Engineered CAR‐T–Derived Exosomes Co‐Delivering miR‐145 and Cytotoxic Proteins for Targeted Solid Tumour Therapy

ABSTRACT

Chimeric antigen receptor T cell (CAR‐T) therapy has demonstrated remarkable efficacy in haematologic malignancies but remains constrained in solid tumours due to limited tumour penetration, immunosuppressive microenvironments and the risk of cytokine release syndrome (CRS). Here, we develop a bioactive, cell‐free therapeutic platform by engineering exosomes derived from B7‐H3‐targeted CAR‐T cells and loading them with miR‐145 (Name this exosome as exo‐CT‐145). The exosomes retain CAR‐specific surface markers and cytotoxic payloads (perforin and granzyme), MiR‐145 is expressed at low levels in various tumours and can inhibit the occurrence and development of tumours through multiple pathways. Exo‐CT‐145 significantly inhibited proliferation, migration, and Epithelial Mesenchymal Transition (EMT) of oesophageal squamous cell carcinoma (ESCC) cells and induced apoptosis in vitro. In vivo, exo‐CT‐145 demonstrated tumour‐targeted accumulation, caspase‐3 activation, EMT reversal, angiogenesis suppression and remodeling the tumor microenviroment while no detectable CRS or systemic toxicity. This study proposes a synergistic nanotherapeutic paradigm integrating antigen‐specific killing and gene regulatory modulation, offering a promising direction for solid tumour treatment with improved safety and efficacy.

1. Introduction

Chimeric antigen receptor T (CAR‐T) cell therapy represents a transformative advance in oncology, empowering genetically modified T cells to precisely identify and eliminate malignant cells through engineered antigen‐specific receptors (June et al. 2018, Cappell and Kochenderfer 2023; Han et al. 2021). CAR‐T cells express synthetic receptors composed of an extracellular antigen‐binding domain, a transmembrane domain and intracellular signalling modules that activate cytotoxic programs upon antigen engagement (Benmebarek et al. 2019). Despite significant successes, exemplified by the approval of seven CAR‐T cell therapies targeting haematological malignancies (Tegenge et al. 2025), translating this efficacy into treatments for solid tumours remains elusive due to multiple biological and clinical challenges (Uslu and June 2025; Yan et al. 2023; Sterner and Sterner 2021). Limited trafficking of CAR‐T cells to tumour sites, impaired infiltration within dense stromal matrices and rapid functional exhaustion in immunosuppressive tumour microenvironments (TMEs) collectively undermine therapeutic effectiveness (Flugel et al. 2023; Labanieh et al. 2022). Moreover, systemic proliferation of activated CAR‐T cells can trigger severe cytokine release syndrome (CRS), posing substantial safety concerns that limit clinical application.

To overcome these constraints, exosomes—naturally occurring extracellular vesicles (30–150 nm in diameter) secreted by virtually all cell types—have emerged as promising therapeutic alternatives (Zhang et al. 2023; Buzas 2023; Herrmann et al. 2021; Witwer and Wolfram 2021). Exosomes inherently mediate intercellular communication by delivering diverse cargos, including nucleic acids, proteins and lipids, protected within a stable lipid bilayer membrane (Kalluri and LeBleu 2020; Wan et al. 2017). Their nano‐scale dimensions enable efficient penetration of biological barriers, selective tumour accumulation and prolonged systemic circulation, attributes highly advantageous for solid tumour therapy (Xie et al. 2019). Exosomes derived from immune cells have shown excellent results in countering tumours, such as dendritic cells (DCs), natural killer (NK) cells and T cells, as well as CAR‐T cells (Zhou et al. 2023; Lu et al. 2022; Haque and Vaiselbuh 2021; Di Pace et al. 2020). Critically, T cell‐derived exosomes intrinsically incorporate cytotoxic molecules such as perforin and granzyme, which potentiate targeted tumour cell apoptosis independent of cellular activation or proliferation (Hu et al. 2024). Similar to other cells, CAR‐T cells also secrete exosomes, studies have shown that exosomes derived from CAR‐T cells carry CAR molecules and contain tumour‐killing factors such as granzymes and perforin (Hu et al. 2024; Fu et al. 2019). Therefore, the exosomes possess some of the tumour‐killing capabilities of CAR‐T cells (Calvo and Izquierdo 2022). CAR‐T exosomes demonstrate exceptional potential as direct attackers in immunotherapy, exosomes generated in vitro and armed with human EGFR and HER2 specific CARs display robust efficacy against EGFR and HER2 human tumour cells (Fu et al. 2019). Yang Pengxiang et al. demonstrated that CAR‐T cell‐derived exosomes targeting mesothelin can significantly inhibit the growth of both endogenous and exogenous mesothelin‐positive triple‐negative breast cancer (Yang et al. 2021). CAR‐T cell‐derived exosomes do not contain programmed cell death protein 1 (PD‐1) protein, and thus can avoid the immunosuppression caused by tumour cells (Yamamoto et al. 2023). In addition, it also has been reported that exosomes derived from CAR‐NK cells also exhibit CAR‐NK cell‐specific cytotoxic activities, showcasing remarkable anti‐tumour efficacy (Tao et al. 2023). This discovery implies that leveraging exosomes sourced from CAR‐based immune cells (including CAR‐T, CAR‐NK et al) represents a pivotal advancement in propelling CAR therapies towards clinical implementation.

Oesophageal squamous cell carcinoma (ESCC) remains a lethal malignancy characterized by aggressive progression and poor prognosis, particularly in advanced stages where treatment options are severely limited (He et al. 2021). Targeting immune checkpoint molecules that are aberrantly expressed on tumour cells yet minimally present in normal tissues represents an ideal therapeutic strategy. Among these, B7‐H3 (CD276), widely overexpressed in ESCC and various solid tumours, has emerged as a compelling target due to its role in tumour progression, immunosuppression and angiogenesis (Xuan et al. 2021). Research findings suggest that abnormal overexpression of the B7‐H3 protein plays a pivotal role in the advancement of ESCC, mechanistically by enhancing tumour cell migration and invasion capabilities (Zhang et al. 2024). Moreover, this overexpression shows a significant correlation with poor prognosis in patients (Xiong et al. 2024). During the malignant transformation process of ESCC, the expression level of B7‐H3 protein exhibits a gradual upregulation trend, among patients with stage II and stage III ESCC, those with high B7‐H3 expression demonstrate a shorter overall survival (OS) (Xiong et al. 2024).

