microRNA-204 as a crucial cancer stem cell related miRNA inhibits stemness of gastric cancer by targeting CD44 and EPCAM

Abstract

Background

Cancer stem cells (CSCs) and normal stem cells share key properties such as self-renewal and differentiation capacity, yet the microRNAs (miRNAs) that are differentially expressed between them remain poorly characterized. This study aimed to identify such miRNAs, with a focus on microRNA-204 (miR-204), and to investigate its functional role and regulatory mechanisms in gastric cancer.

Methods

Using miRNA microarray data from public databases (GEO and TCGA), we identified miR-204 as a stemness-related miRNA. Its expression was validated in clinical gastric cancer tissues and cell lines. Functional assays, including colony formation, Transwell, tumorsphere formation, and extreme limiting dilution analysis, were performed to assess the effects of miR-204 on proliferation, metastasis, and stemness. Luciferase reporter assays confirmed direct targeting of CD44 and EPCAM. The role of TP53 in miR-204 maturation was investigated using TP53-knockout mice and TP53 overexpression cell models.

Results

miR-204 was identified as a crucial miRNA, upregulated in normal stem cells but significantly downregulated in CSCs and multiple human cancers, including gastric cancer. Low miR-204 expression correlated with advanced TNM stage and poor overall survival in gastric cancer patients. Restoring miR-204 expression potently inhibited gastric cancer cell proliferation, metastasis, and stemness properties. Mechanistically, miR-204 directly targeted and suppressed the expression of two key CSC regulators, CD44 and EPCAM. Furthermore, we discovered that TP53 mutation is a major cause of miR-204 downregulation, and demonstrated that TP53 promotes the maturation of miR-204 rather than its primary transcription.

Conclusions

Our study unveils miR-204 as a pivotal tumor suppressor that distinguishes CSCs from normal stem cells. It functions by targeting the CD44/EPCAM axis to inhibit gastric cancer stemness, a process governed by a novel TP53-miR-204 regulatory circuit. These findings position miR-204 as a promising prognostic biomarker and therapeutic target for eradicating CSCs in gastric cancer.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12967-026-08544-w.

Introduction

Since cancer stem cells (CSCs) were first proposed by Bonnet and Dick in leukemia [1], they have been the objects of in-depth study. CSCs have been identified as rare cell populations in many cancers. Owing to the ability of self-renewal, CSCs possess extensive tumor-initiating ability when implanted into immune compromised host mice with limited dilution. Besides, CSCs tend to be more resistant to chemotherapeutic drugs and radiation than differentiated cell subtypes from the same tissue [2]. CSCs are related to initiation, progression, metastasis, and drug resistance of tumors.

The current CSCs hypothesis provides a unified theory of cancer biology, which is mainly promoted by the growing body of work of human normal adult tissue stem cells. Many researchers favor that CSCs are generated by mutation or epigenetic reprogramming in embryonic stem cells, adult stem cells or adult progenitor cells [3]. This hypothesis is supported by results of stem cells transplantation and stem based tumor organoids which involved in the formation of cancer [4, 5]. In addition, a large number of experimental studies have found normal stem cells and CSCs share the same protein molecules and pathways such as Nanog, Notch, Wnt, SOX2, Myc, Hedgehog and KLF4 [6, 7]. Although researchers have identified some compounds, such as salinomycin and narasin, which could selectively reduce the proportion of CSCs, the potential toxicity of these agents to normal stem cells is unavoidable. Hence, identifying genes that have different expression between CSCs and normal stem cells deserve further exploration urgently.

microRNAs are short non-coding RNA molecular (20 to 24 nucleotide) that regulate gene expression at the post-transcriptional level by binding to the 3’-untranslated regions (3’UTR) of target genes, leading to the repression of mRNA translation or degradation of target mRNA [8]. During recent years, researchers have demonstrated miRNAs play an important role in gene regulation during pluripotency, self-renewal and differentiation of many cell types including Embryonic stem cells (ESCs), adult stem cells and induced pluripotent stem cells (iPSCs) [9]. Some miRNAs, such as miR-126, miR-302 and miR-125, are associated with maintaining pluripotency and involved in maintenance of cancer stem cells [10–12]. Some miRNAs including miR-34a, miR-100 and miR-137 which modulate the differentiation of ESC are down-regulated in CSCs [13–15]. Moreover, suppressed expression of miR-200c is required for normal stem cells to form mammary ducts and breast CSCs for oncogenesis [16]. However, some CSCs related miRNAs, which can distinguish CSCs from normal stem cells, have hardly been reported.

TP53 is one of the most vital tumor suppressor which is crucial for cell apoptosis, genome stability and inhibition of angiogenesis. Different isoforms of TP53 due to gene mutation or deletion are responsible for cancer susceptibility and formation [17]. Studies have demonstrated that TP53 is a critical regulator of stem cells [18] and involved in tumorigenesis and pluripotency reprogramming [19]. Mutant TP53 isoforms or tumor-specific splice variants promote tumorigenic potential of CSCs by regulating key pluripotent factors, including SOX2, OCT3/4 and NANOG [20]. Besides, TP53 can regulate cancer stemness through controlling miR-34a, miR-200c expression [21, 22]. Moreover, accumulating somatic mutations and specific splicing variants of TP53 are identified in gastric epithelium with helicobacter pylori infection which is a risk factor for gastric cancer [23].

In this study, we tried to identify some important CSCs related miRNAs by analyzing miRNA microarray, and to explore their function and regulatory mechanism in tumor. Our findings will help us have a better understanding of CSCs and provide a novel therapeutic strategy for tumor targeted therapy.

Methods

Tissue specimens

176 matched gastric cancer and adjacent normal tissues (>5 cm laterally from the edge of cancer tissues) were collected from patients undergoing resection of cancer in the First Affiliated Hospital of China Medical University. Fresh samples were washed and snap frozen in liquid nitrogen immediately and store at −80 °C. For immunohistochemistry assay, 10 pairs of gastric cancer tissue were collected and stained. All samples were obtained with patients’ informed consent, and the study protocol was approved by the Medical Ethics and Human Clinical Trial Committee.

RNA isolation and real-time quantitative PCR (qPCR)

Total RNA from cells or tissues was isolated using Trizol Reagent (Invitrogen, USA), microRNAs were further extracted with miRVana miRNA Isolation Kit (Ambion, USA) following the manufacture’s instruction. 500 ng miRNAs were reverse transcribed to DNA with the PrimeScriptTM RT reagent Kit (Takara, China) using specific reverse primers: miR-204-RT, 5’-GTCGTATCCAGTGCAGGGTCCGAGGTATTCGCACTGGATACGACAGGCATAG-3’; pri-miR-204-RT, 5’-ACAGTTATGGGCCCAATGATG −3’; U6-RT, 5’-AACGCTTCACGAATTTGCGT-3’. The expression of miRNAs was assessed using the following primers: miR-204-forward, 5’-GCGCTTCCCTTTGTCATCCT-3’, miR-204-reverse,5’-CAGTGCAGGGTCCGAGGTC-3’;U6-forward,5’-CTCGCTTCGGCAGCACA-3’, U6-reverse, 5’-AACGCTTCACGAATTTGCGT- 3’; homo-pri-miR-204-forward, 5’-GGACAGGGTGATGGAGAG-3’, homo-pri-miR-204-reverse, 5’-ACAGTTATGGGCCCAATGATG-3’. mice-pri-miR-204-forward, 5’-TCACTCTGTTCAGGACTTGGC-3’, mice-pri-miR-204-reverse,5’-AGTTATGGGCTCAATGATGGT-3’. The expression of mRNAs was assessed using the following primers: CD44-forward, 5’-CCAGAAGGAACAGTGGTTTGGC-3’ and CD44-reverse, 5’-ACTGTCCTCTGGGCTTGGTGTT-3’; Nanog-forward, 5’-CTCCAACATCCTGAACCTCAGC-3’ and Nanog-reverse, 5’-CGTCACACCATTGCTATTCTTCG-3’; EPCAM-forward, 5’-GCCAGTGTACTTCAGTTGGTGC-3’ and EPCAM-reverse, 5’-CCCTTCAGGTTTTGCTCTTCTCC-3’; GAPDH-forward, 5’-GTCTCCTCTGACTTCAACAGCG-3’ and GAPDH-reverse, 5’-ACCACCCTGTTGCTGTAGCCAA-3’.

