Bioinformatics‐based discovery of the involvement of PSAT1 in mediating the anti‐lung adenocarcinoma activity of triptolide

Abstract

Background

Triptolide (TP) exhibits various pharmacological activities. Our previous studies have confirmed the efficacy of TP against lung adenocarcinoma (LUAD). However, the potent pharmacological activity of TP is underpinned by its complex mechanisms. Exploring its potential mechanisms is of great value for promoting the clinical application of TP and extending its clinical use.

Methods

Differentially expressed genes (DEGs) associated with LUAD were analyzed and acquired from the TCGA database, while DEGs related to TP were obtained through RNA sequencing. Hub genes were identified through LASSO and random forest models. The efficacy of TP against LUAD was validated using tumor‐bearing mouse models and A549 cells. The validation of hub genes was conducted using RT‐qPCR. The regulatory effect of hub genes on TP efficacy was validated through overexpression cell models. Furthermore, the potential mechanisms by which TP improves gemcitabine (GEM) resistance were explored using a GEM‐resistant cell line in combination with the overexpression model.

Results

This study validated the therapeutic effect of TP against LUAD in vivo and in vitro. Bioinformatics revealed that the mechanism of TP's effect against LUAD might be associated with amino acid‐related biological processes. Five hub genes were screened and identified by combining bioinformatics methods and experiments. The overexpression model validated that PSAT1 plays an effective role in the efficacy of TP and in alleviating GEM resistance.

Conclusion

This study preliminarily demonstrated that the anti‐LUAD effect of TP was associated with the PSAT1‐regulated serine biosynthesis pathway, and that TP effectively improves GEM resistance by inhibiting PSAT1 expression.

Transcriptome sequencing was performed on tumor tissues from triptolide‐treated tumor‐bearing mice. The sequencing results were integrated with data from public databases for comprehensive analysis, and potential regulatory targets of triptolide were identified. Based on these targets, clinical application scenarios for triptolide in improving gemcitabine resistance were further constructed.

1. INTRODUCTION

Triptolide (TP), one of the major active components of the traditional Chinese medicine Tripterygium wilfordii Hook f., exhibits potent pharmacological activities in treating tumors, rheumatoid arthritis, and other diseases. ¹ , ² , ³ In the field of oncology, TP has been demonstrated to exert strong anti‐tumor effects against multiple cancers, including pancreatic cancer, ⁴ lung adenocarcinoma (LUAD), ⁵ chondrosarcoma, ⁶ and oral squamous cell carcinoma. ⁷ Mechanistically, TP has been shown to regulate many pharmacological processes in tumor cells, such as glycolysis, mitochondrial dysfunction, autophagy, and extracellular matrix activation. ⁴ , ⁸ Specifically, in LUAD treatment, TP effectively inhibits tumor growth in xenograft mice, suppresses cancer cell proliferation and migration, and promotes cell apoptosis. ⁵ The potent pharmacological activity of TP is underpinned by its complex mechanisms. Exploring its potential mechanisms from multiple perspectives using bioinformatics techniques is of significant value for deciphering the anti‐LUAD mechanisms of TP and establishing its clinical application scenarios.

Bioinformatics techniques, as an interdisciplinary field integrating biology, computer science, and other disciplines, enable data‐driven biological analysis, facilitating in‐depth exploration of underlying pathogenic mechanisms of diseases and potential therapeutic targets. For instance, by leveraging databases such as The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) in combination with various computational models for disease mechanism research, these techniques can be used not only to investigate the pathogenesis of single diseases but also to explore shared pathogenic mechanisms across multiple diseases in the context of clinical backgrounds. ⁹ Beyond identifying potential research directions, bioinformatics techniques can significantly reduce experimental costs. In the realm of drug mechanism discovery, they also aid in predicting the potential action mechanisms of drugs and their regulatory targets. By integrating disease sequencing data from existing databases and drug sequencing data, it is possible to construct association networks between drug targets and diseases, enabling the elucidation of complex drug action mechanisms. However, bioinformatics has certain limitations: predictive results need to be validated through molecular biology experiments to effectively enhance their reliability.