MicroRNA (miRNA) is a small RNA molecule, typically 20–24 nucleotides in length. It primarily functions by binding to the 3′‐UTR region of its target sequence, resulting in the inhibition of mRNA expression or sequence degradation. Consequently, miRNA plays a crucial role in a wide range of biological processes (Mohr and Mott 2015; Hayes et al. 2014). The downregulation of microRNA‐145 (miR‐145) expression is observed in a multitude of tumour tissues. It is now widely accepted that miR‐145 plays an inhibitory role in cancer progression (Jin et al. 2019; Shimonosono et al. 2019). MiR‐145 regulates the progression of various tumours by targeting different receptor proteins. It suppresses the expression of AKT3 in oesophageal cancer through direct interaction with the 3′‐UTR region. By directly modulating the PI3K/AKT signalling pathway, miR‐145 enhances the sensitivity of ESCC to cisplatin‐based chemotherapy (Zheng et al. 2019). Zhang et al. demonstrated that miR‐145 plays a tumour‐suppressing role in ESCC (Zhang et al. 2019), miR‐145 has been newly identified as an effective inhibitor of PLCE1 expression by directly targeting the 3′UTR of PLCE1. miR‐145 also inhibits cell proliferation, migration and metastasis, and regulates the cytoskeletal dynamics of oesophageal cancer cells (Cui et al. 2016), Tang et al. demonstrated that miR‐145 exerts tumour‐suppressive effects in human oesophageal squamous cell carcinoma by targeting phospholipase C epsilon 1 (Tang et al. 2019). To translate these biological insights into an effective therapeutic strategy, we engineered CAR‐T exosomes targeted to B7‐H3 and simultaneously loaded with miR‐145.

This dual‐function approach uniquely integrates cytotoxic receptor‐mediated targeting with miR‐145‐driven suppression of tumour‐promoting pathways, aiming to enhance therapeutic outcomes by inducing apoptosis, reversing EMT and inhibiting angiogenesis. Comprehensive characterization and evaluation of these engineered exosomes (exo‐CT‐145) in ESCC models revealed potent antitumour effects, providing strong evidence for their therapeutic potential. Our findings support the feasibility of a modular, cell‐free immunotherapy platform capable of overcoming key barriers inherent to solid tumour treatment, achieving enhanced efficacy with reduced systemic toxicities (Figure 1).

FIGURE 1.

Schematic of exo‐CT‐145 construction and the associated anti‐cancer mechanism (By Figdraw).

2. Materials and Methods

2.1. Cell Lines and Cell Culture, Antibody and Reagents

The cells utilized in this study were obtained from the Cell Bank of the Chinese Academy of Sciences (Shanghai, China). Specifically, 293T cells and HEEC were maintained in DMEM medium (WISENT, Nanjing, China) with 10% foetal bovine serum (WISENT, Nanjing, China), 1% penicillin and streptomycin. KYSE30, KYSE510, TE‐1, TE‐13 and KYSE150 cells were cultured in RPMI‐1640 medium (WISENT, Nanjing, China) with the same supplements. Anti‐B7‐H3 (Cat# 14453‐1‐AP, Proteintech), MMP‐9 polyclonal antibody (Cat# 27306‐1‐AP; Proteintech), Anti‐CD9 antibody (clone EPR23105‐121, Cat# ab236630; Abcam), Anti‐CD81 mouse monoclonal antibody [clone M38] (Cat# ab79559; Abcam)​, Recombinant Anti‐ALIX rabbit monoclonal antibody [clone EPR23653‐32] (Cat# ab275377; Abcam)​, Anti‐Calnexin recombinant rabbit monoclonal antibody (clone EPR3632, Cat# ab92573, Abcam), mouse anti‐CD8A monoclonal antibody conjugated to horseradish‐peroxidase (HRP) was purchase from Sino biological (SEK10980; working concentration: 1 µg/mL), Cleaved Caspase‐3 (Asp175) antibody (clone 5A1E, Cat# 9664S; Cell Signaling Technology) was used at a dilution of 1:1000 for Western blotting, Anti‐E‐Cadherin mouse monoclonal antibody (clone A0‐G11‐2, Cat# EM0502; HUABIO), Anti‐Caspase‐9 recombinant rabbit monoclonal antibody (clone SZ29‐01, Cat# ET1603‐27; HUABIO), Anti‐Bax recombinant rabbit monoclonal antibody (clone SZ3‐07, Cat# ET1603‐34; HUABIO), Anti‐Bcl‐2 recombinant rabbit monoclonal antibody (clone JF104‐8, Cat# ET1702‐53; HUABIO), Anti‐GAPDH rabbit polyclonal antibody (Cat# D110016; BBI), the secondary antibody used was HRP‐conjugated Mouse Anti‐Rabbit IgG (Order NO. D110065; BBI) and HRP‐conjugated Rabbit Anti‐Mouse IgG (Order NO. D110098; BBI). The ELISA kit was also purchased from Sangon Biotech (Shanghai, China) and Wuhan Huamei Biotech Co., Ltd (Wuhan, China). Annexin V‐FITC/PI Cell Apoptosis Detection Kit was purchased from ServiceBio (Wuhan, China). ExoLoad exosomal nucleic acid loading kit was purchased from Echo Biotech Co., Ltd, (Beijing, China). Transfection reagents purchased from Life‐iLab (Shanghai, China). Peripheral blood lymphocyte isolation solution purchased from Beijing Solarbio Co., Ltd. miR‐145 mimics were purchase from Shanghai Genepharma Co., Ltd.

2.2. Lentivirus Packaging and CAR‐T Cell Preparation

Three plasmids (B7‐H3, psPAX2, pMD2.G) were co‐transfected into 293T cells at a mass ratio of 4:2:1, the culture medium was changed 12 h later, subsequently, the cell supernatant was harvested between 24 and 72 h post‐transfection and stored at −20°C for subsequent viral concentration through ultracentrifugation. Pre‐clear cell debris by low‐speed centrifugation (5000 rpm, 30 min at 4°C), discard supernatant and repeat once, concentrate viral particles via ultracentrifugation (100,000 × g, 90 min at 4°C), resuspend the final pellet in serum‐containing RPMI‐1640 medium. Peripheral blood mononuclear cells (PBMCs) isolated from healthy adult donors were activated with anti‐human CD3/CD28 antibody‐coated beads at a bead‐to‐cell ratio of 1:1 (Gibco, 11161D), the cells were cultured in RPMI‐1640 medium supplemented with 10% exosome‐depleted fetal bovine serum (WISENT, Nanjign, China), 200 IU/mL recombinant human IL‐2, and 1% penicillin‐streptomycin. At 48 h post‐activation, the cells were transduced with lentivirus in the presence of 8 µg/mL polybrene to enhance infection efficiency and subsequently expanded for 9–14 days. During expansion, the cell density was maintained between 0.5 and 1 × 10⁶ cells/mL, with the medium being replenished every 2–3 days.

2.3. Lactate Dehydrogenase Release Assay

To evaluate the in vitro killing activity of CAR‐T cells against tumour cells at different effector‐to‐target ratios, the lactate dehydrogenase (LDH) release method was used for detection (Servicebio, G1610). In brief, target tumour cells in the logarithmic growth phase were inoculated at a density of 1 × 10⁴ cells per well into a 96‐well plate 24 h in advance. According to the experimental design, the following groups were set up: the experimental group (tumour cells co‐cultured with different numbers of CAR‐T cells), the sample spontaneous release control well (only containing tumour cells), the sample maximum enzyme activity release control well (only containing tumour cells) and the background blank control well (only containing culture medium). After group treatment, the culture plates were incubated together in a 37°C, 5% CO₂ incubator for 24 h. After incubation, the 96‐well plates were centrifuged at 400 × g for 5 min. For the sample maximum enzyme activity release control well, an appropriate amount of lysis buffer was added before centrifugation, and the samples were further incubated in the incubator for 60 min to ensure that the cells were fully lysed and released all LDH. Subsequently, 80 µL of the supernatant from each well was carefully aspirated and transferred to a new 96‐well plate. Each well was added with an equal volume (80 µL) of LDH detection working solution, and incubated at room temperature in the dark for 30 min. Finally, the absorbance values at 490 nm wavelength were measured using an enzyme analyser. The cytotoxicity percentage was calculated according to the following formula: Cytotoxicity (%) = (Experimental group OD₄₉₀ ‐ Sample spontaneous release control OD₄₉₀) / (Sample maximum enzyme activity release control OD₄₉₀ ‐ Sample spontaneous release control OD₄₉₀) × 100%. All experiments were independently repeated three times, and the data were expressed as mean ± standard deviation.

