Recruiting T-cells toward the brain for enhanced glioblastoma chemo-immunotherapy efficacy by co-delivery of cytokines and temozolomide via ultrasound-gated redox-responsive extracellular vesicles

Abstract

The main limitation of chemo-immunotherapy in glioblastoma (GBM) is the immunosuppressive tumor microenvironment (TME) and the restricted permeability of the blood-brain barrier (BBB). Here, we engineer redox-responsive macrophage-derived extracellular vesicles (M-EVs)-based nanovesicles (TC@MEVs) co-loaded with chemokine CXC chemokine ligand 10 (CXCL10) and temozolomide (TMZ). Combined with ultrasound (US)-mediated BBB opening, TC@MEVs release CXCL10 to recruit CD8⁺ T cells to the GBM region, synergizing with the high concentration of TMZ to amplify the chemo-immunotherapy efficacy of GBM. Consequently, up to 5.52-fold increase in CD8⁺ T cells are observed with US-guided co-delivery of CXCL10 and TMZ, compared to free TMZ and CXCL10. This spatiotemporal combination strategy enhances chemo-immunotherapy by reducing Tregs by 46%, increasing the M1/M2 macrophage ratio by 10.05-fold, achieving 40% tumor elimination, prolonging survival, and establishing long-term immune memory in orthotopic GBM mice. Overall, US-mediated the redox-responsive M-EVs nanovesicles to reverse the immunosuppressive TME by improving the infiltration of CD8⁺ T cells and local release of TMZ, may present a promising strategy for effective GBM chemo-immunotherapy.

Graphical Abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s12951-025-03885-y.

Introduction

Glioblastoma (GBM) is a highly aggressive and treatment-resistant brain cancer, with a median survival of less than 15 months despite multimodal therapies [1, 2]. While chemo-immunotherapy combining temozolomide (TMZ) with immune checkpoint inhibitors has shown potential, its efficacy is fundamentally constrained by the immunosuppressive tumor microenvironment (TME) characterized by sparse CD8⁺ T-cell infiltration and immunosuppressive regulatory T-cell (Treg) populations of GBM [2–4]. The use of chemokines to recruit T cells and reprogramming TME offers promise for enhancing the efficacy of chemo-immunotherapy in GBM [5, 6]. However, this approach faces challenges such as rapid chemokine degradation in circulation and the lack of effective delivery strategies [7, 8].

Despite the development of various nanocarriers, the blood-brain barrier (BBB) limits drug delivery, rendering over 95% of immunomodulators ineffective at reaching tumor sites [7, 8]. Current nanoparticle delivery systems, despite their potential, often struggle to overcome key challenges, including effective BBB penetration without off-target effects, tumor-specific payload release synchronized with immune activation, and the reversal of TMZ-induced lymphopenia [4, 7, 9]. For instance, conventional TMZ formulations inadvertently deplete circulating lymphocytes, compromising immune function and reducing treatment efficacy [9]. These limitations highlight the need for advanced delivery systems that can overcome these obstacles to enhance therapeutic outcomes [4, 9, 10].

Low-intensity focused ultrasound (LIFU) with microbubbles (MBs) is a promising non-invasive method for transiently opening the BBB, improving drug delivery to the brain [11–13]. MBs amplify the ultrasonic energy of LIFU to temporarily disrupt the tight junctions of endothelial cells in the BBB, allowing therapeutic agents to pass through [12, 14]. By precisely controlling parameters such as intensity, frequency, and exposure time, LIFU/MBs ensures safe, reversible BBB opening without causing persistent tissue damage [15]. Once the LIFU is stopped, the BBB restores its normal function within hours, making the US/MBs technology a promising approach for treating conditions like GBM [12, 14, 16].

Macrophage-derived extracellular vesicles (M-EVs) are promising delivery vehicles for targeted GBM therapy [17–19]. Chemokines in M-EVs (such as CCL2 and CX3CL1) enable them to target chemokine receptors highly expressed in GBM (such as CCR2 and CX3CR1) [19–22]. Furthermore, this study demonstrates that M-EVs carry “self-marking” proteins such as cluster of differentiation 47 (CD47), endowing them with the potential to evade immune clearance and prolong their circulation time within the body [23, 24].

Based on these findings, a spatiotemporal combination strategy was developed, integrating US-gated BBB opening with a glutathione (GSH)-responsive M-EVs delivery system to enable effective GBM chemo-immunotherapy. M-EVs were engineered into bionic membrane nanovesicles (TC@MEVs) loaded with chemokine CXC chemokine ligand 10 (CXCL10) and TMZ. US/MBs safely and reversibly opened the BBB for a 4-hours window, allowing TC@MEVs to reach the GBM region. In the presence of high levels of GSH [25, 26], CXCL10 was released from TC@MEVs to recruit CD8⁺ T cells, while TMZ was delivered efficiently, killing tumor cells and releasing immune-activating proteins like high mobility group box 1 (HMGB1) [27, 28] to further enhance the local anti-tumor immune response. This strategy enhances CD8⁺ T-cell infiltration and reverses the immunosuppressive TME, improving chemo-immunotherapy in orthotopic GBM mice (Fig. 1).

Fig. 1.

Schematic diagram of spatiotemporal synergistic chemotherapy-immunotherapy for glioblastoma (GBM) (By Figdraw). (A) Schematic diagram of TC@MEVs synthesis. (B) Ultrasound/microbubbles (US/MBs) open the blood-brain barrier (BBB) combined with intravenous injection of TC@MEVs via the tail vein. (C) US/MBs open the BBB, enabling TC@MEVs to accumulate in the GBM region, where glutathione triggers the release of CXCL10 and temozolomide (TMZ) into the GBM area. (D) Schematic representation of the reversal of the immunosuppressive tumor microenvironment in GBM

This spatiotemporal combination strategy integrates US-triggered BBB opening with redox-responsive TC@MEVs to achieve spatiotemporally controlled delivery, may present a promising strategy for effective GBM chemo-immunotherapy.

Materials and methods

Materials

TMZ were purchased from Med Chem Express. CXCR3 antagonist AMG487 was obtained from Tocris Bioscience (HY-17364, Bristol, United Kingdom). Recombinant Mouse CXCL10/IP-10 Protein is produced by Pichia expression system (RP01625, abclonal, China). Recombinant Mouse GM-CSF was purchased from Novoprotein, (C-6His, Shanghai, China). A specialized medium for CTLL-2 cells was purchased from Prosperity Life Sciences (CM-0331, Wuhan, China). Sparkjade ECL star was purchased from Shandong Sparkjade Biotechnology Co., Ltd (ED0025-A, Sparkjade, Shandong, China). L-7102 Protein Loading Buffer (5X) was purchased from shanghai Biolinkedin biotechnology Co., Ltd (L-7102, Shanghai, China). Tumor necrosis factor-α (TNF-α) Elisa mouse kit (JM-02415M1, Nanjing, China) was purchased from Jiangsu Jingmei Biological Technology Co., Ltd. Interferon-γ (IFN-γ) Elisa mouse kit (JL10967, Shanghai, China) and Interleukin 10 (IL-10) Elisa mouse kit (JL20242, Shanghai, China) were purchased from Jianglai biology Co., Ltd. Interleukin 15 (IL-15) Elisa mouse kit (YM-2908A1, Nanjing, China) and High mobility group box 1 (HMGB1) (YM-66137A1, Nanjing, China) Elisa mouse kit were purchased from Nanjing Youmeng Biotechnology Co., Ltd. GAPDH mouse monoclonal antibody was purchased from Shanghai Epizyme Bio-Pharmaceutical Technology Co., Ltd (LF205S, Shanghai, China). Super sensitive ECL Detection Kit was purchased from Guangzhou Biolight Biotechnology Co., Ltd (ESL003, Guangzhou, China). Subsequently, mIHC multiplex fluorescence immunohistochemistry kit (WAS1504, WASci, China) was used for staining. Fetal bovine serum (FBS) was purchased from Wisent Biotechnology (086–150, Nanjing, China) Co., Ltd. DSPE-PEG-2000 was purchased from Tanshtech (80010201-2000, Guangzhou, China). 1,2-Distearoyl-sn-glycero-3-phosphorylcholine (DSPC) was purchased from AVT Pharmaceutical Tech Co., Ltd. (S01005, Shanghai, China). CCK-8 kit was purchased from Shandong Sparkjade Bio-technology Co (CT0001-A, Shanghai, China). High glucose medium (BC-M-005-500mL DMEM), 1640 medium (BC-1640 M-005-500mL DMEM), 1% penicillin and streptomycin, Fetal bovine serum (BC-SE-FBS01) and streptomycin, and 0.25% trypsin-ethylenediaminetetraacetic acid (BC-CE-005, EDTA) were purchased from BioChannel Biological Technology Co., Ltd. (Nanjing, China). GelNest™ matrix gel was purchased from NEST Biotechnology (211212, Wuxi, China). Lent-EF1a-P2A-luciferase-CMV-coGFP-P2A-Puro lentivirus and polybrene were provided by Banma Biotechnology Co., Ltd. (Changsha, China). Cy5 NHS was purchased from Xi’an Rui Xi Biotechnology Co., Ltd. (146368-14−1, Xi’an Rui Xi, China). IHC staining was offered from Nanjing Youmeng Biotechnology Co., Ltd. Immunohistochemistry and terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) apoptosis detection experiments were performed by Nanjing Youmeng Biotechnology Co. Red blood cell lysate was purchased from ZUNYAN Biological Technology Co., Ltd (ZYFB006-0100, Nanjing, China). 0.1 ml qPCR 96-well plate was purchased from ACCURATE BIOTECHNOLOGY (HUNAN) CO.,LTD, (AG12102, ChangSha, China). Albumin Content Assay Kit (Bromocresol Green Colorimetry) was purchased from Beijing Boxbio Science & Technology Co., Ltd (AKPR040C, Beijing, China). Anti-CD47 antibody was purchased from Shanghai Taoshu Bioscience Co. (T9901A-636, TargetMol, USA). The Reduced GSH Enzyme-Linked Immunosorbent Assay (ELISA) kit (YQ-2794) was obtained from Shanghai Research and Development Biotechnology Co. Live/dead dye, anti-CD3, anti-CD4, anti-CD8a, anti-IFN-γ+, anti-PD1+, anti-CD44, anti-CD62L, anti-CD11c, anti-CD80, anti-CD86, anti-CD11b, anti-CD206 and anti-Foxp3 antibodies (103112, 100203, 100537, 100707, 100537, 100707, 100203, 100537, 100707, 100203, 100537, 126403, Biolegend, USA) for flow cytometry analysis were purchased from Dakewe Biotech Co. Anti-ZO-1 (AB10733242), anti-ALIX (AB2162467), anti-CD9 (AB2783831) and anti-TSG101 (AB2881104) antibodies were provided by proteintech (Chicago, USA). Active ester-disulfide bond-active ester (NHS-S-S-NHS) was purchased from Xi’an Rui Xi Biotechnology Co., Ltd. (1688598-83−5, Xian, China). Analytical grades of all other chemicals were purchased from Sigma-Aldrich (Sigma-Aldrich, USA). The ultrasonic signal generator (AFG3021C) and RF power amplifier (ENI2100L) were purchased from Shanghai Tektronix Technology Co. AMG487 (CXCR3 antagonist) was kindly provided by Amgen (South San Francisco, CA, USA).

