Network toxicology study and key target validation of chlorpyrifos-induced nonalcoholic fatty liver disease

Abstract

Chlorpyrifos (CPF) is a widely used insecticide known for its extended persistence in soil and water. It can induce non-alcoholic fatty liver disease (NAFLD) in mice. However, the specific molecular mechanisms responsible for CPF-induced NAFLD remain partially understood. Utilizing network toxicology analysis, we have identified TP53, HSP90AA1, AKT1, and JUN as pivotal driver genes orchestrating NAFLD progression triggered by CPF. A predictive nomogram constructed on these core genes demonstrates promising translational prospects for NAFLD. Gene-set enrichment analysis (GSEA) underscores the significance of the tricarboxylic acid (TCA) cycle and histidine metabolism as critical pathways associated with CPF-induced NAFLD. Evaluation of immune infiltration patterns reveals notable changes in the immunological milieu and accelerated disease advancement in CPF-induced NAFLD. Molecular docking and dynamic simulations provide support for the formation of stable complexes between CPF and proteins encoded by the identified core genes. Ultimately, biomedical experiments confirmed that CPF exacerbates the pathological progression of NAFLD by stabilizing HSP90AA1 protein and promoting the phosphorylation of TP53 and JUN. This study elucidates the molecular mechanisms underpinning CPF-induced NAFLD, providing a theoretical foundation for comprehending its pathogenesis.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-41592-2.

Introduction

Organophosphate chlorinated compounds, notably Chlorpyrifos (CPF), are widely used as insecticides globally due to their high efficacy, broad-spectrum action, and reduced likelihood of pest resistance, making them popular in agriculture^1,2. Although CPF is cost-effective in pest control, its potential risks to human health and the environment have raised concerns. Exposure to CPF can lead to hepatotoxicity³, disrupt gut microbiota balance⁴, and cause neurotoxicity^5,6, particularly endangering early life stages of aquatic and terrestrial organisms, including humans, by interfering with placental formation and function⁷. Despite stricter regulations in China and other countries, CPF continues to pose a significant ecological risk, with the increasing trends in its annual production and consumption being particularly alarming⁸.

NAFLD, now known as metabolic dysfunction-associated steatotic liver disease, has emerged as one of the most prevalent chronic liver diseases globally, with an incidence rate of nearly 30%⁹. Projections indicate that by 2050, the prevalence of NAFLD and its more severe form, non-alcoholic steatohepatitis, will rise significantly, leading to a substantial increase in the demand for hepatocellular carcinoma treatment and liver transplants, as well as a marked increase in liver-related mortality¹⁰. Granted that insulin resistance, overnutrition, and lack of exercise are common causes of NAFLD, an increasing number of studies have shown that exposure to environmental toxicants also plays a key role in the rising prevalence of obesity and NAFLD^11–14. CPF has been demonstrated to facilitate obesity by inhibiting diet-induced thermogenesis in brown adipose tissue, thereby contributing to NAFLD progression^15,16. Furthermore, its metabolites are positively correlated with the severity of NAFLD, and CPF exposure levels may serve as a potential predictor for the progression of this disease¹². However, the molecular mechanisms underlying CPF-induced NAFLD in humans remain to be elucidated. Elucidating these mechanisms is crucial for a deeper understanding of the disease process of CPF-induced NAFLD.

Network toxicology is an emerging interdisciplinary research paradigm grounded in network biology and pharmacology principles¹⁷. It constructs "chemical-target-disease" network models to systematically analyze the toxicity mechanisms and biological effects of chemical substances¹⁸. Its core lies in leveraging bioinformatics, big data analytics, and multi-omics technologies (e.g., genomics, proteomics, and metabolomics) to comprehensively dissect the complex interactions between chemical substances and biological systems¹⁹.

This study aims to elucidate the mechanism underlying CPF-induced NAFLD. Specifically, it explores the interaction mechanisms between CPF and NAFLD target genes, and their potential role in inducing NAFLD through specific signaling pathways. These findings will provide the theoretical underpinnings for public health policies and prevention strategies, while simultaneously identifying novel avenues for targeted therapy in CPF-associated NAFLD.

Materials and methods

Elucidation and assembly of pivotal gene cohorts in NAFLD triggered by CPF

A dataset encompassing genes associated with CPF was obtained from the Comparative Toxicogenomics Database (CTD, https://ctdbase.org/). Additionally, a collection of genes pertinent to NAFLD was extracted from the GeneCards database (https://www.genecards.org/). An integrated gene dataset was constructed using the VennDiagram package in R to explore the possible toxicological effects of CPF on NAFLD²⁰.

Data collection

Datasets GSE48452 and GSE33814 were retrieved from the Gene Expression Omnibus (GEO) database, and R language combined with the sva package was used to perform batch effect correction on the expression matrix files of GSE33814 and GSE48452 datasets²¹. After merging, these datasets collectively included 27 normal cases and 63 NAFLD cases, forming the discovery cohort. The GSE164760 dataset serves as the validation cohort, which includes 6 healthy samples and 74 NAFLD samples.

GO enrichment analysis and KEGG pathway analysis

To thoroughly elucidate the biological functions of the pivotal gene sets induced by CPF in NAFLD, we employed the clusterProfiler package to conduct comprehensive GO and KEGG pathway enrichment analyses^22–24.

Protein interaction mapping and pivotal gene selection

The pivotal gene sets induced by CPF in NAFLD were uploaded to the STRING database (https://cn. string-db. org/), with Homo sapiens selected as the species under investigation to identify the critical interactions between proteins²⁵. Thereafter, the resultant interaction data were imported into Cytoscape 3.7.1, where the cytoHubba plugin was leveraged to screen for the top 10 hub genes, as determined by descending order of the Degree metric, across multiple algorithmic methods²⁶. Ultimately, the UPSET tool was applied to pinpoint the core genes among them.

Nomogram for NAFLD risk prediction

In the discovery cohort, the Visualization and Missingness (VIM) package was employed to visually address missing data within the dataset, while the logistic regression model was meticulously fitted using the lrm function from the rms package. The nomogramFormula package was subsequently harnessed to calculate the nomogram scores, and the pROC package was utilized to perform a receiver operating characteristic analysis, thereby assessing the discriminatory power of the model²⁷. The rmda package was deployed to conduct a clinical decision curve analysis (DCA), evaluating the clinical utility of the model across various threshold probabilities²⁸. The predictive accuracy and goodness-of-fit of the model were appraised through calibration curves and Hosmer–Lemeshow tests. The model was subjected to testing by a validation cohort, thereby confirming its robustness and reliability in predictive performance.

Single-gene GSEA analysis

Leveraging a suite of bioinformatics packages in R, including ggplot2, limma, pheatmap, ggsci, clusterProfiler, enrichplot, patchwork, and org. Hs. eg. db, a gene set enrichment analysis (GSEA) centered on an individual core gene was executed. Samples were stratified into high and low expression groups based on the expression levels of the core genes. Differential expression analysis was performed using the limma package, followed by statistical testing with the eBayes function, thereby identifying significantly differentially expressed genes. Heatmaps were illustrated using the pheatmap package, and volcano plots were generated with ggplot2 to visualize their expression changes. Additionally, KEGG pathway enrichment analysis was conducted on these differentially expressed genes using the clusterProfiler package, and the top five pathways with the highest and lowest enrichment scores were visualized employing the gseaplot2 function from the enrichplot package.

