Actin gamma smooth muscle 2 drives proliferation, metastasis, and 5-Fluorouracil resistance in gastric cancer: insights from a Recurrence-Related Gene Signature

Abstract

Background

Gastric cancer is characterized by high recurrence rates after curative treatment, significantly contributing to poor survival outcomes. Current clinicopathological parameters inadequately predict recurrence risk, highlighting the need for robust molecular signatures to guide personalized therapy.

Methods

We analyzed gene expression profiles from multiple datasets to identify recurrence-associated genes in gastric cancer. A novel Recurrence-Related Gene Signature (RRGS) was constructed using LASSO regression, and its prognostic value was validated across independent cohorts. Functional validation of actin gamma smooth muscle 2 (ACTG2), which is encoded by the ACTG2 gene, was performed through in vitro knockdown experiments, zebrafish xenograft models, and chemoresistance analyses to assess its role as a key gene product in RRGS.

Results

We identified 72 consistently upregulated and 1 downregulated DEG across the two datasets, which were enriched in focal adhesion, ECM-receptor interaction, and PI3K-Akt signaling. A 32-gene RRGS reliably predicted recurrence risk and correlated with immune cell infiltration patterns. Among these genes, ACTG2 emerged as a key driver of gastric cancer aggressiveness, as its silencing significantly diminished tumor cell proliferation, migration, and invasion, and increased 5-fluorouracil sensitivity both in vitro and in vivo. Elevated ACTG2 protein expression was further validated immunohistochemically in recurrent patient tumors.

Conclusion

Our findings highlight RRGS may serve as a prognostic tool and indicate that ACTG2 may plays an important role in both gastric cancer progression and chemotherapy resistance. Targeting ACTG2 or its downstream pathways may offer novel therapeutic opportunities to enhance treatment efficacy for patients with gastric cancer.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12935-026-04312-4.

Introduction

Gastric cancer remains one of the most common malignant tumors worldwide, ranking as the fifth most frequently diagnosed cancer and the fourth leading cause of cancer-related mortality globally [1–3]. The incidence varies significantly across different geographical regions, with the highest rates observed in Eastern Asia, particularly in Japan, Korea, and China [4–6]. Despite advances in early detection techniques and multimodal treatment strategies, the 5-year overall survival rate for gastric cancer patients remains disappointingly low, primarily due to late-stage diagnosis and high recurrence rates after curative treatment.

Surgical resection combined with adjuvant or neoadjuvant chemotherapy represents the mainstay of treatment for localized gastric cancer [7]. However, even after curative-intent treatment, approximately 40–60% of patients experience disease recurrence, which is often refractory to subsequent therapies and significantly contributes to the poor long-term survival outcomes [8]. Recurrence patterns in gastric cancer are heterogeneous, including locoregional, peritoneal, and distant metastases, suggesting complex underlying biological mechanisms [9].

Current clinical risk stratification for gastric cancer recurrence relies primarily on the TNM staging system, Lauren’s classification, and histopathological features [10]. However, these conventional approaches fail to capture the molecular heterogeneity of gastric cancer and often prove insufficient for accurate individualized recurrence prediction. In recent years, genomic and transcriptomic profiling has facilitated the development of molecular classifications and gene signatures that provide additional prognostic information beyond traditional clinicopathological parameters.

Actin gamma smooth muscle 2 (ACTG2), a smooth muscle-specific actin isoform, has been implicated in cytoskeletal organization and cell motility [11]. While altered expression of ACTG2 has been reported in several malignancies [12–14], its specific role in gastric cancer pathogenesis, particularly in relation to recurrence and chemoresistance, has not been comprehensively investigated. Recent studies have suggested that cytoskeletal proteins might contribute to cancer progression and therapeutic resistance by modulating cellular processes such as migration, invasion, and stress response, making ACTG2 a potentially interesting candidate for exploration in gastric cancer.

In this study, we sought to identify and validate recurrence-associated genes in gastric cancer through comprehensive analysis of multiple independent transcriptomic datasets. We developed a novel Recurrence-Related Gene Signature (RRGS) for predicting recurrence risk and characterized its associations with clinicopathological features and the tumor microenvironment. Furthermore, we focused on ACTG2, a previously uncharacterized gene in gastric cancer from our signature, to investigate its functional roles in tumor progression and chemoresistance both in vitro and in vivo . Our findings aim to provide insights into the molecular mechanisms underlying gastric cancer recurrence and identify potential therapeutic targets to improve patient outcomes.

Methods

In this study, we obtained publicly available gene expression profiles from the GEO database, specifically GSE13861, GSE26901, and GSE66229. We normalized and preprocessed the microarray data using the robust multi-array average (RMA) method implemented in the ‘affy’ package in R. The GSE13861 dataset comprises 65 patients with gastric cancer: 34 remained recurrence-free, while 31 experienced relapse. Among these 65 patients, 49 underwent chemotherapy, with 25 ultimately relapsing post-treatment, and 24 remaining recurrence-free. Similarly, GSE26901 includes data from 109 gastric cancer patients, of whom 49 had no recurrence and 60 relapsed during follow-up. Of these 109 patients, 39 received chemotherapy, resulting in 25 instances of relapse post-therapy, while 14 remained free of recurrence. The GSE66229 dataset includes 300 gastric cancer patients, of whom 80 received chemotherapy. Among these 80 patients, 25 experienced relapse, while 55 remained free of recurrence.

Differentially Expressed Gene (DEG) analysis

Differential expression analysis was performed using the R package ‘limma’. Specifically, we calculated the log2 fold change for each gene. Genes displaying an absolute log2 fold change of at least 0.585 and an adjusted p-value below 0.05 were deemed statistically significant and were subsequently included in further analyses. This rigorous filtering ensured that only the most reliable differentially expressed genes were retained for downstream experimental interpretations.