Cell culture

Mouse T-lymphocyte cell line CTLL-2, mouse dendritic cell line Raw 264.7, mouse brain endothelial cell line bEnd.3, and mouse glioma cells GL261 were obtained from the Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences. The cell lines tested negative for mycoplasma contamination and were regularly treated with mycoplasma removing agent (VivaCell, Shanghai, China). The bEnd.3, GL261 and GL261-luc cells were cultured and grown in high-sugar DMEM with 10% special grade FBS, Raw 264.7 cells were cultured and grown in RPMI-1640 medium with 10% special grade FBS, and the CTLL-2 cells were raised in CTLL-2-specific medium and incubated at 37 °C in air, which containing 5% CO₂. These cells were suspended in serum-free cell cryopreservation solution (CELLSAVING, cat: C40100, New Cell&Molecular Biotech, China).

Animals

Female C57BL/6J mice (6–8 weeks, 20–22 g) were purchased from Hangzhou Qizhen Laboratory Animal Co., Ltd. Before performing experiments, all mice were maintained in a 12-h light-dark cycle with access to food and water. All animal experiments were performed in compliance with the relevant laws and approved by the Institutional Animal Care and Use Committee of Southeast University School of Medicine (NO.SEU-IACUC-20250219004).

Extraction of M-EVs

Initially, Raw 264.7 cells were cultured until they reached the logarithmic growth phase, and the culture supernatant was collected. M-EVs were precipitated using ultra-high-speed centrifugation (100,000 * g for 1–2 h). After discarding the supernatant, PBS was added to resuspend the vesicles, which were then centrifuged again for cleaning. The solution was filtered through a 0.2 μm PTFE membrane (Sigma-Aldrich) to remove larger vesicles such as microvesicles and apoptotic bodies. The extracted vesicles were observed under a transmission electron microscope to assess their morphology, and their size distribution was determined using a nanoparticle size analyzer. Subsequently, lysis buffer was utilized to extract the membrane components of M-EVs, followed by protein quantification. Western blot analysis was performed to detect M-EVs-specific markers such as ALIX, CD9, CD47, and TSG101 [18]. Previous studies have demonstrated that ALIX, CD63, and TSG101 exhibit elevated expression levels in macrophage membranes, whereas CD9 shows reduced expression levels in M-EV membranes [18].

Construction and characterization of TC@MEVs

TMZ and M-EVs were mixed in the ratio of 10%, 15%, 20% or 30% (m/m) and sonicated using Model 505 Sonic Dismembrator with 0.25’’ tip with the following settings: 20% amplitude, 6 cycles of 30 s on/off for 3 min with a 2 min cooling period between each cycle [29]. TMZ was combined with the extracted membrane components of M-EVs and repeatedly extruded 10 times through a 100 nm pore-sized membrane extruder to form T@MEVs. In order to make the M-EVs membrane recovery, the mixture was co-cultured at 37 °C for 1 h. Then un-encapsulated TMZ was removed by centrifugation at 1000 rpm for 10 min, and the T@MEVs pellet was centrifuged at 120,000 g. Subsequently, the CXCL10 was conjugated to the surface of T@MEVs using an active ester disulfide bond crosslinker, along with PEG2000 at a molar ratio of 1:2:1(CXCL10:crosslinker: PEG2000). The conjugation reaction was carried out at pH 7.4 for 2 h, followed by dialysis using a 10 kDa dialysis bag for 24 h to remove small molecules. This dialysis process was repeated three times. The ratios of TMZ and CXCL10 were optimized by adjusting their input amounts. Then, the quantitative fluorescence methods were used to determine the encapsulation efficiency and drug loading capacity of TC@MEVs. Amount of loaded drug: Amount of TMZ or CXCL10 actually encapsulated in TC@MEVs. Amount of drug loaded: Total input amount of TMZ or CXCL10. Amount of drug loaded nano vesicles: Total mass of TC@MEVs.

The Malvern particle size analyzer was utilized to determine the particle size and zeta potential distribution, while transmission electron microscopy (TEM) was employed to observe the morphology and dispersion of TC@MEVs. Each group contains 3 samples. Blinding and randomization treatment groups were allocated in a randomized fashion. Furthermore, equivalent amounts of TC@MEVs were dispersed in PBS (pH 7.4), DMEM, or 10% FBS solutions, and their quantity, particle size, and zeta potential were continuously assessed over a period of 7 days to evaluate the stability of TC@MEVs.

Pharmacokinetic evaluation

Equal amounts of TMZ or CXCL10 loaded in TC@MEVs (TMZ = 10 mg/kg, CXCL10 = 75 µg/kg) were injected into the circulation of 6-week-old female C57 mice for pharmacokinetic evaluation. 3 mice per group, blinding and randomization treatment groups were allocated in a randomized fashion. For easy tracking and detection, TMZ and CXCL10 were fluorescently labeled with Cy5 NHS, facilitating the tracing of its distribution and kinetic process in the bloodstream.

The experimental procedure for labeling TMZ and CXCL10 with Cy5 NHS is as follows: Dissolve Cy5 NHS in DMSO at a concentration of 1.0 mg/mL. TMZ (dissolved in DMSO at 10 mM) and CXCL10 (dissolved in PBS at 1 mg/mL) were mixed with Cy5 NHS at molar ratios of 10:1 (TMZ) or 20:1 (CXCL10) and reacted at room temperature for 1–2 h. Upon reaction completion, add 10 mM glycine to terminate the reaction and remove unreacted Cy5 NHS using a 10 kDa centrifugal filter tube. Finally, assess labeling efficacy by detecting fluorescence at 650 nm wavelength using a fluorometer to confirm successful Cy5 labeling.

In the experiment, blood samples were collected from the mice at 0, 1, 2, 4, 8, 12, and 24 h after injection. Quantitative analysis was performed using a quantitative fluorometric assay to detect changes in blood concentration of TMZ or CXCL10 at each time point. This series of blood concentration data at different time points was used to evaluate the release behavior, absorption rate, and pharmacokinetic characteristics of TMZ and CXCL10 in the TC@MEVs drug delivery system.

TC@MEVs release CXCL10 and TMZ in response to GSH

On the 10th day after establishing the GBM mouse model, extract tumor tissue or brain tissue, weigh it, homogenize it, and place 1 g per sample in 1 mL of double-distilled water. The reduced GSH ELISA kit (YQ-2794, Shanghai Yanqi Biotechnology Co., Ltd.) was used to quantitatively detect GSH levels, assessing the difference in GSH content between tumor tissue and normal tissue. To further investigate the response of TC@MEVs in a GSH environment, CXCL10 was labeled with Cy5 NHS and subsequently added to equal volumes of solutions containing GSH concentrations of 4.5 µmol/g or 1.5 µmol/g, respectively, to simulate brain tumor or normal brain tissue. The release of CXCL10 was monitored in real-time at different time points (0, 1, 2, 3, 4, 5, 6, and 8 h) using an in vivo imaging system (IVIS).

To further compare the effects of GSH and pH on drug release from TC@MEVs, the pH of the GBM tumor tissue suspension was measured using a pH meter and determined to be 6.5. An equal amount of TC@MEVs suspension was divided into four groups and subjected to the following conditions: pH (7.4) + GSH (1.5 µmol/g), pH (7.4) + GSH (4.5 µmol/g), pH (6.5) + GSH (1.5 µmol/g), and pH (6.5) + GSH (4.5 µmol/g). Each group contains 5 samples. Blinding and randomization treatment groups were allocated in a randomized fashion. CXCL10 was labeled with Cy5, and a multifunctional microplate reader (DLK0001622, BioTek, USA) was used for quantitative analysis. The release ratio of CXCL10-Cy5 was measured at different time points (0, 1, 2, 3, 4, 5, 6, and 8 h).

The half-maximal inhibitory concentration (IC₅₀)

To investigate the IC₅₀ of TMZ on the GL261 mouse glioma cell line, cells were seeded in a 96-well plate and treated with a concentration gradient of TMZ ranging from 1 to 10 mM after 24 h. The treatment was maintained for 48 h. Following drug exposure, cell proliferation was assessed using the CCK-8 assay kit, and the OD value of each well was measured using a microplate reader. A drug concentration-cell survival curve was plotted based on the OD values obtained from different concentrations of TMZ treatments. The IC₅₀ value of TMZ was then calculated using nonlinear regression analysis.

