Bone marrow mesenchymal stem cell-derived exosomal miR-93-5p suppresses esophageal squamous cell carcinoma progression and immune escape through targeting PD-L1 Authors: Wenbin Chen, Fangfang Fan, Renliang Fan, Guiju Kong, Yan Wang, Li Zou, Shizhen Wu Published: 2026 ABSTRACT BackgroundEsophageal squamous cell carcinoma (ESCC) is a malignant tumor with poor prognosis, and immune escape and abnormal cell behavior are key factors promoting its progression. Bone marrow mesenchymal stem cell (BMSC)-derived exosomes and microRNAs (miRNAs) play important roles in tumor regulation. This study aimed to explore the regulatory effect and molecular mechanism of BMSC-derived exosomal miR-93-5p on the malignant progression and immune escape of ESCC.MethodsThe BMSC-derived exosomes were extracted and characterized using transmission electron microscopy, size analysis and marker protein detection. miR-93-5p mimic or negative control was transfected into BMSCs. Cell counting kit-8 (CCK-8), 5-ethynyl-2’-deoxyuridine (EDU) staining, Transwell and 2’,7’-dichlorodihydrofluorescein diacetate (DCFH-DA) assays were used to detect the effects of exo-mimic on ESCC cells (KYSE-30, TE-1) viability, proliferation, migration and reactive oxygen species (ROS) accumulation. T cell-ESCC cell co-culture system was established to evaluate immune escape. The luciferase reporter, RNA-pulldown and fluorescence in situ hybridization (FISH) assays verified the targeting relationship between miR-93-5p and programmed death-ligand 1 (PD-L1). Nude mouse xenograft model was used to detect in vivo tumor growth.ResultsBMSCs and their exosomes were successfully isolated; miR-93-5p was upregulated in exosomes. Exo-mimic significantly inhibited ESCC cell viability, proliferation, migration, promoted ROS accumulation. Exo-mimic downregulated PD-L1 expression at both messenger ribonucleic acid (mRNA) and protein levels, enhanced T cell-mediated killing efficiency against ESCC cells, and significantly suppressed xenograft tumor growth in nude mice. Luciferase reporter, RNA-pulldown and FISH assays further confirmed that PD-L1 is a direct downstream target of miR-93-5p.ConclusionsBMSC-derived exosomes carrying miR-93-5p can inhibit ESCC progression via PD-L1 suppression and reverse tumor immune evasion in an in vitro T cell co-culture system, which provides a novel potential therapeutic target and strategy for the clinical treatment of ESCC. Highlight box Key findings • Bone marrow mesenchymal stem cell (BMSC)-derived exosomal miR-93-5p suppresses malignant progression of esophageal squamous cell carcinoma (ESCC) by inhibiting cell viability, proliferation and migration, and elevating reactive oxygen species (ROS) accumulation. Exosomal miR-93-5p directly targets and downregulates programmed death-ligand 1 (PD-L1), reversing ESCC immune escape and restraining in vivo tumor growth. What is known and what is new? (Please discuss this question as two separated points) • ESCC has a poor prognosis due to tumor immune escape and abnormal malignant cell behaviors. BMSC-derived exosomes and microRNAs are critical regulators of tumor development. • This study clarifies the novel anti-tumor function of BMSC-exosomal miR-93-5p in ESCC and identifies PD-L1 as its direct target, revealing a new mechanism for ESCC immune regulation. What is the implication, and what should change now? • This finding provides a new therapeutic target and strategy for ESCC clinical treatment, offering novel insights for reversing ESCC immune resistance and inhibiting tumor progression. Introduction Esophageal cancer (EC) ranks among the most frequent digestive tract malignancies globally, characterized by substantial morbidity and mortality burdens (1). Within the histological spectrum of EC, esophageal squamous cell carcinoma (ESCC) constitutes approximately 90% of all diagnoses (2,3). Despite continuous therapeutic leaps in surgical techniques, radiotherapy, and chemotherapy over recent decades, the clinical prognosis for ESCC patients remains dismal, yielding a 5-year overall survival rate below 30% (4,5). This unfavorable outcome is tightly linked to delayed clinical detection, aggressive infiltrative patterns, propensity for metastasis, and the establishment of immune evasion, all of which compromise therapeutic efficacy (6,7). Consequently, unraveling the molecular machineries driving ESCC pathogenesis and