Post-transcriptional activation of TGF-β signaling via SMAD mRNA stabilization by EIF4A3 promotes EMT and metastasis in gastric cancer

Abstract

Background

Gastric cancer (GC) is characterized by high mortality due to late-stage diagnosis and aggressive metastatic behavior. Epithelial-mesenchymal transition (EMT), primarily driven by the TGF-β/SMAD signaling, critically contributes to GC progression. However, the molecular mechanisms regulating this pathway remain incompletely understood.

Methods

We investigated the expression, function, and regulatory mechanisms of EIF4A3 in GC using patient tissues, GC cell lines, and in vivo mouse models. We performed loss-of-function experiments and xenograft studies to assess the biological effects of EIF4A3. Mechanistic studies were applied for exploring the relation between EIF4A3 and TGF-β/SMAD signaling.

Results

EIF4A3 upregulation was observed in GC samples as well as cell lines compared to normal counterparts, and indicated advanced clinical stages and poor clinical outcomes. EIF4A3 knockdown markedly repressed GC cell growth and DNA synthesis in vitro, while restraining tumor growth and Ki67 expression in xenograft models. Silencing EIF4A3 impaired cell migration, invasion, and lung metastasis, and reversed TGF-β induced EMT marker expression. Mechanistically, TGF-β1 transcriptionally induced EIF4A3 via SMAD2/3 binding to its promoter. In turn, EIF4A3 enhanced TGF-β/SMAD signaling by directly binding to and stabilizing SMAD2, SMAD3, and SMAD4 mRNAs, prolonging their half-lives and promoting their phosphorylation and nuclear accumulation. Furthermore, EIF4A3 knockdown reduced SMAD2/3 phosphorylation, nuclear translocation, and downstream EMT marker expression, thereby attenuating the pro-metastatic effects of TGF-β signaling.

Conclusions

Our study identifies EIF4A3 as a critical post-transcriptional regulator and downstream effector of the TGF-β/SMAD pathway. EIF4A3 forms a positive feedback loop with TGF-β/SMAD signaling, thereby reinforcing EMT and promoting metastatic progression in GC. These findings establish EIF4A3 as a novel post-transcriptional amplifier of TGF-β signaling and a promising candidate for diagnostic and therapeutic intervention in metastatic GC.

Supplementary information

The online version contains supplementary material available at 10.1007/s10565-026-10171-6.

Introduction

Gastric cancer (GC), a prevalent malignancy globally, accounting for the fifth highest incidence and the fourth highest mortality among all malignancies (Sung et al. 2021). Though the incidence shows a decrease over recent decades, GC remains a significant public health burden, particularly in East Asian countries, where it has the highest prevalence and is more commonly diagnosed in men (Zhang et al. 2024). In 2020, China alone accounted for approximately 44.0% of newly diagnosed GC cases and 48.6% of GC-associated death cases worldwide, making it the third most common malignant tumor in the country (Ilic and Ilic 2022; Zou et al. 2015). The miserable outcomes of GC is closely related to its insidious onset and absence of specific early clinical symptoms. As a result, late-stage diagnosis is common among patients, with only 20% being eligible for curative surgical resection due to locally advanced disease or distant metastases. Moreover, GC is marked by aggressive local invasion, frequent recurrence, and high metastatic potential, contributing to poor long-term survival outcomes. Despite significant progress in cancer therapies, the five-year survival rate is still below 30% (Ilic and Ilic 2022; Zou et al. 2015). Thus, elucidating novel mechanistic links with GC progression and aggressiveness is a critical step for innovating therapeutic strategies.

Epithelial-mesenchymal transition (EMT) critically participates in GC progression, contributing to tumor invasion, metastasis, and recurrence.(Ma et al. 2023). TGF-β, Wnt, and Notch pathways are revealed to modulate the EMT process, with TGF-β being a key mediator of tumor fibrosis and progression (Ilic and Ilic 2022; Zou et al. 2015). Upon activation through ligand-receptor binding, TGF-β initiates the canonical signaling cascade, in which SMAD2/SMAD3 are phosphorylated, subsequently form a complex with SMAD4 and translocate into the nucleus, transcriptionally regulating downstream effectors, including fibronectin 1 (FN1) (Zhang et al. 2021a; Itoh et al. 2019). Additionally, specific phosphorylated isoforms of Smad2/3 pSmad2C/L and pSmad3C/L mediate distinct signaling events associated with tumor suppression, fibrosis, or tumorigenesis (Yoshida et al. 2014; Watanabe et al. 2019). Given its central role in EMT and tumor dissemination, the TGF-β/SMAD signaling axis represents a critical therapeutic target in GC.

Despite the well-established role of TGF-β/SMAD pathway in GC, the mechanisms regulating its activation and feedback control remain poorly defined. Emerging evidence suggests that eukaryotic initiation factor 4A3 (EIF4A3) may play a critical role in this regulatory network (Zhao et al. 2021). EIF4A3 is an essential RNA helicase within the exon junction complex (EJC), and participates in multifaceted regulation of mRNA metabolism, including splicing, surveillance, export, and translation (Asthana et al. 2022; Chan et al. 2004). Increasingly, EIF4A3 has been implicated in cancer biology due to its role in stabilizing oncogenic transcripts and regulating alternative splicing. For example, in glioblastoma, EIF4A3 promotes tumor growth by regulating Notch and JAK-STAT3 signal pathways (Wei et al. 2023). In colorectal cancer, it enhances proliferation through the regulation of cell cycle-related splicing events (Kanellis et al. 2021). Similarly, in lung cancer, EIF4A3 enhances tumorigenesis and metastasis via stabilizing FOXM1 mRNA (Li et al. 2024). In GC specifically, EIF4A3 upregulation is observed, correlating with adverse clinical prognosis (Wu and Zhao 2021). Moreover, recent research has shown that EIF4A3 stabilizes oncogenic long non-coding RNAs such as circ_0008126, which promotes GC malignancy via the miR-502-5p/APC/β-catenin axis (Wang et al. 2025). Despite these insights, the potential role of EIF4A3 in modulating the TGF-β/SMAD axis and involvement of EIF4A3 in EMT remain largely unexplored.

In this study, we investigated the role of EIF4A3 in GC progression with a particular focus on its interaction with the TGF-β/SMAD axis and effects on EMT process. Our objective was to elucidate how EIF4A3 mediates EMT and metastatic progression in GC and to assess its value as a potential therapeutic target. We demonstrate that EIF4A3 forms a new positive feedback loop with the TGF-β/SMADs. This feedback loop facilitates EMT and tumor progression. These results offer novel mechanistic insight into the EMT regulation and identify EIF4A3 as a key post-transcriptional modulator and potential candidate for targeted therapy.