Furthermore, we evaluated the IC₅₀ values of CXCL10 and TMZ/CXCL10 loaded in TC@MEVs. Each group contains 5 samples. Blinding and randomization treatment groups were allocated in a randomized fashion. Briefly, lymphocytes were isolated from the spleens of tumor-free C57BL/6J mice using a lymphocyte separation medium (P9000, Solarbio, Beijing, China). Then, 2*10⁶ lymphocytes were seeded into the upper chamber of a 24-well transwell plate with 4 μm pores (3413, Unicorn Biotechnology Co., Ltd., Beijing, China) and incubated at 37 °C for 2 h. GL261 cells were inoculated into the lower chamber at a density of 2*10⁵/mL.

Different treatment groups were established with CXCL10 concentration gradients of 0, 5, 10, 20, 30, 40, 50, 60, and 100 ng/mL. After 24 h, cell viability of GL261 cells was assessed using a CCK-8 assay kit. Similarly, different treatment groups were set up with TC@MEVs concentration gradients of 0, 50, 100, 150, 200, 250, 300, 500, and 1000 ng/mL. After 24 h, the viability of GL261 cells was measured using the CCK-8 assay kit (C0005, TargetMol, USA).

Construction of phospholipid MBs

Lipids DSPC and DSPE-PEG2K were mixed at a 9:1 molar ratio and dissolved in chloroform to prepare a 20 mg/mL DSPC solution and a 16.5 mg/mL DSPE-PEG2K solution, respectively. 538 µL of DSPC solution and 257 µL of DSPE-PEG2K solution were added to a round-bottom flask, and a rotary evaporator was used to evaporate the chloroform for 2 h until it was completely volatilized, forming a milky white film. Subsequently, 5 mL of sterile PBS, 1 mL of propylene glycol, and 1 mL of glycerol were added to prepare the hydrating solution, which was then poured into the round-bottom flask. The lipid suspension was completely dissolved using water bath ultrasonication and dispensed into sterile 3 mL vials, with 1 mL in each vial. After sealing, the vials were vacuumed, and the air inside was replaced with C₃F₈ gas to obtain the MBs stock solution. The stock solution was stored in a refrigerator at 4℃ and shaken with a shaker before use until the liquid turned from transparent to milky white. A small amount of suspension was dropped onto a glass slide, diluted with ultrapure water, covered with a coverslip, and observed under a bright-field microscope to examine the morphology of the MBs [12–14].

Ultrasonic equipment and parameters

In this study, LIFU was employed to facilitate the BBB opening. The ultrasonic signal generator (AFG3021C) and RF power amplifier (ENI2100L), both sourced from Shanghai Tektronix Technology Co., were utilized to administer the LIFU protocol. To optimize the ultrasound exposure parameters and minimize the risk of tissue damage, the following settings were applied: frequency of 1 MHz, ultrasound intensity range of 0–3 w/cm², a total of 10,000 pulses delivered per treatment cycle, duty cycle of 30%, duration of 60 s, inter-pulse interval of 1 s, pulse waveform, and coupling agent for coupling. The combination of low intensity and controlled pulse delivery was essential for achieving a transient, reversible BBB opening, as confirmed through subsequent safety assessments.

CXCL10-induced T-cell recruitment

To assess the efficiency of CXCL10 release from TC@MEVs in response to GSH, we utilized a 24-well transwell chamber with 4 μm pores (3413, Unicorn Biotechnology Co., Ltd., Beijing, China). Initially, TC@MEVs were placed in the lower chamber of the transwell system, along with the addition of 4.5 µmol/L GSH. GL261 cells were inoculated in the lower chamber at a density of 2*10⁵/mL. Each group contains 5 samples. Blinding and randomization treatment groups were allocated in a randomized fashion. Experimental groups were set up as follows: positive control (CXCL10 group, with a CXCL10 concentration of 40 ng/mL), untreated control, TMZ group (TMZ concentration of 2.5 mM), T@MEVs group, TC@MEVs group, T@MEVs combined with GSH group and TC@MEVs combined with GSH group. Lymphocytes were then isolated from the spleens of tumor-free C57BL/6J mice using lymphocyte separation medium (P9000, Solarbio, Beijing, China), and 2*10⁶ lymphocytes were seeded into the upper chamber of the transwell system, followed by incubation at 37 °C for 2 h. After incubation, lymphocytes that had migrated to the lower chamber were collected. These cells were subsequently stained for 60 min using anti-mouse CD3 (100302, BioLegend), CD8a (100708, BioLegend), and CD4 (100537, BioLegend) antibodies. The proportions of CD3⁺CD4⁺ and CD3⁺CD8⁺ T-cell ratio were analyzed by flow cytometry [13, 22]. Additionally, the CCK-8 assay was employed to evaluate the viability of GL261 cells in the lower chamber of the transwell system [13].

To further evaluate the effects of TMZ on bEnd.3 and T cell activity, a BBB model was constructed as described above. TMZ at concentrations of 0, 0.5, 1.0, 2.0, 2.5, 3.0, 5.0, 10.0, 20.0, and 50.0 mM was added to the lower chamber of the transwell chamber. After 24 h, T cells were collected and cell viability was assessed using the CCK-8 assay. Performed TUNEL apoptosis detection on the BBB model constructed by bEnd.3.

To further determine whether T-cell recruitment was specifically attributable to CXCL10 released from TC@MEVs, we designed a CXCR3 receptor blockade experiment. Briefly, isolated lymphocytes were placed in the upper chamber of a 24-well transwell plate with 4 μm pores (3413, Unicorn Biotechnology Co., Ltd., Beijing, China). The experiment included five groups: Control, CXCL10, CXCL10 + AMG487, TC@MEVs + GSH, and TC@MEVs + GSH + AMG487. Each group contained four replicates. Blinding and randomization of treatment groups were implemented. The concentrations used were as follows: AMG487 (a specific CXCR3 receptor inhibitor) at 1 µmol/mL, CXCL10 at 40 ng/mL, and GSH at 4.5 µmol/mL. After a 2-hour incubation, lymphocytes that had migrated to the lower chamber were collected. These cells were then stained for 60 min using anti-mouse CD3 (100302, BioLegend) and CD8a (100708, BioLegend) antibodies. The proportion of CD3⁺CD8⁺ T cells was analyzed by flow cytometry.

Construction and opening of the BBB model

Approximately 10⁵ bEnd.3 cells were seeded on transwell membranes (pore size: 0.4 μm; diameter: 12 mm) in Corning 24-well plates. The trans-epithelial electrical resistance (TEER) between the upper and lower chambers of the transwell system was measured every other day using an epithelial voltohmmeter. The in vitro BBB model was considered stabilized when the TEER value reached its maximum. TEER was calculated using the following formula: Resistance per unit area (Ω·cm²) = Measured resistance (Ω) * Effective membrane area (cm²).

In this experiment, a 24-well transwell chamber with a 4 μm pore size was utilized. bEnd.3 cells were seeded in the upper chamber at a density of 10⁵/mL for routine cultivation. On the 4th day of seeding, GL261-GFP cells were inoculated in the lower chamber at a density of 2*10⁵/mL. After 24 h, six treatments were applied to the upper chamber: PBS, Free TC (TMZ and CXCL10), M-EVs, TC@MEVs, US, US/MBs and TC@MEVs combined with US/MBs (TMZ = 2.5 mM, CXCL10 = 40 ng/mL). Among these, CXCL10 was labeled with Cy5-NHS at a molar ratio of 10:1 (Cy5-NHS: CXCL10). The MBs suspension was used at a concentration of 2 * 10⁷/mL, with a volume of 5 µL.

After 12 h, the BBB model constructed with bEnd.3 cells was removed for immunostaining. Structural changes in the tight junction protein zonulin-1 (Zo-1) were observed. DAPI was used to label the cell nuclei. Fluorescent signals (CXCL10- Cy5) in the culture medium of the lower chamber were observed using a fluorescence microscope, and was quantified with ImageJ software (ImageJ Software Inc., USA). The apoptosis rate of GL261 cells in the lower chamber was detected by flow cytometry. Additionally, the effect of US on the BBB model was further evaluated by adjusting the US intensity (0, 0.1, 0.3, 0.5, 0.7, 1.0, and 2.0 w/cm²) [13, 14].

To further evaluate the differences in the tropism of EVs derived from tumor-associated macrophages (EVs@TAM) or RAW 264.7 cells (EVs@RAW) toward GBM, a BBB model was established. In the lower chamber of the BBB model, GL261 cells were seeded at a density of 2 * 10⁵/mL. After 24 h, 50 µg of DiD-labeled EVs@TAM or EVs@RAW was added to the upper chamber. Following a 12-hour incubation period, the samples were observed using confocal laser scanning microscopy, with cell nuclei stained using DAPI. Quantitative analysis of fluorescence signals was performed using ImageJ software.

Time window and safety range of BBB opening

To further evaluate the safety of BBB opening induced by US/MBs, particularly the effects of US on the BBB of C57 healthy mice, this experiment involved the injection of 50 µL of MBs suspension at a concentration of 2*10⁷/mL via the tail vein. 4 mice per group, blinding and randomization treatment groups were allocated in a randomized fashion. US was then applied at varying intensities (0, 0.1, 0.3, 0.5, 0.7, 1.0, and 2.0 w/cm²) for a duration of 60 s, with a cyclic setting of 10,000 times, a duty cycle of 5%, and a pulse duration of 1 s [13, 14]. Following this, 10 µL of 2% Evans Blue (EB) dye was injected into the tail vein, and the brain was perfused 2 h later to observe the BBB opening effect [13]. Concurrently, HE staining was employed to assess blood cell extravasation after treatment with different US intensities.

TUNEL immunofluorescence staining was used to detect cell apoptosis in brain tissue following exposure to varying sound intensities. Additionally, immunofluorescence staining was utilized to examine the extravasation of inflammatory cytokines, such as interleukin-6 (IL-6) and interleukin-2 (IL-2), in brain tissue after treatment with different sound intensities, and ImageJ software (ImageJ Software Inc., USA) was used for quantitative analysis of fluorescence signals. Furthermore, C57 mice were treated with three different US intensities (0, 0.3, and 0.5 w/cm²) and their brain tissues were extracted at 0, 2, 4, and 6 h post-treatment to observe the recovery of BBB opening [11, 13].

