Surface-engineered milk-derived extracellular vesicles enable oral miRNA therapy against immunosuppressive pancreatic cancer

Abstract

Pancreatic ductal adenocarcinoma (PDAC) shows limited responsiveness to conventional therapies due to an immunosuppressive tumor microenvironment (TME) characterized by programmed death-ligand 1 (PD-L1)-mediated immune escape and oncogenic microRNA-21 (miR-21) signaling. Despite the potential of miRNA therapeutics, their clinical translation is hindered by the lack of stable, non-invasive delivery systems capable of bypassing complex biological barriers. Here, we develop a scalable biogenic delivery platform based on surface-engineered milk-derived extracellular vesicles (MEVs) for systemic transport of anti-miR-21 (α21) after oral administration. MEVs were isolated using a reproducible purification workflow and functionalized with a PD-L1-binding peptide (PBP) to facilitate tumor-associated uptake while preserving vesicle integrity and colloidal stability. The engineered vesicles (α21@MEV_PBP) remained stable under simulated gastrointestinal conditions, underwent efficient transepithelial transport, and exhibited selective accumulation in PD-L1^high pancreatic tumors in vivo, demonstrating a gut-to-tumor delivery pathway. Following cellular internalization, α21 delivery restored tumor-suppressive signaling and induced endoplasmic reticulum stress–associated immunogenic cell death (ICD). Concurrently, modulation of tumor-associated macrophages toward a pro-inflammatory phenotype enhanced CD8⁺ T cell activation and antitumor immune responses, leading to significant tumor growth inhibition without detectable systemic toxicity. These findings establish α21@MEV_PBP as an orally deliverable nanoplatform that bridges biogenic material engineering with targeted immune modulation. Together, the results demonstrate that surface engineering of naturally derived vesicles enables controlled systemic RNA transport via an oral administration route and provides a generalizable strategy for non-invasive delivery of nucleic acid therapeutics to solid tumors.

Graphical abstract

α21@MEV_PBP represents a non-invasive oral immunotherapy that selectively accumulates in PD-L1-overexpressing pancreatic tumors. By delivering anti-miR-21, it suppresses oncogenic signaling, induces immunogenic cell death, and reprograms immunosuppressive macrophages toward a pro-inflammatory state. This dual-action mechanism overcomes immune resistance and restores antitumor immunity in treatment-refractory PDAC.

Highlights

• •Surface-engineered MEV enable oral gut-to-tumor RNA delivery.
• •PD-L1 targeting increases tumor accumulation 2.5-fold in vivo.
• •MEVs exhibit high stability in GI fluids and intestinal transcytosis.
• •MiR-21 silencing restores PTEN/PDCD4 and induces ER-stress-mediated ICD.
• •Dual TME reprogramming restores potent immunity in immune-refractory PDAC.

1. Introduction

Pancreatic ductal adenocarcinoma (PDAC) remains one of the most lethal gastrointestinal (GI) malignancies, with a dismal 5-year survival rate of only ∼10%. It is characterized by a profoundly immunosuppressive tumor microenvironment (TME) that blunts antitumor immunity [1]. Despite some progress with combination chemotherapies and the introduction of immune checkpoint inhibitors, clinical outcomes have improved only marginally, primarily due to a unique constellation of molecular and microenvironmental barriers inherent to PDAC [[2], [3], [4]]. Among these, the concomitant overexpression of programmed death-ligand 1 (PD-L1) and the oncogenic microRNA-21 (miR-21) plays a pivotal role. PD-L1 is abundantly expressed in both tumor and infiltrating immune cells, and its upregulation strongly correlates with poor prognosis and aggressive disease progression. Concurrently, miR-21 is one of the most frequently upregulated oncogenic microRNAs in PDAC, exerting pleiotropic pro-tumorigenic effects by promoting tumor cell proliferation, invasion, metastasis, and chemoresistance while inhibiting apoptosis [5,6]. Mechanistically, miR-21 downregulates multiple tumor suppressor genes—phosphatase and tensin homolog (PTEN), programmed cell death protein 4 (PDCD4), and sprouty homolog 2 (SPRY2), which normally restrain oncogenic signaling and facilitate immunogenic cell death (ICD) [[7], [8], [9]]. The concomitant overexpression of PD-L1 and miR-21 synergistically reinforces an immune-tolerant, treatment-refractory TME, forming a central axis of immune evasion and malignant progression in PDAC. Notably, disrupting this PD-L1/miR-21 axis has been shown to restore antitumor immunity and re-sensitize tumors to immunotherapy; for instance, miR-21 inhibition enhances the efficacy of PD-L1 blockade and suppresses tumor growth in vivo [10].

Current treatment strategies for PDAC rely primarily on intensive multi-agent chemotherapy regimens such as FOLFIRINOX or gemcitabine/nab-paclitaxel [11]. While these combinations can extend survival, their efficacy is often thwarted by PDAC's dense desmoplastic stroma and aberrant vasculature, which together restrict intratumoral drug delivery [[12], [13], [14]]. Furthermore, the invasive nature of intravenous infusion often leads to systemic toxicities and reduced patient compliance, limiting long-term maintenance therapy [15]. Moreover, PDAC is characteristically a “cold” tumor with limited effector T cell infiltration and an abundance of immunosuppressive leukocytes in its milieu [16]. Populations such as myeloid-derived suppressor cells, regulatory T cells, and tumor-associated macrophages release immunosuppressive factors (e.g. IL-10 and TGF-β) that reinforce immune tolerance and further upregulate PD-L1 in the TME, establishing redundant inhibitory circuits [17]. As a result, immune checkpoint blockade therapies (e.g. anti-PD-1/PD-L1 antibodies) have largely failed to yield meaningful clinical responses in PDAC [[18], [19], [20], [21]]. Accordingly, there is an urgent clinical unmet need for an innovative therapeutic approach that can precisely target PDAC by overcoming TME-mediated immune evasion while simultaneously addressing the intracellular oncogenic drivers. However, the effective delivery of nucleic acid therapeutics to PDAC remains a formidable challenge. Conventional systemic administration via intravenous injection often faces rapid renal clearance, unintended systemic toxicity, and, most critically, poor penetration into the dense, hypovascular desmoplastic stroma of PDAC, which severely limits sustained therapeutic exposure at the tumor site [22].

To overcome these fundamental pharmacokinetic barriers, we developed a paradigm-shifting oral nanotherapeutic platform by engineering milk-derived extracellular vesicles (MEVs). While synthetic nanocarriers often trigger immune recognition or lack the structural complexity to survive the harsh GI environment, MEVs represent a naturally evolved, biogenic delivery system. MEVs were chosen based on their exceptional stability in the GI tract and high oral bioavailability [[23], [24], [25]], offering a clinically scalable and naturally biocompatible alternative to synthetic lipid-based vectors. Unlike typical systemic delivery vehicles that demand invasive procedures and suffer from transient half-lives, our MEV-based platform exploits neonatal Fc receptor (FcRn)-mediated transcytosis in the gut to achieve a sustained “gut-to-tumor” delivery route [23]. This approach is particularly advantageous for PDAC, where the dense stroma demands a delivery vehicle capable of prolonged tissue exposure and repeated, non-invasive dosing to improve patient compliance and quality of life.

In this study, we transitioned from passive MEVs to active, tumor-homing modulators through a rigorous isolation and engineering workflow. Specifically, a PD-L1-binding peptide (PBP) was displayed on the surface of MEVs to enable selective penetration by PD-L1^high pancreatic cancer cells, while an anti-miR-21 oligonucleotide (α21) was encapsulated via electroporation. This engineered nanoplatform (α21@MEV_PBP) preserves the intrinsic merits of biogenic extracellular vesicles (EVs)—such as minimal immunogenicity and robust cargo protection—while adding precise navigation capabilities to bypass biological barriers and target the immunosuppressive TME.

Once administered, the α21@MEV_PBP preferentially homes to PD-L1-overexpressing tumor sites, where it binds to PDAC cells through high-affinity PBP–PD-L1 interaction. This tumor-specific binding concentrates the α21 payload at the tumor, where the internalized α21 oligonucleotides suppress the oncogenic activity of miR-21 in cancer cells. As a consequence, tumor-suppressive genes such as PTEN, PDCD4, and SPRY2 are restored, leading to attenuation of oncogenic signaling and induction of immunogenic tumor cell death accompanied by the release of danger signals [26]. In parallel, α21-mediated reprogramming in macrophages enhances phagocytic activity and promotes a pro-inflammatory, tumoricidal phenotype, thereby amplifying local immune activation. Through this integrated modulation of miR-21-driven oncogenic signaling and PD-L1-targeted cellular reprogramming, α21@MEV_PBP effectively transforms the PDAC microenvironment from immune-deserted to immune-active. Therefore, it represents an innovative therapeutic paradigm for PDAC, with the potential to overcome resistance and set a new benchmark for non-invasive cancer immunotherapy.

2. Materials and methods

2.1. Materials

1,2-Distearoyl-sn-glycero-3-phosphoethanolamine-N-[dibenzocyclooctyl(polyethylene glycol)-2000] (DSPE-PEG2K-DBCO) was purchased from Avanti Polar Lipids, Inc. The N-terminal azide–modified PD-L1–binding D-peptide (N₃-NYSKPTDRQYHF) was synthesized by Peptron Co., Ltd. A 5′-DBCO-functionalized α21-5p (DBCO-5′-TCAACATCAGTCTGATAAGCTA-3′), Cy5-labeled α21, and their backbone-modified analogs (locked nucleic acid (LNA)- and phosphorothioate (PS)-modified; underlined bases indicate LNA, italics indicate PS) were obtained from Bioneer Corp.