2.4. CAR‐T Exosome Acquisition

The B7‐H3 recombinant protein (100 ng/mL) was used to coat culture plates overnight at 4°C. CAR‐T cells (1 × 10⁶ cells/mL) were then seeded into the coated plates for antigen‐specific stimulation to enhance the secretion of CAR‐carrying exosomes. CAR‐T cell then expanded for a total of 9–14 days, during expansion, the cell density was maintained between 0.5 and 1 × 10⁶ cells/mL, with medium replenishment every 2–3 days. It is worth noting that during the cultivation of CAR‐T cells, the serum used needs to be subjected to ultracentrifugation to remove exosomes. Subsequently, the culture supernatant was collected and processed for exosome purification through sequential centrifugation. First, use a benchtop centrifuge at 5000 × g, 4°C for 30 min, repeating twice to remove cells and large debris (the difference in the weight of the centrifuge tubes before and after centrifugation should be ≤0.1 g, and the residual cell rate after centrifugation should be confirmed to be <0.1% through a haemocytometer); then, aliquot the pre‐treated supernatant into ultra‐speed centrifuge tubes, and use an ultra‐speed centrifuge (Thermo Scientific Sorvall WX+100) at 100,000 × g, 4°C for 90 min to centrifuge, discard the supernatant, resuspend the precipitate with pre‐cooled PBS (pH 7.4, 0.22 µm filtration), and centrifuge again at the same conditions for 70 min for washing; finally, resuspend the precipitate with 100 µL PBS. The exosomes we obtained were put through endotoxin detection using a dedicated endotoxin detection kit. The conclusive results unequivocally confirmed their freedom from endotoxin contamination. Subsequently, these exosomes were aliquoted and stored at −80°C for future use. Using the BCA Protein Concentration Assay Kit (P0010, Beyotime, Shanghai, China), the concentration of exosomes was preliminarily determined based on the protein content. In 27 mL of culture supernatant, approximately 40 ± 3.5 µg of exosomes could be obtained, corresponding to ∼1.48 ± 0.13 µg/mL.

2.5. MiR‐145 Mimics Loading

MiR‐145 mimics were purchased from GenePharma (www.genepharma.com). CAR‐T cell‐derived exosomes were loaded with miR‐145 mimics using the ExoLoad Exosomal Nucleic Acid Loading Kit (ECHO BIOTECH) according to the manufacturer's instructions. Briefly, 300 pmol of miR‐145 mimics were incubated with approximately 1 × 10^{1 1} exosome particles in the presence of 60 µL of exclusive EV‐Transit Peptide (ETP) under gentle shaking (150 rpm) at 37°C for 2 h. The miR‐145‐loaded exosomes were then transferred to 100 kDa ultrafiltration tubes (Millipore) and diluted to 4 mL with wash buffer. Centrifugation was performed at 4000 × g for 20–30 min to concentrate the exosomes to ∼100 µL. A second wash step was carried out by replenishing the wash buffer to 4 mL and recentrifuging to a final volume of 200 µL. Use nanoparticle tracking analysis to detect the number of exo‐CT‐145 particles. Exosomes were recovered by gentle pipetting of the retentate from the filter membrane and transferred to a fresh tube.

2.6. Real‐Time PCR

For gene expression analysis, cDNA was synthesized from exosomal RNA using the PrimeScript RT Reagent Kit (Takara, Dalian, China) in a 10 µL reaction system (detailed composition and temperature protocol in Table S3). Quantitative PCR was performed using TB Green Fast qPCR Mix with U6 snRNA as the internal control for normalization. Primer sequences are provided in Table S4, and the qPCR reaction setup and cycling conditions are detailed in Table S5.

2.7. ELISA for Quantification of CAR Expression on Exosomes

A sandwich ELISA was used to quantify surface expression of CAR on engineered exosomes. 96‐well plates were coated with recombinant human B7‐H3–Fc protein (2 µg/mL, 200 µL per well) at 4°C overnight. After washing, plates were blocked with 5% BSA (200 µL per well) at 37°C for 2 h. Serial dilutions of recombinant B7‐H3 scFv–CD8α standard (31.25–2000 ng/mL; expressed and purified from HEK293 cells) and exosome samples were then added and incubated at 37°C for 1.5 h. Exosome preparations were first adjusted to a total protein concentration of 1 mg/mL and subsequently serially diluted in PBS containing 1% BSA. After washing, an HRP‐conjugated anti‐human CD8α monoclonal antibody (Sino Biological, SEK10980; 1 µg/mL) was added and incubated at 37°C for 1 h in the dark. TMB substrate was applied for 15 min in the dark, and the reaction was stopped with 2 M H₂SO₄. Absorbance was measured at 450 nm. A four‐parameter logistic standard curve was fitted to the recombinant B7‐H3 scFv–CD8α data, and CAR levels in exosome samples were interpolated from this curve and expressed as ng CAR equivalents per mL.

2.8. Characterization of Exosomes

The morphology of exosomes was characterized using a transmission electron microscope. The specific procedure was as follows: 10 µL of exosome suspension (concentration of 10⁸–10⁹ particles per millilitre) was spread onto a carbon‐coated copper grid (300 mesh, Electron Microscopy Sciences) and allowed to adsorb at room temperature for 5 min. The excess liquid was removed using filter paper (Whatman grade 1), and then the grid was stained with 2% uranyl acetate dye (pH 4.5) for 1 min. The grid was allowed to air dry in white light for 10 min to enhance contrast. Then the exosomes were diluted at a ratio of 1:1000 (volume/volume) with ultrapure water. A total of 1 mL of the diluted suspension was transferred to a disposable polystyrene cuvette (optical path 10 mm, Sartorius Company). The particle size was detected using the Malvern Panaceo Zetasizer Nano ZS90 dynamic light scattering instrument at 25°C. Then, add the diluted exosome suspension to the Malvern Zeta potential sample cell, and conduct the zeta potential measurement of the exosomes according to the instrument operation procedures.

2.9. Uptake of Exosomes

Exosomes were labelled using DiI dye following a specific procedure. Initially, 1 µL of DiI dye (500×) was combined with 500 µL (10 µg/mL) of the exosome solution, vortexed vigorously for 1 min at room temperature, and then incubated on a shaker for 30 min in the dark. Subsequently, excess dye was removed through ultrafiltration. The labelled exosomes were then added to the cell culture dish for co‐incubation. Prior to filming, DAPI dye solution was applied for a 5 min staining period, followed by discarding the dye solution, rinsing twice with PBS and adding fresh culture solution for imaging observation under CLSM.