Materials and methods

Bioinformatics analysis

EIF4A3 levels were compared in STAD (Stomach adenocarcinomna) samples (n = 415) and normal samples (n = 34) or normal samples (n = 34) and different tumor grades (grade 1: n = 12; grade 2: n = 148; grade 3: n = 246) of STAD samples using UALCAN database (https://ualcan.path.uab.edu/). The transcripts per million (TPM) values of EIF4A3 were presented as minimum, 25th percentile, median, 75th percentile maximum, and the interquartile range (IQR). EIF4A3 protein levels in GC samples (n = 70) and normal samples (n = 34) were compared using CPTAC database (https://cprosite.ccr.cancer.gov/). Two GEO datasets (GSE33651, GSE65801) were applied for EIF4A3 expression validation in GC using GEO2R online tool. To assess the prognostic significance of EIF4A3 in GC, we analyzed The Cancer Genome Atlas Stomach Adenocarcinoma (TCGA-STAD) dataset using the GEPIA platform (http://gepia.cancer-pku.cn/). The association between EIF4A3 expression and overall survival (OS) in GC patients was evaluated using Kaplan–Meier survival analysis on the GEPIA database, with between-group comparisons evaluated by the log-rank test.

Clinical tissue collection

Paired GC samples and adjacent non-tumorous samples (at least 5 cm away from the tumor margin) were acquired from 23 patients with gastric adenocarcinoma during surgical resection or endoscopic biopsy at the Second Affiliated Hospital of Soochow University between January 2023 and August 2024. All samples were confirmed by histopathological examination by two independent pathologists. None had received chemotherapy, radiotherapy, or targeted therapy prior to sample collection. All specimens were instantly frozen in liquid nitrogen and stored at −80 °C. The relation between EIF4A3 levels and clinicopathologic parameters of GC patients was exhibited in Table 1. This study was approved by the Ethics Committee of the Second Affiliated Hospital of Soochow University and conducted following the ethical standards of the Declaration of Helsinki. All participants signed the written informed consent prior to sample acquisition.

Table 1.

Correlation between EIF4A3 expression and clinicopathologic features of GC patients

Variables	EIF4A3 expression		P
	High (n=12)	Low (n=11)
Age			0.537
≤55	5	6
> 55	7	5
Gender			0.510
Female	6	4
Male	6	7
Tumor size
≤ 5 cm	7	5	0.537
> 5 cm	5	6
Grade			0.147
I-II	3	6
III	9	5
T			0.552
T1-2	8	6
T3-4	4	5
N			0.022
N0	3	8
N1-3	9	3
M			0.033
M0	6	1
M1	6	10

Cell culture and treatments

The human normal gastric epithelial cell line GES-1 and GC cell lines (HGC-27, AGS, MKN-45) from the American Type Culture Collection (Manassas, VA, USA) were routinely maintained in RPMI-1640 medium (Gibco, USA) containing 10% fetal bovine serum (FBS; Gibco) and supplemented with 1% P/S (Invitrogen, USA) at 37 °C with an atmosphere of 5% CO₂. Cell lines underwent short tandem repeat (STR) identification, and were verified without mycoplasma contamination.

For TGF-β stimulation experiments, GC cells were grown in a 6-well plate and exposed to recombinant human TGF-β1 stimulation (cat# 100–21-10UG; Invitrogen; Thermo Fisher Scientific, USA) at final concentrations of 0.5, 1, 2, or 5 ng/mL for 24 h. To assess the role of TGF-β signaling, cells were co-stimulated by TGF-β1 (5 ng/mL) and the selective TGF-β receptor I kinase inhibitor SB525334 (10 μM; cat# S1476; Selleck Chemicals; TX, USA) for 24 h. Untreated cells cultured under identical conditions served as controls. All treatments were performed in serum-reduced medium (0.5–1% FBS) unless otherwise indicated to minimize confounding effects from serum-derived cytokines.

AGS and MKN-45 cells stably expressing doxycycline-inducible pLVX EIF4A3 (oe-EIF4A3) or empty vector (oe-NC) were generated via lentiviral transduction as previously described (Sakellariou et al. 2021; Tiessen et al. 2019). In some experiments, pLVX EIF4A3 expressing AGS and MKN-45 cells were treated with 100 ng/ml doxycycline (DOX) for 72 h.

Cell transfection

To investigate the functional role of EIF4A3 in GC, AGS and MKN-45 cells received transfection of either a non-targeting control vector (sh-NC) or short hairpin RNAs targeting EIF4A3 (sh-EIF4A3-1 and sh-EIF4A3-2) with Lipofectamine 3000 (Cat# L3000015; Invitrogen, USA). For EIF4A3 overexpression, EIF4A3 overexpression vector (oe-EIF4A3) and empty pcDNA vector (oe-NC) were transfected into GC cells or GES-1 cells. All plasmids were obtained from Geno Meditech (Shanghai, China). For TGF-β pathway modulation, cells were transfected with shRNA constructs targeting TGFβR1, TGFβR2, or SMAD2/3/4, alongside the sh-NC control, under the same transfection conditions. Cells were seeded in 6-well plates at 60–70% confluence 24 h before transfection. Plasmid DNA and Lipofectamine 3000 reagents were separately diluted in Opti-MEM medium (Gibco, USA) and mixed for complex formation before being added to the cells. After 6 h, the medium was replaced with complete growth media, and cells were maintained for an additional 48-h post-transfection. The probe sequences are given in supplementary Table 1.

Actinomycin D RNA stability assay

To elucidate the impact of EIF4A3 on the stability of SMAD2, SMAD3, and SMAD4 mRNAs, GC cells were stimulated by 5 µg/mL actinomycin D (ActD; Cell Signaling Technology, MA, USA; Cat# 15021 s), a transcriptional inhibitor that blocks RNA polymerase II activity.

In the first experimental setup, GC cells were stimulated by 5 ng/mL TGF-β1 with or without actinomycin D (5 µg/mL) to examine the transcriptional regulation and temporal stability of EIF4A3 mRNA. Total RNA was harvested 24 h post-treatment, followed by cDNA synthesis and qRT-PCR analysis to quantify EIF4A3 expression.

In the second assay, after transfection with either sh-EIF4A3 or sh-NC for 48 h., GC cells were treated with actinomycin D (5 µg/mL), and total RNA was isolated at the indicated time points (0, 6, 12, 18, and 24 h), and the decay of SMAD2, SMAD3, and SMAD4 transcripts was quantified using PCR analysis.