Functional enrichment analysis

To explore the underlying biological processes and signaling pathways linked to the DEGs, we conducted enrichment analyses using the Kyoto Encyclopedia of Genes and Genomes (KEGG), Hallmark, and Gene Ontology (GO) databases. Pathways and processes with a p-value of less than 0.05 were deemed statistically significant.

Immune infiltration analysis

Immune infiltration within gastric cancer was evaluated using both the ESTIMATE and CIBERSORT algorithms. Subsequently, we investigated how the RRGS correlates with diverse immune cell subsets.

Cells culture

Human gastric cancer cell lines AGS and SNU-1 were obtained from the American Type Culture Collection (ATCC). Cells were routinely maintained in RPMI 1640 medium supplemented with 10% fetal bovine serum and a 1% penicillin–streptomycin solution, and cultured at 37 °C in a humidified incubator containing 5% CO₂. Culture conditions were carefully monitored to preserve cell viability, and cells were passaged when reaching approximately 80–90% confluence. Periodic morphological checks and mycoplasma testing were performed to ensure the authenticity and stability of the cell lines throughout the experiments.

Quantitative real-time PCR (qRT-PCR)

Total RNA was extracted from cultured cells using TRIzol reagent (Invitrogen), and complementary DNA (cDNA) was subsequently synthesized with the PrimeScript RT reagent Kit (TaKaRa, Japan) following the manufacturer’s instructions. QRT-PCR analysis was then carried out using SYBR Premix Ex Taq (TaKaRa). GAPDH was employed as the internal control, and relative mRNA expression levels were calculated using the 2^−ΔΔCT method. The primer sequences were as follows: ACTG2 forward 5′-GCGTGTAGCACCTGAAGAG-3′, ACTG2 reverse 5′-GAATGGCGACGTACATGGCA-3′; GAPDH forward 5′-GGAAGGACTCATGACCACAGTCC-3′; GAPDH reverse 5′-TCGCTGTTGAAGTCAGAGGAGACC-3′.

Cell proliferation assay

Cell proliferation was assessed using the Cell Counting Kit-8 (CCK-8) assay. Briefly, AGS and SNU-1 cells were each resuspended at 1 × 10⁴ cells/mL, and 2 × 10³ cells were seeded per well in 96-well plates. Cultures were maintained at 37 °C in a humidified incubator with 5% CO₂. At 24, 48, and 72 h, 10 μL of CCK-8 solution was added to each well, followed by a 1-h incubation. Absorbance at 450 nm was then recorded using a microplate reader. Each condition was assayed in triplicate, and the average optical density values were used to generate proliferation curves.

Cell migration assay

A scratch wound-healing assay was conducted to evaluate the migratory potential of AGS and SNU-1 cells. Briefly, cells were seeded into six-well plates at a density of 1 × 10⁶ cells per well and cultured under serum-starved conditions to reduce the confounding impact of proliferation. Upon reaching approximately 90% confluence, a straight scratch was introduced using a 10-µL pipette tip. Floating cells were removed by gentle washes with PBS, and images of the scratch area were taken at 0 and 48 h using an inverted microscope. The extent of wound closure was quantified with ImageJ software to assess migration capacity.

Cell invasion assay

Cell invasion capability was evaluated using Transwell inserts with polycarbonate membranes (Costar; Corning Inc.) in a 24-well plate setup. Briefly, AGS and SNU-1 cells were resuspended at a concentration of 2 × 10⁴ cells per well in 200 μL of serum-free medium and seeded into the upper chamber. The lower chamber contained 200 μL of complete medium supplemented with 10% FBS to serve as a chemoattractant. Cells were incubated at 37 °C in 5% CO₂ for 48 h. After incubation, non-invading cells on the top surface of the membrane were gently removed with a cotton swab. Cells that had invaded to the underside of the membrane were then fixed, stained, and quantified under a microscope.

Cell chemosensitivity assay

A chemosensitivity assay was performed using the Cell Counting Kit-8 (CCK-8), wherein AGS and SNU-1cells were exposed to a range of 5-fluorouracil (5-FU, Sigma-Aldrich) concentrations. This setup enabled us to observe the dose-dependent cytotoxic effects on both cell lines. Each condition was tested in triplicate to ensure experimental reproducibility. The half-maximal inhibitory concentration (IC50)-the concentration of 5-FU required to reduce cell viability by 50%-was calculated using GraphPad Prism 8. These data provided a quantitative basis for evaluating the responsiveness of AGS and CCK-8 cells to 5-FU treatment.

Zebrafish xenograft model

A zebrafish xenograft model was employed to investigate the in vivo tumor growth and metastatic capacity of AGS gastric cancer cells. Briefly, wild-type zebrafish (Danio rerio) were obtained from Fuzhou Bio-Service Biotechnology Co. Ltd (Fuzhou, China) and housed under standardized conditions at 28.5 °C with a 14/10-h light–dark cycle. To label AGS cells, they were digested using trypsin, then incubated in 5 μM 1,1′-dioctadecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate (Dil; Meilun Biotechnology, China), a red fluorescent membrane dye, enabling visualization of subsequent cell tracking. Approximately 200 labeled AGS cells were then transplanted into the central or ventral yolk sac of zebrafish larvae at 2 days post-fertilization using GB100T-8P glass capillaries and FemtoJet 4i microinjectors (Eppendorf, Germany), with each group comprising ten larvae. For proliferation evaluation, fluorescent images of tumor cells were acquired at 2 h and 48 h post-transplantation (hpt), while images at 2 h and 24 h hpt were used to assess metastatic potential. Zebrafish larvae under 5 days of age typically do not require formal ethical approval. This study complied with ARRIVE guidelines for animal research reporting.

