Deciphering the Effect of Methyl 4‑Hydroxybenzoate on Breast Cancer by Bioinformatics and Experiments

Abstract

Methyl 4-hydroxybenzoate (MEP), a widespread environmental contaminant, is suspected to increase breast cancer (BRCA) risk; however, its molecular mechanism remains unclear. Using public health databases, we performed differential expression analysis, network toxicology modeling, and protein–protein interaction mapping to identify candidate biomarkers. Functional interrogation using gene set enrichment analysis, immune infiltration profiling, and molecular docking elucidated the mechanistic roles of these genes in BRCA. Single-cell transcriptomic analysis highlighted the cell populations most relevant to tumor initiation and tracked the biomarker expression patterns within these populations. Subsequently, in vitro assays revealed the direct effects of MEP on malignant behavior. CCNE1, CDK1, E2F1, and EZH2 emerged as core drivers converging on multiple oncogenic pathways. Fifteen immune cell subsets showed markedly altered infiltration patterns in tumors, with macrophages and related populations associated with disease progression. Among the identified biomarkers, CCNE1 exhibited the strongest predicted affinity for MEP (−6.2 kcal mol^–1), highlighting the CCNE1/CDK1 axis as a potential therapeutic target. Fibroblasts were identified as the key cellular context in which these biomarkers are transcriptionally primed during carcinogenesis. Functionally, MEP exposure enhanced malignant behaviors in MCF-7 cells and increased CCNE1, CDK1, E2F1, and EZH2 expression. Collectively, this integrated analysis indicates that MEP may promote BRCA development by acting on four central regulators within fibroblasts, providing mechanistic insights into the environmental origins of BRCA.

1. Introduction

BRCA is one of the most common malignancies, accounting for 23.8% of cancers in female patients. , Characterized by substantial heterogeneity, BRCA exhibits diverse molecular and pathological features, complicating disease management. , Despite advances in early detection and treatment strategies, BRCA exhibits a complex etiology that necessitates a deeper understanding of its risk factors and molecular mechanisms.

MEP, also known as p-oxybenzoic acid methyl ester, is widely used as an antimicrobial agent in cosmetics and personal care products as well as a food preservative. Previous studies have demonstrated that MEP residues are frequently detected in human tissues and environmental components such as soil and water. , Although MEP was considered to have low toxicity in the early assessments, subsequent studies have reported that MEP can bind to estrogen receptors (ERs) and mimic hormonal activity, thereby disrupting the endocrine system and increasing the risk of BRCA. , Further experimental studies corroborated these concerns, showing that BRCA exhibits enhanced growth and metastatic capacity at both cellular and animal levels following chronic MEP exposure. − Thus, further investigation is warranted to assess the health risks and safety implications of MEP in BRCA.

Single-cell RNA sequencing (scRNA-seq) enables the investigation of cellular heterogeneity and dynamic gene expression changes at single-cell resolution, thereby facilitating the characterization of cell types, identification of cellular subpopulations, and elucidation of cellular state transitions. These approaches have been widely used in BRCA to characterize the tumor microenvironment (TME) and track tumor cell evolution. − Network toxicology is an interdisciplinary approach that combines bioinformatics, big data analytics, genomics, and related scientific fields. Combined with molecular docking techniques, the present study uses the “compound–gene–target” framework to systematically elucidate the mechanisms of environmental toxicant–induced pathogenesis and identify potential biomarkers. −

In this study, transcriptomic profiling, scRNA-seq, and network toxicology analyses were integrated to identify MEP-related biomarkers and key cellular drivers in BRCA. Subsequent bioinformatics analyses and experimental validation supported these mechanistic roles and provided new insights into therapeutic development for BRCA.

2. Method

2.1. Data Collection

Relevant data were retrieved from public databases. The transcriptome data set of BRCA (TCGA-BRCA; access date: February 27, 2025) was obtained from the UCSC database (https://xena.ucsc.edu/) as the training set, comprising 1113 BRCA and 113 control tissue samples. The GSE42568 data set from the GPL570 platform (access date: February 27, 2025) was retrieved from the Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/geo/) as a validation set, including 104 BRCA and 17 control samples. The GSE243526 data set (GPL18573), containing 12 BRCA and 4 control samples, was also obtained from the GEO database as a scRNA-seq data set.

2.2. Differential Expression Analysis

In TCGA-BRCA, the “DEseq2” package (v1.40.2) was used to identify differentially expressed genes (DEGs) between BRCA and control samples (|log₂ fold change [FC]| > 2 and adjusted (adj.) [P < 0.05]. After sorting log₂FC values from highest to lowest, all DEGs were presented and the top 10 upregulated and downregulated DEGs were highlighted in a volcano plot generated using the “ggplot2” package (v3.5.1). Expression patterns of these top DEGs were further visualized in a heatmap constructed using the “ComplexHeatmap” package (v2.18.0).

2.3. Network Toxicological Analysis

To evaluate MEP toxicity, its three-dimensional (3D) stereostructure and SMILES notation was obtained by searching for “methyl-4-hydroxybenzoate” in the PubChem database (https://pubchem.ncbi.nlm.nih.gov/). Toxicity properties and basic toxicity reports were evaluated and generated using the ProTox-3.0 database (https://tox.charite.de/protox3) and the SwissADME database (http://www.swissadme.ch/index.php).

The SwissTargetPrediction database integrates the structural and pharmacological data of ligand-target binding, covering over 2000 protein targets in humans. The prediction algorithm has been verified by a large number of clinical drugs, and its prediction accuracy for small molecule compound targets is at the leading level among similar tools. So, using the SMILES notation, potential toxic targets of MEP were identified from the ChEMBL database (https://www.ebi.ac.uk/chembl/) and SwissTargetPrediction database (http://www.swisstargetprediction.ch/) with probability >0. The target species was set to Homo sapiens to ensure relevance to human biology. The targets obtained from the ChEMBL database were converted to gene symbols using the UniProt database (https://www.uniprot.org/). After merging and deduplicating targets from the two databases, potential toxic targets of MEP (MEP targets) were identified. The targets predicted simultaneously by both databases were then used to construct the database–target network using the Cytoscape software (v3.9.1).

In addition, to identify targets associated with BRCA, the OMIM (https://omim.org/) and GeneCards (https://www.genecards.org/) databases were queried using the keyword “breast cancer.” Target genes retrieved from both databases were merged and deduplicated to generate the final set of BRCA-associated targets (BRCA Targets). Targets concurrently predicted by both databases were further used to construct a database–target network using Cytoscape software (v3.9.1).

