A search for all chemical components of P. frutescens was conducted in the traditional Chinese medicine systems pharmacology (TCMSP) database (http://lsp.nwu.edu.cn/index.php) using “zisu” as the key word. Filter criteria included database ADME (absorption, distribution, metabolism, excretion) parameters, select oral bioavailability (OB) ≥30%, and drug-likeness (DL) ≥0.18. In addition, the search focused on components that were effective in the relevant literature but did not meet the above screening criteria to supplement. Finally, the main active components of P. frutescens were identified.

The predicted targets of the main active components were retrieved from the TCMSP database. They were entered into the Uniport database (http://www.uniprot.org/) with the species restricted to “Human.” The official gene name of all target proteins was checked one-by-one. The genes constituted the predicted targets of the main active components.

A search for target genes related to CKD stages 3–5 was performed in the Gene cards database (https://www.genecards.org/) with the keyword “chronic renal failure” (because CKD stages 3–5 are traditionally termed CRF). Due to the excessive number of CRF-related targets, a score ≥10 was used as the threshold to screen the main targets related to CKD stage 3–5 for further research.

Venny 2.1.0 (http://bioinfogp.cnb.csic.es/tools/venny/index.html) software was used to obtain the intersection of the predicted targets of P. frutescens and the main targets of CKD stage 3–5. A Venn diagram was drawn. The intersection genes were possible relevant targets for the treatment of CKD with P. frutescens.

The main active components and the corresponding CKD stage 3–5 related targets were imported into Cytoscape3.6.0 software. The main active components-targets network model was drawn. Then, the Network Analyzer plugin was used to analyze the network topology and to calculate the degree, betweenness centrality, and closeness centrality parameter values. The larger the value, the more important the node is in the network, enabling the clarification of the core components and targets of P. frutescens in the treatment of CKD.

The core targets in the network model were imported into the String11.0 database (https://string-db.org/) with species limited to “homo sapiens.” The highest confidence protein score value was >0.7. Unconnected single proteins were removed to obtain the PPI diagram of these core targets. The relevant action information of these targets was imported into Cytoscape 3.6.0 software to construct a network model and perform topological analysis, and adjust the node size, color, and edge thickness according to the value of the degree and combined score. The degree value was used as a reference to filter out the key nodes in the PPI network and draw a bar graph. At the same time, the MCODE plugin for protein modular analysis was used to set Degree Cutoff = 2, Node Score Cutoff = 0.2, K-Core = 5, Max.Depth = 100, and to screen out key protein modules in the PPI network.

The relevant targets were imported into Cytoscape3.6.0 software. The ClueGO and CluePedia plugins were used to annotate the GO biological functions of these targets, including the enrichment of cellular components (CCs) and molecular functions (MFs). The analysis was set with a p-value ≤0.01 as the criterion to screen out the main biological functions of the core targets. The network model was established and a pie chart drawn.

Target information was imported into the DAVID6.8 online analysis platform (https://david.ncifcrf.gov/) for enrichment analysis of biological processes (BPs) and signaling pathways. The criterion and background were both “homo sapiens.” The GO-BP and Kyoto Encyclopedia of Genes and Genomes (KEGG) databases were used to enrichment analyze the BPs and pathways related to the targets. The smaller the p-value, the higher the degree of enrichment. Therefore, the main BPs and pathways of P. frutescens for CKD stage 3–5 treatment were selected using the criterion of p <0.01. The enrichment results were visualized by software R language 4.0.3 and clusterProfiler package.

The three-dimensional structure of the main active components of P. frutescens was downloaded from the PubChem database (https://pubchem.ncbi.nlm.nih.gov/) and OpenBabel software was used to convert it to the mol2 format. The complex crystal structure containing the original ligand of the key targets in the PPI network from the RSCB PDB database (http://www.rcsb.org/) was imported into the Pymol software to separate the receptor protein and the ligand molecule, followed by pretreatments such as dehydration. After importing the active components and corresponding potential therapeutic targets into Pyrx0.8 software for pretreatment and energy optimization, Autodock vina was used for molecular docking. The crystal box was selected to maximize the coverage of the entire receptor and active site. The binding affinities of receptors and ligands were analyzed and the docking results were compared with those of the original ligand.