2.10. Cell Proliferation Assays

Cell proliferation was assessed using the MTT Cell Proliferation Assay Kit (Invitrogen) according to the instruction. KYSE30 cells and KYSE150 cells were cultured in 96‐well plates (5 × 10³ cells/well). After 48 h of culture, 15 µL MTT solution (5 mg/mL) was added into each well and incubated at 37°C for 4 h adding 150 µL DMSO to each well and shaking the mixture 10 min at room temperature. Absorbance was measured at 490 nm using a microplate reader.

2.11. Wound Healing Migration Assay

The wound healing migration assay was used to evaluate the migration ability of KYSE30 and KYSE150 cells. Cells were plated in six‐well plates then the cell monolayer was scraped using a 2.5 µL micropipette tip. Cells were washed twice with sterile PBS solution, and the medium containing exosomes was added. Cells were observed at 0, 24 and 48 h after wounding, cell migration ratios were calculated using Image J software.

2.12. Transwell Assay

Matrigel was thawed overnight at 4°C, after which Matrigel: 1640 medium was incubated at 1:9. Then 100 µL diluted matrigel (NEST Biotech, Wuxi, China) was added to each chamber (Corning Transwell 24 well plates, Cat. #3422), the assembled 24‐well plates (containing the inserts) were placed in the incubator for film formation. After the Matrigel was solidified, 100 µL medium was added and hydrated for 30 min. 2 × 10⁴ cells were seeded in each well, and then waiting for cell invasion. The chambers were removed and fixed in 4% paraformaldehyde fixative for 1 h. Then the chambers were immersed in 0.1% crystal violet solution and stained for 15 min. The cells in the upper chamber were gently erased, and the number of invaded cells was recorded under a microscope.

2.13. Western Blot

The cells were lysed with RIPA lysate, centrifuged at 12,000 × g for 10 min in a centrifuge, and the supernatant was the total cell protein. Protein loading buffer was added and boiled at 100°C for 10 min. After that, the denatured proteins (∼20 µg of total protein) were separated by electrophoresis in 10% SDS‐PAGE gel and transferred to PVDF membrane. The membrane was blocked with 0.5% skim milk powder for 90 min, incubated overnight at 4°C in primary antibody diluent, washed three times with TBST, incubated in HRP‐conjugated secondary antibody diluent for 1 h at room temperature, washed three times with TBST solution, and exposed in ECL hypersensitive luminescent solution.

2.14. Flow Cytometry With Annexin V/PI Double Staining

The cell culture supernatant was collected, then, after digestion with trypsin without EDTA, the cells were pooled with the cell culture supernatant and collected by centrifugation at 500 × g for 5 min at 4°C. The cells were washed twice with precooled PBS and collected by centrifugation at 500 × g at 4°C for 5 min each time. The cells were gently resuspended in 1× Binding Buffer, and the cell concentration was adjusted to 1–5 ×10⁶/mL. Then, 5 µL Annexin V‐FITC and 5 µL PI (Annexin V‐FITC/PI double‐staining assay, Servicebio, Catalog No. G1510) were added to 100 µL cell suspension, and the mixture was gently mixed and kept in a dark place at room temperature for 8–10 min. Then, 400 µL of precooled 1 × Binding Buffer was added, gently shaken, and detected by flow cytometry or fluorescence microscopy within 1 h, Accordingly, the quadrants were interpreted as: viable (Annexin V⁻/PI⁻), early apoptotic (Annexin V⁺/PI⁻), late apoptotic/secondary necrotic (Annexin V⁺/PI⁺) and necrotic (Annexin V⁻/PI⁺).

2.15. ELISA Assay

ELISA was performed to quantify granzymes and perforin in exosomes using a biotin–streptavidin–HRP–based sandwich immunoassay (Order NO. D711226, Order NO. D715517, BBI). Briefly, 96‐well plates were pre‐coated with capture antibodies (anti‐human perforin or granzyme) overnight at 4°C. After blocking with 5% BSA for 1 h at 37°C, standards (recombinant perforin/granzyme) and exosome samples (normalized by total protein concentration, Exosomes undergo lysis) were added (100 µL/well) and incubated for 90 min at 37°C (Exosomes were diluted to an appropriate concentration‐). Plates were washed four times with PBS containing 0.05% Tween‐20. Subsequently, 100 µL of biotin‐labelled detection antibody was added per well and incubated for 60 min at 37°C, followed by washing. Then, 100 µL of HRP‐conjugated streptavidin working solution was added and incubated for 30 min at 37°C in the dark. After washing, 90 µL of TMB substrate was added and developed for 15 min at 37°C in the dark. The reaction was stopped with 50 µL of 2 M H₂SO₄, and absorbance was measured at 450 nm. A standard curve was generated using serial dilutions of recombinant protein, and the concentrations of perforin and granzymes in exosome samples were interpolated accordingly.

2.16. Construction of ESCC Mice Model

Female nude mice aged 6–8 weeks were selected to construct a subcutaneous ectopic oesophageal carcinoma model. 5 × 10⁶ KYSE‐150 cells were injected on day 0, and the tumour volume was about 75 mm³ after 10 days. The mice were randomly divided into four groups. Equal‐volume PBS, exo‐T, exo‐CT and exo‐CT‐145 were injected into the tail vein every 3 days for seven consecutive times, during which the tumour volume and weight of the mice were detected every other day, and the curves were drawn. All animal‐related experiments and human peripheral blood mononuclear cell experiments are conducted in accordance with the guidelines for evaluation and approval by the Ethics Committee of Henan University of Science and Technology (NO. 2024061280).

2.17. Immunohistochemistry Assay

Tissue samples were collected and subjected to fixation, embedding and sectioning, followed by dewaxing, hydration at room temperature and antigen retrieval. Sections were then blocked with goat serum to reduce nonspecific binding. For primary antibody incubation, sections were incubated with 50–100 µL of optimally diluted primary antibody at 4°C overnight in a humidified chamber to ensure consistent antibody‐antigen binding. After incubation, sections were washed four times with PBS (each wash for 5 min) to remove unbound primary antibodies. Subsequently, 50 µL of biotin‐labelled goat anti‐rabbit IgG working solution (secondary antibody) was added per section and incubated at room temperature for 1–3 h, followed by another PBS wash. Then, 50 µL of diluted streptavidin‐POD working solution was applied for 30 min to 1 h at 37°C to amplify the signal. Freshly prepared DAB chromogenic solution was added dropwise, and staining intensity was monitored under a microscope until the desired signal‐to‐noise ratio was achieved. Sections were counterstained with haematoxylin for 5–10 min, dehydrated through gradient alcohols, cleared in xylene and mounted with neutral balsam, mounted slides were stored under dry conditions.

2.18. HE Stain

Paraffin sections were dewaxed to water, and the sections were washed in xylene I 10 min‐xylene II 10 min‐anhydrous ethanol I, 5 min‐anhydrous ethanol II, 5 min‐95% alcohol, 5 min‐90% alcohol, 5 min‐80% alcohol, 5 min‐70% alcohol and 5 min‐distilled water in turn. After that, the nucleus was stained with haematoxylin, the specific steps were as follows: immerse the sections in Harris haematoxylin for 3–8 min, wash with tap water, differentiation with 1% hydrochloric acid alcohol for a few seconds, rinse with tap water, return blue with 0.6% ammonia water, and rinse with running water. Then the cytoplasm was stained with eosin for 1–3 min. Seal with neutral gum after dehydration. After that, it can be microscopically examined and the images collected and analysed.

