Exploring the Molecular Mechanism of Liuwei Dihuang Pills for Treating Diabetic Nephropathy by Combined Network Pharmacology and Molecular Docking

Background Diabetic nephropathy (DN) is a common and serious complication of diabetes, but without a satisfactory treatment strategy till now. Liuwei Dihuang pills (LDP), an effective Chinese medicinal formula, has been used to treat DN for more than 1000 years. However, its underlying mechanism of action is still vague. Methods Active compounds and corresponding targets of LDP were predicted from the TCMSP database. DN disease targets were extracted from the OMIM, GeneCards, TTD, DisGeNET, and DrugBank databases. Subsequently, the “herbal-compound-target” network and protein-protein interaction (PPI) network were constructed and analyzed via the STRING web platform and Cytoscape software. GO functional and KEGG pathway enrichment analyses were carried out on the Metascape web platform. Molecular docking utilized AutoDock Vina and PyMOL software. Results 41 active components and 186 corresponding targets of LDP were screened out. 131 common targets of LDP and DN were acquired. Quercetin, kaempferol, beta-sitosterol, diosgenin, and stigmasterol could be defined as five crucial compounds. JUN, MAPK8, AKT1, EGF, TP53, VEGFA, MMP9, MAPK1, and TNF might be the nine key targets. The enrichment analysis showed that common targets were mainly associated with inflammation reaction, oxidative stress, immune regulation, and cell apoptosis. AGE-RAGE and IL-17 were the suggested two significant signal pathways. Molecular docking revealed that the nine key targets could closely bind to their corresponding active compounds. Conclusion The present study fully reveals the multicompound's and multitarget's characteristics of LDP in DN treatment. Furthermore, this study provides valuable evidence for further scientific research of the pharmacological mechanisms and broader clinical application.


Introduction
Diabetes mellitus, like COVID-19, is a wicked problem [1]. It may affect 693 million adults by 2045, according to the prediction of the International Diabetes Federation [2]. Diabetes mellitus has emerged as the leading cause of diabetic nephropathy (DN), which is the leading cause of the end-stage renal disease (ESDR) [3]. Epidemiological studies have shown that more than 30% of diabetic patients may develop DN [4]. In some parts of the world, ESDR caused by DN accounts for over 50% of renal replacement therapeutic patients [5]. e current therapies strategies for DN mainly include controlling blood glucose, reducing proteinuria, and managing merging symptoms [6]. However, the understanding of DN continues to increase, and current treatment methods for DN are still not effective enough. Many patients with DN have a poor prognosis, especially those with advanced DN are still unsatisfactory [7]. erefore, novel therapeutic drugs for DN are urgently required. Fortunately, Liuwei Dihuang pills (LDP) may be a potential complementary and alternative therapy medicine for DN.
LDP, a classical prescription first described in Xiaoer Yaozheng Zhijue, has been used to treat DN for more than 1000 years by Chinese people. Qian Yi formulated LDP during the Song dynasty. It is composed of 6 herbs, including Rehmanniae Radix Preparata (Shudihuang), Cortex Moutan (Mudanpi), Rhizoma Dioscoreae (Shanyao), Cornus officinalis (Shanzhuyu), Alisma (Zexie), and Poria Cocos (Fuling). LDP is considered to have the efficacy of nourishing the yin and kidney. In recent years, more and more researchers' interest has been focused on LDP to treat DN. Xu et al. [8] have reported that LDP can protect glomerular mesangial cells and prevent renal fibrosis in the treatment of rats of DN. Lin et al. [9] have documented that Western medicine has better therapeutic efficacy in treating DN when combined with LDP. Shi et al. [10] also have found that LDP has a practical therapeutic effect on diabetic nephropathy. A metaanalysis including 14 RCT studies and 918 study participants has reported that LDP categorized formulas are safe and effective in treating DN proteinuria [11]. However, because of the characteristics of multicompound and multitarget, exploring the underlying molecular mechanism of traditional Chinese medicine (TCM) through cellular or animal studies is relatively tricky [12]. Until now, the pharmacological mechanism of LDP in DN treatment is still vague, which greatly limits its extensive application.
Because of the holism concept of TCM, Chinese medicinal formulas are usually treated diseases through several components and targets rather than a single one [13]. Network pharmacology can abstract the interaction relationship of drugs and targets into a network model and study them via a holistic perspective, consistent with the concept of holism in TCM [14]. In recent years, it has been recognized as an efficient method to study TCM. erefore, our study was set out to reveal the mechanism of LDP in treating DN via the pharmacology network and molecular docking combination approach. It can provide valuable evidence for further basic research and clinical applications. e flowchart of network pharmacology research on LDP in treating DN is shown in Figure 1.