CXCL10 was labeled with Cy5 and exposed to US treatment for 60 s at ultrasonic intensities ranging from 0 to 3.0 w/cm². 4 mice per group, blinding and randomization treatment groups were allocated in a randomized fashion. Following this, brain tissue homogenates were extracted, and the CXCL10 content was quantified using an enzymatic labeling method. The concentration of Cy5-labeled CXCL10 in brain tissue was quantitatively analyzed using this enzymatic assay.

Utilizing IVIS, the distribution of CXCL10 in the mouse brain was observed at 0, 2, 6, 8, 12, 24, 48, and 72 h after four sets of treatments: PBS, Free TC (a combination of TMZ and CXCL10), TC@MEVs, and TC@MEVs combined with US/MBs (where TMZ = 2.5 mM and CXCL10 = 40 ng/mL). 3 mice per group, blinding and randomization treatment groups were allocated in a randomized fashion. Additionally, the distribution of CXCL10 in tissues such as the heart, liver, spleen, lungs, kidneys, and brain tumors were visualized using IVIS, and quantified through fluorescence signal analysis. Immunohistofluorescence sections were employed to examine the distribution of CXCL10 within tumor tissues, with DAPI used to label cell nuclei. Furthermore, TMZ was labeled with Cy5, allowing for the detection of its distribution in various tissues, including the heart, liver, spleen, lungs, kidneys, and brain tumors, via IVIS.

GBM mice models and magnetic resonance imaging (MRI)

EF1α: luciferase-P2A-Puro, CMV: copGFP, lentivirus at a titer of 7*10⁸ TU/mL was used to transduce GL261 cells at 50% confluence (MOI value of 100, polybrene concentration of 4 µg/mL in the medium), and puromycin at a concentration of 3 µg/mL screened for a GL261-luc cell line that stably expresses GFP and luciferase [13]. A mixture of 4 µL matrix gel/PBS with 1*10⁵ GL261 cells or GL261-luc cells was injected intracranially using a microliter syringe (Hamilton, USA) at coordinates of 0.5 mm anterior to the fontanelle, 1.5 mm to the right, and 2 mm deep. MR imaging was performed using a 7.0 T MR small animal MR scanner (Bruker, Germany) and was confirmed by a T2-weighted imaging (T2WI) and fast spin-echo MR imaging sequence (repetition time/echo time, 2000/50 ms, matrix, 256*256, field of view, 20*20 mm, slice thickness, 1.0 mm) to monitor tumor growth) [13, 22, 30]. We used ImageJ (ImageJ Software Inc., USA) to measure glioma-in-situ volume on T2WI images and to obtain the relative volume of glioma-in-situ versus the whole brain (relative tumor volume = glioma-in-situ volume/volume of the whole brain).

IVIS

Bioluminescence imaging was used to quantify tumor load in mice carrying GL261-luc. Tumor-bearing mice were injected with D-luciferin (150 mg/kg body weight) [13, 30], and images were taken using an animal live imaging system (CRi Maestro™ In-Vivo Imaging System; Cri, Hopkinton, MA). 3 mice per group, blinding and randomization treatment groups were allocated in a randomized fashion. Data were subsequently analyzed using Living Image 2.5 software (Caliper Life Sciences, USA).

Antitumor effect on orthotopic GBM mice

To evaluate the antitumor effect in vivo on orthotopic brain tumors, GL261 cells or GL261-luc cells (1*10⁵ cells) were first inoculated intracranially into mice. 10 mice per group, blinding and randomization treatment groups were allocated in a randomized fashion. The inoculation day was designated as Day 0. On Day 10 post-inoculation, mice were randomly assigned to receive the following treatments via tail vein injection: PBS, TMZ, free TMZ&CXCL10, TC@MEVs, US combined with T@MEVs, and US combined with TC@MEVs (where the TMZ dose was 10 mg/kg and the CXCL10 dose was 75 µg/kg). The treatments were administered three times, with two-day intervals between each treatment. The specific method for US combined with TC@MEVs treatment was as follows: First, 2.5 µL/g MBs suspension (2*10⁷/mL) was injected via the tail vein, followed by US application with an intensity of 0.5 w/cm², duration of 60 s, and a frequency of 1 MHz [11–14]. Then, TC@MEVs formulations were intravenously injected at 0.5, 1.0, and 1.5 h, respectively. The GBM mice were observed and their body weights were measured daily for 30 days. Survival was observed daily, until natural death or the end of the experiment. Mouse survival rates were calculated using the logarithmic rank test, and statistical significance was determined.

Two days after the completion of treatment, mouse brain tissues were obtained and fixed with 4% paraformaldehyde, followed by dehydration in alcohol solutions of varying concentrations. 4 mice per group, blinding and randomization treatment groups were allocated in a randomized fashion. After paraffin embedding, glioma sections were prepared according to the manufacturer’s protocol. Tumors were detected using proliferating cell nuclear antigen (PCNA) antibody (abs100392, Absin, Shanghai, China) and Ki67 antibody (ab15580, Abcam, Inc., USA). Tumor sections were dehydrated, rehydrated, and immunostained using an immunofluorescence detection kit (Sangon Biotech, Shanghai, China) and a TUNEL detection kit (Beyotime Biotech, Shanghai, China). Finally, Image-J software was utilized for image analysis.

Additionally, to further evaluate the effects of free MBs or US-only on tumor growth in GBM mice, GL261-luc cells (1*10⁵ cells) were implanted intracranially into 6-week-old C57 mice. Brain fluorescence signals were detected via IVIS imaging at 5, 10, 15, and 20 days post-implantation to assess tumor progression. Specifically, 10 days post-implantation, the corresponding mice in the free MBs group received tail vein injection of 2.5 µL/g MBs suspension (2*10⁷/mL). For the US-only group, corresponding mice underwent US treatment with the following parameters: intensity of 0.5 W/cm², duration of 60 s, and frequency of 1 MHz.

Analysis of TME

Two days after the completion of treatment, the levels of IFN-γ, IL-10, IL-2, TNF-α, and HMGB1 in tumor tissues were determined using corresponding ELISA kits (MM-0182M1, MM-0176M1, MM-0689M1, MM-0132M1, Jiangsu Meilian Biology Co., Ltd.). Each group contains 6 samples. Blinding and randomization treatment groups were allocated in a randomized fashion. Tumor tissues were first collected, homogenized, and then centrifuged. The supernatants were aliquoted and diluted with ELISA assay buffer according to the manufacturer’s instructions.

Furthermore, immunofluorescence staining was performed on tumor tissues from different treatment groups using triple markers (CD3, CD8, and DAPI-labeled nuclei). To compare CD8⁺ T cell infiltration and PD1 expression ratios in GBM mouse tumor tissues across three different treatments—Control, US + C@MEVs, and US + TC@MEVs—tumor tissues from each treatment group underwent immunofluorescence staining using quadruple labeling (PD1, CD3, CD8, and DAPI-stained nuclei).

For T cell analysis, 0.5 µg/mL of collagenase D (Roche, Basel, Switzerland), 0.5 µg/mL of DNaseI (Womei Biotechnology Co., Ltd., China), and 3 µg/mL of DNaseI (Sigma Aldrich, USA) were utilized [13, 22, 30]. Subsequently, brain-infiltrating immune cells were isolated using lymphocyte separation medium (17–5442−03, GE Healthcare, USA). Each group contains 5 samples. Blinding and randomization treatment groups were allocated in a randomized fashion. The cells were then stained for 90 min using live/dead dye, anti-CD3, anti-CD4, anti-CD8a, anti-IFN-γ⁺, anti-PD1⁺ (103112, 100203, 100537, 100707, Biolegend, USA), and anti-Foxp3 antibodies (126403, Biolegend, USA). Flow cytometry analysis was conducted on the stained cells, focusing on the CD3⁺CD8⁺T, IFN-γ⁺ CD3⁺CD8⁺ T, CD8⁺PD1⁺T, CD3⁺CD4⁺T, and CD4⁺Foxp3⁺T cell populations (2*10⁴ events were collected for analysis). Simultaneously, immunofluorescence staining was performed on tumor tissues from different treatment groups using quadruple markers (CD4, CD3, Foxp3, and DAPI-labeled nuclei) [13].

To further analyze the proportions of M1 or M2 TAMs, brain-infiltrating immune cells were isolated and stained for 90 min using live/dead dyes, anti-CD11b, anti-CD206, and anti-CD86 antibodies (103112, 100203, 100537, 100707 from Biolegend, USA). Each group contains 5 samples. Blinding and randomization treatment groups were allocated in a randomized fashion. Flow cytometry analysis was then conducted, focusing on the CD11b⁺, CD11b⁺CD206⁺, and CD11b⁺CD86⁺ cell populations (2*10⁴ events were collected for analysis) [22].

To further analyze the proportion and activity of dendritic cells (DCs), brain-infiltrating immune cells were isolated and stained with Live/Dead dye, anti-CD11c, anti-CD80, and anti-CD86 antibodies (103112, 100203, 100537, 100707 from Biolegend, USA) for 90 min. Each group contains 5 samples. Blinding and randomization treatment groups were allocated in a randomized fashion. Flow cytometry analysis was then performed, focusing on the CD11c⁺, CD11c⁺CD80⁺, and CD11c⁺CD86⁺ cell populations (2*10⁴ events were collected for analysis) [30].

To further analyze memory T cells, brain-infiltrating immune cells were isolated on day 25 after the completion of US + TC@MEVs treatment. The cells were then stained for 90 min using a combination of live/dead dyes and antibodies specific for CD3, CD8, CD44, and CD62L (catalog numbers 103112, 100203, 100537, 100707 from Biolegend, USA). Each group contains 5 samples. Blinding and randomization treatment groups were allocated in a randomized fashion. Following staining, the cells underwent flow cytometry analysis, focusing primarily on the CD3⁺CD8⁺ T, CD44⁺CD62L⁺, and CD44⁺CD62L⁻ T cell populations (with 2*10⁴ events collected for analysis) [13, 22].