P. frutescens leaves (lot no. 200501, place of origin: Liu’an, China) was purchased from Anhui Dabie Mountain Traditional Chinese Medicine Decoction Pieces Co. Ltd (Liu’an, China), and all herbs meet the standards of the Chinese Pharmacopoeia 2015. We weighed out 1200 × g of decoction pieces of P. frutescens leaves and extracted them twice with water. Firstly, the herb was immersed in 12,000 ml water and heated for 1 h at 100°C. 9000 ml of filtrate was obtained. After filtration, boiled herb was added 8000 ml water and heated for 1 h again. Then, we got 6800 ml of filtrate. The two filtrates were mixed and concentrated to obtain P. frutescens Aqueous Extracts (PFAE). Finally, PFAE solvent was prepared into freeze-dried powder by vacuum freeze-drying method. The PFAE freeze-dried powder was dissolved in water for use in the experiments. The quality and component validation were examined with fingerprint analysis via high performance liquid chromatograph (HPLC).

A Waters e2695 HPLC (on-line degasser, quaternary pump, automatic sampler, and UV detector), with a Kh-500de ultrasonic cleaner (Kunshan Hechuang Ultrasonic Instrument), Sartorius CPA225D electronic balance (Nanjing Yimanelli Instrument and Equipment), and a C18 (250 mm, 4.6 mm, 5 m) chromatographic column were used in this analysis. Luteolin (lot no. B20888, HPLC≥98%), rosmarinic acid (lot no. B20862, HPLC≥98%), caffeic acid (lot no. B50428, HPLC≥98%), perilla aldehyde (lot no. B21699, HPLC≥98%) and (+)-catechin (lot no. B21722, HPLC≥98%) were purchased from Shanghai Yuanye Biotechnology (Shanghai, China).

200 mg of PFAE freeze-dried powder was weighed and resolved in 1 ml 80% methanol, then eddy mixed for 10 min. After centrifugation at 13,000 rpm for 10 min, the supernatant was taken for HPLC analysis. Five reference sample substances were accurately weighed and dissolved in methanol to prepare a standard solution. The concentration of each component was 0.5 mg/ml.

The chromatographic method was as follows: A Dubhe C18 chromatographic column (4.6 mm × 250 mm, 5 μm) was employed. The mobile phase was acetonitrile (A) - 0.1% acetic acid (B). Linear gradient elution: 0–17 min, 15%A, 17–20 min, 15%A, 20–25 min, 20%A, 25–30 min, 20%A, 30–38 min, 28%A, 38–40 min, 28%A, 20%A, 40–45 min, 35% A, 45–50 min, 35%A ∼ 40%A, 50–60 min, 40%A ∼ 15%A. The volumetric flow rate was 1 ml/min, the column temperature was 25 C, the detection wavelength was 330 nm, and the operation time was 60 min. After each run, the chromatographic column was balanced with the initial mobile phase for 20 min to ensure the stability of peak extraction.

10 μl of the reference solution and test solution were accurately absorbed and injected into the liquid chromatograph. The chromatographic peaks, retention time and peak area were recorded within 60 min.

