The human endometrial carcinoma cell lines, HEC-1-A and AN3CA, were propagated in McCoy’s 5A (Gibco, USA) and Minimum Essential Medium (MEM, Gibco, USA), respectively, as obtained from the Cell Bank of the Chinese Academy of Sciences. The media were supplemented with 10% fetal bovine serum (FBS, Gibco, USA) and 1% penicillin-streptomycin. Human endometrial epithelial cells (HEEC) were also sourced from the Cell Bank of the Chinese Academy of Sciences and were cultured in DMEM/F12 medium (Gibco, USA) with 10% FBS and 1% penicillin-streptomycin. Human mesenchymal stem cells (hMSCs), purchased from Wuhan Punuosai Life Technology (Wuhan, China), were cultured in serum-free medium specifically designed for hMSCs by Wuhan Punuosai Life Technology. All cells were cultivated in a controlled environment at 37°C with a 5% CO₂ atmosphere. When reaching 80-90% confluence, subculturing was performed using 0.25% trypsin-EDTA, ensuring cell integrity and experimental reproducibility.

The full 3’ untranslated region (3’UTR) of DLL3 was subsequently cloned into the pmirGLO vector (Promega, USA) for post-transcriptional regulation analysis. MiR-508-5p mimics and inhibitors, along with the aforementioned plasmids, were synthesized by Qingke Biotech Co., Ltd. (China). Verification of all constructs was achieved through Sanger sequencing.

The Cell Counting Kit-8 (CCK-8, Vazyme, China) assay was utilized to evaluate cell viability and proliferation post-transfection. Cells were plated in 96-well plates at densities tailored to each cell line’s growth characteristics and incubated for 24 hours to ensure adequate attachment. Following this period, CCK-8 solution was carefully added to each well, and the plates were incubated for a duration of 1-4 hours at 37°C in a 5% CO2 atmosphere. The optical density at 450 nm was determined using a microplate reader (BioTek Instruments, USA), providing a quantitative measure of cell viability. To ensure the reliability of the results, the assay was conducted in triplicate and averaged for statistical analysis.

For the colony formation assay, 500 cells were plated at low density in 6-well plates and cultured for 14 days, with the medium refreshed every 3 days. Colonies were fixed with methanol and stained with 0.1% crystal violet (Beyotime Biotechnology, China). The number of colonies, defined as a group of more than 50 cells, was counted manually using an Olympus light microscope (Olympus Corporation, Japan).

A confluent cell monolayer in a 6-well plate was scratched using a sterile pipette tip to create a wound. Detached cells were removed by washing, and the migration of cells into the wound area was documented at 0 and 48 hours using an Olympus phase-contrast microscope (Olympus Corporation, Japan). The wound closure percentage was calculated as (1 - 48h area/0h area) x 100%, then normalized to negative controls (NC, set as 1). Relative Closure = (Experimental group Closure %)/(NC Closure %).

Cell migration capabilities were assessed using Corning Transwell chambers (8.0 µm pore size). Cells were seeded into the upper chamber in serum-free media, whereas the lower chamber was filled with media supplemented with 10% FBS to serve as a chemoattractant. After a period of 24-48 hours, cells that had migrated through the membrane were fixed with 4% paraformaldehyde, stained with 0.1% crystal violet, and air-dried. For quantification, four randomly selected fields per membrane were imaged under an inverted microscope. Cells were manually counted and the averaged cell count per field was normalized to the negative control (NC) group (set as 1) to determine relative migration capacity.

Cells were lysed and protein concentrations were determined using a BCA protein assay kit (Thermo Fisher Scientific, USA). Equal amounts of protein were separated on SDS-PAGE gels and transferred to polyvinylidene fluoride (PVDF) membranes (Merck Millipore, Germany). Membranes were blocked and then incubated with a primary antibody against DLL3 (Abclonal, China, A18108), CD63 (Abclonal, China, A19023), CD81 (Abclonal, China, A4863), TSG101 (Abclonal, China, A1692), Calnexin (Abclonal, China, A4846). Following primary incubation, membranes were treated with HRP-conjugated secondary antibody (Abclonal, China, AS014). Protein bands were visualized using enhanced chemiluminescence (ECL) reagents (Meilunbio, China) and quantified with image analysis software.