2.2. Isolation of MEVs

MEVs were isolated from commercially available low-fat pasteurized bovine milk. Briefly, residual milk fat was removed by sequential centrifugation at 5000 x g for 30 min and 12,000 x g for 1 h at 4 °C using a Supra R12 high-speed centrifuge (Hanil Science), followed by filtration through a 40-μm cell strainer (SPL, #93040) to yield defatted milk. The defatted milk was diluted with an equal volume of deionized water (DW) (1:1, v/v), and acetic acid (Sigma, #A6283) was added to adjust the pH to 4.6 for casein removal. The precipitated casein was separated by centrifugation at 5000 x g for 30 min using the Supra R12 high-speed centrifuge (Hanil Science) equipped with the A250-6 rotor (Hanil Science). The resulting supernatant was filtered through a 0.22-μm pore-size vacuum bottle-top filter (Millipore, #S2GPT05RE) and subsequently ultracentrifuged at 100,000 x g for 1.5 h using Optima XE-90 ultracentrifuge (Beckman Coulter) equipped with 45Ti rotor (Beckman Coulter). The supernatant was discarded, and the MEV pellet was resuspended in phosphate-buffered saline (PBS) (Fig. S1). Prior to use, MEV protein and particle concentration were determined using a Pierce™ bicinchoninic acid (BCA) Assay Kit (Thermo Fisher Scientific, #A55860) and NanoSight NS300 nanoparticle tracking analysis (NTA) instrument (Malvern Panalytical).

2.3. Nanoparticle tracking analysis

NTA of MEV species was performed using the same measurement conditions. 1 μL of MEV was added to 1 mL of DW and loaded into the NTA instrument. The camera level and gain were fixed at 10 and 8, respectively. Particle tracking was conducted three times for 1 min per run, with the detection threshold set to 10 to determine particle concentration.

2.4. Preparation and characterization of α21@MEV_PBP

Loading of α21 into MEVs via electroporation; 500 pmol of α21 was mixed with 2 x 10¹¹ particles of MEVs suspended in an electroporation buffer (at a MEV-to-electroporation buffer volume ratio of 1:4; Bio-Rad, #1652677) and pre-incubated on ice for 10 min. For electroporation, the mixture was transferred to a 0.4 cm gap GenePulser Electroporation cuvette (Bio-Rad, #1652088) and subjected to eight square wave pulses at 750 V with a pulse length of 5 ms, delivered at 5 s intervals between each pulse, using a Gene Pulser Xcell™ Electroporation System (Bio-Rad, #1652660). After electroporation, samples were incubated on ice overnight (O/N) to allow membrane recovery.

The loading capacity of α21 was calculated based on the amount of α21 successfully loaded into MEVs relative to the number of vesicle particles using the following equation:

$$Loadingcapacity=\frac{Total\alpha 21\left(pmol\right)xLoadingefficiency(\%)}{Particlenumber}=\frac{500pmolx21.35\%}{2x{10}^{11}particles}=\frac{106.75pmol}{2x{10}^{11}particles}=5.34x{10}^{-10}pmol/particle=0.534pmol/{10}^{9}particles$$

Functionalization of MEVs with PBP; DSPE-PEG2K-DBCO (5 mM in 50% EtOH, 50 nmol, 1.0 equiv.) and N-terminal azidoacetylated peptide (10 mg mL⁻¹ in nuclease-free water, 50 nmol, 1.0 equiv.) were conjugated via a copper-free click chemistry reaction at 37 °C for 1 h with shaking at 1100 rpm. After the reaction, α21-loaded MEVs were surface-functionalized with the PBP through a post-insertion process by incubation at 40 °C and 800 rpm for 2 h. Unreacted peptides and free α21 were removed by ultrafiltration using Vivaspin® 2 centrifugal concentrator (molecular weight cut-off (MWCO) 300 kDa; Sartorius, #VS0251).

2.5. Transmission electron microscopy (TEM) imaging

For morphological analysis, the structural characteristics of MEV and MEV_PBP were observed using Titan G2 Cube 60-300 High Resolution-TEM (FEI). 5 μL of MEV suspension (concentration: 1.0 x 10⁹ particles/mL) was placed onto a carbon-coated copper grid (Sigma, #930350) and incubated for 5 min at room temperature (RT). Unbound vesicles were gently removed using filter paper, followed by negative staining with 2% uranyl acetate (w/v) for 1 min. Excess stain was carefully blotted off, and the grid was rinsed with DW for 1 min and air-dried at RT O/N. The grids were imaged at an accelerating voltage of 200 kV.

2.6. PD-L1 binding, cellular uptake, and PD-L1 blocking assays

To evaluate PD-L1 binding ability, 1 x 10¹¹ particles of fluorescein isothiocyanate (FITC)-labeled MEV or MEV_PBP were incubated with 2 μg of recombinant His-tagged PD-L1 protein (Abcam, #ab167713) in 500 μL of PBS (pH 7.4) for 10 min at 4 °C with gentle rotation. Subsequently, 25 μL (1 mg) of pre-washed Dynabeads™ His-Tag Isolation & Pulldown (Thermo Fisher Scientific, #10103D) were added to the mixture and incubated at 25 °C for 10 min to capture the His-tagged PD-L1 complexes. The beads were then pulled down by placing the tube on a magnet for 2 min, and the supernatant was discarded. The beads were washed three times with binding buffer (PBS, pH 7.4) to remove unbound vesicles, and the bound MEVs were eluted in 100 μL of PBS. The fluorescence signal of FITC-labeled MEV_PBP bound to the beads was visualized using a fluorescence microscope (Thermo Fisher Scientific, EVOS M5000) under identical exposure settings.

To examine the cellular binding ability of MEV_PBP, a density of 1 × 10⁶ PD-L1-overexpressing KPC960 cells was treated with Cy5.5-labeled MEVs or MEV_PBP (3 x 10¹⁰ particles) and incubated for 30 min at 4 °C. After washing with Dulbecco's Phosphate-Buffered Saline (DPBS) three times to remove unbound vesicles, the cells were collected and analyzed by flow cytometry (Beckman Coulter). The mean fluorescence intensity (MFI) represented the level of MEV binding to PD-L1-expressing cells.

To assess cellular uptake, KPC960 or NIH3T3 cells (1 x 10⁵ cells/well) were seeded in 6-well plates, incubated O/N, and treated with Cy5.5-labeled MEV or MEV_PBP (3 x 10¹⁰ particles) for 1 h at 37 °C. For PD-L1 blocking, cells were pre-incubated with an anti-PD-L1 antibody (BioXcell, BE0101) for 1 h at 37 °C prior to vesicle treatment. After washing three times with DPBS, cells were collected and the Cy5.5-positive population was quantified by flow cytometry.

2.7. Ex vivo digestion analysis

An ex vivo GI digestion system was established using commercially available customized artificial digestive fluids. Artificial saliva (pH 6.8 ± 0.2; TMALAB, #TB0929) containing α-amylase and lysozyme, simulated gastric fluid (pH 1.5 ± 0.2; TMALAB, #TB1210) supplemented with pepsin, simulated intestinal fluid (pH 6.5 ± 0.2; TMALAB, #TB1212) containing pancreatin, and artificial bile juice (pH 8.2; TMALAB, #TB1220) supplemented with bile salts were sequentially employed. After incubation at 37 °C for 5 min in artificial saliva, 120 min in gastric juice, and 1 h in pancreatic/bile juice, the particle concentration of MEV (before and after digestion) was measured using NTA.

2.8. Cargo integrity assay after in vitro GI digestion

To evaluate whether the encapsulated α21 is protected from GI degradation, free α21 or α21@MEV_PBP containing an equivalent amount of α21 (100 pmol) was subjected to a sequential in vitro digestion protocol as described above. Undigested controls were processed in parallel under identical conditions using enzyme-free buffers. After digestion, MEV membranes were disrupted with 0.1% (v/v) Triton X-100 followed by sonication to release the encapsulated α21. All samples were mixed with 6X DNA loading dye (DyneBio Inc. A750) and resolved on a 2% (w/v) agarose gel prepared in 1X Tris-acetate-EDTA (TAE) buffer at 100 V for 30 min at RT. Gels were visualized using ChemiDoc XRS + System (Bio-Rad). Band intensity was quantified by densitometry using ImageJ software. The remaining amount of α21 was calculated as the ratio of band intensity after digestion to that of the corresponding undigested control.

2.9. Transcytosis of α21@MEV_PBP

For transepithelial transport analysis of α21@MEV_PBP, 3.5 × 10⁵ Caco-2 cells were seeded into the apical chamber of Transwell inserts (SPLInsert™, SPL) and cultured for 21 days to form a differentiated monolayer. Monolayer integrity was verified by measuring transepithelial electrical resistance (TEER) using an epithelial volt-ohmmeter (Millicell® ERS 3.0 Digital Voltohmmeter, Merck). Then, the culture medium was replaced with Hank's balanced salt solution (HBSS; pH 6.5) in the apical chamber and HBSS (pH 7.4) in the basolateral chamber. Cy5.5-labeled MEV species (MEV and MEV_PBP; 5 x 10¹⁰ particles) were added to the apical chamber, and samples from the basolateral chamber were collected at predetermined time points (2, 4, 8, and 12 h). The fluorescence intensity of the transported MEVs was monitored using an In vivo imaging system (IVIS® Lumina, PerkinElmer). MEV species that had traversed to the basolateral chamber were collected and applied to KPC960 cells cultured on 35-mm glass bottom confocal dishes (SPL). After fixation with 4% paraformaldehyde (PFA; Biosesang, #PC2031), the cells were stained with Hoechst 33342 (Thermo Fisher Scientific, #H3570) and imaged using a confocal laser scanning microscope.

2.10. Colloidal stability of α21@MEV_PBP

To assess the colloidal stability of MEV_PBP under physiological conditions, 2 x 10¹¹ particles of MEV_PBP were added and incubated in 1 mL of 10% (v/v) human serum (Sigma) diluted in PBS at 37 °C. At predetermined time points (1, 3, 6, 9, 12, and 24 h), samples were analyzed for hydrodynamic diameter and polydispersity index (PDI) using a Litesizer 500 dynamic light scattering (DLS) analyzer (Anton Paar, Austria).