Quantitative real-time PCR (qrt-PCR)

TRIzol® (L3000075; Invitrogen, USA) was applied for total RNA extraction from tissue samples and cells following the manufacturer’s protocol. NanoDrop spectrophotometer (Thermo Fisher Scientific) was used for assessing RNA integrity and concentration. Then complementary DNA (cDNA) was synthesized using the PrimeScript™ RT Reagent Kit with gDNA Eraser (Takara Bio, Japan), following the manufacturer’s protocol. Gene-specific primers were designed using Primer-BLAST (NCBI) and validated for amplification efficiency and specificity via melt-curve analysis. PCR was carried out using SYBR Green Master Mix (Takara Bio, Japan) on a StepOnePlus Real-Time PCR System (Applied Biosystems, USA). Quantification of gene expression levels was conducted with the 2^−ΔΔCT method, and normalized to endogenous control GAPDH. Sequences of primers were listed in Supplementary Table 1.

Western blotting analysis

Total protein extraction was conducted with a radioimmunoprecipitation assay (RIPA) lysis buffer (89,901; Thermo Fisher Scientific) added with protease and phosphatase inhibitor cocktails (Roche, Switzerland). A bicinchoninic acid (BCA) assay kit (MITOISO2-1KT; Sigma-Aldrich, Burlington, MA, USA) was applied for protein concentration quantification. Equal amount of total protein (20 µg) was denatured in SDS loading buffer, separated via SDS-PAGE, followed by transferring onto nitrocellulose membranes (Millipore, USA) using a wet transfer system. Then a blocking step with 5% non-fat dry milk was conducted. Following blocking, membranes were maintained overnight at 4 °C with specific primary antibodies against the target proteins diluted in TBST containing 5% BSA. The next day, after TBST washing, corresponding secondary antibodies were added for 1-h incubation at ambient temperature. Protein bands were detected using enhanced chemiluminescence (ECL) reagents (89,901; Thermo Fisher Scientific, USA), and images were acquired using a chemiluminescence imaging system (iBright CL750 Imaging System; Thermo Fisher Scientific, USA), with ImageJ software (NIH) for quantification. The antibody sources are provided in Supplementary Table 2.

Cell counting kit-8 (CCK-8) assay

A Cell Counting Kit-8 (CCK-8; Dojindo Laboratories, Cat# CK04) was applied for cell viability assessment following the producer’s protocol. Cells were grown in 24-well plates (1000 cells/well) in complete RPMI-1640 medium overnight at 37 °C. Cells post indicated transfection were cultured for 1, 2, 3, 4, and 5 days, and then each well was added with 10 µL of CCK-8 solution into the culture medium (100 µL), followed by incubating for 4 h at 37 °C. A microplate reader (Benchmark SmartReader™; Sigma-Aldrich, USA) was applied for detecting the OD value at 450 nm.

Colony formation assay

The colony formation assays were carried out for assessing the viability of AGS and MKN-45 GC cells. Cells transfected with indicated vector were inoculated into 6-well plates (2000 cells/well). Cells were maintained in culture for 1–2 weeks, with the culture medium replaced every 3 days to maintain optimal conditions. Next, colonies were immersed for 15 min in 4% paraformaldehyde (PFA), dyed using 0.5% crystal violet solution for thirty min at ambient temperature, and then rinsed using distilled water. The plates were air-dried, and images were captured using an Epson scanner. Colony numbers were counted with ImageJ software, and statistical analysis was performed to compare colony formation between groups.

5-ethynyl-2′-deoxyuridine (edu) incorporation assay

Cell proliferation was evaluated using the EdU incorporation assay. Briefly, AGS and MKN-45 cells post indicated transfection were grown on glass coverslips overnight in a complete culture medium. Next, 50 µM EdU solution (RiboBio, China) was added for incubation with cells for 2 h at 37 °C, following the manufacturer's protocol. After incubation, cells were fastened using 4% PFA for fifteen min, underwent permeabilization for 10 min using 0.5% Triton X-100 and staining with a Click-iT EdU Alexa Fluor 488 Imaging Kit (Thermo Fisher Scientific, USA). Nuclei were counterstained with DAPI (4′,6-diamidino-2-phenylindole). Fluorescence images were captured using a fluorescence microscope (Zeiss Axio Imager or equivalent), and six randomly selected fields per sample were analyzed. ImageJ software was used to quantify the percentage of EdU-positive cells (green fluorescence) relative to the DAPI-stained total cells (blue fluorescence) using.

Wound-healing assay

The wound healing assay was conducted to evaluate cell migration ability. AGS and MKN-45 cells were grown in 6-well plates and maintained until 80–90% confluence. After 24 h of incubation in complete medium, the monolayer was scratched using a sterilized 200-µL pipette tip, creating a uniform wound. The wells were gently rinsed using D-Hank’s solution, and fresh serum-free medium was supplemented. Cells were incubated for 24 h, and images were captured at 0 and 24 h under an inverted microscope, and wound closure was evaluated by detecting the remaining wound area using ImageJ software.

Transwell migration and invasion assays

The 24-well Transwell inserts (pore size: 8 μm, Corning, USA) were applied for Transwell assays. A Transwell migration assay was performed to evaluate the migratory ability of AGS and MKN-45 GC cells. Briefly, cells were trypsinized, centrifuged, followed by resuspending in serum-free RPMI-1640 media. Then 200 μL of GC cell suspension (1 × 10⁶ cells/mL) was supplemented in the top chamber of a Transwell insert, and 800 μL of RPMI-1640 medium supplemented with 30% FBS was filled into bottom chamber. Cells were incubated at 37 °C in a 5% CO₂ humidified incubator for 24 h. After incubation, a cotton swab gently removed cells retained on the upper membrane surface, while the migrated cells were fastened for 15 min with 4% PFA, dyed with crystal violet solution (0.5%) for 30 min at ambient temperature, and rinsed using PBS. Cells were visualized and quantified under an inverted phase-contrast microscope (Keyence, JAPAN) by counting five randomly selected fields per well.

A Transwell invasion assay was carried out similarly to the migration assay, with the addition of a Matrigel-coated barrier to mimic the extracellular matrix. 100 μL of Matrigel (1 mg/mL, Corning, USA) was evenly spread onto the upper chamber membrane and allowed to polymerize at 37 °C for 1 h before cell seeding. Subsequently, cell suspension was added as described above. After incubation for 48 h to allow invasion through the Matrigel layer, non-invading cells were removed, and the invaded cells on the lower surface were fastened, stained, and quantified using the same procedure as described for the migration assay.

Immunofluorescence (IF) staining

GC cells, transfected or untransfected with the sh-EIF4A3 vector, were immersed in 4% PFA for 15 min. Then 0.5% Triton X-100 was added for permeabilization for 10 min, and 5% BSA was used to block nonspecific binding for 60 min. Next, after incubation overnight with anti-SMAD2/3/4 at 4 °C, secondary fluorescently labeled antibodies were added and maintained at ambient temperature for 60 min. DAPI was adopted to counterstain the nuclei, and a fluorescence microscope (DMi8; Leica, Germany) was applied to capture the images. The information of antibodies used was provided in Supplementary Table 2.

