Abstract

Aim

   The application of network analysis algorithms promoted the development
   of network pharmacology. This study aimed to combine network
   pharmacology and signed random walk with restart (SRWR) to reveal the
   mechanism by which Lithospermum erythrorhizon Sieb (LES) exerts effects
   on atopic dermatitis (AD).

Methods

   The compounds and targets of LES were retrieved from Traditional
   Chinese Medicine Integrated Database (TCMID) and Traditional Chinese
   Medicine Systems Pharmacology Database and Analysis Platform (TCMSP),
   and important compounds and targets were identified by intersection
   analysis and protein–protein interaction (PPI) network.

Results

   We found that active LES-derived compounds such as caffeic acid,
   Isovaleric acid, Arnebinol, and Alannan may inhibit PTGS2, HSP90AA1 and
   MAPK14, which are key mediators involved in PI3K-Akt pathway, vascular
   endothelial growth factor signaling pathway, Fc epsilon RI signaling
   pathway, and calcium signaling pathway.

Conclusion

   The application of SRWR could identify potential targets of LES with a
   low false-positive rate and help elucidate the mechanism of action of
   traditional Chinese medicine.

   Keywords: signed random walk with restart, network pharmacology,
   Lithospermum erythrorhizon Sieb, atopic dermatitis

Introduction

   Network pharmacology is a novel method to explore the interaction
   between active components and potential targets of drugs.[34]^1 In
   addition, network pharmacology integrates the interaction into a
   network and analyzes the action of traditional Chinese medicine (TCM)
   from a systematic perspective.[35]^2 Meanwhile, some problems such as
   the difficulty of data integration for network pharmacology remain to
   be overcome. The signed random walk with restart (SRWR) has been
   developed to simulate the propagation of activation or inhibition
   effects of the drug on the signaling network.[36]^3^,[37]^4

   Atopic dermatitis (AD) is a chronic inflammatory skin disease that
   affects around 200 million people in the world. Although AD is more
   common in children (incidence of AD is about 20% in children and 10% in
   adults), approximately 1 in 4 adults with AD have adult-onset
   disease.[38]^5 Recent investigations have revealed the mechanisms that
   contribute to the pathogenesis of AD, such as the role of aryl
   hydrocarbon receptor.[39]^6 Lithospermum erythrorhizon Sieb (LES) has
   been used to treat AD. Naphthoquinone derivatives shikonin,
   acetylshikonin, and isobutylshikonin are widely present in the
   epidermis of the roots of LES.[40]^7 Wang et al showed that shikonin
   reduced inflammatory cell infiltration and collagen deposition by
   inhibiting NF-κB pathway and suppressing the expression of MMP9.[41]^8
   However, the potential mechanisms of anti-inflammatory and anti-allergy
   activities of shikonin derivatives against AD have not been completely
   elucidated. Therefore, this study aimed to combine network pharmacology
   and signed random walk with restart (SRWR) to reveal the mechanism by
   which LES exerts effects on AD.

Methods

Chemical Compounds and Targets of LES

   The compounds and targets of LES were retrieved from Traditional
   Chinese Medicine Integrated Database (TCMID)[42]^9 and Traditional
   Chinese Medicine Systems Pharmacology Database and Analysis Platform
   (TCMSP).[43]^10 The bioactive components were selected based on the
   absorption, distribution, metabolism and excretion (ADME).[44]^11 The
   candidate components were determined based on two parameters: (1)
   Pharmacokinetics GI absorption, (2) Drug-likeness (DL): the Lipinski
   rule of five and the Veber Ghose, Muegge and Egan rules, Compounds that
   meet more than two of Lipinski’s rules will be chosen.[45]^12
   Eventually, 60 compounds were retrieved.

AD Related Genes

   AD-related genes were screened from the DisGeNET^13 and GEO (Gene
   Expression Omnibus) ([46]GSE32924, [47]GSE36842, [48]GSE107361 and
   [49]GSE130588, [50]GSE22255).[51]^14 A total 410 differentially
   expressed genes were identified, including 162 up-regulated and 250
   downregulated genes in AD patients.