2.11. The release profile of α21@MEV_PBP

To evaluate the release profile of Cy5-labeled α21 loaded into MEV_PBP, 100 μL of α21@MEV_PBP was loaded into 20 kDa MWCO Slide-A-Lyzer™ dialysis devices (Thermo Fisher, 69590). Each dialysis device was placed into a microcentrifuge tube (receiver) containing 1 mL of PBS adjusted to either pH 6.5 or pH 7.4 and incubated at 37 °C with continuous shaking at 400 rpm. At predetermined time points (0.5, 1, 2, 4, 8, 12, 24, 48, 72 h), the PBS in the receiver compartment was collected and transferred to a black 24-well plate (SPL, 33224), and Cy5 fluorescence intensity was measured using a microplate reader (Molecular Devices, SpectraMax iD5).

A standard curve was generated using two-fold serial dilution of Cy5-labeled α21 ranging from 100 to 1.56 pmol/mL, with PBS alone serving as a blank. The relationship between α21 concentration and fluorescence intensity was fitted using a log-log linear regression model, and the amount of α21 released at each time point was calculated by interpolating the corresponding fluorescence values against this standard curve. Cumulative release at each time point was calculated by summing the amounts released at all preceding intervals and expressing the total as a percentage of the initial amount of α21 loaded (100 pmol).

2.12. Cell culture

Murine colon carcinoma cell line MC38 (RRID: CVCL_B288), melanoma cell line, B16F10 (RRID: CVCL_0159, #CRL_6475), fibroblast cell line, NIH3T3 (RRID: CVCL_0594; #CRL-1658), and glioblastoma cell line, GL261 (RRID: CVCL_Y003) were obtained from the American Type Culture Collection (ATCC). These cell lines were cultured in Dulbecco's Modified Eagle's Medium (DMEM; Welgene, #LM001-05) supplemented with 10% fetal bovine serum (FBS; Gibco, #A5256701) and 1% antibiotic-antimycotic solution (Gibco, #15240062). Murine breast carcinoma cell line 4T1 (RRID: CVCL_0125; #CRL-2539) and pancreatic cancer cell line KPC960 (RRID: CVCL_1U12) were maintained in Roswell Park Memorial Institute (RPMI)-1640 medium (Welgene, #LM011-01) containing 10% FBS (Gibco, #A5256701) and 1% antibiotic-antimycotic solution (Gibco, #15240062). All cells were maintained at 37 °C in a humidified atmosphere containing 5% CO₂. All cell lines were routinely tested negative for mycoplasma contamination using a PCR-based detection kit (InvivoGen, #rep-mys-10).

To induce differentiation into bone marrow-derived macrophages (BMDMs), bone marrow cells were isolated from the hind limbs of 7-week-old male C57BL/6 mice and plated in 100-mm culture dishes O/N to remove unwanted cells. Following O/N incubation in cell growth media, suspended bone marrow cells were re-plated at a density of 3.5 × 10⁶ cells per dish and cultured for 7 days in growth media supplemented with 20 ng/ml macrophage colony-stimulating factor (M-CSF; Peprotech, #315-02). For polarization, BMDMs were treated with 20 ng/mL interleukin-4 (IL-4; Peprotech, #315-05) to induce M2 macrophages (M2φ) or with 20 ng/mL tumor necrosis factor-alpha (TNF-α; Peprotech) and 100 ng/mL lipopolysaccharide (LPS; Sigma-Aldrich, #L2637) to induce M1 macrophages (M1φ) for 24 h.

2.13. Flow cytometric detection of PD-L1 expression and western blot analysis

To evaluate the surface expression of PD-L1 on different cell lines, MC38, B16F10, NIH3T3, GL261, 4T1, and KPC960 cells were seeded into 35-mm glass-bottom dishes at a density of 3 × 105 cells per dish and incubated at 37 °C for 24 h. The following day, cells were pre-blocked with Fc Block (BD Biosciences, #553141) to prevent nonspecific antibody binding, followed by staining with an allophycocyanin (APC)-conjugated anti-PD-L1 antibody (clone 10F.9G2; BioLegend, #124312) for 15 min at 4 °C in the dark. After staining, cells were washed three times with PBS, resuspended in staining buffer, and analyzed using a CytoFLEX flow cytometer (Beckman Coulter, B53015). The MFI corresponding to PD-L1 expression was quantified using FlowJo software (version 10.0).

For EV protein analysis, equal amounts of MEV and MEV_PBP samples (2 x 10¹⁰ particles each) were used to assess the presence of EV markers and confirm purity.

For the detection of immune-related and miR-21-associated proteins, KPC960 cells were seeded in 6-well plates at a density of 1 × 10⁵ cells per well and incubated O/N at 37 °C. Cells were treated with PBS, α21 (100 pmol), MEV_PBP (2 x 10¹¹ particles), or α21@MEV_PBP (at a concentration of approximately 100 pmol α21) in serum-free medium for 24 h. After treatment, cells were lysed in radioimmunoprecipitation assay (RIPA) buffer (Thermo Fisher Scientific, #89900) supplemented with protease and phosphatase inhibitors, and total protein concentrations were determined using the Pierce™ BCA Protein Assay Kit (Thermo Fisher Scientific, #A55860).

To examine miR-21-related signaling in polarized macrophages, M2φ were treated with PBS, α21 (100 pmol), MEV_PBP (2 x 10¹¹ particles), or α21@MEV_PBP (at 100 pmol α21) in serum-free medium for 24 h. Protein lysates were subsequently collected to analyze miR-21-related signaling proteins.

Cell lysates (20 μg protein per lane) and EV samples (2 x 10¹⁰ particles per lane) were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and transferred onto nitrocellulose membranes. The membranes were blocked with 3% (w/v) bovine serum albumin (BSA) in Tris-buffered saline containing 0.1% Tween-20 (TBS-T) for 1 h at RT, followed by O/N incubation at 4 °C with primary antibodies. After washing with TBS-T, membranes were incubated with horseradish peroxidase (HRP)-conjugated secondary antibodies (1:2000) for 1 h at RT. Protein bands were visualized using an enhanced chemiluminescence substrate (Bio-Rad) and imaged with an iBright Imaging System (Invitrogen). The antibodies used in this study were as follows: CD9 (CST, #98327T, 1:1000), CD81 (CST, #56039, 1:1000), tumor susceptibility gene 101 (TSG101) (CST, #28283, 1:1000), Hsp70 (CST, #4872T, 1:1000) and calnexin (Abcam, #ab22595, 1:1000) as EV and non-EV markers; PTEN (CST, #9959, 1:1000), PDCD4 (Abcam, #ab51495, 1:1000), phospho-Akt (CST, #9271S, 1:500), total-Akt (CST, #9272, 1:1000), LC3B (CST, #2775, 1:1000), SPRY2 (CST, #14954, 1:1000), RAS (CST, #33197, 1:1000), phospho-p44/42 MAPK (Erk1/2) (CST, #9101, 1:1000), total-p44/42 MAPK (Erk1/2) (CST, #9102, 1:1000), phospho-eukaryotic initiation factor 2α (eIF2α) (CST, #9721, 1:1000), eIF2α (CST, #9722, 1:1000), activating transcription factor 4 (ATF4) (Santa Cruz Biotechnology, #sc-390063, 1:500), C/EBP homologous protein (CHOP) (Santa Cruz Biotechnology, #sc-7351, 1:500), inducible nitric oxide synthase (iNOS) (CST, #13120, 1:1000), Arginase-1 (CST, #93668S, 1:1000), CD206 (Abcam, #ab64693, 1:1000), signal transducer and activator of transcription 3 (STAT3) (CST, #4904, 1:1000), phospho-STAT3 (CST, #9145, 1:1000), GAPDH (GeneTex, #GTX100118, 1:1000), anti-rabbit IgG-HRP (GeneTex, #GTX213110-01, 1:2000), and anti-mouse IgG-HRP (GeneTex, #GTX213111-01, 1:2000).

2.14. Quantitative reverse transcription polymerase chain reaction (qRT-PCR)

To quantify the expression level of miR-21, KPC960 cells (1.5 × 10⁵) were seeded into 6-well plates and incubated at 37 °C for 24 h. After stabilization, cells were treated with PBS, α21 (100 pmol), MEV_PBP (2 x 10¹¹ particles), or α21@MEV_PBP (at 100 pmol α21) in serum-free medium for 24 h. Total miRNA was extracted using the miRNeasy Mini Kit (Qiagen, #217004) according to the manufacturer's protocol. The purified miRNAs were reverse transcribed into cDNA using the Mir-X™ miRNA First-Strand Synthesis Kit (Takara Bio, #638313), and the resulting cDNA was eluted in 40 μL of RNase-free water. Quantitative real-time PCR was performed on a StepOnePlus™ Real-Time PCR System (Applied Biosystems) using TOPreal™ SYBR Green qRT-PCR Mix (Enzynomics, #RT501M). U6 small nuclear RNA served as an internal control, and relative miR-21 expression levels were determined using the 2^(-ΔΔCt) method. In addition, the relative mRNA expression levels of inflammatory cytokines (TNF-α, IL-12) in BM-derived M2φ were measured as described above with the following qRT-PCR primers. All experiments were conducted in accordance with the manufacturer's instructions. Primer sequences used in this study are listed below.

Gene	Sequence
miR-21	F: 5′-AGACTAGCTTATCAGACTGA-3′ R: 5′-GTGCAGGGTCCGAGGT-3′
U6	F: 5′-GCT TCG GCA CAT ATA CTA AAA T-3′ R: 5′-CGC TTC ACG AAT TTG CGT GTC AT-3′
TNF-α	F: 5′-CCCTCACACTCAGATCATCTTCT-3′ R: 5′-GCTACGACGTGGGCTACAG-3′
IL-12	F: 5′-TGGTGAAGACGGCCAGAGAA-3′ R: 5′-GGCACAGGGTCATCATCAAA-3′
18S	F: 5′-TTCCGATAACGAACGAGACTCT-3′ R: 5′-TGGCTGAACGCCACTTGTC-3′

(F: Forward; R: Reverse).

