AGS and MKN45 human GC cell lines were acquired from the Procell Life Science & Technology Co., Ltd. and maintained in RPMI-1640 containing 10% FBS (Gibco; Thermo Fisher Scientific, Inc.) at 37°C with 5% CO₂ condition. Senescence was induced by treating AGS cells with 2.5 µM oxaliplatin (OXA) and MKN45 cells with 5 µM OXA for 48 h.

To determine the 50% inhibitory concentration and GC cells were plated in 96-well plates (3.0×10³ cells/well) and treated with a gradient of OXA concentrations: 0, 1.25, 2.5, 5, 10, 20, 40 and 80 µM. After 48 h, 10 µl CCK-8 reagent (Beyotime Institute of Biotechnology) was added per well, followed by a 1 h incubation in the dark at 37°C. Absorbance at 450 nm was recorded with a microplate reader (EPOCH-SN; BioTek Instruments, Inc.).

Cellular senescence was assessed with a commercial kit (Beyotime Institute of Biotechnology). Cells were rinsed with phosphate-buffered saline (PBS), fixed, incubated with SA-β-Gal staining solution overnight at 37°C without CO₂, and then washed with PBS. Images were captured using an optical microscope (IX73; Olympus Corporation), and the percentage of SA-β-Gal-positive cells was quantified.

GC cells were lysed using RIPA buffer (Beijing Solarbio Science & Technology Co., Ltd.) with protease inhibitors (Beyotime Institute of Biotechnology). Protein concentration was then measured with a BCA kit (Beyotime Institute of Biotechnology). Proteins (20 µg per lane) were resolved by 10–15% SDS-PAGE and then transferred onto PVDF membranes (MilliporeSigma). Membranes were blocked with 5% non-fat milk at room temperature (RT) for 2 h, washed with TBST (0.1% Tween-20). They were then incubated with primary antibodies overnight at 4°C and subsequently with secondary antibodies for 1 h at RT. Protein band was developed with ECL reagent (Beyotime Institute of Biotechnology), captured with an Amersham Image Quant 800 imaging system (Amersham; Cytiva), and analyzed for gray-level. Antibody details were listed in Table SI.

Cells (1.0×10⁵) were fixed with 4% paraformaldehyde, blocked with 3% BSA (Wuhan Servicebio Technology Co., Ltd.) for 30 min at RT. Cells were then incubated overnight at 4°C with a primary antibody against phospho-H2A.X (1:100; cat. no. GB111841; Wuhan Servicebio Technology Co., Ltd.), followed by incubation with a Cy3-conjugated goat anti-rabbit IgG secondary antibody (1:300; cat. no. GB21303; Wuhan Servicebio Technology Co., Ltd.) for 50 min at RT. Nuclei were visualized by DAPI (2.0 µg/ml) counterstaining for 10 min. Images were captured using a Nikon Eclipse C1 fluorescence microscope (Nikon Corporation). Antibody details were listed in Table SI.

Total RNA was isolated and reverse transcribed into cDNA using commercial kits (Tiangen Biotech Co., Ltd.) according to the manufacturer's instructions. RT-qPCR was conducted with SYBR Green Master Mix (Tiangen Biotech Co., Ltd.) under the following thermocycling protocol: 95°C for 2 min, 95°C for 5 sec, 60°C for 10 sec and 72°C for 15 sec for 40 cycles. Relative gene expression was calculated using the 2^−ΔΔCq method (16). Primers were listed in Table SII.

AGS cells were divided into two experimental groups: An untreated control group and an oxaliplatin-induced senescent group. For transcriptomic, proteomic and untargeted metabolomics analyses, four independent biological replicates were included for each group (n=4 per group).