pinpointing novel interventional targets are critical imperatives for enhancing clinical management. Immune evasion serves as a principal mechanism allowing neoplastic cells to bypass host defense surveillance, thereby fostering unabated tumor proliferation and dissemination (8). Programmed death-ligand 1 (PD-L1), a pivotal immune checkpoint receptor, exhibits widespread upregulation across multiple solid tumors, including ESCC (9,10). The engagement of tumor-expressed PD-L1 with its corresponding receptor, programmed cell death protein 1 (PD-1) on T lymphocytes, triggers T cell exhaustion and dampens anti-tumor immunity, constituting a core mechanism of immune escape (11). Consequently, disrupting the PD-L1/PD-1 signaling cascade has emerged as a compelling immunotherapeutic strategy. Nevertheless, the precise upstream regulators dictating PD-L1 expression in ESCC require further elucidation to optimize targeted therapies. Bone marrow mesenchymal stem cells (BMSCs), characterized by their multipotent stromal nature, have garnered considerable interest in oncology owing to their robust paracrine functions (12). Among the diverse bioactive factors secreted by BMSCs, exosomes act as primary conduits for intercellular communication (13). These nanoscale lipid-bilayer vesicles (30–150 nm in diameter) shuttle an array of biologically active molecules—including microRNAs (miRNAs), structural proteins, and lipids—from donor to recipient cells, thereby fundamentally altering the recipient’s biological properties (14). Current literature highlights a paradoxical role for BMSC-derived exosomes in oncology, as they can either accelerate or suppress tumorigenesis depending on specific cellular contexts and microenvironmental cues (15). However, their precise functional contribution to ESCC progression, particularly regarding the modulation of tumor immune evasion, remains largely unexplored. MicroRNAs represent a class of short (18–25 nucleotides), non-coding RNA molecules that orchestrate post-transcriptional gene silencing by anchoring to the 3' untranslated regions (3’UTR) of specific target mRNAs, culminating in transcript degradation or hindered translation (16). These epigenetic regulators are intricately involved in dictating tumor trajectories, encompassing cellular multiplication, programmed cell death, motility, and immune subversion (17). Specifically, miR-93-5p, a constituent of the miR-106b-25 cluster, exhibits dysregulated expression profiles across various malignancies and actively participates in oncogenic processes (18). Yet, the definitive expression landscape and mechanistic role of miR-93-5p in ESCC—especially when delivered via BMSC-derived exosomes—warrant comprehensive investigation. The current study hypothesized that BMSC-exosomal miR-93-5p might orchestrate ESCC dynamics by fine-tuning PD-L1 expression. To empirically test this, we isolated and validated BMSC-derived exosomes, subsequently engineering vesicles enriched with miR-93-5p (exo-mimic). We rigorously assessed the impact of these engineered exosomes on ESCC cell (KYSE-30 and TE-1) vitality, replication, migration, and reactive oxygen species (ROS) induction in vitro. Furthermore, a T cell-ESCC co-culture paradigm was established to delineate the influence of exo-mimic on immune evasion. A suite of molecular techniques confirmed the direct regulatory axis between miR-93-5p and PD-L1. Ultimately, an in vivo murine xenograft model was deployed to validate the tumor-suppressive efficacy of exo-mimic. This investigation endeavors to map the therapeutic potential of BMSC-derived exosomal miR-93-5p in combatting ESCC. We present this article in accordance with the MDAR and ARRIVE reporting checklists (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2026-0768/rc). Methods BMSC isolation and verification Bone marrow specimens were procured from clinical donors after securing informed consent. The study was approved by the Institutional Ethics Committee of Nanping First Hospital (approval No. NPSY202607050). The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. Primary BMSCs were isolated via density gradient centrifugation and maintained in Dulbecco’s Modified Eagle Medium (DMEM, Gibco, Carlsbad, CA, USA) supplemented with 10% fetal bovine serum (FBS, Gibco), 100 U/mL penicillin, and 100 µg/mL streptomycin under standard incubator conditions (37 ℃, 