Herbal Compounds and Corresponding Targets in LDP
Extraction. Traditional Chinese Medicine System Pharmacology Database (TCMSP, https://tcmspw.com/tcmsp.php/ ), a systematic pharmacology database, provides information about the active herbal components and related targets [15]. We searched the TCMSP database to acquire the active compounds in LDP. Oral bioavailability (OB) means the rate and percentage of pharmaceutical agents orally absorbed into the systemic circulation [16]. Drug-likeness (DL) is a concept based on existing drugs' physical and chemical properties and can be used to estimate compounds whether reaching the conditions to become new drugs [16]. e active compounds in LDP were further screened with the criteria that OB ≥ 30% and DL ≥ 0.18. en, we also screened the targets related to these herb compounds from the TCMSP database. Taking "Reviewed" and "Human organisms" for filtering conditions, we acquired the related gene symbols of these targets from the UniProt database (https:// www.uniprot.org/) [17]. In this study, the compounds that cannot find corresponding gene symbols were excluded.

"Herbal-Compound-Target" Network Construction.
After intersecting the acquired targets of formula and disease by using R language software, the common targets of LDP in treating DN were screened out. en, we used Cytoscape 3.8.0 software to construct a "Herbal-compound-target" network [23]. e relationship between the active compounds and the anti-DN targets in LDP can be visualized easily on this network.

Protein-Protein Interaction (PPI) Network Construction.
e STRING (https://string-db.org/, version 11.0) is a data platform that collects almost all known and predicted interactions between the expressed proteins [24]. We uploaded the acquired therapeutic targets to this web platform. e species parameter was set as "Homo sapiens," and the confidence score was limited to ">0.7," hide discrete targets, the PPI network was built, and the date of the network was exported.

Screening of Key Targets in the PPI Network.
Cytoscape3.8.0 software was utilized to visualize and analyze the PPI network data. We then used the Cytoscape plugin CytoNCA [25] to screen critical targets in the PPI network. Betweenness centrality, closeness centrality, and degree centrality were chosen as the parameters to calculate topological features of the PPI network. Betweenness centrality was used to assess how much the shortest paths must pass through a given node [26]. Nodes with higher betweenness can be understood as connected to many nodes transmitting information [27]. Closeness centrality indicates the average distance between the nodes in the network [28]. Degree shows the number of edges to node [29]. Nodes with larger degree values are more crucial in the network [30]. e median of the three parameters was set as the thresholds for filtering central nodes to screen key targets.
2.6. Enrichment Analysis. Metascape (http://metascape. org) is an enrichment analysis platform updated monthly rather than longer to keep the data up to date [31]. To further illustrate the underlying mechanism of LDP for treating DN, we uploaded the acquired 131 common targets to Metascape online platform. en, we performed GO and KEGG enrichment analyses on this platform. e "Min Overlap � 3," "Min Enrichment � 1.5," and "P value cutoff <0.01" were set as significant thresholds. Biological processes (BP), cellular components (CC), and molecular function (MF) are included in GO terms. After analysis, the top 10 terms of BP, CC, MF, and KEGG pathways were chosen to be visualized.

Molecular Docking.
e binding capability of nine key targets and their corresponding active compounds were evaluated by molecular docking. To ensure the accuracy of the data, we used the UniProt database to find the UniProt ID of the key targets. According to the Uniport ID, the 3dimensional (3D) protein structures related to the nine key targets were downloaded from the RCSB PDB online tools (http://www.rcsb.org/) [32]. e water molecules and small molecule ligands of the 3D protein structure were removed by PyMOL 2.4.0 software. e 2-dimensional (2D) structures of the compounds corresponding to the nine key targets were acquired from the PubChem database (https:// pubchem.ncbi.nlm.nih.gov/) [33]. en, the 2D structures of active compounds were converted into 3D structures by

Target Prediction.
In total, 186 corresponding targets of active compounds in LDP were screened out (Supplementary Table S1). e number of related targets in Rehmanniae  (Figure 2(a) and Supplementary Table S2). Finally, 131 common targets of LDP and DN were obtained by intersecting the acquired targets ( Figure 2(a), Supplementary Table S3).