Based on this, this study constructed an A549 tumor‐bearing mouse model, obtained transcriptomic data of tumor tissues under TP intervention, integrated with the TCGA database, and predicted and identified five potential hub genes associated with TP efficacy through machine learning models. Combining the results of functional enrichment analysis, PSAT1 was selected as the object for further study. Considering the regulatory characteristics of PSAT1 in tumor drug resistance, this study used A549 cells and GEM‐resistant cells to validate the therapeutic effects of TP in both models. ¹⁰ , ¹¹ Additionally, by constructing overexpression models, the regulatory role of PSAT1 in TP‐mediated anti‐tumor activity was evaluated (Figure 1). This research provides valuable insights for deeply analyzing the mechanism by which TP inhibits LUAD and clarifying the clinical application scenarios of TP and its formulations.

FIGURE 1.

Flow chart of the research design.

2. MATERIALS AND METHODS

2.1. Data collection and differential gene acquisition

At the endpoint of the experiment, tumor tissues were isolated, and three samples from each group (control group and TP group) were collected for transcriptomic analysis. Total RNA was extracted from the tumor tissue samples using an RNA Isolation Kit (Majorivd) and purified with the RNAClean XP Kit (Beckman Coulter) and RNase‐Free DNase Set (QIAGEN). The integrity of the RNA was assessed using an Agilent 2100 Bioanalyzer, and the total RNA quantity and purity were measured with a Qubit 2.0 Fluorometer and NanoDrop ND‐2000 Spectrophotometer. The purified total RNA was then used for cDNA library construction and sequenced on the Illumina NovaSeq 6000 sequencing platform. Differential gene analysis between samples was performed using edgeR, calculating the fold‐change based on FPKM values. A significance threshold for identifying differentially expressed genes (DEGs) was set at p < 0.05 and |log₂ fold change (FC)| ≥1 based on the analysis results. The LUAD gene expression matrix and associated clinical data were obtained from the The Cancer Genome Atlas (https://portal.gdc.cancer.gov/) database, which included 59 normal samples and 541 LUAD samples. DEGs were identified using the edgeR package, with a significance threshold of |log₂ FC| ≥2 and p < 0.05.

2.2. Functional enrichment analysis

Biological functions and pathways associated with DEGs were annotated via Kyoto Encyclopedia of Genes and Genomes (KEGG) and Gene Ontology (GO) enrichment analyses using the clusterProfiler package in R. ¹² Gene Set Enrichment Analysis (GSEA) was also performed using the same package, with a p‐value cutoff set at 0.05. All results were visualized in R using the ggplot2 package. ¹³

2.3. PPI network construction

The construction of the protein–protein interaction network (PPI) was completed using the STRING database (https://cn.string‐db.org/) and Cytoscape (3.9.1). ¹⁴ The DEGs after screening were input into STRING, with the minimum required interaction score set to medium confidence (0.400). The results obtained were then imported into Cytoscape for visualization.

2.4. Machine learning for hub gene identification

The intersection of DEGs from the TCGA dataset (LUAD vs. Normal) and DEGs obtained from transcriptomics (TP vs. Control) was performed to screen and obtain a gene set with inverse expression patterns in both datasets as the candidate hub gene set. Least absolute shrinkage and selection operator (LASSO) and random forest (RF) methods were further utilized to predict potential hub genes for subsequent analysis.

2.5. Tumor‐bearing mouse model construction

An A549 tumor‐bearing mouse model was established to verify the anti‐tumor efficacy of TP and explore its potential mechanism of action. Female Balb/c Nude mice (4–5 weeks) were purchased from Beijing Vital River Laboratory Animal Technology Co., Ltd. (License No. SCXK (Jing)2021–0006). Animal experiments were conducted in strict accordance with the Guide for the Care and Use of Laboratory Animals. The experimental protocol received approval from the Research Ethics Committee of the Institute of Basic Theory of Chinese Medicine, China Academy of Chinese Medical Sciences (IBTCMCACMS21‐2403‐16). A549 cells in the exponential growth phase were subcutaneously inoculated into the right axilla of Balb/c Nude mice at a concentration of 5 × 10⁶ cells/mL per mouse. When the mean tumor volume reached approximately 100 mm³, the mice were randomly divided into a control group (n = 6) and a TP group (n = 6). The control group received daily injections of an equal volume of vehicle, while the TP group was intraperitoneally injected with 500 μg/kg TP solution. Tumor volume changes were recorded every other day using the formula: (length × width²)/2 (mm³). After the last administration, the mice were euthanized, and tumor tissues were dissected for subsequent analysis.