Cell culture and transfection

Human gastric cancer cell lines SGC-7901 and AGS were purchased from the Cell Bank of Type Culture Collection of Chinese Academy of Sciences. To obtain CSCs, cells of 50 ml ascites from 2 peritoneal metastatic gastric cancer were collected and fibroblast cells were discarded using differential attachment, then EPCAM microbeads were employed as a commonly used isolation method for enriching gastric cancer CSCs, EPCAM positive CSCs were enriched by EPCAM microbeads, though it is acknowledged that EPCAM alone is not a definitive CSC marker for gastric cancer (as gastric CSCs may also co-express markers like CD44 or Lgr5). All cells were cultured in proper medium supplemented with 10% FBS (Hyclone), 100 U/ml penicillin and 100 U/ml streptomycin at 37 °C in 5% CO2 incubator.

For transient overexpression of p53, logarithmically growing cell lines were seeded in the 6-well plate, then transfected 3μg pCMV-3×Flag-P53 plasmid or vector with Higene (Applygen, China) according to the manufacturer’s instruction. Replacing fresh medium after 6 h, and cells were harvested after 36 h transfection. In order to generate miR-204 stable cell lines, cells were infected with lentiviral packaged miR-204 according to the manufacturer (GeneChem, China).

AGS and SGC-7901 were selected due to well-characterized cancer stem cell (CSC) populations—efficiently enriched via EPCAM microbeads sorting (a validated approach to isolate gastric CSCs with stemness and metastatic traits [24]) and widespread use in gastric cancer research (encompassing proliferation, autophagy, and epithelial-mesenchymal transition studies). Notably, SGC-7901 forms serum-free floating spheres with stemness features [25, 26], further supporting utility for CSC-related investigations. This combination renders AGS and SGC-7901 ideal models for exploring miR-204 in gastric cancer stemness regulation.

RNA interference and siRNA transfection

Knockdown of target genes was achieved by siRNA transfection. AGS cells at 60–70% confluence were transfected with 100 pmol of siRNA (si-Drosha or negative control siRNA) per well in 6-well plates using Higene transfection reagent (Applygen), following the manufacturer’s instructions. The medium was replaced after 6 h, and cells were collected for analysis 48–72 h later. Knockdown efficiency was verified by Western blot. si- Drosha RNA, and siRNA control were provided by Santa Cruz Biotechnology, Inc.

Western blot analysis

Cell total protein was isolated with RIPA lysis buffer supplemented with protease inhibitor mixture (Roche, USA), then protein was separated with 8% SDS-PAGE and transferred to a PVDF membrane (Amersham, USA).The membranes were blocked with 5% non-fat milk for 1 h, and incubated overnight with primary antibodies including rabbit anti-CD44 monoclonal antibody (ab51037, Abcam), rabbit anti-Drosha (3364T，CST)， rabbit anti-Epcam monoclonal antibody (ab124825, Abcam), rabbit anti-ATM monoclonal antibody(ab32420, Abcam), rabbit anti-snail monoclonal antibody(ab180714, Abcam), rabbit anti-STAT3 monoclonal antibody (ab68153, Abcam), rabbit anti-HDAC1 monoclonal antibody (ab109411, Abcam), rabbit anti-JAK2 monoclonal antibody(ab108596, Abcam), rabbit anti-chk1 monoclonal antibody (ab40866, Abcam)), mouse anti-flag monoclonal antibody(M20008, Abmart), mouse anti-GAPDH antibody (KC-5G4, Kangchen). After being washed with TBST for three times, membranes were incubated with horseradish peroxidase (HRP) conjugated goat secondary antibody (ZSBIO, China) for 1.5 h. Finally, membranes were detected by the enhance chemiluminescence (ECL) system (Thermo pierce, USA).

Colony formation assay

Logarithmically growing cells were seeded into 6-well plates (2 × 10³ /well), and medium was refreshed once three days. After 2 weeks, cells were fixed and stained with crystal violet, and images were captured by camera. Then crystal violet was eluted with 30% ethanoic acid, and the absorbance was measured at OD570.

Transwell assay

1 × 10⁵ Cells suspended with 100 ul serum free medium were seeded onto the transwell chambers (Corning Costar, USA) with or without matrigel matrix (BD, USA). Medium containing 10% FBS was added into the lower 24-well plates. After 24 h incubation, cells were fixed with 4% paraformaldehyde for 20 min, and stained with 0.2% crystal violet. Cells in the upper chambers were removed with cotton swabs carefully and captured with microscope. Then crystal violet was eluted with 33% ethanoic acid, and the absorbance was measured at OD570.

Stem sphere formation and ELDA assays

Cells (1 × 10³/per well) were resuspended in DMEM/F12 medium containing 1×N2 supplement (Invitrogen, USA), 1×B27 supplement (Invitrogen, USA), 20 ng/ml EGF (Protech, USA), 100ng bFGF (protech, USA) and seeded on 6-wells Ultra-Low attachment culture dishes (Corning Costar, USA). After 7 days of culturing, cells were examined under an inverted microscope. Spheres were counted and the efficiency of sphere formation was compared. For extreme limiting dilution analysis (ELDA), different numbers of cancer cells (1000, 500, 100, 50, 10) were seeded into 96-well plates and stem cell frequency was estimated 7 days later according to the method³⁹. For target genes restoring assay, cells were firstly transfected with pCMV-Flag-EPCAM or pcmv-flag-CD44, then seeded into plates for sphere formation and ELDA assays.

Xenograft assays and animals

For xenograft experiment, 4-week old female NOD-SCID mice (Vital River Laboratory Animal Technology Co., Ltd, China), selected for T/B lymphocyte deficiency and impaired NK cell function to avoid human gastric cancer cell immune rejection, were (n = 48, 8 mice/group) randomly allocated via SPSS 23.0-generated random number table (stratified by weight 16-18 g/age) by a blinded investigator before bilateral axillary injection with limiting diluted AGS/SGC-7901 control stable cells or AGS/SGC-7901 miR-204 stable cells (cells resuspended in sterile PBS+10% Matrigel at 1 × 10⁶ −1 × 10⁴ cells/100 μL for consistent dosage). After 8 weeks, the mice were sacrificed via intraperitoneal injection of 100 mg/kg sodium pentobarbital and tumor formation was analyzed with ELDA software.