We have screened and identified some components in P. frutescens through network pharmacology and HPLC. Therefore, we conducted in vitro experiments to further verify whether these components can effectively reduce kidney damage. Luteolin is a key component in network pharmacological model, and molecular docking model shows that it binds well to corresponding therapeutic targets. In addition, it has also been identified and quantified in the PFAE by HPLC. Therefore, luteolin was selected as the representative active component for subsequent in vitro experiments. According to the network pharmacological analysis, luteolin inhibits CKD progression mainly by regulating apoptosis. Previous studies have successfully established a tubular epithelial cell (TEC) injury model using adriamycin (ADR), which promotes TEC apoptosis. The model confirmed that the mechanism is closely related to oxidative stress and the mitogen-activated protein kinase (MAPK) pathway, which is an important form of renal pathology in CKD stage 3–5. Therefore, in vitro, ADR was used to disrupt renal TECs to construct a kidney injury model. Subsequent treatment with luteolin was performed to assess whether it can reduce apoptosis and regulate the expression of AKT, p53, and IL6. Oxidative stress, MAPK pathway and PI3K-AKT pathway are the most common and important ways related to apoptosis. Therefore, we further detected them to explore the mechanism by which luteolin reduces apoptosis.

Luteolin (lot no. HY-N0162, 98.15%) was purchased from MedChemExpress (Shanghai, China). ADR hydrochloride (lot no. S120812, 99.13%) was purchased from Selleck Chemicals (Shanghai, China). Fetal bovine serum (FBS) was purchased from Thermo Fisher Scientific (Waltham, MA, United States), as were phosphate buffered saline (PBS), 0.25% trypsin-EDTA, antibiotic-antimycotic (Anti-Anti, 100×), and Dulbecco’s modified Eagle medium/F-12 (DMEM/F-12). Cell Counting Kit-8 (CCK-8) was obtained from APExBIO (Houston, TX, United States). Dimethyl sulfoxide (DMSO) was purchased from Sigma-Aldrich (St. Louis, MO, United States). Antibodies to phosphorylated (p)-p38 MAPK (Thr180/Tyr182) (lot no. 4511), p-extracellular signal-regulated kinase (ERK)1/2 (Thr202/Tyr204) (lot no. 4370), p-stress-activated protein kinase/c-Jun N-terminal kinase (SAPK/JNK) (Thr183/Tyr185) (lot no. 4668), p-AKT (Ser473) (lot no. 4060), p53 (lot no. 2524), p-p53 (Ser15) (lot no. 9284), Caspase3 (lot no. 9662), and Cleaved Caspase-3 (lot no. 9664) were procured from Cell Signaling Technology (Shanghai, China). Antibodies to AKT (lot no. 10176-2-AP), p38 MAPK (lot no. 14064-1-AP), JNK (lot no. 24164-1-AP), Bcl-2 (lot no. 26593-1-AP), Bax (lot no. 50599-2-Ig), and β-actin (lot no. 66009-1-Ig), horseradish peroxidase (HRP)-conjugated Affinipure goat anti-mouse IgG (H+L) (lot no. SA00001-1), and HRP-conjugated Affinipure goat anti-rabbit IgG (H+L) (lot no. SA00001-2) were purchased from Proteintech (Wuhan, China). Antibodies to p-AKT (Thr308) (lot no. AF0016), p-phosphoinositide 3-kinase (PI3K) p85 (Tyr458) [Tyr467]/p55 (Tyr199) (lot no. AF0016), Interleukin 6 (IL6) (lot no. DF6087), ERK1/2 (lot no. AF0155) and poly (ADP-ribose) polymerase 1 (PARP1) (lot no. DF7198) were from Affinity Biosciences (Cincinnati, OH, United States). N-acetyl-L-cysteine (NAC) (lot no. S0077, >99%), reactive oxygen species (ROS) assay kit, BCA protein assay kit, One Step TdT-Mediated dUTP Nick End Labeling (TUNEL) Apoptosis Assay Kit, and 4,6-diamidino-2-phenylindole (DAPI) staining solution were purchased from Beyotime Biotechnology (Shanghai, China). The Annexin V fluorescein isothiocyanate/propidium iodide (FITC/PI) apoptosis kit was purchased from Multi Sciences (Hangzhou, China).