2.6. RNA extraction and RT‐qPCR experiments

Tumor tissue homogenate was used to extract RNA using the RNAsimple Total RNA Kit (Tiangen Biotech). RNA concentration was determined with a NanoDrop2000 spectrophotometer, followed by reverse transcription to generate cDNA using the First‐Strand Synthesis Master Mix (Lablead Biotech). RT‐qPCR analysis was performed using SYBR Green PCR Fast mixture (Lablead Biotech). Gene expression levels were normalized by GAPDH, and all primer sequences are listed in Table 1.

TABLE 1.

Primers used in qPCR.

Gene	FORWARD	REVERSE
SLC25A10	5′‐TCCCTGACTCGGTTCGCCATC‐3′	5′‐CGGAGCCCAGCAACACCTTC‐3′
PSAT1	5′‐GTCAGCTAAGGCCGCAGAAGAAG‐3′	5′‐AGGAGGCATCTGGGTTGAGGTTC‐3′
GPT2	5′‐CGCCTCCTTCCACTCCACCTC‐3′	5′‐CGCACCGACAGCAGCTTCAC‐3′
FMO2	5′‐CCGCTCACCTGGACAAGTCAAC‐3′	5′‐TCTCTGAGGGCAGGCTACACAAG‐3′
SYT12	5′‐CCGACTCCCTGAACTCCATCTCC‐3′	5′‐ACTCCATGCTCACCTCCACCTG‐3′

2.7. Cell culture and overexpression model construction

The human LUAD cell line (A549 cells) was provided by Shanghai FuHeng Biotechnology Co., Ltd. A549 cells were cultured in complete RPMI 1640 medium supplemented with 10% fetal bovine serum, and maintained at 37℃ in a humidified incubator with 5% CO₂. The gemcitabine (GEM)‐resistant A549 cell line (A549/GR) was obtained from Shanghai Meixuan Biotechnology Co., Ltd. (The 48 h IC50 value of GEM in wild‐type cells (A549) was 35.71 μmol/L, while that in A549/GR cells was 499.43 μmol/L, with a resistance index of 14.0). The cells were cultured in a complete medium consisting of 90% RPMI 1640 and 10% fetal bovine serum, and maintained at 37℃ in a humidified incubator with 5% CO₂.

The PSAT1 plasmid was designed and synthesized by Tsingke Biotechnology Co., Ltd. The protocol for establishing the cell overexpression model followed the manufacturer's instructions. In brief, cells were seeded into a 12‐well plate. Once the cell confluence reached approximately 80%, transfection was performed using LabFect Lipofect5000 transfection reagent (Lablead Biotech). Specifically, 100 μL of Trans Buffer and 1.5 μL (1 μg/μL) of the plasmid were thoroughly mixed in a 1.5 mL centrifuge tube. Subsequently, 2.4 μL of the transfection reagent was added to the mixture and mixed gently. After allowing the resultant mixture to stand at room temperature for 15 min to form transfection complexes, the mixture was slowly pipetted into the wells of the plate. The plate was then gently swirled to ensure uniform distribution of the transfection mixture. The transfection efficiency was evaluated 24 h post‐transfection.

2.8. Cell viability assay

The cell viability experiment was detected using a CCK‐8 kit (Dojindo Laboratories). Briefly, 2 × 10³ cells were cultured in a 96‐well plate. TP was diluted to different concentrations using 1640 medium. After 24 h of culture, the medium was replaced. Then, 10 μL of CCK‐8 reagent was added to each well, and the plate was placed in an incubator. The absorbance was detected at 450 nm using a microplate reader to calculate the cell viability. The experiments were conducted with at least three replicates, and data analysis was performed using GraphPad Prism software.

2.9. Wound healing experiments

A total of 2 × 10⁵ A549 cells were seeded into a 12‐well plate. Once the cells reached confluence, a quick scratch was made using a 10 μL pipette tip to create a scratch area. The experimental groups were treated with culture media containing TP and GEM (A549: TP 60 nmol/L; A549/GR: TP 60 nmol/L, GEM 60 μmol/L), and the scratch healing was observed and recorded at 0 h, 24 h, and 48 h.

