inflammation of COVID-19 [46]. Notably, a report indicated that vascular endothelial growth factor A (VEGFA) antagonized by angiotensin-converting enzyme 2 (ACE2) that is upregulated by COVID-19 infection because COVID-19 inhibits the expression of ACE2. Consequently, VEGFA increases vascular permeability and lessens endothelial damage [47].
Endothelial cell inflammation is a serious symptom of COVID-19 infection, and its uncontrollable cytokine production in tissues and cells causes a severe immune reaction, which is defined as a "cytokine storm" that results in aggravating pneumonia. Moreover, inhibition of the other two targets (TNF and VEGFA) related to the MAPK signaling pathway contribute to anti-proinflammation and anti-vascular permeability against COVID-19. Therefore, the key mechanism of LE against COVID-19 might be the ability to block inflammation and vascular permeability in tissues and/or cells by inactivating the MAPK signaling pathway (Figure 8). However, this research still has some limitations. The incompleteness of the natural products dataset might create a fallacy and COVID-19 data is updated continually as a new version. During the analysis, some results might cause an error unexpectedly, if we are only focused on computational methods.

Conclusions
The bioactive compounds and mechanism(s) of LE against COVID-19 were first explored using network pharmacology. The findings of this study suggest that the final bioactive compounds were methyl 4-prenylcinnamate, tormentic acid, and eugenol and their targets were RELA, TNF, and VEGFA, respectively. The mechanism(s) of LE against COVID-19 might inhibit cell inflammation and permeability against COVID-19 by inactivating the MAPK signaling pathway. This research provides a scientific indication to support LE's therapeutic effects on COVID-19, and thus, the proper application of these three bioactive compounds against COVID-19 might lead to promising synergistic effects such as anti-inflammation and anti-permeability in order to alleviate COVID-19 symptoms. In Thus, we need to validate the pharmacological mechanism via in vitro or in vivo tests. Thereby, the usefulness of computational approach might be represented. As a matter of fact, network pharmacology is a holistic perspective to search for multiple factors against specific diseases, which might provide significant clues about the relationships between components such as signaling pathways, targets, and compounds.

Conclusions
The bioactive compounds and mechanism(s) of LE against COVID-19 were first explored using network pharmacology. The findings of this study suggest that the final bioactive compounds were methyl 4-prenylcinnamate, tormentic acid, and eugenol and their targets were RELA, TNF, and VEGFA, respectively. The mechanism(s) of LE against COVID-19 might inhibit cell inflammation and permeability against COVID-19 by inactivating the MAPK signaling pathway. This research provides a scientific indication to support LE's therapeutic effects on COVID-19, and thus, the proper application of these three bioactive compounds against COVID-19 might lead to promising synergistic effects such as anti-inflammation and anti-permeability in order to alleviate COVID-19 symptoms. In parallel, this study needs to be further verified through in vitro or in vivo models.  Funding: This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.