Glyphosate-Induced Metabolic and Immune Modulation in Hepatoma Cells: Identification of Key Genes as Diagnostic and Therapeutic Targets Using an In Silico Systems Biology Approach

Abstract

Glyphosate, one of the most widely used herbicides worldwide, has raised significant concerns regarding its potential involvement in hepatotoxicity and molecular changes associated with liver cancer biology. These concerns highlight the need to better understand its underlying molecular mechanisms in hepatoma cells. Emerging evidence suggests that glyphosate exposure may increase the risk of liver cancer and chronic liver disease. However, the precise molecular alterations and promising biomarkers associated with glyphosate-induced hepatic toxicity and disease remain largely unexplored. In this study, an RNA-Seq-based in silico systems biology approach was employed to elucidate glyphosate-induced differential transcriptional profiling in hepatoma cells. This analysis revealed significant transcriptional profiling characterized by the upregulated hub genes ATF3, JUNB, ALDOA, FOSB, PFKFB3, G6PD, ENO2, HK2, FOS and PGK1. These genes were primarily associated with glucose metabolism, TNF-α/NF-κB signaling, epithelial–mesenchymal transition (EMT) and cellular stress responses. Conversely, several key genes were significantly downregulated, including PIK3R1, FYN, CEBPA, MLXIPL, PPARA, CD36, PCK2, PNPLA3, NR1H4 and MGLL, which were involved in lipid metabolism, immune regulation and non-alcoholic fatty liver disease (NAFLD) pathways. Notably, all hub genes demonstrated strong diagnostic performance, highlighting their potential as sensitive biomarkers of glyphosate exposure. Collectively, this study provides comprehensive insights into gene expression changes associated with glyphosate exposure in hepatoma cells, linking them to hepatic metabolic dysregulation and immune modulation and suggesting a panel of hub genes with potential diagnostic and therapeutic significance.

1. Introduction

Glyphosate is the most commonly used systemic herbicide in paddy cultivation worldwide and has become deeply embedded in modern agricultural systems due to its broad-spectrum weed control and extensive commercial application [1,2]. Glyphosate has traditionally been considered relatively safe for humans because its primary target, the shikimate pathway, is absent in human cells [3]. In 2004, the Food and Agriculture Organization of the United Nations (FAO) in collaboration with the World Health Organization (WHO) established an acceptable daily intake of 1 mg/kg body weight for glyphosate. This decision was based on the evidence indicating low dietary residues, long-term usage without significant reported health risks, and its chemical and physical characteristics suggesting minimal toxicity [4]. However, this perception began to change when the International Agency for Research on Cancer (IARC) reported glyphosate as a probable human carcinogen based on evidence from animal studies and supportive mechanistic data. This study challenged earlier assumptions regarding its safety profile [5,6,7,8]. Despite continued regulatory review, glyphosate is officially approved for use against approximately 100 annual weed species and over 60 perennial weed species and is currently utilized in more than 130 countries worldwide [9,10].

In 2023, the European Commission extended glyphosate’s approval for the next ten years after conducting a detailed scientific evaluations by the European Food Safety Authority (EFSA) and the European Chemicals Agency (ECHA) [11]. Similarly, the United States Environmental Protection Agency (EPA) has reported that glyphosate is not carcinogenic to humans when used according to label instructions [12,13]. Nevertheless, the renewal was granted with consistent monitoring and the implementation of specific precautionary measures to minimize potential risks. These divergent regulatory conclusions underscore the importance of mechanistic investigations aimed at elucidating glyphosate induced molecular alterations. Glyphosate residues have been consistently detected in environmental samples, indicating the possibility of long-term human exposure [14,15]. Because the liver plays a central role in metabolism and detoxification of xenobiotics, it represents a biologically relevant target for evaluating the potential molecular effects of glyphosate or any other herbicide exposure. Hepatic C3A cells, a clonal derivative of HepG2 hepatocellular carcinoma cells, represents the best characterized human liver cell line; moreover, it is widely used as a model system to study xenobiotic toxicity [16,17]. In this context, investigating glyphosate-induced transcriptional changes in specific liver cancer cell models suggests mechanistic insights that complement epidemiological and toxicological data and help to further justify the relevance of this study.

Previous experimental studies have demonstrated that glyphosate exposure may induce liver dysfunction, disturb lipid metabolism and promote oxidative stress in rats. Hepatotoxicity has been evaluated through alterations in standard liver function tests that include serum alanine aminotransferase, aspartate aminotransferase, lactate dehydrogenase activities, along with measurements of serum lipoproteins, albumin, total protein, triglyceride and cholesterol [18,19]. It has been reported that glyphosate exposure alters hepatic glucose and lipid metabolism in mice through perturbation of the circadian clock system [20]. Moreover, studies have shown that subacute exposure of rats to glyphosate triggers enhanced inflammatory responses. Following glyphosate exposure, levels of C-reactive protein and the cytokines IL-1β, IL-6 and TNF-α were elevated in both liver and adipose tissues. These biochemical changes indicated fatty liver disease progression, systemic inflammation and hepatic fibrosis [21,22]. Beyond traditional rodent models, emerging studies have begun to explore the impact of glyphosate on the liver tumor immune microenvironment [23,24,25,26]. Therefore, identification of glyphosate-specific molecular signatures is essential for distinguishing glyphosate exposure-driven molecular alteration from intrinsic cancer biology and for developing exposure-relevant biomarkers and therapeutic targets.

In this context, high-throughput transcriptomic profiling combined with systems biology approaches provides an effective way to explore the complex molecular interactions involved in glyphosate-associated carcinogenesis. Network- and enrichment-based analyses facilitate identification of key hub genes and pathways influencing disease phenotypes. Moreover, integrating these findings with publicly available cancer datasets and chemical–gene interaction resources further enables validation and assessment of potential clinical relevance.

Recent case studies have reported significantly higher levels of glyphosate and its metabolites in chronic liver disease and liver cancer patients compared with healthy individuals, highlighting potential association between glyphosate exposure and liver pathology [27,28,29]. Despite huge progress in the field of liver cancer genomics and developing targeted therapies, limited studies have systematically characterized the transcriptomic landscape associated with glyphosate exposure in hepatoma cells, thereby emphasizing the need to better understand glyphosate exposure in liver cancer models.

In the present study, an RNA-Seq-based in silico systems biology framework was employed to investigate glyphosate-induced transcriptional alterations in hepatoma cells.

Differential gene expression analysis, protein–protein interaction network prioritization, pathway enrichment, immune gene mapping, toxicogenomic validation and diagnostic performance assessment were integrated to characterize glyphosate-associated gene expression changes and evaluate their potential relevance to liver disease and cancer-related pathways. This comprehensive strategy enables the identification of both previously reported and novel glyphosate sensitive genes, providing mechanistic insight into glyphosate exposure-associated molecular changes relevant to liver cancer biology and highlighting key genes for further evaluation as potential biomarkers.

2. Materials and Methods

2.1. Data Retrieval and Preprocessing

Raw paired-end RNA-Seq data for nine hepatic C3A samples that include three glyphosate-treated and six untreated controls were downloaded from the NCBI Gene Expression Omnibus and Sequence Read Archive (SRA) using accession number GSE292868 [30]. The datasets were subsequently processed for downstream transcriptomic analysis.