Immunohistochemical staining

Gastric tissue sections were cut at a thickness of 4 μm. After dehydration and rehydration through an ethanol gradient, antigen retrieval was performed by heating the sections in 10 mM sodium citrate buffer (pH 6.0) at 100 °C for 10 min, followed by cooling to room temperature. The sections were then incubated overnight at 4 °C with a primary antibody against ACTG2 diluted 1:200. This was followed by incubation with an appropriate secondary antibody for 1 h, treatment with streptavidin–HRP conjugate for 1 h, and visualization with a DAB substrate kit for 30 s. Finally, the sections were counterstained with hematoxylin for 20 s. Images were acquired using a high-resolution whole-slide scanner, and the positive staining rate was quantified using ImageJ software (https://imagej.nih.gov/ij/).

Statistical analysis

All experiments were repeated at least three times, with results presented as mean ± standard deviation (SD). Statistical analysis was carried out using GraphPad Prism 8.0. To compare two groups, Student’s t-test was applied, whereas one-way ANOVA was employed for analyses involving more than two groups. Differences were considered statistically significant when p < 0.05.

Results

Identification of Recurrence-Related Genes in gastric cancer

To investigate genes related to gastric cancer recurrence, we analyzed the GSE13861 and GSE26901 datasets. Comparative analysis revealed 90 significantly upregulated and 24 downregulated genes in recurrent versus non-recurrent tissues from GSE13861 (Fig. 1A&B), while GSE26901 showed 305 upregulated and 20 downregulated genes (Fig. 1C&D). Venn diagram analysis identified 72 consistently upregulated genes and 1 downregulated gene overlapping between both datasets (Fig. 1E&F). KEGG enrichment analysis indicated that the 72 upregulated genes were primarily enriched in focal adhesion, ECM-receptor interaction, and PI3K-Akt signaling pathway (Fig. 1G). Hallmark pathway analysis demonstrated connections to epithelial-mesenchymal transition (EMT) and UV response downregulation (Fig. 1H), while GO-BP terms revealed enrichment in circulatory system development and biological adhesion (Fig. 1I), suggesting potential mechanisms underlying metastatic progression.

Fig. 1.

Differentially expressed genes in recurrence and non-recurrence groups of gastric cancer. (A) Volcano plot of DEGs identified in GSE13861 between recurrence (n = 31) and non-recurrence (n = 34) groups. (B) Heatmap showing the expression patterns of DEGs in GSE13861. (C) Volcano plot of DEGs identified in GSE26901 between recurrence (n = 60) and non-recurrence (n = 49) groups. (D) Heatmap showing the expression patterns of DEGs in GSE26901. (E) Venn diagram indicating the overlapping upregulated genes between the two datasets. (F) Venn diagram indicating the overlapping downregulated genes between the two datasets. Pathway enrichment analyses of the overlapping upregulated genes, including (G) KEGG pathways, (H) Hallmark pathways, and (I) GO-BP terms. DEGs were identified using the limma package with |log2FC|> 0.585 and adjusted p < 0.05. KEGG, Hallmark, and GO enrichment analyses were performed using the clusterProfiler R package with p < 0.05 as the cutoff

Development of a model to predict recurrence in gastric cancer patients

Subsequently, we analyzed the prognostic impact of the 73 DEGs on gastric cancer patient outcomes. Survival analysis revealed that 72 of these genes were significantly associated with prognosis in the GSE13861 dataset (Fig. 2A, Supplementary Table 1), and 70 genes showed statistically significant associations with prognosis in the GSE26901 dataset (Fig. 2B, Supplementary Table 2). A Venn diagram identified 69 intersecting genes shared between the two datasets (Fig. 2C). Based on these 69 intersecting genes, we employed a LASSO regression model to construct a recurrence prediction model for GC, termed the Recurrence-Related Gene Signature (RRGS), using the GSE26901 training cohort. This RRGS model incorporates 32 genes: ACTG2, ANGPTL2, C10orf10, CALD1, CAP2, CASQ2, COL3A1, COL5A1, CYBRD1, DPYSL3, FSTL1, HEYL, HSPB6, HTRA1, INMT, LAMC1, LRRC32, PALLD, PPP1R14A, RGMA, RHOB, SCRG1, SEMA3C, SFRP2, SLIT2, SMOC2, SPARCL1, SVEP1, THBS2, TIMP3, TMEM47, and TNC (Fig. 2D). The coefficients for each gene in the model are shown in Fig. 2E and provided in Supplementary Table 3.

Fig. 2.

Prognostic analysis of differentially expressed genes and construction of the recurrence prediction model for gastric cancer. (A&B) Forest plot of survival-associated genes in the (A) GSE13861 (n = 65) and (B) GSE26901 (n = 109) dataset. (C) Venn diagram identifying 69 overlapping survival-associated genes shared between GSE13861 and GSE26901 datasets. (D) LASSO regression analysis based on the 69 overlapping genes in the GSE26901 dataset to construct RRGS. (E) LASSO coefficient plot for the 32 genes included in the RRGS model. (F) Comparison of RRGS scores between recurrent (n = 60) and non-recurrent (n = 49) gastric cancer patients in the GSE26901 cohort. Data are shown as box plots with median and interquartile range; p value was calculated by Wilcoxon test. (G) ROC curve of RRGS predicting recurrence risk in the GSE26901 dataset (n = 109). (H) Kaplan–Meier survival analysis demonstrating shorter survival in the RRGS-High group compared to the RRGS-Low group within the GSE26901 cohort (n = 109). P value was calculated by Log-rank test. (I) Distribution of RRGS scores in recurrent (n = 31) versus non-recurrent (n = 34) patients in the GSE13861 validation cohort. Data are shown as box plots with median and interquartile range; p value was calculated by Wilcoxon test. (J) ROC analysis of recurrence prediction using RRGS in the GSE13861 dataset (n = 65). (K) Kaplan–Meier survival analysis showing reduced survival in RRGS-High patients relative to RRGS-Low patients within the GSE13861 validation cohort (n = 65). ***, p < 0.001

