Revealing active components, action targets and molecular mechanism of Gandi capsule for treating diabetic nephropathy based on network pharmacology strategy

Background Gandi capsule is a traditional Chinese herbal formula used to promote blood circulation and removing blood stasis in clinical. Our previous study has shown that it reduces proteinuria with routine treatment in diabetic nephrophy (DN), but its pharmacological action mechanism is still unknown. Methods To facilitate the identification of components, a component database of Gandi capsule and target database of DN were established by ourselves. The components absorbed in blood circle were identified in rat plasma after oral administration of Gandi capsule by UHPLC-QQQ-MS/MS. The potential targets were screened by using Libdock tolls in Discovery studio 3.0. Then Pathway and Network analyses were used to enrich the screened targets. The possible targets were verified by using a surface plasmon resonance (SPR) test and the molecular mechanism focusing these targets for treating DN was clarified by western blot. Results Six components in Gandi capsule were identified detected in rat plasma after oral administration by UHPLC-QQQ-MS/MS. After molecular docking analyses in KEGG and Discovery studio, four protein targets including HNF4A, HMGCR, JAK3, and SIRT1, were screened out, and proved as effective binding with baicalin, wogonoside by SPR. And the molecular mechanism was clarified that baicalin and wogonoside inhibit the effect of high glucose (HG)-induced decreased cell viability and podocin expression, and strengthen the activation p-AKT, p-PI3K, and p-AMPK. Conclusion Baicalin and wogonoside were screened out to be the active compounds in Gandi capsule and can ameliorate HG-induced podocyte damage by influencing the AMPK and PI3K-AKT signaling pathways by binding with HNF4A, HMGCR, JAK3, and SIRT1. Graphical abstract Supplementary Information The online version contains supplementary material available at 10.1186/s12906-020-03155-4.


Background
Diabetic nephropathy (DN) is a common diabetic complication that can lead to end-stage renal disease [1], also causes glomerulonephritis [2]. The morbidity and mortality related to DKD has rapidly increased over the past 20 years [3]. According to previous epidemiological data, DKD could be the primary cause for the progression to dialytic end-stage kidney diseases in the urban population in China [4]. The common clinical features of DN are proteinuria, progressive renal impairment, hypertension, and edema [5]. In clinic, hypoglycemic agents, lipid-lowering drugs, and anti-vasoactive drugs are the conventional medicines for DN [6]. Some of these drugs, including the sodium-dependent glucose transporter 2 (SGLT-2) inhibitor and angiotensinconverting enzyme inhibitor (ACEI), have been shown to treat proteinuria [7][8][9].
Gandi capsule (GDC) is derived from traditional Chinese medicine (TCM) prescription and used to promote blood circulation and assist the regular therapy of DN [10]. The GDC is composed of eight varieties of traditional Chinese herbs: Astragali Radix (Astragalus mongholicus Bunge., Huangqi), Corni Fructus (Cornus officinalis Siebold & Zucc, Shanzhuyu), Rehmanniae Radix (Rehmannia glutinosa Gaertn., Dihuang), Leonuri Herba (Leonurus japonicus Houtt., Yimucao), Sophorae Flos (Styphnolobium japonicum (L.) Schott., Huaihua), Scutellariae Radix (Scutellaria baicalensis Georgi., Huangqin), Bombyx Batryticatus (Bombyx scindicus (Stocks) I.Riedl., Jiangcan), and phyllanthi fructus (Phyllanthus emblica L., Yuganzi) and full information of GDC is shown in Table S1. In our previous study, GDC reduced proteinuria in combination with routine medicinal therapy during a 4 month observation period [10]. Through metabonomic analysis on blood and urine of diabetes patients treated by GDC for 6 months, it was found that the levels of urinary albumin excretion was decreased and the levels of estimated glomerular filtration rate (eGFR) was increased after treatment of GDC [11]. Moreover, we have analyzed the compounds in GDC by UHPLC-QQQ-MS/MS in our previous research [12]. Considering the complex of components in GDC, research on the pharmacological mechanism of GDC is still a challenge. There is a urgent need to clarify the active components, action targets and molecular mechanism.
Network pharmacology is based on systems biology and uses bioinformatic methods [13]. It integrates the advantage of systems biology at the level of the organism, is applicable to the analysis of complex mixtures, and has been widely used for the analysis of TCMs [14], especially in drug discovery [15], synergistic pharmacology [16], and compatibility mechanism analysis [17]. In our previous study, we used surface plasmon resonance (SPR) technology to discover the interactions of molecules and protein targets in vitro, which is a tool used to observe the interactions and dynamic changes between biological molecules. When a molecule binds to a protein on the surface of a medium, the refractive index of the medium will change, which can be transferred to an electrical signal. After the analysis of this signal, a series of parameters are displayed to evaluate the combination [18].
In this study, based on the clinical efficacy and component analysis of the Gandi capsule, a predicted analysis method was applied to screen active components and their potential targets via network pharmacology and structural biology, and further clarify the molecular mechanism. The results could provide a better understanding of the pharmacological mechanism of GDC in the treatment of DN.