FDA Approved Drugs and Their Targets

   The FDA-approved drugs for the treatment of AD were used as positive
   controls and their targets were retrieved from the Drugbank
   database.[52]^15

Construction of Signed Human Genomic Signal Network

   A KEGG signaling network was established based on KEGG database. The
   correlation scores between genes were obtained from STRING database,
   and these data were then integrated to construct a network. Active or
   inhibitory relationships were presented as either positive or negative
   links.

   The network is a weighted directed network with the weight of directed
   edges ranging from +1 to −1. The constructed network has 3892 nodes and
   28,459 directed links.

Simulation of the Impact of a Drug by SRWR

   SRWR allowed the restart of the walk in every step at the node with
   probability. SRWR algorithm was used to measure how the activation or
   inhibition of a seed node led to the activation or inhibition of other
   nodes in the human signaling network.[53]^16 The equation for the SRWR
   was defined as:
   graphic file with name Tex001.gif

   where r was the restart probability, W was the column-normalized
   adjacency matrix of the network graph, and Inline graphic was a vector
   of size equal to the number of nodes in the graph where the i -element
   had the probability of being at node i at time step t. The initial
   probability vector Inline graphic was constructed such that equal
   probabilities were assigned to the nodes representing the members of
   the disease, with the sum of the probabilities equal to 1.

Results

Identification of Potential Targets of LES by Intersection Analysis

   We collected two disease gene sets and one drug target set associated
   with AD as reference and investigated their overlap. We found that 69
   of the 751 AD disease genes in DisGeNET were differentially expressed
   in AD patients, while the target genes for anti-AD drugs in Drugbank
   dataset had 11 genes in DisGeNET. Notably, the three datasets had no
   common genes ([54]Figure 1A). [55]Figure 1B shows the overlap of target
   genes of LES with AD disease genes and target genes for anti-AD drugs
   in DrugBank database, and PTGS2, NR3C1 and SLC6A4 were the overlapped
   genes. According to the data from TCMSP, PTGS2 was regulated by more
   than 10 compounds such as caffeic acid,
   (2R)-3-oxo-2-phenylbutanenitrile, acetylshikonin, isobutyryl shikonin,
   deoxyshikonin, palmitic acid, arnebinol, alpha-methyl-n-butylshikonin,
   alkannan, lithospermidin A, Caffeate, arnebinone, isovanillin,
   β-acetoxyisovalery lshikonin, lithospermidin B, shikonofuran B.

Figure 1.

   [56]Figure 1
   [57]Open in a new tab

   Overlapping of different gene sets. (A) Overlapping of disease-related
   genes in DisGeNET (Disease genes), DEGs in GEO dataset (Diff genes), AD
   drug target genes in DrugBank database (DrugBank targets). (B)
   Overlapping of disease-related genes in DisGeNET- DRUG target genes in
   DrugBank database, and LES target genes (LES targets).

PPI Network Analysis of the Targets of LES

   We constructed a compound–compound target-AD target PPI network,
   including 77 nodes (23 compounds, 37 intersection targets between
   compound targets and AD targets, and 17 AD targets) and 441 edges
   ([58]Figure 2). Nine nodes with an average value of degree ≥25 could be
   considered as major nodes, including Interleukin-6 (IL-6), PTGS2,
   nuclear receptor subfamily 3, group C (NR3C1), estrogen receptor 1
   (ESR1), interleukin-1β (IL-1β), androgen receptor (AR), interleukin-10
   (IL-10), interleukin-4 (IL-4) and heat shock protein 90 alpha family
   class A member 1 (HSP90AA1).

Figure 2.

   [59]Figure 2
   [60]Open in a new tab

   Compound-compound target-AD target human protein PPI network. Green
   circles indicate compounds with effects on intersection targets between
   compound targets and AD targets. Red circles indicate the intersection
   targets between compound targets and AD targets. Blue circles indicate
   targets related to AD.

Prediction of Genes Significantly Inhibited and Activated by LES

   Anti-AD drugs from DrugBank were used to validate the effects of the
   drugs. There were 36 targets of DrugBank anti-AD drugs, and they were
   divided into three groups based on the corresponding drugs ([61]Table
   1). AD-related disease genes extracted from GEO datasets were used as
   controls. In [62]Table 1, we listed the genes in common between the top
   genes of each compound and the three validation gene datasets. We found
   that most of genes in the validation sets were inhibited by some
   compounds, especially Acetylshikonin, Deoxyshikonin,
   Isobutyrylshikonin, Alkannan, Caffeic acid. These results suggest that
   LES exerts anti-AD effects by inhibiting the genes associated with AD.