2.19. Statistical Analysis

Statistical analysis was performed using Graphpad prime 10.1.2 software and SPSS 22.0. All values are presented as mean ± standard deviation. The one‐way analysis of variance, the chi‐square test and two factors analysis of variance were used for statistical analysis with p < 0.05 considered to be statistically significant (The statistical methods for each statistical chart are shown in Table S6).

3. Results

3.1. Expression of B7‐H3 in ESCC

The Cancer Genome Atlas (TCGA) and Genotype‐Tissue Expression (GTEx) databases were queried (Text S1). The analysis of gene expression revealed that B7‐H3 expression was significantly elevated in oesophageal cancer tissues compared to adjacent tissues (Figure 2a). By querying the ESCCdb database (http://cailab.labshare.cn/ESCCdb/search), it was revealed that among the 16 analysed gene expression datasets of ESCC, as many as 12 datasets clearly demonstrated that the expression level of B7‐H3 protein in ESCC tissues was significantly higher than that in normal tissues, and this expression difference exhibited a high degree of statistical significance (Figure S1). B7‐H3 expression was significantly upregulated in ESCC tissues compared to paracancerous tissues (70.5% vs. 42.6% positivity; χ ² = 9.64, p < 0.01), as assessed by immunohistochemistry in 61 paired samples (Figure 2b and Table S1). Moreover, various ESCC cell lines and human oesophageal epithelial cells (HEEC) were used to verify B7‐H3 expression. The results indicated that B7‐H3 expression was remarkably increased in cancer cell lines compared to HEEC cells (Figure 2c,d).

FIGURE 2.

B7‐H3 is highly expressed in ESCC. (a) Differential expression of B7‐H3 between ESCA and normal tissues (Unpaired t‐test, data points are plotted from maximum to minimum value, ****p < 0.0001); (b) Immunohistochemical (IHC) analysis of B7‐H3 expression in ESCC and para‐carcinoma tissue; (c) Western blot results of total B7‐H3 protein in different ESCC cell lines and HEEC cell; (d) Quantitative analysis of B7‐H3 expression in various cell lines (ANOVA, n = 3, data represents the mean ± SD, ****p < 0.0001).

3.2. Construction and Characterization of CAR‐T Cells

The CAR lentiviral vector targeting B7‐H3 (Text S2 and Figure 3a), along with psPAX2 and pMD2.G, was transfected into HEK‐293T cells. Green fluorescence confirmed the successful transfection (Figure S2). In parallel, T cells were stimulated in vitro with anti‐CD3 and anti‐CD28 antibodies (Figure S3), demonstrated by a significant upregulation of CD25 (Figure S4), a key T cell activation marker. The harvested lentiviral particles successfully transduced T cells, as evidenced by overexpression of fluorescence proteins (Figure 3b). The expression of CAR on T cells was verified by western blot (Figure 3c). The transduction efficiency of CAR‐T cells was 93.5% (Figure 3d). Lactate dehydrogenase (LDH) experiments verified the killing efficacy of CAR‐T cells against B7‐H3 expressing KYSE30 cells at ratios of 1:1, 2:1, 5:1 and 10:1 (Figure 3e). High effector cells to target cells (ET) ratios correlated with increased target cell killing. KYSE150 cells also exhibited similar outcomes (Figure 3f). While the cytotoxic efficiency of untransduced T cells significantly diminished when co‐cultured with tumour cells (Figure 3g,h).

FIGURE 3.

Construction and characterization of CAR‐T cells. (a) Schematic diagram of structure of CAR; (b) Intrinsic fluorescence of T cells successfully transduced with B7‐H3 CAR lentivirus; (c) The expression of CAR in T cells was detected by western blot; (d) Transduction efficiency of CAR‐T cells measured by flow cytometry; (e) Determination of KYSE30 cell mortality at varying effector‐to‐target ratios via LDH release assay; (f) Determination of KYSE150 cell mortality at varying effector‐to‐target ratios via LDH release assay; (g) Determination of KYSE30 cell mortality at varying effector‐to‐target ratios via LDH release assay; (h) Determination of KYSE150 cell mortality at varying effector‐to‐target ratios via LDH release assay; (ANOVA, n = 3, data represents the mean ± SD, *p < 0.05, **p < 0.01, ****p < 0.0001).

3.3. Characterization of Exosomes

Under transmission electron microscope (TEM), both exo‐T and exo‐CT exhibit the characteristic saucer‐shaped morphology. No significant difference was observed in their average size and ζ‐potential (Figure S5). Approximately three copies of miR‐145 per exosome were loaded into exo‐CT to generate exo‐CT‐145. RT‐PCR results showed that miR‐145 in exo‐CT‐145 increased by >30 times compared to exo‐CT (Figure S6). exo‐CT‐145 also exhibited a ‘saucer‐shaped structure’ with an average size of 115 nm and an average ζ‐potential of ‐20.6 mV (Figure 4a–c). Nanoparticle tracking analysis measured the concentration of exo‐CT‐145 particles to be 5.97 × 10¹¹ particles/mL. The ELISA experiment confirmed the approximate content of CAR: approximately 0.56 ng/µg of exosomes (Figure S7). The expression of CAR, CD9, TSG101, ALIX and calnexin in each group was assessed by western blot (Figure 4d). ELISA quantified the levels of GzmB and perforin amount in each respective group (Figure 4e,f). The concentrations of GzmB and perforin were 214 and 176 ng/mL in exo‐CT‐145, 205 and 175 ng/mL in exo‐CT, and 126 and 133 ng/mL in exo‐T, respectively. GzmB and perforin levels in exo‐CT and exo‐CT‐145 were higher than in exo‐T, likely due to B7‐H3 stimulation. It is notable that due to the use of nucleic acid loading kits and the loading of miR‐145 into exosomes through transport peptides (ExoLoad–Extracellular Vesicle RNA Loading Kit), the loading of miR‐145 did not lower GzmB and perforin levels in Exo‐CT‐145. After evaluating the uptake of exosomes by HEEC cells and KYSE30 cells, it was found that HEEC cells exhibited relatively weak exosome uptake capacity, with their uptake amount being significantly lower than that of KYSE30 cells (Figure 4g). KYSE30 cellular uptake assays revealed that the presence of the CAR significantly enhanced the internalization of exo‐CT and exo‐CT‐145 by recipient cells compared to exo‐T (Figure 4h). A similar uptake trend was observed in KYSE150 cells as in KYSE30 cells (Figure S8).

FIGURE 4.

Characterization of exosomes. (a) The morphology of exo‐CT‐145 under TEM; (b) The size distribution of exo‐CT‐145; (c) The average ζ potential of exo‐CT‐145. (d) CAR, CD9, ALIX, TSG101 and calnexin in each group and HEK293T cell; (e) The level of GzmB in each group; (f) The level of perforin in each group; (g) After 4 h of culture, DiL‐labelled exosomes were internalized in HEEC cells (Scale bar = 20 µm); (h) After 4 h of culture, DiL‐labelled exosomes were internalized in KYSE30 cells (Scale bar = 20 µm). (ANOVA, n = 5, data represents the mean ± SD, ****p < 0.0001).