Experimental animals
All experimental procedures complied with the Guidelines on the Care and Use of Laboratory Animals (National Institutes of Health Publication no.85- 23, revised 1996)  All animals were fed in a SPF environment in the Laboratory Animal Center of Xinhua Hospital affiliated to Shanghai Jiaotong University School of Medicine maintained at the temperature of 25 ± 2°C and relative humidity of 45 ± 5% and provided with water and food freely.

Establishment of component database and target protein database
According to the composition of the Gandi capsule, we built a compound database with 136 components using a Traditional Chinese Medicine and Active Ingredient Database (http://www.organchem.csdb.cn/). After analyzing the oral bioavailability (OB) and drug-likeness (DL) in the traditional Chinese medicine systems pharmacology online database (http://lsp.nwu.edu.cn/). Filter parameters were OB > 30% and DL > 0.18. Finally, 97 components were considered for this study.
We also founded a Target Protein Database to contain the protein targets related to DN. These targets are divided into two parts: confirmed targets and predicted targets. Confirmed targets are from the online databases the Therapeutic Target Database (TTD) (http://bidd.nus. edu.sg/group/cjttd/) and DisGeNET (http://www.disgenet.org/) by searching the keyword "diabetic nephropathy". Predicted targets came from microarray data GSE30528, which includes 9 DN samples and 13 diabetes mellitus samples. This data is from the Gene Expression Omnibus (GEO) database (https://www.ncbi. nlm.nih.gov/geo/). To select more accurate targets, we set the filter parameter to P ≤ 0.05 and log fold change (logFC) (differential multiple) ≥ 2 or logFC ≤ − 2. Finally, we synthesized two different targets for diabetes mellitus from the target protein database.