The C3A human hepatoma cells, a clonal derivative of the HepG2 cell line, are commonly used as an in vitro model for studying liver-specific xenobiotic metabolism and toxicological responses. According to the original experimental design described in the source study, C3A hepatoma cells were exposed to increasing glyphosate concentrations (0.1–10,000 µM) for 24 and 48 h to assess cytotoxicity and determine IC50 at 48 h. For transcriptomic profiling, a subcytotoxic concentration of 5.6 µM was selected and applied for 48 h. The present study utilized the raw RNA-Seq data generated under these predefined exposure conditions for downstream analysis.

2.1.1. Raw Data Quality Check

The raw sequencing data from nine RNA-Seq samples were downloaded in FASTQ format using the NCBI SRA Toolkit with four threads enabled to improve processing efficiency [31]. The RNA-Seq data analysis begins with evaluating the quality of the retrieved raw sequencing data, which is essential for ensuring accurate downstream interpretation. The quality of the reads obtained in FASTQ format was explored to assess the data characteristics such as per-base sequence quality, GC content, sequence length distribution, duplicated reads and adapter contamination using FASTQC version v0.12.1 [32].

2.1.2. Read Alignment to Reference Genome

High-quality reads were then used to map and align to the latest human genome assembly GRCh38/hg38 in FASTA format, downloaded from the ENSEMBL database [33]. The alignment and mapping of the raw sequencing data were performed using the Linux-based software HISAT2 version 2.2.1 [34]. HISAT2 is widely used as a preferred method for efficient and accurate genome assembly. This software employs indexed reference genome files to facilitate rapid and accurate alignment of sequencing reads to the human genome. The alignment process initially generated output in the text-based Sequence Alignment Map (SAM) format. In order to minimize memory usage, the output SAM stream was directly converted into a sorted BAM file using SAMtools version 1.19.2 [35].

2.1.3. Transcript Assembly and Quantification

Furthermore, the aligned reads to the reference genome are subsequently assembled into transcripts to quantify gene expression levels. Read quantification was performed using the FeatureCounts program from the Subread package [36]. The resulting count matrix was then utilized for differential expression analysis to identify differentially expressed genes (DEGs) between glyphosate-treated samples and untreated control samples.

2.2. Differential Gene Expression Analysis

To identify the genes exhibiting significant expression changes between the glyphosate-treated hepatic cell samples and untreated samples, differential gene expression analysis was performed using DESeq2 package (version 1.44.0) [37] in R software (version 4.4.1) [38]. The DESeq2 method uses negative binomial distribution and shrinkage estimators to robustly identify statistically significant DEGs observed between the given treatment types.

Significant DEGs were initially filtered based on an adjusted p-value (padj) < 0.05. Subsequently, genes meeting this statistical threshold were further filtered using a fold change cutoff of |log2FC| > 1 to identify biologically meaningful expression changes.

To provide a clear overview of the differential expression results, the volcano plot was generated by EnhancedVolcano package in R [39]. The plot displays log2 fold changes against the −log10 of the adjusted p-values obtained from DESeq2 output. Color coding was applied to distinguish significantly upregulated (red) and downregulated (blue) genes in the plot.

2.3. Construction of Protein–Protein Interaction (PPI) Network and Hub DEGs Analysis

To investigate protein–protein interactions (PPIs) and functional associations among the identified significant DEGs, the StringApp plugin in Cytoscape software version 3.10.3 was utilized [40,41]. A confidence score threshold of >0.4 was used as the cutoff value. Two separate PPI networks for upregulated and downregulated DEGs were constructed.

In order to identify the highly interconnected gene clusters (modules) within the (PPI) network, Molecular Complex Detection (MCODE) plugin in Cytoscape software was applied [42]. The analysis was performed using a Degree Cutoff ≥ 3 while other parameters were kept at their default settings: Node Score Cutoff ≥ 2, K-Core ≥ 2, and Maximum Depth = 100. The top three most significant modules were selected from each PPI network for further analysis. Finally, the CytoHubba, a Cytoscape plugin, was used to extract the top ten hub genes from the PPI network based on the Maximal Clique Centrality (MCC) network topological parameter [43].

2.4. Expression Analysis of the Identified Hub DEGs Using GEPIA2

The expression patterns of the identified hub differentially expressed genes (DEGs) were further crosschecked using the GEPIA2 (Gene Expression Profiling Interactive Analysis 2) web server, where the RNA-sequencing data from The Cancer Genome Atlas (TCGA) database was selected [44,45]. Gene expression patterns in liver hepatocellularcarcinoma (LIHC) samples were analyzed by comparing 369 tumor samples with 50 normal liver samples from the TCGA dataset. Boxplots were generated with the help of GEPIA2 to visualize the differential expression patterns of the selected hub genes between tumor and normal samples. To identify significantly dysregulated hub genes in hepatocellular carcinoma, statistically significant thresholds were set as |log2 fold change| ≥ 1 and p-value < 0.05. This approach helped us in the assessment of altered expression trends of hub genes in hepatocellular carcinoma vs. normal liver tissue samples.

2.5. Gene Set Enrichment Analysis (GSEA)

To uncover pathways and hallmark signatures significantly associated with gene expression changes, GSEA was conducted. GSEA was performed in R using the clusterProfiler package (version 4.16.0) to identify significantly enriched biological pathways associated with differential gene expressions [46,47]. Pre-ranked gene lists based on the log2fold change in all expressed genes were also created. Gene sets were obtained from the MSigDB version (2025.1) using the msigdbr package, and the primary focus was to retrieve the C2 (KEGG legacy) and hallmark signature collections [48,49]. Results were visualized using enrichment plots to highlight key pathways and hallmark processes. Gene sets with false discovery rate q-values < 0.05 and normalized enrichment scores |NES| > 1 were considered significantly enriched.

2.6. Functional Enrichment Analysis

Next, enrichment analysis was performed to investigate the biological roles of the DEGs identified within the significant modules of the PPI network. Gene Ontology (GO) terms, which comprise biological process (BP), cellular component (CC) and molecular function (MF), and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses, were carried out using the DAVID (Database for Annotation, Visualization, and Integrated Discovery) online database [50,51,52]. In addition, disease-associated enrichment analysis was conducted using the DisGeNET database through the DAVID platform [53]. Default parameters were used for each database with statistical significance evaluated using a p-value threshold of <0.05.

2.7. Screening of Glyphosate-Associated Genes

Glyphosate-related genes were retrieved using chemical gene association data from the Comparative Toxicogenomics database (CTD) (http://ctdbase.org/). This database is publicly available and is a curated database of chemical gene and chemical disease interactions from the scientific literature. The downloaded table included gene symbols, gene IDs, interaction types, evidence sentences and references for Homo sapiens. Data preprocessing was performed to obtain a non-redundant list of human genes associated with glyphosate. The steps included removal of duplicate rows and selection of only curated interactions linking glyphosate to the gene with the supporting literature.