Tumor Rechallenge study

In the tumor rechallenge experiment, four GBM mice that were cured after being subjected to US combined with TC@MEVs treatment were randomly selected and reinjected with 1*10⁵ GL261 cells (4µL). Meanwhile, C57BL/6J mice of the same age and weight, which had not undergone any treatment, were set as the control group and inoculated with the same GL261 cells. Each group contains 4 samples. Blinding and randomization treatment groups were allocated in a randomized fashion. Ten days later, MRI was used to monitor tumor size, memory T cells were analyzed, and the survival time of the mice was recorded.

Safety evaluation

To assess the safety of six different treatments on C57BL/6J mice, including PBS, TMZ, free TC (TMZ&CXCL10), TC@MEVs, US combined with T@MEVs, and US combined with TC@MEVs, various tests were conducted after the experiment. Each group contains 4 samples. Blinding and randomization treatment groups were allocated in a randomized fashion. Among these, the CXCL10 concentration was 40 ng/mL, and the TMZ concentration was 2.5 mM. Firstly, histological changes in organs such as the heart, liver, spleen, lungs, and kidneys were observed through HE staining to evaluate the toxic effects on major organs. Secondly, blood biochemical analysis was performed, focusing on key indicators reflecting liver and kidney function, including alanine aminotransferase (ALT), aspartate aminotransferase (AST), blood urea nitrogen (BUN), and creatinine (CRE). Additionally, routine blood tests were conducted, including red blood cell (RBC), white blood cell (WBC), and platelet (PLT) counts, as well as white blood cell subpopulation analysis, to assess the impact of treatment on immune function or the hematopoietic system.

Statistical analysis

Prior to analysis, the Kolmogorov-Smirnov test and Levene’s test were employed to assess the normality and homogeneity of variance of the raw data. Necessary data transformations were performed if the data did not conform to a normal distribution or if the variances were not homogeneous. Blinding and randomization treatment groups were allocated in a randomized fashion. Differences in CD8⁺ T cell ratios, cytokine activity, tumor volume, body weight, blood biochemistry, blood routine, inflammatory cytokine levels, as well as results from Western blotting, immunostaining, and flow cytometry analysis among different treatment groups were evaluated using one-way analysis of variance (ANOVA) followed by Tukey’s Honestly Significant Difference (HSD) post-hoc test. In cases of inhomogeneous variances, the rank sum test was applied. For the rechallenge study, independent sample t-tests were used to analyze and compare tumor volumes and memory T cell ratios in immune memory analysis. The statistical analysis of overall survival curve (Kaplan–Meier) was conducted using the log-rank test. Statistical significance was defined as a P-value less than 0.05. All statistical tests were performed using SPSS software (version 19.0, SPSS Inc.). Significance compared to the control group was denoted as follows: *p < 0.05, **p < 0.01, ***p < 0.001; significance in comparisons between groups was indicated by #p < 0.05, ##p < 0.01, ###p < 0.001; ns indicated no significance.

Results and discussion

Preparation and characterization of TC@MEVs

M-EVs, due to their inherent biological properties, can be regarded as ideal delivery vectors for targeting GBM [17, 18]. Previous research has indicated that M-EVs exhibit low expression of CD9 and high expression of TSG101 protein [18]. As shown in Fig. 2A, the low expression of CD9 and high expression of TSG101 protein in M-EVs provide further support for the M-EVs we obtained. Meanwhile, the expression of CD47 protein in M-EVs sends a “don’t eat me” signal, further enhancing their circulation stability [31, 32]. Pharmacokinetic experiments demonstrate that, compared to free CXCL10 or TMZ, the circulation time of T@MEVs is significantly prolonged (Figure S1A-B). The half-life values for pharmacokinetic data of CXCL10 and TMZ are 0.72 ± 0.07 µg/mL (2 h) and 94.47 ± 4.58 µg/mL (2 h), respectively (Figure S1C). For T@MEVs encapsulated CXCL10 and TMZ, the values are 0.79 ± 0.05 µg/mL (12 h) and 102.78 ± 1.61 µg/mL (24 h), respectively (Figure S1C).

Fig. 2.

Characterization of TC@MEVs. (A) The expression levels of ALIX, CD47, TSG101, and CD9 proteins in M(macrophage membranes), M-EVs, T@MEVs, and TC@MEVs. (B) Transmission electron microscopy (TEM) images of M-EVs, T@MEVs, TC@MEVs, and TC@MEVs. C-F) The average particle size of M-EVs (69.42 ± 11.16 nm), T@MEVs (72.84 ± 5.59 nm), and TC@MEVs (100.71 ± 8.12 nm) (n = 3). G) The average potential of M-EVs (−28.43 ± 0.75 mV), T@MEVs (−26.44 ± 1.18 mV), and TC@MEVs (−20.98 ± 1.37 mV) (n = 3). H) Concentration changes of TC@MEVs resuspended in PBS (pH 7.4), DMEM, or 10% FBS (n = 3). I-J) 7-day particle size and potential changes of TC@MEVs (n = 3). K-N) The encapsulation efficiency and drug loading capacity of temozolomide (TMZ) and CXCL10 were determined at different input ratios (n = 3). O-Q) TC@MEVs released CXCL10 in response to GBM suspension with 4.5 µmol/mL GSH or mouse brain tissue suspension with 1.5 µmol/mL GSH (n = 3). All statistics are expressed as mean ± standard deviation. Statistical significance was calculated by one-way ANOVA with the Tukey post hoc test. *p < 0.05, ***p < 0.001, ns: no significance

To further assess the biological characteristics of TC@MEVs, TEM analysis was performed. The results showed that, compared to M-EVs, no significant morphological changes were observed in T@MEVs and TC@MEVs (Fig. 2B). However, TC@MEVs underwent lysis and morphological alterations when exposed to a GSH solution at a concentration of 4.5 µmol/mL (Fig. 3B). This suggests that redox-responsive TC@MEVs undergo lysis in response to GSH, leading to drug release. This effect is primarily due to the cleavage of disulfide bonds in the active ester crosslinker [30].

Fig. 3.

Ultrasound/microbubbles-mediated opening of the blood-brain barrier model. (A) Schematic diagram illustrating TC@MEVs-responsive CXCL10 release and T cell recruitment in vitro. (B) CCK-8 assay to detect GL261 cell viability in the lower chamber of the transwell system (n = 4). C-F) Flow cytometry analysis of the proportion of CD3⁺CD8⁺ T cells and CD3⁺CD4⁺ T cells in the lower chamber of the transwell system (n = 4). G) Schematic diagram illustrating BBB opening via ultrasound in vitro. H-I) Fluorescence microscopy observation of fluorescence signals (Cy5 NHS-labeled CXCL10) in the lower chamber of the transwell system, with fluorescence signal quantitation using ImageJ (n = 4). J-K) Flow cytometry detection of GL261 cell apoptosis rate (n = 4). L) Immunofluorescence detection of ZO-1 expression in a BBB model constructed with bEnd.3 cells. All statistics are expressed as mean ± standard deviation. Statistical significance was calculated by one-way ANOVA with the Tukey post hoc test. **p < 0.01, ***p < 0.001, ns: no significance

According to Malvern particle size analyzer measurements, the average particle size of M-EVs was determined to be 69.42 ± 11.16 nm, T@MEVs exhibited an average size of 72.84 ± 5.59 nm, and TC@MEVs had an average size of 100.71 ± 8.12 nm (Fig. 2C-F). Furthermore, the average potential of M-EVs was found to be −28.43 ± 0.75 mV, with T@MEVs carrying a charge of −26.44 ± 1.18 mV, and TC@MEVs having a charge of −20.98 ± 1.37 mV (Fig. 2G).

The stability of TC@MEVs was assessed over a 7-day period in PBS (pH 7.4), 

DMEM, or 10% FBS, showing negligible changes in particle size and concentration (Fig. 2H-J). Additionally, efficient delivery of the nano-preparation relies on high drug loading capacity and encapsulation efficiency [7]. Modifying the input ratios of TMZ and CXCL10 in TC@MEVs revealed that both maintained high drug loading capacity and high encapsulation efficiency when TMZ was at 15% and CXCL10 at 20% (Fig. 2K-N). IVIS imaging showed that in tumor tissue homogenates from GBM mice with a GSH concentration of 4.5 µmol/mL, TC@MEVs completed the full release of CXCL10 within 8 h. Conversely, in normal brain tissue homogenates with a GSH concentration of 1.5 µmol/mL, there was no significant release of CXCL10 within 5 h (Fig. 2O-Q). These findings indicate that TC@MEVs exhibit excellent dispersibility, stability, and high drug loading capacity. Furthermore, under conditions of high GSH concentration, TC@MEVs show promise for selectively releasing CXCL10 and TMZ into the GBM region.

Further detection results revealed that, compared to the pH (6.5) + GSH (1.5 µmol/g) group, the CXCL10 release efficiency in the pH (6.5) + GSH (4.5 µmol/g) group increased by 164.31%. Similarly, compared to the pH (7.4) + GSH (4.5 µmol/g) group, the CXCL10 release efficiency in the pH (6.5) + GSH (4.5 µmol/g) group was enhanced by 109.02% (Figure S1D). These findings indicate that GSH, rather than pH, serves as the primary trigger facilitating the release of CXCL10 from TC@MEVs. CCK-8 assays revealed that the IC₅₀ value of TMZ for GL261 cells was 2.5 mM (Figure S2A), while the IC₅₀ value for free CXCL10 was 40 ng/mL and that for the TC@MEVs formulation was 250 ng/mL (Figure S2B-C).

TC@MEVs release CXCL10 to recruit T cells

To evaluate the effect of CXCL10 on T-cell recruitment [5, 6], we designed an in vitro simulation experiment based on the transwell system (Fig. 3A). Results of the CCK-8 assay indicated that the viability of GL261 cells in the TC@MEVs + GSH (4.5 µmol/mL) group was reduced by 2.06-fold compared with that in the free TC@MEVs treatment group (Fig. 3B).