2.10. Immunohistochemistry experiments

For immunohistochemistry, paraffin sections were dewaxed and rehydrated, followed by heated antigen retrieval. Endogenous peroxidase was blocked with 3% hydrogen peroxide, and non‐specific binding was blocked using 3% BSA within tissue areas. Sections were incubated with PSAT1 primary antibody (Proteintech) overnight at 4℃, washed, and then incubated with secondary antibody. DAB staining was performed for visualization, followed by hematoxylin counterstaining. Sections were dehydrated, cleared, and imaged under a microscope. PSAT1‐positive areas were quantified using ImageJ software.

2.11. Statistical analysis

Data analysis was conducted with GraphPad Prism 7 along with relevant R packages. The experimental results were expressed as mean ± standard deviation. Comparisons between two groups were performed with unpaired Student's t tests, whereas one‐way ANOVA was utilized to compare means among three or more groups. A p < 0.05 was regarded as statistically significant.

3. RESULTS

3.1. Validation of the therapeutic efficacy of TP and analysis of transcriptomic results

In the A549 tumor‐bearing mouse model, the in vivo efficacy of TP was verified. As shown in Figure 2A, with increasing administration time, the tumor volume in the control group showed a significant upward trend over time, while tumor growth in TP‐treated mice was effectively suppressed. Pathological analysis revealed vacuolation in tumor cytoplasm, obvious nuclear heterogeneity, and visible cell necrosis in tissues under TP treatment (Figure 2B). These results indicate that TP exhibits favorable therapeutic effects against LUAD models.

FIGURE 2.

Validation of the therapeutic efficacy of TP and analysis of transcriptomic results. (A) Tumor volume change curve and images of tumor tissues (n = 6). (B) HE staining results of mouse tumor tissues. (C) Volcano plot of transcriptomic analysis results of tumor tissues (TP vs. Control). (D) GO analysis results of DEGs. (E) KEGG analysis results of DEGs. (F) GSEA analysis results.

To further explore the potential mechanism of TP in the treatment of LUAD, transcriptomics was used to detect gene‐level changes in tumor tissues under TP treatment. A total of 15 706 genes were identified. With p < 0.05 and |log₂ fold change (FC)| ≥1 as the screening criteria, a total of 1002 DEGs were obtained in the TP vs. Control group. Among them, 210 genes were upregulated and 792 genes were downregulated (Figure 2C).

The DEGs in the GO analysis revealed enrichment in the Biological Process (BP) category, particularly in the fatty acid metabolic process and small molecule catabolic process. In the Cellular Component (CC) category, they were primarily enriched in the collagen‐containing extracellular matrix, and in the Molecular Function (MF) category, they were mainly enriched in extracellular matrix structural constituents and cargo receptor activity (Figure 2D). The KEGG analysis indicated that DEGs were predominantly enriched in pathways related to fatty acid metabolism, biosynthesis of amino acids, carbon metabolism, and biosynthesis of unsaturated fatty acids (Figure 2E). The GSEA analysis results demonstrated that TP treatment affects pathways associated with biosynthesis of amino acids, carbon metabolism, fatty acid metabolism, glycine, serine and threonine metabolism, glycolysis/gluconeogenesis, and pyruvate metabolism (Figure 2F).

3.2. Common gene screening and PPI network construction

To identify potential therapeutic targets for TP (treatment purpose) in combination with clinical samples, gene expression profiles of LUAD samples were retrieved from the TCGA database. A total of 20 565 genes were identified in the LUAD dataset. Using |log₂ FC| ≥2 and p < 0.05 as the criteria, 2487 DEGs were obtained, including 2037 upregulated and 450 downregulated genes (Figure 3A). KEGG pathway analysis showed that DEGs were significantly enriched in pathways such as Neutrophil Extracellular Trap Formation, Cell Cycle, and Transcriptional Misregulation in Cancer (Figure 3B). GO enrichment analysis indicated that in the Molecular Function (MF) category, DEGs were primarily associated with extracellular matrix organization, regulation of mitotic nuclear division, and immunoglobulin‐mediated immune response. In the Cellular Component (CC) category, they were observed in immunoglobulin complex, collagen‐containing extracellular matrix, and nucleosome. In the MF category, they were mainly enriched in serine‐type endopeptidase activity, serine hydrolase activity, and extracellular matrix structural constituent (Figure 3C).

FIGURE 3.