2.8. Screening of Immune-Related Gene Sets

To explore the association of the identified hub genes with the immune system, curated genes related to innate and adaptive immunity were collected from multiple publicly available databases. Innate immune genes were retrieved from the InnateDB online database (www.innatedb.com). Conversely, adaptive immune system like B cell- and T cell-specific gene sets were retrieved from the MSigDB C7 Immunologic Signatures dataset using the msigdbr R package (version 4.5.1) [49]. Additionally, Gene Ontology (GO) biological process terms and Reactome pathways related to immune system function were fetched using msigdbr and biomaRt R packages [54].

2.9. Receiver Operating Characteristic (ROC) Analysis

To evaluate the diagnostic potential of the identified influential hub DEGs, ROC curve analysis was performed. ROC curves were generated using the “pROC” package in R software [55]. The area under the ROC curve (AUC) was calculated to assess the diagnostic performance of these genes in distinguishing glyphosate-treated samples from untreated hepatoma cells.

3. Results

3.1. Quality Check and Preprocessing Analysis

The RNA-Seq data analysis pipeline commenced with a comprehensive quality assessment of the raw data to identify low-quality reads that could affect downstream analyses and lead to inaccurate results. All samples showed adequate sequencing depth with read counts ranging from ~20 to 33 million reads per sample across the dataset. Per-base sequence quality (≥Q30 bases) remained uniformly high across the entire read length for both forward and reverse reads, which reflects high accuracy and very low sequencing noise. The GC content distribution was 48–49, showing strong consistency across all samples and closely matching with the expected GC profile of the human transcriptome. Adapter contamination was not detected in any library, which further confirms the high integrity of library preparation and sequencing. Sequence duplication levels were found at a moderate level, that is, ~27–30%, a characteristic feature of RNA-Seq datasets affected by natural transcript abundance.

Additionally, minor per-base sequence composition and K-mer enrichment biases were observed in a few samples. Importantly, these parameters do not indicate low quality of data and are not expected to impact downstream differential expression or systems-level analyses (Supplementary Table S1). Finally, the quality control parameters further confirmed that all samples passed quality thresholds and were used in subsequent transcriptomic profiling and systems biology analyses.

3.2. Differentially Expressed Genes (DEGs) Analysis

To investigate the gene expression changes between the glyphosate-treated hepatoma cells with untreated control samples, differentially expressed gene analysis was performed. A total of nine hepatoma cell samples retrieved from the GEO dataset were analyzed, comprising three glyphosate-treated samples and six untreated samples. Boxplot visualization of normalized gene expression values demonstrated comparable distribution patterns across all samples, indicating appropriate normalization and overall data consistency between treatment groups (Figure 1A). Additionally, no significant outliers were detected, further supporting the high quality of the dataset for downstream analyses.

A heatmap representing the top 50 significantly differentially expressed genes (DEGs) revealed a clear distinction in expression profiles between glyphosate-treated and untreated hepatoma cells (Figure 1B). Genes were hierarchically clustered based on their expression similarities, with treated samples grouping separately from control samples, which leads to robust transcriptomic changes induced by glyphosate exposure. The observed clustering patterns and identified notable differences in gene expression suggest that the obtained DEGs may have meaningful biological significance.

Further, differential gene expression analysis was performed using the DESeq2 package to identify significantly differentially expressed genes (DEGs) between glyphosate-treated and untreated hepatoma cells. Genes with |log2FC| ≥ 1 and adjusted padj < 0.05 were considered significantly differentially expressed between the groups. Overall, 1569 significant DEGs were identified, which include 841 upregulated and 728 downregulated genes. In Figure 1C, the volcano plot depicts the pattern of all identified DEGs in treated vs. untreated samples. Among the obtained differentially expressed genes, LDHAP7, which is a protein-coding gene, was the most highly upregulated (log2FC = 6.0), whereas FP671120.6, a miRNA-coding gene, was observed to be the most significantly downregulated (log2FC = −5.7).

The list of identified significant DEGs is given in Supplementary Table S2. This distinct transcriptome expression profile suggests that glyphosate exposure might induce notable transcriptomic alterations in hepatoma cells.

3.3. Protein–Protein Interaction (PPI) Network Analysis

To gain insight into the interaction between the significant DEGs, the PPI network was constructed and analyzed using the STRING database, restricting the species to Homo sapiens. Separate PPI networks were generated for upregulated and downregulated DEGs. Among the 841 upregulated and 728 downregulated DEGs identified, 304 protein-coding upregulated genes and 173 protein-coding downregulated genes were recognized and mapped by STRING, respectively. As shown in Figure 2, nodes represent protein-coding genes and edges represent correlations between the proteins. The network analysis revealed 908 edges among 304 nodes in the upregulated DEG network and 336 edges among 173 nodes in the downregulated DEG network (Figure 2A,D).

Subsequently, the three most significant modules from each network were identified utilizing MCODE plugin in Cytoscape (Figure 2C,F). Modules are considered as densely connected gene clusters potentially involved in related biological processes. In the upregulated PPI network, module 1 consisted of 19 nodes and 76 edges, module 2 included 20 nodes and 49 edges, and module 3 comprised 15 nodes and 29 edges. Similarly, in the downregulated PPI network, the top three significant modules were the following: module 1 comprised 5 nodes and 10 edges, module 2 was composed of 5 nodes and 10 edges, and module 3 contained 7 nodes and 13 edges.

Moreover, to identify hub genes within the PPI networks, the Maximal Clique Centrality (MCC) algorithm was applied using the CytoHubba plugin in Cytoscape with default parameters. The top 10 genes with the highest MCC scores were identified as hub genes in both networks, as shown in Figure 2B and 2E, respectively. Ten significantly upregulated hub genes were FOSB, PFKFB3, ENO2, HK2, JUNB, FOS, G6PD, ATF3, ALDOA and PGK1. Similarly, ten downregulated hub genes were FYN, CD36, MGLL, PIK3R1, CEBPA, PCK2, NR1H4, MLXIPL, PNPLA3 and PPARA (

Table 1. Identified differentially expressed hub genes in glyphosate-treated vs. untreated hepatoma cells using DESeq2.

Upregulated Hub DEGs			Downregulated DEGs
Genes	Log2Fold Change	padj	Genes	Log2Fold Change	padj
FOSB	2.00	3.96 × 10−22	FYN	−3.53	2.93 × 10−2
PFKFB3	1.87	6.44 × 10−84	CD36	−1.85	7.12 × 10−4
ENO2	1.41	7.87 × 10−28	MGLL	−1.49	2.86 × 10−2
HK2	1.39	8.86 × 10−15	PIK3R1	−1.43	3.55 × 10−48
JUNB	1.34	6.25 × 10−50	CEBPA	−1.24	5.20 × 10−30
FOS	1.30	2.23 × 10−14	PCK2	−1.11	2.80 × 10−79
G6PD	1.26	9.76 × 10−28	NR1H4	−1.10	2.90 × 10−29
ATF3	1.08	2.72 × 10−34	MLXIPL	−1.08	9.03 × 10−28
ALDOA	1.06	1.53 × 10−8	PNPLA3	−1.05	1.62 × 10−6
PGK1	1.02	5.04 × 10−15	PPARA	−1.00	1.10 × 10−28

). Notably, all upregulated hubs were also present in the top three identified modules of the upregulated DEGs. Additionally, out of ten downregulated hubs, FYN, CD36, PIK3R1, CEBPA, PCK2 and NR1H4 genes were common in the modules of the downregulated DEGs.