Total RNA was extracted using the FastPure Cell/Tissue Total RNA Isolation Kit V2 (Vazyme, China), and miRNA was specifically extracted with the miRNeasy Kits (Qiagen, Germany), following the protocols provided by the manufacturers. Reverse transcription for mRNA was performed using the HiScript III 1st Strand cDNA Synthesis Kit (+gDNA wiper) (Vazyme, China), while miRNA reverse transcription utilized the miRNA 1st Strand cDNA Synthesis Kit (by stem-loop) (Vazyme, China).

qRT-PCR analyses were executed on a Bio-Rad CFX system using SYBR Green Master Mix (Vazyme, China). The primers used targeted DLL3, GAPDH, U6, and miR-508-5p, respectively. DLL3 primer F: CGTAGATTGGAATCGCCCTGAAG, R: CGTAGATGGAAGGAGCAGATATGAC, GAPDH primer F: ACAACTTTGGTATCGTGGAAGG, R:GCCATCACGCCACAGTTTC, U6(F: CTCGCTTCGGCAGCACA, R: AACGCTTCACGAATTTGCGT), miR-508-5p F: TACTCCAGAGGGCGTCACTCATG.

Gene expression data, corresponding survival data, and clinical information for patients were downloaded from the UCSC Xena database (TCGA-UCEC cohort) (https://xenabrowser.net/). For survival analysis, patients were stratified into high and low DLL3 expression groups based on the median expression value of DLL3. Kaplan-Meier survival curves were generated and compared using the log-rank test, implemented in the ‘survival’ R package (23). Univariate Cox proportional hazards regression analysis was also conducted using the ‘survival’ package to assess the prognostic significance of DLL3 expression (24). Differential gene expression analysis between the high and low DLL3 expression groups was performed using the ‘edgeR’ package in R. Genes with an adjusted p-value (FDR) < 0.05 and |log2FoldChange| > 2 were considered differentially expressed. A volcano plot was generated using the ‘ggplot2’ package (25) to visualize the differentially expressed genes. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses were performed using the ‘clusterProfiler’ package to identify significantly enriched biological processes and pathways associated with the differentially expressed genes (26). Gene Set Enrichment Analysis (GSEA) was performed using the ‘clusterProfiler’ package and the ‘ReactomePA’ package to further investigate the functional implications of DLL3 expression in UCEC (26, 27). The R software version used for all of the above analyses is 4.4.1.

Paraffin-embedded tissue sections underwent deparaffinization and rehydration, followed by antigen retrieval in a citrate buffer solution. The sections were then blocked to prevent non-specific binding and incubated with a primary antibody against DLL3 (Abclonal, China, A18108) overnight at 4°C. This step was succeeded by the application of HRP-conjugated secondary antibodies (Abclonal, China, AS014). For visualization, DAB (3,3’-Diaminobenzidine) was utilized as the chromogen, with sections subsequently counterstained with hematoxylin to highlight the nuclei. The prepared slides were examined under a light microscope. Quantitative assessment of staining intensity was meticulously performed using specialized image analysis software ImageJ, ensuring precise and objective evaluation of the expression levels of DLL3 in the tissue samples.

The 3’UTR regions of DLL3 were cloned into the pmirGLO vector, alongside control vectors, and co-transfected into target cells with miR-508-5p mimics or control sequences. Luciferase assays were conducted using the Dual-Luciferase Reporter Assay System (Beyotime, China), following manufacturer’s protocols. Firefly luciferase activity was measured and normalized against Renilla luciferase to assess relative luciferase expression. Measurements were performed using a microplate luminometer (Bio-Rad, USA).

The RIP experiment was carried out using an RIP Kit (Thermo Fisher Scientific, USA) following the provided protocol. Briefly, cells were lysed, and the lysates were incubated with magnetic beads conjugated to antibodies against Argonaute 2 (Ago2, Abclonal, China, A19709) or non-specific IgG (AC005, Abclonal, China) as a control. After incubation, RNA bound to the beads was isolated. The expression of miR-508-5p and DLL3 mRNA in the immunoprecipitated complexes was detected via qRT-PCR.

From 2023 to 2024, clinical samples were collected from Wuhan Children’s Hospital. The collection of clinicopathological samples was approved by the Medical Ethics Committee of Wuhan Children’s Hospital in accordance with the Declaration of Helsinki (approval number: 2024R045), with approval granted from May 6, 2024, to May 5, 2025. Immediately after surgery, the samples were submerged in liquid nitrogen, and subsequently, Wuhan Servicebio Technology Co., Ltd., was entrusted with the paraffin embedding of the samples.