Assessing immune cell infiltration

The CIBERSORT algorithm was leveraged for estimating the relative abundances of different cell types in samples²⁹. Barplots were constructed using the barplot function to illustrate the distribution of relative percentages across various samples, with the rainbow function used to assign colors to each sample. Box plots were generated with the ggplot2 package to further reveal the distributional traits of the data.

Construction and analysis of core gene-miRNA interaction networks

The miRNAs associated with core genes were procured via the networkanalyst platform, whereupon the interaction network linking core genes to miRNAs was meticulously constructed (https://www.networkanalyst.ca/)³⁰. Subsequently, the Cytoscape software was engaged to refine the visual appeal of the network diagram. The FunRich was subsequently harnessed to ascertain the transcription factors associated with the miRNAs, and to execute GO enrichment analyses, which served to elucidate the roles of the miRNAs and the biological pathways with which they are engaged^31,32.

Molecular docking

The SDF file of CPF (CID: 2730) was downloaded from PubChem (https://pubchem.ncbi.nlm.nih.gov/). The protein PDB files (3DCY for TP53; 2QF6 for HSP90AA1; 7WM2 for AKT1; 1JUN for JUN) were obtained from the RCSB PDB (https://www.rcsb.org/). All proteins and molecular files were converted to PDBQT format, with water molecules removed and polar hydrogen atoms added. The grid box was centered to cover the protein domains and allow for free molecular movement. The docking pocket was set as a 30 Å × 30 Å × 30 Å cube with a grid spacing of 0.05 nm. Molecular docking was performed and visualized by AutoDock Vina 1.2.2³³.

Molecular dynamics simulation

Molecular dynamics simulations were performed using the GROMACS 2022 software package³⁴. The force field parameters for the small molecules were based on the General Amber Force Field, while the protein was described using the AMBER14SB force field^35,36. The solvent water molecules were modeled using the TIP3P model. The simulation system was constructed by merging the files of the protein and the small molecule ligand to form a complex.

The simulations were conducted under constant temperature and pressure (NPT) conditions with periodic boundary conditions. The LINCS algorithm was employed to constrain all atoms involved in hydrogen bonds, and an integration time step of 2 fs was used. Electrostatic interactions were calculated using the Particle-Mesh Ewald method with a cutoff value of 1. 2 nm. Nonbonded interactions were truncated at 10 Å and updated every 10 steps.

Temperature control was achieved using the V-rescale method, maintaining the simulation temperature at 298 K. Pressure control was performed using the Berendsen method, keeping the pressure at 1 bar. The system was first equilibrated for 100 ps under NVT (constant number of particles, volume, and temperature) and NPT conditions at 298 K, followed by a 100 ns MD simulation of the complex. The simulation trajectory was saved every 10 ps.

After the simulation, the trajectories were analyzed using the VMD and PyMOL software for visualization. Binding free energy analysis was performed using the g_mmpbsa program, based on the Molecular Mechanics Poisson-Boltzmann Surface Area (MMPBSA) method to calculate the binding free energy between the protein and the small molecule ligand³⁷.

Animal handling

Male wild-type C57/BL6 mice were maintained at the Animal Facility Center of Harbin Medical University School of Public Health under specific pathogen-free (SPF) conditions. Humane care was provided in compliance with the "Principles of Laboratory Animal Care" and the "Guide for the Care and Use of Laboratory Animals." Male six-week-old mice weighing approximately 18–22 g were used to mimic NAFLD conditions, while male eight-week-old mice weighing approximately 22–25 g were utilized for primary hepatocyte isolation. All mice were purchased from Liaoning Changsheng Biotechnology Co., Ltd (License No.: SCXK(Liao)2020–0001). Mice were fed a high-fat diet (HFD) from Research Diets (catalog no. D09100310) for 16 weeks. Rodent Diet With 40 kcal% Fat (Mostly Palm Oil), 20 kcal% Fructose and 2% Cholesterol. Anesthesia was induced by intraperitoneal injection of pentobarbital sodium (30 mg/kg). Euthanasia was performed via cervical dislocation, with animal death confirmed through continuous monitoring of respiratory cessation and cardiac arrest to ensure compliance with internationally recognized ethical standards.

The experimental protocol was approved by the local Institutional Animal Ethics Committee (Approval No: IRB3025725) in accordance with the Animal Experimentation Law. This study adheres to the ARRIVE reporting guidelines, and we confirm that all experimental procedures have been conducted in compliance with relevant guidelines and regulations.

Cell acquisition and culture

HepG2 cells were purchased from Procell (CL-0103); Isolation of primary hepatocytes was performed as follows: Eight-week-old C57BL/6 mice were anesthetized, and inferior vena cava cannulation was carried out. The liver was first perfused with HBSS buffer containing 0.1% EDTA (37 °C,pre-perfusion fluid), followed by continuous perfusion with digestive fluid once the liver was fully perfused. When the liver turned pink, the hepatic hilum was clamped, and the digestive fluid was perfused at a constant speed. At the end of perfusion, the liver was quickly removed and placed into serum-free DMEM. After hepatocytes were released, the cell suspension was filtered through a 70 µm mesh, centrifuged at 1000 rpm for 2 min, and further resuspended and centrifuged using DMEM containing PERCOLL. The cells were then washed with serum-free DMEM.

CPF was purchased from MCE (HY-B0815). For lipid intervention, palmitic acid (PA, obtained from MCE, cat. no. HY-N0830) and oleic acid (OA, acquired from Aladdin, cat. no. O108484) were prepared at a 1:2 ratio, resulting in final concentrations of 0.33 mM and 0.66 mM, respectively. All cells were cultured at 37 °C with 5% CO₂ in DMEM medium supplemented with 1% penicillin–streptomycin-gentamicin antibiotics and 10% fetal bovine serum. HepG2 cells were passaged approximately every 2 days.

Cell proliferation and cytotoxicity analysis

The CCK-8 kit (Abbkine, KTA1020) was used. Briefly, 100 µL of culture medium containing cells at 30%-50% confluence was added to each well of a 96-well plate and incubated for 2 days at 37 °C in a 5% CO₂ incubator. On the third day, 10 µL of CCK-8 reagent was added to each well and incubated for an additional 2 h. The absorbance of each well was measured at 450 nm using a microplate reader.

Evaluation of liver functionality

Reagent kits for aspartate aminotransferase (AST), alanine aminotransferase (ALT), total cholesterol (TC), and triglyceride (TG) from Nanjing Jiancheng Bioengineering Institute were utilized to assess liver function. Following a 2-day exposure to CPF, the corresponding reagents were sequentially administered in accordance with the manufacturer’s protocol. The absorbance of AST and ALT was determined at 505 nm, while that of TC and TG was ascertained at 500 nm through microplate spectrophotometry. Concurrently,the protein concentration was determined using the BCA method, with absorbance measured at 562 nm.

Oil red O staining

The Oil Red O staining solution was purchased from Nanjing Jiancheng Bioengineering Institute. The culture medium was removed from the culture plate, followed by three PBS washes to dryness. Subsequently, the red Oil Red O staining solution was added for staining for 5 min. After staining, the plate was washed with PBS, treated with counterstaining solution for 1 min, and then washed with PBS until the washings were colorless. After microscopic imaging, quantitative analysis was performed using ImageJ software.