Table 1.

   Genes Significantly Inhibited by LES
    Drug Class or Compound  Targets^a AD Disease Genes^b AD Drugs Targets^c
   GEO Upregulated Genes
   Anti-histamine           22        2                  11                 0
   Anti-inflammatory        7         1                  2                  0
   Immunity                 5         1                  1                  0
   Caffeic acid             28        11                 1                  4
   Arnebinol                19        4                  1                  2
   Palmitic acid            26        4                  0                  1
   Acetylshikonin           28        3                  0                  8
   Deoxyshikonin            16        3                  0                  6
   Isobutyryl shikonin      18        2                  0                  5
   Isovaleric acid          15        2                  0                  4
   Lithospermidin A/B       13        2                  0                  4
   α-methyl-n-butylshikonin 15        2                  0                  4
   Lobelanidine             21        2                  2                  3
   Senecic acid             6         2                  0                  2
   [63]Open in a new tab

   Notes: ^aTargets identified by SRWR of KEGG. ^bAD disease genes
   extracted from GEO datasets. ^cAD drugs targets extracted from
   DrugBank.

KEGG Pathway Enrichment Analyses

   Pathway enrichment analysis showed that 107 KEGG pathways were
   significantly enriched with inhibited genes corresponding to LES
   targets (p<0.05), including all three anti-AD associated pathways.
   Similarly, 13 KEGG pathways were significantly enriched with anti-AD
   drugs ([64]Figure 3), including metabolism pathways (Tyrosine
   metabolism, Tryptophan metabolism, Sphingolipid signaling pathway,
   Arginine and proline pathway), Cancer pathways (Small cell lung cancer,
   Prostate cancer, Chemical carcinogenesis), PI3K-Art pathway,
   Neuroactive ligand–receptor interaction, Fc epsilon RI pathway, VEGF
   pathway ([65]Table 2).

Figure 3.

   [66]Figure 3
   [67]Open in a new tab

   KEGG analysis of major targets of LES. X-axis showed the Rich factor
   (FDR<0.01), and Y-axis showed significantly enriched KEGG pathway of
   the targets.

Table 2.

   Pathways Associated with Major Targets of LES
   Pathways Gene Compounds
   Metabolism Tyrosine metabolism MAOB, GOT1, ADH1C, MAOA, ADH1B, ADH1A
   Isovanilli, Isovaleric acid, caffeic acid
   Tryptophan metabolism MAOB, MAOA, CYP1A1 Isovanilli, caffeic acid
   Sphingolipid signaling pathways OPRD1, PRKCB, BCL2, RAC1, MAPK14, CTSD
   Lobelanidine, caffeic acid, Palmitic acid, Alkannan,
   Arginine and proline pathway GATM, MAOB, GOT1, MAOA Isovanilli,
   Isovaleric acid
   PI3K-Art signaling pathway COL1A1, GSK3B, BCL2, HSP90AA1, RXRA, RAC1,
   CHRM1, CDK2, KDR Palmitic acid, Alkannan, α-methyl-n-butylshikonin,
   Arnebinol, caffeic acid
   VEGF signaling pathway PRKCB, KDR, RAC1, MAPK14, PTGS2 Caffeic acid,
   Arnebinone, α-methyl-n-butylshikonin, Alkannan
   Fc epsilon RI signaling pathway PRKCB, BTK, RAC1, MAPK14 Caffeic acid,
   Alkannan
   Calcium signaling pathway CHRM3, CHRM1, PRKCB, ADRA1D, ADRB1, ADRB2,
   ADRA1B Arnebinol, Caffeic acid, Alkannan, Lobelanidine,
   [68]Open in a new tab

   Notably, 3 signaling pathways were associated with cancer, indicating
   that LES has the potential to treat diverse cancers such as lung cancer
   and prostate cancer as reported previously.[69]^17–19

   Phosphoinositide 3-kinase (PI3K) acts as a key signaling protein
   in T cell activation, proliferation, and migration. The major
   components of LES may inhibit IL-6 and IL-10 through PI3K/Akt pathway.
   Therefore, we hypothesized that PIK3-Akt pathway may mediate the
   effects of LES on AD. Overproduction of VEGF in AD lesions has been
   known as a potential marker of AD severity.[70]^20 Major compounds
   (Caffeic acid, Alkannan, Arnebinone) may inhibit VEGF signaling
   pathways through MAPK14 and PTGS2.