Chromatin immunoprecipitation (chip) assay

MKN-45 and AGS GC cells, treated or untreated with TGF-β1, were subjected to ChIP assay. In short, cells were fixed by incubation with 1% formaldehyde for 10 min at ambient temperature. The crosslinking reaction was terminated by adding 125 mM glycine, and lysed to obtain chromatin fragments via sonication. Chromatin extract was immunoprecipitated for one night at 4 °C with anti-SMAD2/3/4 or anti-IgG as a negative control. Immunoprecipitates were collected via Protein A/G magnetic beads, washed, and subjected to reverse crosslinking at 65 °C overnight. Purified DNA was amplified using polymerase chain reaction (PCR) with EIF4A3-specific primers. The antibodies used were presented in Supplementary Table 2.

Luciferase reporter assay

The transcriptional regulation of EIF4A3 by the TGF-β/SMAD axis in GC cells was assessed using luciferase reporter assays. Two separate assays were conducted to evaluate EIF4A3 promoter activity and its regulation by SMAD proteins. In the first assay, AGS and MKN-45 cells were grown into a 24-well plate with 1 × 10⁵ cells/well and transfected with the EIF4A3 promoter luciferase reporter vector (500 ng/well) using Lipofectamine 3000 following the producer’s guide. After a 6-h incubation, the medium was refreshed, and cells were exposed to 2 ng/mL TGF-β1 for 24 h to assess EIF4A3 transcriptional activation.

In the second assay, to investigate the role of SMAD proteins in regulating EIF4A3 transcription, cells were co-transfected with shRNA vectors targeting SMAD2, SMAD3, or a non-targeting shRNA control (sh-NC) along with the EIF4A3 luciferase reporter vector (500 ng/well) using Lipofectamine 3000. After 6 h, the transfection medium was replaced, and cells were incubated for an additional 18 h in the presence or absence of TGF-β1 (2 ng/mL) stimulation. Following a total incubation period of 24 h, cells were subjected to passive lysis buffer (Promega, USA), with a GloMax 20/20 Luminometer (Promega, USA) for luciferase activity quantification.

RNA immunoprecipitation (RIP) assay

To analyze EIF4A3 interactions with SMAD2/3/4 mRNA, RIP was performed. Lysates of AGS and MKN-45 cells (1 × 10⁷) were prepared, and centrifuged for harvesting the supernatant. Protein A/G magnetic beads were pre-cultured with 5 µg of anti-EIF4A3 or IgG control antibody at 25 °C for 1 h. The cell lysate was then used to incubate with the antibody-bead complex for a whole night at 4 °C. After three washes, 10% of the beads were collected to assess the efficiency of immunoprecipitation. The remaining immunoprecipitate was exposed to proteinase K for 30 min at 55 °C, followed by RNA purification using TRIzol reagent. SMAD2/3/4 mRNA enrichment was analyzed using qRT-PCR. The antibody sources are provided in Supplementary Table 2.

Xenograft mouse model

C57BL/6 J nude mice (6 weeks) were provided by the Qinglongshan Experimental Animal Centre (Jiangsu, China), and housed in a specific pathogen-free (SPF) facility under constant ambient temperature and humidity. Animals with free access to food and water ad libitum were allowed to acclimate for one week before experimentation. For subcutaneous xenograft tumor models, AGS cells were resuspended in PBS and 200 µL cell suspension (5 × 10⁶ cells), under indicated transfection (sh-EIF4A3, sh-NC), were administrated via subcutaneous injection into the abdominal cavity of mice (n = 5 per group). Tumor volume was observed every 7 days using vernier calipers, and quantified as V = (length × width²)/2. After four weeks, the mice were euthanized via CO₂ asphyxiation followed by cervical dislocation, and the tumors were excised. The tumors were photographed, weighed, and subsequently frozen at −80 °C for future analyses.

For a lung metastasis model, cell suspension containing 5 × 10⁶ AGS cells in 200 µL of PBS, transfected or untransfected with sh-EIF4A3, was administrated via lateral tail vein injection into C57BL/6 J nude mice (n = 4 per group). Mice were maintained for four weeks, after which lung metastases were assessed. The IVIS Spectrum Imaging System (PerkinElmer, USA) was applied for in vivo fluorescence imaging and detecting metastatic tumor burden. Mice were then euthanized, and lungs were excised. The number of lung metastatic nodules was manually quantified by two blinded investigators using a microscope. All animal studies were performed in compliance with the ARRIVE guidelines under the approval of the Animal Ethics Committee of our hospital.

Histopathological analysis

Lung tissue sections from mouse models were immersed in 4% PFA, paraffin-embedded, and 4-μm thick sections were prepared. After deparaffinizing with xylene and processing through a series of graded alcohol solutions for hydration, sections were subjected to hematoxylin dyeing for five min, followed by tap water rinsing, and eosin staining for 3 min. After dehydration, the stained sections were mounted using neutral resin. Microscopic examination was performed using a light microscope, and metastatic nodules were quantified using ImageJ software.

Immunohistochemistry (IHC) analysis

Paraffin-embedded tumor tissue slices were applied for immunohistochemical staining. Tissue Sects. (4-µm thick) after dewaxing and rehydration were heating in sodium citrate buffer (pH 6.0) at 95 °C for 15 min. Subsequently, 3% H₂O₂ in methanol was added for a 10-min incubation. After washing with phosphate-buffered saline (PBS, pH 7.4), sections were incubated overnight at 4 °C with primary antibodies specific to EIF4A3 and Ki67. Next, sections were washed and then maintained with corresponding secondary antibodies at ambient temperature for thirty min. Then DAB chromogenic substrate was added, followed by hematoxylin counterstaining. Finally, the slides were dehydrated, cleared, and mounted with neutral balsam for long-term preservation. A light microscope was applied for capturing the images, with ImageJ software for image analysis. The final IHC staining was scored on a scale of 0 to 4 considering positive staining and staining intensity, with higher scores denoting stronger immunostaining intensity. Supplementary Table 2 listed the antibody sources.

Statistical analyses

SPSS 21.0 (IBM, USA) and GraphPad Prism 8 (GraphPad Software, USA) were applied for statistical analyses and data presentation. Continuous data were presented as mean ± standard deviation (SD) of ≥ 3 independent replicates. Inter-group differences were evaluated with the Student’s t-test or one-way analysis of variance (ANOVA) with Tukey's test. Statistically significance was set as p < 0.05 (two-tailed).