5% CO2). Cells at 80–90% confluency were utilized for downstream applications. Cellular identity was confirmed via flow cytometry (FACSCalibur, BD Biosciences) by assessing the presence of typical mesenchymal markers (CD73, CD105) and the absence of the hematopoietic marker CD45. Exosome purification and characterization Extracellular vesicles were extracted from BMSC conditioned media utilizing a rotary ultrafiltration approach. Initial centrifugation steps were applied to the collected supernatant to discard cellular debris. Following filtration through a 0.22-µm membrane (Millipore), the fluid underwent ultracentrifugation at 100,000 ×g for 2 hours in Ultra-Clear™ tubes (Beckman Coulter). This ultracentrifugation step was performed twice. The isolated exosomal pellets were resuspended and stored at −80 ℃. Vesicle morphology was visualized utilizing transmission electron microscopy (TEM; Hitachi H-7600, Japan), whereas particle size distribution was quantified via Nanoparticle Tracking Analysis (NTA; NanoSight LM10, Malvern Instruments). The abundance of exosomal hallmark proteins (CD9 and CD63) was verified through immunoblotting. Cell culture and treatment protocols Human ESCC cell lines (KYSE-30 and TE-1) were supplied by the Shanghai Cell Bank (Chinese Academy of Sciences). The cells were propagated in RPMI-1640 medium containing 10% FBS and 1% antibiotic solution (Gibco). To model immune interactions, T cells derived from peripheral blood mononuclear cells were primed with an anti-CD3 antibody (Abcam). These activated T cells were subsequently co-cultured with ESCC cells that had been pre-conditioned with either Exo-NC or Exo-mimic. For genetic modifications, miR-93-5p mimics, inhibitors, and corresponding controls (Sangon, Shanghai, China) were transfected into BMSCs or ESCC cells utilizing Lipofectamine 3000 (Invitrogen) following standard manufacturer guidelines. Transfected cells were harvested post-48 hours. Evaluation of cell proliferation and viability ESCC cells seeded in 96-well culture plates were exposed to Exo-NC or Exo-mimic for a duration of 48 hours. Cellular viability was quantitatively measured via a cell counting kit-8 (CCK-8 assay, Solarbio, Beijing, China) by recording absorbance at 450 nm on a microplate spectrophotometer (Shimadzu, Kyoto, Japan) after a 1-hour incubation with the reagent. Proliferative capacity was further assessed using a 5-Ethynyl-2’-deoxyuridine (EDU) incorporation assay kit (Beyotime, Shanghai, China). Positively stained proliferating cells were observed and quantified using an inverted fluorescence microscope (Olympus, Tokyo, Japan). Apoptosis assay (TdT mediated dUTP Nick End Labeling, TUNEL) Cell apoptosis was detected using the TUNEL Apoptosis Assay Kit (40306ES60, Yeasen, Shanghai, China). Briefly, cells were fixed with 4% paraformaldehyde, permeabilized with 0.1% Triton X-100 (BL934B, Biosharp, Hefei, China), and incubated with TdT enzyme and biotin-dUTP at 37 ℃ for 60 minutes. After washing, cells were incubated with HRP-conjugated streptavidin for 30 minutes at room temperature. Apoptotic cells were visualized using a fluorescence microscope, and the percentage of apoptotic cells was calculated. Flow cytometry assay T cells co-cultured with ESCC cells that had been pre-conditioned with either Exo-NC or Exo-mimic were collected, washed with PBS, and stained with specific fluorochrome-conjugated antibodies to evaluate their functional status. T cell activation was assessed using anti-CD25-PE and anti-CD69-FITC antibodies. T cell exhaustion was assessed using anti-PD-1-FITC and anti-TIM3-PE antibodies. All antibody reagents were purchased from BioLegend, including anti-CD69 (Cat. No. 310903), anti-CD25 (Cat. No. 302605), anti-PD-1 (Cat. No. 379205), and anti-TIM-3 (Cat. No. 364805). Flow cytometric data acquisition was conducted using flow cytometry (FACSCalibur, BD Biosciences), and analyzed with FlowJo software. Cell migration assessment Migratory behavior was analyzed utilizing 8-µm pore Transwell inserts (Corning, USA). Post-exosome exposure, approximately 1×105 ESCC cells in serum-free media were introduced to the upper compartments, while the lower chambers contained complete culture medium serving as a chemoattractant. Following a 48-hour incubation, non-migrating cells were wiped from the upper surface. Cells that traversed the membrane were fixed in 4% paraformaldehyde, stained