3.4. Treatment Effect In Vitro

In B7‐H3 overexpressing KYSE30 and KYSE150 cells, exo‐T, exo‐CT and exo‐CT‐145 exhibited dose‐dependent cytotoxicity (Figure 5a). After 48 h of exposure to 1 mg/mL exosomes, exo‐CT and exo‐CT‐145 significantly decreased KYSE150 cell survival compared to exo‐T, with survival rates of 32% and 29% versus 58%. A similar trend was observed in KYSE30 cells, where exo‐CT and exo‐CT‐145 resulted in survival rates of 64% and 42%, respectively, compared to 81% for exo‐T. At the same concentration, their cytotoxicity to B7‐H3 low expressing HEEC cells was minimal, with survival rates of 93% for exo‐T, 91% for exo‐CT and 83% for exo‐CT‐145. Invasion and migration assays further verified the inhibitory effect of exosomes on B7‐H3 overexpressing cells (Figure 5b–e). In KYSE150 cells, exo‐T, exo‐CT and exo‐CT‐145 reduced invasive cell counts by 27%, 40% and 70%, respectively. Similarly, KYSE30 cell invasion decreased by 31%, 45% and 75%, respectively. Migration assays revealed a reduction in KYSE150 cell mobility from 60% to 38%, 22% and 13%, respectively, after 48 h. Under the same conditions, KYSE30 cell mobility decreased from 49% to 37%, 20% and 12%, respectively. Based on these findings, subsequent experiments used an exosome concentration of 1 mg/mL.

FIGURE 5.

Effects of exosomes on cell proliferation, invasion, and migration. (a) Dose‐dependent effect of exosomes on viability of HEEC, KYSE150 and KYSE30 cells (two‐way ANOVA, n = 3, Each data represents mean ± SD, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001); (b–c) Inhibitory effect of 1 mg/mL exosomes on KYSE150 and KYSE30 cell invasion (ANOVA, n = 3, Each data represents mean ± SD, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001); (d and e) Inhibitory effect of 1 mg/mL exosomes on KYSE150 and KYSE30 cell mobility (two‐way ANOVA, n = 3, each data represents mean ± SD, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001).

3.5. Exosome Infiltration in ESCC Spheroids

To assess exosome infiltration in solid tumours, KYSE150 spheroids were prepared. Exosome penetration into spheroids was visualized and confirmed by a confocal microscopy using exosomal fluorescence signals (Figure 6a). Compared to exo‐T, exo‐CT showed increased infiltration, and exo‐CT‐145 largely retained this enhanced ability. Notably, size and density of spheroids influence exosomal penetration. Despite our efforts to standardize spheroid size and density, subtle differences between groups may have influenced infiltration. Consequently, while the data show exosome penetration at least 50 µm, these results should be viewed with some reservation. Moreover, apoptosis assays showed that 48‐hour exosome treatment increased cell death in KYSE150, KYSE30 and HEEC cells (Figure 6b–e and Figure S9). Compared to the control, exo‐T, exo‐CT and exo‐CT‐145 increased cell death by 5%, 20% and 34% in KYSE150; 6%, 24% and 41% in KYSE30; and 11%, 12% and 21% in HEEC cells, respectively.

FIGURE 6.

Exosome penetration into KYSE150 spheroids and induction of ESCC cell apoptosis. (a) Confocal images showing exosome infiltration into KYSE150 tumour spheroids at different depths (scale bar = 50 µm). (b) Representative flow‐cytometry dot plots of Annexin V–FITC/PI–stained KYSE150 cells after treatment with Ctrl., exo‐T, exo‐CT or exo‐CT‐145. The x‐axis shows Annexin V–FITC fluorescence intensity (phosphatidylserine externalization, early apoptosis), and the y‐axis shows PI fluorescence intensity (loss of plasma membrane integrity, late apoptosis/necrosis). The percentages of cells in each quadrant are indicated: viable (Annexin V⁻/PI⁻, lower left), early apoptotic (Annexin V⁺/PI⁻, lower right), late apoptotic/secondary necrotic (Annexin V⁺/PI⁺, upper right), and necrotic/damaged cells (Annexin V⁻/PI⁺, upper left). (c) Quantification of total apoptosis (early + late apoptotic cells) in KYSE150 cells (one‐way ANOVA, n = 3 independent experiments; data are shown as mean ± SD; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001). (d) Representative flow‐cytometry dot plots of Annexin V–FITC/PI–stained KYSE30 cells after treatment with Ctrl., exo‐T, exo‐CT or exo‐CT‐145. (e) Quantification of total apoptosis in KYSE30 cells (one‐way ANOVA, n = 3 independent experiments; data are shown as mean ± SD; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001).

3.6. Exosomes Inhibit Angiogenesis, Activate Caspase‐3 Apoptotic Pathway and Reverse EMT Process

The anti‐angiogenic effects of exosomes were evaluated using human umbilical vein endothelial cells (HUVECs). exo‐CT and exo‐CT‐145 exhibited a more potent anti‐angiogenic effect compared to exo‐T, with exo‐CT‐145 demonstrating the strongest inhibition (Figure 7a). To understand the mechanisms underlying the therapeutic effects of these exosomes on ESCC, we examined the expression of key apoptosis and EMT markers, including cleaved caspase‐3, caspase‐9, bax, bcl‐2, E‐cadherin and N‐cadherin. Our findings demonstrated that exo‐CT and exo‐CT‐145 induce apoptosis by activating the caspase‐3 and caspase‐9 pathways, accompanied by an upregulation of the pro‐apoptotic protein bax and a downregulation of the anti‐apoptotic protein bcl‐2. Exosomal GzmB activate both caspase‐3 and caspase‐9, triggering apoptosis. The activation of caspases further regulates bax and bcl‐2. Moreover, miR‐145 targets the 3′‐UTR region of N‐cadherin, leading to its inhibition and subsequent reversal of the EMT in ESCC cells, as evidenced by decreased N‐cadherin levels and increased E‐cadherin levels (Figure 7b,c).

FIGURE 7.

Exosomes inhibit angiogenesis, activate the caspase‐3–dependent apoptotic pathway, and reverse the EMT process. (a) Representative images (left) and quantitative analysis (right) of HUVEC tube formation after treatment with Ctrl., exo‐T, exo‐CT or exo‐CT‐145 (one‐way ANOVA, n = 3 independent experiments; data are presented as mean ± SD; ns, not significant; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001). (b and c) Representative Western blots (upper panels) and densitometric analysis (lower panels) of apoptosis‐ and EMT‐related proteins in KYSE150 (b) and KYSE30 (c) cells treated with Ctrl., exo‐T, exo‐CT or exo‐CT‐145. For each lane, 20 µg total protein was loaded. Band intensities were quantified using ImageJ, background‐subtracted and normalized to GAPDH. Relative expression levels are expressed as fold change versus the Ctrl. group, which was set to 1.0 (two‐way ANOVA, n = 3 independent experiments; data are presented as mean ± SD; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001).