Reverse transcription quantitative polymerase chain reaction (RT-qPCR)

Total RNA was extracted from cells, animal tissues and human tissues. All human liver specimens originated from surgically resected waste tissues. cDNA was routinely synthesized. The expression levels of core genes in various tissues and different treated cell groups were analyzed using the Step-one Real-time Fluorescent Quantitative PCR instrument from ABI Company and the SYBR Premix Ex Taq kit from Takara Company, with β-actin as the internal reference. The primers used are listed in Supplementary Table 1.

Protein and stability analysis

Total proteins were extracted using RIPA lysis buffer (Beyotime Biotechnology, Shanghai, China) supplemented with protease inhibitors (Roche, Switzerland). After extraction, the protein concentration was measured using a BCA protein quantification kit (Beyotime Biotechnology, Shanghai, China). The protein samples were separated by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE), transferred onto nitrocellulose membranes, and subsequent blocking was performed with 5% non-fat milk.

Following blocking, the nitrocellulose membranes were incubated with primary antibodies against each target protein at 4℃ overnight. The detailed information of the primary antibodies is as follows: β-Tubulin (Proteintech, Wuhan, China, Cat No.: 10094–1-AP, dilution ratio: 1:1000), AKT1 (Wanlei Biotechnology, Shenyang, China, Cat No.: WL01652, dilution ratio: 1:1000), JUN (Wanlei Biotechnology, Shenyang, China, Cat No.: WL02863, dilution ratio: 1:1000), TP53 (Wanlei Biotechnology, Shenyang, China, Cat No.: WL01919, dilution ratio: 1:1000), phosphorylated AKT1 (Ser473 site, Wanlei Biotechnology, Shenyang, China, Cat No.: WLP001, dilution ratio: 1:1000), phosphorylated JUN (Ser243 site, Wanlei Biotechnology, Shenyang, China, Cat No.: WL05439, dilution ratio: 1:1000), phosphorylated TP53 (Ser315 site, Wanlei Biotechnology, Shenyang, China, Cat No.: WL02504, dilution ratio: 1:1000), and HSP90AA1 (Proteintech, Wuhan, China, Cat No.: CL488-60,318, dilution ratio: 1:1000).

After primary antibody incubation, the nitrocellulose membranes were washed three times with TBST buffer, 5 min per wash. Subsequently, horseradish peroxidase (HRP)-conjugated anti-rabbit/mouse IgG secondary antibodies (LI-COR Bioscience, USA) were added, and the membranes were incubated in the dark for 50 min with the secondary antibody diluted at a ratio of 1:10,000. The Odyssey infrared imaging system (LI-COR Bioscience, USA) was used for imaging and quantitative analysis of protein bands, with β-Tubulin as the internal reference protein to calibrate the experimental results.

For protein stability analysis, cells were treated with cycloheximide (CHX, MCE, China, Cat No.: HY-12320) at a working concentration of 30 μg/mL. Cell samples were collected at three time points: 0 h, 3 h, and 6 h after treatment. The subsequent protein extraction and detection procedures were performed with reference to the aforementioned Western Blot (WB) experimental method.

Raw protein data are provided in the supplementary materials. The CHX assay was performed using the 26,616 protein marker for single‑side labeling, with two experimental groups between the two marker bands. To ensure the authenticity and reliability of the raw data, the original raw protein blot images with dual‑side markers are shown in the supplementary data.

Statistical analysis

In this study, data analysis was performed using GraphPad Prism 10.1.2 software. For comparisons between two independent groups, the independent samples t-test was applied to assess whether there was a significant difference in means. For comparisons involving three or more groups, one-way analysis of variance (ANOVA) was used to evaluate differences in means among the groups. A p-value of less than 0.05 was considered statistically significant.

Results

Identification of CPF-induced NAFLD genes

The Venn diagram demonstrated 582 overlapping genes between the CPF-related genes obtained from the CTD database and the NAFLD genes collected from the GeneCards database (Supplementary Table 2, Fig. 1A). GO enrichment analysis reveals enrichment in BP, including the fatty acid metabolic process, the regulation of small molecule metabolic process, and the regulation of lipid metabolic process, which are processes that contribute to the formation of NAFLD. Within the context of MF,enrichments were observed in signaling receptor activator activity, receptor ligand activity, DNA-binding transcription factor binding, ubiquitin-like protein ligase binding, RNA polymerase II-specific DNA-binding transcription factor binding, ubiquitin protein ligase binding, and antioxidant activity. CC analysis provides evidence of significant enrichment within vesicle lumens, cytoplasmic vesicle lumens, secretory granule lumens, and endoplasmic reticulum lumens. Additionally, endocytic vesicles, membrane rafts, and ficolin-1-rich granule lumens also exhibit significant enrichment, suggesting their roles in intracellular trafficking and signaling pathways (Fig. 1B). Employing a KEGG-based approach, we further confirmed the significant role of overlapping genes in insulin resistance, diabetes, cancer, NAFLD, and atherosclerosis, which are consistent with NAFLD caused by overeating and insulin resistance^22–24. Similarly,these genes are enriched in multiple key signaling pathways, including the FoxO signaling pathway, HIF-1 signaling pathway, AMPK signaling pathway, TNF signaling pathway, PI3K-Akt signaling pathway, Adipocytokine signaling pathway, PPAR signaling pathway, Longevity regulating pathway, Toll-like receptor signaling pathway, and IL-17 signaling pathway (Fig. 1C).

Fig. 1.

Identification of overlapping genes between NAFLD and CPF. (A) Venn Diagram showing overlapping gene sets between CPF and NAFLD. (B,C) GO enrichment analysis and KEGG pathway analysis of overlapping genes between CPF and NAFLD^22–24.

TP53, HSP90AA1, AKT1 and JUN are core genes of NAFLD mediated by CPF

To further ascertain the core genes mediated by CPF in the pathogenesis of NAFLD,PPI network was constructed for these core genes, and seven distinct algorithms were employed to identify the top 10 most significant genes as determined by each method (Fig. 2A,B). The Radiality algorithm revealed key genes, including RELA, STAT3, TP53, MAPK8, ESR1, EGFR, HSP90AA1, AKT1, HIF1A, and JUN. The MNC algorithm identified STAT3, TP53, ESR1, TNF, HSP90AA1, IL6, AKT1, HSP90AB1, NFKB1, and JUN as critical genes. The EPC algorithm ascertained RELA, STAT3, TP53, ESR1, HSP90AA1, IL6, AKT1, TLR4, NFKB1, and JUN as significant. The Degree algorithm determined STAT3, TP53, CTNNB1, TNF, EGFR, HSP90AA1, IL6, AKT1, TLR4, and JUN to be of paramount importance. The Closeness algorithm yielded RELA, STAT3, TP53, ESR1, EGFR, HSP90AA1, IL6, AKT1, HIF1A, and JUN as key genes. The Betweenness algorithm identified PPARG, HSPA8, FN1, TP53, EGFR, HSP90AA1, IL6, AKT1, GAPDH, and JUN as crucial. The Stress algorithm disclosed FN1, PPARG, STAT3, TP53, ESR1, EGFR, HSP90AA1, IL6, AKT1, and JUN as essential genes. Subsequent to performing an intersection analysis of the genes filtered through various algorithms using the UPSET method, TP53, HSP90AA1, AKT1 and JUN were ultimately pinpointed as the core genes in CPF-mediated NAFLD (Fig. 2C). Conclusively, the interplay among the four genes was highlighted (Fig. 2D).