C57BL/6 p53^+/+ and p53^-/- mice were provided by professor Liu Cao, selected for identical C57BL/6 genetic background to eliminate inter-strain variability for TP53-related assays. All animals used in this study were maintained under specific pathogen-free conditions in Laboratory Animal Center of China Medical University, and all treatments were in accordance with the NIH Animal Care and Use Committee regulations.

Immunohistochemistry

To detect the protein expression of EPCAM, CD44 and ATM in gastric cancer tissues, 4 μm paraffin sections of gastric cancer tissues were deparaffinized, rehydrated and heated for antigen retrieval. Then sections were incubated overnight at 4 °C with primary antibody after blocking with normal goat serum. The sections were stained with EliVision^TM plus and DAB kits (Maxin Biotechnologies) according to the instructions. Slides were photographed with microscope and the grayscale analysis of protein expression were analyzed by Image J software.

Dual luciferase reporter assay

The luciferase reporter assay was performed in AGS and SGC-7901 cell lines using pMIR-report^TM system (Ambion, USA). Vectors of CD44-WT-3’UTR, CD44-deletion-3’UTR, EPCAM-WT-3’UTR, EPCAM-deletion-3’UTR were constructed. Cells were seeded at 1 × 10⁵ per well of a 24-well plate, and co-transfected with 200 ng WT or deletion reporter plasmids and 20 ng pMIR-Report^TM-βgal control plasmid in the presence of 100 nM miRNA mimics or negative control. The luciferase activity was measured with Dual Luciferase Reporter Assay System (Promega, USA). The relative luciferase activity was obtained by normalizing luciferase expression with β-gal expression. All experiments were performed in triplicate at 3 times independently.

Data and bioinformatics analysis

GSE42446, GSE36046, GSE62145, GSE2564, GSE58004, GSE32174, GSE19769, GSE61079, and GSE61080 were obtained from GEO (https://www.ncbi.nlm.nih.gov/geo/). Their matrix data were analyzed for differentially expressed genes using a false discovery rate (FDR) threshold of 0.05. All microarray data were processed with R (v4.3.1) tools (including RMA normalization, gene annotation tools for ID mapping, and batch effect correction tools) or GEO2R software; outlier samples (outside 95% CI in PCA) were excluded, with |log2 fold change| >1 as an additional filter.

TCGA datasets were obtained from https://tcga-data.nci.nih.gov/tcga/. Data were processed to exclude samples with incomplete clinical data, retaining 389 qualified samples. Cancer genomics were analyzed by cBioPortal. Protein-protein interactions were predicted via STRING with confidence score ≥ 0.7. Target genes of miR-204 were analyzed by StarBase V2.0 (filters: pan-cancer conserved sites, CLIP-seq/degradome-seq support); genes overlapping with cancer stem cell-related genes (http://www.sabiosciences.com/rt_pcr_product/HTML/PAHS-176Z.html) were analyzed with Venny software. All tests were two-tailed, FDR corrected by Benjamini-Hochberg.

Aldehyde dehydrogenase (ALDH) activity assay

ALDH enzymatic activity was measured using a commercial Aldehyde Dehydrogenase (ALDH) Activity Assay Kit (Solarbio, #BC0755) according to the manufacturer’s instructions. Briefly, EPCAM-positive and EPCAM-negative cells were collected and lysed. The supernatant was collected after centrifugation. The reaction mixture, containing assay buffer, substrate, and coenzyme, was added to the samples. After incubation at 37 °C for the specified duration, the increase in absorbance at 340 nm, corresponding to NADH production, was measured using a microplate reader. ALDH activity was calculated based on the NADH standard curve and normalized to the total protein concentration of each sample, which was determined using a BCA protein assay kit. Results are expressed as units per milligram of protein.

Statistical analysis

All statistical analyses were performed using SPSS Version 23.0 and GraphPad Prism software. Data are presented as mean ± standard deviation. For datasets involving fold-change comparisons that required parametric analysis, data were log2-transformed prior to testing to approximate a normal distribution. All other continuous data were analyzed based on raw values. Normality and homogeneity of variances were assessed using the Shapiro-Wilk test and Levene’s test, respectively. Comparisons between two groups were performed using unpaired or paired two-tailed Student’s t-test for parametric data, or the Mann-Whitney U test for non-parametric data. Comparisons among multiple groups were conducted using one-way analysis of variance followed by Tukey’s post-hoc test. Survival analysis was performed using the Kaplan-Meier method with the log-rank test; hazard ratios and 95% confidence intervals were derived from Cox proportional hazards regression. Correlation was assessed using Pearson’s correlation coefficient. Extreme limiting dilution assay data were analyzed using the online ELDA software based on a Poisson distribution model. The exact sample size for each experiment is provided in the corresponding figure legend. A p-value < 0.05 was considered statistically significant. To address potential bias from imbalanced groups in survival analysis, a robustness check was performed using propensity score weighting (Inverse Probability of Treatment Weighting, IPTW). Key clinical covariates were balanced using the WeightIt package in R (v4.4.3). Survival analysis was repeated on the weighted cohort.

Results

MiR-204 is identified as a crucial CSCs related miRNA

To identify CSCs enrich microRNAs, we searched and collected stemness-related microarray data from GEO database. As a result, we got one ESC microarray data (GSE42446) which contains 10 ESC, 6 soma tissues and 3 cancer cell lines; one CSCs microarray data (GSE61079) including CSCs and cancer cells; one CSCs microarray data (GSE61080) including neural stem cell (NSC) and NSC derived glioma-initiating cells (GICs or Glioma CSCs). By comparing embryonic stem cells (ESCs) vs. somatic cells, ESCs vs. cancer cells, cancer cells vs. cancer stem cells (CSCs), and normal stem cells (NSCs) vs. CSCs, the aim is to identify miRNAs that are highly expressed in normal stem cells but downregulated in CSCs. Firstly, we identified differentially expressed miRNAs between ESC and soma tissues (Fig. 1A, Table S1); ESC and cancer cells (Fig. 1B, Table S2); CSCs and cancer cells (not CSCs) (Fig. 1C, Table S3); normal stem cells and derived CSCs (Fig. 1D, Table S4). Subsequently, we performed further analysis of overlapping miRNAs among these groups using Venny software, retaining only those microRNAs satisfying all three criteria: upregulation in embryonic stem cells vs. somatic tissues; downregulation in cancer stem cells vs. non-cancer stem cells; and downregulation during induced pluripotent stem cell differentiation. Any microRNA that failed to meet all these criteria was excluded. miR-204 was identified as a significantly differentially expressed miRNA, demonstrating upregulation in ESCs and downregulation in both cancer cells and CSCs. (Fig. 1E，Figure S5). Subsequently, we evaluated the expression of miR-204 in differential iPSCs GEO profiles (GSE62145). As expected, miR-204 was down-regulated when iPSCs differentiated to neurons by expressing neurogenin1/2 (NGN1/2) (Figure S1A). To further validate this finding, CSCs from peritoneal metastatic gastric cancer tissues were enriched by EPCAM microbeads and real-time qPCR assay was performed. As shown in Fig. 1F, miR-204 level was significantly decreased in EPCAM positive CSCs compared with the negative cells. To assess the stemness of EPCAM⁺ cells, we measured mRNA expression of the stemness markers CD44 and Nanog in EPCAM⁺ and EPCAM⁻ cells isolated from two gastric cancer ascites donors. As shown in the figure, both markers were significantly upregulated in EPCAM⁺ cells from each donor (Fig. 1G). We also evaluated ALDH1 enzymatic activity, a functional readout of cancer stemness. Consistent with the gene expression data, EPCAM⁺ cells exhibited markedly elevated ALDH1 activity compared to EPCAM⁻ cells in both donors (Fig. 1H). Together, these results confirm that the EPCAM⁺ population is enriched with cancer stem-like properties, providing a suitable model for studying miR-204 downregulation in CSCs. All these results indicate miR-204 was a CSCs related miRNAs which may have a different role in normal stem cells.