"Herbal-Compound-Target" Network Construction.
A "herbal-compound-target" network ( Figure 3), including 172 nodes (41 active compound nodes of herb and 131 common target nodes) and 293 edges, was constructed. We obtained the "degree" parameter of the herbal-compoundtarget network by using an analysis network tool in Cytoscape software. After the network analysis, we found that the MOL000098 (quercetin, degree � 104), MOL000422 (kaempferol, degree � 42), MOL000358 (beta-sitosterol, degree � 16), MOL000546 (diosgenin, degree � 16), and MOL000449 (stigmasterol, degree � 13) were the top 5 compounds in the 41 active compounds. ese five compounds may play the most significant role in treating DN. is result indicates that LDP played anti-DN roles mainly through these compounds. Overall, the relationships among herbs, active compounds, and disease targets can be observed through the network graph.

PPI Network Construction and Key Targets Analysis.
For the purpose of further elucidating the mechanisms of LDP treatment of DN in vivo, we introduced 131 common targets into the STRING online service platform, and the date and figure of a PPI network were acquired (Figure 4). e PPI network data were visualized and analyzed via the "Analysis network" tool in Cytoscape 3.8.0 software. 122 nodes and 914 edges were included in the PPI network ( Figure 5(a)). We tried to find the key targets by the topological features and set the median of betweenness, closeness, and degree centrality as the screening criteria. e thresholds of the first screening were betweenness ≥ 38.641242955, closeness ≥ 0.445673703, and degree ≥ 11. A new network including 43 nodes and 440 edges was acquired after the first screening ( Figure 5(b)). Subsequently, betweenness ≥ 13.17520385, closeness ≥ 0.65625, and degree ≥ 20 were set as the thresholds of second screening. A center network that includes 20 nodes and 155 edges was constructed ( Figure 5(c)). For the purpose of finding the most critical targets in the PPT network, betweenness ≥ 3.3646742145, closeness ≥ 0.826086957, and degree ≥ 15 were set as the last screening thresholds. Eventually, JUN, MAPK8, AKT1, EGF, TP53, VEGFA, MMP9, MAPK1, and TNF were screened as the key targets of the PPI network ( Figure 5(d)).

Enrichment
Analysis. GO and KEGG enrichment analyses of 131 common targets were performed on the Metascape data platform. is study found 2231 GO terms, including 2008 BP terms, 83 CC terms, and 143 MF terms. e top 10 most important terms of BP, CC, and MF are shown in Figure 6(a). BP terms mainly include response to the apoptotic signaling pathway, oxygen levels, lipopolysaccharide, wounding, organic cyclic compound, and nitrogen compound. CC terms were mainly enriched in membrane raft, protein kinase complex, vesicle lumen, transcription factor complex, plasma membrane protein complex, ficolin-1-rich granule lumen, organelle outer membrane, and cell body. MF terms mainly involved cytokine receptor binding, antioxidant activity, protein domain specific binding, transcription factor binding, nuclear Evidence-Based Complementary and Alternative Medicine receptor activity, protein homodimerization activity, repressing transcription factor binding, and protein kinase binding. 339 pathways were acquired through the KEGG pathway enrichment analysis. e top 10 most important terms of the KEGG pathway enrichment analysis were chosen for the visual analysis (Figure 6(b)). e potential targets of LDP in treating DN were principally enriched in pathways in cancer, AGE-RAGE signaling pathway in diabetic complications, and L-17 signaling pathway. Besides, MAPK, HIF-1, platinum drug resistance, cellular senescence, hepatitis C, and leishmaniasis pathways were also included. e AGE-RAGE signaling pathway is closely connected with inflammatory response and diabetic complications and is selected to be further visualized as an example. e pink labeled nodes are the common targets of LDP and DN (Figure 7). It suggests that the AGE-RAGE signaling pathway is significant in LDP's anti-DN efficacy. More details of the GO functional and KEGG pathway analyses results are shown in additional file 1.  Figure 3: e "herbs-active compounds-disease targets" network. e circular nodes indicate active compounds, and the square nodes indicate the possible therapeutic targets in LDP. Different colors in the circular nodes represent that active compounds are included in different herbs. e yellow is Shanzhuyu, the green is Shanyao, the blue is Mudanpi, the red is Fuling, the pink is Zexie, and the purple is Shudihuang. 6 Evidence-Based Complementary and Alternative Medicine corresponding active compounds was performed. When the binding energy is <0 kJ mol, the small molecule ligand can spontaneously bind to the protein receptor. If the binding energy is <−5.0 kJ mol or lower, it indicates that the two have the better binding ability [34]. rough docking simulations, 16 pairs of docking results were yielded ( Table 2). eir binding energy is all <−5 kJ mol, which means all of them can bind very well. e detailed information of the four best molecular docking targets and their corresponding active compounds is shown in Figure 8