Common gene screening and PPI network construction. (A) Volcano plot of the LUAD dataset. (B) KEGG analysis of DEGs in the LUAD dataset. (C) GO analysis of DEGs in the LUAD dataset. (D) Volcano plot combining A549 tumor‐bearing mice (TP vs. Control) and TCGA database analysis (LUAD vs. Normal), identifying 81 common DEGs. (E) PPI network of the 81 common DEGs. (F) KEGG analysis results of the 81 common DEGs.

Combining transcriptomic results from A549 tumor‐bearing mice (TP vs. Control) and TCGA database analysis (LUAD vs. Normal), we identified candidate hub genes for TP efficacy by selecting DEGs significantly regulated by TP and showing opposite expression trends in LUAD versus Normal (Figure 3D). A total of 81 candidate hub genes were obtained, including 80 genes upregulated in LUAD versus Normal but downregulated in TP versus Control, and 1 gene downregulated in LUAD versus Normal but upregulated in TP versus Control. PPI network analysis of the candidate hub gene set is shown in Figure 3E. KEGG pathway analysis of the 81 genes revealed significant enrichment in pathways such as Biosynthesis of amino acids, Phenylalanine Metabolism, Carbon Metabolism, Glycolysis/Gluconeogenesis, Histidine Metabolism, and Glycine, Serine and Threonine Metabolism (Figure 3F).

3.3. Screening and validation of hub genes associated with TP

In order to further screen and obtain the core genes related to the efficacy of TP, based on the 81 candidate hub genes obtained previously, two machine learning algorithms (LASSO and RF) were selected to predict the potential hub genes. Using the ten‐fold cross‐validation procedure, LASSO regression identified a total of 17 hub genes that are potentially related to the efficacy of TP (Figure 4A). Similarly, 20 hub genes (importance greater than 1) were screened out by the random forest method, and the results are shown in Figure 4B,C. By integrating the two machine learning algorithms, a total of 5 hub genes were obtained, namely SYT12, FMO2, GPT2, PSAT1, and SLC25A10 (Figure 4D). In the tumor‐bearing mouse model, the expression trends of the five hub genes were verified using the RT‐qPCR method, and the results were shown in Figure 4E. Compared with the control group, the expression trends of SLC25A10, PSAT1, GPT2, FMO2, and SYT12 were consistent with the results of transcriptomics. TP effectively inhibited their expression.

FIGURE 4.

Screening and validation of hub genes associated with TP. (A) Coefficient distribution plots for the log(lambda) sequence were generated using the LASSO regression model with 10‐fold cross‐validation and the lambda 1 standard error (lambda 1SE) criterion. (B) Twenty genes with an importance score exceeding 1 were selected through the random forest algorithm. (C) Evaluation of the random forest model's prediction accuracy. (D) Venn diagram depicting the overlap of genes identified by the two machine learning methods. (E) Validation of hub gene expression via RT‐qPCR (n = 4). **p < 0.01, ***p < 0.001.

3.4. PSAT1 expression detection and its correlation analysis with clinical characteristics

Through the analysis of the TIMER database, it was found that PSAT1 was significantly upregulated in a variety of tumors such as LUAD, bladder urothelial carcinoma (BLCA), colon adenocarcinoma (COAD), esophageal carcinoma (ESCA), head and neck squamous cell carcinoma (HNSC), lung squamous cell carcinoma (LUSC), rectal adenocarcinoma (READ), prostate adenocarcinoma (PRAD), uterine corpus endometrial carcinoma (UCEC), etc. (Figure 5A). Similarly, in the TCGA dataset, the mRNA expression of PSAT1 was also significantly upregulated compared with the Normal group (Figure 5B,C). Based on the analysis of the expression level of PSAT1 and the corresponding clinical characteristics from the TCGA database, it was found that there was no significant association between the expression level of PSAT1 and Age or Gender (Figure 5D,E), while obvious differences were observed between the clinical stage and lymph node metastasis (Figure 5F,G). In the tumor‐bearing mouse model, immunohistochemical detection revealed that TP could effectively inhibit the protein expression of PSAT1 (Figure 5H,I), and the results were consistent with the trend of mRNA expression. Similarly, TP also effectively inhibited the expression of PSAT1 in vitro (Figure 5J,K).

FIGURE 5.