Fig. 1.

miR-204 was identified as a crucial cancer stem cell related microRNA. A, differentially expressed microRNAs between ESCs (embryonic stem cells) and adult tissue cells (GSE42446). B, differentially expressed microRNAs between ESCs and cancer cells (GSE42446). C, differentially expressed microRNAs between cancer cells and cancer stem cells (CSCs) (GSE61079). D, differentially expressed microRNAs between normal stem cells and cancer stem cells (CSCs) (GSE61080). E, overlap analysis identifying miR-204 as a candidate CSCs-related miRNA. Venn diagram shows miRNAs commonly upregulated in embryonic stem cells versus somatic cells (GSE42446) and downregulated in cancer stem cells versus non-stem cancer cells (GSE61079) or normal neural stem cells versus glioma-initiating cells (GSE61080). Differential expression was defined as |log2 Fold change| >1 with FDR < 0.05. The underlying microarray data were log2-transformed and normalized prior to analysis. F, relative expression of miR-204 in CSCs enriched from peritoneal metastatic gastric cancer ascites. EPCAM-positive cells (n = 2 independent patient samples) were compared to EPCAM-negative cells by qPCR. Data are presented as log2 (fold change) (mean±SD). Statistical significance was assessed using a paired Student’s t-test. G, qPCR detection of CD44 and Nanog mRNA expression levels in EPCAM-positive cells and EPCAM-negative cells(n = 2 independent patient samples); data are presented as log2 (fold change) (mean±SD). Statistical significance was assessed using a paired Student’s t-test(**p < 0.01, ***p < 0.001). H, ALDH1 enzymatic activity assay in EPCAM⁺ and EPCAM⁻ cells; absorbance was measured at 450 nm, data represent mean±SD from three independent experiments. Statistical significance was assessed using a paired Student’s t-test(***p < 0.001)

MiR-204 is down-regulated in diverse human cancers

In order to clarify whether the abnormal expression of miR-204 was universal in cancers, the GSE2564 microarray which contains 51 normal tissues, 116 cancer tissues and 35 cancer cell lines was analyzed. As shown in Figure S1B, the expression of miR-204 was significantly decreased in most cancers, indicating miR-204 may function as a crucial cancer suppressor. To further validate these results, we analyzed the expression of miR-204 in a large amount of tumor tissues collecting from TCGA database. As expected, miR-204 was significantly decreased in 449 colon adenocarcinoma (COAD), 414 bladder urothelial carcinoma (BLCA), 480 stomach adenocarcinoma (STAD), 366 esophageal carcinoma (ESCA), 369 liver hepatocellular carcinoma (LIHC), 1353 breast invasive carcinoma (BRCA), 480 head and neck squamous cell carcinoma (HNSC), 512 thyroid carcinoma (THCA), 488 prostate adenocarcinoma (PRAD) and 487 lung adenocarcinoma (LUAD). The expression of miR-204 was also significantly down-regulated in 261 kidney renal clear cell carcinoma (KIRC, also called ccRCC) and 62 kidney chromophobe carcinoma (KICH, also called chRCC) (Fig. 2).

Fig. 2.

Analysis of miR-204 expression in different tissues and tumors. Analysis of miR-204 expression across multiple human cancers. Expression levels of miR-204 (log2-transformed) are shown for tumor samples compared with adjacent normal tissues in colon adenocarcinoma (COAD, n = 449), bladder urothelial carcinoma (BLCA, n = 414), stomach adenocarcinoma (STAD, n = 480), esophageal carcinoma (ESCA, n = 366), liver hepatocellular carcinoma (LIHC, n = 369), breast invasive carcinoma (BRCA, n = 1353), head and neck squamous cell carcinoma (HNSC, n = 480), kidney chromophobe carcinoma (KICH, n = 62), kidney renal clear cell carcinoma (KIRC, n = 261), thyroid carcinoma (THCA, n = 512), prostate adenocarcinoma (PRAD, n = 488), and lung adenocarcinoma (LUAD, n = 487). Data are presented as log2(RPM or intensity) (mean±SD). For each cancer type, statistical significance was assessed using an unpaired Student’s t-test (*p < 0.05)

MiR-204 is a potential tumor repressor in gastric cancer tissues

Gastric cancer (GC) is one of the most common cancers in the world and a great threat to human health. In this study, we used GC as a model to study the function and molecular mechanism of miR-204 in cancers. Firstly, we evaluated the expression of miR-204 in 176 pairs of human GC with cancer and paracancer tissues. As expected, miR-204 expression significantly decreased in cancer tissues compared with the matched normal tissues (Fig. 3A). Overall survival of patients with a high-expression level of miR-204 was significantly higher than those with low-expression level (HR = 0.43, 95% CI: 0.20–0.94, p = 0.034) (Fig. 3B). To address potential bias from imbalanced sample sizes and baseline characteristics, we performed a robustness analysis using Inverse Probability of Treatment Weighting (IPTW). The analysis balanced key clinical and pathological covariates including Sex, Age, Histologic grade, Tumor size, Histological type, pT stage, pN stage, Lymph node metastasis, Venous invasion, Peripheral organ involvement, TNM stage, and Implantation metastasis. In this adjusted analysis, high miR-204 expression remained a significant and independent predictor of favorable overall survival (HR = 0.41, 95% CI: 0.19–0.86, p = 0.019)(Figure S6 and Table S5). This result confirms the robustness of miR-204 as a prognostic biomarker in gastric cancer. Moreover, the association of miR-204 expression with clinicopathologic characteristics in patients was further evaluated. As shown in Table 1, the expression of miR-204 in cancer tissues was negatively correlated with age, tumor size, depth of invasion, implantation metastasis and pathological stage. The lower expression is in larger size tumor (Fig. 3C). Moreover, miR-204 expression level is higher in TNM stage I-II than TNM stage III-IV (Fig. 3D), indicating abnormality of miR-204 occurs at the early stage of tumor development. All these findings indicated miR-204 was a potential tumor repressor in gastric cancer tissues.

Fig. 3.

miR-204 is a potential tumor repressor in gastric cancer tissues. A, relative expression of miR-204 in 176 gastric cancers compared with precancers were detected by real-time qPCR, data are presented as log2 (fold change) and were log-transformed before statistical analysis (t-test). B, Kaplan-Meier survival analysis of gastric cancer patients stratified by miR-204 expression. Patients were divided into a high-expression group (log2 (fold change) >0, n = 26) and a low-expression group (log2 (fold change) <0, n = 149). Differences were compared using the log-rank test (p = 0.034). The hazard ratio was 0.43 (95% CI: 0.20–0.94). C, Association between miR-204 expression and tumor size. Data are presented as log2 (fold change) (mean±SD; <5 cm group n = 76, ≥5 cm group n = 100). Statistical significance was assessed using unpaired Student’s t-test. D, Association between miR-204 expression and TNM stage. Data are presented as log2 (fold change) (mean±SD; stage I-II n = 21, stage II-IV n = 155). Statistical significance was assessed using unpaired Student’s t-test

Table 1.

