Mechanism of Astragalus Membranaceus in the Treatment of Diabetic Nephropathy Based on Network Pharmacology

Background: Diabetic nephropathy (DN), which affects more than 40% of diabetic patients, is still the main cause of end-stage renal disease in most countries. Prescription containing Astragalus Membranaceus (AM) in the treatment of early stage DN with signicant effect, but the mechanism is unclear, which we would explore based on network pharmacology. Method: First, we searched the database of China National Knowledge Internet, Wanfang Database, Chinese Biomedical Literature Database, PubMed, EMBASE, Cochrane database about the randomized, single (double) blind, controlled clinical studies of prescription containing AM in the treatment of DN, determined the effectiveness of prescription containing AM in the treatment of DN. Then, the effective components of AM in Traditional Chinese Medicine systems taxonomy database and analysis platform (TCMSP), Traditional Chinese medicine integrated database (TCMID) and Bioinformatics analysis tool for molecular mechanism of traditio Internal Chinese medicine (BATMAN-TCM) database were searched. According to the oral utilization ≥ 30% and drug-like ≥ 0.18, the effective components were screened. PubChem and health information technology (HIT) databases were searched for query validation targets and Simplied molecular input line entry specication (SMILES) of the effective ingredients, and Swisstarget prediction database was used to obtain prediction targets (possibility > 0). Drugbank, transient triple differential (TTD), and DisGeNET were searched for the targets related to DN. Finally, the data were integrated by the Cytoscape software, the drug-target-disease network was drawn, the protein interaction network was drawn with String database, and the signal pathway enrichment analysis was carried out with ClueGO. Result: The results showed that prescription containing AM could effectively reduce the urinary protein excretion rate [95% md-43.30 (- 57.00, - 29.61)]. 51 active components, 396 veried targets and 2330 predicted targets were searched. A total of 120 DN related targets were retrieved. 21 targets were found with Cytoscape. The main pathways were Interleukin-4 and 13 signaling, Activation of Matrix Metalloproteinases, EPH-Ephrin signaling. Conclusion: Prescription containing AM could effectively reduce the urinary protein excretion rate of DN patients. We speculated that AM mainly treated DN by regulating Angiotensin-converting enzyme (ACE), Vascular endothelial growth factor A (VEGFA), Janus Kinase 1 (JAK1), and interleukin-4 (IL-4) and interleukin-13 (IL-13) signaling, activation of matrix metalloproteins, ephrin signaling and other signaling pathways based on network pharmacology, which would be veried by animal experiments or in vitro experiments in the later.

according to the Mogensen diagnostic criteria of DN, the urinary protein excretion rate was selected as the main data for analysis to determine the effectiveness of prescription containing AM in the treatment of DN.
The key word of search was "Astragalus Membranaceus or huangqi". According to the characteristics of pharmacokinetics, the active ingredients were searched with the oral availability ≥ 30% and the drug like ≥ 0.18.
1.2 Acquisition of veri ed targets and Simpli ed molecular input line entry speci cation (SMILES) of active components of AM PubChem (https://pubchem.ncbi.nlm.nih.gov/, 2019 Jan 8) (21)and health information technology (HIT) database (http://lifecenter.biosino.org/hit/) (22) were used to obtain the validation targets of active ingredients of AM, and the SMILES of active ingredients from PubChem database.

Predicted targets of effective components of AM
Swisstarget prediction (http://www.swisstargetprediction.ch) (23) was used to predict drug targets based on ligand structure, which simulated the binding of receptor and ligand by computer and predict their a nity. Firstly, the SMILES of the effective components of AM was collected from PubChem database. For components that cannot get the SMILES directly, ALOGPS (http://www.vccllab.org/lab/logips/, version: 2.1) was used to calculate (24). In the Swisstarget prediction database, we selected the species "Homo sapiens", then inputed the smile structure to save the output le as a CSV format le.

Gene information standardization
The obtained targets information of drugs and diseases were normalization in UniProt database (https://www.uniprot.org/, 2019) (28), which was a protein database with the most abundant information and resources. Finally, the network of "drug-target-disease" was constructed with Cytoscape (http://cytoscapeweb.cytoscape.org/, version: 3.2.1) (29).