3.4. Expression Analysis of the Identified Hub DEGs in Tumor vs. Normal Liver Samples

The purpose of this analysis was to examine whether the upregulated and downregulated DEGs identified in glyphosate-treated cancerous hepatocytes compared with untreated hepatocytes exhibit similar expression patterns in human liver tumor versus normal tissues. For this validation, gene expression data from The Cancer Genome Atlas Liver Hepatocellular Carcinoma (TCGA-LIHC) dataset were utilized, where samples were collected from different individuals and geographical locations. This approach allows assessment of the diagnostic relevance and therapeutic potential of the identified DEGs beyond the experimental glyphosate exposure model.

The TCGA data-based analysis revealed heterogeneous expression patterns across tumor and normal liver tissues. Notably, PFKFB3 and HK2 showed significantly higher expression in tumor vs. normal, as well as in the glyphosate-treated dataset. Conversely, ATF3, ENO2 and FOS exhibited significantly lower expression in tumor liver tissues vs. normal tissues, which indicates an opposite expression trend compared with the glyphosate-treated dataset. Similarly, JUNB, ALDOA, FOSB, PGK1 and G6PD showed no significant or only marginal differences between tumor and normal tissues, but these DEGs are significantly upregulated in glyphosate-exposed samples (Figure 3A).

To further evaluate the relevance of hub genes that were downregulated in glyphosate-treated hepatoma cells compared with untreated hepatoma cells, their expression patterns were analyzed in TCGA LIHC tumor vs. normal liver tissues. The analysis revealed that downregulated DEGs, except CD36, showed non-significant differential expression in tumor vs. normal liver samples, which is in contrast with their downregulation observed in glyphosate-exposed hepatoma cells (Figure 3B). Collectively, these findings indicate that although the selected hub genes were dysregulated in glyphosate-treated hepatoma cells, such significantly altered expression is not consistently observed in liver tumor vs. normal liver. This suggests that the observed transcriptional alterations primarily reflect glyphosate-driven molecular changes in vitro rather than consistent tumor associated expression signatures.

3.5. Gene Set Enrichment Analysis (GSEA) Results

To explore alterations in key biological pathways associated with glyphosate exposure, all expression data were subjected to Gene Set Enrichment Analysis (GSEA). Enrichment analysis was performed using the MSigDB hallmark and KEGG pathway databases. The GSEA results showed the five significant upregulated hallmark pathways, i.e., epithelial–mesenchymal transition (EMT), KRAS signaling, myogenesis, TNF-α signaling via NF-κB and hypoxia (Figure 4A).

Furthermore, a total of seven KEGG pathways were found to be significantly altered in the GSEA results, which include four upregulated and three downregulated pathways. The four upregulated pathways were ribosome, cardiac muscle contraction, Parkinson’s disease and Huntington’s disease. Among these, only the ribosome and cardiac muscle contraction pathways are shown in Figure 4B, as the genes of interest were not present in the other two upregulated pathways. On the other hand, significant downregulated pathways included glycerolipid metabolism, butanoate metabolism and histone metabolism, pointing to suppressed lipid metabolic activity and chromatin-related processes (Figure 4B).

Moreover, FOS, ATF3, ALDOA, PFKFB3, ENO2, HK2 and PGK1 upregulated hub genes were found to be associated with hypoxia hallmark. TNF-α signaling via NF-κB hallmark was associated with FOS, JUNB, FOSB, PFKFB3 and ATF3, whereas epithelial–mesenchymal transition was associated with the ENO2 hub gene. Similarly, PNPLA3 and MGLL downregulated hub DEGs were associated with glycerolipid metabolism (Supplementary Table S3A,B).

3.6. Biological Function and Disease Enrichment Analysis of DEGs

To explore the biological functions, molecular pathways and disease associations of the module specific DEGs identified between glyphosate-treated and untreated hepatoma cells, comprehensive functional enrichment analyses were performed using GO, KEGG pathway analysis and disease enrichment analysis. These analyses were carried out to elucidate the potential roles and implications of the key hubs in relevant biological processes, molecular functions, cellular components, signaling pathways and molecular mechanisms in association with diseases.

Enrichment analysis revealed that module 1-associated upregulated DEGs were significantly enriched in glucose metabolism-related biological processes, including glycolytic processes, gluconeogenesis, and fructose metabolic processes. Conversely, enriched cellular components included the AP1 transcription factor complex and chromatin, while enriched molecular functions comprised DNA binding transcription activator activity and D-glucose binding (Figure 5A).

Similarly, module 2-associated upregulated DEGs showed significant enrichment in biological processes, i.e., skeletal and cardiac muscle contraction and sarcomere organization, while module 3-associated upregulated DEGs were enriched in processes including response to unfolded protein, positive regulation of cell migration and protein refolding. Consistent with their involvement in muscle-related biological processes, the enriched cellular components for module 2-associated DEGs included the sarcomere and the cardiac troponin complex. Enriched molecular functions included actin binding and binding to troponin T, I and C in module 2, whereas module 3-associated DEGs were enriched in functions related to unfolded protein binding and protein folding chaperone activity (Supplementary Table S4).

To evaluate the potential disease relevance of the identified module specific DEGs, disease enrichment analysis was conducted using DisGeNET. This study revealed that upregulated DEGs of module 1 were associated with autoimmune, cardiovascular, metabolic and cerebrovascular disorders, i.e., juvenile polyarthritis, myocardial ischemia and glycogen storage disease. Upregulated DEGs of module 2 showed enrichment for cardiac and musculoskeletal conditions, such as hypertrophic cardiomyopathy and arthrogryposis. Meanwhile, upregulated DEGs of module 3 showed associations with liver diseases, autoimmune arthritis and gastrointestinal cancers (Figure 5A).

KEGG pathway enrichment analysis displayed that upregulated DEGs of module 1 were significantly enriched in carbon, fructose and mannose metabolism, glycolysis/gluconeogenesis and the HIF-1 signaling pathway. Upregulated DEGs of module 2 were enriched in pathways related to hypertrophic cardiomyopathy, motor proteins and the cytoskeleton in muscle cells. Upregulated DEGs of module 3 showed enrichment in the VEGF signaling pathway, bladder cancer, proteoglycans in cancer and the MAPK signaling pathway. Furthermore, enrichment analysis of the upregulated hub DEGs revealed significant involvement in the biosynthesis of amino acids, fructose and mannose metabolism, glycolysis/gluconeogenesis, the HIF-1 signaling pathway and carbon metabolism (

Table 2. Identified KEGG pathways associated with gene modules and hub genes.