2.15. Rescue analysis of α21@MEV_PBP

To verify that the molecular and ICD-associated effects of α21@MEV_PBP were mediated by miR-21 inhibition, KPC960 cells were seeded in 6-well plates at a density of 1 × 10⁵ cells per well and incubated O/N at 37 °C. After stabilization, the cells were treated with α21@MEV_PBP in serum-free medium for 24 h. The cells were subsequently transfected with a miR-21 mimic using Lipofectamine™ 3000 (Thermo Fisher Scientific, #L3000001) according to the manufacturer's instructions. After 4 h of transfection, the medium was replaced with fresh complete medium, and the cells were incubated for an additional 20 h. PTEN and PDCD4 protein expression were analyzed by western blotting. To assess cell-surface calreticulin (CRT) exposure, cells were washed with PBS and incubated with an Alexa Fluor 647-conjugated anti-CRT antibody (Abcam, #ab196159) at 4 °C for 30 min. Following two additional DPBS washes, CRT fluorescence intensity was analyzed using a CytoFLEX flow cytometer (Beckman Coulter).

2.16. Damage-associated molecular pattern (DAMP) analysis

To investigate α21@MEV_PBP-induced ICD, the cell surface exposure of CRT and the extracellular release of high mobility group box 1 (HMGB1) and ATP were analyzed. For CRT detection, 1 × 10⁵ KPC960 cells were seeded in six-well plates and incubated with PBS, α21, MEV_PBP, or α21@MEV_PBP in serum-free medium for 24 h. After treatment, cells were washed with PBS and incubated with an Alexa Fluor 647-conjugated anti-CRT antibody (Abcam, #ab196159) at 4 °C for 1 h. Following two additional PBS washes, CRT fluorescence intensity was analyzed using a CytoFLEX flow cytometer (Beckman Coulter). For HMGB1 and ATP release assays, cells were treated under identical conditions for 48 h, and supernatants were collected and centrifuged at 300 × g for 10 min to remove debris. The clarified supernatants were concentrated using a 3 kDa MWCO centrifugal filter (14,000 × g, 15 min, 4 °C), mixed with 5× SDS sample buffer, and subjected to western blotting using an anti-HMGB1 antibody (Abcam, #ab18256). HMGB1 levels were normalized to total protein content determined by Coomassie blue staining (Abcam). For ATP analysis, supernatants were analyzed using an ATP Assay Kit (Promega) following the manufacturer's instructions.

2.17. In vitro phagocytosis assay

To assess the phagocytic activity of α21@MEV_PBP, BMDMs were isolated and differentiated as described in Section 2.12. BMDMs (2 × 10⁵ cells) were stained with CellTracker™ Green CMFDA (1 μM; Thermo Fisher Scientific, #C7025) for 30 min at RT and seeded into 35-mm glass-bottom dishes. The macrophages were then treated with or without α21@MEV_PBP. Separately, KPC960 pancreatic cancer cells (2.5 × 10⁵ cells) were seeded into 60-mm glass-bottom dishes and incubated at 37 °C for 24 h. After stabilization, the cells were treated with or without α21@MEV_PBP in serum-free medium for 24 h, followed by staining with CellTracker™ Deep Red (1 μM; Thermo Fisher Scientific, C34565) for 30 min at RT. Subsequently, the stained cancer cells were co-cultured with BMDMs at a cancer cell to BMDM ratio of 1:4 for 2 h. After co-culture, cells were washed three times with DPBS to prepare single-cell suspensions. Phagocytic activity was quantified by flow cytometry using a CytoFLEX flow cytometer (Beckman Coulter).

2.18. Ex vivo biodistribution of MEV_PBP

All animal experiments were conducted in accordance with the guidelines of the Institutional Animal Care and Use Committee (IACUC) of Sungkyunkwan University (approval number: SKKUIACUC2025-04-51-1). 6-week-old C57BL/6 male mice were obtained from Orient Bio. To investigate ex vivo biodistribution, MEV and MEV_PBP were labeled with Cy5.5 fluorophore. Labeling of MEVs with Cy5.5-NHS ester (BioActs, #PNS1515) was performed as previously described, one day prior to use. To prepare disrupted MEV_PBP, Cy5.5-labeled MEV_PBP was treated with 0.1% Triton X-100 to disrupt vesicle membranes, followed by sonication. The resulting disrupted MEV_PBP was used as a structurally disrupted control as shown in Fig. S15A. For time-dependent blood fluorescence analysis, mice were orally administered Cy5.5-labeled intact MEV_PBP, disrupted MEV_PBP, or free dye (adjusted to match the fluorescence intensity of the vesicle formulation). Blood samples and tumor tissues were collected at 1, 6, 12, and 24 h after oral administration, and fluorescence intensity was measured to compare systemic exposure among groups. For 24 h ex vivo biodistribution analysis, mice were anesthetized with 2.5% isoflurane. The major organs (liver, lung, spleen, heart, and kidney) and tumor tissues were excised for fluorescence imaging. The harvested organs were imaged using an IVIS Lumina Series III Imaging System (PerkinElmer). Quantitative fluorescence analysis was performed based on total radiant efficiency.

2.19. In vivo models and antitumor efficacy of orally administered α21@MEV_PBP

To investigate the antitumor efficacy of α21@MEV_PBP, subcutaneous PDAC mouse models were established using KPC960 cells. 5 × 10⁶ KPC960 cells suspended in 80 μL of cell medium were implanted subcutaneously into the right flank of male C57BL/6 mice (6 weeks old). Mice were randomly divided into four groups (n = 6 per group): PBS, α21, MEV_PBP, and α21@MEV_PBP. Beginning on day 3, treatments were administered orally every two days at a dose equivalent to 1 x 10¹² particles/mouse, for a total of five doses (days 3, 5, 7, 9, and 11). Tumor volumes and body weights were monitored every two days using digital calipers. The antitumor efficacy was evaluated by sacrificing mice one day after the final administration (day 12). Tumor tissues were dissociated into single-cell suspensions using a MACS Tumor Dissociation Kit (Miltenyi Biotec, #130-096-730) and a gentleMACS Octo Dissociator with Heaters (Miltenyi Biotec). Tumor-draining lymph nodes (TDLNs) and spleen were harvested from the same mice and mechanically dissociated by passing through a 40-μm cell strainer (SPL, #93040) in cold DPBS. Red blood cells were removed using red blood cell (RBC) lysis buffer (BioLegend, #420301). After cell counting, Fc Block (BD Biosciences, #553141) was added to prevent nonspecific antibody binding for 15 min at 4 °C. Cells were then washed three times with DPBS and resuspended for flow cytometric analysis. Flow cytometry was performed using a CytoFLEX flow cytometer (Beckman Coulter), and data were analyzed with FlowJo software (v10). CRT⁺ cancer cells were detected using Alexa Fluor 647-conjugated anti-CRT antibody (Abcam, #ab196159). Tumor-infiltrating immune cells were analyzed using the following antibodies (BioLegend): APC-anti-CD11b (clone M1/70, #101212), peridinin-chlorophyll protein (PerCP)-anti-CD206 (clone C068C2, #141715), FITC-anti-CD86 (clone GL-1, #105005), PerCP/Cy5.5-anti-CD3 (clone 17A2, #100218), FITC-anti-CD3 (clone 17A2, #100204), APC-anti-CD8 (clone 53-6.7, #100711), phycoerythrin (PE)-anti-Ki67 (clone 11F6, #151209), PerCP/Cy5.5-anti-CD44 (clone IM7, #103032), and FITC-anti-Granzyme B (GrB) (clone GB11, #515403).

2.20. Histological analysis

For histological analysis, tumor tissues were harvested from PDAC-bearing mice one day after the final oral administration of α21@MEV_PBP and fixed in 4% PFA at 4 °C O/N. All fixed samples were equilibrated in 30% sucrose (in PBS), embedded in optimal cutting temperature compound (Leica Biosystems, #HLE-01480), and cryosectioned at a thickness of 10 μm using a CM1850 cryostat (Leica Biosystems). To prevent nonspecific binding, sections were blocked with 3% BSA and 0.1% Triton X-100 in PBS for 30 min at RT. To evaluate intratumoral CD8⁺ T cell activation, the sections were incubated O/N at 4 °C with primary antibodies against CD8 (APC-anti-CD8, clone 53-6.7, #100711), and GrB (FITC-anti-Granzyme B, clone GB11, #515403). To assess macrophage polarization within TME, the sections were stained with Arginase-1 (APC-anti-Arginase-1, clone A1exF5, Invitrogen, #17-3697-82) and iNOS (Alexa Fluor 488-anti-iNOS, clone CXNFT, Invitrogen, #53-5920-82) as M2-like and M1-like markers, respectively. Nuclei were counterstained with Hoechst 33342 (Thermo Fisher Scientific, #H3570) for 10 min at RT.

A fluorescent terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) assay (Promega, #G3250) was conducted according to the manufacturer's protocol to identify apoptotic cells in tumor tissues. Fluorescence images were analyzed using ImageJ software to quantify relative fluorescence intensity compared with the PBS group.

2.21. In vivo toxicity assessment

The major organs (heart, liver, spleen, lung, and kidney) were collected from treated mice, fixed in 4% PFA, embedded in paraffin, and sectioned into 6 μm-thick slices using a rotary microtome. The tissue sections were stained with hematoxylin and eosin (H&E) to examine potential histopathological alterations and imaged using an Olympus BX51 microscope. For biochemical analyses, blood samples were collected from the abdominal vena cava into heparin vacutainer tubes (Becton Dickinson, #367871). Plasma was separated by centrifugation at 3500 x g for 20 min, and the levels of aspartate aminotransferase (AST), alanine aminotransferase (ALT), alkaline phosphatase (ALP), blood urea nitrogen (BUN), creatinine (CRE), and total protein were analyzed by DKKorea Co. (Republic of Korea).