Results

Eif4a3 is highly expressed in GC and is related to adverse clinical outcomes

We first delved into clinical relevance of EIF4A3 with GC by analyzing its expression in 23 paired GC and adjacent non-tumorous tissues. Both EIF4A3 mRNA and protein levels exhibited a significant upregulation in tumor samples compared to paracancerous counterparts (p < 0.01) (Fig. 1A-B). IHC staining further verified the elevated EIF4A3 protein expression in GC (p < 0.01) (Fig.  1 C). Consistent with our findings, analysis with CPTAC database displayed upregulated EIF4A3 protein levels in GC samples relative to normal samples (p < 0.0001) (Figure S1). Data from the UALCAN database also demonstrated higher levels of EIF4A3 in GC tumors, particularly in stage II and III patients (p < 0.01) (Fig. 1D-E). Data from two GEO datasets (GSE33651, GSE65801) also validated the upregulated levels of EIF4A3 in GC compared to the normal control samples (Figure S2A-B). Moreover, the high expression of EIF4A3 was correlated with reduced overall survival (OS) (p = 0.047) (Fig.  1 F). We also found that GC patients with high EIF4A3 level were also related to elevated N and M stage (Table 1). Furthermore, EIF4A3 mRNA expression and protein levels were significantly elevated in GC cell lines in comparison with normal gastric epithelial cells (p < 0.01) (Fig. 1G-H). These results indicate the upregulation of EIF4A3 is in GC, potentially linked with metastasis and poor prognosis.

Fig. 1.

EIF4A3 is highly expressed in gastric cancer and is associated with poor clinical outcomes. (A-B) qRT-PCR and Western blot analysis of EIF4A3 expression in 23 paired gastric cancer and adjacent non-tumor tissues.(C) Immunohistochemical (IHC) analysis of EIF4A3 expression in 23 gastric cancer tissues and corresponding paracancerous tissues; expression score was calculated by combining staining intensity and the percentage of positive cells (score range: 0–4).(D-E) UALCAN database analysis of EIF4A3 expression in gastric cancer (TCGA-STAD), stratified by tumor stage.(F) GEPIA survival analysis showing the correlation between EIF4A3 expression and overall survival in gastric cancer patients.(G-H) EIF4A3 expression in gastric epithelial cell line GES-1 and gastric cancer cell lines (HGC-27, AGS, and MKN-45) were analyzed by qRT-PCR and Western blot (n=3). Data are shown as the mean ± SD, and statistical analyses were performed using either unpaired Student’s t-test or one-way analysis of variance (ANOVA) with Tukey’s post-hoc test. *P < 0.05, **P < 0.01, ***P < 0.001

Eif4a3 promotes GC cell growth and tumorigenesis

We then delved into the functional role of EIF4A3 in GC, an EIF4A3 knockdown model was established in GC cells. Quantitative PCR analysis confirmed that EIF4A3 expression was significantly reduced following knockdown, validating the efficiency of the model (Fig.  2 A). Functionally, EIF4A3 deficiency reduced GC cell proliferation and DNA synthesis, as well as the clonogenic potential (p < 0.01) (Fig. 2B-D). Meanwhile, we explored the effects of EIF4A3 overexpression on GC cells and normal GES-1 cells. The overexpression of EIF4A3 was verified using qRT-PCR (p < 0.01) (Figure S3A). Notably, the viability and proliferation of GES-1 cells were not significantly changed by EIF4A3 overexpression (p > 0.05), while those of GC cells were significantly increased by overexpressing EIF4A3, suggesting the phenotype-dependent role of EIF4A3 (p < 0.01) (Figure S3B-D). To validate the impact of EIF4A3 on observed GC cell behaviors, we used a doxycycline-inducible expression system to reconstitute EIF4A3 in the knockdown cells, and the results showed that ectopic re-expression of EIF4A3 almost completely rescued the suppressive impact of sh-EIF4A3 on GC cell growth, migration and invasion, suggesting the causal link between EIF4A3 knockdown and inhibited malignant phenotypes of GC cells (p < 0.01) (Figure S4A-D). In vivo, xenograft mouse models injected with EIF4A3-deficient AGS cells exhibited smaller tumor volumes and reduced tumor weights compared to controls (p < 0.01) (Fig. 2E-G). IHC analysis further revealed decreased EIF4A3 and Ki67 expression in xenografts from sh-EIF4A3 group (p < 0.01) (Fig. 2H-I). Overall, EIF4A3 promotes both in vitro GC cell growth and in vivo tumorigenesis.

Fig. 2.

EIF4A3 promotes gastric cancer cell proliferation and colony formation in vitro and tumor growth in vivo. (A) Western blot detection of EIF4A3 expression in MKN-45 and HGC-27 cells after indicated transfection. (B) EdU assay was performed to analyze the effect of the knockdown of EIF4A3 on cell proliferation. (C) CCK-8 assay was used to analyze the effect of the knockdown of EIF4A3 on cell viability. (D) Clone formation assay was used to analyze the effect of the knockdown of EIF4A3 on cell proliferation. (E) Tumor size in nude mice injected with AGS cells transfected with either sh-NC (negative control) or sh-EIF4A3 (n=5). (F) Tumor growth curve. (G) Tumor weight. (H-I) IHC analysis of EIF4A3 and Ki67 expression in mouse tumor tissues (n=5). Data are shown as the mean ± SD, and statistical analyses were performed using either unpaired Student’s t-test or one-way analysis of variance (ANOVA) with Tukey’s post-hoc test. **P < 0.01

Eif4a3 enhances GC cell aggressiveness as well as lung metastasis

Given the association between EIF4A3 and adverse prognosis, we next explored its role in cancer cell motility and metastasis. Wound healing assays as well as Transwell assays showed reduced GC cell migration rate and invasiveness after EIF4A3 knockdown (p < 0.01) (Fig. 3A-C). IHC of xenograft tumors revealed that EIF4A3 deficiency elevated E-cadherin expression while reduced N-cadherin levels, suggesting a reversal of EMT (p < 0.01) (Fig. 3D-E). To further evaluate its role in metastasis in vivo, we employed a lung metastasis model. Mice injected with EIF4A3-deficient AGS cells exhibited significantly reduced metastatic lung nodules and lower bioluminescence intensity than control mice (p < 0.01) (Fig. 3F-I). Overall, EIF4A3 boosts GC cell metastatic potential through the promotion of EMT.

Fig. 3.