Fig. 2.

Building protein networks and identifying core genes. (A) The PPI network map constructed from overlapping genes. (B) The top ten key genes detected with seven algorithms. (C) The upset plot of genes identified based on seven different algorithms. (D) The protein interaction network among the four core genes.

CPF-mediated NAFLD diagnostic model with efficacy evaluation

Initially, to facilitate subsequent research, cohorts were meticulously constructed, integrating diverse datasets and ensuring comprehensive representation of the study population. The Discovery cohort was meticulously assembled by merging GSE33814 and GSE48452, employing batch-correction techniques to harmonize the datasets, and ultimately comprising 27 normal cases and 63 NAFLD cases (Supplementary Fig. 1). The Validation cohort consisted of GSE164760, containing 6 normal cases and 74 NAFLD cases (Supplementary Table 3). Leveraging the core genes TP53, HSP90AA1, AKT1 and JUN, a diagnostic nomogram model for NAFLD has been meticulously crafted using both the Discovery cohort and the Validation cohort (Fig. 3A,B). In the ROC curve analysis, when the risk scoring was conducted using the Discovery cohort, the AUC value was 0.7625 (Fig. 3C). Interestingly, when generating the ROC curve with the Validation cohort, the AUC value associated with the risk scoring rose to 0.9189 (Fig. 3D). Through calibration curve analysis, the model we developed for predicting the risk of NAFLD has demonstrated commendable predictive accuracy, exhibiting a high degree of congruence between the predicted probabilities and the actual observed probabilities. Furthermore, the Hosmer–Lemeshow test p-values for the two models, standing at 0.6216 and 0.7321 respectively, exceed the threshold of 0.05, thereby indicating there is no significant discrepancy between the model predictions and the actual observations (Fig. 3E,F). DCA reveals a positive net benefit is consistently observed across the majority of high-risk threshold levels. It is particularly noteworthy that the peak net benefit of the model is attained when the high-risk threshold is set between 0.2 and 0.6, thereby substantiating the clinical applicability of the model (Fig. 3G,H).

Fig. 3.

Constructing a nomogram model to assess the diagnostic value of NAFLD. (A,B) Risk prediction nomograms for NAFLD constructed based on core genes in the discovery cohort and the validation cohort. (C,D) ROC curves for the discovery cohort and the validation cohort. (E, F) The calibration curves for the two groups. (G,H) Decision curve analysis constructed for the discovery cohort and the validation cohort.

Single-gene GSEA analysis of core genes

For single-gene GSEA analysis, differentially expressed genes (DEGs) between high and low expression groups of TP53, HSP90AA1, AKT1 and JUN was calculated, and heatmaps and volcano plots were generated (Supplementary Fig. 2). In Fig. 4, GSEA analysis emphasizes the Citrate cycle (TCA cycle), Hepatocellular carcinoma, Histidine metabolism, and Nicotine addiction pathways are consistently present across three distinct gene sets. The pathways of Chemical carcinogenesis-receptor activation, Human immunodeficiency virus 1 infection, Propanoate metabolism, Protein export, and Ubiquinone and other terpenoid-quinone biosynthesis are observed in only two gene sets. In addition, the TP53 gene was found to be uniquely involved in DNA replication, Focal adhesion, mTOR signaling pathway, Other glycan degradation, Proteasome, Proteoglycans in cancer, Ribosome, and Taste transduction. The HSP90AA1 gene is linked with pathways like Chemical carcinogenesis-receptor activation, Human immunodeficiency virus 1 infection, Propanoate metabolism, Protein export, Tuberculosis, and Ubiquinone and other terpenoid-quinone biosynthesis. The AKT1 gene is exclusively associated with pathways such as Chemical carcinogenesis-receptor activation, Human immunodeficiency virus 1 infection, NOD-like receptor signaling pathway, Propanoate metabolism, Protein export, and Ubiquinone and other terpenoid-quinone biosynthesis. Lastly, the JUN gene is singularly involved in pathways including Apelin signaling pathway, Ascorbate and aldarate metabolism, Diabetic cardiomyopathy, Fatty acid degradation, Hepatitis C, Thermogenesis, Tryptophan metabolism, and Virion-Ebolavirus, Lyssavirus and Morbillivirus.

Fig. 4.

Single-gene GSEA analysis with core genes: TP53, HSP90AA1, AKT1, JUN.

Immune cell landscape and core gene associations

The CIBERSORT algorithm was used to conduct a comprehensive survey of the abundance of immune cells, in order to elucidate the correlation between these core genes and the immune cell infiltration in NAFLD induced by CPF (Fig. 5A). The onset of NAFLD is significantly associated with marked alterations in immune cell infiltration, particularly characterized by an increased infiltration of naive B cells, macrophages of the M0 subtype, and resting mast cell, whereas there is a notable decrease in the infiltration of macrophages of the M2 subtype (Fig. 5B). Plasma cells and T cells CD4 memory activated exhibit positive correlations with AKT1, HSP90AA1, and TP53, with a particularly strong positive correlation between T cells CD4 memory activated and TP53, where the correlation coefficient is greater than or equal to 0.06. Additionally, positive correlations are observed between resting mast cells and AKT1, between monocytes and HSP90AA1, and between M2 macrophages and TP53 (Fig. 5C).

Fig. 5.

Immune cell infiltration and correlation analysis. (A) Stacked plot of the infiltration levels of 22 immune cell types in each liver tissue sample within the discovery cohort. (B) Statistical comparison of infiltration levels of 22 immune cell types between NAFLD and Control groups in the discovery cohort. (C) Correlation heatmap depicting the correlation between core genes and immune cell infiltration levels. * p < 0.05, *** p < 0.001.

Identification of miRNAs associated with core genes

miRNA plays an indispensable role in both the transcriptional and translational processes of genes, and identifying miRNAs associated with CPF is particularly crucial for understanding their mechanisms of action in NAFLD. A miRNA network was constructed based on 318 miRNAs associated with core genes, deducing 29 miRNAs that are most likely to be influenced by CPF (Fig. 6A, Supplementary Table 4). The GO analysis discloses substantial enrichment across various realms of BP, encompassing the regulation of nucleobase, nucleoside, nucleotide and nucleic acid metabolism, signal transduction, cell communication, transport, apoptosis, cell growth modulation, gene expression and epigenetic mechanisms regulation, cell organization and biogenesis, translation regulation, and cell cycle governance. In the MF aspect, influence is exerted through activities such as transcription factor and regulator roles, ubiquitin-specific protease function, protein serine/threonine kinase function, GTPase function, receptor signaling complex scaffold function, transmembrane receptor protein tyrosine kinase function, cytoskeletal protein binding, guanyl-nucleotide exchange factor function, and receptor binding. The CC domain predominantly involves locations within the nucleus, cytoplasm, Golgi apparatus, lysosome, early endosome, endosome, actin cytoskeleton, perinuclear region, heterogeneous nuclear ribonucleoprotein complex, and lamellipodium. These enriched findings are predominantly and intricately linked to the pathogenesis of NAFLD (Fig. 6B). Leveraging miRNAs, a series of transcription factors were anticipated, with Fig. 6C delineating the top ten, which are identified as RREB1, FOXA1, NKX6-1, ZFP161, MEF2A, NFIC, SP4, POU2F1, SP1, and EGR1.