KEGG Pathway Term	Count	Genes	FDR
Upregulated DEGs of Module 1
hsa01100:Metabolic pathways	9	PFKFB4, G6PD *, PFKFB3 *, PGK1 *, PYGM, PCK1, ALDOA *, ENO2 *, HK2 *	4.00 × 10−2
hsa01200:Carbon metabolism	5	G6PD *, PGK1 *, ALDOA *, ENO2 *, HK2 *	8.97 × 10−4
hsa00051:Fructose and mannose metabolism	4	PFKFB4, PFKFB3 *, ALDOA *, HK2 *	7.72 × 10−4
hsa00010:Glycolysis/Gluconeogenesis	5	PGK1 *, PCK1, ALDOA *, ENO2 *, HK2 *	2.02 × 10−4
hsa04066:HIF-1 signaling pathway	6	PFKFB3 *, PGK1 *, ALDOA *, ENO2 *, HK2 *, PDK1	8.01 × 10−5
Upregulated DEGs of Module 2
hsa05410:Hypertrophic cardiomyopathy	4	EDN1, TPM2, TNNT2, TNNC1	1.00 × 10−2
hsa04814:Motor proteins	7	ACTA1, TNNT1, TPM2, TNNT2, TNNC1, TNNT3, TNNI2	3.79 × 10−5
hsa04820:Cytoskeleton in muscle cells	8	ACTA1, TNNT1, TPM2, TNNT2, TNNC1, TNNT3, TNNI2, TCAP	8.35 × 10−6
Upregulated DEGs of Module 3
hsa04370:VEGF signaling pathway	3	CDC42, HSPB1, HRAS	4.70 × 10−2
hsa05219:Bladder cancer	3	HRAS, THBS1, HBEGF	3.70 × 10−2
hsa05205:Proteoglycans in cancer	4	CDC42, HRAS, THBS1, HBEGF	3.70 × 10−2
hsa04010:MAPK signaling pathway	5	CDC42, HSPA6, HSPB1, HRAS, EREG	1.80 × 10−2
Upregulated Hub DEGs
hsa01230:Biosynthesis of amino acids	3	PGK1 *, ALDOA *, ENO2 *	2.60 × 10−2
hsa00051:Fructose and mannose metabolism	3	PFKFB3 *, ALDOA *, HK2 *	6.00 × 10−3
hsa00010:Glycolysis/Gluconeogenesis	4	PGK1 *, ALDOA *, ENO2 *, HK2 *	5.37 × 10−4
hsa04066:HIF-1 signaling pathway	5	PFKFB3 *, PGK1 *, ALDOA *, ENO2 *, HK2 *	6.96 × 10−5
hsa01200:Carbon metabolism	5	G6PD *, PGK1 *, ALDOA *, ENO2 *, HK2 *	6.96 × 10−5
Downregulated DEGs of Module 1
hsa04666:Fc gamma R-mediated phagocytosis	3	VAV3, PIK3R1 *, LAT	2.10 × 10−2
hsa04660:T cell receptor signaling pathway	4	VAV3, FYN *, PIK3R1 *, LAT	4.10 × 10−4
hsa04664:Fc epsilon RI signaling pathway	4	VAV3, FYN *, PIK3R1 *, LAT	1.10 × 10−4
hsa04650:Natural killer cell mediated cytotoxicity	5	VAV3, KLRC2, FYN *, PIK3R1 *, LAT	3.66 × 10−6
Downregulated DEGs of Module 2
hsa04514:Cell adhesion molecules	4	CLDN20, CLDN14, CLDN23, CLDN2	2.46 × 10−5
hsa04670:Leukocyte transendothelial migration	4	CLDN20, CLDN14, CLDN23, CLDN2	9.69 × 10−6
hsa04530:Tight junction	5	CLDN20, CLDN14, CLDN23, CLDN2, TJP3	3.04 × 10−7
Downregulated DEGs of Module 3
hsa03320:PPAR signaling pathway	4	PLIN4, CD36 *, PLIN5, PCK2 *	1.81 × 10−4
Downregulated Hub DEGs
hsa04932:Non-alcoholic fatty liver disease	4	CEBPA *, MLXIPL *, PIK3R1 *, PPARA *	2.90 × 10−2
hsa04931:Insulin resistance	5	MLXIPL *, CD36 *, PIK3R1 *, PPARA *, PCK2 *	3.81 × 10−4

* Represents the hub genes identified through network analysis.

).

All ten upregulated hub genes, ATF3, JUNB, ALDOA, FOSB, PFKFB3, G6PD, ENO2, HK2, PGK1 and FOS, were observed to be involved in key metabolic and signaling processes. Notably, genes such as PGK1, ALDOA, ENO2, PFKFB3, HK2 and G6PD were consistently enriched across pathways related to glycolysis, carbon metabolism and the HIF-1 signaling pathway.

Gene Ontology enrichment analysis of the downregulated genes of module 1 revealed significant involvement in several biological processes, particularly related to immune signaling and intracellular communication that include intracellular signal transduction and the adaptive immune response. For cellular components, the enriched terms highlight localization to immune-related membrane structures. These include the plasma membrane, immunological synapse, cell–cell junctions and membrane rafts, which suggests that these genes are functionally active in membrane-associated immune signaling phenomena (Figure 5B).

Downregulated DEGs in module 2 revealed strong involvement in cell–cell adhesion processes, particularly those associated with tight junctions. According to the biological process terms enrichment result, the genes were found to be involved in calcium-independent cell–cell adhesion and bicellular tight junction assembly processes. Moreover, molecular function enrichment showed that the downregulated genes were significantly enriched in structural molecule activity and protein binding, further supporting their role in maintaining structural integrity of intercellular junctions.

The downregulated DEGs of module 3 were associated with lipid metabolism, energy regulation and inflammation-associated processes. Specifically, these biological processes include long-chain fatty acid transport, response to fatty acids and lipid storage. CEBPA, NR1H4 and PCK2 downregulated hub genes were enriched in biological processes like cellular response to tumor necrosis factors, notch signaling pathway and response to lipopolysaccharide, which highlight their potential roles in immune and inflammatory signaling. Cellular component enriched terms included the lipid droplet, mitochondrion, RNA polymerase II complex and receptor complex, suggesting these genes function in both metabolic and transcriptional regulation pathways. Molecular function analysis indicated enrichment in transcription cis-regulatory region binding (Supplementary Table S3).

Furthermore, NR1H4 and UCP2 downregulated DEGs were identified to be associated with fatty liver and steatohepatitis in disease enrichment analysis, which suggests their involvement in metabolic liver disorders (Figure 5B).

Pathway enrichment analysis revealed that downregulated DEGs of module 1, i.e., VAV3, FYN, PIK3R1, KLRC2, and LAT, were significantly associated with immune-related pathways. In contrast, downregulated DEGs of module 2 were significantly enriched in pathways related to tight junctions, cell adhesion molecules and leukocyte trans-endothelial migration. CLDN20, CLDN14, CLDN23, CLDN2 and TJP3 downregulated DEGs were the key contributors in the above pathways. In module 3, the downregulated DEGs showed significant enrichment in the PPAR signaling pathway, and the involved DEGs were PLIN4, PLIN5, CD36, and PCK2, all of which play key roles in lipid metabolism pathways.

KEGG pathway enrichment analysis of downregulated hub genes revealed significant associations with metabolic disorder related pathways. Non-alcoholic fatty liver disease (NAFLD) included CEBPA, MLXIPL, PIK3R1 and PPARA, while insulin resistance included MLXIPL, CD36, PIK3R1, PPARA and PCK2, suggesting these genes may contribute to liver metabolic dysregulation (Table 2).