Normal rat tubular proximal epithelial cell lines (NRK-52E) were obtained from the University of Yamanashi (Yamanashi, Japan). The cells were maintained in DMEM/F-12 supplemented with 5% FBS, 100 U/ml penicillin G, and 100 mg/ml streptomycin, and incubated at 37°C in 5% CO₂. The cells at approximately 80–90% confluency were digested and passaged.

NRK-52E cells were digested and seeded in 96-well plates at 1 × 10⁴ cells per well. After 24 h of incubation, the cells were cultured in a serum-free medium and stimulated with ADR (2 μg/ml) in the absence and presence of PFAE (5 mg/ml) or luteolin (10 μM/L). After 24 h of treatment, changes in cell morphology and number were observed by inverted phase-contrast microscopy using model CKX41 microscope (Olympus, Tokyo, Japan). Then, the medium was discarded, and 100 μl serum-free medium containing 10% CCK-8 was added to each well. After incubating for another 1 h, the absorbance was measured using a model ELx800 spectrometer (BioTek Instruments, Winooski, VT, United States) at a wavelength of 450 nm. Cell viability is expressed as a percentage of the control.

ROS production was detected using an ROS detection kit. Briefly, cells were seeded in a 12-well plate and incubated for 24 h. ADR, luteolin, and NAC were used to stimulate the cells according to the experimental protocol. The fluorescent probe 2ʹ,7ʹ-dichlorofluorescin diacetate (DCFH-DA) was diluted with serum-free medium at a ratio of 1:1000 to obtain a final concentration of 10 μmol/L. The medium was removed and 500 μl of diluted DCFH-DA was added to each well. Cells were incubated at 37°C for 20 min and washed thrice with serum-free medium. Fluorescent images were visualized and captured with a standard green fluorescence cube using a model Ts2R inverted fluorescence microscope (Nikon, Tokyo, Japan). Fluorescence images were analyzed quantitatively using ImageJ software (1.52a; NIH, Bethesda, MD, United States).

TUNEL staining kit and DAPI staining solution were used to assess the presence of apoptotic cells. Briefly, the cells were incubated and treated in accordance with the experimental protocol, washed once with PBS, and fixed with 4% paraformaldehyde for 30 min. Then, they were permeabilized with 0.3% Triton X-100 in PBS for another 30 min and washed three times with PBS. Fifty microliters of TUNEL staining solution were added to each well and incubated in the dark for 1 h at 37°C. Then, cells were washed three times with PBS and 100 μl of DAPI staining solution was added to each well, followed by incubation in the dark for 10 min and washing with PBS. Finally, cell fluorescent images were captured using a fluorescence microscope (Nikon Ni-U, DS-Qi2) equipped with standard red and ultraviolet fluorescence cubes. Fluorescent images were analyzed quantitatively using ImageJ software (1.52a).

An Annexin V-FITC/PI apoptosis assay kit was used to detect apoptosis. Briefly, the cells were washed with PBS and digested with trypsin for 3 min. Then, serum-containing medium was added to terminate the digestion. Each sample was centrifuged and the supernatant was discarded. Five hundred microliters of 1× binding buffer were added to resuspend the cells, followed by 5 μl Annexin V-FITC and 10 μl PI to each tube. After incubation for 10 min protected from light on ice, a FACSCelesta flow cytometer (BD, Santa Clara, CA, United States) was used to analyze the percentage of apoptotic cells.

After incubation and intervention, the cells were lysed in RIPA buffer supplemented with 2% protease and phosphatase inhibitor cocktail for 20 min on ice. After sonication, the samples were centrifuged at 4°C for 20 min, and each supernatant was collected for subsequent experiments. The total protein concentration was determined using a BCA protein assay kit. Equal amounts of protein (20 µg) were separated by 10 or 12% sodium-dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE). The resolved proteins were electrophoretically transferred to polyvinylidene difluoride (PVDF) membranes (Millipore, Billerica, MA, United States). The membranes were blocked with 5% non-fat dried milk in PBS containing Tween (PBST) buffer for 1 h at room temperature, washed thrice with PBST, and incubated with primary antibodies overnight at 4°C. The membranes were washed again and incubated with HRP-conjugated anti-rabbit or anti-mouse IgG for 1 h at room temperature. The bands were flushed with chemiluminescent HRP substrate and imaged using a chemiluminescence system (ChemiDoc™ XRS+ with Image Lab Software; Bio-Rad, Hercules, CA, United States). Western blot images were analyzed quantitatively using ImageJ software (1.52a). β-actin was also assayed to confirm equal loading of proteins.