Further analysis demonstrated that the RRGS scores were significantly higher in recurrent GC patients compared to non-recurrent patients within the GSE26901 cohort (Fig. 2F). Receiver operating characteristic (ROC) analysis showed that RRGS exhibits excellent accuracy in predicting recurrence risk for gastric cancer patients, with an area under the curve (AUC) value of 0.94 (Fig. 2G). Kaplan–Meier survival analysis further revealed that patients in the RRGS-High group had significantly shorter overall survival compared to those in the RRGS-Low group (Fig. 2H). To validate the predictive performance of this model, we tested it in an independent validation cohort (GSE13861). Similar to the training cohort, RRGS scores were significantly higher in recurrent gastric cancer patients than in non-recurrent patients in the GSE13861 dataset (Fig. 2I). ROC analysis for the validation cohort demonstrated good predictive accuracy, with an AUC of 0.75 (Fig. 2J). Furthermore, Kaplan–Meier survival analysis confirmed that RRGS-High group patients had significantly shorter survival times when compared to the RRGS-Low group (Fig. 2K).

To further validate the RRGS model, we incorporated data from the GSE66229 dataset. Kaplan–Meier survival analysis revealed that patients in the RRGS-High group exhibited significantly shorter overall survival compared to the RRGS-Low group (Supplementary Fig. 1A), confirming the prognostic value of RRGS in this independent cohort. Furthermore, RRGS values were significantly higher in patients with stage III&IV disease (Supplementary Fig. 1B) and T3&4 stage (Supplementary Fig. 1C) compared to stage I&II and T2 stage respectively, indicating a correlation between RRGS score and tumor progression. Notably, RRGS scores did not significantly differ based on N stage, M stage, sex, or age (Supplementary Fig. 1D-G). However, RRGS values were significantly lower in chemotherapy-non-recurrent patients than in chemotherapy-recurrent patients (Supplementary Fig. 1H), further supporting the role of RRGS in predicting gastric cancer recurrence.

RRGS demonstrates robust risk stratification across clinical and molecular subgroups

To further evaluate the robustness of our RRGS model, we conducted stratified analyses in patients with gastric cancer based on sex, age, tumor location, Lauren classification, and clinical stage. Using the GSE29601 dataset, we performed subgroup analyses for Sex = Female (Fig. 3A), Sex = Male (Fig. 3B), Age ≤ 60 (Fig. 3C), Age > 60 (Fig. 3D), Location = Body (Fig. 3E), Location = Antrum (Fig. 3F), Lauren = Diffuse (Fig. 3G), Lauren = Intestinal (Fig. 3H), Stage = I&II (Fig. 3I), and Stage = III&IV (Fig. 3J). In all these subgroups, the RRGS-based score consistently showed independent prognostic value, with patients classified as RRGS-High exhibiting worse survival outcomes.

Fig. 3.

Stratified survival analysis of RRGS in clinical and molecular subgroups of gastric cancer. Subgroup analyses were conducted using the GSE29601 dataset for (A) female (n = 34), (B) male (n = 75), (C) age ≤ 60 (n = 48), (D) age > 60 (n = 61), (E) tumors located in the body (n = 42), (F) tumors located in the antrum (n = 51), (G) diffuse-type Lauren classification (n = 53), (H) intestinal-type Lauren classification (n = 46), (I) stage I&II (n = 51), and (J) stage III&IV (n = 58). Analyses in the validation cohort GSE13861 were performed for (K) female (n = 20), (L) male (n = 45), (M) age ≤ 60 (n = 27), (N) age > 60 (n = 38), (O) tumors in the body (n = 28), (P) tumors in the antrum (n = 30), (Q) diffuse-type Lauren classification (n = 31), (R) intestinal-type Lauren classification (n = 29), (S) stage I&II (n = 31), and (T) stage III&IV (n = 34)

We further validated these findings using the GSE13861 dataset. Stratified analyses for Sex = Female (Fig. 3K), Sex = Male (Fig. 3L), Age ≤ 60 (Fig. 3M), Age > 60 (Fig. 3N), Location = Body (Fig. 3O), Location = Antrum (Fig. 3P), Lauren = Diffuse (Fig. 3Q), Lauren = Intestinal (Fig. 3R), Stage = I&II (Fig. 3S), and Stage = III&IV (Fig. 3T) again revealed that the RRGS-High subgroup generally had poorer survival. However, in the Sex = Female, Location = Antrum, and Stage = I&II categories, these differences did not reach statistical significance. Taken together, these results highlight the strong risk stratification capability of this model.

