1. Introduction
Radiation-induced lung injury (RILI) is a common complication of radiation therapy and is characterized by inflammation, fibrosis, and lung dysfunction. It results from inflammatory activation, oxidative stress, and immune cell infiltration, leading to cellular damage and tissue remodeling [[
1], [
2], [
3]]. RILI can present acutely as radiation pneumonitis and progress to chronic pulmonary fibrosis. Radiation therapy is crucial for managing lung cancer, particularly for patients with localized tumors unsuitable for surgery [[
4], [
5], [
6]]. However, although effective against cancer cells, radiation also damages healthy lung tissue, leading to RILI [[
7], [
8], [
9]]. The immune and inflammatory responses triggered by radiation further complicate the treatment of both lung cancer and RILI, presenting a significant challenge for its management.
Compound Kushen injection (CKI) is an injection [[
10], [
11], [
12], [
13]] composed of traditional Chinese medicine (TCM) Kushen (KS;
Sophora flavescens Aiton) and Baitulin (BTL;
Heterosmilax japonica Kunth), which was launched in 1995. It has the effects of clearing heat, relieving dampness, cooling blood, detoxifying, dispersing masses, and relieving pain and is used for the treatment of cancer pain and bleeding. Clinically, it is commonly used for the treatment of radiation-induced injuries [[
14], [
15], [
16], [
17], [
18]], as well as for the treatment and adjuvant treatment of various cancers [[
19], [
20], [
21], [
22]]. However, the mechanism of CKI in the treatment of radiation-induced diseases remains unclear.
Network pharmacology [
23] is an interdisciplinary approach to study the complex interactions between drugs, targets, and diseases, aiming to identify potential therapeutic mechanisms and optimize drug development strategies [
24,
25]. As the core theory of TCM network pharmacology, network target is proposed to address complex systemic challenges, such as in TCM research [[
26], [
27], [
28], [
29], [
30]]. Through the integration of systems biology and multi-omics technologies, network target provides genome-wide target predictions for TCM herbs and formulas and construct multilevel modular biomolecular networks, offering a panoramic theoretical foundation for elucidating the mechanisms of TCM.
In this study, we improved a previously proposed statistical strategy [
31] for calculating the holistic targets of TCM herbs or formulas by integrating quantitative compound detection data. Based on the improved strategy and the quantitative compound information of CKI, we modeled the holistic targets of CKI. By integrating transcriptomics data, we predicted the key mechanisms of CKI in the intervention of RILI, which was experimentally validated in irradiated animals.
2. Methods and materials
2.1. Network target analysis of CKI in the treatment of RILI
2.1.1. Weighted holistic target model
Based on published and unpublished studies conducted by Shanxi Zhendong Pharmaceutical Co., Ltd., China, we collected 107 compounds in total, including 42 flavonoids, 30 alkaloids, 5 sugars, 4 triterpenoid saponins, 2 phenolic glycosides, 2 phenylpropanoids, and 22 unclassified compounds. The results of the quantitative analysis of several types of compounds, such as alkaloids, sugars, and flavonoids, were also obtained, as were the results of the quantitative analysis of certain main alkaloids. All compounds were subsequently annotated in the PubChem database for structural information. A network-based algorithm from the Intelligent and Quantitative Analysis Technology and System for Traditional Chinese and Western Medicine Based on Network Targets (UNIQ) system [
24], drugCIPHER [
32], was subsequently used to predict the genome-wide biological effect profile of every compound in CKI.
A previously described statistical strategy [
31] was proposed for predicting the holistic targets of a given complex system, such as TCM herbs or TCM formulas. To integrate the quantitative information of certain compounds, we have improved the original calculation strategy. For cases where the contents of certain compounds are known, we employ an improved holistic target calculation method. Assuming that there are a total of
n compounds in the TCM herb or formula and that
m of these compounds have a content ratio of
pk, based on the given number of samples
N, the composition of the new TCM/TCM formula is determined by sampling the corresponding compounds probabilistically, with the sampling weight being as follows:
Alternatively, if the detected information indicates that the content ratio of the compound category to which compound k belongs is Pk. The composition of the new TCM herb or formula, which is based on the given number of samples N, is determined by probabilistically sampling the corresponding compounds, with the sampling weight being as follows:
where, mk represents the total number of compounds in the category to which compound k belongs.
Finally, for the composition of the new TCM herb or formula (considered as a total of n compounds), the holistic target model described above is applied to the sampling process, and the quantitative holistic target of the TCM herb or formula is calculated. For one specific class of compounds in a TCM herb or formula, we also applied this strategy to calculate holistic targets:
where is the probability of the target appearing in the k th component target profile, Fk is all the subsets of the k components, A is the subset of the k components, Ac is the complement of A, and pi and pj are the probabilities of the target appearing in the predicted target profile of component k.
2.1.2. Network target analysis
To investigate the mechanisms underlying RILI, we performed an extensive review of the literature and systematically gathered molecular evidence. This process involved obtaining the RILI-related dataset GSE206426 from the Gene Expression Omnibus (GEO) database, collecting RILI-associated molecules using the Comparative Toxicogenomics Database (CTD) [
33] and the Gendoo algorithm [
34], and reviewing additional relevant studies for a more in-depth understanding [[
1], [
2], [
3]]. First, we performed enrichment analyses to identify the pathway-level mechanism of RILI and potentially targeted pathways or biological processes of CKI and its components. The intersection pathways or biological processes related to the literature reviews were subsequently retained to construct a modular regulatory effect of CKI in the treatment of RILI. Multilayer biological networks were constructed based on the key molecules in these pathways or biological processes, as well as the holistic targets of CKI in these pathways or biological processes.
In detail, Kyoto Encyclopedia of Genes and Genomes (KEGG) and Gene Ontology (GO) enrichment analyses for the holistic targets of CKI- and RILI-associated molecules were performed with the R package clusterProfiler [
35]. Multilayer biological networks were constructed based on protein-protein interaction (PPI) networks collected from Ref. [
36]. Single-cell RNA sequencing (scRNA-seq) analysis of GSE206426 was conducted using the R package Seurat [
37] to find the differentially expressed genes (DEGs) between the control group and radiation group in two kinds of animal models (C57BL/6N and C3H/HeN mice). The quantitative estimate of drug-likeness (QED) score, which was used to estimate the drug likeness of every compound, was calculated by the QED function implemented by rdkit. The similarities between RILI-related molecules and the targets of given compounds were predicted using a network-based algorithm [
38] implemented in the R package GoSemSim [
39].