Analysis of components absorbed in blood by UHPLC-QQQ-MS/MS
Before experiment, rats were fasted 12 h before experiment and water was taken freely. And then the contents of the GDC were made into a suspensions with sterile water-5%Tween 80.Total 12 rats were administered with 0.27 g of drug suspensions per 1 kg of bodyweight or sterile water randomly by way of stomach, each group had six samples. This dose was identified based on the clinic dosage (0.3 g/per capsule* 3 capsules/per time* 3 times/ per day) of GDC in patients weighting 60 kg. And according to the result of pre-expreiment, 2 h later, after all rats receving CO 2 gently and sacrificing, blood samples were collected by heart exposing and punctures. Then the addition of 500 μL of acetonitrile and 10 μL of internal standard per 90 μL, the mixture was vortexed for 30 s and centrifuged at 13,000 rpm for 4 min. Only the supernatant was used. The standard powders, morroniside, loganin, baicalin, wogonoside, wogonin, and sweroside, were purchased from Nature Standard Co., Ltd. (Shanghai, China). Every drug powder was high purity (≥ 98%). The standard samples were added to a blank sample and operated by the same process as the capsule treated rats. The concentration of each standard compound in the mixture was 500 ng/mL.
To accurately identify the compounds, we compared the samples with standards via UHPLC-QQQ-MS/MS, which consisted of a Waters UPLC system with an Applied Biosystem 5500 QTRAP® hybrid triple-quadrupole mass spectrometer (Applied Biosystems/MDS Sciex, Foster City, CA, USA). This system was equipped with a turbo ion spray source. Chromatographic separation was performed on a ZORBAX Eclipse Plus C18 (50 mm × 2.1 mm, 1. Nitrogen was used in four different ways: the curtain (CUR) gas, nebulizer (GS1), heater (GS2), and the collision activation dissociation (CAD) gas. Firstly, all targeted analytes were tested in the multiple reaction monitoring (MRM) acquisition mode by MS/MS for simultaneous detection and two precursor-to-product ion transitions were monitored for the two compounds. Then, all source settings and instrument parameters for each MRM transition were optimized for the generation of as much protonated analyte molecules ([M + H] + ) as possible for each targeted compound to efficiently produce its characteristic fragment/product ions. The ion spray needle voltage was set at 5500 V and the source temperature was 550°C.
The following setting parameters were optimized; electrospray ionization source was operated with the CUR, GS1, and GS2 set at 40, 35, and 35 psi, respectively; independent precursor-product ion transitions specific to each analyte that depended on the instrumental parameters of the compound, such as the precursor ion, two product ions, declustering potential (DP), entrance potential (EP), collision energy (CE), and collision cell exit potentials (CXP); and the parameters were 50 V, 19 eV, 10 V, and 12 successively for the lactones. To detect osthol, the transition from precursor ion [M + H] + at m/z 233.2 to product ion at m/z 187.1 was monitored. As the transition from the precursor was m/z 245.1 to the product at m/z 189.1, the dwell time was set at 550 ms.

Molecular docking
To establish the correlation between drug molecules and targets, we used the docking software in Discovery Studio 3.0. Firstly, all target names were standardized using the data from the Uniprot database (https://www.uniprot.org/). Then, by selecting the minimum value from each resolution, a parameter used to describe the difference between the real and virtual target structures, the best target structures were obtained. All target structures were crystal structures determined by X-ray. This provided another filter for our study. A total of 373 protein targets and their structures were imported into Discovery Studio 3.0. After initial preparation, all structures were optimized by hydrogenation, water removal, or structural simplification. The canonical simplified molecular-input line-entry system data were imported into the software to build the molecular structures. Finally, every molecule was docked with the targets above to acquire the docking parameter X libdockscore . In addition, the ligand of each target was docked to acquire the parameter X ligand . All operations were performed under the CHARMM force field.

Kyoto encyclopedia of genes and genomes (KEGG) enrichment analysis
To reduce the false-positive result of docking, two calculation formulas were applied. One was "D-value = X libdockscore -X ligand and D-value ≥ 0", which was applied to confirm the differential binding affinity of molecules with targets compared to that with the ligand. The other was " − X libdockscore = Σ(X libdockscore )/(number of targets) and X libdockscore ≥ − X libdockscore ". This formula filtered the term near the critical value. A calculation was performed for every molecule. After analyzing the docking data, all targets were subjected to KEGG enrichment analysis. This step was achieved using the online analysis website David (https://david.ncifcrf.gov/). We acquired enrichment pathways easily using the Function Annotation module. All enriched pathways were displayed in turn by comparing the P values.

Network building and analysis
To establish the connection of molecules, targets, and pathways all selected terms were imported into Cytoscape 3.7 based on the results of molecular docking and KEGG enrichment analysis. Following the network analysis, an attached software analysis module was applied. Afterward, the parameter "Neighborhood Connectivity" (NC) was applied. As for the targets and pathways, if their NC was more than that of others, it was considered to be more enriched.

Target validation by SPR assay
To verify the effectiveness of the above method, 11 representational molecule-target pairs were chosen for the SPR test. We installed 3 regulars to choose these 11 molecule-target pairs. The first one was 'greater value of "Neighborhood Connectivity"'. The second one was 'greater value of "X libdockscore "'. The last one was 'target that docked with more molecules'. In the SPR test, the solution of protein standards was prepared and the concentration of the protein solution was 500 μg/mL. After combining with chips, a series of concentrations of compounds, which were double diluted step to step, were tested for their binding affinity to the proteins. Finally, the binding level between the protein and molecule was determined by the parameter Kd value.