Detection of PSAT1 expression and its correlation analysis with clinical characteristics. (A) Pan‐cancer analysis of the expression of PSAT1 in the TIMER database. (B) The expression of PSAT1 mRNA in LUAD tissues in the TCGA database. (C) The expression of PSAT1 mRNA in paired tumor tissues and adjacent normal tissues in the TCGA database. (D–G) The clinical correlations between the expression of PSAT1 and age, gender, stage, and lymph node metastasis. (H) The immunohistochemical results of PSAT1 in the tumor tissues of tumor‐bearing mice. (I) The results of quantitative analysis of immunohistochemistry (n = 3). (J) Representative PSAT1 immunofluorescence results in A549 cells. (K) PSAT1 relative fluorescence intensity (n = 5). *p < 0.05, **p < 0.01, ***p < 0.001.

3.5. Evaluating the anti‐tumor efficacy of TP in PSAT1‐overexpressing cells

To verify the regulatory role of PSAT1 in the anti‐LUAD effects of TP, a plasmid was used to construct PSAT1‐overexpressing A549 cells (Figure 6A). Cell viability assessments showed that TP inhibited A549 cell activity in a dose‐dependent manner, with more pronounced anti‐tumor effects observed at 48 h (IC₅₀: 88.41 nmol/L) over time (Figure 6B,C). Wound healing assays demonstrated that TP effectively suppressed the migration ability of A549 cells, especially at 48 h. Upon PSAT1 overexpression, cell migration ability significantly increased, particularly at 48 h, and the anti‐migration effect of TP was notably inhibited (Figure 6D–F). These results suggest that PSAT1 may play an important regulatory role in the anti‐ LUAD activity of TP.

FIGURE 6.

Evaluation of the anti‐tumor efficacy of TP in PSAT1‐overexpressing cells. (A) RT‐qPCR detection of PSAT1 mRNA expression in the A549 overexpression model. (B) Cell viability of A549 cells after 24 h of TP treatment at different concentrations. (C) Cell viability of A549 cells after 48 h of TP treatment at different concentrations. (D) Representative results of the wound healing assay (TP: 60 nmol/L, n = 3, scale bar: 200 μm). (E) Cell migration rate at 24 h (n = 3). (F) Cell migration rate at 48 h (n = 3). *p < 0.05, **p < 0.01, ***p < 0.001.

3.6. Validation of the potential of TP in GEM chemoresistance

Previous studies have indicated that PSAT1 plays an important regulatory role in GEM resistance of LUAD. ¹¹ Based on this, this study further explored the application value of TP in the context of GEM resistance. Cell viability analysis showed that compared with A549 cells, the antitumor efficacy of TP at 24 h and 48 h was relatively reduced in the A549/GR model (Figure 7A,B). In A549/GR cells, even at the 24 h maximum effective concentration of GEM (1000 μmol/L), cell viability remained above 50%, with an IC50 value of 193.2 μmol/L at 48 h (Figure 7C,D). In contrast, the IC50 of GEM in A549 cells at 24 h was 1.17 μmol/L (Figure 7E). These results suggest that the efficacy of both TP and GEM was suppressed in A549/GR cells. The scratch assay showed that in A549/GR cells, the effect of GEM in inhibiting cell migration was significantly suppressed. However, when combined with TP, the inhibitory effect increased significantly, especially after 48 h of intervention. Conversely, after overexpressing PSAT1, the inhibitory effects of both TP and GEM were suppressed, but the combination of TP and GEM still showed a significant inhibitory effect (Figure 7F–H).

FIGURE 7.

Validation of TP's potential role in GEM chemoresistance. (A) Cell viability of A549/GR cells treated with different TP concentrations for 24 h. (B) Cell viability of A549/GR cells treated with different TP concentrations for 48 h. (C) Cell viability of A549 cells treated with different GEM concentrations for 24 h. (D) Cell viability of A549/GR cells treated with different GEM concentrations for 24 h. (E) Cell viability of A549/GR cells treated with different GEM concentrations for 48 h. (F) Representative results of wound healing assay (TP: 60 nmol/L, GEM: 60 μmol/L, n = 3, scale bar: 200 μm). (G) Cell migration rate at 24 h (n = 3). (H) Cell migration rate at 48 h (n = 3). **p < 0.01, ***p < 0.001.