Total RNA was isolated with TRIzol reagent (Beyotime Institute of Biotechnology). RNA concentration was measured using a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific, Inc.). RNA samples with a concentration ≥10 ng/µl and a total amount ≥1 µg were considered qualified for subsequent library construction. RNA integrity was verified by 1% RNase-free agarose gel electrophoresis stained with GelRed. cDNA libraries were constructed with the Hieff NGS^® Ultima Dual mode mRNA Library Prep Kit (cat. no. 12309ES; Shanghai Yeasen Biotechnology Co., Ltd.) and were then amplified by PCR. Library concentration was evaluated using the DNA1000 assay Kit (cat. no. 5067-1504; Agilent Technologies, Inc.), and the final library loading concentration was 3 ng/µl. Sequencing was performed on an Illumina NovaSeq X Plus platform with a paired-end 150 bp strategy by Gene Denovo Biotechnology Co., Ltd.

Next, using DESeq2 software, differentially expressed genes (DEGs) were selected based on the criteria of |fold change| ≥2 and an adjusted P<0.05. Volcano plots and heatmaps were plotted with the online Omicsmart platform (http://www.omicsmart.com). Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis was performed on DEGs.

Following protein extraction and enzymatic digestion, the peptides were separated by an UltiMate 3000 liquid chromatography system (Thermo Fisher Scientific, Inc.) and analyzed using a timsTOF Pro2 mass spectrometer from Bruker Daltonics. DIA proteomics data were analyzed using Spectronaut 18 (Biognosys AG) under default parameters. The ideal extraction window was dynamically determined by the Spectronaut using iRT calibration and gradient stability. A false discovery rate threshold of 1% was applied at both the precursor and protein levels. Peptides passing this filter were quantified using the MaxLFQ algorithm to generate normalized protein group abundances. Mass spectrometric data were acquired in DIA mode using diaPASEF. Differentially expressed proteins were defined by a |fold change| >1.2 and a Benjamini-Hochberg adjusted P<0.05.

The samples were mixed with methanol/acetonitrile/H₂O (2:2:1, v/v/v). Next, mixtures were sonicated twice on ice for 30 min each and centrifuged at 14,000 × g (20 min, 4°C). The supernatant was collected and evaporated under vacuum. Dried extracts were redissolved in an acetonitrile aqueous solution, centrifuged, and the supernatant was collected for injection. Separation was conducted on an Agilent 1290 Infinity UHPLC system interfaced with an AB Sciex TripleTOF 6600 Q-TOF mass spectrometer, and quality control was conducted using QC samples. Data from the positive (pos) and negative (neg) ion modes were processed independently. Raw data were transformed into the mzML format utilizing ProteoWizard. Subsequent processing, including peak alignment, retention time correction, and extraction of peak area, was carried out with XCMS software.

Metabolite identification and data preprocessing were then performed; the steps were as follows: Features with missing values exceeding 50% were removed; remaining missing values were filled using k-nearest neighbor imputation; and features with a relative standard deviation greater than 50% were filtered out. Significantly altered metabolites were selected with the variable importance in projection (VIP) score ≥1 in OPLS-DA model, followed by filtering with a univariate t-test (P<0.05).

The prognostic significance of NR4A1 expression in GC was evaluated via the GEPIA2 platform (http://gepia2.cancer-pku.cn/), which combines RNA-seq data from The Cancer Genome Atlas (TCGA) and the Genotype-Tissue Expression projects (17). For the OS analysis, TCGA-stomach adenocarcinoma (STAD) cases were segregated into ‘high’ and ‘low’ NR4A1 expression groups according to the median transcripts per million values as the cut-off. Survival outcomes between groups were compared using Kaplan-Meier analysis with the log-rank (Mantel-Cox) test. The hazard ratio along with 95% confidence intervals was calculated.

The short hairpin RNA (shRNA) NR4A1 lentiviral vector was constructed by Shanghai GeneChem Co., Ltd., using the pFU-GW backbone. The sequences for the shRNAs are as follows: A non-targeting control (sh-NC, 5′-TTCTCCGAACGTGTCACGT-3′) and three NR4A1-specific sequences (sh-NR4A1#1 sense: 5′-TGGTGAAGGAAGTTGTCCGAA-3′, antisense: 5′-TTCGGACAACTTCCTTCACCA-3′; sh-NR4A1#2 sense: 5′-GATTGACAGTATCCTGGCCTT-3′, antisense: 5′-AAGGCCAGGATACTGTCAATC-3′; sh-NR4A1#3 sense: 5′-CTCCTTCAAGTTCGAGGACTT-3′, antisense: 5′-AAGTCCTCGAACTTGAAGGAG-3′).