2.2. Establishment of mouse irradiation model
2.2.1. Experimental animals
Wild-type (WT) male C57BL/6 mice (aged 8-10 weeks, weighing 20-25 g) were housed under specific pathogen-free (SPF) conditions. The mice were purchased from the Guangdong Medical Laboratory Animal Center (license No. SCXK (Yue) 2022-0002). They were housed at the Experimental Animal Center of Guangdong Pharmaceutical University with the animal ethics approval number GDPULACSPF2022234. The animals were kept at room temperature (23 ± 2) °C and a relative humidity of 40%-52%, with a 12-h light/12-h dark cycle, and were provided with free access to food and water.
2.2.2. Experimental animal groupings
The C57BL/6 mice were divided into groups (ten mice per group): nonirradiated group (CON group), 15 Gy irradiation model (NC group), CKI low-dose group (CKI-L group), CKI middle-dose group (CKI-M group), CKI high-dose group (CKI-H group), and the amifostine control group (150 mg∙kg−1; ALT group).
2.2.3. Drug preparation
NC group: 100 μL 0.9% NaCl; CKI-L group: 75 μL 0.9% NaCl + 25 μL CKI; CKI-M group: 50 μL 0.9% NaCl + 50 μL CKI; CKI-H group: 100 μL of CKI; ALT group: 100 μL amifostine solution (150 mg∙kg−1).
2.2.4. Establishment of a mouse lung irradiation model
The mice were anesthetized by intraperitoneal injection of 2% pentobarbital sodium solution, placed in the supine position, and fixed to an X-ray biological irradiation apparatus. Custom lead plates were used to shield the head, neck, abdomen, and upper and lower limbs, leaving only the mouse lung exposed (3 cm width × 2 cm height). A dose of 15 Gy was applied. The total irradiation time was calculated based on the dose rate and irradiation speed to ensure an accurate irradiation dose (160 kV, 25 mA; default dose rate 6.031 Gy∙min−1) and the irradiation time was 149 s.
2.3. Sample collection
Blood collection: C57BL/6 mice were euthanized, and blood was collected by eye enucleation. After the blood clotted, the serum was separated and stored at −80 °C for subsequent enzyme-linked immunosorbent assay (ELISA) and superoxide dismutase (SOD) experiments.
Lung tissue collection: Mice were fixed in the supine position on the operating table, and their chests were disinfected with 75% medical alcohol-soaked cotton balls. The sternum was cut open to expose the chest cavity, and the lungs were carefully removed for macroscopic observation. The left upper lung was fixed in 4% paraformaldehyde for histopathological analysis. The left lower lung and right lung were rapidly frozen in liquid nitrogen.
2.4. Lung tissue histopathological examination
2.4.1. Preparation of lung tissue paraffin sections
Lung tissue was fixed in 4% paraformaldehyde for more than 24 h at 4 °C. The tissue was then washed with running tap water for 12-16 h. Following gradient dehydration, the tissue was soaked in 50%, 70%, 80%, and 90% ethanol solutions for 30 min each, followed by soaking in 95% ethanol solution 1, 95% ethanol solution 2, 100% ethanol solution 1, and 100% ethanol solution 2 for 1 h each. The tissue was cleared by soaking in xylene solutions 1 and 2 for 1 h each. For paraffin embedding, the tissue was soaked in low-, medium-, and high-melting-point paraffin for 1 h each. The paraffin-embedded lung tissue was placed into a paraffin embedding machine to form wax blocks.
The embedded wax blocks were sectioned at a thickness of 3-5 µm using a microtome. The wax block was trimmed with a blade until smooth, and the sections were cut. The cut samples were placed into a water bath set at 42 °C to spread them out. The tissue sections were collected on prelabeled slides, air-dried on newspaper, and then dried in a 37 °C oven to remove excess moisture.
2.4.2. Hematoxylin and eosin (H&E) staining
The tissue sections were dewaxed and rehydrated through a series of ethanol washes. They were then stained with hematoxylin solution for 3-5 min, followed by washing with running tap water and distilled water for 1 min each. Afterward, the sections were stained with an eosin solution for 1 min and rinsed with a color-enhancing solution for a few seconds. A drop of neutral gum was placed on the stained tissue, and a coverslip was gently applied while avoiding bubbles. Finally, the morphology of the lung tissue was examined under a microscope.
2.4.3. Lung tissue immunofluorescence staining
Paraffin sections were dewaxed and rehydrated through a series of xylene and ethanol washes. Afterward, antigen retrieval was performed by heating the tissue sections in citrate buffer solution containing citric acid and trisodium citrate in a pressure cooker for 15 min. The slides were then cooled to room temperature and washed with phosphate-buffered saline (PBS). Blocking was performed with normal goat serum for 45 min at room temperature. Primary antibodies were added, and the samples were incubated overnight at 4 °C. The following day, the slides were washed with 0.05% Tween 20 solution (PBST) and incubated with secondary antibodies at 37 °C for 1 h in the dark. Afterward, the slides were counterstained with 4′,6-diamidino-2-phenylindole (DAPI) for 5 min, washed, and mounted with anti-fade reagent. Images were captured using an inverted fluorescence microscope.
2.5. Lung tissue sequencing and analysis
2.5.1. Sample detection
To ensure the use of qualified samples for transcriptomic sequencing, the purity, concentration, and integrity of the extracted total RNA were detected using advanced molecular biology equipment, with strict monitoring. The total RNA quality testing methods were as follows: ① A NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific, USA) was used to determine the purity and concentration of the RNA, and ② an Agilent 2100 (Agilent Technologies, USA)/LabChip GX (PerkinElmer, USA) was used for the precise detection of RNA integrity.