3. Results

3.1. Characterization of PBP-functionalized MEVs

Before surface functionalization, we first evaluated the reproducibility and vesicular characteristics of MEVs isolated from independent milk batches. MEVs were prepared using acid precipitation, sequential centrifugation, filtration, and ultracentrifugation, a process designed to remove major non-vesicular milk components including casein-rich aggregates and soluble milk proteins. To assess batch-to-batch reproducibility, MEVs isolated from three independent milk batches were characterized by NTA and western blotting. NTA showed comparable size distribution profiles across Batch 1–3 (Fig. S2A). Total particle yields and particle/protein ratios were consistent across batches (Fig. S2A), further supporting the reproducibility of the isolation protocol.

Western blot analysis further showed consistent detection of the EV-associated marker CD9 and the absence of calnexin, an endoplasmic reticulum (ER)-resident protein used as a negative marker for EV preparations, across all three batches (Fig. S2B). To confirm the removal of milk-specific contaminants, casein levels were compared across whole milk, an intermediate pre-purification fraction (defatted milk), and the final MEV preparation. Marked depletion of casein-associated signals was observed in the final MEV fraction, supporting the effective removal of residual casein-rich contaminants through our purification process (Fig. S2C). These results support the reproducible preparation, vesicular characteristics, and purity of MEVs obtained using our isolation protocol.

To formulate PBP-functionalized MEVs, we sequentially applied click chemistry and post-insertion techniques. First, lipid-conjugated DBCO (DSPE-PEG2K-DBCO) and azido-conjugated PBP (azido-PBP; peptide sequence: NYSKPTDRQYHF) were reacted via click chemistry to synthesize lipid-conjugated PBP (DSPE-PEG2K-PBP). Subsequently, the DSPE-PEG2K-PBP was incubated with MEVs, facilitating the insertion of the PBP into the MEV membrane to yield the desired surface-functionalized MEVs (Fig. 1A). To confirm the structural and biochemical integrity of the MEVs following functionalization, we performed NTA, TEM imaging, and western blot analysis. NTA demonstrated that the particle size of the MEVs remained similar after PBP insertion compared to intact MEVs (Fig. 1B, Fig. S3). Corroborating the NTA results, TEM imaging also confirmed that the spherical morphology and membrane structure of the MEVs were well maintained post-PBP insertion. In addition, we performed zeta-potential analysis to physically validate the surface modification. While naïve MEV exhibited a negative surface charge (approximately −18.3 mV), a distinct positive shift to −13.2 mV was observed following PBP functionalization (Fig. S4). This change is attributed to the presence of positively charged amino acid residues (lysine (K), arginine (R), and histidine (H)) within the PBP sequence displayed on the MEV surface. Importantly, the MEV_PBP retained a net negative charge to maintain electrostatic repulsion, thereby ensuring colloidal stability and preventing aggregation. Following the Minimal Information for Studies of Extracellular Vesicles (MISEV) guidelines [[27], [28], [29]], western blot analysis was performed to verify MEV purity and integrity post-functionalization. We observed robust expression of transmembrane (CD81) and cytosolic (TSG101 and Hsp70) markers in both MEV and MEV_PBP groups, whereas these proteins were undetectable in the supernatant collected post-ultracentrifugation (MEV-free milk). The comparable band intensities between MEV and MEV_PBP (Fig. 1C) indicate that vesicular integrity was maintained throughout the PBP insertion process. Conversely, negative markers, including calnexin (cellular contaminant), were successfully depleted in the purified MEV and MEV_PBP, confirming the high purity of our isolation protocol.

Fig. 1.

Characterization of PBP-functionalized MEV (MEV_PBP). (A) Schematic illustration of the MEV_PBP preparation through sequential click chemistry and post-insertion of azide-PBP into DSPE-PEG2K-DBCO-modified MEVs. (B) NTA results showing size distribution and representative TEM images of MEV and MEV_PBP. Scale bar, 100 nm. (C) Western blot analysis of EV markers (CD81, TSG101, Hsp70) and the impurity marker (casein). (D) Fluorescence image of the bead suspensions incubated with FITC-labeled MEV or MEV_PBP; right, quantification of bead-bound FITC-positive particles (n = 6/group). (E) Optimization of MEV: PBP conjugation ratio (1:0.1–1:0.5) by flow cytometry in PD-L1-overexpressing KPC960 cells (n = 3/group). (F) NTA and TEM images of MEV_PBP following simulated GI digestion, showing preserved colloidal stability. (G) PD-L1 binding capability of MEV_PBP after exposure to simulated digestive fluids (n = 3/group). (H) Gel electrophoresis analysis of α21 stability following digestive treatment and quantification of α21 retention in MEV_PBP after simulated GI digestion. The remaining amount of α21 was quantified by ImageJ densitometry relative to the corresponding undigested control. (I) Near-infrared fluorescence imaging of Cy5.5-labeled MEV or MEV_PBP transported across a Caco-2 monolayer at 2, 4, 8, and 12 h; right, cumulative fluorescence quantification. (J) Representative fluorescence images showing Cy5.5-labeled MEV_PBP (red) targeting KPC960 cells after trans-epithelial transport; nuclei stained with Hoechst 33342 (blue). Scale bar, 50 μm. (K) Serum stability of MEV_PBP after incubation in 10% human serum for up to 24 h. Statistical significance was calculated by unpaired two-tailed t-test or one-way analysis of variance (ANOVA) with Tukey's post-hoc multiple comparisons, n. s.: not significant; **P < 0.01, ****P < 0.0001. All data are presented as mean ± standard deviation (SD). (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

Next, to confirm the specific binding affinity of the MEV_PBP for PD-L1, a pull-down assay was conducted using His-tagged PD-L1 protein. FITC-labeled MEV_PBP demonstrated a 3.97-fold higher binding capacity for PD-L1 compared to unmodified control MEVs (particle number, MEV: 193 ± 38.6; MEV_PBP: 766 ± 78.2; Fig. 1D). Prior to evaluating the cellular uptake of MEV_PBP, we screened several murine cell lines, including KPC960 (PDAC), GL261 (glioma), MC38 (colon adenocarcinoma), and NIH3T3 (fibroblast), for PD-L1 and miR-21 expression using flow cytometry and qRT-PCR, respectively. KPC960 exhibited the highest PD-L1 and miR-21 expression levels among all tested lines, closely recapitulating the dual immunosuppressive features of PDAC (Fig. S5). Based on these findings, KPC960 was selected as the representative in vitro and in vivo model for subsequent experiments. We then optimized the PBP-to-MEV ratio by evaluating different MEV: PBP reaction ratios (1:0.1, 1:0.2, and 1:0.5, w/w) for targeting efficiency toward PD-L1-overexpressing KPC960 cells. Flow cytometric analysis revealed that a 1:0.5 ratio yielded the highest cell-targeting ability (Fig. 1E), which was adopted for all subsequent studies.

To further validate whether the enhanced cellular uptake of MEV_PBP was mediated by PD-L1-specific interactions, we performed PD-L1 blocking experiments and included NIH3T3 cells as a PD-L1^low control cell line. In NIH3T3 cells with low PD-L1 expression, MEV and MEV_PBP showed comparable uptake levels of 1.97% and 2.11%, respectively, with no statistically significant difference (Fig. S6A), indicating that PBP modification alone does not specifically enhance cellular uptake. By contrast, MEV_PBP exhibited enhanced uptake in PD-L1^high KPC960 cells, increasing the positive population to 21.1% compared with 2.01% for unmodified MEVs (Fig. S6B). Importantly, this enhancement was substantially reduced to 4.36% upon PD-L1 blockade, supporting a PD-L1-mediated uptake mechanism. These findings demonstrate that PBP functionalization facilitates preferential uptake of MEVs by PD-L1^high tumor cells.

To evaluate GI stability, MEV_PBP was subjected to an ex vivo digestive system simulation. TEM imaging showed that MEV_PBP retained its spherical morphology without detectable structural disruption after digestion (Fig. 1F). Particle size slightly increased after digestion, while particle concentration remained stable (Before: 13.9 x 10⁸ particles/mL; After digestion: 13.0 x 10⁸ particles/mL), suggesting that MEV_PBP resists degradation throughout the harsh GI tract. To further verify that the PBP displayed on the MEV surface retained its PD-L1 binding ability after digestion, KPC960 cells were treated with the Cy5.5-labeled MEV_PBP, and cellular binding was measured using flow cytometry. The results confirmed that MEV_PBP preserved its PD-L1 binding capability even after digestion (Fig. 1G).

We next evaluated whether MEV_PBP protects the encapsulated α21 cargo from GI-associated degradation. α21@MEV_PBP and free α21 were subjected to the same simulated GI digestion conditions, after which MEV membranes were disrupted and the recovered α21 was analyzed by 2% agarose gel electrophoresis. No detectable α21 band was observed in the free α21 group following GI digestion, whereas α21 encapsulated within MEV_PBP retained 85.2% of its initial loading after digestion (Fig. 1H). These findings demonstrate that MEV_PBP effectively protects encapsulated α21 from enzymatic degradation under simulated GI conditions, supporting its suitability for oral delivery.

Prior to the transcytosis assay, Caco-2 monolayer integrity was verified by measuring TEER. Caco-2 monolayers were incubated with PBS, free α21, MEV_PBP, or α21@MEV_PBP for up to 24 h. TEER values remained stable throughout the incubation period and were consistently maintained within the range of 300–400 Ω cm² across all treatment groups, indicating that neither native nor engineered MEVs compromised epithelial barrier integrity (Fig. S7). These results confirm that the transcytosis assay was performed under intact monolayer conditions.