2.4. Identification and Functions of Candidate Genes

After intersecting DEGs, MEP Targets, and BRCA Targets using the “ggvenn” package (v0.1.10), the overlapping genes were defined as candidate genes. Subsequently, gene symbols were converted to Entrez IDs using the “org.Hs.eg.db” package (v3.18.0). Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses were then performed using the “clusterProfiler” package (v4.10.1). For GO analysis, the “enrichGO” function was used to assess functional enrichment of candidate genes (adj. P < 0.05), and the top five terms in each category were visualized using the “GOplot” package (v1.0.2) based on adj. P values. Similarly, KEGG analysis pathway enrichment was conducted using the “enrichKEGG” function (adj. P < 0.05), with the top 10 pathways ranked in descending order according to gene count.

2.5. Protein–Protein Interaction (PPI) Network and Expression Level Verification

First, to acquire hub genes and explore the interaction of candidate genes at the protein level, candidate genes with confidence scores >0.7 were visualized in a protein–protein interaction (PPI) network using the STRING database (http://string-db.org) and Cytoscape. Dense connection regions were identified using the MCODE plugin (degree cutoff = 2, node score cutoff = 0.2, k-core = 2, maximum depth = 100), and the top two modules were defined as remarkable modules. Genes in these modules were termed module genes. The Radiality, Closeness, and Degree algorithms in Cytoscape were used to select the top 15 key genes, which were intersected with module genes via the “ggvenn” package (v0.1.10) to define hub genes.

Expression differences of hub genes between BRCA and control samples in TCGA-BRCA and GSE42568 were assessed using the Wilcoxon test (P < 0.05). Genes showing marked intergroup differences were labeled as biomarkers and visualized using “ggplot2.”

2.6. Gene Set Enrichment Analysis (GSEA)

To explore biomarker mechanisms, the “c2” (C2: CP: KEGG) data set from the “msigdbr” package (v7.5.1) was used as the reference gene set based on the MSigDB database (https://www.gsea-msigdb.org/gsea/msigdb). Spearman correlation analysis between each biomarker and all other genes was performed in TCGA-BRCA using the “psych” package (v2.4.3). Genes were ranked by correlation coefficients, and gene set enrichment analysis (GSEA) was conducted using the “clusterProfiler” package to identify pathways associated with biomarkers (|normalized enrichment score [NES]| > 1, adj. P < 0.25, and P < 0.05).

2.7. Immune Infiltration Analysis

In TCGA-BRCA, CIBERSORT was used to estimate infiltration of 22 immune cell types. Cells with zero proportion were excluded, and Wilcoxon test was used to compare immune cell infiltration between BRCA and control samples. Immune cells showing significant differences (P < 0.05) were considered as differential immune cells. Correlations among differential immune cells were evaluated using the “psych” package, and correlations between biomarkers and differential immune cells (|correlation [cor]| > 0.3 and P < 0.05) were visualized using “corrplot” (v0.92) and “ggplot2” (v3.5.1) packages. The main purpose of this analysis was to explore the differential immune cells associated with biomarkers in BRCA.

2.8. Molecular Docking and Molecular Dynamics (MD) Simulation

Molecular docking was performed to determine binding of MEP to biomarkers. First, the 3D structures of MEP and biomarkers were obtained from PubChem (https://pubchem.ncbi.nlm.nih.gov/) and RCSB (http://www.pdb.org/), respectively. CB-Dock2 (https://cadd.labshare.cn/cb-dock2/php/index.php) automatically identify potential binding cavities and predict docking positions of MEP within these pockets in the target proteins, and docking simulations were visualized using PyMOL (v3.0.3) (binding energy < −5 kcal/mol). Interaction profiling was conducted using PLIP to determine atomic distances (Å) and donor angles, while Discovery Studio 4.5 was used to categorize contacts into conventional hydrogen bonds, π–π stacking, and alkyl/π-alkyl interactions.

To validate docking rationality and reliability, the protein sequences of biomarkers were retrieved from the National Center for Biotechnology Information (https://www.ncbi.nlm.nih.gov/) for human species, and 100 ns molecular dynamics (MD) simulations were performed using GROMACS (2024.2). The protein was parametrized with the Amber99sb-ildn force field, while the MEP ligand adopted the GAFF2 force field combined with RESP charges. The solvent was modeled using the TIP3P water model. The system was placed in a dodecahedral box with a solvent layer thickness of 1.0 nm, and 0.15 M Na⁺/Cl^– was added to neutralize the system charge. Energy minimization was performed in two consecutive steps: first, 5000 steps of steepest descent method with protein constraints, followed by 20,000 steps of conjugate gradient method for full-system refinement. Subsequently, 100 ps NVT equilibration (with V-rescale thermostat at 300 K) and 200 ps NPT equilibration (with Parrinello–Rahman barostat at 1 bar) were carried out sequentially. Root mean square deviation (RMSD) revealed positional changes in the protein during the simulation, and root-mean-square fluctuation (RMSF) was used to characterize residue flexibility and mobility. Additionally, the radius of gyration, reflecting protein structure compactness, was determined.

2.9. Quality Control Analysis and Cell Communication

To ensure accuracy and reliability, all analyses were conducted using the “Seurat” package (v5.1.0).

After batch correction and integration via the “cca” function, quality control was performed on the sample data. Cells were retained if they met the following criteria: (1) total gene expression was ≤30,000 (nCount_RNA); (2) detected gene count between 200 and 1500 (nFeature_RNA); and (3) mitochondrial proportion ≤10% (percentage). Results were visualized using a violin plot. Filtered data were normalized using the “NormalizeData” function. The 2000 highly variable genes (HVGs) were selected, with the top 10 HVGs labeled using the “FindVariableFeatures” function and displayed using the “LabelPoints” function. To further reduce data dimensionality, the “ScaleData” function was used to scale the data set. Principal components (PCs) for clustering were computed using the “RunPCA” function based on the top 2000 HVGs. Statistically significant PCs were identified using the “JackStrawPlot” function (P < 0.05), and principal component analysis results were visualized using the “Elbowplot” function. Unsupervised clustering was performed through “FindNeighbors” and “FindClusters” to determine cell cluster counts (resolution = 0.2, dimension = 20), and clusters were visualized via uniform manifold approximation and projection using the “RunUMAP” function. Cell clusters were annotated according to established methods, , and marker gene expression was presented using a bubble plot. Finally, cell–cell interactions and ligand–receptor pairs were analyzed using the “CellChat” package (v1.6.1) in the GSE243526 data set.