3.7. Retrieval of Glyphosate Associated Genes

Next, a comparative analysis was done to explore whether any of the identified key genes have previously been known to be associated with glyphosate exposure. A total of 496 genes were retrieved which are known to be associated with glyphosate herbicide in humans, using the CTD (Supplementary Table S5). This analysis revealed a panel of key DEGs, which include the significant module DEGs and the MCC ranked hub genes that are strongly associated with glyphosate exposure.

The analysis indicated that GADD45B, FOS, DUSP1, PGK1, THBS1 and HSPA6 were upregulated DEGs, and PIK3R1 was a downregulated gene associated with glyphosate exposure according to the CTD (Figure 6). Out of these mapped glyphosate-associated DEGs, FOS, PGK1 and PIK3R1 genes were also among the identified hub genes.

3.8. Retrieval of Immune Related Gene Sets

To further investigate that glyphosate exposure may influence immune-related pathways, immune-associated differentially expressed genes were analyzed among the identified influential DEGs. To perform this analysis, immune-related genes were retrieved from publicly available resources (Supplementary Table S6).

Among the upregulated hub DEGs, ATF3, JUNB, ALDOA, FOSB, PFKFB3, G6PD, HK2, PGK1 and FOS were common among the immune-related genes. Additionally, the downregulated hub DEGs consisted of CEBPA, NR1H4, PCK2, FYN, PNPLA3, MGLL, PIK3R1, PPARA and CD36 and were found to be common among immune-related genes in humans.

3.9. Diagnostic Performance of Hub Genes

The ROC result showed that all analyzed upregulated hub genes, including FOSB, FOS, ENO2, ATF3, ALDOA, G6PD, PGK1, HK2, JUNB and PFKFB3, exhibited excellent diagnostic performance with AUC values of 1.000 (Figure 7A).

Similarly, the downregulated hub genes CEBPA, NR1H4, PCK2, FYN, PNPLA3, PIK3R1, PPARA, MLXIPL and CD36 showed excellent diagnostic potential with AUC values of 1.000. Notably, one downregulated gene, MGLL, exhibited a slightly lower but still high diagnostic accuracy, with an AUC value of 0.944. Overall, the ROC analysis suggests that the selected hub DEGs displayed strong diagnostic potential and may serve as potential biomarkers for glyphosate-treated hepatoma cells (Figure 7B).

4. Discussion

Liver cancer remains a major global health challenge due to its high rate of mortality, late-stage diagnosis, limited early detection tools and complex molecular heterogeneity [56,57,58]. Emerging evidence further indicates that glyphosate exposure induces distinct metabolic pathway alterations, particularly those related to metabolic reprogramming, oxidative stress and immune signaling in liver [22,23]. Consequently, identifying glyphosate exposure-associated molecular signatures and pathway perturbations is crucial for understanding glyphosate-driven alterations in liver cancer and for developing more precise biomarkers and targeted therapeutic strategies.

Based on the observed transcriptomic alterations, a comprehensive investigation of the molecular mechanisms by which glyphosate may contribute to liver injury, metabolic dysregulation and immune modulation is provided. Multiple bioinformatics approaches were applied, including raw count preparation, differentially expressed gene identification, influential gene identification, pathway enrichment analysis, immune-related gene identification and ROC analysis. Unlike traditional methods that focus on a single gene target, this approach looks at multiple targets and their connected pathways, offering a broader picture of how glyphosate affects biological systems. Although several dysregulated pathways associated with glyphosate exposure have been reported earlier, the involved panel of key genes remains poorly defined. This study aimed to identify critical genes as putative therapeutic targets and potential biomarkers specific to glyphosate-exposed hepatoma cells.

Overall, the analysis identified statistically significant panels of differentially expressed genes in glyphosate-treated vs. untreated hepatoma cells. Among these, influential genes were prioritized as hub genes based on PPI network analysis. Functional enrichment analysis further demonstrated that these hub genes were significantly associated with metabolic reprogramming, immune modulation and liver dysfunction-related pathways, which were distinctly altered in glyphosate-treated cells.

Metabolic reprogramming is a well-established hallmark of cancer [59], whereby tumor cells usually adapt their metabolic pathway to meet the increased energy demands required for rapid growth and proliferation [60]. One of the most common metabolic phenotypes of tumor cells is the Warburg effect, in which cells produce ATP via glycolysis even in the presence of oxygen, instead of relying on oxidative phosphorylation [61]. Also, it is important to understand how cancer cells integrate fructose into glucose metabolic pathways, with the growing consumption of fructose over recent decades [62,63]. This insight may support the development of therapeutic strategies that target glucose- and fructose-related metabolic pathways to limit cancer progression. The enrichment of the HIF-1 signaling pathway further supports a hypoxia-driven metabolic adaptation in glyphosate-treated hepatoma cells. It has been studied that hypoxia inducible factor (HIF)-1 signaling pathway, in association with the downstream factor BCL-2, can induce cellular autophagy which, in turn, contributes to metastasis and liver cancer [64].

Six upregulated hub genes, HK2, ALDOA, ENO2, PGK1, PFKFB3 and G6PD, exhibited their association with metabolic pathways, i.e., carbon metabolism, fructose and mannose metabolism, glycolysis/gluconeogenesis and HIF-1 signaling pathway (Table 2). Among them, Hexokinase 2 (HK2) is known to catalyze the first step of glycolysis, which links glucose uptake to downstream carbon flux and serves as a key regulator of metabolic reprogramming in hepatocellular carcinoma [65,66]. Its observed enrichment with glycolytic processes and the HIF-1 signaling pathway suggests a glycolysis-dominant biochemical state and hypoxia-induced metabolic adaptation in glyphosate-treated hepatoma cells [67]. ALDOA (fructose-bisphosphate aldolase A) catalyzes the reversible cleavage of fructose-1,6-bisphosphate into glyceraldehyde-3-phosphate and dihydroxyacetone phosphate. This is known as a key step in glycolysis that can contribute to cancer initiation and progression when ALDOA is overexpressed [68,69]. Glucose-6-phosphate dehydrogenase (G6PD), which is a rate-limiting enzyme of the pentose phosphate pathway, is frequently overexpressed in cancers and is associated with poor prognosis, immune cell infiltration and regulation of proliferation and apoptosis processes [70,71]. G6PD generates NADPH, a crucial reducing agent for biosynthesis and protection against oxidative stress. Altered NADPH homeostasis may amplify oxidative stress, triggering inflammation through pathways such as NF-κB and contributing to metabolic disorders such as NAFLD [72,73]. In addition, PFKFB3 has emerged as an independent prognostic factor in hepatocellular carcinoma with elevated expression, which further relates to poor overall survival, specifically under hypoxic conditions [74]. From a biochemical perspective, PFKFB3 drives glycolysis by elevating fructose-2,6-bisphosphate levels, which in turn stimulates phosphofructokinase-1 activity [75]. This mechanism strengthens glycolytic flux and enables cells to maintain energy production under hypoxic stress. In addition to its glycolytic activity, Enolase (ENO2) contributes to tumor cell proliferation, invasion and migration [76,77,78]. Similarly, other hub genes such as phosphoglycerate kinase 1 (PGK1) usually catalyzes the transfer of a high-energy phosphate group from 1,3-bisphosphoglycerate to ADP, which generates ATP through substrate-level phosphorylation. Under hypoxic conditions, when mitochondrial oxidative phosphorylation is not sufficient, this reaction becomes particularly important for sustaining cellular ATP levels and supporting metabolic adaptation associated with tumor progression [79]. Notably, the comparative analysis further suggests that PGK1 is linked not only to tumor associated metabolic processes but also to glyphosate exposure and immune-related pathways. Moreover, ALDOA, PFKFB3, G6PD, HK2 and PGK1 were identified as overlapping genes within the immune-related genes (Supplementary Table S6), supporting their potential association with immune modulation in glyphosate exposed to hepatoma.