with crystal violet (Beyotime), and visually enumerated under a light microscope. Intracellular ROS quantification Intracellular ROS accumulation was monitored utilizing the fluorescent probe 2’,7’-dichlorodihydrofluorescein diacetate (DCFH-DA, Beyotime). Following designated exosome treatments, ESCC cells were rinsed and incubated with the DCFH-DA solution in the dark for 30 minutes at 37 ℃. Subsequent to multiple washes with serum-free media, the intracellular fluorescence intensity was captured via fluorescence microscopy. Lactate dehydrogenase (LDH) cytotoxicity assay The cytolytic activity of T cells against ESCC populations was determined using an LDH release assay (Beyotime). Post co-culture, the supernatant was harvested and clarified via centrifugation. The LDH working solution was introduced, and optical density was recorded using a microplate reader. Cytotoxicity percentages were calculated based on the ratio of experimental LDH release relative to maximum cellular LDH content. Enzyme-linked immunosorbent assay (ELISA) Concentrations of IFN-γ and IL-10 within culture supernatants and tumor homogenates were quantified using specific ELISA kits (Solarbio). Prepared samples were added to specific antibody-coated microplates. Following incubation and washing steps, HRP-conjugated secondary antibodies were introduced. The subsequent colorimetric reaction was halted and measured spectrophotometrically to deduce cytokine concentrations based on standard curves. Fluorescence in situ hybridization (FISH) To determine the spatial relationship between miR-93-5p and PD-L1 transcripts, a FISH assay was conducted. FITC-conjugated miR-93-5p probes and cy3-labeled PD-L1 probes were hybridized with cellular samples overnight at 37 ℃ in a specialized buffer. Nuclei were subsequently counterstained utilizing 4’,6-diamidino-2-phenylindole (DAPI, Solarbio), and the fluorescent signals were analyzed utilizing a confocal laser-scanning microscope (Zeiss, Germany). Quantitative real-time PCR (RT-qPCR) Total RNA isolation was executed utilizing TRIzol reagent (Invitrogen), and concentrations were verified via NanoDrop. RNA was reverse-transcribed into cDNA using the PrimeScript RT Kit (Takara, Dalian, China). RT-qPCR was conducted utilizing SYBR Green master mix (Sigma-Aldrich) on an ABI 7900HT platform. Relative gene transcript levels were determined via the 2⁻ΔΔCt mathematical model. Primer sequences are detailed in Table S1. Immunoblotting analysis Proteins extracted using RIPA lysis buffer containing protease inhibitors were quantified via a BCA assay (Yeasen, China). Equivalent protein loads (30 µg) were resolved on SDS-PAGE gels and electro-transferred onto PVDF membranes. The membranes were blocked using skim milk and probed overnight at 4 ℃ with primary antibodies: anti-CD9 (1:1500, AF5139, Affinity, Shanghai, China), anti-CD63 (1:1,000, AF5117, Affinity, China), anti-PD-L1 (1:2,000, DF6526, Affinity, China), and anti-GAPDH (1:2,500, ab181602, Abcam, China). Post-washing, HRP-linked secondary antibodies were applied. Protein bands were visualized utilizing an enhanced chemiluminescence system (Pierce, Rockford, IL, USA). Dual-luciferase reporter assay Sequences of the PD-L1 3’UTR containing putative miR-93-5p binding motifs (wild-type, WT) or mutated variants (MUT) were inserted into luciferase reporter plasmids. ESCC cells were co-transfected with these plasmids alongside miR-93-5p mimics or NC oligonucleotides. Forty-eight hours post-transfection, a dual-luciferase reporter system (Vazyme, Nanjing, China) was employed to quantify luminescence, normalizing Firefly signals against Renilla activity to ascertain direct target engagement. RNA pulldown assay Biotinylated miR-93-5p and control probes (Sangon Biotech) were conjugated with C-1 magnetic beads (Life Technologies) at 25 ℃ for 2 hours. Cell lysates, obtained from formaldehyde-fixed and sonicated ESCC cells, were incubated with the probe-bead complexes for 4 hours. Following rigorous washing protocols, the bound RNA species were eluted, and the enrichment of PD-L1 mRNA within the complex was quantified via RT-qPCR. Xenograft tumor model Animal experiments were performed under a project license (No. IACUCFJABR2025060901) granted by the Institutional Animal Care and Use Committee of Fujian Anburui Biotechnology, in compliance with institutional guidelines for the care and