2.10. Identification and Metabolic Activity of Key Cells and Pseudotime Analysis

Key cells in BRCA were identified by comparing cell abundances between BRCA and control samples in GSE243526 using the chi-squared test (P < 0.05). Cells showing significant proportional differences were defined as differential cells. Expression differences of biomarkers in these cells were assessed using the Wilcoxon test (P < 0.05) and presented using “ggplot2.” Cells with significantly altered biomarker expression levels between the two groups were defined as key cells.

The VISION algorithm in the “scMetabolism” package (v0.2.1) was used to evaluate the metabolic activities of key cells, with results visualized using the “DotPlot.metabolism” function. To explore developmental trajectories, key cells were dimensionally reduced, clustered, and annotated using marker genes from previous studies − (resolution = 0.1), and marker gene expression was displayed via bubble plots. The differentiation trajectories of key cells were analyzed using the “monocle” package (v2.30.1). Changes in cell quantities during differentiation were assessed using the “ggpubr” package (v0.6.0), and biomarker expression was plotted using the “plot_pseudotime_heatmap” function in “monocle”.”

2.11. Statistical Analysis

All analyses were performed in R (v4.3.1). The Wilcoxon test was used to compare differences between groups (P < 0.05). Data are expressed as means ± standard deviations (n = 3). Significant differences were determined using the t-test, one-way ANOVA, or two-way ANOVA using GraphPad Prism (v8.0). Statistical significance was set at P < 0.05.

2.12. Cell Culture

The human BRCA cell line MCF-7 was obtained from ATCC and cultured in minimum essential medium (Gibco, Carlsbad, CA, USA) in an incubator at 37 °C and 5% CO₂.

2.13. Quantitative Real-Time PCR (qPCR)

The samples were treated with DMEM, 2.5 and 5 μg/L MEP for 24 h, and then the total RNA was extracted. Total RNA was extracted from the cells using TRIzol reagent (Vazyme, Nanjing, China), and cDNA was synthesized using a reverse transcription kit (Vazyme, Nanjing, China). The qPCR reaction mixture (20 μL) contained 10 μL SYBR Green Master Mix (Vazyme, Nanjing, China), 0.8 μM each of forward and reverse primers, and 2 μL cDNA template. Each sample was analyzed in triplicate, along with a no-template control. Amplification was carried out on a real-time PCR system under the following conditions: initial denaturation at 95 °C for 30 s, followed by 40 cycles at 95 °C for 5 s and 60 °C for 30 s (fluorescence acquisition). A melting curve analysis (60 to 95 °C) was performed to verify the amplification specificity. Relative gene expression was calculated using the ΔΔCt method, with GAPDH as the internal reference gene. Primer sequences for the biomarkers are shown in Table .

1. Primer Sequence of Biomarkers.

gene	forward (5′-3′)	reverse (5′-3′)
CCNE1	CTGGATGTTGACTGCCTTGAAT	TTCTCTATGTCGCACCACTGA
CDK1	AGGTCAAGTGGTAGCCATGAA	GCATAAGCACATCCTGAAGACT
E2F1	CGTGGACTCTTCGGAGAACTT	GAGATGATGGTGGTGGTGACA
EZH2	CCGCTGAGGATGTGGATACT	AAGGCTGCCGTGGATGAT
GAPDH	AGATCCCTCCAAAATCAAGTGG	GGCAGAGATGATGACCCTTTT

2.14. Cell Cycle

MCF-7 cells were seeded in 6-well plates at an appropriate density and treated with DMEM, 2.5 μg/L MEP, and 5 μg/L MEP for 24 h, respectively. Cell cycle staining kits (MULTI SCIENCES, Hangzhou, China) were used to examine the effects of MEP on BRCA cell cycle progression. The cells were processed using flow cytometry, and the data were analyzed using NovoExpress (v 1.6.2.0).

2.15. Wound Healing Assay

A wound healing assay was performed to evaluate the migratory ability of MCF-7 cells. Cells were seeded in 6-well plates at a density of 5 × 10⁵ cells/well and treated with MEP for 48 h. When the cells reached confluence, a uniform scratch was made in the monolayer using a 200 μL pipet tip. After washing to remove debris, MEP of 2.5 and 5 μg/L was added to the experimental group, and the cells were maintained in 1% FBS-supplemented medium for 48 h to monitor their migration. Wound images were captured at 0 (postscratch), 24, and 48 h using an EVOS M7000 imaging system (10× magnification), and gap closure was quantified using ImageJ.

2.16. Transwell Invasion Assay

Cell invasion was assessed using Matrigel-coated Transwell chambers (24-well plates; Corning Inc., NY, USA). MCF-7 cells (3 × 10⁴ cells in 100 μL of serum-free medium) were pretreated with 2.5 and 5 μg/L MEP for 48 h, seeded into Matrigel-coated upper chambers (Biozellen, NE, USA), and allowed to invade toward the lower chambers containing 800 μL of 10% FBS-supplemented medium for 72 h at 37 °C. Noninvading cells were removed from the upper chamber using a cotton swab, and invading cells were fixed with 4% paraformaldehyde for 30 min, stained with 0.1% crystal violet for 30 min, and imaged at 10× magnification using an EVOS M7000 microscope (Invitrogen). Five randomly selected fields per well were analyzed, with cell counts quantified using ImageJ.

2.17. Cytoskeleton Staining

MCF-7 cells were pretreated with 2.5 and 5 μg/L MEP for 24 h, respectively, before the next operation. Following fixation with 4% formaldehyde and permeabilization with 0.1% Triton X-100, the cytoskeletal structure was stained with Actin-Tracker Red-Rhodamine (Beyotome, C2207S; 1:200 dilution) for 30 min at room temperature in the dark. Samples were washed three times (5 min per wash) in phosphate-buffered saline supplemented with 0.1% Triton X-100. Nuclei were counterstained with DAPI, and fluorescence images were captured using a laser-scanning confocal microscope.