Fig. 6.

miRNA-core gene regulation network. (A) Orange indicates core genes; green indicates association with only 1 core gene; pink indicates association with ≥ 2 core genes. (B) Enrichment analysis of pink miRNAs. (C) Top 10 predicted transcription factors for pink miRNAs.

Molecular docking confirm the key role of core genes in CPF-induced NAFLD

The lower the binding energy, the better the binding between the ligand and the receptor. Typically, when the binding energy dips below -5 kcal/mol, it suggests a favorable binding capability between them. In Fig. 7, the binding energies derived from molecular docking analyses conducted via AutoDock Vina 1.2.2 between all core genes and CPF molecules are presented, with the top ten values highlighted. The resultant binding energies, which were uniformly indicative of robust interactions, collectively furnished compelling evidence that CPF exerts a significant influence on the core genes. In Fig. 7A, while the top ten binding energies of the TP53 protein do not uniformly demonstrate exceptional binding efficacy, there remain eight sites with binding energies ranging from -5.028 kcal/mol to -5.345 kcal/mol, which still possess a relatively favorable binding capacity. This is sufficient to indicate a significant interaction between TP53 and CPF molecules. The top ten binding energies for AKT1 are all less than -5 kcal/mol, ranging from -5.079 kcal/mol to -5.519 kcal/mol (Fig. 7C). The binding energy values of HSP90AA1 and JUN proteins are particularly striking, with binding energies ranging from -5.735 kcal/mol to -6.182 kcal/mol and -6.269 kcal/mol to -7.303 kcal/mol, respectively. This relatively strong binding affinity highlights their role in NAFLD progression facilitated by CPF (Fig. 7B,D).

Fig. 7.

Molecular docking of core genes with CPF and their binding energies. The color gradient reflects hydrogen bond strength, transitioning from crimson (strong hydrogen bonds) to green (weak hydrogen bonds), while also indicating hydrophobicity, shifting from magenta (strong hydrophobicity) to blue (weak hydrophobicity).

Molecular dynamics highlights the tight association between core genes and CPF

Considering the limitations of current semi-flexible docking techniques in molecular docking, which fail to adequately account for the dynamism of protein structures, temperature, pressure, solvent effects, and other factors, this study employs molecular dynamics simulations to delve into the interactions between core genes-encoded proteins and CPF, thereby aiming to more vividly reveal the tightness and stability of their binding.

RMSD (Root Mean Square Deviation) is a commonly used indicator to measure the difference between two molecular structures. It is calculated by comparing the spatial coordinates of corresponding atoms in the two molecules, reflecting the similarity of their conformations, and the lower the RMSD value is, the closer the two structures are. As shown in Fig. 8A, the RMSD of CPF fluctuates in sync with that of the TP53 protein, indicating the RMSD fluctuations of the complex originate from the protein itself. The RMSD of both the HSP90AA1-CPF and JUN-CPF complex structures gradually stabilizes as the simulation progresses, which suggests the structures of these complexes are becoming increasingly stable (Fig. 8B,D). The RMSD of the AKT1-CPF complex is remarkably prominent, with the complex showing nearly complete overlap (Fig. 8C).

Fig. 8.

Molecular dynamics exploration of core genes. (A–D) The RMSD curves corresponding to the core genes-CPF complexes. Blue curve: RMSD values of the entire complex; red curve: RMSD of CPF; purple curve: RMSD for the protein. (E–H) The Rg curves of the core genes-CPF complexes. (I–L) Constructing FEL using RMSD and Rg. (M–P) The charts depict the RMSF curves of the core genes-CPF complexes. (Q–T) The number of hydrogen bonds in the core genes-CPF complexes. (U–X) Energy changes for all complexes without considering solvation effects. Blue curve: represents binding energy; Purple curve: represents van der Waals forces; Red curve: represents electrostatic energy. (Y-AB) Residues with substantial contributions to binding energy in each protein, where greater negative values indicate larger contributions.(AC-AF) Conformations stabilized during molecular dynamics simulations were selected to illustrate the three-dimensional structures of protein-CPF complexes, with key amino acid residues interacting with CPF and their interaction types highlighted.

Radius of Gyration (Rg) is a key parameter measuring the compactness of protein-small molecule complexes. It reflects the distribution of all atoms in the molecule relative to the center of mass. A smaller Rg value indicates a more compact structure, while a larger Rg value suggests a looser, possibly more expanded structure. Compared to the stable structures of HSP90AA1 and JUN complexes, the Rg values of TP53 and AKT1 initially exhibit significant downward fluctuations but subsequently stabilize, indicating the complex structures’ destined return to stability (Fig. 8E–H).

In the realm of molecular simulation, the Free Energy Landscape (FEL) serves as a sophisticated graphical representation that vividly delineates the distribution of free energy across diverse conformational states of a molecular system. With the RMSD and Rg of the complex serving as the coordinate axes, this representation not only clearly reveals the various possible conformations that a molecule or system may adopt but also visually reflects the relative stability among these conformations. In Fig. 8I–L, where the FEL of all complexes are examined, a low-energy state is found to exist, implying high overall structural stability of all complexes.

The Buried Solvent Accessible Surface Area (Buried SASA), a metric employed to appraise the regions within a molecule or molecular complex that are sequestered from direct solvent contact, is utilized in this context. As demonstrated in Supplementary Fig. 3A-D, where the Buried SASA values exhibit stability, it is inferred the contact surface area between CPF and the proteins encoded by core genes remains constant. In order to ascertain the initial docking site of CPF, an examination of the distances between the centroids of the residues at said sites and the centroid of CPF was conducted, alongside assessing the distance between CPF and the protein’s centroid. This analytical approach to the distances enabled us to establish that CPF forms a stable bond with the protein at the initial binding site, and that this bond strengthens as time elapses (Supplementary Fig. 3E-H). These constancies further substantiate the stability of the complexes formed between CPF and these proteins.

Root mean square fluctuation (RMSF) can measure the flexibility of amino acid residues in proteins. By calculating the motion of each atom or residue in a molecular dynamics trajectory, it accurately identifies flexible and rigid regions within the molecule. The B-Factor map, generated using RMSF values as B-factors, demonstrates the surrounding amino acids of small molecules in all core gene-CPF complexes have low flexibility, indicating stable binding between CPF and proteins (Fig. 8M–P, Supplementary Fig. 3I-L).

Hydrogen bonds are crucial for protein–ligand binding, closely related to and reflecting electrostatic interactions’ strength. The TP53-CPF complex, HSP90AA1-CPF complex, and JUN-CPF complex all have very few hydrogen bonds, fluctuating mainly between 0 and 1. Notably, the AKT1-CPF complex has the most hydrogen bonds among them, with its count typically varying between 0 and 3 (Fig. 8Q–T). Without considering solvation effects, for all complexes, the van der Waals forces and electrostatic interactions remain relatively stable (Fig. 8U–X).