EIF4A3 promotes migration, invasion, and lung metastasis of gastric cancer cells.(A) Wound-healing assay assessing cell migration following EIF4A3 knockdown (n=3). (B-C) Transwell migration and invasion assays demonstrating reduced motility and invasiveness in EIF4A3-silenced cells (n=3). (D-E) IHC staining of E-cadherin and N-cadherin in subcutaneous tumors (n=5). (F) Fluorescence imaging of lung metastases in mice following tail vein injection of AGS cells transfected with either sh-NC (negative control) or sh-EIF4A3 (n=4). (G-I) Quantification of metastatic nodules in lung tissue sections by H&E staining (n=4).Data are shown as the mean ± SD, and statistical analyses were performed using either unpaired Student’s t-test or one-way analysis of variance (ANOVA) with Tukey’s post-hoc test. **P < 0.01

Eif4a3 expression is transcriptionally induced by the TGF-β/SMAD axis

Given the established link between TGF-β and EMT as well as metastasis (!!! INVALID CITATION!!! xxxx), we hypothesized that EIF4A3 might be regulated by this pathway. qRT-PCR analysis revealed that dose-dependent elevation in EIF4A3 levels in response to TGF-β1 treatment, with the highest expression observed at 5 ng/μL TGF-β1 (p < 0.001) (Fig. 4A-B). Time-course experiments further showed that EIF4A3 expression peaked at 24 h post TGF-β1 treatment (p < 0.01) (Fig. 4C-D). Correspondingly, luciferase reporter assays showed that TGF-β1 stimulation significantly increased EIF4A3 promoter activity in both dose- and time-dependent manners (p < 0.01) (Fig. 4E-F). To assess whether TGF-β induced EIF4A3 upregulation is dependent on transcriptional activation, cells were treated with actinomycin D, a transcriptional inhibitor. The results demonstrated that actinomycin D blocked TGF-β-induced EIF4A3 upregulation, suggesting that EIF4A3 is transcriptionally induced by TGF-β signaling (p < 0.01) (Fig. 4G-H). Furthermore, pharmacological inhibition of TGFβR1 using SB525334 or genetic knockdown of TGFβR1 or TGFβR2 significantly inhibited the TGF-β/SMAD activation (p < 0.01) (Figure S5A-B) and downregulated EIF4A3 expression (p < 0.01) (Fig. 4I-L), confirming that EIF4A3 is transcriptionally induced through the canonical TGF-β/SMAD pathway.

Fig. 4.

EIF4A3 transcription is induced by TGF-β/SMAD signalling. (A-B) qRT-PCR and Western blot analysis of EIF4A3 expression in gastric cancer cells treated with increasing doses of TGF-β for 24 h. (C-D) EIF4A3 expression after TGF-β (5 ng/mL) treatment for various time points. (E-F) Luciferase reporter assays showing dose- and time-dependent transcriptional activation of EIF4A3. (G-H) qRT-PCR and Western blot analysis of EIF4A3 mRNA and protein levels in AGS and MKN-45 cells left untreated or treated with TGF-β1 (5 ng/mL) alone or in combination with actinomycin D (5 µg/mL) for 24 h. (I-J) EIF4A3 expression in GC cells co-treated with TGF-β1 (5 ng/mL) and the selective TGF-β receptor I kinase inhibitor SB525334 (10 μM) for 24 h. (K-L) EIF4A3 expression in TGF-β untreated or treated cells (5 ng/mL for 24 h) after TGFβR knockdown.Data are shown as the mean ± SD of three independent experiments, and statistical analyses were performed using one-way analysis of variance (ANOVA) with Tukey’s post-hoc test. *P < 0.05, **P < 0.01, ***P < 0.001

Smad2/smad3 directly binds to eif4a3 promoter and activates eif4a3 transcription

To determine which SMADs mediate EIF4A3 induction, we selectively knocked down SMAD2, SMAD3, or SMAD4. The results revealed that SMAD2 and SMAD3 knockdown reversed TGF-β induced EIF4A3 expression, whereas SMAD4 knockdown had no significant effect (p < 0.01) (Fig. 5A-B). ChIP assay confirmed that TGF-β1 stimulation enhanced SMAD2 and SMAD3 binding to the EIF4A3 promoter, while SMAD4 did not bind to this region (p < 0.01) (Fig. 5C-E). These findings were further corroborated by promoter luciferase assays, where the knockdown of SMAD2 or SMAD3 significantly reduced the ability of TGF-β to activate the EIF4A3 promoter (p < 0.01) (Fig. 5F-G). Altogether, this data clearly demonstrates that SMAD2/SMAD3 directly binds to and activates the EIF4A3 promoter upon TGF-β stimulation.

Fig. 5.

SMAD2 and SMAD3 directly bind to and activate the EIF4A3 promoter. (A-B) qRT-PCR and Western blot analysis of EIF4A3 expression after knockdown of SMAD2, SMAD3, or SMAD4 in TGF-β-treated cells (5 ng/mL for 24 h). (C-E) ChIP-qPCR assays confirming SMAD2/3/4 binding to the EIF4A3 promoter, with or without TGF-β1 treatment. (F-G) Luciferase reporter assays showing that SMAD2 and SMAD3 are required for TGF-β1-mediated transcriptional activation of EIF4A3. Data are shown as the mean ± SD of three independent experiments, and statistical analyses were performed using one-way analysis of variance (ANOVA) with Tukey’s post-hoc test. ns; not significant; **P < 0.01

Eif4a3 contributes to TGF-β induced GC cell migration, invasion, and metastasis

We next explored whether EIF4A3 functionally mediates the pro-metastatic effects of TGF-β signaling, particularly its role in promoting EMT as well as cell aggressiveness in GC. In a series of wound healing and Transwell invasion and migration assays, TGF-β1 treatment significantly increased the invasive potential of AGS and MKN-45 cells, as expected. However, when EIF4A3 was knocked down, the TGF-β induced enhancement of invasion and migration was substantially suppressed. Specifically, fewer EIF4A3-deficient cells were able to migrate through the Matrigel-coated membrane despite TGF-β stimulation, indicating that EIF4A3 is required for the invasive phenotype induced by TGF-β (p < 0.01) (Fig. 6A-F). This suggests that EIF4A3 is not merely a downstream marker of TGF-β activation but a critical functional mediator of the pathway’s pro-invasive effects. To further substantiate this, we then examined changes in EMT-associated protein expression under TGF-β stimulation, with or without EIF4A3. As expected, TGF-β treatment upregulated mesenchymal markers while decreasing the E-cadherin levels, consistent with EMT induction. However, when EIF4A3 was knocked down, these TGF-β-driven EMT changes were largely reversed: epithelial markers were restored, and mesenchymal markers were suppressed (p < 0.01) (Fig. 6G-H). Taken together, these findings confirm that EIF4A3 is a key downstream effector of TGF-β/SMAD2/3 in GC.

Fig. 6.