use of animals. Female BALB/c nude mice (4–6 weeks old) obtained from Cavens (China) were housed in SPF facilities. Prior to inoculation, KYSE-30 cells were pre-treated in vitro for 24 hours with either exo-NC or exo-mimic (100 µg/mL). Mice were randomly allocated into three cohorts (n=5): a control group (untreated cells), an exo-NC group, and an exo-mimic group. Each mouse received a subcutaneous injection of 5×106 cells (100 µL PBS suspension) into the right flank. Tumor dimensions were recorded weekly utilizing digital calipers, calculating volume as (length × width2)/2. On day 35, mice were euthanized via sodium pentobarbital overdose, and the excised tumors were photographed and weighed for subsequent pathological evaluation. Immunohistochemistry (IHC) Resected tumor tissues were formaldehyde-fixed, paraffin-embedded, and sectioned at 4-µm thickness. Following deparaffinization, antigen unmasking was achieved using a boiling citrate buffer. Endogenous peroxidases were quenched before overnight incubation with anti-Ki67 (1:200, Affinity) and anti-PD-L1 (1:500, Abcam) primary antibodies at 4 ℃. Signal detection was facilitated by an HRP-conjugated secondary antibody and a DAB substrate kit (Beyotime), accompanied by hematoxylin counterstaining. Statistical analysis Quantitative data are presented as the mean ± standard deviation (SD). GraphPad Prism 7.0 was utilized for all statistical computations. Two-group comparisons were executed utilizing Student’s t-tests, whereas multigroup comparisons were evaluated via one-way ANOVA followed by appropriate post-hoc testing. The threshold for statistical significance was defined as P<0.05. Results Validation of BMSCs, derived exosomes, and miR-93-5p enrichment Primary BMSCs were successfully cultured, displaying typical morphology (Figure 1A). Flow cytometric profiling (Figure 1B) confirmed that the cells were highly positive (>95%) for mesenchymal markers CD73 and CD105, while negative for the leukocyte marker CD45, confirming their identity. TEM analysis of the isolated vesicles revealed classical cup-shaped, double-membrane architectures (Figure 1C). NTA profiling demonstrated a size distribution primarily spanning 90 to 120 nm (Figure 1D). Immunoblotting further indicated the enhanced enrichment of exosome markers CD9 and CD63 in the vesicle fractions compared to parent BMSCs (Figure 1E). Notably, basal miR-93-5p expression was higher in exosomes than in the originating BMSCs (Figure 1F). Following BMSC transfection with specific oligonucleotides, RT-qPCR detected elevated miR-93-5p abundance in exosomes harvested from mimic-treated cells (Figure 1G). Characterization of BMSCs and BMSC-derived exosomes, and expression of miR-93-5p. (A) Morphology of isolated and cultured MSCs. (B) Flow cytometry analysis of BMSC-associated markers (CD73, CD105) and hematopoietic marker (CD45) (n=3). (C) TEM image of BMSC-derived exosomes (n=3). Scale bar =100 nm. (D) NTA analysis of BMSC-derived exosomes (n=3). (E) Expression of exosome markers (CD9, CD63) in MSCs and exosomes were detected via Western blot assay (n=3). (F) The miR-93-5p expression in MSCs and BMSC-derived exosomes detected by RT-qPCR (n=3). (G) The miR-93-5p expression in exosomes from MSCs transfected with miR-93-5p mimic or NC (n=3). **, P<0.01. All the data are presented as the means ± SD. BMSC, bone marrow mesenchymal stem cell; MSC, mesenchymal stem cell; NC, negative control; NTA, Nanoparticle Tracking Analysis; RT-qPCR, real-time quantitative polymerase chain reaction; SD, standard deviation; TEM, transmission electron microscopy. Exosomal miR-93-5p suppresses ESCC malignant phenotypes and induces ROS To investigate cellular impacts, ESCC lines (KYSE-30 and TE-1) were cultured with exosomes extracted from transfected BMSCs for 48 hours. CCK-8 assessments (Figure 2A) and EdU incorporation assays (Figure 2B,2C) revealed that exo-NC slightly reduced cell viability and proliferation relative to untreated cells, whereas exo-mimic produced a stronger inhibitory effect. Similarly, Transwell experiments (Figure 2D,2E) indicated an attenuated migratory capacity in exo-NC-treated cells, with further suppression observed following exo-mimic treatment. Furthermore, intracellular ROS tracking via DCFH-DA (Figure 2F,2G) demonstrated that while exo-NC alone elevated ROS production relative to controls, while exo-mimic treatment increased