To construct engineered MSCs, lentivirus overexpressing miR-508-5p was used to establish stable miR-508-5p overexpressing MSCs. The lentivirus was purchased from GeneChem (Shanghai, China). Briefly, MSCs were infected with the lentivirus, followed by puromycin selection 72 hours post-infection. Single MSC clones were picked, expanded, cultured, and confirmed for miR-508-5p expression.

The supernatant from the cultured engineered MSCs was collected into centrifuge tubes. The first centrifugation was performed at 500g for 10 minutes at 4°C, and the supernatant was carefully transferred to new centrifuge tubes. A second centrifugation was done at 2,000g for 20 minutes at 4°C, followed by supernatant collection. The third centrifugation was at 12,000g for 20 minutes at 4°C, with the supernatant collected and filtered using a 0.22μm filter.

For ultracentrifugation, a Beckman ultracentrifuge with an SW 32 Ti rotor was used at 120,000g for 90 minutes at 4°C. After centrifugation, the pellet was resuspended in PBS, aliquoted, and stored at -80°C.

Exosome samples were negatively stained for TEM imaging. The exosome samples were placed on 200-mesh copper grids. They were then stained with 1% phosphotungstic acid (Aladdin, China) for 1 minute. Subsequently, the samples were washed twice with PBS, each for 1 minute, and the grids were dried before visualization using a Hitachi HA7100 transmission electron microscope, operated at 80 kV.

Purified engineered exosomes overexpressing miR-508-5p were analyzed for particle size and concentration using a flow nanoanalyzer (nanoFCM, China). S23M-SEV (nanoFCM, China) was utilized as the particle size standard, while the Quality Control Nanospheres Series (nanoFCM, China) served as the concentration standard. The data were processed using NF Profession 1.16 software.

Nude mice (6-8 weeks old, female) were obtained from the Laboratory Animal Center at Wuhan University of Science and Technology, Wuhan, China. For the subcutaneous xenograft model, 5 × 10⁶ UCEC cells were resuspended in 100 µL of medium and injected subcutaneously into each mouse. In the metastasis model, 1 × 10⁶ UCEC cells were injected via the tail vein using a murine tail vein injector in a volume of 100 µL. The study was conducted in accordance with the guidelines of the Wuhan University of Science and Technology’s Animal Care and Use Committee, with ethical approval (2024116). Tumor growth was monitored by caliper measurements, and mice were euthanized when tumors reached 1500 mm³ in compliance with humane endpoints. Tumors were subsequently collected for further examination. In the metastasis model, euthanasia was performed 45 days post-injection, followed by isolation of lung tissue for HE staining.

Data were presented as mean ± standard deviation (SD) from at least three independent experiments. Statistical significance between groups was evaluated using Student’s t-test or one-way ANOVA, followed by post-hoc tests when applicable. A p-value < 0.05 was considered statistically significant. All analyses were performed using SPSS or GraphPad Prism software.

In an in-depth analysis of the expression profiles for 33 cancer types within the TCGA database, a lack of adjacent non-tumor tissue data precluded statistical analysis for 10 types of cancer. Among the remaining 22 cancer types, DLL3 expression was found to be downregulated in thyroid cancer (THCA) but significantly increased in several other cancers compared to adjacent non-tumor tissues ( Figure 1A ). Further survival data analysis of cancer types with significant changes in expression revealed that the expression level of DLL3 was significantly associated with overall survival (OS) in patients with renal cell carcinoma (KIRC) and UCEC, as indicated by Log-rank tests. However, results from the Cox proportional hazards model showed that high DLL3 expression was significantly associated with poor survival prognosis only in UCEC patients ( Figures 1B, C ). Additionally, for UCEC patients, those with high DLL3 expression had significantly worse disease-free interval (DFI) and disease-specific survival (DSS) ( Figures 1D, E ). Further analysis revealed no significant correlation between DLL3 expression and the patients’ grade or stage ( Figure 1F ), suggesting that DLL3 expression is independent of the clinical staging and has potential as an independent prognostic factor for cancer.