Statistical analyses were performed using SPSS version 19.0 (IBM, Armonk, NY, United States). All values are expressed as mean ± standard deviation (SD). Comparisons of two populations were made using unpaired 2-tailed Student’s t-test. For multiple groups, statistical significance was determined using one-way analysis of variance (ANOVA) followed by Dunnett’s test. Statistical significance was set at p<0.05.

A total of 328 chemical components of P. frutescens were retrieved from the TCMSP database. Use of OB ≥30% and DL ≥0.18 as the screening criteria yielded 14 active components. An analysis of the published literature revealed that although (s) – carvone (Alasanea and Liu, 2017), caffeic acid (Tyszka-Czochara et al., 2018), rosmarinic acid (Colica et al., 2018), caryophellene (Sharma et al., 2016), perilla aldehyde (Song et al., 2018), and perilla ketone (Roellecke et al., 2017) did not meet the aforementioned OB and DL screening criteria, in vivo and in vitro experiments were validated. Therefore, these components were included to result in 19 main active components (C20 Supraene did not retrieve the predicted target, and so was excluded in further research) (Table 1 and Supplementary Figure S1). The predicted targets of the active components were searched for in the TCMSP database, and the predicted targets in the Uniport database were checked to obtain 225 targets of the main active components. Of these, 141 unique targets were finally obtained after deduplication.

The relevant targets of CKD stage 3–5/CRF were searched for in the Gene cards database with a score ≥10 as the screening criterion. The search resulted in 2263 main targets.

The predicted targets of P. frutescens were mapped to the main targets related to CKD stage 3–5 using Venny2.1.0 software. A Venn diagram was constructed (Figure 1A). A total of 108 intersection targets were obtained. These may be the key targets of P. frutescens for CKD treatment.

The main active components and key targets of P. frutescens were imported into Cytoscape 3.6.0, to build a main active component-target network model. The model included 160 nodes and 225 edges. In this model, 19 active components are shown as diamond-shaped yellow nodes, 141 targets are blue square nodes, and the edges represent the interaction between the components and the targets (Figure 1B). The Network Analyzer plugin was used to perform network topology analysis and to calculate the degree, betweenness centrality and closeness centrality parameter values. The median of the degree was one. Therefore, we chose degree >2 (i.e., more than twice the median) as the criterion to filter out the core nodes in the network model, including 12 components and 18 targets (Supplementary Table S1).

Further analysis of the nodes in the table revealed that the highest degree among the main active components was C15 (luteolin), with 57 targets. In descending order, C5 (β-sitosterol) corresponded to 37 targets, C14 (rosmarinic acid) had 32 targets, C13 (caffeic acid) corresponded to 22 targets, and C6 (β-carotene) had 21 targets. The target is the key to the pharmacological effect of the component, so it is speculated that luteolin, β-sitosterol, and rosmarinic acid are the key components of P. frutescens in the treatment of CKD stage 3–5. Among the target nodes, the highest degree was prostaglandin-endoperoxide synthase 2 (PTGS2), which targeted 11 components. Subsequently, in descending order, PTGS1 and nuclear receptor coactivator 2 (NCOA2) all targeted eight components, heat shock protein 90 (HSP90) is connected to five active components. The biological functions related to these key targets were mainly related to inflammation, immunity, metabolism, and others, suggesting that P. frutescens may exert its pharmacological effects on CKD through the aforementioned approaches.