Considering solvation effects, stable complex conformations are selected by analyzing RMSD, Rg, Distance, Buried SASA, and interaction energy. Then, MM-PBSA method calculates binding energy-related terms (Supplementary Table 5). In the TP53-CPF and HSP90AA1-CPF complexes, ΔE_vdw exceeds ΔE_nonpol, with both being greater than ΔE_ele. Thus, in the binding energy composition, van der Waals interactions predominate, hydrophobic interactions are secondary, and electrostatic interactions are supplementary. In the AKT1-CPF and JUN-CPF complexes, the binding energy is predominantly contributed by van der Waals interactions. The difference lies in the fact that, in the AKT1-CPF complex, electrostatic interactions play an auxiliary role, while hydrophobic interactions contribute the least. In contrast, in the JUN-CPF complex, both electrostatic and hydrophobic interactions serve as subsidiary contributors. Overall, as evidenced by ΔG_bind*, all core genes were found to stably bind with CPF. Among them, the stability of the HSP90AA1-CPF complex was the most prominent, followed by the TP53-CPF complex, the AKT1-CPF complex, while the JUN-CPF complex showed the weakest stability. ΔE_MMPBSA was decomposed to analyze the contributions of individual amino acid residues to the overall binding energy, thereby evaluating the roles of key amino acids in the protein. In the TP53 protein, amino acid residues such as LEU-100 and VAL-229 play crucial roles in small molecule binding (Fig. 8Y); in the HSP90AA1 protein, residues like PHE-138 and LEU-107 serve as key binding sites (Fig. 8Z); for the AKT1 protein, residues including TYR-131 and TYR-135 are vital for small molecule binding (Fig. 8AA); and in the JUN protein, residues such as LEU-23 and ARG-75 are essential binding amino acids (Fig. 8AB).

During the simulation of the TP53 protein, stable conformations were selected to analyze its structural features and interaction patterns. As shown in Fig. 8AC, amino acid residues such as ALA-68, LEU-23, LEU-30, and ALA-71 formed hydrophobic interactions with the CPF, while residues like ASN-64 and THR-26 engaged in van der Waals interactions with CPF. In HSP90AA1, amino acids including PHE-138, VAL-186, LEU-103, TRP-162, LEU-107, ALA-111, and VAL-136 participate in Pi-Cation, Pi-Pi Stacked, Alkyl, and Pi-Alkyl hydrophobic interactions with CPF. Additionally, LEU-48, TYR-139, and other residues interact with CPF through van der Waals forces (Fig. 8AD). Within the AKT1 protein, ARG-169 forms hydrogen bonds with CPF. Other residues, TYR-131, VAL-110, LYS-130, TYR-112, TYR-135, and ILE-137, engage in Pi-Pi T-shaped, Alkyl, and Pi-Alkyl hydrophobic interactions with CPF. Furthermore, amino acids like THR-107 and ASP-103 interact with CPF via van der Waals interactions (Fig. 8AE). In the JUN protein, amino acid residues including ALA-68, LEU-23, LEU-30, and ALA-71 form Alkyl and Pi-Alkyl hydrophobic interactions with CPF. Residues such as ASN-64 and THR-26 also engage in van der Waals interactions with CPF (Fig. 8AF).

Biomedical experiments emphasize that CPF accelerates NAFLD progression

Previous research has shown when HepG2 cells is treated with 2.5 × 10³ μmol/L CPF, it significantly impacts fatty acid biosynthesis³⁸. Exposure to CPF led to a marked decrease in cell viability of both HepG2 cells and primary hepatocyte (Fig. 9A,F). Further liver function tests have revealed CPF treatment not only promotes the deposition of TG in HepG2 cells but also exacerbated hepatocellular damage induced by AST and ALT. However, when PA and OA are added in conjunction with CPF treatment, no significant change in AST levels is observed (Fig. 9B–E). Similar findings in primary hepatocytes further confirmed CPF’s impact on NAFLD (Fig. 9G–J). As illustrated in Fig. 9K, Oil Red O staining was harnessed to visually delineate the extent of lipid droplet accumulation. Within the context of HepG2 cell studies, treatment with CPF precipitated a substantial augmentation of lipid droplet formation, an effect that was further accentuated upon the introduction of PA/OA. In stark contrast, primary hepatocytes exhibited a diminished responsiveness to CPF, manifesting only a modest increase in lipid droplet accumulation following CPF treatment alone (Fig. 9L). When CPF was co-administered with PA/OA, the Oil Red O staining area ratios were notably elevated compared to observed in the PA/OA group. Taken together, we conclude that CPF exerts a promoting effect on NAFLD progression.

Fig. 9.

CPF affects liver function and lipid droplet accumulation in hepatocytes. (A) Comparison of cell viability between Control group and CPF group in HepG2 cells, with n = 12. (B–E) Sequential variations in TC, TG, AST, and ALT levels across Control, CPF, PA/OA, and CPF + PA/OA groups, with n = 6. (F) Assessment of cell viability between Control group and CPF group in primary hepatocytes from C57BL/6 mice, with n = 12. (G–J) Comparison of TC, TG, AST, and ALT levels across four groups: Control, CPF, PA/OA, and CPF + PA/OA, with n = 6. (K) Oil Red O staining and Oil Red area ratio in HepG2 cells treated with Control, CPF, PA/OA, and CPF + PA/OA groups, with n = 9, (Scale bar: 50 μm). (L) Oil Red O staining and the corresponding area ratio in primary hepatocytes across four groups: Control, CPF, PA/OA, and CPF + PA/OA, with n = 9, (Scale bar: 50 μm).* p < 0.05, ** p < 0.01, **** p < 0.0001;# indicates comparison with the CPF group, ## p < 0.01, ### p < 0.001, #### p < 0.0001;† indicates comparison with the PA/OA group, † p < 0.05, †† p < 0.01, †††† p < 0.0001.

CPF promotes NAFLD via stabilizing HSP90AA1 and inducing JUN/TP53 phosphorylation

Upon comparing the RT-qPCR results of livers from C57BL/6 mice fed CD versus HFD, it emerged the expression of TP53, AKT1, and JUN was significantly upregulated following HFD feeding (Fig. 10A,C,D). While HSP90AA1 exhibited an upward trend, it did not attain statistical significance (Fig. 10B). In the livers of NAFLD patients, the mRNA expression levels of TP53, HSP90AA1, AKT1, and JUN were up-regulated (Fig. 10E–H). Most unexpectedly, in cell cultures mediated by CPF, the RT-qPCR results of the four core genes showed no significant changes in either HepG2 cells or primary hepatocytes (Supplementary Fig. 4). Molecular dynamics simulations have demonstrated that CPF can bind to the domains of four proteins. To further explore the effects of CPF on the stability of these proteins, we employed CHX in our analysis. WB results revealed that CPF significantly enhances the stability of the HSP90AA1 protein (Fig. 10J). In contrast, CPF did not exert a notable effect on the stability of TP53, AKT1, or JUN proteins (Fig. 10I, K, L). The phosphorylated forms of TP53 (P-TP53), AKT1 (P-AKT1), and JUN (P-JUN) represent the primary active states through which these proteins fulfill their biological functions. By assessing the levels of protein phosphorylation, we observed that CPF significantly up-regulates the expression of P-JUN (S243) and P-TP53 (Ser315) (Fig. 10N,O). However, the expression level of P-AKT1 (Ser473) remained unchanged (Fig. 10M). Therefore, we propose that CPF exacerbates NAFLD primarily by enhancing the stability of HSP90AA1 protein and upregulating the protein levels of P-JUN and P-TP53.

Fig. 10.