We then examined whether PBP functionalization interferes with FcRn-mediated MEV transcytosis. Cy5.5-labeled MEV and MEV_PBP were applied to the apical side of a Caco-2 cell monolayer in a trans-well system. The fluorescence intensity measured in the basolateral chamber was similar between groups (Fig. 1I), confirming that surface PBP did not compromise FcRn-mediated transcellular transport. Interestingly, when KPC960 cells were treated with the translocated MEVs collected from the basolateral chamber, the MEV_PBP-treated group exhibited a significantly stronger fluorescence intensity compared to the MEV-treated group (Fig. 1J), demonstrating that MEV_PBP retains its targeting functionality even after intestinal absorption. Finally, to evaluate stability under physiological conditions, MEV_PBP was incubated in 10% human serum for up to 24 h. Particle size and polydispersity remained unchanged throughout the incubation period (Fig. 1K and Fig. S8), confirming that MEV_PBP maintains structural stability in circulation and is suitable for oral systemic delivery.

3.2. Validation of α21@MEV_PBP-mediated ICD in KPC960 cells

Before confirming the ICD induced by miR-21 suppression in KPC960 cells, we first optimized the α21 loading process into MEV_PBP. Following a previously reported electroporation-based method for miRNA encapsulation into plasma-derived EVs [30], we systematically tested four modified electroporation conditions. The optimized parameters (square wave pulsed at 750 V for 5 ms, repeated eight times with 5 s intervals) yielded a loading efficiency of approximately 21.35% (Fig. S9A–B). Based on the loading efficiency and vesicle particle number determined by NTA, the α21 loading capacity was calculated to be 0.534 pmol per 10⁹ particles. DLS analysis confirmed that particle size and PDI remained unchanged after electroporation, indicating that vesicle integrity was preserved (Fig. S9C).

The in vitro release profile of α21 from α21@MEV_PBP was next evaluated to determine whether the encapsulated cargo is retained until cellular uptake. Because α21@MEV_PBP is expected to encounter both physiological pH and mildly acidic TME after systemic distribution, release was monitored at pH 7.4 and pH 6.5 over 72 h. Cumulative α21 release remained limited under both conditions, reaching 11.21% at pH 7.4 and 9.98% at pH 6.5, with comparable release kinetics between the two conditions (Fig. S10). These results indicate that α21 is stably retained within the vesicle across the physiological-to-tumor pH range without premature leakage, supporting that intracellular α21 delivery proceeds through vesicle internalization rather than extracellular diffusion [31].

As illustrated in Fig. 2A, α21@MEV_PBP is designed to restore tumor-suppressor function in PDAC cells by silencing oncogenic miR-21. To verify target engagement, KPC960 cells were treated with PBS, free α21, MEV_PBP, and α21@MEV_PBP. α21@MEV_PBP treatment led to a significant reduction in miR-21 expression compared with control groups (Fig. 2B), confirming efficient intracellular delivery and activity of the α21 payload.

Fig. 2.

α21@MEV_PBP suppresses oncogenic signaling and induces ER stress-mediated ICD in PD-L1^high PDAC cells. (A) Schematic illustration showing how α21@MEV_PBP restores tumor suppressor genes (PTEN, PDCD4, and SPRY2) and triggers ER stress-driven ICD in PDAC cells. (B) Quantitative RT-PCR analysis of miR-21 expression levels in KPC960 cells. All samples were normalized to U6 expression (n = 3/group). (C) Flow cytometric analysis of CRT surface exposure as an ICD marker (n = 3/group). (D) Quantification of extracellular ATP release (n = 3/group). (E) Western blot analysis of extracellular HMGB1 abundance. (F) Western blot analysis of PI3K/Akt signaling-related proteins (PTEN, pAkt, tAkt, and LC3-II) with corresponding densitometric quantitative analysis (n = 3/group). (G) Western blot analysis of RAS/ERK signaling-related proteins (PDCD4, SPRY2, pERK, RAS) with quantitative analysis. (H) Western blot analysis of ER stress-related proteins (peIF2α, ATF4, CHOP) with quantitative analysis. Statistical significance was calculated by one-way ANOVA with Tukey's post-hoc multiple comparisons, n. s.: not significant; *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. All data are presented as mean ± SD.

We next examined whether miR-21 inhibition induces ICD. Because ICD is characterized by cell-surface translocation of CRT and extracellular release of ATP and HMGB1, these hallmark ICD markers were assessed in KPC960 cells following α21@MEV_PBP treatment. Flow cytometry analysis showed that α21@MEV_PBP increased surface CRT exposure (Fig. 2C), and extracellular ATP and HMGB1 were correspondingly elevated (Fig. 2D and E), indicating that α21@MEV_PBP triggers ICD in PDAC cells.

To determine whether these responses are sequence-specific rather than attributable to the vesicle carrier or the oligonucleotide loading process, MEV_PBP was loaded with an equimolar amount of a scrambled oligonucleotide (scOligo@MEV_PBP) or a miR-21 mimic (m21@MEV_PBP) and applied under identical conditions. The scOligo@MEV_PBP did not alter miR-21 levels, whereas m21@MEV_PBP increased them by 2.64-fold (Fig. S11A). The same distinction was observed at the phenotypic level. α21@MEV_PBP increased cell-surface CRT exposure from 4.9% in control cells to 13.5%, whereas neither scOligo@MEV_PBP (5.0%) nor m21@MEV_PBP (3.4%) induced CRT exposure above the control level (Fig. S11B). These results indicate that both miR-21 suppression and the ensuing ICD-associated response require the specific α21 sequence and are not elicited by the MEV_PBP carrier or the electroporation-based loading procedure.

To determine whether the miR-21-PTEN/PDCD4-ER stress axis we previously characterized for a PD-L1-targeted peptide conjugate [26] is engaged in PDAC cells following oral delivery, we examined tumor suppressor gene expression downstream of miR-21 inhibition. Western blot analysis showed that α21@MEV_PBP treatment markedly restored PTEN, a direct target of miR-21 and a critical negative regulator of the phosphoinositide 3-kinase (PI3K)/Akt pathway (Fig. 2F). Restoration of PTEN expression resulted in a pronounced decrease in phosphorylated Akt (pAKT) without affecting total Akt (tAKT), confirming suppression of PI3K/Akt oncogenic signaling. This signaling attenuation was accompanied by a significant upregulation of the autophagy marker LC3-II, suggesting that α21@MEV_PBP simultaneously induces autophagic activity, a known amplifier of ICD responses.

The suppression of miR-21 also restored the expression of other key miR-21 targets, SPRY2 and PDCD4 (Fig. 2G). Both act as negative regulators of the RAS/ERK signaling cascade, and their re-expression corresponded with a marked decrease in phosphorylated ERK (pERK) and RAS levels. This cumulative inhibition of PI3K/Akt and RAS/ERK signaling axes imposed strong cellular stress, leading to ER stress. Indeed, western blot analysis showed a significant upregulation of the protein kinase RNA-like endoplasmic reticulum kinase (PERK)/eIF2α/ATF4/CHOP pathway components, confirming ER stress-mediated pro-apoptotic signaling (Fig. 2H).

To determine whether these effects are causally mediated by miR-21 inhibition rather than merely associated with it, KPC960 cells treated with α21@MEV_PBP were subsequently transfected with a miR-21 mimic to restore intracellular miR-21 activity. α21@MEV_PBP increased PTEN and PDCD4 expression by 2.5-fold and 2.2-fold relative to control, respectively, and miR-21 reintroduction partially reversed these increases to 1.53-fold and 1.52-fold (Fig. S12A). A corresponding reversal was observed at the phenotypic level: α21@MEV_PBP raised cell-surface CRT exposure from 1.21% in control cells to 12.1%, whereas subsequent miR-21 reintroduction reduced it to 2.12%, approaching the control level (Fig. S12B). Transfection of the miR-21 mimic alone did not induce CRT exposure (1.07%), excluding a nonspecific contribution of the transfection procedure. Together, these results demonstrate that both the restoration of tumor-suppressive signaling and the induction of ICD-associated responses are causally dependent on miR-21 inhibition.

3.3. α21@MEV_PBP-mediated reprogramming of tumor-associated macrophages and restoration of macrophage-mediated antitumor immunity

Tumor-associated macrophages (TAMs) are a critical component of the immunosuppressive TME, with the M2-like subtype predominating in the PDAC milieu [32]. M2φ exhibit high levels of both miR-21 and PD-L1, which together reinforce their immunosuppressive phenotype [32,33]. miR-21 suppresses tumor-suppressive genes such as PTEN and PDCD4 while sustaining STAT3 signaling, thereby promoting M2 polarization [34]. Elevated miR-21 expression also enhances IL-10 secretion and arginase-1 activity, contributing to tumor progression and immune evasion [35]. In parallel, upregulated PD-L1 facilitates immune tolerance by dampening cytotoxic T cell activation within the TME.

By targeting PD-L1, α21@MEV_PBP enables selective delivery of α21 into macrophages via PD-L1-mediated endocytosis, resulting in intracellular miR-21 suppression. Consequently, PTEN and PDCD4 expressions are restored, while STAT3 phosphorylation is inhibited, effectively reprogramming TAMs from an M2-like immunosuppressive state to a pro-inflammatory M1-like phenotype (Fig. 3A). Based on this rationale, we hypothesized that α21@MEV_PBP could reinvigorate macrophage-mediated antitumor immunity through (1) targeting PD-L1-overexpressing TAMs to deliver α21 and suppress miR-21-driven M2 polarization, and (2) reprogramming TAMs toward an M1-like phenotype, thereby enhancing cytokine secretion, phagocytic activity, and T cell-mediated antitumor responses within the TME.

Fig. 3.