2.18. Western Blotting

MCF-7 cells were pretreated with 2.5 and 5 μg/L MEP for 24 h and then lysed on ice. Total cellular proteins were extracted and quantified using a bicinchoninic acid (BCA) assay. Protein samples (20 μg per lane) were separated via sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and electrotransferred onto poly­(vinylidene fluoride) (PVDF) membranes (Millipore, Bedford, MA, USA). After blocking with 5% skim milk, the membranes were probed with primary antibodies (1:1000–1:2500 dilution) overnight at 4 °C, followed by incubation with HRP-conjugated secondary antibodies (1:5000 dilution) for 1 h at room temperature. Protein bands were visualized using the Image Lab system. Primary antibodies against Cyclin E1 (HUABIO, ET1612–16; 1:2000), EZH2 (HUABIO, HA722095; 1:2000), E2F1 (HUABIO, ET1701–73; 1:2000), CDK1 (Affinity, DF6024; 1:2000), E-cadherin (Abcam, ab314063; 1:1000), vimentin (Abcam, ab8978; 1:2000), MMP2 (Abcam, ab92536; 1:2000), MMP9 (HUABIO, ET1704–69; 1:4000), and GADPH (HUABIO, HA721136; 1:20,000) were used. Secondary antibodies included HRP-conjugated goat antirabbit IgG polyclonal antibody (HUABIO, HA1001; 1:50,000), and HRP-conjugated goat antimouse IgG polyclonal antibody (HUABIO, HA1006; 1:50,000).

3. Results

3.1. Identification and Functions of Candidate Genes

Compared with the control group, 3258 DEGs were identified in the BRCA group, including 2251 upregulated and 1007 downregulated DEGs. Based on log₂FC values, Figure A shows a volcano plot of all DEGs and the top 10 upregulated and downregulated DEGs. Figure B shows a volcano plot illustrating the expression patterns of the top 10 upregulated and downregulated DEGs between the BRCA and control groups.

1.

Identification and functions of candidate genes. (A, B) The DEGs analyzed by a volcano plot and a heatmap plot between BRCA and control. (C) Oral toxicity prediction results of MEP. (D, E) PPI network of potential MEP toxicity targets from ChEMBL and Swiss Target Prediction database (D), and GeneCards and OMIM database (E). Node size reflects degree centrality. (F) Venn diagram of intersecting DEGs, MEP-related targets, and BRCA-related targets. (G, H) GO enrichment analysis (G) and KEGG analysis indicated the candidate genes.

Database–based information related to MEP was obtained, including a median lethal dose of 1190 mg/kg and a toxicity class of 4. The MEP chemical structure was established and related structural information was acquired (Figure C). Notably, 1067 potential toxic targets of MEP were obtained after merging and removing duplicates from the two databases. Targets predicted by both databases included CA4 and CYP1A2 (Figure D). After merging and deduplicating BRCA-related target genes from the two databases, 6045 BRCA-associated targets were identified. The overlapping targets predicted by both databases, including androgen receptor (AR), are shown in the network (Figure E). Finally, 72 candidate genes were identified by intersecting DEGs, MEP-related targets, and BRCA-related targets (Figure F).

GO analysis indicated that the candidate genes were enriched in 282 terms, including 180 biological process terms such as primary alcohol metabolic process, 22 cellular component terms such as ion channel complex, and 80 molecular function terms such as growth factor binding (adj. P < 0.05; Figure G and Table S1). KEGG analysis revealed the enrichment of candidate genes in 24 pathways, including cellular senescence (adj. P < 0.05; Figure H and Table S2). These findings suggest that BRCA development may be associated with specific metabolic-related pathways.

3.2. Identification and Biomarkers

In the PPI network, 23 candidate genes were identified as isolated targets, whereas 49 such genes showed direct interactions (Figure A). CCNE1 and UGT1A6 were located at the network center, indicating that they may exert substantial effects on BRCA development. After identifying densely connected regions, the top two highest-scoring modules contained five and eight module genes, respectively (Figure B,C). Using the Radiality (Figure D), Closeness (Figure E), and Degree algorithms (Figure F), the top 15 genes from each method were obtained. Intersecting these genes with the module genes identified five hub genes (Figure G).

2.

Identification and biomarkers (A–C) PPI network: The interaction of 23 candidate genes (A), CCNE1 (B), and UGT1A6 (C). (D–F) The top 15 genes acquired by Radiality (D), Closeness (E), and Degree algorithms (F). (G) Venn diagram of intersecting genes from the above 3 algorithms with the module genes. (H–I) Box plot analysis indicated the hub genes in the training set and validation set.

In TCGA-BRCA and GSE42568, except for CDKN2A, the remaining four hub genes showed significantly higher expression in BRCA samples than controls, indicating that CCNE1, CDK1, E2F1, and EZH2 are MEP-related biomarkers in BRCA (P < 0.001; Figure H,I).

3.3. Pathways Enriched with Biomarkers and Immune Microenvironment in BRCA

GSEA revealed that CDK1 was significantly enriched in 80 pathways, including cell cycle and DNA replication pathways (Figure A and Table S3). E2F1 was significantly enriched in 88 pathways, including cell cycle, spliceosome, and DNA replication pathways (Figure B and Table S4). EZH2 was significantly enriched in 109 pathways, including cell cycle and proteasome pathways (Figure C and Table S5). CCNE1 was significantly enriched in 79 pathways, including cell cycle and spliceosome pathways (Figure D and Table S6). Overall, the top pathways in which all biomarkers were strongly enriched were largely overlapping, demonstrating functional consistency among the biomarkers and indicating that these pathways contribute to the development of BRCA.

3.

Pathways enriched with biomarkers in BRCA (A–D) GSEA revealed the enriched pathway of (A) CDK1, (B) E2F1, (C) EZH2, and (D) CCNE1.

The proportional distribution of immune cell infiltration across 22 immune cell types in BRCA and control samples is shown in Figure A. Notably, 15 immune cell types, including activated CD4 memory T cells, eosinophils, and memory B cells, exhibited significant differences in the infiltration between BRCA and control groups (P < 0.01; Figure B). Compared with controls, M2 macrophage infiltration was significantly reduced in BRCA samples (P < 0.0001), whereas M1 macrophage infiltration was significantly increased (P < 0.0001). Among immune cells, plasma cells showed the strongest positive association with naïve B cells (odds ratio [OR] = 0.52, P < 0.001; Figure C). In contrast, monocytes exhibited the strongest negative association with M0 macrophages (OR = −0.57, P < 0.001). Specifically, CCNE1 demonstrated the highest positive correlation with M0 macrophages (cor = 0.36, P < 0.0001; Figure D and Table S7). In summary, BRCA development may be closely related to macrophages or other immune cell populations. However, whether this correlation was caused by MEP still requires further verification.

4.

The proportional distribution of immune cell infiltration across 22 immune cell types in BRCA (A) The proportional distribution of 22 immune cell infiltration in both the BRCA and control samples. (B) The infiltration level of 15 immune cells with significant in both the BRCA and control samples. (C) Heatmap plot showing the association between four hub genes and 15 immune cell types. (D) Bubble plot showing the relative abundance of indicated immune cell subsets in the BRCA cohort versus. controls. Bubble size corresponds to the mean proportion of each cell population.