CPF exacerbates NAFLD by stabilizing HSP90AA1 protein and promoting the phosphorylation of JUN and TP53. (A–D) Hepatic TP53, HSP90AA1, AKT1, and JUN mRNA expression levels in CD and HFD groups , with n = 8. (E–H) TP53, HSP90AA1, AKT1, and JUN mRNA expression levels in NAFLD patients and control subjects, with n = 8. (I–L) Effect of CPF on the protein stability of AKT1, HSP90AA1, JUN, and TP53 in HepG2 cells detected by CHX, with n = 3. (M) Impact of CPF on AKT1 phosphorylation (P-AKT1) at Ser473 in HepG2 cells, with n = 3. (N) Regulation of JUN phosphorylation (P-JUN) at S243 by CPF in HepG2 cells, with n = 3. (O) Effect of CPF on TP53 phosphorylation (P-TP53) at Ser315 in HepG2 cells, with n = 3. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Discussion

Over the past decades, the extensive application of organophosphorus pesticides, while propelling the vigorous growth of crops, has concurrently imposed a significant burden on the environment³⁹. The exposure to insecticides is not confined merely to the environs of production sites; rather, it is extensively disseminated due to their persistence in water, with CPF serving as a quintessential exemplar of such enduring contaminants^40–43. Given its pivotal role in metabolic detoxification and the elimination of pesticides within the body, the liver naturally emerges as the primary target organ susceptible to the influence of these chemical substances. With increasing exposure levels, hepatic CPF accumulation may progressively induce a cascade of pathological effects, including hepatocellular damage, hepatic inflammation, and NAFLD development^3,4,11,12,15. The potential toxicological targets and mechanisms of CPF-induced NAFLD have not yet been reported. This study aimed to employ network toxicology to explore these mechanisms, with a particular focus on the impact of CPF on signaling pathways and cellular functions, thereby addressing the current gaps in the research landscape.

There exist 582 overlapping genes between those associated with CPF retrieved from CTD and those related to NAFLD collated from the GeneCards database. The biological processes previously identified, such as fatty acid metabolic process, regulation of lipid metabolic process, signaling receptor activator activity, and receptor ligand activity, have been confirmed to play a critical role in the pathogenesis of NAFLD, which is consistent with our GO enrichment results⁴⁴. In the KEGG signaling pathway enrichment analysis, insulin resistance, a hallmark of the pathophysiology of NAFLD, was significantly enriched⁴⁵. In NAFLD, the disruption of lipid metabolism is manifested as an increase in hepatic fat accumulation, primarily due to an imbalance in fatty acid uptake, synthesis, oxidation, and secretion. Concurrently, the inflammatory response is exacerbated, with elevated levels of pro-inflammatory cytokines such as TNF-α, IL-1β, and IL-6, leading to hepatocellular injury and fibrosis. CPF has been shown to significantly activate key signaling pathways such as the PI3K-Akt pathway, AMPK pathway, and TNF pathway. These pathways are closely related to lipid metabolism and inflammatory responses, which are critical in NAFLD progression^15,46,47. Based on enrichment analysis and existing research, we speculate that CPF may play a crucial role in NAFLD progression by modulating insulin resistance, lipid metabolism, and inflammatory responses.

To further delineate the mechanisms by which CPF exerts its effects in NAFLD, we employed the STRING platform and Cytoscape software to construct a PPI network. Following the application of seven distinct algorithms for comprehensive analysis, we conclusively identified TP53, HSP90AA1, AKT1, and JUN as the core genes implicated in CPF-induced NAFLD.

Based on the core genes identified, we formulated clinical nomograms to predict the probability of CPF-mediated NAFLD occurrence. ROC curves manifested robust diagnostic performance in both the discovery cohort and the validation cohort. The calibration curves, validated by the Hosmer–Lemeshow test, demonstrated no significant discrepancies between the model-predicted probabilities and the actual observed probabilities. Moreover, DCAs ascertained the core genes possess substantial clinical utility in the context of NAFLD. The GSEA analysis underscores the consistent presence of several key pathways, including TCA cycle, Hepatocellular carcinoma, Histidine metabolism, and Nicotine addiction. Amino acids have been identified by prior research as the principal carbon source for hepatic fat synthesis⁴⁸. The formation of NAFLD further propels the progression of the TCA cycle, thereby creating a vicious cycle akin to a self-perpetuating spiral of TCA cycle-NAFLD-TCA cycle, which accelerates the rapid advancement of the disease⁴⁹. Moreover, histidine, produced under the action of hypoxia-inducible factor 2α, also promotes NAFLD progression⁵⁰. Simultaneously, histidine metabolism in the gut can serve as a biomarker for NAFLD⁵¹. Nicotine is also involved in the pathogenesis of NAFLD, where it can mediate the suppression of hepatic triglyceride accumulation, thereby influencing disease progression, as evidenced in rat models⁵². The GSEA results provide corroborating evidence that CPF exacerbates the pathological progression of NAFLD.

Considering multiple pathways are associated with inflammation, we explored the immune landscape of the discovery cohort and found an increased infiltration of naive B cells, M0 macrophage subtype, and resting mast cells, while the infiltration of the M2 macrophage subtype was significantly reduced. Correlation analysis revealed plasma cells and CD4⁺ memory-activated T cells correlated with core genes such as AKT1, HSP90AA1, and TP53, suggesting in the CPF-mediated NAFLD process, core genes may exert their effects through immune cells.

Previous studies have established that miR-15a-5p, miR-16-5p, miR-139-5p, miR-19a-3p, miR-26b-5p, miR-375, miR-30a-5p, miR-125b-5p, miR-155-5p, miR-25-3p, miR-125a-5p, miR-101-3p, miR-17-5p, miR-34a-5p, miR-429, and miR-223-3p modulate NAFLD progression, and these miRNAs also emerged as the most prominent miRNAs in our own screening^53–67. We collected miRNAs related to core genes and performed enrichment analysis. On this basis, we further predicted the target transcription factors of these core gene-related miRNAs. The prediction results showed that EGR1, SP1 and FOXA1 were their major target regulatory molecules. Meanwhile, existing studies have confirmed that EGR1, SP1 and FOXA1 are all key regulatory factors involved in the pathogenesis of NAFLD^68–70. The regulatory network formed by these miRNAs and transcription factors elucidates the molecular mechanism underlying CPF-induced NAFLD from multiple dimensions.

In vitro experimental data demonstrate that CPF significantly induces hepatotoxicity, as evidenced by elevated AST/ALT levels and triglyceride accumulation, with Oil Red O staining further confirming its NAFLD-inducing effects. Notably, while core gene mRNAs show upregulation trends in both NAFLD patients and animal models, CPF exposure does not significantly alter their transcriptional levels. These findings suggest that CPF primarily promotes NAFLD progression through direct interactions with core gene-encoded proteins rather than through transcriptional regulation of these genes.

TP53, also known as P53, which is regarded as the “tumor suppressor”, participates in many signaling pathways that lead to apoptosis and cell cycle arrest⁷¹. In the realm of non-cancerous TP53-related diseases, a multitude of relevant research efforts have been initiated. As an illustration, TP53 plays a pivotal role in NAFLD progression^72,73. TP53 exerts its influence by inhibiting the activity of pyruvate dehydrogenase kinase 2, thereby facilitating the conversion of pyruvate into acetyl-CoA, which subsequently participates in TCA cycle⁷⁴. In addition, TP53 is capable of inducing the expression of Sirtuin1, an NAD-dependent deacetylase that modulates the activity of malonyl-CoA, thereby playing a role in fatty acid metabolism⁷⁵.