RRGS associates with stromal-enriched immune microenvironment and metastatic signaling activation

To evaluate the potential of the RRGS as a surrogate marker for the tumor immune microenvironment, we further examined its correlations with ESTIMATE-derived scores, including the Immune score, Stromal score, and the overall ESTIMATE score. In the GSE26901 dataset, patients in the RRGS-high group exhibited a significantly higher Stromal score than those in the RRGS-low group (Fig. 4A). Similarly, in the GSE13861 dataset, patients with high RRGS not only demonstrated markedly elevated Stromal scores but also showed significantly increased ESTIMATE scores compared to the RRGS-low group (Fig. 4B). These findings indicate that a pronounced stromal component is associated with high RRGS across both datasets. Subsequently, we employed the CIBERSORT algorithm to assess immune cell infiltration in gastric cancer samples. The results from the GSE26901 dataset revealed that, relative to the RRGS-low group, the RRGS-high group had a significant increase in the infiltration of M2 macrophages and resting mast cells, while the levels of plasma cells, follicular helper T cells, and M0 macrophages were significantly decreased (Fig. 4C). Consistent with these findings, analysis of the GSE13861 dataset showed that the RRGS-high group exhibited a significant augmentation in M2 macrophages and resting mast cells, accompanied by a notable reduction in regulatory T cells (Tregs) infiltration (Fig. 4D). Collectively, these results suggest that the RRGS-high group is characterized by a distinct immune microenvironment, where the accumulation of pro-tumorigenic immune cells may contribute to adverse clinical outcomes.

Fig. 4.

Characterization of tumor microenvironment and signaling pathways associated with RRGS in gastric cancer. (A&B) Comparison of Stromal, Immune, and ESTIMATE scores between RRGS-high and RRGS-low groups in (A) GSE26901 (n = 109) and (B) GSE13861 (n = 65) datasets. (C&D) Relative abundance of immune cell populations in RRGS-high and RRGS-low groups, analyzed using CIBERSORT algorithm in (C) GSE26901 and (D) GSE13861 datasets. (E) Heatmap showing correlation between RRGS and key Hallmark signaling pathways, highlighting associations with Hypoxia, EMT, and UV response pathways in both datasets

To further elucidate the molecular mechanisms underpinning the RRGS signature, we performed Gene Set Variation Analysis (GSVA) to investigate its association with key Hallmark signaling pathways. Notably, in both the GSE26901 and GSE13861 datasets, the Hypoxia, EMT, and UV response downregulation pathways exhibited correlation coefficients (R) exceeding 0.5 with the RRGS (Fig. 4E). This robust correlation suggests that these three pathways may play key roles in mediating the unfavorable prognosis observed in patients with high RRGS.

ACTG2 is identified as a potential prognostic biomarker from RRGS signature in gastric cancer

Building upon the identification of the 32-gene RRGS signature, we conducted a prioritization analysis of the top 10 most differentially expressed genes within this signature. Venn diagram analysis revealed five overlapping genes (SFRP2, THBS2, ACTG2, SCRG1, and SPARCL1) consistently ranked among the top DEGs across both the GSE26901 and GSE13861 datasets (Fig. 5A). Notably, four of these genes (SFRP2, THBS2, SCRG1, and SPARCL1) have been previously implicated in gastric cancer pathogenesis, whereas ACTG2 remains uncharacterized in this malignancy. Given this knowledge gap, we selected ACTG2 for focused mechanistic exploration.

Fig. 5.

Identification of ACTG2 as a prognostic biomarker in gastric cancer. (A) Venn diagram showing the overlap of top 10 differentially expressed genes between GSE13861 and GSE26901 datasets. (B) ACTG2 expression in normal (n = 15) versus cancer (n = 134) tissues (GSE29272). (C-D) ACTG2 expression across tumor stages in GSE13861 and GSE26899 datasets. (E-N) Kaplan–Meier survival analyses showing the association between ACTG2 expression and overall survival in multiple independent datasets (GSE16861, n = 65; GSE14208, n = 80; GSE15459, n = 200; GSE26899, n = 90; GSE26901, n = 109; GSE62254, n = 300; GSE84426, n = 80; GSE84433, n = 150; TCGA, n = 350; GSE84437, n = 400). (O) Progression-free survival analysis by ACTG2 expression in GSE14208 (n = 80). (P-R) Relapse-free survival analyses by ACTG2 expression in GSE13861 (n = 65), GSE26899 (n = 90), and GSE26901 (n = 109) datasets

Transcriptomic analysis using the GSE29272 dataset demonstrated significantly elevated ACTG2 expression in gastric tumor tissues compared to adjacent normal tissues (Fig. 5B). Furthermore, ACTG2 exhibited a stage-dependent expression pattern, with progressive upregulation observed in advanced clinical stages across both the GSE13861 (Fig. 5C) and GSE26899 (Fig. 5D) datasets. To evaluate the clinical relevance of ACTG2 dysregulation, we performed survival meta-analysis spanning 10 independent datasets (GSE16861, GSE14208, GSE15459, GSE26899, GSE26901, GSE62254, GSE84426, GSE84433, TCGA, GSE84437). Strikingly, high ACTG2 expression correlated with significantly worse overall survival in all analyzed datasets (Figs. 5E-N). The prognostic significance extended to progression-free survival in the GSE14208 dataset (Fig. 5O) and relapse-free survival in the GSE13861, GSE26899, and GSE26901 datasets (Figs. 5P-R). These consistent findings across diverse clinical endpoints suggest that ACTG2 may serve as a robust prognostic biomarker in gastric cancer.

Functional validation of ACTG2 as a pro-tumorigenic driver in gastric cancer progression

To further clarify the role of ACTG2 on gastric cancer proliferation and metastasis in vitro and in vivo , we utilized two gastric cancer cell lines, AGS and SNU-1, to perform functional validations. Our goal was to understand the influences of ACTG2 silencing on the biological behaviors associated with gastric cancer malignancy. Experimental results indicated that ACTG2 knockdown markedly suppressed proliferation, migration, and invasion abilities of AGS cells (Fig. 6A-D). Consistent results were observed in SNU-1 cells, where diminishing ACTG2 expression similarly reduced proliferation, migration, and invasion capacities (Fig. 6E-H). We also performed ACTG2 overexpression experiments. As shown in Supplementary Fig. 2A-D, ectopic expression of ACTG2-OS in AGS cells significantly increased ACTG2 expression, proliferation, migration, and invasion compared to negative control cells. Similarly, ACTG2 overexpression in SNU-1 cells led to a significant increase in ACTG2 expression, proliferation, migration, and invasion (Supplementary Fig. 2E-H).