Association between the MiR-204 expression and clinicopathologic features in patients with gastric cancer

parameters	No.	MiR-204 (Fold Change)1	p value2
Age(yrs)
<65	113	−3.65±3.50	0.012 *
≥65	63	−5.13±4.06
Gender
Male	131	−4.07±3.56	0.521
Female	45	−4.49±4.34
Family history
NO	161	−4.24±3.78	0.513
Yes	15	−3.57±3.68
Histologic grade
Well & Moderate	63	−4.66±3.82	0.204
Poor & Signet	113	−3.91±3.73
Size (maximal diameter)
<5 cm	76	−3.54±3.80	0.049 *
≥5 cm	100	−4.67±3.69
Histological type
Intestinal	82	−4.44±4.03	0.396
Diffuse	94	−3.95±3.53
Depth of invasion (pT)
T1, T2	35	−2.82±3.33	0.017 *
T3, T4	141	−4.52±3.81
Lymph node metastasis
No	48	−4.13±4.04	0.909
Yes	128	−4.20±3.67
Lymph node status (pN)
N0, N1	84	−4.23±3.59	0.858
N2, N3	92	−4.13±3.94
Peripheral organ injury
No	134	−3.93±3.72	0.119
Yes	42	−4.97±3.85
Implantation metastasis
No	166	4.36 ± 3.72	0.008 **
Yes	10	−1.13±3.32
Pathological stage (pStage)
Stage I, II	21	−2.31±2.95	0.015 *
Stage III, IV	155	−4.43±3.80

¹Mean of relative expression by log2(fold change), with mean±SD

² Student’s t-test was used for analysis. * p < 0.05, ** p < 0.01

MiR-204 inhibits cell proliferation, metastasis and stemness of gastric cancer

To further clarify the function of miR-204, we constructed miR-204 overexpressing stable cell lines (Figure S2A) and performed colony formation assay to examine the effect of miR-204 on cell proliferation. As shown in Fig. 4A, increasing miR-204 obviously inhibited cell growth. Transwell assay revealed restoration of miR-204 expression in gastric cancer cells significantly restrained the capability of migration and invasion (Fig. 4B). Considering miR-204 was identified as a CSCs related miRNA, so the effect of miR-204 on the stemness of gastric cancer cells was assessed. As shown in Fig. 4C&Figure S2B, overexpressing miR-204 hardly formed a tumor sphere compared with cells transfected with the empty vector. Data of extreme limiting dilution assay (ELDA) in vitro also confirmed the CSCs frequency was inhibited by miR-204 (Fig. 4D). Subsequently, in vivo assay that implanting differently diluted cancer cells into immune compromised NOD-SCID mice was conducted. Results showed regaining miR-204 significantly inhibited tumor formation (Fig. 4E–F and Figure S2C).

Fig. 4.

Overexpression of miR-204 inhibits gastric cancer proliferation, metastasis and stemness. A, cell proliferation of miR-204 overexpressing cell lines (miR-204) and control stable cell lines transfected with empty vector (control) were detected by colony formation assay, and the absorbance (OD570) of crystal violet was measured. Quantitative data (mean ± SD) are from three independent experiments. Statistical significance was assessed by unpaired Student’s t-test. B, Transwell assays evaluating cell migration (upper panels) and matrigel invasion (lower panels). Representative images (left) and quantitative analysis (right) show the number of AGS and SGC-7901 miR-204 stable cells that migrated or invaded through the membrane, relative to the vector control stable cells, statistical significance was assessed by unpaired Student’s t-test. C, Representative images of tumorspheres formed by AGS and SGC-7901 vector control and miR-204 stable cells after 7 days of culture under ultra-low attachment conditions. D, cancer stem cell frequency assessed by extreme limiting dilution analysis (ELDA) in vitro. The plot and frequency table are derived from pooled data of three independent experiments. Statistical analysis was performed using the ELDA software based on a likelihood ratio test. E-F, limiting diluted miR-204 overexpressing and control gastric cancer cells were bilaterally axillary injected in NOD-SCID mice, then tumors incidence were evaluated after 8 weeks

MiR-204 attenuate the stemness of CSCs by targeting CSCs related genes

To further investigate the potential mechanism of miR-204 in suppressing the stemness property of CSCs, 84 CSCs related genes (gene list from Qiagen RT² Profiler™ PCR Array Human Cancer Stem Cells product PAHS-176Z) and 1423 miR-204 interacting target genes (predicted by StarBase software) were overlapped by Venny diagram. As displayed in Figure 5A, 21 potential CSCs related genes regulated by miR-204 were identified. Protein interaction analysis with STRING software showed these genes were enriched in tissue and epithelium development (Figure S3).

Although a series target genes of miR-204 had been reported and involved in tumor progress (Figure S4A), few genes that are closely related to cancer stem cells were identified. Among the 21 potential targets, six genes (ATM, Snail, EPCAM, HDAC1, JAK2, CD44) were chosen for their well-documented roles in gastric CSC differentiation [24, 27]. The inclusion of STAT3 and Sirt1 [28, 29] was driven by a dual rationale: their established status as miR-204 targets in other malignancies, coupled with their critical involvement in CSC stemness maintenance-exemplified by STAT3‘s regulation of NANOG expression in gastric CSCs [30]. Western Blot showed that overexpression of miR-204 in gastric cancer cells obviously impaired ATM, CD44, EPCAM protein levels and slightly inhibited Sirt1, JAK2 and Snail levels, while had no effect on STAT3 and HDAC1 protein levels (Fig. 5B). Besides, our study also demonstrated miR-204 inhibited the GC EMT related genes (Figure S4C).

Fig. 5.

miR-204 as a CSCs suppressor by targeting CSCs related genes. A, overlap analysis between miR-204 target genes and CSC related genes. B, Western blot analysis of potential target genes in AGS and SGC-7901 cells after transfection with miR-204 mimics (miR-204) or negative control miRNA (control). C, immunohistochemistry assay was performed to assess relative protein expression of ATM, CD44 and EPCAM in gastric cancer tissues. D. Correlation analysis between log2-transformed miR-204 expression and protein levels of ATM, CD44, and EPCAM (n = 10 patient samples). Pearson correlation coefficient was used. E-F, luciferase reporter assays validating direct targeting of CD44 (E) and EPCAM (F) 3’UTRs by miR-204. Data are presented as relative luciferase activity (mean ± SD) from three independent experiments. Statistical significance was assessed by one-way ANOVA with Tukey’s post-hoc test. G, sphere formation assay in AGS and SGC-7901 miR-204 stable cells following transfection with empty vector, CD44, or EPCAM expression plasmids. H, cancer stem cell frequency assessed by ELDA after restoring target gene expression. Data are from three independent experiments analyzed by the ELDA software

Subsequently, IHC assay was performed to check the protein expression of ATM, CD44 and EPCAM in gastric cancer tissues. As shown in Fig. 5C, both CD44 and EPCAM had elevated expression in tumor tissues compared with paracancer tissues, while there was no significant increased expression of ATM. Correlation analysis revealed CD44 and EPCAM were negatively correlated with miR-204, but no obvious relevance between miR-204 and ATM was observed (Fig. 5D). To validate CD44 and EPCAM were the target genes, luciferase reporter plasmids containing WT-3’UTR (wild type) and Deletion-3’UTR (binding site deletion) were constructed (Figure S4B) and luciferase report assay was conducted. As expected, exogenous expression of miR-204 markedly decreased the reporter luciferase activity of WT-3’UTR, but the activity had no significant difference when co-transfected with deletion- 3’UTRs (Fig. 5E–F, and Figure S4D-E). Moreover, restoring CD44 and EPCAM expression in miR-204 overexpressing cells could enhance stemness of cancer cells (Fig. 5G–H and Figure S4F-H). All these findings confirmed CSCs regulators CD44 and EPCAM were novel target genes of miR-204.