Additionally, the network model found that 18 active components targeted at least two targets, and 41 targets were connected to at least two components. There was a large amount of cross-coincidence between each component and target. These results reflect the multi-component, multi-target effects of P. frutescens.

The relevant targets of P. frutescens for treating CKD stage 3–5 were imported into String 11.0, under the condition of “homo sapiens.” The highest confidence protein score value was set as >0.7, a PPI network was constructed, and unrelated individual proteins were removed. The network included 105 targets and 736 interactions. The PPI network information was imported into Cytoscape3.6.0 to build a network model, and the Network Analyzer plugin was used to analyze network characteristics. By adjusting the node color and size according to the degree, the greater the degree, the larger was the node. The thickness of the side was revised according to the Combine score: the larger the Combine score, the thicker the side (Figure 2A).

We selected degree ≥20 as the criterion. Twenty-seven key proteins in the PPI network were selected (Figure 2B), including AKT1, tumor protein 53 (TP53), IL6, tumor necrosis factor (TNF), MAPK1, JUN, vascular endothelial growth factor A (VEGFA), epidermal growth factor receptor (EGFR), caspase 3 (CASP3), cyclin D1 (CCND1), MYC, RELA, matrix metalloproteinase 9 (MMP9), IL2, PTGS2, and others, implicating these targets as the main targets of P. frutescens in the treatment of CKD stage 3–5.

The PPI network is modular, and the functional module is the basic unit of protein function. Modular analysis can simplify the complex protein network, which is conducive to further research on the biological and pharmacological functions of the protein. Therefore, the MCODE plugin in the Cytoscape software was used to perform modular analysis of the PPI network, with Degree Cutoff = 2, Node Score Cutoff = 0.2, K-Core = 5, and Max. Depth = 100. Two key protein modules were identified. Protein module A contained 20 targets and 101 interactions. Protein module B contained 25 targets and 99 interactions (Figure 2C).

We imported 105 targets in the PPI network into Cytoscape3.6.0 software and used ClueGO and CluePedia plugins for enrichment analysis of CC and MF. A p ≤ 0.01 was the screening parameter. A pie chart was drawn (Figures 3A,B). CCs mainly included membrane microdomains, integral components of presynaptic membranes, protein kinase complexes, platelet alpha granules, nuclear euchromatin, basal plasma membrane, multivesicular body, and pore complex. MFs mainly involved phosphatase binding, estrogen 2-hydroxylase activity, catecholamine binding, cysteine-type endopeptidase activity involved in the apoptotic signaling pathway, cytokine receptor binding, BH3 domain binding, protein kinase C activity, HSP70 protein binding, nitric oxide synthase regulator activity, virus receptor activity, cyclin-dependent protein serine/threonine kinase regulator activity, disordered domain-specific binding, nitric oxide synthase binding, chemokine receptor binding, copper ion binding, and insulin-like growth factor receptor binding. Enrichment analysis of CCs was selected as a representative to construct a network model (Figure 3C). Multiple targets were connected to at least two CCs, which are the key nodes in the network. The findings provided further evidence of the multi-target, multi-effect characteristics of P. frutescens.