3.7. Treatment Effect In Vivo

DiR‐labelled exosomes were injected intravenously into tumour‐bearing mice to investigate their biodistribution and tumour targeting at various time points. Compared to exo‐T, more exo‐CT and exo‐CT‐145 can circulate in mice in 12‐h period, and a significant difference in quantified fluorescence signal intensity can be observed after 1 h (Figure 8a). Fluorescence signals increased over time in the exo‐CT and exo‐CT‐145 groups, indicating effective tumour cell binding. In contrast, exo‐T showed limited tumour tropism, with fluorescence primarily concentrated in the liver. At 12‐hour, ex vivo imaging of organs and tumours confirmed greater accumulation of exo‐CT and exo‐CT‐145 in tumour tissue, with signal intensity approximately six‐fold higher than exo‐T (Figure 8b). Next, we investigated their therapeutic efficacy against tumours in vivo over a 21‐day periods (Figure 8c). Tumour volumes in mice treated with PBS (average 1078 mm³) and exo‐T (average 854 mm³) rapidly increased, showing no significant difference between the two groups (Figure 8d,e). Conversely, exo‐CT and exo‐CT‐145 significantly restrained tumour growth, decreasing the volume to 427 and 205 mm³, likely due to their active targeting effect. At the end of treatment, the average tumour weights in the four groups were 0.934, 0.814, 0.57 and 0.34 g, respectively, corroborating the tumour volume data (Figure 8d,e). Tumour growth curves show that exo‐CT‐145 treatment significantly restricted tumour volume increase compared to other groups over the 21‐day treatment period (Figure 8f). Mouse body weight remained stable throughout the 3‐week administration (Figure 8g). Over the 65‐day observation period, the exo‐CT‐145 group exhibited the longest median survival time of 59 days (Figure 8h). Overall, the in vivo therapeutic efficacy of exo‐CT‐145 was further improved compared to exo‐CT by ∼1.5‐fold. After a continuous 21‐day drug administration process, in order to deeply investigate the possible pathological changes in the main organs of the experimental animals during the drug's action in the body, we carefully selected key organs such as the heart, liver, spleen, lungs and kidneys as the research objects. Subsequently, we used haematoxylin‐eosin (HE) staining technology to stain the tissue of these organs, and then conducted detailed observation and analysis of their tissue morphology and structure under an optical microscope, thereby comprehensively evaluating the degree of the drug's impact on the main organs. Histological analysis demonstrated the absence of significant inflammation or tissue damage, indicating the high biocompatibility of the exosomes (Figure S10). Furthermore, we focused on detecting liver and kidney function‐related indicators, including ALT, AST, CRE, BUN and WBC. The detection was carried out using standard instruments and followed standardized procedures. The results showed that all the indicator values were within the normal reference ranges, and there was no significant difference compared with the control group (p > 0.05), indicating that the experiment did not cause obvious damage to the liver and kidney functions of the mice (Table S2). Moreover, Furthermore, during the 21‐day continuous administration period, in order to deeply and accurately observe the impact of exosomes on the cytokine levels in the mice's bodies, we conducted a rigorous screening and focused on selecting three key indicators closely related to cytokine storms—IL‐2, IL‐8 and TNF‐α. We used a highly sensitive enzyme‐linked immunosorbent assay (ELISA) technique for continuous and dynamic observation. During the observation period, we strictly followed the experimental operation procedures and regularly collected mouse serum samples to ensure the accuracy and reliability of the data. The results showed that the serum levels of IL‐2, IL‐8 and TNF‐α did not show any elevation indicative of a cytokine storm. Specifically, at each time point during the entire 21‐day administration period, the concentrations of IL‐2, IL‐8 and TNF‐α in the mouse serum remained within a relatively stable low‐level range, and there was no statistically significant difference compared to the control group (p > 0.05), indicating that exosomes did not cause abnormal elevation of cytokine levels in the mice's bodies under this administration condition and did not induce cytokine storms (Figure S11).

FIGURE 8.

In vivo therapeutic effects of engineered exosomes. (a) Representative whole‐body fluorescence images (left) and corresponding quantitative analysis of tumour fluorescence intensity (right) in mice treated with exo‐T, exo‐CT or exo‐CT‐145 at the indicated time points within 12 h after injection (two‐way ANOVA, n = 3; data are shown as mean ± SD; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001). Images show one representative mouse per group at each time point; bar graphs show integrated tumour ROI fluorescence (mean ± SD, n = 3). The upper focus marks the flank tumour, and the lower focus represents liver/spleen uptake (see panel b). (b) Ex vivo fluorescence images (left) and corresponding quantitative analysis (right) of fluorescence intensity in tumours and major organs from mice treated with exo‐T, exo‐CT or exo‐CT‐145 (two‐way ANOVA, n = 3; mean ± SD; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001). (c) The timeline of drug administration. (d) Tumour obtained from each group over a 21‐day treatment period. (e) The weight of the tumour in Figure (d) (ANOVA, n = 5, data are shown as mean ± SD, *p < 0.05, **p < 0.01, ***p < 0.0001); (f) The volume changes of tumours in each group within 21 days (ANOVA, n = 3, data are shown as mean ± SD, *p < 0.05, **p < 0.01, ***p < 0.0001); (g) Body weight change curve of mice over a 21‐day treatment period (n = 3); (h) Survivorship curve.

3.8. Mechanism of Exosomes Inhibit ESCC

Post‐treatment analysis revealed a significant reduction in the expression levels of MMP9, CD31 and VEGF within the tumour tissues (Figure 9a). Given their association with angiogenesis, the observed reduction in expression suggests that the exosomes may have an inhibitory effect on blood vessel formation. Moreover, assessment of cleaved caspase 3, E‐cadherin, N‐cadherin, bax and bcl‐2 revealed altered EMT‐related protein expression following exosome treatment (Figure 9b), suggesting exosomes suppressed EMT and activated the caspase 3 apoptosis pathway in vivo. Notably, compared to control, significant differences were detected between exo‐CT and exo‐CT‐145, whereas no difference was observed for exo‐T. Overall, the in vivo results were consistent with the in vitro observations.

FIGURE 9.

Inhibitory mechanisms of exosomes in vivo. (a) Immunostaining of CD31, VEGF and MMP9 in tumour tissues and their quantitative data (Scale bar = 50 µm, ANOVA, n = 3, data presented as mean ± SD, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001). (b) Western blot analysis of MMP9, E‐cadherin, N‐cadherin, Bax, Bcl‐2, Cleaved caspase 3 protein in tumour tissues and their quantitative data (two‐way ANOVA, n = 3, data presented as mean ± SD, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001).

4. Discussion

CAR‐T cell therapy, while offering rapid and durable therapeutic benefits, is associated with acute toxicities, moreover, the efficacy of CAR‐T cells in solid tumours can be severely hampered by the immunosuppressive tumour microenvironment. CAR‐T cell in vivo expansion can trigger excessive cytokine release, leading to uncontrolled effects and adverse reactions like CRS and on‐target off‐tumour response (Zhang et al. 2022; Schubert et al. 2021). In contrast, exosome‐based, acellular effectors may mitigate some product‐driven toxicities and may be less susceptible to certain checkpoint‐mediated inhibitory pathways (e.g., PD‐1/PD‐L1) reported for CAR‐T cells (Hu et al. 2024). Exosomes derived from cell types such as dendritic cells, natural killer cells, T cells and CAR‐T cells have shown significant antitumour efficacy, exhibiting enhanced tumour binding and antitumour effects (Jung et al. 2024). Moreover, the nanoscale size of exosomes allows them to effectively penetrate the dense extracellular matrix and fibroblast barriers characteristic of solid tumours (Xu et al. 2020; Zhang et al. 2023).