2.5.2. Library construction
After confirming the quality of the samples, library construction was performed. The main process was as follows: ① Eukaryotic messenger RNA (mRNA) was enriched using magnetic beads with oligo(dT). ② The mRNA was randomly fragmented using fragmentation buffer. ③ Using the mRNA as a template, the first and second complementary DNA (cDNA) strands were synthesized, followed by cDNA purification. ④ The purified double-stranded cDNA was then end-repaired, an A-tail was added, and sequencing adapters were ligated. The fragments were size-selected using AMPure XP beads. ⑤ Finally, polymerase chain reaction amplification was performed to obtain the cDNA library.
2.5.3. Library quality control
After library construction, initial quantification was performed using a Qubit 3.0 fluorometer, ensuring that the concentration reached at least 1 ng∙μL−1. Afterward, the inserted fragments of the library were analyzed using a Qsep400 high-throughput analysis system. After confirming the insert size, quantitative polymerase chain reaction (qPCR) was used to accurately quantify the effective concentration of the library (effective concentration > 2 nmol∙L-1) to ensure library quality.
2.5.4. Sequencing
After passing quality control, the library was subjected to sequencing using a high-throughput sequencing platform in PE150 mode.
2.5.5. Analysis workflow
After sequencing, the raw data were processed using the bioinformatics analysis workflow provided by BMKCloud. The data were filtered to obtain clean data, which were subsequently mapped to the specified reference genome to obtain mapped data. Library quality assessment, structural-level analysis, differential expression analysis, gene functional annotation, and functional enrichment were subsequently performed. Additionally, through BMKCloud, in-depth data mining and interactive analysis of the transcriptome were conducted, including gene search and plotting, unique and shared gene analysis, PPI network construction, gene set enrichment analysis (GSEA), differential gene coexpression analysis, and gene coexpression network construction (WGCNA).
2.6. Establishment of mouse irradiation model with nuclear factor erythroid 2-related factor 2 (NRF2) inhibition
2.6.1. Experimental animal grouping
C57BL/6 mice were divided into groups (ten mice per group): CON group, NC group, CKI-H group, Solvent (dimethyl sulfoxide (DMSO)) group, and ML385 (NRF2 inhibitor) group.
2.6.2. Drug preparation
NC group: 100 μL of 0.9% NaCl; CKI-H group: 100 μL of CKI; Solvent group: 5 μL of DMSO + 40 μL of polyethylene glycol (PEG) 300 + 5 μL of Tween 80 + 50 μL of physiological saline; ML385 group: 5 μL of stock solution (100 mg ML385 + 0.76 mL DMSO) + 40 μL of PEG 300 + 5 μL of Tween 80 + 50 μL of physiological saline.
2.6.3. Lung tissue RNA extraction and real-time fluorescence qPCR analysis
Lung tissue RNA was extracted using TRIzol reagent. Approximately 50 mg of lung tissue was placed in 500 μL of TRIzol and incubated for 5 min. The tissue was homogenized using a tissue homogenizer until no fragments remained. Afterward, 100 μL of chloroform was added, followed by vortexing for 20 s and manual inversion to mix. The sample was left on ice for phase separation, followed by centrifugation at 12 000 r∙min
−1 at 4 °C for 15 min. The upper colorless liquid was carefully transferred to a new Eppendorf (EP) tube. Next, 500 μL of precooled 75% ethanol was added, and the sample was vortexed. The sample was centrifuged at 4 °C and 7500 r∙min
−1 for 5 min, after which the supernatant was discarded. The pellet was resuspended in 1 mL of a freshly prepared 75% ethanol-diethyl pyrocarbonate (DEPC) water mixture, vortexed until the precipitate floated, and then centrifuged at 4 °C and 7500 r∙min
−1 for 5 min. After the supernatant was discarded, the EP tube was placed upside down on filter paper to dry for 3-5 min. Finally, 10-25 μL of DEPC water was added to dissolve the RNA, and the RNA purity and concentration were measured using a spectrophotometer. The RNA samples were stored at −80 °C. The primer sequences are provided in
Table 1.
2.7. Western blot analysis of lung tissue
Total protein was extracted from lung tissues with the lysis buffer and phenylmethylsulfonyl fluoride (PMSF) added at a ratio of tissue:lysis buffer:PMSF = 1:10:0.1. The tissue was homogenized using an electric grinder on ice and then incubated for 30 min for complete lysis. The homogenate was transferred to a new 1.5 mL EP tube, and the protein concentration was determined using the bicinchoninic acid (BCA) method. Equal amounts of protein (30 μg) were mixed with 5× loading buffer, boiled for 15 min, and stored at −80 °C. For Western blotting, proteins were separated by sodium dodecyl-sulfate polyacrylamide gel electrophoresis (SDS-PAGE) electrophoresis using concentration and separation gels and transferred onto polyvinylidene fluoride (PVDF) membranes. The membranes were activated with methanol and soaked in transfer solution for 15 min. After electrophoresis, the gel was cut according to the molecular weight of the target protein, and the transfer was performed at 320 mA for 45 min in an ice bath. The membranes were subsequently washed three times with washing solution for 5 min each and blocked with 8% nonfat milk for 2 h. Primary antibodies were added, and the samples were incubated overnight at 4 °C. After the samples were washed, secondary antibodies were added, and the samples were incubated at room temperature for 1 h. The membranes were then incubated with enhanced chemiluminescence (ECL) substrate for 30 s and visualized under a fluorescence imaging system.
Serum SOD levels were measured using an SOD test kit (Nanjing Jiancheng Bioengineering Institute, China). Serum samples were thawed on ice and placed in labeled centrifuge tubes. The test wells were prepared as follows: one blank, one control, one sample well, and one test blank. In the blank well, 20 μL of distilled water and 20 μL of enzyme dilution were added; in the control well, 20 μL of distilled water and 20 μL of enzyme working solution were added; in the sample well, 20 μL of serum sample and 20 μL of enzyme working solution were added; and in the test blank well, 20 μL of serum sample and 20 μL of enzyme dilution were added. SOD test substrate (200 μL) was added to each well, and the plate was incubated at 37 °C for 20 min. The absorbance was measured at 450 nm using a microplate reader to calculate the SOD levels.