Flow cytometry analysis showed that, compared to the control group, the proportion of CD3⁺CD8⁺ and CD3⁺CD4⁺ T cells in the lower chamber of the transwell system increased by 3.25-fold and 3.19-fold, respectively, in the free CXCL10 treatment group (Fig. 3C–F). In the TC@MEVs + GSH (4.5 µmol/mL) treatment group, the proportion of CD3⁺CD8⁺ and CD3⁺CD4⁺ T cells increased by 1.98-fold and 2.16-fold, respectively, compared to the free TC@MEVs treatment group (Fig. 3C–F). No significant difference in T cell proportions was observed between the free CXCL10 and TC@MEVs + GSH (4.5 µmol/mL) treatment groups (Fig. 3C–F). These findings suggest that TC@MEVs respond to GSH (4.5 µmol/mL), releasing both TMZ and CXCL10 while maintaining CXCL10’s T cell-recruiting activity.

The results from the CXCR3 receptor blockade assay demonstrated that, compared to the control group, the proportion of CD3⁺CD8⁺ T cells in the lower chamber was significantly increased in both the CXCL10 group and the TC@MEVs + GSH group (Figure S3A–B). In contrast, no significant change in the proportion of CD3⁺CD8⁺ T cells was observed in the lower chamber of the CXCL10 + AMG487 group or the TC@MEVs + GSH + AMG487 group, relative to the control group (Figure S3A–B). These findings suggest that T cell recruitment is specifically mediated by CXCL10.

To further investigate the effects of US/MBs on the BBB, we designed an in vitro BBB model based on bEnd.3 cells (Fig. 3G). We prepared a phospholipid MBs suspension, and the MBs had a uniform size distribution with a diameter of approximately 1 μm (Figure S3C). Previous studies have shown that in vitro BBB models, constructed using bEnd.3 cells, commonly validate the integrity of tight junctions in the BBB model by assessing the expression of the tight junction protein ZO-1 [11–14]. Additionally, we verified the barrier function of the model by measuring trans-endothelial electrical resistance (TEER) to assess the permeability of the monolayer, which further supported the presence of functional tight junctions (Figure S4A).

Immunofluorescence experiments demonstrated that US/MBs treatment disrupted the tight junction protein ZO-1, allowing TC@MEVs to penetrate the BBB model and enter the lower chamber of the transwell system (Figure S4B). However, compared with the PBS group, US alone failed to induce significant alterations in the BBB model (Figure S4B). Compared to the free TC@MEVs-treated group, the TC@MEVs + US-treated group showed a 2.46-fold increase in CXCL10-Cy5 levels in the lower chamber of the BBB model (Fig. 3H-I). Additionally, EVs derived from TAMs (EVs@TAM) or RAW 264.7 cells (EVs@RAW) both possess the ability to migrate toward GL261 cells, with no significant difference observed (Figure S4C-D).

CC-K8 and TUNEL assay results indicate that a TMZ concentration of 2.5 mM in the lower chamber of the transwell system had no significant effect on bEnd.3 cell viability or T cell activity (Figure S4E-G). Compared to the control group, TMZ concentrations exceeding 5 mM induced a significant increase in bEnd.3 cell apoptosis and a significant decrease in T cell activity (Figure S4E-G).

Flow cytometry analysis revealed that, compared to the free TC@MEVs-treated group, the TC@MEVs + US-treated group showed a 2.78-fold increase in the apoptosis rate of GL261 cells in the lower chamber of the BBB model (Fig. 3J-K). Additionally, observation of structural changes in the tight junctions (Zo-1 protein) of the BBB model showed that the opening effect of US/MBs on the BBB model increased with increasing ultrasound intensity (Fig. 3L).

These findings suggest that US/MBs have the potential to overcome the BBB obstacle and promote the penetration of TC@MEVs through the BBB model. However, the extent of this effect may depend on various factors such as ultrasonic intensity and duration [13]. Further studies are needed to fully evaluate the effectiveness of this approach in enhancing chemo-immunotherapy by recruiting T cells.

US/MBs safely and reversibly open the BBB in vivo

To avoid potential toxic effects associated with BBB opening, it is crucial to further evaluate the safety and reversibility of the US/MBs method for BBB opening [11, 12, 15, 16]. In vivo studies have demonstrated that the use of US or MBs alone does not achieve BBB opening (Figure S5). Additionally, BBB opening with US/MBs requires a sound intensity of at least 0.3 W/cm², and the opening effect becomes more pronounced as the sound intensity increases (Figure S5). Compared with the control group (0 w/cm²), HE staining results indicate significant blood cell extravasation when the sound intensity exceeds 0.7 w/cm² (Figure S6A-B). Simultaneously, TUNEL apoptosis immunofluorescence detection reveals notable brain tissue apoptosis at sound intensities above 0.7 w/cm², compared with the control group (0 w/cm²) (Figure S6A, C).

Furthermore, immunofluorescence experiments for inflammatory cytokines IL-6 and IL-2 show significant inflammatory cytokine infiltration at sound intensities greater than 0.7 w/cm², compared with the control group (0 w/cm²) (Figure S7A-B and Figure S8A-B). Conversely, at sound intensities ranging from 0.3 to 0.5 w/cm², there is no significant blood cell extravasation, brain tissue apoptosis, or inflammatory cytokine infiltration observed, providing a 4-hour window for reversible BBB opening (Figure S5-S9). These findings suggest that the safe sound intensity range for BBB opening with US/MBs is between 0.3 and 0.5 w/cm², and that the BBB opening may remain reversible within a 4-hour window. However, the long-term effects of this technique on BBB integrity and brain tissue remain to be fully investigated [11–14].

US-mediated BBB opening enhances TC@MEVs delivery

Encouraged by the above experimental results, we further conducted in vivo experiments to explore the effect of US-guided TC@MEVs penetration through the BBB into the brain. The detection results from the multifunctional microplate reader showed that the efficiency of US-guided TC@MEVs delivering CXCL10 into the brain increased with higher sonic intensity, compared with the control group (0 w/cm²) (Figure S10). US/MB-induced BBB opening, as evidenced by EB dye penetration in brain tissue sections, suggests that ultrasound-guided TC@MEV delivery can potentially achieve deep penetration throughout tumor tissue (Fig. 4A).

Fig. 4.

Ultrasound/microbubbles improve the brain penetration efficiency of TC@MEVs. A) Observation of the effect of ultrasound/microbubbles (US/MBs) on the in vivo opening of the blood-brain barrier (BBB) through brain tissue sections in vivo (20 µL, 2% Evan’s blue dye). B-C) Evaluation of the efficiency of US-guided TC@MEVs delivery into the brain using the in vivo imaging system (IVIS) (Cy5 NHS-labeled CXCL10), with fluorescence signal values analyzed and quantified using ImageJ software (n = 3). D-E) Assessment of the organ distribution of CXCL10 using IVIS, with fluorescence signal values analyzed and quantified by ImageJ software (n = 3). F) Distribution of CXCL10-Cy5 in tumor tissue (T: tumor, N: normal). All statistics are expressed as mean ± standard deviation. Statistical significance was calculated by one-way ANOVA with the Tukey post hoc test. ***p < 0.001

IVIS imaging results indicated that compared to the Free CXCL10 treatment group, the TC@MEVs treatment group enhanced the brain delivery efficiency of CXCL10, suggesting that TC@MEVs retained a certain degree of M-EVs’ ability to penetrate the BBB (Fig. 4B-C). Compared to the TC@MEVs treatment alone, the combination of US and TC@MEVs enhanced both the brain delivery efficiency and duration of CXCL10, primarily due to the improved TC@MEVs penetration through the BBB during US/MBs-induced BBB opening (Fig. 4B-C). IVIS imaging of tissue organs demonstrated that the US combined with TC@MEVs treatment group significantly improved the brain delivery efficiency of CXCL10 compared to other groups, while the residual amount in the liver was significantly reduced (Fig. 4D-E).

Brain tissue sections analysis showed significantly higher CXCL10-Cy5 levels in the GBM region of the US + TC@MEVs group compared to the free TC@MEVs group (Fig. 4F). Similarly, IVIS was used to detect the efficiency of US-guided TC@MEVs delivering TMZ into the brain. The results showed that compared to other treatment

groups, the combination of US and TC@MEVs significantly enhanced the brain delivery efficiency of TMZ (Figure S11A-B). The US + TC@MEVs group showed a significant reduction in TMZ content in liver tissue compared to other treatment groups (Figure S12A-B). GBM mice inherently exhibit some degree of BBB disruption, but this disruption only slightly improves drug penetration efficiency and remains limited in terms of deep tissue penetration [2, 11, 13]. In contrast to free PC, TC@MEVs exhibit chemotactic activity toward the GBM, leading to an increase in CXCL10 concentration within the tumor. The combined administration of TC@MEVs with US/MBs-mediated BBB opening further enhances the efficiency of TC@MEVs entering the GBM and releasing CXCL10.

These results demonstrate that US/MBs effectively open the BBB, enabling the release of CXCL10 and TMZ from TC@MEVs into the GBM region. However, given that GSH concentrations vary across different stages of GBM progression, this variability may lead to inconsistent drug release rates [2, 26, 33]. Therefore, further validation of the antitumor efficacy using an orthotopic GBM mouse model is warranted [2, 7, 26].

Anti-tumor effects

To further evaluate the anti-tumor effects in vivo, an orthotopic GBM mouse model was constructed [1]. MRI was used to measure the proportion of the tumor relative to the entire brain volume every 5 days to track tumor progression, and a treatment regimen combining US with TC@MEVs was designed (Fig. 5A-B). IVIS imaging results indicate that tumor growth in GBM mice treated with free MBs or US alone showed no significant difference compared to the PBS group (Figure S13A-B). This suggests that tail vein injection of MBs or US alone does not induce GBM suppression.

Fig. 5.