Gene status of miR-204 and the potential regulators

miR-204 is an intragenic miRNA which locates in the large intron 6 of transient receptor potential Cation channel 3 (TRPM3) gene. So, we firstly investigated the gene status of TRPM3 and miR-204 in cancer tissues. Surprisingly, TCGA cancer genomics data showed that mutation of TRPM3 can occur in some samples, but the deletion status of TRPM3 and miR-204 only accounted for less than 1% tumor tissues (Fig. 6A–B), indicating that genomic change may not be the major factor for missing miR-204 expression. It was reported that von hippel-lindau tumor-suppressor gene (VHL) which was lost in most ccRCC (clear cell renal cell carcinoma) played an important role in miR-204 regulation^{21, 22}. However, gene status analysis showed that mutation or deletion of VHL occurred only in ccRCC, and was rarely observed in other tumors including breast cancer, colorectal cancer, head and neck cancer, stomach cancer and chromophobe renal cell carcinoma (chRCC) (Fig. 6C). Subsequently, we further analyzed the mRNA level of VHL in a series cancers, and confirmed that there was no significant difference between normal tissue and cancer tissues (Fig. 6D), indicating some other factors could cause fatal deficiency of miR-204.

Fig. 6.

Gene status of TRPM3 (miR-204 host gene), miR-204, VHL and TP53. A, gene status including mutation, deletion, amplification and multiple alterations of TRPM3 are analyzed in various tumors using cbioportal database. B, gene status including mutation, deletion, amplification and multiple alterations of miR-204 in various tumors. C, gene status including mutation, deletion, amplification and multiple alterations of VHL in various tumors. D, the mRNA expression of VHL in normal gastric tissues and gastric cancer tissues by analyzing GSE2685 dataset. E, potential interaction proteins of VHL were analyzed with STRINF software. F, gene status including mutation, deletion, amplification and multiple alterations of TP53 in various tumors

Protein interaction analysis by STRING software revealed that VHL could interact with transcriptional factors such as TP53, HIF1A and ATM (Fig. 6E). Considering TP53 as one of the most important tumor suppressor genes which played a crucial role in preventing cancer formation, we firstly analyzed its gene status in tumors. Cancer genome analysis revealed that most cancer tissues have TP53 mutation (Fig. 6F). There were more than 70% tumor samples of colorectal and Head and Neck cancer with TP53 mutation, and more than 50% tumor samples of stomach cancer had TP53 mutation. Impressively, TP53 mutation occurred in chRCC samples was more than 30%, while this mutation was barely found in ccRCC samples, indicating VHL may be the major factor in ccRCC and TP53 may possess key effect on other cancers including chRCC.

TP53 is involved in the maturation of miR-204

Studies showed that mutation or alternative splicing of TP53 was accumulate in gastric epithelium with Helicobacter pylori (HP) infection [31, 32], our bioinformatics analysis revealed the expression of miR-204 was significantly down-regulated in HP positive gastric tissues, hinting miR-204 may be a regulatory factor in the origin of GC (Fig. 7A). To further validate TP53 was a regulator of miR-204, we detected the expression of miR-204 in tissues from 10 kinds of organs including brain, lung, liver, stomach, colon, intestine, kidney, spleen, skin and bladder in TP53 knockout mice. As shown in Fig. 7B, miR-204 (22nt) was obviously down-regulated in the brain, lung, stomach, colon, spleen and bladder tissues from knockout mice. Considering TP53 could regulate gene splicing, microRNA transcription and maturation [33, 34], we further measured the expression of primary miR-204 (pri-miR-204) (368 nt) in mice tissues. Impressively, the pri-miR-204 expression had no significant difference in brain, stomach, colon and bladder tissues except the lung and bladder tissues (Fig. 7C), indicating that TP53 may play a key role in the maturation of miR-204 rather than promote its transcription. This hypothesis was validated by regaining the expression of TP53 in primary knocking-out cells. As shown in Fig. 7D–E, mature miR-204 (22nt) significantly elevated in p53 overexpression group, while the pri-miR-204 had no impressive alteration. These findings were also confirmed in TP53 overexpressing gastric cancer cells (Fig. 7F–G). Drosha, the core component of the nuclear microprocessor complex, mediates the rate-limiting cleavage of pri-miRNAs into pre-miRNAs during miRNA maturation. Previous studies have linked TP53 to Drosha/DGCR8 complex interaction in modulating miRNA biogenesis[33]. To clarify the molecular link between TP53 and miR-204 maturation, we performed Drosha knockdown and rescue experiments in AGS cells, Western blot confirmed efficient Drosha knockdown and Flag-p53 expression (Fig. 7H, J). QPCR results showed that compared with the si-NC group, si-Drosha significantly downregulated mature miR-204 (p < 0.001) without altering pri-miR-204 levels (Fig. 7I). Notably, TP53 overexpression failed to reverse mature miR-204 downregulation in Drosha-knockdown cells, and pri-miR-204 levels remained unchanged (Fig. 7K). Taken together, all these data proved that TP53 could regulate the expression of miR-204 mainly by promoting its maturation.

Fig. 7.

TP53 is a key regulator of miR-204 in cancer. A, most differentially expressed microRNAs between HP negative and HP positive gastric tissue (noncancer) in GSE58004 (mouse model), GSE32174 (human), GSE19769 (human) datasets. B, the expression of miR-204 in brain, lung, liver, stomach, colon, intestine, spleen, kidney and bladder tissues of TP53 knockout mice, TP53 expression was checked by Western blot assay. C, the expression of primary miR-204 (pri-miR-204) in the brain, lung, stomach, colon, spleen, and bladder tissues of TP53 knockout mice. D-G, validation of TP53’s role in miR-204 maturation in primary cells (D, E) and gastric cancer cell lines (F, G). Data are presented as relative expression (mean ± SD) from three independent experiments. Statistical significance was assessed by unpaired Student’s t-test. H-I, in AGS cells transfected with si-NC or si-Drosha, (H) Western blot detection of Drosha protein expression, (I) qPCR analysis of mature miR-204 and precursor miR-204 expression; data represent the mean ± standard deviation from three independent experiments(**p < 0.01). J-K, (J) Western blot validation of Drosha knockdown efficiency and flag-p53 expression in each group of AGS cells. (K) qPCR analysis of mature miR-204 and pri-miR-204 expression in AGS cells transfected with si-NC, si-Drosha, or si-Drosha co-transfected with flag-p53; data represent the mean ± standard deviation from three independent experiments. Statistical significance was determined using the unpaired Student’s t-test(***p < 0.01)

Discussion

CSCs represent a rare cell subset across multiple human malignancies and drive tumor initiation, metastasis, therapeutic resistance and local recurrence. Accumulating in vivo and in vitro evidence has validated CSCs as promising therapeutic targets for innovative anticancer strategies. At the molecular regulatory level, microRNAs constitute a family of endogenous non-coding RNAs with a length of 20 to 24 nucleotides, which have emerged as central regulators in stem cell biology over the past decade [35]. These molecules function by binding to the 3’-untranslated region of target genes via base complementary pairing, thereby repressing mRNA translation or inducing transcript degradation to achieve precise modulation of gene expression. In normal stem cells, miRNAs govern cell fate decisions, with the miR-302 family contributing to the maintenance of embryonic stem cell pluripotency, and miR-125 participating in the differentiation balance of hematopoietic stem cells [36, 37]. In cancer stem cells, miRNAs broadly modulate stemness maintenance, cell proliferation and apoptosis resistance by targeting stemness-related signaling pathways, tumor suppressor genes and oncogenes [38–41].