2.8. Biolayer interferometry (BLI) assay
The Octet RED96e system (Sartorius, Germany) is ideally suited for the characterization of protein-small-molecule binding kinetics and binding affinity. NRF2 protein was diluted to 30-50 μg∙mL−1. The proteins were then immobilized on glutathione S-transferase (GST) sensors. The sensors were then blocked, washed, and moved into wells containing various concentrations of the test compounds of CKI in kinetic buffer. Program settings default to programs that bind proteins and small molecules in the system. The binding signals were identified, and the results were analyzed using OctetHT V10.0 software. R2 is an estimate of the goodness of fit, with values close to 1 indicating an excellent fit.
2.9. Statistical analysis
The results were plotted using GraphPad Prism 8.0 software. All the data are presented as the mean ± standard error of the mean (SEM). Statistical comparisons between two groups were made using a t test (Student’s t test), and comparisons between three or more groups were made using one-way analysis of variance (ANOVA). A p value < 0.05 was considered to indicate statistical significance.
3. Results
3.1. Identification of compounds and quantitative assay-based holistic target modeling of CKI
In our study, we explored the mechanisms underlying the therapeutic effects of CKI on RILI. We first integrated compound identification and quantitative analysis of CKI, applying a weighted approach to model its network target. A multilayered modular biomolecular network was constructed to elucidate CKI’s intervention across various stages of RILI, followed by transcriptomic sequencing and validation using a mouse irradiation model.
To predict the potential impact of CKI on RILI from a molecular perspective, we curated 107 compound components from CKI sourced from both published and unpublished fingerprinting and content studies conducted by Shanxi Zhendong Pharmaceutical Co., Ltd. (China) [[
40], [
41], [
42]]. These compounds included 42 flavonoids, 30 alkaloids, 5 sugars, 4 triterpenoid saponins, 2 phenolic glycosides, 2 phenylpropanoids, and 22 unclassified compounds (
Fig. 1(a)). We also analyzed the compound profiles of BTL and KS to provide a comprehensive classification of the compounds derived from CKI. Quantitative analysis revealed that the highest contents in CKI were carbohydrates and alkaloids, accounting for 42.71% of the total sugars and 18.90% of the total alkaloids, respectively. Among the alkaloids, oxymatrine and matrine had the highest contents (42.96% and 15.20%, respectively), followed by oxysophocarpine (13.45%), sophoridine (5.02%), and sophocarpine (4.89%) (
Fig. 1(b)). All compounds were subsequently annotated in the PubChem database for structural information.
Using one of our previously developed algorithm in the UNIQ system, drugCIPHER [
32], we predicted the genome-wide biological effects of each identified compound. Building on this, we applied an improved statistical method suitable for analyzing quantitative compound detection data to calculate the targets of CKI, KS, and BTL compounds. This strategy allowed us to integrate quantitative data from CKI’s constituent compounds into the calculation of its holistic targets (
Fig. 1(c)). Notably, pathway analysis revealed that the primary targets of CKI were involved in signal transduction, immune system modulation, and energy metabolism (
Fig. 1(d)).
3.2. Network target analysis revealing the modular regulatory effect of CKI on RILI
To explore the mechanisms of RILI, we conducted a comprehensive review of the relevant literature and systematically collected molecular evidence. This included obtaining the RILI-related dataset GSE206426 from the GEO database, gathering RILI-associated molecules via the CTD database [
33] and the Gendoo algorithm [
34], and further reviewing pertinent studies for deeper insights [[
1], [
2], [
3]]. KEGG enrichment [
43] analysis of these RILI-related molecules indicated that RILI is closely linked to processes such as the immune response, inflammation, and apoptosis (
Fig. 1(e)). Based on our literature review, we identified three distinct stages in the disease mechanism of RILI. The early phase is centered on DNA damage, the acute phase is marked by the onset of pneumonia, and the late phase is characterized by the development of fibrosis. This complex pathological progression, from initial DNA damage and inflammation to late-stage fibrosis, is driven by the combined effects of oxidative stress, immune cell infiltration, inflammatory responses, and profibrotic factors. This mechanism accounts for the transition from acute pneumonia to irreversible pulmonary fibrosis in RILI.
Based on the predicted holistic targets of CKI and the mechanisms of RILI, network analysis revealed that CKI regulates RILI through six key modules: DNA damage in the early stage, reactive oxygen species (ROS), the immune response, inflammation during the pneumonitis phase, and fibroblast activation and extracellular matrix (ECM) remodeling during the fibrosis stage (Fig. S1(a) in Appendix A). KS and BTL were found to have potentially synergistic effects on ROS levels and immune-related pathways, such as oxidative stress response and T-cell activation regulation (
Fig. 2(a)). Specifically, alkaloids in CKI appear to play dominant roles in these pathways, with significantly enriched scores across various pathways (
Fig. 2(b)). Additionally, flavonoids, phenolic glycosides, and triterpenoid saponins contributed significantly to multiple modules. Building on these insights, we constructed three distinct biological networks to characterize the regulatory effects of CKI on RILI across different stages (
Fig. 2(c), Figs. S1(b) and (c) in Appendix A). Notably, the regulation of immune responses, inflammation, and ROS-related processes during the pneumonitis stage seems to be the most critical aspect of CKI activity in RILI treatment. These pathways are intricately connected: immune cells generate ROS through inflammatory processes, while ROS further exacerbate inflammation and tissue damage. In this network, alkaloids, flavonoids, and triterpenoid saponins appear to exert synergistic effects on molecules across multiple modules.
Furthermore, to gain more comprehensive insights into RILI-related molecular mechanisms, we analyzed the scRNA-seq dataset GSE206426 (
Fig. 2(d)). KEGG enrichment analysis revealed that biological processes within the ROS-related module, including oxidative stress response and ROS regulation, were significantly enriched by DEGs identified from the scRNA-seq data (
Fig. 2(e)). Key biomolecules involved in ROS and oxidative stress, such as nuclear factor erythroid 2 like 2 (NFE2L2), NAD(P)H quinone dehydrogenase 1 (NQO1), superoxide dismutase 1 (SOD1), catalase (CAT), peroxiredoxin 1 (PRDX1), and thioredoxin-interacting protein (TXNIP), exhibited consistent expression differences between normal and irradiated samples across the two types of animal models (
Fig. 2(f)). Finally, by integrating all the molecular mechanisms of RILI and the predicted holistic targets of CKI and its compounds, we constructed a multilayered network to characterize the potential effects of different CKI compounds on three central modules: ROS-related processes, the immune response, and inflammation (
Fig. 2(g)). Among the CKI compounds, alkaloids were identified as the most influential due to their broad effects across multiple modules, while flavonoids and phenolic glycosides also played significant roles in modulating several key processes.