Antitumor effects in orthotopic glioblastoma (GBM) mice. A-B) Illustrates the treatment schedule for the GL261 mouse brain tumor model, detailing the treatment sequence and specific time points. C-D) Presents brain images of mice from each group, captured by T2 MRI after undergoing various treatments. The tumors are outlined with red dashed lines (n = 5). E) Depicts survival curve (Kaplan–Meier) for GBM mice treated with different preparations (n = 10). Statistical significance was determined through the log-rank test. F-I) Shows immunohistochemical images of tumor sections examined at 17 days post-rhabdomyolysis, stained for Ki67, PCNA, and TUNEL (n = 4). J-K) Assessment of GBM mouse tumor growth in different treatment groups using the in vivo imaging system (IVIS) (n = 5). Data are expressed as mean ± standard deviation. Statistical significance was calculated by one-way ANOVA with the Tukey post hoc test. *p < 0.05, ***p < 0.001, and ^##p < 0.01, ^###p < 0.001, ns: no significance

As shown in Fig. 5C-D, mice treated with PBS, free TMZ, and the Free TC group exhibited rapid tumor progression, indicating that systemic delivery of TMZ and CXCL10 failed to achieve an effective T-cell response and tumor elimination. However, compared to the Free TC group, the TC@MEVs group showed a certain degree of tumor growth inhibition in GBM mice (Fig. 5C-D). When compared to the TC@MEVs group, the US + TC@MEVs group significantly inhibited tumor growth and achieved 40% tumor elimination (Fig. 5C-D). The US + T@MEVs group demonstrated some degree of tumor growth inhibition but failed to achieve partial tumor elimination (Fig. 5C-D). This may be due to the efficient delivery of TMZ, which boosts its antitumor efficacy, but its anti-GBM activity is limited by lymphocyte depletion and a highly immunosuppressive TME [13, 22, 34].

On the other hand, GBM mice treated with Free TMZ or Free TC had a short survival time of less than 35 days and significant weight loss (Fig. 5E and Figure S13C). In contrast, the survival period of GBM mice in the US + TC@MEVs group was significantly extended, and their body weight remained largely unchanged (Fig. 5E and Figure S13C). Histological analysis of brain tumor tissues revealed that, compared to the PBS group, the TC@MEVs group exhibited a 6.64-fold increase in the apoptosis rate and a 1.90-fold decrease in the proliferation rate of the corresponding tumor tissues (Fig. 5F–I). In comparison with the TC@MEVs group, the US + TC@MEVs group showed a 1.84-fold increase in the apoptosis rate and a 4.17-fold decrease in the proliferation rate of the corresponding tumor tissues (Fig. 5F–I and Figure S14).

Transfection of GL261 cells with the Lent-EF1α-P2A-luciferase-CMV-coGFP-P2A-Puro lentivirus revealed that the optimal multiplicity of infection (MOI) was 100 (Figure S15). Following this, GL261-GFP-Luc cell lines were screened and purified using 3 µg/mL puromycin (Figure S16A–B). The dual-luciferase GBM mice constructed from the GL261-GFP-Luc cell line were further identified by the IVIS system, and the optimal detection time window was determined to be 7–13 min (Figure S17).

IVIS detection results further showed that tumor progression was significantly inhibited in the TC@MEVs treatment group compared to GBM mice treated with Free TC (Fig. 5J–K). The US + TC@MEVs group significantly inhibited tumor progression and achieved partial tumor elimination compared to GBM mice treated with TC@MEVs (Fig. 5J–K). These results confirm that the US + TC@MEVs group significantly inhibits tumor progression, suppresses proliferation, promotes apoptosis, and extends survival in GBM mice.

Enhancing tmz’s antitumor effects by recruiting T cells

Five days after the completion of treatment in GBM mice, we performed flow cytometry and immunofluorescence analysis to assess changes in T cells within the TME of GBM mice. Flow cytometry showed a 1.50-fold increase in the proportion of CD3⁺ CD8⁺ T-cells and a 1.46-fold increase in tumor-specific T-cells in the tumor tissue of GBM mice treated with TC@MEVs compared to Free TC-treated mice (Fig. 6A, B, E, F). Furthermore, in GBM mice subjected to US combined with TC@MEVs treatment, the proportion of CD3⁺CD8⁺ T-cells increased by 5.52-fold, and tumor-specific T-cells increased by 3.01-fold compared to Free TC group(Fig. 6A, B, E, F). These findings suggest that TC@MEVs enhance the brain delivery of TMZ and CXCL10, while US/MBs improve TC@MEVs penetration through the BBB.

Fig. 6.

T-cell infiltration in tumor tissues of glioblastoma (GBM) mice. A, E) Flow cytometry analysis of the proportion of CD3⁺CD8⁺ T cells in tumor tissues (n = 5). B, F) Flow cytometry detection of tumor-specific T-cell ratios in tumor tissues. C-D) Quantification of IFN-γ and TNF-α in tumor tissues by Elisa (n = 5). G) Flow cytometry examination of PD1 expression on CD3⁺CD8⁺ T cells in tumor tissues. H) Immunofluorescence observation of CD3⁺CD8⁺ T-cell infiltration in tumor tissues (blue: DAPI; green: Alexa Fluor 488-CD3; red: CY3-CD8). Data are expressed as mean ± standard deviation. Statistical significance was calculated by one-way ANOVA with the Tukey post hoc test. *p < 0.05, **p < 0.01, ***p < 0.001, and ^###p < 0.001, ns: no significance

ELISA results indicated that the levels of IFN-γ and TNF-α in tumor tissues of GBM mice treated with US + TC@MEVs were increased by 3.58-fold and 3.57-fold, respectively, compared with the free TC group (Fig. 6C, D). Compared with the free TC group, the proportion of PD1-expressing CD3⁺CD8⁺ T cells in tumor tissues was reduced by 1.61-fold in the US + TC@MEVs group (Fig. 6G and Figure S18A). Meanwhile, ELISA results showed that the level of HMGB1 in tumor tissues of GBM mice treated with US + TC@MEVs was increased by 4.84-fold compared with the free TC group (Figure S18B). Immunofluorescence staining revealed a significant increase in CD3⁺ CD8⁺ T-cell infiltration in the tumor tissue of GBM mice treated with US + TC@MEVs, compared to other treatment groups (Fig. 6H).

Compared to the control group, the US + C@MEVs group exhibited increased infiltration of CD8⁺ T cells (Figure S18C). Relative to the US + C@MEVs group, the US + TC@MEVs group showed a reduced proportion of CD8⁺ T cells expressing PD-1 (Figure S18C). These results suggest that CD8⁺ T cells recruited by CXCL10 may be influenced by the immunosuppressive TME of GBM, leading to a certain degree of exhaustion. When both CXCL10 and TMZ were delivered simultaneously to the GBM region, CD8⁺ T cell exhaustion could be partially rescued.

These results suggest that following the opening of the BBB by US/MBs, TC@MEVs enter the GBM region and respond by releasing CXCL10 and TMZ. CXCL10 recruits T-cells to the GBM region, generating anti-tumor activity and releasing proinflammatory cytokines like IFN-γ and TNF-α [35, 36]. Efficient TMZ delivery accelerates tumor necrosis, releasing antigens such as HMGB1 that activate the immune system [27]. The recruitment of T-cells amplifies the anti-tumor effects of TMZ.

Immunosuppressive TME of GBM is reprogrammed

Based on alterations in the quantity and state of T cells within the TME, as well as the surge in proinflammatory cytokines, it is inevitable that the entire TME will be further activated [33, 36, 37]. We comprehensively evaluated the proportions and activation states of DCs, macrophages, and Treg cells in the TME of GBM mice treated with US + TC@MEVs (Fig. 7A).

Fig. 7.

Reprogramming of the tumor microenvironment in glioblastoma (GBM) mice. A) Schedule for in vivo tumor immune profiling in the GL261 mouse brain tumor model. B-C) Quantification of IL-2 and IL-10 levels in tumor tissues determined by Elisa (n = 5). D, J) Flow cytometry analysis of Treg cell proportion in the TME (n = 5). E) Changes in the ratio of CD8⁺ T cells to Treg cells in the TME (n = 5). F, G, K, L) Flow cytometry evaluation of dendritic cell (DC) proportion and activity changes in the TME (n = 5). H) Flow cytometry assessment of macrophage proportion in the TME (n = 5). I) Flow cytometry assessment of M1/M2 macrophage ratio in the TME (n = 5). M) Flow cytometry assessment of M2-like macrophage proportion and activity alterations in the TME. N) Flow cytometry assessment of M1-like macrophage proportion and activity variations in the TME. Data are expressed as mean ± standard deviation. Statistical significance was calculated by one-way ANOVA with the Tukey post hoc test. *p < 0.05, **p < 0.01, ***p < 0.001, and ^#p < 0.05, ^##p < 0.01, ^###p < 0.001, ns: no significance

According to the ELISA results, compared with the Free PC group, the IL-2 level in GBM tumor tissues in the US + TC@MEVs group increased by 4.32-fold, while the interleukin-10 (IL-10) level decreased by 4.10-fold (Fig. 7B–C). Flow cytometry analysis revealed that the proportion of CD4⁺ T cells in GBM showed no significant difference between the US + TC@MEVs group and the PBS group (Figure S19A–B). Flow cytometry showed a 1.84-fold decrease in Treg cells and a 9.84-fold increase in the CD8 + T/Treg ratio in the TME of GBM mice treated with US and TC@MEVs compared to Free TC-treated mice (Fig. 7D, E, J). Additionally, immunofluorescence staining showed a notable reduction in Treg cell infiltration in the TME of mice receiving US + TC@MEVs group, compared to other groups (Figure S20).

We further evaluated the activation and proportions of DCs, M1-like TAMs, and M2-type TAMs within the TME [33, 37]. Flow cytometry results demonstrated that the proportion of DCs in the TME of GBM mice treated with TC@MEVs increased by 1.18-fold, compared to Free TC group (Figure S21A-B). This proportion was further augmented by 2.62-fold in mice receiving US + TC@MEVs group, compared to TC@MEVs alone (Figure S21A-B).