ELISA results showed that CAR‐T cell‐derived exosomes contain high levels of granzyme and perforin. Granzymes induce apoptosis through multiple mechanisms. Specifically, granzyme A (GrA) disrupts mitochondrial function, leading to reactive oxygen species release, nuclear translocation of the SET complex, and DNA damage through the release of nucleases like NM23‐H1 (Martínez‐Lostao et al. 2015; Prager and Watzl 2019; Voskoboinik et al. 2015). GrA also impairs DNA repair by cleaving HMGB2 and Ape1 (Prager and Watzl 2019). On the other hand, GzmB activates the caspase cascade, targeting caspase‐3 and caspase‐7 via mitochondria, resulting in DNA fragmentation and apoptosis (Belizário et al. 2018). In caspase‐deficient cells, GzmB can activate the intrinsic mitochondrial apoptosis pathway by cleaving bid and activating bax/bak, ultimately leading to apoptotic body formation (Catalán et al. 2015). In this study, CAR‐T exosome treatment of ESCC cells resulted in increased cleaved caspase‐3 and caspase‐9 expression, decreased anti‐apoptotic bcl‐2, and increased pro‐apoptotic bax, suggesting that CAR‐T exosomes induce apoptosis via the caspase‐3 pathway.

Exosomes offer a promising platform for delivering therapeutic drugs (Abhange et al. 2021; Wang et al. 2023), miRNA or siRNA, due to their ability to target both superficial and deeply located tumour cells. Their phospholipid bilayer enhances drug delivery efficiency, facilitates tissue penetration and extends the circulation time of cargo (Zhang et al. 2023; Elsharkasy et al. 2020; Chen et al. 2022; Quinn et al. 2021; Wan et al. 2022). Our results demonstrated that miR‐145 directly targets the 3′UTR of the N‐cadherin gene, leading to a reduction in N‐cadherin protein levels. N‐cadherin, frequently upregulated in tumours, is known to promote tumour invasion, induce EMT and enhance cell migration and invasiveness (Cao et al. 2019; Yu et al. 2019). The observed miR‐145‐mediated downregulation of N‐cadherin resulted in the reversal of EMT (Mo et al. 2017). Angiogenesis, the formation of new blood vessels, is critical for tumour growth and progression (Zuazo‐Gaztelu and Casanovas 2018), especially after tumours reach a size of 1–2 mm, requiring adequate oxygen and nutrient supply (He et al. 2019). In subcutaneous ESCC tumour models using nude mice, CAR‐T exosomes effectively targeted tumour tissues, significantly suppressed tumour progression, and reduced tumour vascularization, as indicated by decreased expression of CD31, VEGF and MMP9 proteins.

Nucleic acid small molecule drugs present several advantages, including high specificity for target genes, reduced resistance development, a streamlined development process, and a favourable research and development success rate (Zhu et al. 2022; Childs‐Disney et al. 2022; Sani et al. 2024). miRNA, with its long half‐life, holds promising therapeutic potential, the key to the success of small nucleic acid drugs currently lies in delivery technology. A promising therapeutic strategy involves combining the advantages of small nucleic acid drugs with the natural targeting and cytotoxicity of CAR‐T exosomes. Exosomal genetics suggest that exosomes derived from multi‐targeted CAR‐T cells may exhibit multiple targets on their surface. Therefore, modified CAR‐T cell‐derived exosomes not only enhance effectiveness in patient treatment but also improve safety. These novel ‘CAR‐T’ products are expected to possess greater potency in the exosomes they release, offering a balance between safety and efficacy through uniquely designed cars that address the limitations of CAR‐T cell therapies (Herrmann et al. 2021).

Lastly, beyond cancer immunotherapy, the similar engineering strategies used for CAR‐T exosomes can be extended to exosomes derived from other immune cells and stem cells (Chen et al. 2024; Zhu et al. 2023). These exosomes could potentially be used to suppress unwanted immune responses. This versatility underscores the exciting possibilities of exosomes in modulating the immune system, offering the potential to both stimulate and dampen immune activity as needed.

Although acknowledging the positive findings of this study, it is crucial to delineate the scope of its safety assessment. Since the CAR construct in this study was designed to recognize human B7‐H3, and its binding to murine homologs was not tested, the standard mouse xenograft model cannot accurately predict the risk of on‐target, off‐tumour toxicity in humans that may arise from B7‐H3 expression in normal tissues. Consequently, all references to ‘improved safety’ in this work are strictly limited to the advantage of exo‐CT/145, compared to live CAR‐T cells, in mitigating product‐driven (rather than target‐dependent) systemic CRS‐like inflammatory responses. Thus, a more precise conclusion is that, under the experimental conditions used here, exo‐CT/145 demonstrates a favourable profile in attenuating treatment‐related hyperinflammation. To fully define its clinical safety, future studies must incorporate systematic testing of the CAR's species cross‐reactivity and utilize more advanced models that better recapitulate the human immune microenvironment and target expression patterns.

In this study, it should be noted that using athymic nude mice was a no‐alternative choice to avoid immune rejection of immunocompetent BALB/c or C57BL/6 mice to human‐derived ESCC cells and assess the direct anti‐tumour effect of CAR‐T cell derived exosomes (exo‐CT‐145). It is well known that immunocompetent models like BALB/c or C57BL/6 mice better mimic human immunity, which are key for studying tumour‐exosome‐immune system interactions. The current model gives insights into exo‐CT‐145's anti‐tumour activity, but future use of immunocompetent systems will deepen understanding of interactions and aid in developing better, translatable cancer therapies.

5. Conclusion

This study demonstrates that CAR‐T cell‐derived exosomes, enriched with cytotoxic molecules and surface CARs, effectively target and kill solid tumours both in vitro and in vivo. These exosomes, loaded with miR‐145, inhibited tumour growth, angiogenesis, and EMT progression in ESCC cells, while also proving safe and avoiding immunosuppressive TME and CRS. This research highlights CAR‐T exosomes as a promising, safe, and effective nanotherapeutic strategy for cancer treatment, offering a potential alternative to traditional CAR‐T cell therapy.

Author Contributions

Ruyue Yang: data curation, formal analysis, writing – original draft. Haitao Wang: methodology, writing – review and editing. Guidan Wang: software. Gankun Yuan: data curation. Zhaoyi Wei: data curation. Junling An: writing – review and editing. Dong Hu: software. Wenjing Wen: data curation. Shegan Gao: investigation, project administration. Yuan Wan: formal analysis, writing – review and editing. Gaofeng Liang: conceptualization, formal analysis, supervision, funding acquisition, writing – original draft, writing – review and editing.

Conflicts of Interest

The authors declare no conflict of interest, financial or otherwise in this paper.

Supporting information

Supplementary Materials: jev270245‐sup‐0001‐SuppMat.docx

Data Availability Statement

The data that support the findings of this study are available in the supplementary material of this article.