To delve deeper into the transcriptomic characteristics and potential biological functions of DLL3 in UCEC, a bioinformatics analysis was conducted. The results, using TCGA data, divided the samples into high and low expression groups based on the median value of DLL3 expression. Differential gene analysis (LogFC>2) identified a total of 735 genes with significant changes, including 166 upregulated and 569 downregulated genes ( Figures 2A, B ). Furthermore, GO functional enrichment analysis revealed the top three enriched terms as epidermis development, neuronal cell body, and monoatomic ion channel activity ( Figure 2C ). KEGG enrichment analysis highlighted Salivary secretion, Bile secretion, and Neuroactive ligand−receptor interaction as the top three pathways ( Figure 2D ). Reactome enrichment analysis showed Formation of the cornified envelope, GPCR ligand binding, and Keratinization as the leading categories ( Figure 2E ). Additionally, GSEA c5.go functional enrichment pinpointed significant enrichment in GOBP KERATINIZATION, GOCC SYNAPTIC MEMBRANE, and GOMF GATED CHANNEL ACTIVITY. GSEA c2.cp.kegg functional enrichment revealed predominant enrichment in KEGG NEUROACTIVE LIGAND RECEPTOR INTERACTION, KEGG ASCORBATE AND ALDARATE METABOLISM, and KEGG PENTOSE AND GLUCURONATE INTERCONVERSIONS ( Figure 2F ). These results suggest that DLL3 may regulate cancer development and progression by influencing various biological functions.

To confirm whether the expression level of DLL3 in UCEC aligns with database predictions, this study utilized immunohistochemistry, Western blotting, and other methods to assess DLL3 expression in collected samples from UCEC patients and corresponding adjacent non-tumor tissues. As indicated in Figures 3A–D , DLL3 expression was significantly higher in UCEC tissues compared to adjacent non-tumor tissues ( Figures 3A–D ), Besides, IHC results also showed higher PCNA and Ki67 expression in patients in the DLL3 high expression group or tumor group ( Figures 3F, G ). Further analysis through cell line studies, comparing HEEC with endometrial cancer cell lines (HEC-1-A and AN3CA), revealed that DLL3 expression levels in HEC-1-A and AN3CA cell lines were significantly elevated compared to HEEC ( Figure 3E ).

Previous functional enrichment results suggested that DLL3 may regulate the biological functions of UCEC by modulating the functionality of the intermediate filament cytoskeleton, which plays a crucial role in maintaining cell structure, crucial for the invasiveness and migration abilities of tumor cells. While previous studies have shown DLL3 regulates proliferation and migration in cancers such as gastric and prostate cancer, reports on its regulation in UCEC are absent. Therefore, DLL3 expression was knocked down in UCEC cell lines HEC-1-A and AN3CA. Knockdown efficiencies are presented in Figures 4A, C and Supplementary Figure S1 . The results indicated that shRNA-2 had the highest knockdown efficiency ( Figure 4A ). And the western blot analysis further confirmed that shRNA-2 effectively reduced the expression of the DLL3 protein ( Figure 4C , Supplementary Figure S1 ). Consequently, it was chosen for further experiments. Subsequently, the impact of DLL3 expression alteration on UCEC proliferative abilities was analyzed using CCK-8 and colony formation assays. Results indicated a significant reduction in UCEC cell proliferation following DLL3 knockdown ( Figures 4B, D, E ). Similarly, wound scratch and Transwell assays demonstrated reduced cell migration capabilities with decreased DLL3 expression ( Figures 4F–I ).

To explore the involvement of DLL3 in the in vivo progression of UCEC, nude mice were subcutaneously inoculated with HEC-1-A cells, wherein DLL3 was knocked down, along with a matched control group. Tumor volume was meticulously measured every six days, culminating in the euthanization of the mice 24 days after cell inoculation. Comparative analysis revealed a notable deceleration in the growth rate of tumors with DLL3 knockdown compared to controls, but the body weight exhibited no significant differences between them ( Figures 5A, B ). Subsequent examination of the excised tumors confirmed a significant reduction in both tumor weight and volume in the DLL3 knockdown group ( Figures 5C–E ). IHC analysis further underscored this finding, showing reduced proliferation markers Ki67 in the DLL3-suppressed tumors ( Figure 5F ), collectively indicating that DLL3 downregulation curtails tumor proliferation in vivo.

The HEC-1-A cells with diminished DLL3 expression were utilized alongside control group cells to evaluate their potential for metastasis in vivo. A pulmonary metastasis model was established by intravenously injecting tumor cells into nude mice. After 45 days post-injection, euthanasia was performed on the animals. Lung tissues were isolated and subjected to H&E staining. Subsequent examination under a microscope determined the presence of metastatic lesions in the lungs. As depicted in Figures 5G, H , within the control group, 6 out of 9 animals (66.6%) displayed pulmonary tumor nodules, whereas only 3 out of 9 animals (33.3%) injected with DLL3 suppressed HEC-1-A cells exhibited pulmonary tumor nodules. Furthermore, the nodules observed in animals from the sh-DLL3 group were notably smaller in size compared to those in animals inoculated with control group cells, indicative of diminished metastatic potential ( Figures 5G, H ). In summary, these findings suggest that inhibition of DLL3 expression attenuates the ability of HEC-1-A tumor cells to establish pulmonary metastases.