Cell line and cell culture
Human podocyte cell line was obtained from XYBiotechnology (Shanghai, China) and was cultured as previous research [19]. Briefly, cells were cultured in a Roswell Park Memorial Institute (RPMI) 1640 medium (Hyclone, Shanghai, China) mixed with 10% fetal bovine serum (FBS) and 1% penicillin-streptomycin at 5% CO 2 and 37°C. Before the experimental operation, the cells were cultured in RPMI 1640 medium supplemented with 10 U/mL recombinant interferon gamma at 33°C for 2 weeks after seeding in culture plates. Only the differentiated cells were used in our research.

Western blot analyses
After incubation with the indicated conditions, the cells were washed twice with 1× phosphate-buffered saline and lysed with radioimmunoprecipitation assay lysis buffer. After vortexing 15 s and centrifugation in 14,000 g, the extracted protein content was detected using a bicinchoninic acid protein assay kit. Afterward, various groups of protein were electrophoresed using 10% sodium dodecyl sulfate-polyacrylamide gel electrophoresis and transferred onto a nitrocellulose membrane via electroblotting. To prevent non-specific binding, the membrane was blocked with 5% non-fat milk dissolved in 1× Tris-buffered saline and 0.1% Tween 20. Then, the membranes were incubated with primary antibodies against rabbit podocin, PI3K, p-PI3K, AKT, p-AKT, AMPK, and p-AMPK at 4°C overnight. The membrane was then washed with 1 × TBST three times and incubated with DyLight™ 680 conjugated secondary antibodies (CST, Massachusetts, USA) for 1 h. Finally, after washing three times with 1 × TBST, the membrane was scanned.

Statistical analysis
All experiments were repeated for three times. Statistical analysis was conducted using GraphPad Prism 7.0 software (La Jolla, CA, USA). The Student's t test was used to examine the differences between two groups. Differences were considered significant if P < 0.05.

Construction of molecule database and protein database
There were 97 components in the filtered molecule database (shown in Table S2). After comparison with 52 ingredients in Gandi capsules in vitro which we detected before [12], only 12 potential components were used, as shown in Table 1. As for predicted targets, there were 119 genes that changed after filtration, which means that these protein targets might be related to DN. The differential expression of these genes is shown in Fig. 1a. Additionally, there were 209 confirmed targets in TTD and 724 in Disgenet. After removing duplicates, there were 918 targets in the target protein database and the source constitution of the target protein database is shown in Fig. 1b.

Identification of components absorbed into blood circle via UHPLC-QQQ-MS/MS
Both positive and negative ion detection modes responded well in quadrupole full scans by a 100 ng/mL tuning solution in methanol, which was achieved in positive ionization mode for isoalantolactone, alantolacton. The protonated molecular ions revealed peaks of morroniside, loganin, baicalin, wogonoside, wogonin, sweroside and internal standard (IS) at m/z 451.0, 408. An acetonitrile-water mobile phase system from 30 to 95% acetonitrile without any acid in 3.10 min was applied because it can enhance the mass spectrum response of isoalantolactone and alantolactone. Zorbax Eclipse Plus-C18 (2.1 × 50 mm, 1.8 μm) analytical column was selected to produce more satisfactory chromatography results than other commercially available C18 columns, such as Poroshell 120EC-C18 (3.0 × 50 mm, 2.7 μm) and Acquity UPLC BEH C18 (2.1 × 50 mm, 1.7 μm). Three column temperatures (30°C, 40°C, and 50°C) were tested and 50°C was selected as it had the shortest retention time. Based on the above analysis, morroniside, loganin, baicalin, wogonoside, wogonin, sweroside were eluted at retention times of 3.106, 3.407, 4.592, 5.100, 6.300, and 3.490 min, respectively, and IS was eluted at 2.60 min. The parameters of target ions were set at m/z 451.0 → 179.0 for morroniside, m/z 408.2 → 229.0 for loganin, m/z 445.1 → 269.0 for After the analysis of the sample, six representational molecules, morroniside, loganin, baicalin, wogonoside, wogonin, and sweroside, were selected. The total ion chromatograms for the compounds are shown in Fig. 2a-d Table 2.