Fig. 6.

ACTG2 knockdown inhibits proliferation, migration, and invasion in gastric cancer cells. (A) ACTG2 expression levels after shRNA-mediated knockdown in AGS cells. (B) Cell proliferation assay of AGS cells after ACTG2 knockdown. (C) Wound healing assay showing reduced migration ability in ACTG2-silenced AGS cells. (D) Transwell invasion assay demonstrating decreased invasion capacity in ACTG2-knockdown AGS cells. (E–H) Corresponding experiments in SNU-1 cells showing similar effects of ACTG2 knockdown on expression levels (E), cell proliferation (F), migration (G), and invasion (H). n = 3 independent experiments; data are mean ± SD. *, p < 0.05; **, p < 0.01; ***p, < 0.001

Furthermore, we conducted experiments to assess the effects of ACTG2 overexpression following its knockdown. As shown in Supplementary Fig. 3A, ectopic expression of ACTG2-OS restored ACTG2 levels significantly and enhanced the proliferation, migration, and invasion (Supplementary Fig. 3B-D) abilities of AGS cells compared to the ACTG2-shRNA group. Similarly, in SNU-1 cells, ACTG2-OS overexpression led to a significant increase in ACTG2 expression and significantly promoted proliferation, migration, and invasion compared to the ACTG2-shRNA group (Supplementary Fig. 3 E–H).

These results collectively confirm that ACTG2 functions as a pro-tumorigenic driver in gastric cancer progression, whereby ACTG2 knockdown hinders malignant behaviors, and its restoration through overexpression reverses these effects.

To further validate our in vitro results using an animal model, we developed a zebrafish xenograft model implanted with AGS cells, aiming to evaluate tumor growth and metastatic tendency. Early observations at 2 h post-transplantation showed no evident difference in fluorescence intensity between ACTG2 knockdown (ACTG2-shRNA) and negative control (NC) groups, suggesting initially comparable tumor cell distribution. Nonetheless, at 48 h post-transplantation, fluorescence intensity markedly decreased in the ACTG2-shRNA group compared with the NC group (Fig. 7A), implying that silencing ACTG2 significantly inhibited tumor growth in the zebrafish model. Additionally, tail fluorescence imaging was conducted to assess metastatic capabilities of AGS cells (Fig. 7B). At the early stage (2 h post-transplantation), no significant difference was noted in fluorescence area at tail regions between the ACTG2 knockdown and NC groups. However, by the 24 h time point, the ACTG2 knockdown group exhibited a significant decline in tail fluorescence area compared with NC controls, indicating a clear inhibition of tumor cell migration following ACTG2 depletion. These in vivo data further reinforce our in vitro findings, suggesting ACTG2 may represent a promising therapeutic target to impede gastric cancer progression.

Fig. 7.

In vivo evaluation of ACTG2 knockdown in a zebrafish xenograft model. (A) Representative images and quantification showing decreased yolk fluorescence area at 48 h post-transplantation in ACTG2-shRNA groups compared to controls, indicating reduced tumor growth. (B) Tail fluorescence images and quantification reveal a significant reduction in metastatic spread at 24 h in ACTG2-knockdown groups relative to NC. n = 10 larvae per group; data are mean ± SD. NS, p > 0.05; *, p < 0.05; **, p < 0.01

ACTG2 overexpression correlates with chemoresistance and promotes 5-FU resistance in gastric cancer

In the GSE13861, GSE26901 and GSE66229 datasets, 49, 39, and 80 patients received chemotherapy treatment, respectively. We conducted a comparative analysis of ACTG2 expression in gastric cancer tissues from patients who experienced tumor recurrence after chemotherapy versus those who remained recurrence-free. Our findings revealed significantly elevated ACTG2 expression in tumor tissues from patients with recurrence compared to those without recurrence across both datasets (Fig. 8A&C). These results suggest a potential association between ACTG2 overexpression and chemoresistance in gastric cancer patients.

Fig. 8.

ACTG2 expression correlates with chemotherapy resistance in gastric cancer. (A-C) ACTG2 expression in tumor tissues from patients with or without recurrence after chemotherapy in GSE13861 (A, recurrence n = 25, non-recurrence n = 24), GSE26901 (B, recurrence n = 25, non-recurrence n = 14), and GSE66229 (C, recurrence n = 25, non-recurrence n = 55) datasets. (D-E) Effects of ACTG2 knockdown on 5-FU sensitivity in AGS (D) and SNU-1 (E) cells, showing dose–response curves (left, n = 6 technical replicates; data are mean ± SD) and IC50 values (right, n = 6 technical replicates; data are mean ± SD). *, p < 0.05; **, p < 0.01

To further investigate the functional role of ACTG2 in chemoresistance, we evaluated the effect of ACTG2 on 5-FU resistance in gastric cancer cells. Experimental results demonstrated that knockdown of ACTG2 significantly reduced the IC50 values of 5-FU in both AGS and SNU-1 cell lines (Fig. 8D&E), indicating increased sensitivity to 5-FU treatment. These findings provide compelling evidence that ACTG2 may contribute to chemoresistance mechanisms in gastric cancer and suggest that targeting ACTG2 could potentially enhance the efficacy of 5-FU-based chemotherapy regimens.