Compared to the Free PC group, tumor tissues from GBM mice treated with US + TC@MEVs showed enhanced DCs activity [38], with CD80 and CD86 expression of DCs increasing 2.77-fold and 3.74-fold, respectively (Fig. 7F, G, K, L). Flow cytometry analysis showed a significant increase in macrophage proportion, with a notable shift from M2-type to M1-like TAMs [39–41], in the TME of mice treated with US combined with TC@MEVs, compared to other groups (Fig. 7H, I, M, N, and Figures S22, S23A-B).

CXCL10 recruits circulating T cells (e.g., CD4⁺ and CD8⁺ T cells) into the GBM region, resulting in a relative decrease in the proportion of Tregs [6, 13]. Furthermore, alterations in the TME, such as increased levels of cytokines TNF-α and IFN-γ and decreased IL-10, further suppress the recruitment and expansion of Tregs [13, 22]. By binding to the CXCR3 receptor, CXCL10 promotes the polarization of M1-type macrophages [13, 32]. These M1 macrophages exhibit enhanced antitumor immune activity and facilitate the activation of effector T cells (e.g., CD8⁺ T cells) through the secretion of cytokines such as IL-2 and TNF-α, thereby augmenting the immune response against the tumor [13, 32].

These findings suggest that the immunosuppressive TME in GBM mice treated with US combined with TC@MEVs is activated, resulting in increased activity of antitumor immune cells and a decrease in immunosuppressive cells such as Tregs [33, 37].

Establishment of antitumor immune memory

Increased T-cell infiltration in the TME not only activates the environment but also promotes the formation of immunological memory [37, 42, 43]. Therefore, we further examined the proportions of effector memory T cells (Tem) and central memory T cells (Tcm) in the TME of GBM mice.

Flow cytometry results showed that, compared to the free TC group, the TC@MEVs group exhibited a 1.35-fold increase in the proportion of Tem and a 1.09-fold decrease in the proportion of Tcm in GBM mouse tumor tissues (Figures S24A-B and S25A-D). Compared to the TC@MEVs group, the US + TC@MEVs group showed a further 1.45-fold increase in Tem proportion and a 1.41-fold decrease in Tcm proportion in their TME (Figure S25A-D). It is noteworthy that, compared to the Free TC group, the Tem/Tcm ratio in the TME of GBM mice treated with US + TC@MEVs increased by 2.38-fold (Figure S25A-D).

Additionally, GBM mice that were treated with US combined with TC@MEVs and then cured were re-inoculated with GL261 cells [34, 44]. MRI results revealed that, compared to native mice inoculated with GL261 cells for the first time, tumor growth in the re-challenge group was significantly inhibited, and their survival was notably prolonged (Figure S26A-C). Flow cytometry analysis further demonstrated that, relative to naïve mice, the proportion of Tem cells in the tumor tissue of the re-challenge group increased by 6.49-fold, while the proportion of Tcm cells decreased by 1.42-fold, resulting in a 9.10-fold increase in the Tem/Tcm ratio (Figure S26D-F).

Following treatment with US + TC@MEVs, the increase in the proportion of Tem cells indicates that the therapy successfully stimulated an immune response in T cells and promoted the differentiation of more T cells into Tem cells, which possess rapid-response characteristics [22, 30, 43]. Tem cells exhibit strong immediate immune response capacity [42, 43]. Upon re-encountering the same antigen, they can rapidly exert effector functions, directly killing target cells or secreting cytokines to mediate immune responses [13, 42, 43]. In contrast, Tcm cells have a weaker immediate response capability [13, 42, 43]. As long-term memory reservoirs, Tcm cells may not undergo a sharp increase in number during the initial phase of the antitumor immune response, but rather play a role in sustaining long-term immunity [13, 22, 30, 42, 43].

These results suggest that GBM mice treated with US combined with TC@MEVs not only achieved partial tumor elimination but also developed robust, long-term immune memory.

Safety evaluation

The advantage of nano-preparations lies in their ability to enhance drug delivery efficiency while reducing toxic side effects on tissues and organs [7, 34, 45]. Therefore, we further conducted HE staining on the heart, liver, spleen, lung, and kidney of GBM mice subjected to different treatments [45, 46]. The results revealed that compared to the control group, GBM mice treated with Free TMZ or Free TC exhibited slight structural changes in their liver and kidney, suggesting the presence of hepatorenal toxicity (Figure S27). However, no structural alterations were observed in the liver and kidney of GBM mice treated with TC@MEVs alone or US + TC@MEVs (Figure S27).

Additionally, blood biochemical tests indicated that the levels of alanine aminotransferase (ALT), aspartate aminotransferase (AST), total bilirubin (TBIL), Urea, and creatinine (Crea) in the plasma of GBM mice treated with Free TMZ or Free TC were beyond the normal range, indicating hepatorenal toxicity (Figure S28A-F). Conversely, all blood biochemical parameters remained within the normal range for GBM mice treated with TC@MEVs alone or US + TC@MEVs. No significant changes were observed in the blood cell components among the different treatment groups, all remaining within normal limits (Figure S29A-F and Figure S30A-D).

These results demonstrate that US/MBs combined with TC@MEVs effectively deliver TMZ and CXCL10 to the GBM region. Potent antitumor effects were observed, and preliminary safety data indicate minimal hepatic and renal toxicity. However, long-term safety studies are needed to confirm these findings.

Conclusion

In this study, we designed a redox-responsive M-EVs-based nanoplatform (TC@MEVs) that co-delivers CXCL10 and TMZ, combined with US/MBs-mediated BBB opening. This approach aims to address the challenges posed by the immunosuppressive TME and the limited permeability of the BBB, with the potential to enhance GBM chemo-immunotherapy. However, further clinical studies are needed to validate the therapeutic benefits and safety of this strategy [1–3]. In vitro experiments demonstrated that TC@MEVs exhibited excellent stability, high drug-loading capacity, and GSH-responsive release of CXCL10 and TMZ [32]. In the BBB model experiments, US/MBs facilitated TC@MEVs penetration, and further releases CXCL10 effectively recruited CD8⁺ T cells, synergizing with TMZ to induce tumor cell apoptosis.

In vivo experiments revealed that US-guided TC@MEVs achieved efficient BBB penetration. This approach significantly enhanced CD8⁺ T-cell infiltration (5.52-fold higher than free CXCL10), reduced immunosuppressive Tregs and M2-like TAMs, and activated DCs and M1-like TAMs [38–41, 47]. These immune changes were associated with tumor suppression and prolonged survival in orthotopic GBM mice. However, further studies are needed to explore the full extent of the immune response and its impact on tumor recurrence. The treatment elicited long-term immune memory. This was evidenced by elevated Tem-cell proportions and resistance to tumor rechallenge.

Safety assessments underscored the translational potential of this strategy. Optimized US parameters (1 MHz, 0.5 w/cm²) achieved transient BBB opening (< 4 h) without microhemorrhage or cytokine exudation. Histopathological analysis of major organs showed no evidence of hepatic or kidney injury in the short term, and blood chemistry parameters (ALT, AST, BUN) remained within normal ranges. Long-term monitoring of these parameters is necessary to assess potential delayed toxic effects [45, 46]. TC@MEVs’ intrinsic CD47 “self-marker” prevented off-target phagocytosis and avoided inflammatory cytokine storms [16, 25, 38, 46, 48].

Despite promising results, several limitations should be considered [2, 7–9, 12, 13, 22, 30, 34]. First, the in vivo experiments were conducted on a relatively small sample size of mice, and further studies with larger sample sizes are needed to confirm the reproducibility and generalizability of these findings [2, 7]. Second, while the effects of US combined with TC@MEVs were observed at various time points, the long-term therapeutic outcomes, including potential side effects or risks associated with repeated treatments, remain unclear [7–9]. Third, the study primarily focused on acute responses, and further research is needed to assess the chronic effects of US/MBs treatment on the BBB integrity and overall brain health [6, 7, 13]. Additionally, although significant immune modulation was observed, the precise mechanisms underlying the activation of immune cells and their interactions within the tumor microenvironment require more detailed investigation [8, 9, 13, 22]. Lastly, while our study shows promising preclinical results, clinical trials are needed to validate the safety and efficacy of this therapeutic strategy in human subjects [30, 34, 44–46].

In conclusion, this spatiotemporal combination of US-gated BBB opening and GSH-responsive M-EVs delivery system to reverse the immunosuppressive TME by improving the infiltration of CD8⁺ T cells and local release of TMZ, offer a promising strategy for effective GBM chemo-immunotherapy.

Author contributions

Author contributionsL.D. conducted the synthesis, characterization, cell studies, animal experiments, and data analysis. L.D., Q.X., Z.G., H.G., and Z.W. performed the construction of an orthotopic brain tumor model. L.D., Q.X., R.L., H.J., Z.Y., and H.L. wrote the manuscript. L.D., J.W., R.L., H.J., Z.Y., and H.L. contributed to the data analysis and discussed the data. H.Z., D.L., R.L., and H.Y. conceived and designed the experiments. H.Z., and D.L. supervised the entire project. L.D., J.W., Z.G. and H.Z. revised the manuscript. We thank the Home for Researchers editorial team (www.home-for-researchers.com) for the language editing service. Thanks to Wuhan Ruisaiqi Biotechnology Co., Ltd for the technical support of multicolor immunofluorescence. All authors reviewed the manuscript. We extend our gratitude to Zhongmo Biology for their technical support.

Funding

This work was supported by the National Natural Science Foundation of China, National Major Scientific Instrument Development Project (82227808). National Natural Science Foundation of China (81371678). Science and Technology Plan Project of Jiangsu Province, China (BE2019716, BE2019738). Henan Medical Science and Technology Research Program (LHGJ20210527). Sanming Project of Medicine in Shenzen Municipality (SZZYSM202311016). SEU Innovation Capability Enhancement Plan for Doctoral Students (CXJH_SEU 24220).

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval

Female C57BL/6 mice (6–8 weeks old) were purchased from Hangzhou Qizhen Laboratory Animal Co., Ltd. All animal experiments were performed in compliance with the relevant laws and approved by the Institutional Animal Care and Use Committee of Southeast University School of Medicine (NO.SEU-IACUC-20250219004).

Competing interests

The authors declare no competing interests.