Virtual docking and compound-target-pathway network analysis
Molecular docking, KEGG enrichment analysis, and network analysis were applied. Molecular docking was a high-through method and a first filter tool for our study.
In total, 373 targets were docked with 6 compounds after X-ray structures were filtered. After 2238 docking experiments, we acquired 748 positive results of molecule-target pairs, including 141 with morroniside, 141 with loganin, 163 with baicalin, 114 with wogonoside, 54 with wogonin, and 135 with sweroside. In virtual docking, molecules bound to targets via a series of interactive forces, such as hydrogen bonds, van der Waals forces, and conjugated bonds. The docking statistical results are shown in Fig. 3a-c. These targets were filtered in the form of molecular-target pairs, then 131 targets were used in KEGG enrichment analysis. After analysis and classification, 30 pathways entered the next step as potential active pathways. These enrichment pathways are shown in Fig. 3d. Fig. 1 The molecular and protein databases. a Heatmap of microarray data "GSE30528". The filter parameter is "p ≤ 0.05 and logFC ≥ 2 or logFC ≤ -2". b The source constitution of the protein target database. TTD targets are from the online database "http://bidd.nus.edu.sg/group/cjttd/" and DisGeNET is from "http://www.disgenet.org/". Moreover, the target using the term "GEObase" is from microarray data GSE30528 By analyzing the representative pathways and targets under each term, a relationship of compound-targetpathway is displayed in a network shown in Fig. 3e. In this network, 97 nodes are marked, including 6 compounds, 61 target proteins, and 30 pathways. HNF4A, HMGCR, JAK3, and SIRT1 were outstanding according to the parameter "Neighborhood Connectivity".

Verification of the binding relationship of baicalin and wogonoside via SPR
In total, 11 compound-target pairs were enrolled in the SPR test and they were shown in Table 3, including SIRT1 with morroniside, loganin, baicalin, wogonoside, and sweroside; HMGCR with baicalin and wogonoside; HNF4A with baicalin and wogonoside; and JAK3 with baicalin and wogonoside. Eight compound-target pairs, including baicalin and wogonoside with HNF4A, HMGCR, JAK3, and SIRT1, interacted with each other in vitro. All positive binding results are shown in Fig. 4a to h and the information and parameters of this result are shown in Table 4. Based on the result of SPR and KEGG enrichment analysis, we discovered that these four targets were clustered in the PI3K-AKT and AMPK signaling pathways. Baicalin and wogonoside were the most significantly compounds in our study.  Baicalin and wogonoside enhanced cell viability in the HG condition in human podocyte cells To discover the influence of these compounds, we investigated the cytotoxicity of baicalin and wogonoside and discovered that they had little cytotoxicity to human podocytes at concentrations below 10 mg/L. This is shown in Fig. 5a and b. Furthermore, pretreating with baicalin and wogonoside at different doses offset the effect of HG on cell viability. The result suggested that HG induced the decreased cell viability; however, baicalin and wogonoside reduced this influence on human podocyte cells, as shown in Fig. 5c. We also verified this hypothesis using podocin, a marker of DN via western blotting, as shown in Fig. 5d, e.
Baicalin and wogonoside altered the phosphorylation and expression of AMPK, PI3K, and AKT in the HG condition in human podocyte cells To verified influence of baicalin and wogonoside in PI3K-AKT pathway and AMPK pathway in cells, we  used a western blot method. As a result, pretreatment with HG activated the phosphorylation of AKT and reduced the phosphorylation of AMPK and PI3K (Fig. 6a,  b). However, baicalin and wogonoside blunted the HGinduced reduction of AMPK and PI3K phosphorylation.
They also activated the expression of p-AKT directly, as shown in Fig. 6c, d.