Elevated ACTG2 protein expression is associated with gastric cancer recurrence

IHC analysis was performed on an independent cohort to validate ACTG2 expression at the protein level. Consistent with transcriptomic data, ACTG2 immunoreactivity was significantly stronger in tumors from recurrent patients versus recurrence-free controls (Fig. 9A). Quantitative analysis confirmed a markedly higher positive staining rate in the recurrence group (Fig. 9B).

Fig. 9.

Immunohistochemical analysis of ACTG2 expression in recurrent vs. non-recurrent gastric cancer tissues. (A) Representative IHC images showing ACTG2 expression in tissue samples from recurrence-free (left, n = 3) and recurrent (right, n = 3) patients. (B) Quantitative analysis of the positive staining rate. Scale bar: 200 μm. data are mean ± SD, *, p < 0.05

Discussion

Gastric cancer remains an aggressive malignancy characterized by high recurrence rates and dismal long-term survival outcomes. In this study, we performed a comprehensive analysis of multiple gene expression datasets to identify recurrence-related genes, construct a predictive model for recurrence risk, and explore the biological significance of ACTG2-a gene highlighted by our RRGS. Our results provide additional insights into the pathophysiological mechanisms underlying gastric cancer progression, highlight the prognostic utility of RRGS, and suggest a potential role of ACTG2 in both tumorigenesis and chemoresistance.

While numerous gene signatures for gastric cancer have been developed in recent years, our RRGS may offer certain advantages in several aspects. Unlike many existing signatures that focus solely on overall survival, our 32-gene panel specifically targets recurrence prediction, a critical clinical challenge following curative treatment. This approach aligns with recent trends in molecular profiling for recurrence risk stratification, as highlighted in reviews of prognostic biomarker development [15]. The integration of our signature with immune infiltration analysis further distinguishes it from conventional prognostic tools by providing mechanistic context to the recurrence phenotype.

Molecular signatures like RRGS may contribute to precision oncology paradigms, where treatment decisions are guided by individual tumor biology rather than population-based statistics. As emphasized in recent frameworks for biomarker-based strategies [16], translating such signatures into clinical practice requires rigorous validation across diverse patient populations. Our validation of RRGS across independent cohorts and multiple clinical subgroups demonstrates its robust applicability, though further prospective studies will be necessary for clinical implementation.

The clinical utility of biomarker discovery extends beyond prognostication to drug development and therapeutic selection [17]. By identifying ACTG2 as a key component of the recurrence signature, our study not only provides a potential prognostic marker but also suggests that ACTG2 may represent a candidate for therapeutic exploration.

Our functional validation suggests that ACTG2 may function as a critical driver of gastric cancer progression and 5-FU resistance. These findings align with recent studies implicating actin cytoskeleton components in cancer aggressiveness across multiple tumor types. For instance, ACTG2’s role in hepatocellular carcinoma migration and its involvement in bladder cancer cell behavior suggest conserved functions across different malignancies.

The relationship between ACTG2 expression and chemoresistance warrants particular attention, especially in light of emerging concepts of cancer cell plasticity. Tumor cells can dynamically switch phenotypes to survive therapeutic pressures [18]. ACTG2-mediated cytoskeletal remodeling may facilitate such adaptive responses, enabling subpopulations to transiently adopt drug-tolerant states. This perspective is further supported by recent CRISPR screening studies identifying novel resistance mediators [19], highlighting the importance of cytoskeletal proteins in therapeutic resistance across cancer types.

The clinical relevance of ACTG2 in chemoresistance is further supported by our immunohistochemical analysis of patient tissues. In an independent cohort, we confirmed that ACTG2 protein expression is significantly higher in gastric cancer specimens from patients who experienced recurrence after adjuvant chemotherapy, compared to matched recurrence-free patients. This histopathological validation at the protein level supports the clinical phenotype and corroborates our transcriptomic and functional findings, suggesting that ACTG2 may serve not only as a molecular driver but also as a tangible tissue biomarker associated with treatment failure.

The mechanisms by which ACTG2 confers 5-FU resistance in gastric cancer likely involve multiple cellular processes. While our study demonstrates a clear phenotypic effect, future investigations should explore whether resistance occurs through altered drug uptake, enhanced efflux, modified apoptotic thresholds, or metabolic adaptations. Beyond classical apoptotic pathways, ACTG2 might influence alternative regulated cell death modalities that are increasingly recognized as critical in cancer biology and treatment response.

Ferroptosis, an iron-dependent form of regulated cell death characterized by lipid peroxidation, represents one such pathway that may intersect with ACTG2-mediated effects, particularly given the role of cytoskeletal proteins in cellular redox homeostasis [20]. Similarly, recent pan-cancer analyses have highlighted the relevance of copper-induced cell death (cuproptosis) in tumor progression and therapeutic response [21, 22]. The newly described disulfidptosis, driven by aberrant disulfide stress, may also intersect with cytoskeletal regulation [23]. Additionally, autophagy represents another critical pathway potentially modulated by cytoskeletal dynamics [24]. Investigating these non-apoptotic death mechanisms in relation to ACTG2 expression could reveal novel therapeutic vulnerabilities in recurrent gastric cancer.

Our findings demonstrated that high RRGS scores, including elevated ACTG2 expression, correlated with distinct immunological and stromal profiles. Specifically, the enrichment of M2 macrophages in high-RRGS tumors suggests an immunosuppressive microenvironment that facilitates tumor progression and therapeutic resistance. This observation aligns with the growing recognition that cancer progression relies not only on intrinsic tumor cell properties but also on complex interactions with surrounding stromal components [25].