3.4. The Binding Ability of MEP to Biomarkers

The binding energies between MEP and CDK1, E2F1, EZH2, and CCNE1 were −5.8 kcal/mol (Figure A), −5.7 kcal/mol (Figure B), −5.4 kcal/mol (Figure C), and −6.2 kcal/mol (Figure D), respectively, all of which belong to moderate binding strength (−5.0∼−7.0 kcal/mol) (Table ). Among these biomarkers, CCNE1 showed the most favorable predicted binding energy among the four targets, suggesting that it may represent a potential interaction partner of MEP. To further interpret the docking poses at the residue level, PLIP and Discovery Studio analyses suggested that MEP binding to CCNE1, CDK1, E2F1, and EZH2 was mainly stabilized by hydrophobic contacts and hydrogen bonds (Figure S2, Table S8). Key interacting residues included ILE10/VAL64/LEU134/GLU51 in CCNE1, TYR223/VAL226/LEU300 in CDK1, multiple hydrogen-bonding residues in E2F1, and ASN567/TYR574 in EZH2, supporting the predicted docking poses and binding stability.

5.

The binding ability of MEP to biomarkers Molecular docking models of (A) CDK1, (B) E2F1, (C) EZH2, and (D) CCNE1 with MEP. (E) RMSD analysis of protein backbone atoms during 100 ns MD simulations. (F) RMSF analysis of protein backbone atoms during 100 ns MD simulations. (G) Rg analysis of protein backbone atoms during 100 ns simulations.

2. Binding Energy of Molecular Docking.

gene	PDB ID	drug	binding energy (kcal/mol)
CDK1	4yc3	MEP	–5.8
E2F1	7toa	MEP	–5.7
EZH2	4mi0	MEP	–5.4
CCNE1	5I2w	MEP	–6.2

Subsequently, MD simulations were performed for 100 ns. The RMSD analysis of biomarkers showed that CCNE1, CDK1, E2F1, and EZH2 exhibited RMSD values of 0.15–0.25, 0.15–0.30, 0.15–0.35, and 0.15–0.35 nm, respectively, indicating that both proteins and their ligands reached a stable state between 20 and 100 ns of the simulation (Figure E). RMSF analysis revealed prominent peaks in regions 4000–5000 and 6000–7000 for CCNE1, 500–600 for CDK1, 2000–2100 for E2F1, and 500–700 for EZH2, suggesting increased flexibility in these regions, whereas other regions remained relatively stable (Figure F). Finally, radius of gyration analysis revealed that protein peptide chains maintained structural stability throughout the simulation (Figure G). The RMSF curve of CDK1 showed a significantly higher fluctuation in the 500–600 residue region compared to other proteins. As a cyclin-dependent kinase, the core structure of CDK1 consists of an N-terminal β-sheet and a C-terminal α-helix. Between the two lobes were the ATP/substrate binding pocket, and the hinge region and activation loop connecting the two lobes are highly flexible areas. The T-loop was in an unordered and nonactivated conformation when not bound to the cyclin B, and thus underwent intense conformational fluctuations during simulation. The 100 ns molecular dynamics simulation results showed that in the CCNE1-MEP system, MEP was stably bound to the dimerization interface of CCNE1, with Glu56 (hydrogen bond, 78%) and Leu60 (hydrophobic interaction, 82%) being the key interacting residues; in the CDK1-MEP system, the T-loop of CDK1 presented a closed activated conformation, and MEP occupied the ATP pocket, with Lys33 (hydrogen bond, 85%) and Arg150 (electrostatic interaction, 88%) being the key interacting residues; in the E2F1-MEP system, MEP bound to the transcription activation region of E2F1, with Asn30 (hydrogen bond, 75%) and Phe34 (hydrophobic interaction, 80%) being the key interacting residues; in the EZH2-MEP system, MEP was located in the SET domain catalytic pocket of EZH2, with Tyr641 (hydrogen bond, 92%) and Trp642 (hydrophobic interaction, 89%) being the key interacting residues. Overall, these molecular docking and simulation results support the plausibility of the predicted MEP–biomarker interactions.

3.5. Cell Types in the scRNA-seq Data

Before quality control, 139,841 and 26,234 genes were detected in the data set (Figure S1A). After quality control, 110,619 and 26,234 genes remained for subsequent analysis (Figure S1B). Using standardized scRNA-seq data, the top 2000 HVGs were identified and the top 10 HVGs were labeled (Figure A). The JackStraw plot showed that the top 30 PCs were significant (P < 0.05) (Figure S1C). Because variance explained by the PCs gradually stabilized after the first 20 components, the top 20 PCs were selected for subsequent analysis (Figure S1D). Clustering analysis grouped cells into 12 distinct clusters (Figure S3). Based on marker genes, these clusters were annotated into eight cell types, including fibroblasts, epithelial cells, endothelial cells, natural killer T cells, smooth muscle cells, macrophages, B cells, and mast cells (Figure B). Finally, expression levels of marker genes across annotated cell types were visualized (Figure C).

6.

Cell types in the scRNA-seq data (A) Mean–variance plot showing the identification of HVGs across standardized scRNA-seq data sets. (B) Cluster analysis showing the 8 cell subsets based on marker genes. (C) Dot plot displaying the averaged expression levels (color scale) and percentage expression levels (dot size) for selected marker genes across the 8 cell clusters.

3.6. The Communication Situation Among Cells

In control samples, endothelial cells interacted primarily with epithelial cells and fibroblasts (Figure A), with notable interaction weights between endothelial cells and fibroblasts (Figure B), whereas interactions between epithelial cells and macrophages occurred through MIF-(CD74 + CD44) signaling (Figure C). In BRCA samples, the number of interactions between endothelial cells and fibroblasts was the highest (Figure D), fibroblasts showed the strongest interactions with macrophages and endothelial cells (Figure E), and epithelial cell–macrophage interactions also occurred via MIF-(CD74 + CD44) signaling (Figure F).

7.

The communication situation among cells (A, B) the interaction number (A) and the interaction weight (B) of 8 cell subsets in the control samples. (C) The interaction between epithelial cells and macrophages in the control samples. (D–F) The interaction number (D) and the interaction weight (F) of 8 cell subsets in the BRCA samples. (F) The interaction between epithelial cells and macrophages in the BRCA samples.