GC cells plated in 6-well plates were exposed to viral particles in medium containing 5 µg/ml polybrene at a multiplicity of infection value of 20. Following incubation under standard conditions (37°C, 5% CO₂), medium was replaced with a fresh complete medium and cultured the cells for another 48 h. The lentiviral transduction was conducted over 72 h. Successfully transduced cells were selected with 2.0 µg/ml puromycin for 48 h, followed by an additional 48 h recovery period in fresh complete medium prior to subsequent experiments.

Following the instructions of a commercial lactate content detection kit (Wuhan Servicebio Technology Co., Ltd.), for lactate assay, L-lactic acid was oxidized by a specific lactate oxidase to pyruvate and H₂O₂. In the presence of peroxidase and chromogens, a colored dye was produced, which was measured photometrically at 546 nm. The absorbance of the dye, measured photometrically, was directly proportional to the L-lactic acid concentration. The results were normalized according to the total protein content.

Following the instructions of the ATP content detection kit (Wuhan Servicebio Technology Co., Ltd.), ATP assay was based on the principle that luciferase catalyzes the chemiluminescence reaction of luciferin under the energy provided by ATP, and within a certain range, the luminous intensity was directly proportional to the ATP content. The results were normalized based on the cell count.

Statistical analyses and data visualization were performed using GraphPad Prism software (v9.5.0; Dotmatics) and shown as the mean ± standard deviation (SD). Significance was determined using an unpaired two-tailed Student's t-test or one-way analysis of variance, followed by Tukey's multiple comparisons test. P<0.05 was considered to indicate a statistically significant difference.

To verify whether OXA induce senescence in GC cells, an in vitro model was established using AGS and MKN45 cells lines. Briefly, cells were assigned to control (Ctrl) and OXA groups. CCK-8 assays revealed that OXA reduced AGS and MKN45 cell viability in a concentration-dependent manner (Fig. S1A and B). Based on these results, AGS cells were treated with 2.5 µM OXA, while MKN45 cells were treated with 5 µM OXA, both for 48 h. Senescence was assessed through SA-β-Gal staining, western blotting, immunofluorescence detection of γ-H2AX, and qPCR analysis of SASP factors.

SA-β-Gal staining detects cellular senescence by measuring the SA-β-Gal activity at pH 6.0, which hydrolyzes X-Gal to generate a blue product (18). As presented in Fig. 1A, compared with the Ctrl group, OXA group displayed a greater percentage of SA-β-Gal-positive GC cells, along with enlarged and flattened morphology. Quantitative analysis confirmed that the average SA-β-Gal positivity rates were 88.3% in AGS cells from the OXA group and 89.32% in MKN45 cells from the OXA group (Fig. 1B).

As a cyclin-dependent kinase inhibitor, p21 is a key mediator of cellular senescence (19). Western blotting indicated a significant upregulation of p21 protein in the OXA group compared with the Ctrl group, suggesting the induction of cell cycle arrest (Fig. 1C). Since γ-H2AX is a classic marker of DNA damage and replication stress and is a central driver of senescence induction (20), its expression was assessed. Immunofluorescence showed a stronger enhancement of γ-H2AX foci in OXA group than in Ctrl group (Fig. 1D and E).

SASP comprises the secretion of diverse inflammatory cytokines, chemokines and proteases, which affects the tumor microenvironment (TME) (21). Next, the mRNA expression levels of major SASP components were evaluated using RT-qPCR. RT-qPCR analysis indicated a significant upregulation of various SASP factors, such as interleukin (IL)-6, IL-8 and IL-1β in the OXA group relative to the Ctrl group in AGS and MKN45 cells (Fig. 1F).