A total of 105 targets were imported into the DAVID6.8 online analysis platform for the enrichment analysis of BP and KEGG signaling pathways. With p < 0.01 as the screening criterion, 254 BPs and 101 pathways were obtained. The first 20 BPs and pathways were selected using the clusterProfiler package in R language to draw bar graphs and bubble diagrams (Figures 4A,B and Supplementary Figures S2A,B). In the bubble diagram, the node size represents the gene number: the larger the node, the more genes corresponding to the node. The color of the node represents the p-value. The smaller the p-value, the redder the color; otherwise, the values were depicted in shades of blue. The ordinate represents the BP or pathway name corresponding to the node, and the abscissa represents the fold enrichment. The greater the fold enrichment, the higher the degree of gene enrichment. BPs were observed to be mainly related to the response to drugs, negative regulation of apoptotic process, response to estradiol, inflammatory response, positive regulation of protein phosphorylation, positive regulation of nitric oxide biosynthesis process, aging, positive regulation of transcription, response to ethanol, positive regulation of cell proliferation, cellular response to organic cyclic compounds, response to toxic substances, positive regulation of transcription from RNA polymerase II promoter, positive regulation of nuclear factor-kappa B (NF-κB) transcription factor activity, and positive regulation of ERK1 and ERK2 cascade. Enriched pathways related to CKD stages 3–5 mainly included the apoptosis signaling pathway, T cell receptor (TCR) signaling pathway, PI3K-AKT signaling pathway, hypoxia inducible factor-1 (HIF-1) signaling pathway, Toll-like receptor (TLR) signaling pathway, and TNF signaling pathway. Thus, P. frutescens may play a therapeutic role in CKD stages 3–5 by regulating the above multiple pathways.

The key components in the component-target network were selected as the ligand, and the corresponding core node as the receptor. Imported Pyrx was used for preprocessing, and Autodock vina was used for molecular docking (Table 2). Binding affinity <0 meant that the receptor and ligand bound spontaneously: the lower the value, the greater the binding affinity. The binding energies of the targets, PTGS2, AKT1, p53, TNF-α, MAPK1, IL6, JUN, and their original ligands were −11.6, −8.7, −4.3, −9.7, −9.8, −4.5, −9.6 kJ/mol, respectively. The binding capacity of each key component and corresponding target was similar to that of the original ligand. We selected luteolin, the most critical component in the network model, and performed molecular docking and visual analysis through Schrödinger Maestro and Pymol2.1 (Figures 5A,B). These results verified that the main active components of P. frutescens may play a therapeutic role in CKD by acting on their corresponding key targets.

As shown in Figure 6, five chemical components in PFAE were detected by HPLC, including luteolin (C₁₅H₁₀O₆; CAS: 491-70-3), rosmarinic acid (C₁₈H₁₆O₈; CAS: 20283-92-5), caffeic acid (C₉H₈O₄; CAS: 501-16-6), perilla aldehyde (C₁₀H₁₄O; CAS: 18031-40-8) and (+)-catechin (C₁₅H₁₄O₆; CAS: 154-23-4). Luteolin, rosmarinic acid and caffeic acid are the main active components of P. frutescens, and the content is large. Therefore, the concentrations of these three components in different preparations were measured for quality control.

In addition, these components identified by HPLC were coincident with those screened from the TCMSP database. And according to the results of network pharmacological analysis, these five components were also the main active components in the network model of P. frutescens for treating CKD.

According to the HPLC results, we calculated that the concentration of luteolin in the PFAE was about 0.15 mg/ml, which provided the basis for further study of its therapeutic effects. Luteolin-pretreated NRK-52E cells were examined to assess the effect on cell morphology and viability. Compared with the control group, the number of cells in the ADR (2 μg/ml) group was reduced, the cells were shrunken and deformed, and some damaged cells floated in the culture (Supplementary Figures S3A,B). Compared with the ADR group, the number of cells was increased in the PFAE (5 mg/ml) and luteolin (10 μM/L) pretreatment groups, and the cell morphology was normal (Figure 7A). Cell viability was measured using the CCK-8 assay. Compared with the ADR group, the cell viability in the PFAE+ADR and luteolin+ADR groups was significantly increased. Therefore, PFAE and luteolin (the key component of P. frutescens) can attenuate ADR-induced cell injury (Figure 7B).

AKT participates in a variety of physiological processes in the body, including cell proliferation, apoptosis, protein translation, and glucose metabolism. It plays an important role in various kidney diseases. AKT is the core protein in the PPI network, and molecular docking has revealed good binding ability between luteolin and AKT. We used ADR to construct an in vitro renal tubular injury model and found that it can significantly increase the phosphorylation of AKT. Upon the addition of luteolin was added, and this promotion was effectively inhibited (Figure 7C).