Discussion
DN is a common diabetic complication that threatens the health and lives of diabetic patients. Unfortunately, most diabetic patients cannot obtain good treatment effects with routine therapies. Growing evidence has suggested that LDP may be a potential adjuvant or alternative medicine for DN [8][9][10][11]. However, the detailed mechanism of action remains obscure. As a relatively new approach in drug discovery, network pharmacology can illustrate the interaction between diseases, drugs, and targets [35,36]. To some extent, the characteristics of network pharmacology coincide with Evidence-Based Complementary and Alternative Medicine the "multi-compounds, multi-targets, and multi-pathways" theory of traditional Chinese medicinal formula. erefore, we explored the mechanism of action of the Chinese medicine formula LDP as an adjuvant treatment of DN through a network pharmacology approach. Moreover, the inner links between LDP and DN were further verified via molecular docking. e present study improves the understanding of the molecular mechanism of LDN in treating DN, which is of great importance for further basic research and clinical application.
Based on the results of the network analysis, quercetin, kaempferol, beta-sitosterol, diosgenin, and stigmasterol can be defined as the crucial compounds of LDP in treating DN. It is reported that quercetin could inhibit inflammatory cell infiltration, alleviate renal oxidative stress injury, relieve the pathological damage of the kidney, and improve renal function in DN [37]. Kaempferol has anti-inflammatory, antioxidant, and antifibrotic properties in DN [38,39]. Beta-sitosterol has been identified as a potential herbal nutraceutical for DN because it has anti-inflammatory, lipidlowering, antioxidant, and antidiabetic activities [40]. Diosgenin plays a protective role in DN through lowering oxidative stress and inflammation [41]. Stigmasterol has the function of regulating the glucose metabolism [42]. ese main LDP compounds collectively exert anti-inflammation, antioxidant, antifibrotic, antihyperglycemic, and antihyperlipidemic effects which can form a pharmacological basis for the anti-DN function of LDP.
rough the PPI network analysis, JUN, MAPK8, AKT1, EGF, TP53, VEGFA, MMP9, MAPK1, and TNF were the key targets of LDP in treating DN. ese targets are mainly connected with inflammation, vascular permeability, and oxidative stress. In some ways, this is consistent with the disease characteristics and pathogenesis of DN. To further reveal LDP's possible anti-DN molecular mechanism, we conducted molecular docking of 9 key targets with their    corresponding active compounds. Study results have shown that the nine key targets have an excellent ability to bind their related active compounds in LDP. Among them, JUN, MAKP1, and AKT1 had a more stable binding ability than others. A recent study showed that c-Jun could be progressively elevated, and it could activate the expression of TGFβ1 via ross-activation and autoregulation during renal fibrosis in DN [43]. MAPK1 can increase many inflammatory and adhesion factors in glomerular cells and exacerbate the damage in the pathological state of DN [44]. AKT1 is closely associated with the immune regulation and inflammation reaction of DN. It plays a vital role in basement membrane thickening, mesangial proliferation, and podocyte injury [45]. e GO functional enrichment analysis of the 131 common targets was carried out (Figure 6(a)). e 10 most meaningful enriched BP terms were principally associated with apoptosis, response to oxygen levels, and response to lipopolysaccharide. Related research demonstrated that the initiation and progression of DN are closely associated with apoptosis, oxidative stress, and lipopolysaccharide level [46,47]. MF terms mainly included transcription factor binding, protein homodimerization activity, cytokine receptor binding, and antioxidant activity. e targets primarily enriched in the above MF terms were JUN, TNF, VEGFA, SOD1, DPP4, and AKT1. ey are principally included in inflammatory regulation, immune response, and oxidative stress. Inflammation and immune response play essential roles in the progression of DN [48]. CC terms were mainly enriched in membrane raft, protein kinase complex, extracellular matrix, transcription factor complex, and vesicle lumen. e key targets, such as TNF, MAPK1, and VEGFA, were included in these terms. ese finds indicated that DN is very complex, and the LDP could be used to treat DN by interfering with various molecular functions and cellular components.
Associated with the GO enrichment analysis, we found out that the main pathways of LDP on DN might be the AGE-RAGE signaling pathway and IL-17 signaling pathway based on the enrichment results of KEGG. As we all know, the AGE-RAGE signaling pathway is of great significance to diabetic complications. e upregulation of AGEs levels and RAGE expression can aggravate the progression of DN [49]. When the kidney is subjected to long-term stimulation of glycosylation of reducing sugars, the AGEs are gradually accumulated and increase the risk of extracellular matrix migration, renal tubular dysfunction, and glomerular proliferative lesion. Furthermore, AGEs can bind to RAGE receptors to cause chronic inflammation reaction, oxidative stress, kidney tissue damage, and the loss of kidney function [50]. Sharma et al. have reported that AGE-RAGE interaction promotes DN's progression because of the release of fibronectin, TGF-β, and inflammatory cytokines [51]. If the IL-17 pathway is activated in many kidney diseases, it can promote inflammatory cytokines [52]. Inflammatory cytokines can cause glomerulosclerosis and kidney tissue damage in DN via inflammatory response [53]. Mohamed et al. have shown that low-dose recombinant IL-17 might prevent and reverse DN [54].
To some extent, our results are supported by the previous studies, which made them more reliable. However, there are also some limitations in this study. For example, it is difficult to ensure that the drug's active ingredients are identical to those absorbed into the patient's bloodstream; We are still vague about the interaction effects of different nodes in the network analysis. e functions and pathways highly researched may cause departures from expected results.
us, further experimental and clinical studies are warranted to verify our theoretical prediction.