Finally, to clarify the cell cycle regulatory network associated with senescence, the expression of key regulatory proteins was examined (22). Compared with the Ctrl group, CDK4 and cyclin D1 were significantly downregulated in OXA group; however, total RB protein levels remained unchanged, and p-RB levels decreased. These results suggested that after OXA induction, cell cycle progression was inhibited, and RB was maintained in a hypo-phosphorylated active state (Fig. 1G).

In summary, it was established that OXA induces senescence in AGS and MKN45 cells, as evidenced by elevated SA-β-gal activity, p21 upregulation, cell cycle regulator inhibition, DNA damage accumulation and enhanced SASP secretion.

To systematically elucidate the molecular mechanisms of OXA-induced senescence in GC cells, integrative transcriptomic and proteomic analyses of AGS cells were performed by comparing Ctrl and OXA groups.

As demonstrated in Fig. 2A, the transcriptomic volcano plot revealed 1,536 DEGs in the OXA group (|log₂FC| >0.585, P<0.05) with 1,193 upregulated and 343 downregulated genes. Proteomic analysis further identified 2,047 proteins that were significantly altered (|log₂FC| >0.26, P<0.05), including 1,040 upregulated and 1,007 downregulated proteins. Venn diagram analysis exhibited that 224 molecules were consistently differentially expressed at both transcriptomic and proteomic levels (Fig. 2B), implicating a central role in OXA-induced senescence.

KEGG enrichment of the overlapping differentially expressed molecules revealed marked enrichment in pathways closely associated with senescence, tumor progression and metabolism, including ‘p53 signaling pathway’, ‘PI3K-AKT signaling pathway’, ‘platinum drug resistance’ and ‘metabolic pathways’ (Fig. 2C). Notably, the PI3K-AKT signaling pathway was significantly enriched at both omics’ levels, highlighting its critical role in chemotherapy-induced senescence. The expression of PI3K-AKT pathway associated genes and their corresponding proteins, was then visualized using heatmaps (Fig. 2D).

Among these candidates, NR4A1 was explored, a stress-responsive nuclear receptor that is regarded as a key molecule linking signal transduction with metabolic fate. Emerging evidence indicates that NR4A1 regulates cellular metabolism and fate by suppressing PI3K/AKT signaling (23,24). It was discovered that in GC, NR4A1 expression was elevated at the transcriptional and protein levels in the OXA group. Consistent with this, qPCR (Fig. 2E) and western blotting (Fig. 2F) confirmed elevated NR4A1 expression in OXA group compared with the Ctrl group. Moreover, to enhance the clinical value of NR4A1, its prognostic significance was further analyzed using the TCGA-STAD dataset via GEPIA2. Notably, Kaplan-Meier survival analysis showed a shorter OS in patients with high relative to low NR4A1 expression (Fig. 2G).

Collectively, integrative multi-omics analyses indicate that OXA induces cellular senescence in GC cells by upregulating NR4A1 and suppressing the PI3K/AKT pathway. Notably, NR4A1 expression was induced during OXA-induced senescence in vitro, whereas intrinsically high NR4A1 expression was correlated with poor prognosis, indicating a context-dependent role of NR4A1 in GC.

To explore the role of NR4A1 in chemotherapy-induced senescence, NR4A1-knockdown AGS and MKN45 cell lines were generated using shRNA technology. Western blotting confirmed that NR4A1 protein levels were decreased in the sh-NR4A1 groups compared with the sh-NC group, with sh-NR4A1#2 exhibiting the most efficient knockdown (Fig. 3A). Consequently, sh-NR4A1#2 was selected for subsequent experiments.

The senescence marker p21 expression was first detected. As expected, p21 expression was significantly upregulated in the sh-NC + OXA group relative to the sh-NC group. However, this induction was significantly weakened in the sh-NR4A1#2 + OXA group (Fig. 3B). These results indicated that NR4A1 knockdown partially reversed the p21-mediated senescence signal triggered by OXA. Consistently, SA-β-Gal staining indicated that senescent cells increased substantially in the sh-NC + OXA group compared with sh-NC, while this increase was significantly reduced in the sh-NR4A1#2 + OXA group (Fig. 3C). Next, the SASP phenotype was assessed using RT-qPCR. High expression of typical SASP factors was found in the sh-NC + OXA group, but this upregulation was significantly reduced in the sh-NR4A1#2 + OXA group (Fig. 3D).