α21@MEV_PBP enhances macrophage phagocytic activity via the miR-21/STAT3 axis. (A) Schematic illustration depicting the immunomodulatory mechanism of α21@MEV_PBP, which silences miR-21 and delivers α21 selectively to PD-L1^high TAMs, leading to restoration of PTEN, PDCD4, and suppression of pSTAT3 signaling. (B) qRT-PCR analysis of miR-21 expression levels in M2-polarized macrophages treated with α21@MEV_PBP (n = 4/group). (C) Representative confocal microscopy images of macrophages showing M1 marker (iNOS, green) and M2 marker (Arginase-1, red) expression after treatment. Nuclei were stained with Hoechst 33342 (blue). Scale bar, 100 μm. (D) Western blot analysis of PTEN, PDCD4, and phosphorylated STAT3 (pSTAT3) in M2φ following treatment with α21@MEV_PBP, confirming restoration of tumor-suppressive signaling. (E-F) qRT-PCR analysis of M1-associated cytokines (TNF-α and IL-12) mRNA expression (n = 3/group). (G) Flow cytometric analysis of macrophage phagocytosis. Macrophages were pretreated with α21@MEV_PBP or untreated for 24 h, followed by co-culture with untreated or α21@MEV_PBP-treated KPC960 cells for an additional 2 h. (H) Quantification of phagocytic activity (%), defined as the percentage of macrophages that engulfed cancer cells (n = 3/group). Statistical significance was calculated by unpaired two-tailed t-test, one-way ANOVA with Tukey's post-hoc multiple comparisons, or two-way ANOVA with Sidak's multiple comparisons. n. s.: not significant; *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001, ^##P < 0.01. All data are presented as mean ± SD. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

To validate this hypothesis, we first confirmed that miR-21 is enriched in M2φ. BMDMs were differentiated into M2-like cells, verified by increased expression of the M2 markers CD206 and arginase-1 (Fig. S13A). qRT-PCR analysis revealed markedly elevated miR-21 levels in M2φ compared with unpolarized M0 macrophages (M0φ) and M1 controls (Fig. S13B). Following α21@MEV_PBP treatment, miR-21 expression was significantly reduced (Fig. 3B). As expected, fluorescence confocal imaging revealed a clear phenotypic shift in macrophages following α21@MEV_PBP treatment. Specifically, the expression of the M2-associated marker arginase-1 was markedly reduced, whereas the expression of the M1-associated marker iNOS was substantially increased (Fig. 3C and Fig. S14). Consistent with this finding, western blot analysis demonstrated restoration of PTEN and PDCD4 expression accompanied by reduced STAT3 phosphorylation (Fig. 3D), confirming that α21@MEV_PBP effectively counteracts the miR-21-mediated M2-polarizing pathway. Functionally, α21@MEV_PBP significantly upregulated the expression of pro-inflammatory cytokines TNF-α and IL-12 (Fig. 3E and F), confirming transcriptional and functional reprogramming of TAMs toward an M1-like phenotype.

We next investigated whether this phenotypic and molecular reprogramming translated into enhanced phagocytic activity. Flow cytometry-based phagocytosis assays were performed using co-cultures of macrophages and KPC960 PDAC cells under four different conditions. The basal phagocytic activity of untreated M2φ against KPC960 cells was approximately 2.93%. Pre-treatment of macrophages with α21@MEV_PBP increased phagocytic activity to 6.48%, demonstrating effective reactivation of their intrinsic antitumor function. Interestingly, when untreated M2φ were co-cultured with α21@MEV_PBP-treated KPC960 cells, which release higher levels of DAMPs such as HMGB1 and ATP, phagocytic activity further increased to 7.40%. This observation suggests that α21@MEV_PBP treatment in tumor cells not only induces ICD but also amplifies macrophage recognition through DAMP-mediated activation. The highest phagocytic rate (9.01%) was observed when both macrophages and tumor cells were treated with α21@MEV_PBP, although the increase relative to tumor cell treatment alone did not reach statistical significance, indicating a trend toward additional enhancement when both compartments are targeted (Fig. 3G and H). Collectively, these findings demonstrate that α21@MEV_PBP simultaneously silences miR-21 and inhibits STAT3 signaling while inducing ICD-associated DAMP release from tumor cells. This coordinated dual action reactivates macrophage phagocytic function and remodels the PDAC TME from an immunosuppressive to an immune-active state.

3.4. In vivo tumor targeting, antitumor efficacy, and safety of α21@MEV_PBP

To determine whether PD-L1 functionalization enhances tumor-specific delivery after oral administration, we first investigated the in vivo biodistribution and tumor-targeting capability of MEV_PBP in KPC960 tumor-bearing mice. Cy5.5-labeled MEVs and MEV_PBP were orally administered, and whole-body fluorescence was monitored using an IVIS at various time points. As shown in Fig. 4A, both vesicle types exhibited gradual signal appearance in the abdominal region, consistent with GI uptake and systemic distribution. However, MEV_PBP showed substantially stronger and more persistent fluorescence in the tumor region compared with unmodified MEVs, suggesting PD-L1-dependent tropism toward PDAC lesions.

Fig. 4.

Tumor-targeted oral delivery and therapeutic efficacy of α21@MEV_PBP in a KPC960 pancreatic tumor syngeneic model. (A) Representative in vivo fluorescence images showing the biodistribution of Cy5.5-labeled MEV and MEV_PBP in KPC960 tumor-bearing mice at predetermined time points following oral administration. (B)Ex vivo fluorescence images of major organs (liver, lung, spleen, kidney, heart, and tumor) harvested 24 h after oral administration of Cy5.5-labeled MEV or MEV_PBP in KPC960 tumor-bearing mice (n = 4). (C) Representative ex vivo images of excised tumors and corresponding quantitative ROI analysis of PBS, Cy5.5 free dye, Cy5.5-labeled MEV and MEV_PBP (n = 4). (D) Experimental schedule for in vivo evaluation of α21@MEV_PBP therapeutic efficacy following 5 times oral administration. (E) Tumor growth curves of KPC960 tumor-bearing mice treated with PBS, free α21, MEV_PBP, and α21@MEV_PBP. (F) Tumor growth curves of individual mice in each group. (G) Representative images of excised tumors at the end of therapy. (H) Representative TUNEL-stained tumor sections from each group (left) and quantification of the TUNEL⁺ area (%) (right; n = 4). Nuclei were stained with Hoechst (blue), and apoptotic cells are shown in green (TUNEL⁺). Scale bar: 100 μm. Statistical significance was calculated by one-way ANOVA with Tukey's post-hoc multiple comparisons, or two-way ANOVA with Sidak's multiple comparisons. n. s.: not significant; ***P < 0.001, ****P < 0.0001. All data are presented as mean ± SD. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

Ex vivo fluorescence imaging of excised organs further confirmed this preferential tumor accumulation of MEV_PBP (Fig. 4B). Although MEV and MEV_PBP signals were detected in the liver and spleen, the primary biodistribution and clearance organs for nanoparticle formulations, the MEV_PBP group accumulated at significantly higher levels in tumors. Quantitative region-of-interest (ROI) analysis showed that tumor-associated fluorescence intensity in the MEV_PBP group was approximately 2.5-fold higher compared with unmodified MEVs (Fig. 4C), demonstrating that PD-L1 surface targeting markedly improves the tumor-homing capacity of orally delivered MEVs.

We next asked whether the tumor-associated signal required an intact vesicular structure, using a PBP-engineered disrupted MEV control (Fig. S15A). Disrupted MEV_PBP showed minimal tumor accumulation compared with intact MEV_PBP, indicating that the fluorescence detected in tumors was not attributable to degraded vesicle components (Fig. S15B–C). Consistently, time-dependent blood fluorescence analysis following oral administration of Cy5.5-labeled intact MEV_PBP, disrupted MEV_PBP, or free dye demonstrated that vesicle integrity is critical for efficient systemic bioavailability. Intact MEV_PBP maintained significantly higher blood fluorescence at all measured time points (1, 6, 12, and 24 h), exhibiting sustained systemic exposure with maximal levels between 6 and 12 h before gradually declining by 24 h due to systemic clearance and tissue distribution (Fig. S16). In contrast, signals from free dye and disrupted MEV_PBP remained substantially lower throughout the observation period. Together, these findings support a sequential delivery process in which intact MEV_PBP undergoes intestinal transcytosis, enters systemic circulation, and subsequently selectively accumulates in PD-L1-overexpressing tumors, validating its suitability as an α21 delivery vehicle for therapeutic evaluation.

Having confirmed tumor-selective biodistribution, we next assessed the in vivo antitumor efficacy of α21@MEV_PBP according to the dosing schedule shown in Fig. 4D. Tumor-bearing mice received five oral administrations of PBS, free α21, MEV_PBP, or α21@MEV_PBP at two-day intervals, and tumor growth and body weight were monitored throughout the study. As shown in Fig. 4E and F, all KPC960 tumor-bearing mice orally administered α21@MEV_PBP exhibited a marked attenuation in tumor progression, displaying significantly reduced tumor volumes beginning on day 9 and reaching a 76.5% reduction relative to the PBS group by day 11. This therapeutic effect was not observed in mice treated with free α21 or MEV_PBP alone, indicating that tumor suppression was driven by the α21 payload rather than by the vesicle scaffold or surface ligand alone. Consistently, the excised tumors from the α21@MEV_PBP group were consistently smaller than those from the control groups (Fig. 4G). The TUNEL assay revealed markedly increased apoptosis in the α21@MEV_PBP-administered group, with a TUNEL-positive area of 11.23% compared with PBS (0.51%), free α21 (0.37%), and MEV_PBP (2.97%), corresponding to a 22.1-fold increase relative to the PBS control (Fig. 4H).

To evaluate the contribution of PBP functionalization to the therapeutic efficacy, α21-loaded unmodified MEVs (α21@MEV) were included as a non-targeted control formulation and administered on the same schedule (Fig. S17A). α21@MEV_PBP produced significantly greater tumor growth inhibition than α21@MEV, as demonstrated by tumor growth curves, individual growth profiles, and endpoint tumor weights (Fig. S17B–D). Consistent with the enhanced tumor accumulation observed for PBP-functionalized MEVs (Fig. 4A–C), these findings support that PBP functionalization improves the antitumor efficacy of orally delivered α21-loaded MEVs, likely through enhanced PD-L1-mediated tumor targeting.