Additionally, we predicted key compound combinations from the compositions of CKI by considering two factors: drug-likeness (estimated by QED scores) and their proximity to the molecular mechanisms of RILI as predicted by a network-based algorithm [
38]. Among the compounds deemed most important, alkaloids exhibited both high QED and high closeness scores (
Fig. 2(h)), suggesting that they are likely to contribute significantly to the regulatory effects of CKI on RILI. More specifically, by analyzing the top 10 compounds according to the network-based closeness scores, we identified four alkaloids—matrine, oxymatrine, sophoridine, and oxysophocarpine—which also demonstrated favorable drug-likeness with high QED scores (
Fig. 2(i)). Finally, to validate the compound combination we predicted, we prepared solutions based on the mass percentage of each alkaloid (matrine (3.39 mg∙mL
−1), oxymatrine (9.58 mg∙mL
−1), sophoridine (1.12 mg∙mL
−1), and oxysophocarpine (3 mg∙mL
−1)). In animal experiments, we observed, via H&E staining, that on Days 7 and 14, the lung tissue of normal mice maintained a clear and organized alveolar structure. In contrast, the 15 Gy irradiation group (NC group) showed significant infiltration of inflammatory cells, primarily neutrophils and mononuclear macrophages, in the alveolar walls. The alveolar walls displayed varying degrees of collapse, resulting in lung tissue damage, disappearance of alveolar cavities, and mild to moderate degeneration of bronchial epithelial cells with a disordered arrangement. After treatment with the alkaloid combination (
Fig. 2(j), Figs. S1(d) and (e) in Appendix A) or CKI, the lung tissue structure of the mice remained relatively intact, with reduced inflammatory cell infiltration and a clearer, more complete alveolar structure, particularly in the CKI-H group, demonstrating a dose-dependent therapeutic effect.
3.3. Assistance of transcriptomic data in the discovery of the key regulatory mechanism of CKI in RILI
To perform a more comprehensive analysis of the mechanism by which CKI influences the treatment of RILI, we conducted a transcriptomic assay on both irradiated animals and those subjected to CKI intervention postirradiation (Fig. S2(a) in Appendix A). DEGs were identified using the DESeq2 algorithm (Figs. S2(b) and (c) in Appendix A), with a total of 175 genes exhibiting notable differential expression between pre- and post-intervention conditions (
Fig. 3(a)). Specifically, in the irradiated samples compared with the nonirradiated controls, 371 genes were downregulated, whereas 912 genes were upregulated (
Fig. 3(b), Fig. S2(d) in Appendix A). In contrast, after CKI intervention, 235 genes whose expression levels were downregulated and 23 genes whose expression levels were upregulated were detected (
Fig. 3(c), Fig. S2(e) in Appendix A).
A closer examination revealed that several GO terms, including immune response, inflammatory response, lymphocyte activation and differentiation, and leukocyte activation and differentiation, were significantly enriched among the DEGs in the irradiated samples (
Fig. 3(d), Fig. S2(f) in Appendix A). Following CKI intervention, there was a marked enrichment in GO terms related to the activation of the innate immune response and cellular responses to interferon (IFN)-α, IFN-β, and IFN-γ (
Fig. 3(e), Fig. S2(g) in Appendix A). The results of the KEGG enrichment analysis further elucidated the involvement of key signaling pathways. In irradiated animals, the DEGs were predominantly associated with cytokine-cytokine receptor interactions, mitogen-activated protein kinase (MAPK) signaling, and phosphoinositide 3-kinase (PI3K)-protein kinase B (Akt) signaling pathways. However, after CKI intervention, the DEGs were enriched primarily in pathways such as MAPK signaling, cytokine-cytokine receptor interactions, and Janus kinase (JAK)-signal transducer and activator of transcription (STAT) signaling. Additionally, several immune system-related signaling pathways, including natural killer cell-mediated cytotoxicity and NOD-like receptor signaling, were also significantly enriched (Figs. S3(a) and (b) in Appendix A).
To gain deeper insights into the functional impact of these gene expression changes, we performed GSEA on the DEGs derived from the transcriptomic data. Analysis revealed that molecular functions related to antioxidant activity were significantly enriched in the top enriched GO terms and KEGG pathways (
Fig. 3(f)). Conversely, the biological process of the innate immune response was significantly suppressed. Furthermore, GSEA revealed that in the irradiated samples, the inflammatory response, immune response, and NOD-like receptor signaling pathways were significantly activated (Fig. S3(c) in Appendix A). After CKI intervention, immune-related biological processes and pathways, such as the immune response, activation of the innate immune response, and NOD-like receptor signaling, were significantly suppressed (Fig. S3(d) in Appendix A).
Similar to the findings from the above computational prediction, CKI showed multiple modular regulatory effects on many aspects, such as immune activity, inflammation, oxidative stress, and cellular processes. The results of this more intricate analysis suggest that CKI has a profound effect on antioxidant stress and modulates immune-related pathways, which might contribute to its therapeutic effects in the treatment of RILI. We also obtained hints from both computational predictions and transcriptomics, suggesting that the antioxidant stress effect, as its dominant role in these analyses, may be a core mechanism of CKI in the treatment of RILI.
3.4. Regulation of CKI against RILI is centered on antioxidative stress
After irradiation, mouse body weight decreased within the first 7 d but gradually recovered and increased. Compared with the NC group, the CKI-H and ALT groups experienced significant weight gain (
p < 0.01) (
Fig. 4(a)). ELISA revealed that compared with those in the nonirradiated CON group, serum tumor necrosis factor (TNF)-α and IL-6 cytokine levels were significantly elevated in the NC group (
p < 0.01). Treatment with different doses of CKI for 7 and 14 d reduced TNF-α and IL-6 levels in a dose-dependent manner, with the effect in the CKI-H group being most pronounced (
p < 0.01). Treatment with ALT, a positive control drug, significantly decreased the expression of inflammatory cytokines (
Fig. 4(b)).