Based on the aforementioned findings, the PI3K-AKT pathway was highly enriched in transcriptomic and proteomic analyses, and it was further explored whether NR4A1 was involved in regulating this pathway. For this purpose, levels of phosphorylated AKT at Ser473 (p-AKT) and total AKT, core indicators of PI3K/AKT pathway activity, were analyzed by western blotting. It was found that the sh-NC + OXA group significantly suppressed AKT phosphorylation (decreased p-AKT/AKT ratio). This suppression was effectively reversed in the sh-NR4A1#2 + OXA group, in which the p-AKT/AKT ratio was restored (Fig. 3E).

In conclusion, NR4A1 is a critical regulator of OXA-induced senescence in GC cells. NR4A1 knockdown weakened p21 upregulation, SA-β-Gal positivity and SASP secretion, and restored PI3K/AKT-associated pathway activity. Collectively, these results highlight a central role for the NR4A1-PI3K/AKT axis in chemotherapy-induced cellular senescence.

To assess metabolic alterations during chemotherapy-induced senescence, untargeted metabolomic profiling of AGS cells was performed in Ctrl and OXA groups. Score plots demonstrated an obvious separation between Ctrl and OXA groups in neg and pos ion detection modes, indicating favorable reproducibility (Fig. 4A). Permutation tests further confirmed that the model fits well in both modes (Fig. 4B), supporting the metabolomic analysis stability. The aforementioned results prove the stability of the experiment and the reliability of the observed metabolic differences. Differential metabolite analysis revealed pronounced metabolic alterations in the OXA group under both neg and pos ion modes. Specifically, 41 metabolites were upregulated and 66 downregulated in the neg mode, while 59 were upregulated and 76 downregulated in the pos mode (VIP >1, P<0.05) (Fig. 4C). Following integration of the differential metabolites from both neg and pos modes, KEGG pathway enrichment analysis was performed (Fig. 4D). The analysis indicated that these metabolites were substantially enriched in key metabolic pathways, including ‘OXPHOS’, ‘HIF-1 signaling pathway’, ‘metabolic pathways’ and ‘nucleotide metabolism’ (Fig. 4D). Notably, the significant enrichment of OXPHOS and HIF-1 signaling, two central energy metabolism pathways, was significantly enriched. This result suggests that OXA-induced senescence in GC cells may drive metabolic reprogramming by inhibiting OXPHOS and promoting glycolysis.

To verify the role of NR4A1 in metabolic reprogramming, the mRNA expression levels of key glycolytic enzymes were measured. RT-qPCR analysis demonstrated significant upregulation of lactate dehydrogenase A (LDHA) and hexokinase 2 (HK2) in the sh-NC + OXA group relative to the sh-NC group, whereas this effect was significantly reversed in the sh-NR4A1#2 + OXA group (Fig. 4E). The functional experiments supported these findings. Lactate levels were substantially increased in the sh-NC + OXA group and were significantly reduced in the sh-NR4A1#2 + OXA group (Fig. 4F). Concurrently, ATP detection showed a notable decrease in cellular ATP content in the sh-NC + OXA group relative to sh-NC, reflecting impaired OXPHOS function. Notably, ATP production was restored in the sh-NR4A1#2 + OXA group, exceeding that in the sh-NC + OXA group (Fig. 4G).

In conclusion, OXA-induced GC cells triggered extensive metabolic reprogramming characterized by OXPHOS inhibition and glycolysis enhancement (‘Warburg effect’). NR4A1 is a central mediator of OXA-induced metabolic reprogramming. NR4A1 knockdown effectively reversed the upregulation of glycolytic genes (LDHA and HK2) by OXA, alleviated ATP generation suppression, and reduced lactate accumulation. Building on previous findings, these data demonstrate that NR4A1-mediated inhibition of the PI3K/AKT pathway serves as a novel bridge connecting chemotherapy-induced stress to metabolic reprogramming (Fig. 5).