Further, GSEA results highlight the involvement of major crucial pathways that include epithelial–mesenchymal transition (EMT), KRAS signaling, myogenesis, TNF-α/NF-κB signaling and hypoxia, which may collectively reflect glyphosate-associated tumor progression, inflammation and metabolic stress responses. Importantly, hub genes such as FOS, FOSB, JUNB, ATF3, ALDOA, PFKFB3, ENO2, HK2 and PGK1 were present within these enriched pathways. This observation further supports the fact that our results were not influenced by focusing exclusively on modules or hub genes only. Instead, the GSEA results prove the robustness of these findings and highlight the central role of these identified hub genes in connecting metabolic reprogramming, hypoxia, epithelial–mesenchymal transition (EMT) and inflammatory signaling in glyphosate-treated hepatoma cells.

The dataset used did not include normal liver cell lines, which made it challenging to directly compare glyphosate induced transcriptional changes in a cancerous liver cell line with a non-cancerous cell line. To investigate whether the in silico identified hubs reflect patterns that might be relevant in human liver tissue, their expression was examined in the TCGA-LIHC dataset, comparing tumor and normal samples. Therefore, a hub gene showing similar trends in both settings may represent general features of liver cancer, while differences in expression could suggest transcriptional changes specifically associated with glyphosate exposure that are not typically present in tumors. This method suggests insight into the possible relevance of these hub genes in human liver tissue, despite the difference in biological context between the two comparisons.

Except for HK2 and PFKFB3, the remaining hubs were either not significantly altered or showed significant opposite expression trends upon TCGA-LIHC validation. Notably, ATF3, ENO2 and FOS exhibited opposite expression patterns in glyphosate-treated hepatoma cells compared with human liver tumors from TCGA-LIHC, highlighting a potential glyphosate-driven effect on these regulatory hubs. For instance, studies have shown that Activating transcription factor 3 (ATF3) is usually downregulated in hepatocellular carcinoma, where low levels are linked to tumor progression, but its overexpression can cause oxidative stress and fat accumulation and is associated with nonalcoholic fatty liver disease (NAFLD) [80,81,82]. Also, ATF3 activation has been found to be associated with increased reactive oxygen species (ROS) production, resulting in suppression of antioxidant defense pathways and dysregulation of lipid-handling genes [83]. These observations suggest that glyphosate may trigger a stress-induced upregulation of ATF3 in hepatoma cells, activating pathways distinct from those in untreated liver tumors.

Similarly, ENO2 catalyzes the reversible conversion of 2-phosphoglycerate to phosphoenolpyruvate (PEP), a critical intermediate between glycolysis and gluconeogenesis [84]. PEP serves as the immediate substrate for pyruvate kinase, which further generates pyruvate and ATP via substrate-level phosphorylation. The resulting pyruvate can either enter the tricarboxylic acid (TCA) cycle through conversion to acetyl-CoA or be reduced to lactate under hypoxic conditions [85,86]. Furthermore, glycolytic intermediates derived from ENO2 activity can participate in amino acid biosynthesis pathways, supporting anabolic requirements and metabolic alteration [87]. Collectively, ENO2 upregulation reflects coordinated metabolic reprogramming involving glycolysis, gluconeogenesis, carbon metabolism and hypoxia in glyphosate-treated hepatoma cells. Also, high FOS expression is associated with tumor progression, inhibition of apoptosis, and poor prognosis in HCC [88]. Taken together, this observation suggests that glyphosate exposure may alter cellular pathways distinct from those typically dysregulated in liver cancer.

The identified upregulated hubs, ATF3, JUNB, ALDOA, FOSB, PFKFB3, G6PD, ENO2, HK2, FOS and PGK1, not only highlight their importance as potential therapeutic targets but also suggest that they may serve as diagnostic biomarkers of glyphosate exposure because of their discriminatory potential (AUC = 1.0) for distinguishing glyphosate-exposed hepatoma cells.

Furthermore, the analysis revealed that several immune- and metabolism-related pathways were significantly associated with the downregulated hub genes. In this study, PIK3R1 and FYN downregulated hub genes were enriched in immune signaling pathways that include T-cell receptor, Fc epsilon RI and natural killer cell mediated cytotoxicity, which highlights altered tumor–immune interactions. In parallel, CD36 and PCK2 downregulated hubs were associated with the PPAR signaling pathway, while CEBPA, MLXIPL, PIK3R1, PPARA, CD36 and PCK2 were enriched in non-alcoholic fatty liver disease and insulin resistance pathways. These findings collectively highlight a promising link between metabolic dysregulation and immune modulation.

Although several studies and TCGA-based observations in this study showed no significant downregulation of PIK3R1 in liver tumors, a significant decrease in its expression in the used dataset was observed [89]. Furthermore, CTD analysis indicated that PIK3R1 is linked to glyphosate exposure, suggesting that its downregulation may be exposure-driven rather than solely tumor-related. Also, reduced PIK3R1 expression was found to be associated with poor prognosis in hepatocellular carcinoma based on Kaplan–Meier analysis [89]. Dysregulated phosphoinositide 3-kinase (PI3K) activity is recognized as a hallmark of cancer, where PIK3R1 and PIK3R2 are the main regulatory subunits of PI3K [90,91]. The most notable distinction between PIK3R1 and PIK3R2 genes emerges in cancer biology, where PIK3R1 functions as a tumor suppressor and PIK3R2 promotes tumor progression [92]. PIK3R1 plays a central role in insulin signaling by mediating downstream activation of AKT, which regulates glucose uptake, glycogen synthesis and suppression of gluconeogenic genes. Hence, altered expression of PIK3R1 can modulate insulin signaling and may contribute to metabolic disturbances that also include insulin resistance [93]. Insulin resistance is a key driver of NAFLD because, when insulin signaling in the liver is impaired, this further results in glucose and lipid metabolism disruption, resulting in fat accumulation in hepatocytes [94]. Also, studies have suggested that PI3K activity is strongly maintained in T and B lymphocytes and that alteration in the PI3K-associated pathway can cause primary immunodeficiencies in humans [95]. Moreover, studies have shown that overexpression of FYN inhibits proliferation, migration and epithelial-to-mesenchymal transition (EMT) through downregulating the AKT axis in various cancers [96,97]. Additionally, reduced expression of FYN is likely to impair T-cell receptor (TCR)-mediated signaling. FYN low expression may lead to decreased T-cell activation, proliferation and cytokine production, which ultimately may result in a weakened cell-mediated immune response, thereby compromising immune surveillance [98,99,100]. In this study, the observed downregulation of PIK3R1 and FYN indicates a potential disruption of tumor-suppressive regulation, which may facilitate oncogenic signaling and immune modulation in glyphosate-treated hepatoma cells.