EIF4A3 contributes to TGF-β induced gastric cancer cell migration and invasion. (A-B) Wound-healing assays evaluating the effect of EIF4A3 knockdown on TGF-β-induced cell migration. (C-D) Transwell migration assays in cells transfected with sh-EIF4A3 and treated with TGF-β (5 ng/mL for 24 h). (E-F) Transwell invasion assays under the same conditions. (G-H) Western blot analysis of epithelial–mesenchymal transition (EMT) markers (E-cadherin, N-cadherin, Vimentin, and Snail) in transfected GC cells left untreated or treated with TGF-β (5 ng/mL for 24 h). Data are shown as the mean ± SD of three independent experiments, and statistical analyses were performed using one-way analysis of variance (ANOVA) with Tukey’s post-hoc test. **P < 0.01

Eif4a3 promotes TGF-β/SMAD signaling activation by stabilizing smad2/3/4 mrnas

We next asked whether EIF4A3 could also feedback to enhance TGF-β/SMAD signaling. To this end, we used a luciferase reporter construct containing SMAD-binding elements (pSBE) to measure overall TGF-β/SMAD transcriptional activity. As expected, TGF-β1 stimulation robustly increased luciferase activity, reflecting enhanced SMAD signaling. Interestingly, EIF4A3 knockdown significantly reduced this activation (p < 0.01) (Fig.  7 A), suggesting that EIF4A3 might enhance SMAD pathway activity beyond merely being a transcriptional target. To explore this further, we assessed the gene expression and protein levels of SMAD2, SMAD3, and SMAD4 following EIF4A3 knockdown. Remarkably, all three SMAD gene expression and protein levels were reduced in EIF4A3-deficient cells, with upstream receptors TGFβR1 and TGFβR2 not significantly altered (p < 0.01) (Fig. 7B-D). This indicates that EIF4A3 specifically supports the expression of SMAD proteins, potentially forming a positive feedback loop that amplifies TGF-β/SMAD signaling. Additionally, we examined SMAD2/3 phosphorylation, which is required for their activation and nuclear translocation. Upon EIF4A3 knockdown, phosphorylated SMAD2 and SMAD3 levels were significantly reduced, even under TGF-β stimulation (p < 0.01) (Fig. 7E-F), indicating impaired pathway activation. Immunofluorescence imaging also revealed reduced nuclear accumulation of SMAD2, SMAD3, and SMAD4 in EIF4A3-silenced cells, further confirming that EIF4A3 is required for SMAD protein activation and nuclear translocation (Fig. 7G-I). Taken together, these findings suggest that EIF4A3 supports TGF-β/SMAD pathway activity by maintaining SMAD protein levels and facilitating their activation and nuclear localization.

Fig. 7.

EIF4A3 promotes TGF-β/SMAD signalling by upregulating SMAD2/SMAD3/SMAD4 levels. (A) Luciferase reporter assay of SMAD-binding element (SBE) activity in cells treated with TGF-β and sh-EIF4A3. (B-C) Western blot analysis of SMAD2, SMAD3, SMAD4, TGFβR1, and TGFβR2 expression following EIF4A3 knockdown. (D) qRT-PCR analysis of SMAD2/3/4 mRNA levels after EIF4A3 silencing. (E-F) Western blot analysis of total and phosphorylated SMAD2/3 following EIF4A3 knockdown and TGFβ treatment (5 ng/mL for 24 h). (G-I) Immunofluorescence staining showing nuclear localization of SMAD2/3/4 in TGF-β-treated gastric cancer cells after 6 h. Data are shown as the mean ± SD of three independent experiments, and statistical analyses were performed using either unpaired Student’s t-test or one-way analysis of variance (ANOVA) with Tukey’s post-hoc test. **P < 0.01

EIF4A3 stabilizes SMAD2/SMAD3/SMAD4 transcripts via direct binding

To dissect the underlying mechanism by which EIF4A3 regulates SMAD protein levels, we investigated its effect on SMAD mRNA stability. RIP assays revealed that EIF4A3 directly binds to SMAD2, SMAD3, and SMAD4 mRNAs in GC cells (p < 0.01) (Fig. 8A-C), suggesting a role in transcript stabilization. This was tested functionally using actinomycin D, a transcriptional inhibitor, to block new RNA synthesis and observe mRNA decay over time. EIF4A3 knockdown accelerated the degradation of SMAD2, SMAD3, and SMAD4 transcripts and significantly shortened their half-lives, as shown in decay curve analyses (p < 0.01) (Fig. 8D-F). These results demonstrate that EIF4A3 stabilizes SMAD mRNAs, thereby maintaining their protein expression and supporting active TGF-β/SMAD signaling. Moreover, rescue assays were carried out to explore the effects of SMAD2/SMAD3/SMAD4 overexpression on EIF4A3-silenced GC cells. We demonstrated that the reduction in GC cell migration rate was significantly restored by overexpressing SMAD2/SMAD3/SMAD4 (p < 0.01) (Figure S6A). Meanwhile, we evaluated the changes of EMT-related proteins in EIF4A3-silenced GC cells with SMAD2/SMAD3/SMAD4 overexpression. The E-cadherin protein levels were upregulated, while those of N-cadherin and Vimentin were downregulated in EIF4A3-silenced GC cells, and SMAD2/SMAD3/SMAD4 overexpression was revealed to reverse these alterations (p < 0.01) (Figure S6B), suggesting the critical involvement of this signaling in EIF4A3-meidated oncogenic effects in GC.

Fig. 8.

EIF4A3 interacts with and enhances the stability of SMAD2/SMAD3/SMAD4 mRNAs. (A-C) RNA immunoprecipitation (RIP) assays confirming EIF4A3 binding to SMAD2, SMAD3, and SMAD4 mRNAs. (D-F) mRNA stability assays showing reduced half-life of SMAD2/3/4 transcripts in AGS and MKN-45 cells upon EIF4A3 knockdown. Data are shown as the mean ± SD of three independent experiments, and statistical analyses were performed using unpaired Student’s t-test. **P < 0.01, ***P < 0.001

Discussion

This study identified EIF4A3 as a critical regulator of GC progression, functioning within a new positive feedback loop that reinforces the TGF-β/SMAD signaling. We provide comprehensive in vitro and in vivo evidence that EIF4A3 promotes GC cell EMT and other malignant phenotypes. Mechanistically, our data reveal that EIF4A3 is transcriptionally induced by TGF-β signaling via SMAD2/3-mediated promoter activation, and in turn, EIF4A3 amplifies this pathway by stabilizing SMAD2, SMAD3, and SMAD4 mRNAs, thereby reinforcing EMT and metastatic potential.

EIF4A3 as a key EJC component is a known RNA-binding protein that modulates mRNA splicing, transport, and decay. It has been increasingly recognized for its oncogenic role in various cancers through the post-transcriptional regulation of cancer-related transcripts (Sakellariou and Frankel 2021; Qiu et al. 2022). Prior studies have reported EIF4A3-mediated promotion of invasion and metastasis in various cancers via different targets. For instance, EIF4A3 facilitates c-Myc translation via hsa_circ_0000467 in colorectal cancer, promoting the growth, migration and invasiveness of cancer cells (Jiang et al. 2024). EIF4A3 drives glycolysis and invasion through the circWAC/miR-599/E2F3 axis in breast cancer (Huang et al. 2022). EIF4A3 inhibits the expression of tumor-suppressor circ_0087429 by direct binding, contributing to cell migration, invasion as well as EMT in cervical cancer (Yang et al. 2022). Similarly, in NSCLC, Chen et al. (Li et al. 2024) have shown that EIF4A3 stabilizes FOXM1 mRNA, thereby promoting proliferation and metastasis. In line with these findings, our data demonstrates that EIF4A3 enhances GC cell growth, and tumorigenesis. These results further extend the oncogenic repertoire of EIF4A3 to GC, a malignancy in which its role has remained largely underexplored.