4-Hydroxynonenal (4-HNE), a harmful metabolite from fatty acid peroxidation, was significantly elevated in the lung tissue of irradiated Model mice compared with that in the Control group. Treatment with different doses of CKI for 14 d reduced 4-HNE expression, with better results in the CKI-M and CKI-H groups than in the CKI-L group (Fig. 4(c), Figs. S4(a) and (b) in Appendix A). Similarly, the level of 8-hydroxy-2′-deoxyguanosine (8-OHdG), a marker of oxidative DNA damage, was significantly greater in the NC group than in the CON group. After CKI treatment for 14 d, 8-OHdG expression decreased in a dose-dependent manner, with the greatest reduction occurring in the CKI-H group (Fig. 4(d), Figs. S4(c) and (d) in Appendix A). Myeloperoxidase (MPO), an enzyme that is indicative of inflammation, was significantly elevated in the NC group. After 14 d of CKI treatment, MPO expression decreased in a dose-dependent manner, with the best therapeutic efficacy observed in the CKI-H group. Inducible nitric oxide synthase (iNOS) levels were elevated in the NC group, indicating inflammation. Dual-fluorescence labeling (red for F4/80 and green for iNOS) revealed a significant increase in iNOS
+F4/80
+ cells in the lung tissue of the NC group. CKI treatment reduced these cells in a dose-dependent manner, with CKI-H showing the greatest effect (
Fig. 4(d)).
At the transcript level,
iNOS mRNA was significantly increased, whereas Yeast extract-induced macrophage protein 1 (
YM-1) and arginase-1 (
ARG-1) mRNA levels were decreased in the NC group (
p < 0.01). After CKI treatment,
iNOS mRNA levels decreased, whereas
YM-1 and
ARG-1 levels increased in a dose-dependent manner, with CKI-H being the most effective (
p < 0.01). CKI upregulated the expression of some M2 macrophage-related genes and downregulated the expression of certain M1 macrophage-related genes, potentially suggesting that CKI interferes with macrophage-associated biological processes in RILI (
Fig. 4(e)). Compared with those in the Control group, the mRNA levels of the inflammatory cytokines
IL-1β,
IL-6, and
TNF-α (
Fig. 4(f)) were significantly elevated in the NC group (
p < 0.01). After CKI treatment, these cytokine levels were reduced in a dose-dependent manner, with CKI-H showing the best efficacy (
p < 0.01). Additionally,
IFN-γ mRNA levels were elevated, whereas
IL-10 mRNA levels were decreased in the NC group. CKI treatment reduced
IFN-γ mRNA levels and increased
IL-10 mRNA levels, with the best effect observed in the CKI-H group (
p < 0.01). The anti-apoptotic effect of CKI has also been confirmed (
Fig. 4(g)), as it can reduce mRNA expression of
BAX while increasing the expression of
BCL-2 (
p < 0.01).
In regard to oxidative stress, the expression levels of NQO1, heme oxygenase-1 (HO-1), and glutathione peroxidase-1 (GPX-1), which are key enzymes for maintaining redox homeostasis, were significantly lower in the lung tissue of irradiated Model mice than in that of Control mice. CKI treatment increased the levels of these enzymes in a dose-dependent manner, with the best therapeutic effect observed in the CKI-H group (
Fig. 4(h)). Similarly, the mRNA levels of
NQO1 and
HO-1 were significantly decreased in the NC group but increased with CKI treatment in a dose-dependent manner (
Fig. 4(i)). The CKI-H group again had the most significant effect (
p < 0.01). These findings collectively suggest that CKI has the potential to regulate oxidative stress damage in irradiated lung tissue cells.
The whole-genome perspective of transcriptomics, combined with experiments such as qPCR and ELISA, partially validated the multiple modular regulatory effects of CKI on RILI, including apoptosis, immunity, inflammation, and oxidative stress. It also highlights the important role of the biomolecular network centered on immunity/inflammation/oxidative stress in CKI intervention for RILI, especially in the pneumonia phase.
3.5. NRF2-mediated regulation of CKI in the oxidative stress response
Building upon our previous research, we focused on NRF2, a pivotal regulator of oxidative stress [
44], to further validate its role and explore its therapeutic potential. In the CON group, the lung tissue exhibited well-preserved alveolar architecture with a regular and organized arrangement. In contrast, the lung tissue of irradiated mice in the NC group exhibited significant pathological alterations, including the infiltration of inflammatory cells, predominantly neutrophils, and mononuclear macrophages, into the alveolar walls. These alterations led to varying degrees of alveolar wall collapse, resulting in the loss of alveolar spaces and mild to moderate degeneration of bronchial epithelial cells, accompanied by a disorganized tissue structure. Following treatment with CKI-H, the lung tissue exhibited relative structural preservation, with a marked reduction in inflammatory cell infiltration and clearer, more intact alveolar structures. The pathological features of the DMSO-treated group were similar to those of the NC group, including prominent inflammatory cell infiltration and alveolar collapse. ML385 is a selective NRF2 inhibitor that has been used to block NRF2 activation and assess its roles in various cellular processes, including oxidative stress and inflammation. At the molecular level, BLI was performed to analysis the dissociation constant (
Kd) in molarity and revealed that key compounds in CKI, including matrine (
Kd = 1.59 × 10
-4 mol∙L
-1) and sophoridine (
Kd = 1.27 × 10
-4 mol∙L
-1), weakly bound to NRF2 (Figs. S5(a)-(d) in Appendix A), and NRF2 nuclear translocation experiments revealed that CKI-H increased NRF2 nuclear entry (Figs. S5(e) and (f) in Appendix A), whereas ML385 inhibition significantly decreased NRF2 translocation. Notably, treatment with the NRF2 inhibitor ML385 substantially diminished the protective effects of CKI on lung tissue pathology (
Fig. 5(a), Fig. S5(g) in Appendix A).