Conclusion
is research proved that the therapeutic mechanism of LDP on DN might be realized by multitargets, multiactive compounds, and multipathways. We found that quercetin, kaempferol, beta-sitosterol, diosgenin, and stigmasterol can be defined as five crucial compounds. JUN, MAPK8, AKT1, EGF, TP53, VEGFA, MMP9, MAPK1, and TNF may be the nine most important therapeutic targets. AGE-RAGE and IL-17 are two key signal pathways of LDP for the treatment of DN. e potential pharmacological mechanism mainly associates with inflammation reaction, oxidative stress, immune regulation, and cell apoptosis. Also, some active compounds, target genes, and pathways in our study have few reports, which may be the clues for further research in the mechanism of LDP in treating DN. In summary, our study provides valuable evidence for further basic research and clinical applications.

Data Availability
e relevant data and code used to support the findings of this study are available from the corresponding author upon request for two years after publication. Disclosure e authors are immensely thankful to the Research Square platform for posting a preprint of an earlier version of this article to the following link: https://www.researchsquare. com/article/rs-463825/v1 [55].

Conflicts of Interest
e authors declare that they have no conflicts of interest.

Authors' Contributions
HL conceived and designed the study. XK and SC involved in technical or material support. ZL and SR searched the databases. GW, QH, and LZ analyzed and prepared the manuscript. DL, XL, and HZ provided suggestions for data analysis and figure design. HL, XK, and SC revised the final manuscript. All authors read and approved the final manuscript. Gaoxiang Wang, Lin Zeng, and Qian Huang contributed equally to this study.