CEBPA is a known key transcription factor that normally regulates hepatocyte differentiation, glucolipid regulation and cell-cycle-control-related processes [101]. Additionally, downregulated CEBPA has been correlated with tumor-promoting immunity, demonstrating a functional role in altering the tumor immune microenvironment [102,103]. Downregulation of CEBPA has been associated with progression of NAFLD. In both human NAFLD livers and diet-induced NAFLD mouse models, reduced hepatic CEBPA expression correlates with severity of steatosis and fibrosis, indicating that loss of CEBPA may contribute to altered lipid metabolism [101]. Studies have report that reduced PPARα is associated with worse clinicopathological features and poorer prognosis [104]. Studies indicated that hepatocyte PPARα is essential for preventing NAFLD [105]. Utilizing a mouse model, it has further been demonstrated that loss of PPARα promotes hepatocarcinogenesis and activates pro-inflammatory pathways such as NF-κB, which further supports its tumor-suppressive role in the liver [106]. MLXIPL (carbohydrate response element–binding protein, ChREBP), a glucose-sensitive transcription factor, is reported to regulate key hepatic metabolic pathways such as glycolysis and lipogenesis [105]. Dysregulation of such metabolic regulators has been shown to influence CD8⁺ T cell and myeloid cell recruitment, potentially leading to altered antitumor immunity and immune differentiation within the tumor microenvironment [106].

TCGA LIHC data show significant upregulation of CD36 in HCC, also supported by various studies, consistent with its role in fatty acid uptake and metabolic reprogramming [107,108]. In contrast, CD36 was downregulated in glyphosate-treated hepatoma cells, potentially suggesting glyphosate-induced suppression of PPAR signaling (Table 2) and consequent alterations in hepatocyte metabolic function. CD36 is a well-established transcriptional target of PPARα and plays a critical role in facilitating long-chain fatty acid uptake into hepatocytes [109]. Phosphoenolpyruvate carboxykinase 2 (PCK2) has been reported to function as a tumor suppressor in hepatocellular carcinoma [110]. Downregulation of PCK2 in liver cancer reflects a shift away from mitochondrial metabolic flexibility toward glycolytic reprogramming and also predicts poor prognosis and is related to the immune invasion in hepatocellular carcinoma [111]. Therefore, its downregulation may contribute to liver cancer progression by disrupting normal metabolic programming and hampering PPAR metabolic pathways [112]. Nuclear Receptor Subfamily 1 Group H Member 4 (NR1H4) encodes the nuclear receptor Farnesoid X receptor (FXR), a key regulator of hepatocyte glucose and lipid metabolism, and inhibits inflammation responses [113,114]. Decreased expression of NR1H4 (FXR) has been consistently associated with lipid accumulation and metabolic dysregulation in NAFLD [115]. Therefore, downregulation of NR1H4 observed in the present study may reflect disruption of metabolic and immune regulatory pathways in glyphosate-exposed hepatic carcinoma cells.

Furthermore, pathway enrichment analysis revealed that the non-alcoholic fatty liver disease (NAFLD) pathway might be influenced by CEBPA, MLXIPL, PIK3R1, NR1H4 and PPARA downregulated hub genes in glyphosate-treated liver cancer cells. These genes are crucial regulators of hepatic metabolism and inflammatory control, and their significant downregulation indicates metabolic dysregulation, as discussed above.

Additionally, downregulated glycerolipid metabolism was identified by GSEA, which indicated suppressed lipid metabolic activity in glyphosate treated hepatoma cells. The downregulated hub genes Patatin-like phospholipase domain-containing 3 (PNPLA3) and Monoacylglycerol Lipase (MGLL) were found to be associated with glycerolipid metabolism, further supporting glyphosate-induced disruption of lipid metabolism. PNPLA3 encodes a lipid droplet-associated protein with acyltransferase activity, thereby regulating triglyceride remodeling and storage [116]. In contrast, MGLL catalyzes the hydrolysis of monoacylglycerols into free fatty acids and glycerol, representing the final step of triglyceride breakdown [117]. Dysregulation of these enzymes may therefore disturb the balance between lipid storage and lipolysis, contributing to altered glycerolipid turnover and metabolic reprogramming in hepatocytes. Reduced PNPLA3 and MGLL expressions indicate altered hepatic lipid processing, which have been linked to non-alcoholic fatty liver-associated metabolic dysregulation [118,119,120].

CEBPA, PCK2, FYN, PNPLA3, PIK3R1, PPARA, NR1H4, MLXIPL and CD36 were observed to exhibit perfect discriminatory ability with an AUC of 1.0, whereas MGLL showed strong diagnostic performance with an AUC of 0.944. These results suggest that these genes may serve as potential diagnostic biomarkers in glyphosate-associated hepatocellular carcinoma.

Being an in silico investigation, this study has inherent limitations, First, experimental validation of the identified hub genes was not performed, and normal hepatocyte models were not included for direct comparison. As an in silico investigation, the analysis relied on RNA-seq data derived from in vitro hepatoma cell lines, which may not fully recapitulate the biological complexity of normal liver tissue.

In summary, the transcriptomic and systems biology analyses reveal that glyphosate exposure induces distinct molecular alterations in hepatoma cells in comparison to untreated hepatoma cells. These alterations are characterized by the upregulation of hub genes (ATF3, JUNB, ALDOA, FOSB, PFKFB3, G6PD, ENO2, HK2, FOS and PGK1) associated with glucose metabolism, immune signaling and cellular stress. In contrast, downregulated hub genes were involved in lipid metabolism, immune regulation and non-alcoholic fatty liver disease pathways (PIK3R1, FYN, CEBPA, MLXIPL, PPARA, CD36, PCK2, PNPLA3, NR1H4 and MGLL). These hub genes show promising discriminatory capability as diagnostic biomarkers and may be suggested as potential therapeutic targets for glyphosate-exposed hepatoma cells.

Pathway analyses indicate that glyphosate exposure induces widespread metabolic reprogramming in hepatoma cells. Enriched pathways include glycolysis and gluconeogenesis, fructose and mannose metabolism, carbon metabolism, NAFLD, lipid metabolism, immune-related pathways, and HIF-1 signaling, which reflects adaptation to hypoxic conditions. These results suggest that glyphosate triggers coordinated changes in energy production, biosynthetic processes and hypoxia-driven responses, which together may support rapid cell proliferation and survival under stress in glyphosate-exposed hepatoma cells. Collectively, these findings provide mechanistic insights into glyphosate exposure-associated molecular alterations relevant to liver cancer and highlight the potential of the identified hub genes as candidate biomarkers and therapeutic targets for further investigation. This study is limited by the lack of experimental validation of the hub genes and absence of normal hepatocyte models. Future studies including primary hepatocytes or immortalized normal liver cell lines are required. Multi-omics integration and functional assays will further clarify the role of glyphosate-induced metabolic and immune dysregulation in liver cancer-related molecular mechanisms.