This study discovered that CPF can form a stable bond with the TP53 protein without notably impacting the protein’s stability. Since phosphorylated TP53 (P-TP53) plays a crucial role in the onset and progression of NAFLD, we proceeded to assess the protein expression of P-TP53⁷⁶. Our findings revealed that CPF markedly increased the expression of P-TP53. Consequently, we infer that CPF primarily affects TP53 by modulating its phosphorylation status.

HSP90AA1 is implicated in a myriad of pivotal biological processes, which encompass the facilitation of EMT, the regulation of cellular signal transduction, DNA repair, angiogenesis, immune response, and cell migration, among others⁷⁷. Notably, HSP90AA1 protein stabilization promotes HCC, while inhibiting its degradation also contributes to NAFLD development^78,79. consistent with the CPF-mediated effects observed in our study. Serum HSP90AA1 levels typically correlate with NAFLD severity and occur concomitantly with increased hepatic HSP90AA1 mRNA levels⁸⁰. However, our study in HFD-induced mouse models revealed no significant increase in hepatic HSP90AA1 mRNA, presenting a discordant outcome. This discrepancy may stem from the absence of severe inflammation in the HFD group, given that HSP90AA1 expression is regulated by pro-inflammatory cytokines⁸¹. Furthermore, correlation analysis indicated an association between HSP90AA1 and immune cells, such as monocytes, reinforcing the notion inflammation is a crucial factor in HSP90AA1-driven NAFLD progression. Finally, in this study, CHX intervention experiments confirmed that CPF can promote NAFLD progression by maintaining the structural stability of the HSP90AA1 protein.

Upon stimulation by insulin, growth factors, energy sources, and cytokines, the phosphoinositide 3-kinase (PI3K)/AKT1 signaling pathway is activated, subsequently modulating glucose and lipid metabolism as well as protein synthesis⁸². AKT1 activation is predominantly achieved through phosphorylation. Recent studies confirm AKT1 activation can trigger NAFLD and further facilitate the progression to non-alcoholic steatohepatitis (NASH) and even HCC⁸³. Treatment with AKT1 inhibitors can markedly reduce the pathological burden of NAFLD⁸⁴. Regrettably, the present study did not detect any alterations in the protein stability or phosphorylation level of AKT1 induced by CPF. We hypothesize that CPF may regulate AKT1 via other protein post-translational modifications, such as palmitoylation, and this hypothesis warrants further experimental validation⁸⁵.

We have observed JUN is notably enriched within the Apelin signaling pathway. Meanwhile, serum Apelin can function as a biomarker for NAFLD⁸⁶. In the livers of HFD-fed mice, JUN phosphorylation and nuclear translocation are elevated, which is closely associated with its activation state in hepatic pathological alterations⁸⁷. JUN activation additionally influences AMPK complex assembly and activation, consequently impacting hepatic metabolic processes⁸⁸. Our results demonstrated that CPF treatment significantly increased the protein level of P-JUN, which clearly confirmed that promoting JUN phosphorylation is the core mechanism underlying the biological effects of CPF. In the progression of NAFLD mediated by JUN, miRNAs also exert significant influence. Among them, miR-139-5p, associated with three core genes, targets the JUN/SREBP-1c pathway to reduce lipid accumulation in NAFLD⁵⁵.

In summary, our integrated network toxicology approach has systematically elucidated that CPF induces NAFLD progression through direct protein-level interactions with four core proteins, bypassing transcriptional modulation as evidenced by: (i) molecular dynamics simulations demonstrating stable binding conformations of CPF-protein complexes, (ii) pathway enrichment in TCA cycle dysregulation and histidine metabolism, (iii) Immune cell-mediated pathogenic cascades were observed, and (iv) in vitro experiments confirmed that CPF exacerbates hepatic lipid accumulation without inducing significant alterations in the mRNA expression levels of core genes. Conversely, CPF stabilizes the protein structure of HSP90AA1 and promotes the phosphorylation of TP53 and JUN. The established clinical nomograms exhibited robust predictive value (AUC > 0.76 across cohorts), positioning these targets as promising candidates for therapeutic development. However, critical limitations warrant attention: firstly, to comprehensively assess the relationship between CPF and NAFLD, there is an urgent need to conduct large-scale sampling of individuals from various populations worldwide in order to identify the “threshold” between CPF concentration in the circulatory system and the onset of NAFLD. This will provide crucial clues for our subsequent research. Moreover, there are no targeted therapeutic regimens for CPF-induced NAFLD, and the efficacy of conventional medications remains uncertain. Finally, this study only confirms that CPF can bind to the AKT1 protein, but the specific role of AKT1 in this process remains unclear, and further in-depth studies are still required.

In this study, employing network toxicology as our “compass” and molecular docking technology as our "probe," we embarked on an in-depth exploration of the intricate relationship between CPF and NAFLD. We not only successfully identified key targets CPF may act upon during the pathogenesis of NAFLD but also preliminarily unveiled its underlying mechanisms, thereby offering a novel perspective on how this environmental toxin affects liver health. More importantly, by ingeniously integrating network toxicology with molecular docking technology, our study has established a new approach in toxicological research. This integration, akin to bestowing a pair of “discerning eyes” upon toxicity studies, not only enables the rapid and efficient screening of potential toxicological targets but also delves into the complex mechanisms behind them, significantly enhancing the depth and accuracy of research.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

In this study, Xiaoning Chen, Na Li, Ying Zhang, and Shibo Sun were responsible for conceptualization, project administration, and supervision. Na Li and Xiaoning Chen developed the methodology. Yapeng Li was primarily responsible for manuscript writing, network pharmacology analysis, initial data analysis, software application, and also participated in formal analysis and investigation; Zhengwei Zhang and Guixin Zhang focused on the design and implementation of biological experiments and contributed to manuscript revisions; Hongji Li conducted data verification for all experiments, participated in formal analysis, investigation, and manuscript review and editing. Junting Ren, Manyu Gong and Haodong Li performed comprehensive checks on visualization. All authors actively participated in research discussions and reviewed and approved the final manuscript.

Funding

The author received no specific funding for this work.

Data availability

The original contributions to this study are included in the article and Supplementary Material. For further inquiries, please contact the corresponding authors.

Declarations

Competing interests

The authors declare no competing interests.

Ethics approval

The research complied with the ethical principles outlined in the Declaration of Helsinki. The animal studies were reviewed and approved by the Ethics Committee of the School of Pharmacy at Harbin Medical University (Approval No: IRB3025725). All animal experiments in this study adhered to the guidelines of the Animal Research: Reporting of In Vivo Experiments (ARRIVE) guidelines. All human liver tissues were derived from surgical discarded tissues and met the relevant requirements of the Medical Ethics Committee of the Second Affiliated Hospital of Harbin Medical University (Approval No: KY2024-235). Patient information was uploaded to Supplementary Table 6. We confirm that the protocols for human tissue-related experiments have been approved by the Medical Ethics Committee of the Second Affiliated Hospital of Harbin Medical University.Prior informed consent has been obtained from all participants for their involvement in the study.