The tumor-stroma interaction represents a critical determinant of gastric cancer behavior, particularly in the context of local invasion and metastatic dissemination. Cancer-associated fibroblasts (CAFs) play central roles in modifying the extracellular matrix to create permissive niches for tumor cell migration and invasion [26]. Extracellular vesicles facilitate bidirectional communication between tumor cells and CAFs, potentially involving ACTG2 or its downstream effectors. Given the enrichment of focal adhesion and ECM-receptor pathways in our recurrence-associated gene set, ACTG2 may function at the interface between tumor cells and their microenvironment to promote invasion and therapy resistance.

The tumor microenvironment in gastric cancer is further shaped by Helicobacter pylori infection, a well-established risk factor for gastric carcinogenesis. Recent studies have demonstrated that H. pylori facilitates cancer cell migration and impacts clinical outcomes [27]. The inflammatory milieu induced by H. pylori may synergize with ACTG2-mediated cytoskeletal changes to enhance tumor cell motility and invasive capacity. Interestingly, natural products targeting H. pylori, such as compounds derived from Terminalia chebula Retz, have shown promise in inhibiting adhesion and inflammatory responses [28, 29]. These agents may counteract the adhesive and invasive phenotypes associated with high ACTG2 expression, offering novel therapeutic strategies for high-risk patients.

The identification of ACTG2 as a driver of gastric cancer progression and chemoresistance suggests potential directions for future therapeutic exploration. Similar to other biomarker-driven approaches [30], targeting ACTG2 or its downstream effectors could sensitize tumors to conventional chemotherapy or prevent recurrence after curative treatment.

Recent advances in understanding the non-classical roles of membrane-associated proteins in cancer highlight potential synergies with ACTG2-directed therapies. For instance, sodium channel proteins have emerged as important mediators of cancer cell motility and invasion, potentially interacting with cytoskeletal components like ACTG2 to facilitate metastatic dissemination. Similarly, the identification of mitotic checkpoint kinases as pan-cancer biomarkers [31]and CENPA as a potential therapeutic target [32] suggests that combining ACTG2 inhibition with cell cycle modulators might enhance therapeutic efficacy.

An intriguing therapeutic approach might involve repurposing existing drugs that influence cytoskeletal dynamics. Local anesthetics, for example, have shown promise in targeting cancer stem cells and modulating invasive behavior [33, 34]. These agents could potentially disrupt ACTG2-dependent processes and enhance the efficacy of conventional chemotherapy in gastric cancer patients.

Despite the comprehensive nature of our analyses, several limitations warrant consideration. First, although we utilized multiple independent datasets, retrospective analyses are inherently limited by potential biases in data collection, patient selection, and variations in microarray or sequencing platforms. Prospective validation of the RRGS signature in large-scale multicenter clinical trials would be essential before full clinical integration. Second, our functional experiments focused on a subset of genes from the signature (chiefly ACTG2), leaving open the possibility that other RRGS components could exert equally significant or synergistic effects on tumor progression and recurrence. Future studies are needed to determine the relative contributions of these additional genes. Third, we acknowledge that the observed phenotypic effects of ACTG2 knockdown were modest and that rescue experiments using shRNA-resistant ACTG2 re-expression were not performed. This omission limits our ability to definitively establish that the phenotypes are solely attributable to ACTG2 loss rather than off-target effects. Fourth, while our zebrafish xenograft model recapitulates certain aspects of tumor growth and dissemination, it differs from mammalian systems in anatomy and immune composition, and drug metabolism. Confirmatory studies in mouse xenograft models or patient-derived organoids would strengthen the translational relevance of our findings. Fifth, our mechanistic interrogation of ACTG2-driven chemoresistance was primarily limited to 5-FU; it remains unclear whether ACTG2 confers resistance to other commonly used agents, such as platinum-based or taxane-based therapies. Additionally, while we observed phenotypic changes following ACTG2 modulation, the precise molecular mechanisms through which ACTG2 exerts its effects-particularly its downstream signaling pathways and potential interactions with other cytoskeletal proteins-remain to be fully elucidated. Lastly, gastric cancer is a highly heterogeneous disease, and our findings may not be generalizable to all molecular subtypes. Further studies incorporating diverse patient populations and molecular classifications are needed to validate the broader applicability of our observations.

Conclusion

In conclusion, this study provides additional insights into the molecular determinants of gastric cancer recurrence and suggests that RRGS signature, especially ACTG2, may have prognostic and therapeutic relevance. Further prospective studies are needed to validate these findings and explore their clinical applicability.

Author contributions

Xiuli Zhang, Yining Xu, Shan Nie, Minghan Huang, Ruochen Sun carried out the bioinformatic analyses. Xiuli Zhang, Xin Yang, Qingshui Wang conducted the experiments. Xiuli Zhang, Qingshui Wang, Yao Lin supervised the work and wrote the manuscript.

Funding

This work was funded by the National Natural Science Foundation of China (82174365), the Fujian Province Science and Technology Innovation Joint Funding Program (2024Y9487), the Education and Research Project for Young and Middle-aged Teachers of Fujian Education Department (JAT231036), the Youth Science and Technology Innovation Talent Cultivation Program of FJTCM (XQC2023007), the Fujian Province Higher Education Industry-University-Research Joint Innovation Project (2023N5012), the major Scientific Research Special Project of Fujian Province (2024YZ034026) and Fujian University of Traditional Chinese Medicine (X2024037).

Data availability

The datasets used to support the conclusion of this study was collected from publicly available database in GEO database (https://www.ncbi.nlm.nih.gov/geo/).

Declarations

Competing interests

The authors declare no competing interests.