Importantly, our findings elucidate a previously uncharacterized regulatory connection between EIF4A3 and the canonical TGF-β/SMAD signaling axis, a major driver of EMT and metastatic dissemination in GC (Zheng et al. 2022; Motizuki et al. 2024). While TGF-β induced EMT has been well documented, the factors that modulate its strength and persistence are not fully elucidated. In the canonical pathway, TGF-β receptor activation initiates phosphorylation and subsequent nuclear translocation of a SMAD2/3-SMAD4 complex, regulating EMT-associated transcriptional programs (Zheng et al. 2022). These include the upregulation of transcriptional repressors such as Snail, Slug, and ZEB1/2, which repress epithelial markers and stimulates mesenchymal markers (Motizuki et al. 2024; Tavakolian et al. 2019; Serrano-Gomez et al. 2016). Although inhibitory SMADs such as SMAD7 act as negative feedback regulators by preventing SMAD2/3 phosphorylation and complex formation (Zhang et al. 2021b; Fan et al. 2023), persistent pathway activation in advanced cancers often overrides these suppressive mechanisms. Our data reveals a previously unrecognized positive feedback loop wherein TGF-β signaling activates SMAD2/3, which transcriptionally induces EIF4A3. EIF4A3, in turn, stabilizes SMAD2/3/4 mRNAs, thereby reinforcing TGF-β/SMAD signaling output. This loop amplifies EMT and metastasis in GC cells. Notably, TGF-β stimulation upregulated EIF4A3 expression in a dose- and time-dependent manner, while EIF4A3 knockdown suppressed TGF-β-induced EMT markers, including Snail-a central transcriptional repressor of E-cadherin (Tavakolian et al. 2019; Serrano-Gomez et al. 2016). Moreover, EIF4A3 silencing impaired SMAD2/3 phosphorylation and nuclear translocation despite active TGF-β stimulation, highlighting its essential role in sustaining pathway activity through transcript stabilization.

Specifically in GC, our study expands on prior work by Wang et al. (Wang et al. 2025), who reported that EIF4A3 stabilizes circ_0008126, thereby promoting GC cell growth, invasiveness, and EMT via the miR-502-5p/APC/β-catenin axis. However, their study primarily focused on non-coding RNA regulation, whereas we uncover a broader mechanistic framework in which EIF4A3 not only acts as a downstream effector of the TGF-β/SMAD axis but also acts as a feedback enhancer of this pathway by stabilizing key SMAD transcripts. This is a significant advancement over previous reports, as it demonstrates a dual regulatory role of EIF4A3, both as a transcriptional target and a functional amplifier of the TGF-β/SMAD axis, explaining the sustained EMT, and metastatic behavior often observed in advanced GC despite transient upstream stimuli.

From a clinical perspective, our results hold translational relevance. Elevated EIF4A3 indicates adverse survival outcomes, supporting its utility as a prognostic biomarker. More importantly, EIF4A3 may serve as a viable therapeutic target. Given the complexity and pleiotropic nature of TGF-β signaling, direct inhibition of the pathway often leads to unwanted systemic effects. EIF4A3 inhibition, by selectively destabilizing SMAD transcripts, could provide a more targeted strategy to dampen TGF-β driven EMT and metastasis while preserving physiological TGF-β functions. Indeed, pharmacologic inhibition of the TGF-β receptor I kinase (with SB525334) has shown promise in preclinical models (Heo et al. 2021; Grygielko et al. 2005). Our findings suggest that combined inhibition of both upstream TGF-β signaling and its downstream post-transcriptional amplifiers, such as EIF4A3, may produce synergistic therapeutic effects. Future studies should explore the development of EIF4A3-targeting compounds, particularly helicase inhibitors, or nanoparticle-encapsulated siRNA, and evaluate their combinatorial efficacy with TGF-β pathway inhibitors in preclinical GC models.

We should also acknowledge the limitations of our study. First, the small clinical sample size restrains the statistical power as well as generalizability of our results. Second, only two cell lines were used for exploring the function of EIF4A3 in vitro, which could not accurately reflect the complexity of tumor heterogeneity, and its role in other cells especially with lower EIF4A3 expression should be investigated in future research. Third, though we have demonstrated the oncogenic effects of EIF4A3 on promoting the EMT and metastasis in GC via TGF-β/SMAD pathway, we do not exclude the possibility that EIF4A3 may cooperate with additional targets or cofactors in GC. Moreover, the subcutaneous xenografts were used for in vivo studies, which lacks a normal stromal microenvironment for cancer cells and cannot model the interactions with the tumor microenvironment.

Conclusions

Our study identifies EIF4A3 as a key post-transcriptional regulator of the TGF-β/SMAD axis and a driver of EMT and metastasis in GC. By establishing a novel positive feedback loop with SMAD signaling, EIF4A3 reinforces oncogenic TGF-β signaling and promotes GC progression (graphical abstract). Our work expands the understanding of post-transcriptional regulation in EMT and metastasis and suggests that EIF4A3 may serve as both a biomarker and a promising molecular target for therapeutic intervention in GC. Further studies are warranted to investigate EIF4A3 inhibitors and their efficacy in preclinical GC models and to determine whether similar feedback mechanisms exist in other SMAD-regulated contexts.

Supplementary information

Below is the link to the electronic supplementary material.

Author contributions

YW conceived and designed the experiments. YL, YZ, and WD contributed significantly to the experiments and arranging data. WP, MC, YZ, and QS contributed to project administration, data analyses, visualization, and validation. YL and YZ wrote the initial draft of the manuscript. YW revised the manuscript. All authors read and approved the final manuscript.

Funding

This work did not receive any funding support.

Data availability Statement

The datasets generated during and/or analyzed during the current study are included in the manuscript and are available from the corresponding author upon reasonable request.

Declarations

Conflicts of interest

The authors declare no competing interests.

Ethical approval

All the experimental procedures involving animals were conducted following ARRIVE guidelines and were approved by the Ethics Committee of the Second Affiliated Hospital of Soochow University. The study involving human subjects complied with the Declaration of Helsinki and was approved by the ethical committee of our institution. All participants provided written informed consent.

Clinical trial number

Not applicable.