MiRNAs, endogenous RNA molecules approximately 22 nucleotides in length, play pivotal roles in gene expression regulation by targeting mRNAs. In the quest to identify miRNAs that regulate DLL3, bioinformatics analysis unveiled MiR-508-5p as a candidate with a potential binding site on the 3’ untranslated region (3’UTR) of DLL3 ( Figures 6A, B ). Subsequent validations using luciferase assays and RIP experiments established that DLL3 expression is modulated by MiR-508-5p targeting its 3’UTR ( Figures 6C–E ). Specifically, transfection of UCEC cell lines with MiR-508-5p mimic significantly attenuated the luciferase activity of the DLL3 wild type (WT), while the activity of a mutant DLL3 (MUT) lacking the binding site remained unchanged. Additionally, overexpression of MiR-508-5p led to a reduction in DLL3 expression within these cells ( Figures 6F–H ).

Empirical evidence has underscored that restoring MiR-508-5p expression notably hampers the proliferation and migration of UCEC cells ( Figures 6I–K ). Utilizing clinical UCEC samples, further investigation into the clinical significance of MiR-508-5p in UCEC revealed that the levels of MiR-508-5p were significantly reduced in UCEC tissues compared to adjacent non-tumor tissues ( Figure 6L ). And a clear negative correlation between MiR-508-5p and DLL3 expression was also revealed ( Figure 6M ). These findings reveal a potential mechanism by which MiR-508-5p inhibits DLL3 expression by targeting DLL3, thereby inhibiting the proliferation and migration tendency of UCEC cells.

To further investigate whether the deletion of MiR-508-5p leads to increased expression of DLL3, thereby promoting the proliferation and migration of UCEC cells, we conducted rescue experiments. Western blotting results revealed that, contrary to the previous findings with the introduction of MiR-508-5p mimic into UCEC cells, the addition of MiR-508-5p inhibitor significantly increased the expression levels of DLL3 ( Figures 7A, B , Supplementary Figure S2 ). Concurrently, CCK-8, wound scratch, colony formation and Transwell assays further demonstrated a significant increase in the proliferation and migration abilities of UCEC tumor cells exposed to the MiR-508-5p inhibitor compared to the negative control group ( Figures 7C–I ). However, this enhanced capability was partially reduced upon blockade of DLL3 ( Figures 7C–I ). These results collectively suggest that MiR-508-5p inhibits DLL3 expression by directly targeting DLL3, consequently suppressing the proliferation and migration of UCEC cells.

To efficiently deliver miRNA to tumor tissues, we developed engineered exosomes overexpressing MiR-508-5p based on MSCs. The specific isolation method is illustrated in Figure 8A . The extracted exosomes were observed using transmission electron microscopy, which revealed the characteristic bilayer membrane of exosomes ( Figure 8B ). Nano-flow cytometry showed that the particle size of the prepared exosomes ranged from 60 to 120nm, with a mean size of 84.6 ( Figure 8C ). Western blotting confirmed that exosome markers were higher in exosomes compared to donor cells ( Figure 8D ). Furthermore, after 4 hours of co-culture, MiR-508-5p expression in UCEC cells was significantly increased compared to the PBS control group ( Figure 8F ), indicating that the engineered exosomes effectively and precisely delivered miRNA to the tumor cells and restored MiR-508-5p expression. After 48 hours of co-culture, DLL3 expression was also significantly reduced ( Figure 8E ). Functional assays demonstrated that treatment with engineered MiR-508-5p overexpressing exosomes significantly decreased UCEC cell proliferation and migration ( Figure 8G ).

Additionally, to better evaluate the therapeutic efficacy of the engineered exosomes in vivo, a subcutaneous xenograft model of HEC-1-A was established. Once the tumors reached a volume of 100 mm³, tumor-bearing mice were treated by intratumoral injection with 10 μg of exosomes per mouse, administered every other day for a total of five injections. The endpoint for euthanasia was set at a tumor volume of 1500 mm³. Results showed that engineered MiR-508-5p overexpressing exosomes significantly reduced tumor burden in mice ( Figures 8H–K ).