intracellular oxidative stress levels. To further verify whether ROS is the cause of the inhibition of proliferation and apoptosis, we supplemented ROS inhibition rescue assay with ROS scavenger NAC (N-acetylcysteine) to neutralize intracellular ROS. The result of CCK-8 (Figure 2H) and TUNEL (Figure 2I,2J) assays demonstrated that NAC partially rescues exo-mimic-induced growth arrest and apoptosis. BMSC-derived exosomes carrying miR-93-5p regulated the viability, proliferation, migration and ROS accumulation of ESCC cells. ESCC cells (KYSE-30, TE-1) were treated with exosomes from NC-transfected (exo-NC) or miR-93-5p mimic-transfected (exo-mimic) BMSCs for 48 h. (A) Cell viability detected by CCK-8 assay (n=5). (B,C) Cell proliferation detected by EDU staining (n=5). Scale bar =20 µm. (D,E) Cell migration detected by Transwell assay (n=5). Stained with crystal violet; scale bar =50 µm. (F,G) Intracellular ROS accumulation detected by DCFH-DA fluorescent probe (n=5). Scale bar =50 µm. (H) Cell viability detected by CCK-8 assay (n=5). (I,J) Cell apoptosis detected by TUNEL assay (n=5). Scale bar =20 µm. **, P<0.01. All the data are presented as the means ± SD. BMSC, bone marrow mesenchymal stem cell; CCK-8, cell counting kit-8; DCFH-DA, 2',7'-dichlorodihydrofluorescein diacetate; EDU, 5-ethynyl-2'-deoxyuridine; ESCC, esophageal squamous cell carcinoma; NC, negative control; ROS, reactive oxygen species; SD, standard deviation; TUNEL, TdT mediated dUTP nick end labeling. Exo-mimic antagonizes ESCC immune evasion An in vitro co-culture system comprising activated T cells and exosome-pretreated ESCC cells was constructed to evaluate immune interplay. Molecular analysis revealed that ESCC cells conditioned with Exo-mimic exhibited a substantial downregulation of PD-L1 at both the transcript (Figure 3A,3B) and protein levels (Figure 3C-3F) compared to the exo-NC cohort. Crucially, LDH cytotoxicity assays showed that exo-mimic pretreatment increased the sensitivity of KYSE-30 and TE-1 cells to T cell-dependent lysis (Figure 3G,3H). Cytokine profiling of the co-culture media via ELISA supported this immunostimulatory shift; Exo-mimic significantly increased the secretion of the pro-inflammatory cytokine IFN-γ while suppressing the immunosuppressive factor IL-10 (Figure 3I,3J). Moreover, Flow cytometry was used to characterize T cell subsets following co-culture and the result of Figure 3K-3N indicated that exo-mimic promoted T cell activation through upregulating CD69+/CD25+ cells proportion, while suppressing T cell exhaustion with decreasing PD-1+/TIM-3+ cells proportion. These findings highlight the capability of exo-mimic to modulate ESCC immune resistance. BMSC-derived exosomes carrying miR-93-5p regulated the immune escape of ESCC cells. ESCC cells were pretreated with exo-NC or exo-mimic, then co-cultured with activated T cells. (A,B) PD-L1 mRNA expression detected by RT-qPCR (n=3). (C-F) PD-L1 protein expression detected by Western blot (n=3). (G,H) T cell killing efficiency evaluated by lactate dehydrogenase release assay (n=5). (I,J) IFN-γ and IL-10 levels in co-culture supernatant detected by assay kits (n=5). (K-N) Flow cytometry detection of T cell phenotypes after co-culture (N=5). Markers for T cell activation: CD69, CD25; Markers for T cell exhaustion: PD-1, TIM-3. **, P<0.01. All the data are presented as the means ± SD. BMSC, bone marrow mesenchymal stem cell; ESCC, esophageal squamous cell carcinoma; IFN-γ, interferon γ; IL-10, interleukin-10; NC, negative control; PD-1, programmed cell death protein 1; PD-L1, programmed death-ligand 1; RT-qPCR, real-time quantitative polymerase chain reaction; SD, standard deviation. Identification of PD-L1 as a direct target of miR-93-5p Bioinformatic alignment suggested a putative binding interface between miR-93-5p and the 3’UTR of PD-L1 (Figure 4A). Dual-luciferase reporter assays confirmed this interaction: co-transfection of miR-93-5p only suppressed the luminescence of the wild-type PD-L1 construct, with no change observed for mutant constructs (Figure 4B,4C). RNA-pulldown techniques further supported this physical interaction, demonstrating a significant enrichment of PD-L1 mRNA by biotinylated miR-93-5p pull-down fractions in both ESCC cell lines (Figure 4D,4E). FISH imaging confirmed overlapping intracellular localization of miR-93-5p and PD-L1 transcripts (Figure 