Protein level validation revealed that the expression levels of NQO1 and HO-1, crucial antioxidant enzymes, were significantly lower in the lung tissue of irradiated mice in the NC group than in that of normal controls. After 14 d of CKI-H treatment, the protein levels of NQO1 and HO-1 significantly increased (
p < 0.01) in the irradiated mice, demonstrating the restorative effect of CKI on oxidative stress pathways. However, ML385 treatment led to a marked reduction in the expression of these proteins, effectively reversing the protective effect of CKI on oxidative stress and cellular damage (
Fig. 5(b)). At the transcriptional level, we observed significant upregulation of
iNOS mRNA expression and a concurrent reduction in
ARG-1 mRNA expression in the NC group compared with those in control mice (
p < 0.01). After 14 d of CKI-H treatment, a notable shift was observed, with reduced
iNOS expression and increased
ARG-1 expression, highlighting the potential of CKI to modulate macrophage polarization toward an anti-inflammatory phenotype. In contrast,
iNOS expression increased and
ARG-1 expression decreased in the DMSO group, which is consistent with the trends observed in the NC group, whereas the NRF2 inhibitor ML385 further exacerbated this imbalance, increasing
iNOS expression and suppressing
ARG-1 expression. Additionally, the expression levels of key antioxidant genes, such as
NQO1,
HO-1, and
GPX-4, were significantly lower in the lung tissue of irradiated mice than in that of control mice (
p < 0.05). CKI treatment led to the significant upregulation of these genes (
p < 0.01), providing further evidence for its protective role against oxidative damage. ML385 treatment reversed these effects, suppressing the expression of
NQO1 and
HO-1 and thereby disrupting the modulation of oxidative stress by CKI (
Fig. 5(c)).
The oxidative stress markers 4-HNE and 8-OHdG were significantly elevated in the lung tissue of irradiated mice compared with nonirradiated controls. Following CKI-H treatment, the expression levels of both markers significantly decreased, indicating a reduction in oxidative damage. The levels of 4-HNE and 8-OHdG increased in the DMSO group, which is consistent with the trends observed in the NC group. Treatment with ML385 effectively blocked the protective effects of CKI, leading to substantial increases in 4-HNE and 8-OHdG expression, further underscoring the role of NRF2 in mediating the antioxidant effects of CKI (
Fig. 5(d), Figs. S6(a) and (b) in Appendix A). The expression of MPO, a marker of neutrophil infiltration, was significantly greater in the lung tissue of irradiated mice than in that of control mice. Following high-dose CKI treatment, MPO expression was markedly reduced, reflecting the attenuation of inflammatory cell infiltration. MPO expression increased in the DMSO-treated group, which is consistent with the pattern observed in the NC group. Treatment with ML385 reversed the protective effects of CKI on MPO expression, highlighting the involvement of NRF2 in regulating the inflammatory response in lung tissue. Immunofluorescence analysis revealed that the number of iNOS
+F4/80
+ cells was significantly greater in irradiated mice than in control mice. Following CKI-H treatment, the number of these cells significantly decreased, whereas ML385 treatment led to an increase in iNOS
+F4/80
+ cells (
Fig. 5(e), Figs. S6(c) and (d) in Appendix A). The levels of the inflammatory cytokines IL-6 and TNF-α were significantly greater in the lung tissue of irradiated mice than in that of control mice (
p < 0.05). After 14 d of CKI-H treatment, the mRNA expression levels of
IL-6 (
p < 0.05), and
TNF-α (
p < 0.01) were significantly decreased. In contrast, compared with those in the CON group, the cytokine levels in the DMSO group were elevated. Treatment with ML385 resulted in significant increases in
IL-6 and
TNF-α expression, effectively reversing the anti-inflammatory effects of CKI (
Fig. 5(f)). ELISA analysis revealed that serum levels of TNF-α and IL-6 were significantly greater in the NC group than in control mice (
p < 0.05). CKI-H treatment for 14 d significantly reduced serum cytokine levels, whereas ML385 treatment reversed this reduction, further supporting the central role of NRF2 in modulating the inflammatory response (
Fig. 5(g)). Finally, compared with that in normal controls, the activity of SOD, a key antioxidant enzyme downstream of NRF2, was significantly lower in irradiated mice. After CKI-H treatment, serum SOD levels significantly increased (
p < 0.01), suggesting the restoration of antioxidant capacity. ML385 treatment resulted in a significant decrease in serum SOD levels, effectively reversing the protective effects of CKI on oxidative stress (
Fig. 5(h)).
These findings highlight the critical role of NRF2 in mediating the protective effects of CKI in RILI (Fig. (i)). CKI treatment restored lung tissue structure and reduced oxidative stress and inflammation, offering a promising therapeutic approach for managing radiation-induced lung damage. Our findings also underscore the importance of NRF2 signaling in mediating the protective effects of CKI, as evidenced by the reversal of these effects following NRF2 inhibition with ML385. These results suggest that CKI could serve as a potential therapeutic agent for alleviating radiation-induced pulmonary injury by targeting oxidative stress and inflammation pathways.
4. Discussion and conclusion
RILI is a common complication of lung cancer radiotherapy [[
7], [
8], [
9]] and involves DNA damage, oxidative stress, inflammation, and immune changes during the pneumonia phase, followed by fibrosis and ECM production during the fibrosis phase [[
1], [
2], [
3]]. CKI, a TCM formula consisting of KS and BTL, is widely used in clinical practice for treating radiation-induced damage [[
45], [
46], [
47]] and various cancers [[
48], [
49], [
50]] and as an adjunctive cancer therapy. Clinical reports indicate that CKI is effective at treating RILI [[
51], [
52], [
53], [
54]], but its mechanism of action remains unclear. In this study, we employed network target analysis to elucidate the modular regulatory mechanisms of CKI in RILI treatment, utilizing an improved holistic target model that integrates compound quantitative data. By constructing multilayer biological networks, we systematically mapped the regulatory landscape of CKI, identifying alkaloids as the key active components and oxidative stress as the central therapeutic axis. Validation in irradiated animal models confirmed that NRF2 and its downstream molecules, such as HO-1 and NQO1, play crucial roles in the modulation of oxidative stress, immunity, and inflammation by CKI. This integrated approach combining machine learning, transcriptomics, and experimental validation provides new insights into the therapeutic potential of CKI.