4. DISCUSSION

TP is a diterpenoid epoxide isolated from the traditional Chinese medicine Tripterygium wilfordii Hook. f. Since its isolation in 1972, it has been proven to possess various pharmacological activities such as anti‐inflammatory, anti‐tumor, and immunosuppressive effects. ¹⁵ In terms of pharmacological activities, we investigated the potential mechanisms underlying the anti‐rheumatoid arthritis effects ¹⁶ and anti‐tumor effects ⁵ of TP. Additionally, a systematic summary has been conducted on the toxicity mechanisms of TP and potential detoxification strategies. ¹ To address its hepatotoxicity, a detoxification strategy combining glycyrrhizic acid and TP has been proposed, with its efficacy validated at both in vitro and in vivo levels. ¹⁷ Focusing on the antitumor activity of TP, we investigated its efficacy against LUAD using in vivo models and found that TP exhibited potent pharmacological activity in A549 tumor‐bearing mice, effectively inhibiting tumor growth (Figure 2A,B). However, the potent pharmacological activity of TP underpins its complex mechanism. Bioinformatics can integrate existing datasets to explore the potential mechanisms of the drug from different perspectives, offering significant value in elucidating the mechanisms of drug efficacy and identifying clinical application scenarios.

To investigate the potential anti‐LUAD mechanisms of TP, transcriptomic profiling was performed on mouse tumor tissues, and the analysis was combined with the LUAD dataset from the TCGA database. Analysis of the transcriptomic profiling results from mouse tumor tissues identified a total of 1002 DEGs between TP and Control groups, including 210 upregulated and 792 downregulated genes (Figure 2C). Functional analysis of these DEGs revealed associations with pathways such as biosynthesis of amino acids, glycine, serine and threonine metabolism, glycolysis/gluconeogenesis, and pyruvate metabolism (Figure 2D–F). Analyzing the LUAD dataset from the TCGA database, a total of 2487 DEGs were identified, including 2037 upregulated genes and 450 downregulated genes (Figure 3A). After integration, a total of 81 potential target genes were identified. Functional enrichment analysis revealed that the 81 candidate hub genes were associated with pathways such as Biosynthesis of Amino Acids, Phenylalanine Metabolism, Carbon Metabolism, Glycolysis/Gluconeogenesis, Histidine Metabolism, and Glycine, Serine and Threonine Metabolism (Figure 3D,F). These findings suggest that the anti‐LUAD mechanism of TP could be associated with the regulation of amino acid‐related biological processes.

To further elucidate the potential regulatory hub genes, the candidate gene set was re‐screened using LASSO and random forest methods. LASSO regression identified a total of 17 candidate genes, while the random forest analysis selected the top 20 genes. The intersection of results from both algorithms resulted in 5 potential hub genes (SYT12, FMO2, GPT2, PSAT1, and SLC25A10), which were validated using RT‐qPCR (Figure 4D,E). Among them, PSAT1 has been shown to be closely associated with amino acid biological processes. ¹⁸

Tumor cells exhibit a highly metabolic state and rely on exogenous serine, glycine, and other amino acids to support their initiation, proliferation, metastasis, and resistance. ¹⁹ , ²⁰ , ²¹ , ²² Among them, serine plays a crucial role in one‐carbon metabolism and makes significant contributions to various cellular processes, including nucleotide synthesis, glutathione production, regulation of the NADPH/NADP⁺ ratio, and maintenance of the cellular redox balance. ¹⁹ , ²³ , ²⁴ , ²⁵ Serine metabolism, as an important aspect of tumor metabolic reprogramming, plays a significant regulatory role in tumor growth and chemoresistance. ¹⁹ , ²³ , ²⁴ Studies have found that a serine/glycine‐restricted diet helps regulate the systemic immunity and promotes the anti‐tumor efficacy. ²¹ Restricting exogenous serine intake in colorectal cancer can effectively inhibit tumor growth and reduce the resistance to 5‐fluorouracil. ²⁴ Under conditions of exogenous serine deprivation, approximately 70% of cellular serine is derived from the serine synthesis pathway. ²⁶ The serine synthesis pathway begins with the glycolytic intermediate 3‐phosphoglycerate, which undergoes a series of enzymatic transformations via key serine synthesis enzymes—3‐phosphoglycerate dehydrogenase (PHGDH), phosphoserine transaminase (PSAT1), and phosphoserine phosphatase (PSPH) to generate serine. ¹⁹ , ²³ , ²⁴ PSAT1, a key enzyme in the serine synthesis pathway, has been demonstrated to be closely associated with tumor metastasis in LUAD. ²⁷ Inhibition of PSAT1 expression enhances the chemosensitivity of GEM in non‐small cell lung cancer (NSCLC), while PSAT1 expression also influences the efficacy of serine/glycine‐restricted diet therapy. ¹¹ Based on this, combined with the results of functional enrichment analysis, this study selected PSAT1 for further exploration of the anti‐tumor mechanism of TP and its potential clinical application scenarios.