3.7. Key Cells in the scRNA-seq Data

Comparison of cell abundances between BRCA and control samples showed that all cell types except endothelial cells were defined as differential cells (Figure A,B). Compared with controls, the proportions of certain cell types, such as macrophages, were significantly higher in BRCA samples (P < 0.0001), whereas the proportions of other cell types, such as fibroblasts, were significantly higher in control samples (P < 0.0001). Similarly, comparison of biomarker expression among these differential cell types between BRCA and control samples yielded the following results: CDK1 expression was significantly higher in fibroblasts from control samples than in those from BRCA samples (P < 0.05; Figure C); compared with controls, E2F1 was significantly upregulated in fibroblasts (P < 0.001) and smooth muscle cells (P < 0.01) but significantly downregulated in epithelial cells in both groups (P < 0.01; Figure D); EZH2 exhibited significant differential expression across most cell types, with notably higher levels in control-derived fibroblasts (P < 0.01; Figure E); CCNE1 showed significant differential expression in fibroblasts and smooth muscle cells between groups, with significantly higher expression in control samples (P < 0.001; Figure F). In summary, all biomarkers exhibited significant differential expression in fibroblasts between groups; therefore, fibroblasts were identified as key cells in the scRNA-seq data set. Notably, fibroblasts from BRCA samples were enriched in pathways such as amino sugar and nucleotide sugar metabolism, whereas fibroblasts from control samples were enriched in pathways such as pyruvate metabolism (Figure G).

8.

Key cells in the scRNA-seq data (A, B) Relative comparison of spatial clusters between the BRCA and the control samples. (C–F) The violin plot revealed the expression level of (C) CDK1, (D) E2F1, (E) EZH2, and (F) CCNE1 in different cell subsets between the BRCA samples versus. controls. (G) KEGG results for fibroblasts between the BRCA samples versus. controls.

3.8. The Expression of Biomarkers in Key Cells

Following reannotation, fibroblasts were reclassified into six subtypes: IL6, PI16, LRRC15, CD74, MYH11, and CTNNB1 fibroblasts (Figure A). Expression of corresponding marker genes across these six cell subtypes is shown in Figure B. During differentiation, fibroblasts transitioned from dark to light blue and were distributed across seven states (Figure C). In the early stage of cell differentiation, PI16 fibroblasts showed a pronounced peak; during the intermediate stage, multiple fibroblast subtypes coexisted; and during the late stage, CTNNB1 fibroblasts predominated (Figure D). Moreover, during fibroblast differentiation, CDK1 and CCNE1 were highly expressed in early and intermediate stages, whereas EZH2 and E2F1 were predominantly expressed in intermediate and late stages (Figure E).

9.

The expression of biomarkers in key cells (A) Annotation of fibroblasts subclusters based on marker gene expression. (B) The marker gene expression in 6 cell subtypes. (C) Annotation of fibroblasts in the differentiation process. (D) Cell trajectory analysis of 6 cell subtypes. (E) Relative expression of 4 genes in the differentiation process of fibroblasts.

3.9. Assessment of the Effect of MEP In Vitro

To validate MEP effects, qRT-PCR and Western blotting were used to assess biomarker expression. Results showed that in MCF-7 cells treated with MEP, mRNA expression levels of CCNE1, CDK1, and EZH2 were significantly increased compared with controls, whereas E2F1 mRNA expression remained unchanged (Figure A). In contrast, CCNE1, CDK1, E2F1, and EZH2 protein levels were all upregulated (Figure B).

10.

MEP promote the malignant biological behavior of MCF-7 (A) qRT-PCR data showing CDK1, EZH2, E2F1, and CCNE1 mRNA expression. (B) Western blots showing CDK1, EZH2, E2F1, and CCNE1 expression. (C) Macrographs of colony formation and quantitative analysis. (D) Flow cytometry showing cell cycle distribution and cell percentages in different phases. (E) Wound healing assay with quantitative analysis was performed to assess the migratory capacity. Scale bar 650 μm. (F) Representative cell migration images and migration index. Scale bar 275 μm. (G) Phalloidin staining was employed to evaluate F-actin reorganization. Scale bar 25 μm. (H) Western blots showing E-cadherin, Vimentin, MMP2, and MMP9 expression. Data are presented as the mean ± standard deviation (n = 3), *P < 0.05, **P < 0.01, ***P < 0.001 and ****P < 0.0001 when compared with control cells.

Colony formation assays showed that MEP treatment significantly enhanced the clonogenic ability of MCF-7 cells (Figure C). Flow cytometry revealed that MEP significantly altered cell cycle distribution, characterized by shortened G1 phase and prolonged S phase (Figure D), indicating accelerated G1/S transition. Wound healing and Transwell assays showed significantly enhanced MCF-7 cell migration and invasion following 48 h MEP exposure compared with controls (Figure E,F). Consistent with these results, phalloidin staining demonstrated marked reorganization of the actin cytoskeletal architecture in MEP-treated cells (Figure G). Furthermore, analysis of epithelial–mesenchymal transition (EMT) markers revealed MEP-mediated modulation of key molecular signatures, including increased E-cadherin expression (epithelial marker), reduced vimentin expression (mesenchymal marker), and elevated levels of matrix metalloproteinases MMP2 and MMP9 (Figure H).

Collectively, these findings provide strong evidence that MEP promotes malignant behaviors in MCF-7 cells through coordinated regulation of cell cycle progression, cytoskeletal remodeling, and EMT-associated molecular reprogramming.

4. Discussion

With its high incidence and therapeutic heterogeneity, BRCA represents a global health burden for women. The prevention and management of BRCA remain long-term challenges that warrant continued investigation. MEP, a widely used component of paraben preservatives, is ubiquitous in everyday consumer products. Concerns have increasingly been raised about the potential carcinogenic effects of MEP, as residual MEP has been detected in both in vivo and in vitro settings. Previous studies have suggested that MEP may function as an estrogen receptor agonist, potentially elevating BRCA risk. Therefore, this study sought to validate the effects of MEP on BRCA and further investigate its underlying toxicological mechanisms.