Discussion
DN, a diabetic complication, has become a worldwide challenge due to the lack of an effective treatment. Increased proteinuria is a typical clinical manifestation of DN and is mostly due to the damage of podocytes [20].
Podocytes are specialized epithelial cells that are present in the lateral glomerular basement membrane and retain the integrity of the glomerular filtration barrier [21]. Podocytes may be an important cell for DN research. In DN, unbalanced glucose metabolism causes abnormal expression of protein kinase C-alpha and mediates the reduction of nephrin in podocytes [22]. Moreover, HG may modulate matrix-related cell adhesion and integrin subunit synthesis, thereby affecting changes in podocyte and glomerular basement membrane interactions [23].  Under the pressure of HG, progressive podocyte injury continues and leads to the degeneration of podocyte foot processes, cell apoptosis, and loss of proteins [24]. The GDC may aid proteinuria reduction in combination with common treatments and has been used as a DN therapy for many years in China. However, to date, the pharmacological mechanism of GDC has not been defined due to the complex composition of compounds. Here, we used the methods of UHPLC-QQQ-MS/MS and network pharmacology to uncover the potential active compounds in GDC and preliminarily demonstrate the results in vitro. In our study, we built a molecular database including six compounds that could be absorbed into the blood of rata according to UHPLC-QQQ-MS/MS. A target database was also built. Molecular docking, KEGG enrichment analysis, and network analysis were performed and four notable targets, HNF4A, HMGCR, JAK3, and SIRT1, were observed. The binding relationship of molecules and these four targets were demonstrated via SPR. Finally, based on the result above, baicalin and wogonoside plobablely the active ingredients in GDC. And they might act via PI3K-AKT and AMPK signaling pathways. It also was observed that baicalin and wogonoside alleviated HG-induced decreased cell viability, protein marker podocin damage and adjust the protein expression of PI3K-AKT and AMPK signaling pathways in human podocyte cells.
Baicalin is a plant-derived flavonoid that has a potential antidiabetic effect [25]. Baicalin protects kidney epithelial cells against methylglyoxal-induced cytotoxicity via anti-inflammatory and antioxidant processes [26]. Baicalin can also reduce mitochondrial damage in streptozotocin-induced diabetic rats [27]. Meanwhile, wogonoside has anti-dyslipidemia, anti-obesity, and antidiabetic effects as a traditional Chinese medicine prescription [28]. Obviously, as we know, baicalin and wogonoside had protective action in DN progress. And in SPR test, to our knowledge, four protein targets act on the progress of DN. HNF4A is a transcription factor that regulates glucose metabolism and nutrient-induced insulin secretion in pancreatic β-cells in mice [29]. It can also inhibit vascular endothelial growth factor (VEGF)-mediated endothelial proliferation and migration and reduce VEGF-stimulated in vitro angiogenesis by directly repressing the transcript of FLK1 [30], which is closely correlated to the occurrence of DN. HMGCR is a therapeutic target and key enzyme in lipid metabolism [31]. Moreover, HMGCR inhibition, such as by atorvastatin [32], is one of the routine medicines for DN and can alleviate vascular oxidative stress [33]. JAK3 is an important molecular target for the treatment of autoimmune insulin-dependent (type 1) diabetes mellitus [34]. As a member of the JAK family, it also participates in the tyrosine-protein kinase Janus kinase /signal transducer and activator of transcription (JAK/STAT) pathway. The JAK/STAT pathway is involved in glucose metabolism [35] and T cell-reduced immune function [36]. SIRT1 has been shown to connect to β cell function in insulin secretion in vivo [37]. Moreover, some studies indicate that SIRT1 can alleviate diabetic albuminuria in DN and SIRT1 expression can be an early sign of DN [38]. In our study, the SPR test results indicated a correlation between baicalin and wogonoside with HNF4A, HMGCR, JAK3, and SIRT1 and DN, it might be the basis of baicalin and wogonoside's protective action for cell damage in HG-induced podocyte cells.
Considering the pathways discovered to be involved in this research, PI3K-AKT and AMPK signaling pathways have a wide range of functions in the progress of DN.