In RILI, NRF2 is a potential mediator of oxidative stress [
44] and acts as a crucial transcription factor that regulates oxidative stress and suppresses inflammation [
55,
56]. When activated, NRF2 is translocated to the nucleus and binds to antioxidant response elements (AREs) in the promoter regions of various genes involved in the detoxification and antioxidative response. Among its downstream targets, HO-1 and NQO1 are two critical enzymes involved in protecting cells from oxidative damage [[
57], [
58], [
59]]. HO-1 helps break down heme into biliverdin, free iron, and carbon monoxide, all of which exert antioxidative and anti-inflammatory effects. NQO1 plays a crucial role in protecting cells from oxidative stress by catalyzing the reduction of quinones to prevent the formation of harmful ROS. The regulation of NRF2, HO-1, and NQO-1 by CKI, along with its modulation of various molecules, such as SOD, 4-HNE, and 8-OHdG, demonstrates the role of CKI in inhibiting oxidative stress and oxidative stress-related inflammatory responses in RILI. In addition, after NRF2 is inhibited, the suppressive effects of CKI on molecules such as TNF-α, iNOS, and IL-6 are weakened, indicating that these molecules are influenced primarily by CKI through its roles as downstream targets of NRF2 in the regulation of RILI. In addition to TNF-α, iNOS, and IL-6, CKI also regulates molecules such as IL-1β, YM-1, ARG-1, IL-10, and IFN-γ, which play important roles in inflammation and immune regulation. Finally, BAX and BCL-2 were also regulated by CKI, demonstrating the antiapoptotic effects of CKI and its protective role in cells.
In our study, we confirmed that NRF2 is a pivotal mediator of the therapeutic effects of CKI on RILI and plays a pivotal role in mediating cellular responses to oxidative stress and inflammation, which are central to the pathogenesis of RILI and other types of radiation-induced injury. In our study, we confirmed that CKI regulates oxidative stress in RILI and exerts therapeutic effects through NRF2-mediated regulation through NRF2 inhibition and nuclear translocation experiments. NRF2 activation has been shown to regulate the expression of antioxidant enzymes and detoxifying proteins, thereby protecting cells from radiation-induced oxidative damage and reducing inflammation. Studies have demonstrated that NRF2 activation can mitigate the pathological effects of RILI, including inflammation, fibrosis, and apoptosis, by enhancing the cellular antioxidant capacity and inhibiting proinflammatory cytokine production [
54]. Furthermore, the ability of NRF2 to modulate macrophage function in irradiated lung tissues has been highlighted as a key mechanism through which it alleviates radiation-induced lung fibrosis and promotes tissue repair. Given its central role in these protective mechanisms, NRF2 presents a promising therapeutic target for RILI, offering potential strategies for treatment via NRF2 activators or modulators that enhance its antioxidant and anti-inflammatory effects [
60,
61]. Thus, targeting NRF2 in RILI therapy could provide a novel and effective approach to mitigate lung damage induced by radiation. In addition to the lung, NRF2 confers protection against radiation-induced oral mucositis by diminishing ROS levels and thickening the keratin layer [
62], mitigates ferroptotic death in hippocampal neurons via the pyruvate kinase M2 (PKM2)/NRF2/GPX-4 pathway [
63], and preserves epithelial integrity against ionizing radiation through
Lycium barbarum polysaccharide-glycoprotein-mediated NRF2 activation [
64]. Additionally, we conducted binding experiments between key compounds in CKI and NRF2. Based on the results of the BLI assay, we found that sophoridine and matrine weakly bound to NRF2 (at a concentration of approximately 100 µmol∙L
-1). This weak binding may be a common phenomenon in multicompound systems such as TCM. This could be due to the intrinsic structural characteristics of these natural products, which, when binding to multiple target proteins, may operate in a noncovalent binding mode known as allosteric regulation [
65]. This type of noncovalent allosteric regulation can facilitate the dissociation of a molecule from a protein. Given that prolonged protein activation or inhibition is not always beneficial—as overexpression or excessive suppression can lead to adverse effects—this specific type of noncovalent allosteric regulation may serve as a valuable reference or theoretical foundation for future drug development.
Our study has several limitations. With respect to compound data collection, we were able to obtain quantitative detection data for only a subset of alkaloid compounds. For other alkaloids or compounds from different categories, we could only gather information on their total content within broad groups, such as total sugars and total flavonoids. Fortunately, our improved strategy mitigates the lack of quantitative data by performing a large-scale random sampling of compounds based on their weights, which reduces potential errors. By integrating the available quantitative data, this enhanced method provides more accurate predictions of holistic targets than our previous version does.
In conclusion, we provided a paradigm for research on TCM herbs or formulas such as CKI, integrating an improved strategy that could combine both compound information and quantitative assay results, transcriptomics technologies and animal validations. Through this paradigm, we constructed a modular biological network centered on the regulation of oxidative stress, immunity, and inflammation to elucidate the underlying mechanism of CKI intervention in RILI. CKI targets NRF2 and regulates downstream molecules such as HO-1 and NQO1, which control oxidative stress. On the one hand, it inhibits oxidative stress and modulates the levels of molecules such as SOD and 4-HNE. On the other hand, it further affects immune/inflammation-related molecules such as TNF-α, IL-6, and iNOS, exerting anti-inflammatory effects and immune regulation.
CRediT authorship contribution statement
Boyang Wang: Writing - review & editing, Writing - original draft, Visualization, Methodology, Formal analysis, Conceptualization. Defei Kong: Writing - original draft, Visualization, Validation, Investigation, Data curation. Zhiru Yang: Validation. Jun Kang: Validation. Deyang Sun: Writing - original draft, Investigation. Xiumei Duan: Writing - original draft, Data curation. Jing Jin: Investigation. Tingyu Zhang: Visualization, Investigation. Qingyuan Liu: Visualization, Formal analysis. Hui Yin: Writing - review & editing, Supervision, Project administration, Data curation. Shao Li: Writing - review & editing, Supervision, Project administration, Funding acquisition, Conceptualization.
Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgments
This work was supported by the Innovation Team and Talent Support Program Project of Traditional Chinese Medicine (ZYYCXTD-D-202405) from National Administration of Traditional Chinese Medicine, the National Natural Science Foundation of China (T2341008), and the Pilot Project for Disciplinary Breakthroughs of Ministry of Education (Prevention and Treatment of Multi-System Comorbid Diseases with Traditional Chinese Medicine).
Appendix A. Supplementary data
Supplementary data to this article can be found online at
https://doi.org/10.1016/j.eng.2025.09.018.
PDF
(4917KB)