4F). Subsequent gain- and loss-of-function experiments confirmed the regulatory axis. MiR-93-5p mimic transfection reduced PD-L1 mRNA (Figure 4G,4H) and protein levels (Figure 4I-4L), whereas miR-93-5p inhibitors produced the opposite effect. Furthermore, the correlation analysis of the expression of miR-93-5p and PD-L1 in ESCC patient samples (Figure 4M) exhibited a significant negative correlation between miR-93-5p and PD-L1 in ESCC clinical samples. In addition, the rescue assays were carried out to confirm that PD-L1 overexpression partially reversed effects of exo-miR-93-5p. The results indicated that PD-L1 overexpression significantly rescues the suppressive cell viability (Figure 4N,4O) and migration ability (Figure 4P,4Q) induced by exo-miR-93-5p, proving PD-L1 is the key downstream effector of exosomal miR-93-5p. PD-L1 was validated as a downstream target of miR-93-5p. (A) Predicted binding region between miR-93-5p and PD-L1 3’UTR. (B,C) Luciferase activity in cells co-transfected with miR-93-5p and wild-type or mutant PD-L1-3’UTR plasmid (n=3). (D,E) PD-L1 enrichment by biotin-labeled miR-93-5p detected by RNA-pulldown assay (n=3). (F) Co-localization of miR-93-5p and PD-L1 detected by FISH assay (n=3). Scale bar =20 µm. (G,H) PD-L1 mRNA expression in ESCC cells transfected with miR-93-5p mimic or inhibitor. (I-L) PD-L1 protein expression in ESCC cells transfected with miR-93-5p mimic or inhibitor. (M) Pearson correlation analysis of the expression of miR-93-5p and PD-L1 in ESCC patient samples (N=20). (N-O) Cell viability detected by CCK-8 assay (n=5). (P-Q) Cell migration detected by Transwell assay (n=5). Stained with crystal violet; scale bar =50 µm. *, P<0.05; **, P<0.01; ns, not significant. All the data are presented as the means ± SD. DAPI, 4',6-diamidino-2-phenylindole; ESCC, esophageal squamous cell carcinoma; FISH, fluorescence in situ hybridization; MUT, mutated variants; NC, negative control; PD-L1, programmed death-ligand 1; SD, standard deviation; WT, wild-type. Exo-mimic restricts xenograft tumor progression in vivo To translate our in vitro findings, mice were implanted with KYSE-30 cells previously exposed to exo-NC or exo-mimic. Over a 35-day observation window, the exo-NC cohort exhibited a moderate deceleration in tumor volumetric expansion and mass compared to untreated controls. Remarkably, tumors originating from exo-mimic-treated cells displayed the most significant growth restriction (Figure 5A-5C). Immunohistological evaluation of the excised tumors (Figure 5D-5G) revealed that exo-mimic treatment significantly decreased the proportion of Ki67-positive proliferating cells and suppressed PD-L1 tissue expression. Furthermore, intra-tumoral cytokine analysis mirrored the in vitro data, showing elevated IFN-γ and reduced IL-10 concentrations in the exo-mimic group compared to the exo-NC and control counterparts (Figure 5H,5I). Effects of BMSC-derived exosomes carrying miR-93-5p on ESCC xenograft tumor growth in nude mice. Nude mice were transplanted with KYSE-30 cells pretreated with exo-NC or exo-mimic, and tumor growth was monitored for 35 days. (A-C) Morphology of tumors, tumor volume and weight (n=5). (D-G) Ki67 and PD-L1 expression in tumor tissues detected by immunohistochemistry (n=5). Scale bar =50 µm. (H,I) IFN-γ and IL-10 levels in tumor tissues detected by assay kits (n=5). *, P<0.05; **, P<0.01. All the data are presented as the means ± SD. BMSC, bone marrow mesenchymal stem cell; ESCC, esophageal squamous cell carcinoma; IFN-γ, interferon γ; IL-10, interleukin-10; NC, negative control; PD-L1, programmed death-ligand 1; SD, standard deviation. Discussion ESCC constitutes a formidable clinical challenge, characterized by robust invasiveness, early metastatic dissemination, and an intrinsic capacity to subvert host immune surveillance (19). Despite the integration of multimodal therapeutic regimens encompassing surgery and contemporary immunotherapies, achieving durable remission in advanced ESCC cohorts remains exceptionally difficult (20). Recently, intercellular communication orchestrated by extracellular vesicles has emerged as a crucial determinant of the tumor microenvironment. Notably, exosomes secreted by BMSCs present a compelling platform for targeted oncological interventions due to their inherent biocompatibility, negligible immunogenicity, and proficiency in delivering therapeutic