Further investigations have elucidated that aberrant expression of pivotal miRNAs is intimately linked to the pathophysiological traits of CSCs [8]. Downregulation or deletion of tumor-suppressive miRNAs weakens their inhibitory effects on CSCs, leading to abnormal activation of downstream CSC-related genes [42, 43]. Such molecular events promote the generation, survival and stemness maintenance of CSCs. Changes in miRNA expression patterns constitute a core molecular basis for sustaining the malignant phenotype of CSCs, and also provide a theoretical basis for developing CSC-targeted therapeutic strategies through modulating miRNA expression [44]. In the present study, miR-204 was highly expressed in normal stem cells and participated in stem cell differentiation, suggesting its essential role in maintaining stem cell physiological properties. This viewpoint is supported by recent evidence that high miR-204 expression contributes to the undifferentiated state of neural stem cells [45]. Intriguingly, distinct from its profile in normal stem cells, miR-204 is markedly downregulated in cancer tissues, particularly in CSCs. Public data from TCGA datasets confirmed miR-204 downregulation occurs in multiple human malignancies including colon adenocarcinoma, bladder urothelial carcinoma, stomach adenocarcinoma, esophageal carcinoma, liver hepatocellular carcinoma, breast invasive carcinoma, head and neck squamous cell carcinoma, thyroid carcinoma, prostate adenocarcinoma, lung adenocarcinoma, kidney renal clear cell carcinoma and kidney chromophobe carcinoma. In gastric cancer, low miR-204 expression correlates with advanced TNM stage and unfavorable patient survival. Restoration of miR-204 expression in gastric cancer cells significantly suppresses cell proliferation, migration, invasion and stem cell traits. Compared with other well-recognized CSC-related miRNAs, miR-204 presents a unique bidirectional expression pattern with high levels in normal stem cells and low abundance in CSCs. It directly targets two crucial surface molecules CD44 and EPCAM, thus forming a core regulatory axis. Its prognostic value remains stable after rigorous statistical adjustment, highlighting its specific functional role in gastric cancer. These results are in line with previous reports indicating abnormal miR-204 expression modulates cell migration, proliferation, epithelial mesenchymal transition and glioma stem cell characteristics [46–48].

Several studies have confirmed that miR-204 suppresses cancer stemness by targeting stemness-associated genes including SOX4, Slug, HDAC1 and JAK2 [46, 49–51]. Nevertheless, other comprehensive analyses have reported elevated miR-204 expression in breast cancer stem cell spheroids [52]. In this work, we screened 21 CSC-related genes as potential downstream targets of miR-204. Further experimental validation showed that miR-204 downregulated the expression of CD44, ATM, EPCAM and JAK2 in gastric cancer [53, 54], whereas the expression of STAT3, HDAC1 and Snail remained unchanged. Our data also identified CD44 and EPCAM as two novel target genes of miR-204, with their protein expression levels negatively correlated with miR-204 abundance in gastric cancer tissues.CD44 is a transmembrane glycoprotein and a major cell surface receptor for hyaluronic acid. It is widely accepted as a core marker of cancer stem cells in multiple malignancies including gastric cancer [55, 56]. CD44 maintains CSC stemness by facilitating self-renewal, mediating crosstalk with the microenvironmental niche, and activating pro-survival cascades such as EGFR, Ras and MAPK signaling. Its expression is closely linked to tumor onset, metastatic spread and therapeutic resistance. EPCAM is a glycosylated surface protein that acts not only as a reliable marker for epithelial-originated cancer stem cells but also as an active regulator sustaining stemness phenotype [55, 57]. The intracellular fragment released by proteolytic cleavage of EPCAM can modulate intracellular signaling and regulate the expression of proliferation and pluripotency related genes such as c-Myc and cyclins. In gastric cancer, CD44 and EPCAM have been widely applied to isolate and characterize CSC subpopulations with high tumorigenic capacity [58, 59]. Our data revealed restoration of CD44 and EPCAM in miR-204 overexpressing cells significantly enhanced the stemness of gastric cancer cells, underscoring their functional importance in miR-204–mediated regulation of CSC properties.

Beyond verifying the inhibitory effect of miR-204 on gastric cancer stemness, our data uncover novel molecular mechanisms governing this regulatory process. The identification of CD44 and EPCAM as direct downstream targets of miR-204 constructs an unreported regulatory connection between this miRNA and key surface molecules that sustain the stem-like traits of gastric cancer cells. Moreover, our finding that TP53 facilitates miR-204 maturation further enriches the regulatory network modulating cancer stem cell behaviors in gastric cancer. The TP53-miR-204-CD44/EPCAM signaling axis acts as a vital mechanism to maintain cellular homeostasis. Disruption of this axis may drive the generation and sustained existence of treatment-resistant cancer stem cell populations. Accumulating evidence indicates that ambient pollutants and endocrine perturbations can interfere with signaling cascades involved in stem cell function and non-coding RNA regulation. Ambient particulate matter alters gene expression profiles associated with cell fate determination [60]; ambient sulfur dioxide was found to impact pathways involved in cellular differentiation [61]; and subclinical hyperthyroidism disturbs molecular networks responsible for cellular homeostasis [62]. These external factors may also interfere with the TP53-miR-204-CD44/EPCAM axis, though relevant exploration remains to be conducted.

VHL binds Elongin B, Elongin C and Cullin-2 to form a protein complex with E3 ubiquitin ligase activity. VHL mutation serves as a major cause driving miR-204 dysregulation in clear cell renal cell carcinoma. Gene profiling analysis revealed that VHL mutation and deletion occur frequently in clear cell renal cell carcinoma but are rarely detected in other tumor types. Protein interaction evidence indicated that VHL can interact with the tumor suppressor TP53 and further stabilize and transactivate TP53³⁶. Bioinformatic analysis showed that TP53 harbors a high mutation frequency in colorectal cancer, breast cancer, gastric cancer, head and neck cancer, and chromophobe renal cell carcinoma, whereas its mutation and deletion rate is less than 5% in clear cell renal cell carcinoma [63]. Such distinct gene alteration patterns of TP53 and VHL suggest that VHL acts as the key regulator of miR-204 in clear cell renal cell carcinoma, while TP53 dominates miR-204 regulation in chromophobe renal cell carcinoma and other malignancies. This hypothesis is supported by previous studies reporting that VHL upregulates miR-204 expression by inducing short transcripts of TRPM3 in clear cell renal cell carcinoma [64, 65]. Multiple studies have confirmed that TP53 modulates the processing, biogenesis and secretion of various miRNAs such as miR-34a and miR-1246 [66]. In this study, we verified that TP53 participates in miR-204 maturation. TP53 deletion markedly repressed mature miR-204 expression across multiple tissues, whereas TP53 restoration elevated mature miR-204 levels without affecting pri-miR-204 abundance. Importantly, Drosha knockdown experiments demonstrated that TP53-mediated miR-204 maturation is strictly Drosha-dependent. TP53 overexpression failed to rescue the reduction of mature miR-204 induced by Drosha depletion, indicating that TP53 facilitates pri-miR-204 processing mainly through the Drosha/DGCR8 microprocessor complex. In a TP53-deficient mouse glioma model, miR-204 expression was significantly downregulated in glioma stem cells derived from neural stem cells compared with normal neural stem cells. We also observed decreased miR-204 levels in gastric tissues with Helicobacter pylori infection, implying that miR-204 acts as a key regulator in cancer stem cell maintenance and gastric tumorigenesis.