Analysis using the TIMER database showed that PSAT1 was significantly upregulated in multiple tumor tissues, including LUAD. Correlation analysis of clinical characteristics revealed that PSAT1 expression levels were closely associated with the clinical stage and lymph node metastasis of LUAD. Additionally, RT‐qPCR results from tumor tissues in mice showed that TP significantly inhibited PSAT1 mRNA levels (Figure 4E), while immunohistochemistry results indicated that TP effectively suppressed PSAT1 protein expression in tumor tissues (Figure 5H,I). Cell viability assays showed that TP exerted a concentration‐dependent inhibitory effect on A549 cells (Figure 6B,C). The scratch assay indicated that TP effectively inhibited tumor cell proliferation; however, after PSAT1 overexpression, the cell proliferation rate significantly increased, and the inhibitory effect of TP was notably attenuated (Figure 6D,F). These results indicate that PSAT1 plays an important regulatory role in LUAD and is crucial in TP‐mediated antitumor efficacy. In terms of drug resistance, although the cytotoxicity of TP and GEM toward A549/GR cells was lower compared to A549 cells, their combination significantly inhibited the proliferation of A549/GR cells (Figure 7F–H). These findings suggest that TP and its related formulations have potential application value in clinical scenarios of GEM resistance. Additionally, beyond its role in GEM resistance in LUAD, previous studies have indicated that PSAT1 plays critical roles in sunitinib resistance in advanced renal cell carcinoma, 5‐FU resistance in colorectal cancer, and EGFR inhibitor resistance in LUAD. ²⁴ , ²⁸ , ²⁹ Based on the findings of this study, further exploration of the clinical applications of TP and its formulations is crucial for addressing clinical challenges and uncovering the antitumor potential of TP.

5. CONCLUSIONS

Taken together, the results of this study identified that the PSAT1‐regulated serine biosynthesis pathway is closely associated with the development of LUAD, and PSAT1 plays a critical regulatory role in TP‐mediated inhibition of LUAD activity. Additionally, in a GEM‐resistant LUAD model, the combined antitumor efficacy of TP and GEM was significantly enhanced, with PSAT1 exerting a clear regulatory effect in this process. Thus, based on the regulatory mechanism of PSAT1 in TP's antitumor activity, further investigation of the underlying mechanisms and exploration of the clinical applications of TP hold significant value for both addressing clinical needs and uncovering TP's full antitumor potential.

AUTHOR CONTRIBUTIONS

Zhiwen Cao: Conceptualization; writing – original draft. Lulu Zhang: Conceptualization; writing – original draft. Wenqiang Zhang: Data curation. Rong Wan: Data curation. Xiaogang Peng: Data curation. Jinyan Xie: Data curation. Ruru Bai: Data curation. Jiejing Jin: Data curation. Changqi Shi: Visualization. Lan Yan: Validation; visualization. Xiangyu Guo: Writing – review and editing. Yang Shen: Writing – review and editing. Cheng Lu: Writing – review and editing.

FUNDING INFORMATION

This research was funded by the Beijing Science and Technology New Star Program Cross‐cooperation Project (No. 20240484711), Jiangxi Provincial Natural Science Foundation (20252BAC200586) and National Natural Science Foundation of China (No. 82560858).

CONFLICT OF INTEREST STATEMENT

The authors declare that they have no competing interests.

ETHICS STATEMENT

The experimental protocol received approval from the Research Ethics Committee of the Institute of Basic Theory of Chinese Medicine, China Academy of Chinese Medical Sciences (IBTCMCACMS21‐2403‐16).

CONSENT FOR PUBLICATION

All authors gave consent for the publication of the article.

Cao Z, Zhang L, Zhang W, et al. Bioinformatics‐based discovery of the involvement of PSAT1 in mediating the anti‐lung adenocarcinoma activity of triptolide. Anim Models Exp Med. 2026;9:115‐127. doi: 10.1002/ame2.70120

DATA AVAILABILITY STATEMENT

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.