payloads directly to neoplastic cells (21,22). Prior investigations have firmly established that BMSC-derived exosomes can transfer a repertoire of regulatory molecules—most notably miRNAs—to modulate recipient tumor cell behavior (23). Our research aligns with this paradigm, utilizing BMSC-exosomes as biological vehicles for targeted miRNA delivery. Literature indicates that exosomal delivery of miR-1-3p can potently restrict the proliferation and metastatic colonization of EC by directly targeting E2F5 and dampening the MAPK/ERK signaling cascade (24). Similarly, BMSC-derived exosomal miR-10527-5p has been proven to hinder lymphatic metastasis in ESCC by disrupting Wnt/β-catenin signaling via Rab10 modulation (25). These established precedents strongly validate the premise of our study: engineered BMSC-exosomes can serve as potent delivery systems for tumor-suppressive miRNAs in ESCC management. The microRNA miR-93-5p acts as a context-dependent regulator across different cancer landscapes. In ovarian carcinoma, its overexpression precipitates cellular ferroptosis and blunts malignant behaviors via the targeted suppression of SLC7A11 (26). Conversely, in breast cancer models, miR-93-5p induces cell cycle arrest and sensitizes tumors to paclitaxel by downregulating the E2F1 and CCND1 networks (27). Our findings significantly expand upon this existing knowledge base. We demonstrate that BMSC-exosomal delivery of miR-93-5p exerts profound tumor-suppressive effects in ESCC. More importantly, we delineate a novel immune-regulatory axis wherein miR-93-5p directly targets and silences PD-L1, thereby reversing T cell exhaustion and restoring anti-tumor cytotoxicity. While our data present a compelling therapeutic strategy, certain limitations warrant consideration. Primarily, all in vitro functional assays were performed on two immortalized ESCC cell lines (KYSE-30, TE-1) rather than primary patient-derived ESCC cells or organoids. Tumor cell lines undergo long-term in vitro adaptive evolution and cannot fully recapitulate the heterogeneity, genetic background and stromal interactions of primary human ESCC lesions. Meanwhile, our immune evaluation system adopts a simplified two-cell co-culture system consisting only of purified T cells and ESCC cells, lacking other critical stromal components in the tumor microenvironment (cancer-associated fibroblasts, macrophages, myeloid-derived suppressor cells). This simplified platform cannot completely reflect complex multi-cellular immune crosstalk within native ESCC lesions. Subsequent research should employ patient-derived organoids co-cultured with autologous peripheral immune cells to validate our immune regulatory axis in a more physiologically relevant system. In addition, our in vivo model relies on athymic nude mice and although we successfully validated the direct inhibitory effect of exo-miR-93-5p on ESCC proliferation and PD-L1 expression in vivo, this model cannot evaluate the physiological interplay between exosomal miR-93-5p and endogenous tumor-infiltrating T cells. The immunomodulatory and immune-escape-reversing functions of exo-miR-93-5p are exclusively supported by our in vitro activated T cell co-culture system. Future studies using immunocompetent syngeneic ESCC mouse models are required to verify the in vivo immune regulatory axis under intact adaptive immunity. Furthermore, the pharmacokinetics, long-term biosafety, and immunogenic profiling of engineered BMSC-exosomes necessitate exhaustive pre-clinical evaluation. Future investigations should prioritize optimizing exosomal loading and targeted delivery techniques, alongside exploring synergistic combinations of exosome-based miRNA therapeutics with currently approved immune checkpoint blockades, aiming to maximize clinical benefits for ESCC patients. Conclusions Collectively, the findings in this study elucidate the exosomal miR-93-5p/PD-L1 axis as a critical modulator linking tumor malignant phenotypes and immune escape. This study provides significant mechanistic insights into the tumor-regulatory roles of BMSC-derived exosomal miRNAs in EC. Notably, it offers a promising preclinical basis for the development of exosome-based targeted therapeutics and immune combination strategies for ESCC treatment. Supplementary The article’s supplementary files as Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2026-0768/prf