While this study provides insights into the role of miR-204 in regulating gastric cancer stemness through the TP53–miR-204–CD44/EPCAM axis, several limitations warrant careful consideration. Regarding CSC identification, we used EPCAM-based sorting to enrich tumor-initiating cells, but gastric CSCs are highly heterogeneous and also encompass subtypes such as CD44⁺ CD24⁻ and Lgr5⁺ cells. Our findings on miR-204 therefore primarily reflect the EPCAM⁺ subset. Future multi-parameter sorting or single-cell sequencing studies are needed to fully delineate miR-204 expression and function across the broader spectrum of gastric CSC phenotypes. Beyond CSC heterogeneity, our conclusion regarding miR-204’s enrichment in normal stem cells and downregulation during differentiation is based primarily on public database analyses and extrapolation from tumor models, with no direct functional validation in normal gastric stem cells or organoid models. Given the technical challenges of isolating primary normal gastric stem cells, direct functional evidence—such as that from gain- or loss-of-function assays in gastric organoids—remains a critical gap in our understanding. While our bioinformatic analyses of public datasets suggest miR-204 contributes to normal stem cell differentiation, experimental validation of this causal relationship is currently restricted to cancer models. Establishing such normal gastric stem cell or organoid models in future research will be vital to delineate the physiological functions of miR-204 from its tumor-suppressive roles and to clarify causal relationships rather than merely correlative associations. Regarding the regulatory mechanism, we confirmed that TP53 regulates miR-204 maturation in a Drosha-dependent manner, but the precise molecular details remain incompletely resolved. It is not yet clear whether TP53 directly binds the Drosha/DGCR8 complex or regulates its transcription. Although our data support a post-transcriptional regulatory model, further studies using RNA Immunoprecipitation quantitative real-time polymerase chain reaction, Co-Immunoprecipitation, and Chromatin Immunoprecipitation are required to verify direct physical interactions and to distinguish validated functional dependence from unconfirmed structural mechanisms. Concerning preclinical evaluation, our in vivo data were obtained from subcutaneous xenograft models that do not fully recapitulate the native tumor microenvironment, limiting assessment of miR-204 effects on metastasis and stem cell niche interactions. Orthotopic transplantation or patient-derived xenograft models will be critical to evaluate the therapeutic potential of this axis under more pathophysiologically relevant conditions. Regarding clinical generalizability, all specimens were from a single center, which may introduce regional selection bias despite the relatively large cohort. Although TCGA data provide supportive evidence, multicenter and cross-regional validation remains indispensable to confirm miR-204 as a reliable prognostic biomarker before clinical application. Finally, from analytical and statistical standpoints, the bioinformatic identification of miR-204 relied on public GEO datasets that may not encompass all stemness-related miRNAs, potentially introducing selection bias. Statistical limitations also exist, including imbalanced sample sizes in survival analysis and a relatively small number of early-stage patients. Despite statistically significant associations, these imbalances call for cautious interpretation and further validation in larger, better-balanced cohorts.

Addressing these limitations in future studies will help clarify the biological and clinical significance of miR-204 in gastric cancer.

Conclusions

In this study, we identified miR-204 as a key differential factor between cancer stem cells (CSCs) and normal stem cells, was downregulated in CSCs but upregulated in normal stem cells. Critically, overexpression of miR-204 significantly repressed gastric cancer (GC) cell proliferation, metastasis, and stemness by directly targeting the CSC-related genes CD44 and EPCAM. Furthermore, we demonstrate that mutations in the tumor suppressor TP53 drives miR-204 dysregulation in cancer. These findings collectively proposed a novel model of CSC regulation (Fig. 8), enhancing our understanding of the fundamental distinctions between normal and malignant stem cells.

Fig. 8.

Schematic model illustrating the role of the TP53-miR-204-CD44/EPCAM axis in gastric cancer stem cells. A, in normal gastric epithelial cells or stem cells (left), wild-type TP53 promotes the maturation of miR-204. Mature miR-204 inhibits the translation of CD44 and EPCAM mRNA, thereby suppressing stemness properties and maintaining differentiation. In gastric cancer stem cells (right), TP53 mutation or loss-of-function leads to impaired maturation of miR-204. The subsequent downregulation of miR-204 releases its repression on CD44 and EPCAM, resulting in their aberrant overexpression. Elevated CD44 and EPCAM levels contribute to the acquisition and maintenance of cancer stemness features

These findings implicate miR-204 downregulation as a promising prognostic biomarker in gastric cancer, given its correlation with advanced TNM stage and poorer survival. Moreover, restoring miR-204 function to suppress tumorigenicity underscores its potential as a novel therapeutic target against therapy-resistant CSCs. To translate these insights, future work must delineate the molecular mechanism of TP53-mediated miR-204 maturation, develop efficient in vivo delivery systems for miR-204 mimics, and evaluate their therapeutic efficacy and synergy with standard chemotherapy in advanced models, ultimately paving the way for clinical trials. This work holds significant promise for developing novel targeted cancer therapies.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Abbreviations

CSC

Cancer stem cell

MTT

3-(4,5-dimethyl-2-thiazolyl)-2–5-diphenyl-2-H-tetrazolium bromide.

PCR

Polymerase chain reaction

Author contributions

Z.C, D.L and F.L designed the experiments and drafting the manuscript. D.L, Z.C, R.Z, H.Z and X.L carried out most experiments. D.L, Z.C and R.Z performed data analysis. H.X, L.C provided mice with different p53 genotypes, L.C help revised the manuscript. All authors have read and approved the final version of this manuscript.

Funding

This study was supported by grants from the National Natural Science Foundation of China (82003296) and Post-doctor Research Initiation Project of 2023 Joint Fund Project (2023-BSBA-362).

Data avaliability

No datasets were generated or analysed during the current study.

Declarations

Ethical approval

All animal experiments in our study were carried out in accordance with the Helsinki Declaration, and approved by the Ethics Committee of The First Affiliated Hospital of China Medical University. Patients were informed that the resected specimens were stored by the hospital and potentially used for scientific research, and that their privacy would be maintained. All patients provided informed consent prior to undergoing surgery. Our study protocol was approved by the Ethics Committee of China Medical University. Clinical trial number: not applicable.

Consent for publication

Not applicable.

Competing intersets

The authors declare no competing interests.