Results
2.1 Results of prescription containing AM reducing urinary albumin excretion rate 7 high-quality randomized controlled trials (RCTs) were included in the treatment of DN with prescription containing AM. The basic information included the author (date of publication), research methods, diagnostic criteria (diabetes, DN), intervention drugs, composition of prescription containing AM, duration of intervention, number of patients included and results of urinary protein excretion rate ( Table 1). The results of one study was not included in the statistics because of the inconsistent result, which showed that prescription containing AM could effectively reduce the urinary protein excretion rate (32). The total results showed that prescription containing AM could signi cantly reduce the urinary protein excretion rate [95% MD-43.30 (− 57.00, − 29.61)] (Fig. 2).

Active ingredients of AM
A total of 51 active components of AM were retrieved, including Adenine, Calycosin, Astragaloside VII, and the SMILES of active components was recorded ( Table 2).

Veri cation and prediction targets of AM
According to the screening criteria, 396 veri cation targets, and 2330 prediction targets were searched in the Swiss target prediction database.

Protein Interaction
The STRING database was used to construct the interaction between proteins, in which the red line represented the evidence of fusion, the green line represented the evidence of proximity, the blue line represented the evidence of coexistence, the purple line represented the experimental evidence, the yellow line represented the evidence of text mining, the light blue line represented the evidence of database, and the black line represented the evidence of co expression. (Fig. 4)

Signal pathway enrichment results
ClueGO was used to analyze the signal pathway enrichment of common targets of AM and DN, the results showed that the signal pathways such as interleukin-4 and 13 signaling, activation of matrix metalloproteases, ephrin signaling are most related to AM treatment of DN (Fig. 5). Among them, AM was most likely to play a therapeutic role by acting on Interleukin-4 (IL-4) / Interleukin-13 (IL-13), Matrix metalloproteinase (MMPs), autoantigen / Tumor necrosis factor -α (TNF -α), collagen degradation, c-Jun N-terminal kinase (JNK) and p-38 mitogen-activated protein kinase (p38MAPK) pathway (Fig. 6). found. A total of 120 DN related targets were retrieved. 21 targets were found to be shared by DN and AM with Cytoscape. The related targets mainly involved in interleukin-4 and 13 signaling, activation of matrix meta Lloproteinases, ephrin signaling and other signaling pathways. Therefore, AM may play a therapeutic role through these targets, which providing an alternative treatment for DN.
For interleukin-4 and 13 signaling, IL-4 and IL-13 receptors were expressed in human cells, and constantly changeed in the course of disease, which was related to cell in ammation and cell proliferation (39). For activation of matrix metalloproteases, MMP was involved from embryonic development to apoptosis of cells, which was closely related to in ammation and atherosclerosis. The MMP modulator could be used as a preventive treatment for individuals with risk of atherosclerosis and vascular weakening (40). Eph / ephrin signaling was a two-way signaling pathway, which include in developmental and homeostatic environments. The interaction between Eph / ephrin signaling and cells and tissues in the urinary system was also important in the damage response of podocytes and glomeruli (41). These studies showed that AM may play a therapeutic role by regulating chronic in ammation of DN, improving microvascular circulation and maintaining cell function.
At present, there is no study to predict the possible targest of AM for DN through network pharmacology. A pharmacological study showed that AM could reduce proteinuria, reverse glomerular ltration and improve early or other stages of DN in streptozotocin induced diabetic animal models, which may be related down regulating the expression of TGF-β1, Smad2/3 and α-SMA through affecting TGF-β/SMADs signal pathway, delaying renal brosis in diabetic mice (16). In clinical practice and statistics of clinical research, we found that prescription containing AM could signi cantly reduce the urinary protein excretion rate of DN patients. Based on the theory of system biology, we applied the network analysis of biological system, and with more comprehensive potential mechanism, such as AM may act on IL-4 / IL-13, MMPs, autoantigen/TNF-α, collagen degradation, JNK and p38MAPK pathway to treat DN, which provides guidance for the later research direction.
Only high-quality clinical studies were selected in the included our study, which may lead to publication bias. At the same time, the results of AM and DN targets were seached from the currently recorded data in the database, which may also lead to publication bias. At the same time, animal experiments and in vitro studies are still lacking.

Conclusion
Prescription containing AM could effectively reduce the urinary protein excretion rate of DN patients, which may be used as alternative treatment for DN, and Availability of data and materials The data used to support the results of this study can be obtained from the corresponding author.
Author contributions FML designed the protocol. XDA, LYD, DJ, RRZ, and YYD carried out the active ingredient, and relevant targets search. FML and XDA contributed to data extraction and results analysis. YRZ and SHZ corrected the related data. All authors approved the nal version of the manuscript.

Not Applicable
Funding section    Signal pathway enrichment Figure 6