   FcεRI is the high-affinity receptor of IgE.[71]^21 Omalizumab, cyclic
   peptides, and some small proteins of FcεRI signaling pathway are drug
   candidates against AD In this study, Caffeic acid and Alkannan have the
   potential to be candidate natural drugs to bind PRKCB, BTK, RAC1, and
   MAPK14 to inhibit FcεRI signaling pathway. Inflammation suppressing
   role of tacrolimus could be related to its effects of inhibiting T cell
   activation and blocking calcineurin signaling.[72]^22^,[73]^23
   ORAI1/NFAT-calcium pathway may regulate TSLP in AD.[74]^24 Combined
   with the PPI network, the results indicate that Arnebinol, Caffeic
   acid, Alkannan, Lobelanidine may inhibit TSLP through
   ORAI1/NFAT-calcium pathway.

Discussion

   Using bioinformatics approach, we systematically investigated anti-AD
   mechanism of LES. We retrieved 23 active compounds from LES and
   predicted 99 targets. Network analysis showed that LES exerted effects
   on AD by regulating such targets as IL-6, PTGS2, ESR1, IL-1β, AR,
   IL-10, HSP90AA1 and IL-4. Analyses based on GEO database and DrugBank
   suggest that the active LES-derived compounds caffeic acid, Isovaleric
   acid, Arnebinol, Alannan may inhibit PTGS2, HSP90AA1, MAPK14, which are
   key mediators involved in anti-inflammation and anti-histamine
   pathways.

   Target-pathway network analysis of potential LES targets showed that
   LES could modulate several signaling pathways implicated in the
   pathogenesis of AD, including PI3K-Akt pathway, VEGF signaling pathway,
   Fc epsilon RI signaling pathway and calcium signaling pathway. These
   pathways crosstalk to regulate inflammation, immunity, and allergy. Our
   results suggest that LES may exert anti-AD effects by inhibiting the
   proteins involved in each of the four pathways.

   This study employed a computational pharmacology approach to elucidate
   the mechanisms of action of LES. In view of the interactions between
   proteins/genes, the impact of a natural drugs on targets can spread to
   other proteins/genes through protein–protein interaction networks or
   downstream signaling pathways. Several propagation-like algorithms such
   as SRWR have been employed to identify protein targets of
   drugs.[75]^16^,[76]^25 However, most of these algorithms could not
   reflect the complete picture of drug-induced activation and/or
   inhibition on proteins. SRWR algorithm was used to simulate the spread
   of active compound-induced activation or inhibition on a group of
   targets on a signed human signaling network. One difficulty in the
   application of SRWR algorithm is the lack of a PPI network. In this
   study, we built a new network combining KEGG and STRING databases as a
   background network to conduct SRWR.

   Using SRWR algorithm, we identified the top 5% genes inhibited by LES
   and predicted the targets of active LES compounds with the highest
   confidence scores. Using this strategy, we identified LES targets with
   a low false-positive rate. However, this study has limitations. First,
   this study is based on online datasets. The detailed mechanisms of
   action of LES on AD should be investigated by functional studies.
   Second, the constructed human signaling network may not include all
   known protein–protein interactions.

   In conclusion, we constructed a signed network by combining KEGG and
   STRING databases and ran SRWR based on the background network. Our
   results show that active LES-derived compounds, such as caffeic acid,
   Isovaleric acid, Arnebinol, Alannan may inhibit PTGS2, HSP90AA1,
   MAPK14, which are key mediators involved in PI3K-Akt pathway, VEGF
   signaling pathway, Fc epsilon RI signaling pathway and calcium
   signaling pathway. These compounds may have potential application in
   clinical treatment of AD.

Funding Statement

   This study was supported by Scientific research projects in key fields
   of Tianjin Health Commission (2022003).

Data Sharing Statement

   All data and material are available upon request.

Disclosure

   The authors declare that there are no competing interests associated
   with this work.

References