In this study, an integrative approach combining transcriptomic analysis, network toxicology, and molecular docking was applied to identify and validate four putative MEP-associated biomarkers in BRCA: CCNE1, CDK1, E2F1, and EZH2. GSEA revealed that these key genes predominantly modulate the cell cycle and biological process pathways. Although the methodological approach employed by Yang et al. to explore the relationship between MEP and BRCA partially overlaps with that of the present investigation. Their study aims to establish a causal relationship between MEP and BRCA through Mendelian randomization, followed by the integration of network toxicology, multiomics analyses, and molecular docking to identify key genes associated with nonhormone-dependent pathways. Contrary, the present study focused on a distinct set of key genes with further investigations compared to prior research. Immune infiltration analysis indicated the involvement of immune cell populations in BRCA progression. Additionally, scRNA-seq analysis identified fibroblasts as potential key cellular mediators of toxicant-driven carcinogenesis. In vitro experiments demonstrated that MEP altered the expression of these four pivotal genes and disrupted cell cycle progression along with extracellular matrix remodeling, thereby promoting malignant phenotypes in BRCA cells. Overall, this study provides a more comprehensive and in-depth examination chain spanning molecular, cellular, and phenotypic levels between MEP and BRCA.

Previous studies have demonstrated that parabens promote BRCA development through processes such as enhanced proliferation, cell cycle dysregulation, and metastasis. In vitro studies have further confirmed that parabens increase MCF-7 cell viability at specific concentrations. , Peinado et al. suggested that parabens and their analogs may disrupt cell cycle progression in endometrial tissues via CDK1 interference. Another study indicated that propylparaben exposure induces DNA damage in mouse oocytes and subsequently impairs CDK1 function, resulting in G2/M phase arrest and inhibition of oocyte maturation. However, few studies have directly examined the relationship between MEP and key regulators such as CCNE1, E2F1, CDK1, and EZH2, which presents both a challenge and a distinctive innovative aspect of the present study. Nevertheless, cellular experiments and zebrafish models have provided evidence that MEP disrupts the expression of cell cycle–related genes. , Consistent with these reports, GSEA results confirmed enrichment of cell cycle pathways associated with the four candidate genes.

CCNE1 exhibits dynamic variation in protein expression across the cell cycle. Evidence indicates that CCNE1 is aberrantly expressed in multiple malignancies. In BRCA, particularly triple-negative breast cancer (TNBC), CCNE1 overexpression has been identified as an early event in tumor progression, with elevated expression levels consistently associated with poor clinical outcomes. Although direct evidence linking MEP exposure to specific BRCA subtypes is lacking, its estrogen-mimicking activity suggests a potential role in promoting hormone receptor–positive BRCA development. Clinical studies have demonstrated that patients with ER + /HER2– BRCA treated with palbociclib exhibit significant CCNE1 downregulation. Conversely, CCNE1 amplification may drive acquired resistance to CDK4/6 inhibitors, including palbociclib and ribociclib, thereby positioning CCNE1 as a predictive marker for reduced progression-free survival following CDK4/6 inhibitor therapy. , Interestingly, tumors showing CCNE1 overexpression demonstrate enhanced responsiveness to immunotherapy in clinical cohorts. Therefore, CCNE1 may serve as an alternative biomarker for immunotherapeutic stratification.

CDK1 (cyclin-dependent kinase 1) functions as a master regulator of essential cellular processes, including DNA replication, transcription, repair, and morphogenesis. Studies have established CDK1 as key regulator of the G2/M transition, controlling mitotic entry via phosphorylation of diverse substrates. Elevated CDK1 expression correlates strongly with poor prognosis in BRCA. , Studies have shown that NFIX-mediated downregulation of CDK1 induces mitotic delay and effectively suppresses BRCA cell proliferation. Conversely, the deubiquitinase YOD1 stabilizes CDK1 protein levels, thereby accelerating TNBC progression in PDX models. Moreover, disruption of the CDC25C–CDK1 interaction promotes DNA damage accumulation and cell cycle arrest, leading to effective suppression of TNBC growth. Collectively, these findings highlight CDK1 as a promising therapeutic target for BRCA.

E2F1 (E2F transcription factor 1), a master regulator of the eukaryotic cell cycle, is significantly upregulated in primary breast tumors compared with normal mammary tissue. E2F1 is known to drive cell cycle progression through the G1/S transition and S-phase entry, thereby facilitating oncogenesis and TNBC progression. Mechanistically, E2F1 transcriptionally activates CDCA5, which supports cancer cell proliferation. In turn, increased CDCA5 expression enhances E2F1 binding to the FOXM1 promoter, inducing FOXM1 expression and accelerating BRCA progression. Additionally, E2F1 exhibits functional interplay with CDK1. The chromosomal condensin complex component NCAPD2 promotes the proliferation and metastasis of MDA-MB-231 cells via the E2F1–CDK1 activation axis.

EZH2 (enhancer of zeste homologue 2) is frequently overexpressed in multiple malignancies, including prostate cancer and BRCA. EZH2 has been shown to promote therapeutic resistance in BRCA through multiple mechanisms, including TME remodeling, cell cycle regulation, EMT, and dysregulation of critical signaling pathways, such as the PTEN/PI3K/AKT/mTOR and RAS/RAF/MEK/ERK pathways. EZH2 inhibition is also known to sensitize tamoxifen-resistant BRCA cells through regulation of cell cycle regulation. Thus, EZH2 holds substantial therapeutic potential for overcoming drug resistance in BRCA and exhibits oncogenic properties; furthermore, its ectopic expression drives proliferation in otherwise nontransformed cells. Additionally, EZH2-mediated activation of integrin β1–FAK signaling converges with TGF-β signaling to drive BRCA bone metastasis.

Previous findings demonstrate that targeting the cell cycle regulatory network represents a promising antitumor strategy. Inhibition of E2F1 degradation increases CCNE1 expression levels. E2F1, EZH2, and CDK1 engage in reciprocal regulatory interactions that collectively drive malignant proliferation. − These findings suggest the therapeutic rationale for simultaneously targeting this regulatory network to disrupt BRCA progression. PCR and Western blot analyses further demonstrated that MEP upregulated the expression of key genes. Flow cytometry confirmed that MEP disrupted cell cycle progression in BRCA cells. Phenotypic and molecular assays consistently indicated that MEP enhances the metastatic potential of MCF-7 cells.