At the same time, TP53 and IL6 are also key targets of the PPI network model. P53 (the expressed protein of TP53) is a key protein that regulates the cell cycle. It is closely related to cell self-repair and cell apoptosis. IL6 is regarded as a pro-inflammatory cytokine, which plays a key role in the immune response and inflammation. Both p53 and IL6 are involved in the development of CKD, and can aggravate renal injury by inducing apoptosis, promoting inflammation, accelerating cell senescence and other ways (Amdur et al., 2016; Juvet et al., 2020; Wang et al., 2020). Therefore, we also detected the phosphorylation level of p53 and the expression of IL6. The results showed that luteolin can significantly reduce ADR-induced p53 phosphorylation, but had no significant effect on IL6 (Figure 7D). These results indicate that AKT and p53 may be the targets of luteolin in reducing ADR-mediated kidney damage.

TUNEL + DAPI staining and flow cytometry analysis were performed to verify the anti-apoptotic role of luteolin in NRK-52E cells treated with ADR. The apoptosis rate in the luteolin+ADR group was significantly lower than that in the ADR group (Figures 8A,B). We further tested the expression of apoptosis-related proteins, including PARP1, Caspase3, Bcl-2, and Bax. Compared with the ADR group, these apoptotic proteins were significantly down-regulated in the luteolin+ADR group (Figure 8C). These data show that luteolin mitigated ADR-induced apoptosis.

Oxidative stress is one of the important mechanisms in cell apoptosis and plays a key role in the pathological process of CKD. The overproduction of ROS is the initial factor of oxidative stress. We tested ROS production using an immunofluorescence probe. As shown in Figure 9A, after 4 h of incubation, ADR significantly increased the generation of ROS (green), whereas pretreatment with luteolin or NAC antioxidant suppressed the ADR-induced overproduction of ROS. Furthermore, compared with the NAC pretreated group, there was no significant difference in the inhibitory effect of luteolin on ROS induced by ADR. These outcomes suggest that luteolin effectively inhibits ADR-induced apoptosis via suppression of ROS overproduction.

According to the results of network pharmacology analysis, luteolin inhibits apoptosis to protect the kidneys against multiple insults. The MAPKs are critical in apoptosis and usually activated by oxidative stress. Thus, we focused on the MAPK pathway to assess the role of luteolin. As shown in Figure 9B, ADR significantly promoted the activation of MAPK families (ERK1/2, p38, and JNK) in renal tubular cells, whereas pretreatment of cells with luteolin effectively reduced the phosphorylation of p38, ERK1/2, and JNK. These results suggest that the regulation of MAPK pathways may mediate the protective effect of luteolin on ADR-induced kidney damage, and can be regarded as a link between oxidative stress and apoptosis.

PI3K-Akt pathway plays an important role in cell apoptosis, and it is also one of the top pathways in the KEGG enrichment analysis. We have found that luteolin can reduce the phosphorylation of AKT. Therefore, we further detected the phosphorylation of PI3K. The results showed that the effect of luteolin on PI3K phosphorylation did not reach statistical significance (Figures 9C,D). According to the results, the role of PI3K-Akt pathway in luteolin’s reduction of apoptosis has not been clarified, and the reason why the changes of PI3K and AKT are not completely parallel needs to be further studied.

The p53 pathway is also closely related to cell apoptosis. In Figure 7D, we have found that luteolin can inhibit the phosphorylation of P53. Therefore, down-regulation of the p53 pathway may also be an important mechanism by which luteolin alleviates apoptosis.

Finally, we summarized the possible mechanism of luteolin to reduce the apoptosis of renal tubular cells (Figure 10). Additionally, the key pathway maps of MAPK and P53 were also retrieved from KEGG database to provide reference for further in-depth exploration of the mechanism (Supplementary Figures S4A,B).