Immune infiltration analysis revealed significant differences in the abundances of 15 distinct immune cell populations. Correlation analysis demonstrated significant positive associations between CCNE1, CDK1, E2F1, and EZH2 expression levels and the infiltration of T follicular helper cells, regulatory T cells (Tregs), M1 and M0 macrophages, and activated CD4+ memory T cells. Conversely, negative correlations were observed with resting CD4+ memory T cells, naïve B cells, monocytes, and M2 macrophages. A previous study reported that high infiltration of T follicular helper cells in BRCA tumor-draining lymph nodes is associated with poor prognosis. However, other studies have indicated that T follicular helper cells can enhance tumor sensitivity to immunotherapy by mediating B-cell–dependent responses. , Tregs contribute to the formation of an immunosuppressive TME, thereby facilitating TNBC progression and reducing therapeutic efficacy. CAF-S1, a carcinoma-associated fibroblast (CAF) subset, enhances the immunoregulatory function of Tregs to suppress T-effector cell proliferation. In general, proinflammatory M1 macrophages exhibit antitumor activity, whereas anti-inflammatory M2 macrophages are associated with tumor-promoting functions. Nevertheless, evidence suggests that M1 macrophages may exert protumorigenic effects in specific contexts. Increased infiltration of activated CD4+ memory T cells and naïve B cells has been associated with improved prognosis, whereas elevated resting CD4+ memory T cells correlate with poorer outcomes in BRCA. − Unexpectedly, immune infiltration analysis yielded results partially inconsistent with our initial hypothesis. We propose three potential explanations: (1) substantial heterogeneity may exist among immune cell profiles across distinct BRCA subtypes, requiring more refined comparative analyses; (2) the four identified genes may not directly regulate immune cell phenotypes or functions; and (3) these findings potentially highlight novel biological phenomena that warrant further mechanistic investigation.

Based on the scRNA-seq data analysis, fibroblasts were identified as key cellular contributors to BRCA progression. CAFs are central regulators of the TME, coordinating extracellular matrix remodeling, mediating bidirectional communication with tumor cells, and shaping immune cell crosstalk. Notably, CD74+ antigen-presenting fibroblasts in BRCA exhibit context-dependent functions, ranging from immunosuppression to enhancement of antitumor immunity, depending on the microenvironmental context. BRCA-associated fibroblasts promote tumor cell migration and invasion through paracrine secretion of multiple soluble factors, including TGF-β, HGF, bFGF, FSP-1, CCL11, CXCL14, CCL2, and IL-6. Furthermore, patients with elevated LRRC15 + /PI16+ fibroblast ratios show reduced T-cell and antitumor cytokine expression levels, consistent with an immunosuppressive phenotype. Accordingly, characterizing fibroblast heterogeneity between healthy and pathological mammary tissues is of considerable importance for early detection of environmental toxicants and advancing precision oncology.

Although this study achieved substantial progress in investigating the association between MEP and BRCA-related biomarkers, several limitations warrant consideration. First, by integrating bioinformatics analyses, molecular docking, and foundational cellular experiments, potential biomarkers and key cell types were identified; however, more extensive experimental validation is necessary to confirm their reliability and clinical relevance. Second, although immune infiltration analysis revealed significant differences in 15 immune cell populations between BRCA and control samples, the correlations between these immune subsets and the identified biomarkers did not fully align with the initial hypotheses, highlighting the need for further mechanistic investigation. Third, while this study linked MEP to four core genes via integrated computational and in vitro approaches, their potential role as “pan-cell cycle regulators” remains a concern. Additionally, since transcription factors like E2F1 often lack defined binding pockets, our docking and MD results are exploratory predictions rather than proof of direct binding. Further validation using biophysical assays (e.g., SPR/ITC), mutagenesis, and functional studies is required to confirm these interactions and their biological significance. Fourthly, although scRNA-seq primarily enabled cell type characterization and biomarker expression profiling in critical cellular subsets, it provided limited insight into complex intercellular communication and signaling networks. Moreover, complex cell–cell crosstalk within the TME plays a pivotal role in BRCA progression, highlighting the importance of detailed investigation into fundamental cellular communication mechanisms and their modulation via MEP exposure and associated biomarkers. Finally, reliance on publicly available data sets introduces potential biases and inherent limitations that should be acknowledged.

5. Conclusion

Using bioinformatics analyses and experimental validation, this study identified four biomarkers (CCNE1, CDK1, E2F1, and EZH2) and fibroblasts as key contributors to the effects of MEP in BRCA, providing important scientific evidence for defining therapeutic targets, refining molecular intervention strategies, and fully elucidating the association between MEP and BRCA.

All data are available within the manuscript and Supporting Information files. All analysis scripts involved in this study have been deposited in the Science Data Bank (ScienceDB), an open-access research data repository operated by the Computer Network Information Center, Chinese Academy of Sciences. A permanent DOI has been assigned: 10.57760/sciencedb.35995

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsomega.5c11680.

• Quality control of the scRNA-seq data (Figure S1); two-dimensional ligand–protein interaction diagrams of methyl 4-hydroxybenzoate (MEP) with the four biomarkers (Figure S2); unlabeled cluster analysis of all samples (Figure S3); GO analysis (Table S1); KEGG analysis (Table S2); GSEA analysis of CDK1 (Table S3); GSEA analysis of E2F1 (Table S4); GSEA analysis of EZH2 (Table S5); GSEA analysis of CCNE1 (Table S6); proportional distribution of 22 immune cell infiltrations (Table S7); PLIP interaction profiles of MEP with CCNE1, CDK1, E2F1, and EZH2 (Table S8) (PDF)

#.

Y.M., X.Z., and Z.Y. contributed equally to this work and share cofirst authorship. Y.M.: Writingreview and editing, writing–original draft, visualization, investigation, conceptualization. X.Z.: Writing–review and editing, investigation, formal analysis, data curation. Z.Y.: Writing–review and editing, investigation, formal analysis, data curation. Y.D.: Writing–review and editing, and data curation. H.S.: Writingreview and editing, data curation. L.Z.: Writing–review and editing, visualization, investigation. H.L.: Writing–review and editing, visualization, investigation. M.X.: Visualization, investigation. B.W.: Writingreview and editing, funding acquisition. W.H.: Writingreview and editing, validation, supervision, resources, conceptualization. X.M.: Writingreview and editing, validation, supervision, resources, project administration, funding acquisition, and conceptualization. D.Q.: Writing–review and editing, validation, supervision, resources, project administration, funding acquisition, and conceptualization.

This research was supported by the National Natural Science Foun dation of China (Grant No. 82404685), the Zhejiang Science and Technology Department “vanguard” “leading goose” research (Grant No. 2023C03044), the Sanming Project of Medicine in Shenzhen (SZSM202301035), the Research Project of Jiangsu Association of Chinese Medicine (CYTF 2024050), and the Suzhou Science and Technology Development Program (SKYD2023087).

The data used in this study are sourced from publicly available databases and do not involve the direct collection of data from human participants. According to the guidelines of the local ethics review board, research using publicly available data is exempt from requiring ethical approval. Therefore, no additional ethical approval is needed for this study.

The authors declare no competing financial interest.