Importantly, the average body weight of all groups increased by 5.67% on the 11th day compared with day 3 (the day of the first administration; Fig. 5A), suggesting that oral administration did not induce overt systemic toxicity. To further evaluate biosafety, serum biochemical analyses were performed. Levels of key hepatic markers (AST, ALT, and ALP) and renal function indicators (BUN and CRE) in the α21@MEV_PBP-treated mice were comparable to those in the PBS control group, with no statistically significant deviations (Fig. 5B and C). Additionally, H&E-stained tissue sections from five major organs, including the liver, lung, spleen, kidneys, and heart, showed no evidence of organ toxicity (Fig. 5D). These findings confirm the absence of liver or kidney toxicity and support the high biocompatibility of the formulation following repeated oral dosing. Together, these data demonstrate that α21@MEV_PBP exerts a significant antitumor effect following non-invasive oral administration, achieving tumor growth suppression while maintaining systemic safety. This in vivo efficacy strongly supports the therapeutic potential of miR-21 silencing combined with PD-L1–targeted delivery in treatment-refractory PDAC.

Fig. 5.

In vivo biocompatibility of orally administered α21@MEV_PBP. (A) Body weight profiles of KPC960 tumor-bearing mice during repeated oral administration of PBS, free α21, MEV_PBP, and α21@MEV_PBP. (B, C) Serum biochemical analysis of total protein, ALT, AST, ALP, BUN, and CRE levels after treatment (n = 3). (D) Representative H&E-stained images of major organs (heart, liver, spleen, lung, and kidney) harvested after treatment. Scale bars: 200 μm. Statistical significance was determined by one-way ANOVA with Tukey's post-hoc multiple comparisons; n. s.: not significant; *P < 0.05, **P < 0.01, ***P < 0.001. All data are presented as mean ± SD.

3.5. miR-21 silencing and ICD-mediated immune activation

To determine whether the antitumor efficacy of α21@MEV_PBP was driven by successful miR-21 silencing and subsequent immune activation, we analyzed tumors collected after the final treatment dose. qRT-PCR analysis confirmed a significant reduction in intratumoral miR-21 levels exclusively in the α21@MEV_PBP group, whereas free α21 or MEV_PBP alone showed no measurable effect (Fig. 6A). This selective downregulation validates that PD-L1–guided vesicle uptake is required for effective delivery of the α21 cargo in vivo. Western blot analysis further demonstrated the functional consequence of miR-21 knockdown: expression of its downstream tumor-suppressor targets PTEN, SPRY2, and PDCD4 were restored in α21@MEV_PBP-treated tumors, whereas all remained repressed in all control groups (Fig. 6B). These results indicated that α21@MEV_PBP not only reaches the tumor but also reverses miR-21-driven oncogenic signaling at the protein level.

Fig. 6.

In vivo antitumor immune response of α21@MEV_PBP in a KPC960 pancreatic tumor syngeneic model. (A) Relative miR-21 expression levels in tumor tissues analyzed by qRT-PCR after oral administration of PBS, free α21, MEV_PBP, or α21@MEV_PBP (n = 3). (B) Western blot analysis and quantification of PTEN, SPRY2, and PDCD4 in tumor lysates (n = 3). (C) Flow cytometric analysis of CRT exposure on tumor cells from each treatment group and quantification of CRT⁺ tumor cell populations (n = 3). (D) Western blot analysis and quantification of HMGB1 release in tumor lysates (n = 3). (E-F) Percentage of DC maturation (CD11c⁺CD40⁺, CD86⁺, or CD80⁺) in TDLNs and spleen. (G) Percentage of total CD3⁺CD8⁺ T cells and (H–J) activated CD8⁺ T cell subsets expressing CD44, GrB, or Ki67 in tumor tissues. (K) Representative immunofluorescence images of tumor sections stained for CD8 and GrB. Orange arrows indicate extracellular GrB. Scale bars: 100 μm. Fluorescence colocalization intensity of CD8 (red) and GrB (green) was measured across the A-B line. Statistical significance was determined by one-way ANOVA with Tukey's post-hoc multiple comparisons; n. s.: not significant; **P < 0.01, ***P < 0.001, ****P < 0.0001. All data are presented as mean ± SD. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

We next investigated whether miR-21 suppression and PTEN/PDCD4 restoration were accompanied by induction of ICD and recruitment of cytotoxic immune cells within the tumor. α21@MEV_PBP treatment induced key hallmarks of ICD, as shown by enhanced CRT surface exposure in tumor cells (Fig. 6C) and increased extracellular HMGB1 release, indicating enhanced immunogenicity of dying tumor cells (Fig. 6D).

We then assessed whether these signals were accompanied by remodeling of the antigen-presenting compartment. Analysis of TAMs revealed that α21@MEV_PBP treatment decreased the proportion of CD206⁺ macrophages from 41.1% to 16.2%, while increasing the proportion of CD86⁺ macrophages from 5.25% to 25.0% within the tumor (Fig. S18A–B). Immunofluorescence analysis further showed decreased Arginase-1 and increased iNOS signals (Fig. S18C), indicating TAM reprogramming toward a more activated, M1-like phenotype [36]. Dendritic cell (DC) maturation was then examined in tumor-draining lymph nodes (TDLNs) and the spleen to further characterize the immune responses associated with ICD [37]. In both the TDLNs and spleen, α21@MEV_PBP significantly increased the proportions of CD11c ⁺ CD40⁺, CD11c⁺CD80⁺, and CD11c⁺CD86⁺ DCs, indicating enhanced DC maturation and costimulatory capacity (Fig. 6E–F and Fig. S19-20). In parallel, activated CD8⁺ T cell populations were significantly increased in the TDLNs (Fig. S21), suggesting enhanced T cell priming following α21@MEV_PBP treatment.

Flow cytometric analysis of tumor tissues revealed a marked increase in tumor-infiltrating CD3⁺CD8⁺ cytotoxic T lymphocytes, together with significantly increased CD44⁺, GrB⁺, and Ki67⁺ CD8⁺ T cell populations (Fig. 6G–J and Fig. S22), indicating enhanced effector activation, cytolytic function, and proliferative expansion. Consistent with these findings, co-immunostaining of CD8 and GrB demonstrated increased infiltration of functionally activated cytotoxic T cells within tumors, with a strong positive correlation between CD8 and GrB signals (r = 0.7222) (Fig. 6K). Collectively, these findings support a coordinated immune mechanism in which α21@MEV_PBP suppresses miR-21, restores tumor-suppressive signaling, induces ICD-associated DAMP release, and is associated with macrophage repolarization, DC maturation, and tumor-localized cytotoxic T cell responses, consistent with conversion of the PDAC microenvironment toward a more immune-responsive state.

4. Discussion

MEVs have emerged as promising oral nanocarriers owing to their intrinsic biocompatibility, GI stability, and ability to transport therapeutic cargo across the intestinal barrier. Previous studies have established the feasibility of orally administered MEVs for the delivery of chemotherapeutics and nucleic acids, demonstrating improved GI stability, systemic bioavailability, and gut-to-tumor delivery following oral administration [23,38,39]. However, these platforms have primarily relied on the intrinsic biodistribution of MEVs or focused on improving oral absorption, with limited emphasis on active tumor targeting or therapeutic remodeling of the TME.

In contrast, α21@MEV_PBP integrates oral RNA delivery with active PD-L1-mediated targeting, thereby combining GI stability, systemic delivery, and receptor-mediated tumor targeting within a single platform. While PD-L1-targeted α21 therapy has previously demonstrated therapeutic potential using injectable peptide-conjugate systems [10,26], the present study extends this therapeutic strategy to a non-invasive oral EV platform that achieves effective tumor targeting and coordinated immune remodeling following oral administration. The superior therapeutic efficacy of α21@MEV_PBP over the non-targeted α21@MEV formulation (Fig. S17) further supports the functional contribution of PBP-mediated targeting after systemic absorption.

Beyond peptide-conjugate systems, most α21 nanocarriers, including lipid and polymeric nanoparticles, have relied on parenteral administration and primarily focused on suppressing tumor-intrinsic oncogenic signaling [40,41]. In contrast, α21@MEV_PBP simultaneously restored PTEN and PDCD4 signaling, induced ICD, promoted macrophage repolarization, enhanced DC maturation, and strengthened tumor-specific CD8⁺ T cell responses. Rather than functioning solely as an anti-miRNA carrier, the present platform coordinates molecular and immune reprogramming across multiple cellular compartments within the pancreatic TME.

Although further validation in orthotopic or genetically engineered PDAC models, together with quantitative pharmacokinetic analysis of intact α21 and longer-term safety evaluation, will be important for clinical translation, the present findings demonstrate the feasibility of combining gastrointestinally stable EV-mediated delivery with receptor-guided targeting after systemic absorption.

5. Conclusions

In this study, we developed α21@MEV_PBP, an orally administered, PD-L1-targeted MEV platform for α21 therapy in PDAC. Following oral administration, α21@MEV_PBP exhibited favorable GI stability, sustained systemic exposure, and preferential tumor accumulation, enabling efficient suppression of miR-21 and restoration of PTEN/PDCD4 signaling. This molecular and immune reprogramming was accompanied by DC maturation and enhanced cytotoxic CD8⁺ T cell responses, resulting in significant tumor suppression without detectable systemic toxicity.

Collectively, α21@MEV_PBP establishes a strategy for integrating oral administration, receptor-mediated targeting, and immunomodulatory RNA therapy within a single EV-based nanoplatform. By coupling targeted anti-miRNA delivery with coordinated remodeling of malignant and immune-cell compartments, this approach broadens the therapeutic potential of orally administered EVs for RNA-based treatment of pancreatic cancer.

CRediT authorship contribution statement

Youngri Ryu: Writing – review & editing, Writing – original draft, Visualization, Validation, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Hochung Jang: Writing – review & editing, Writing – original draft, Visualization, Validation, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Eun Hye Kim: Validation, Methodology, Investigation, Formal analysis. Junha Lee: Methodology, Investigation. Byeongmin Park: Methodology, Investigation. Yewon Cho: Investigation. Wonjun Kim: Investigation. Hong-Yeol Yoon: Supervision. Sun Hwa Kim: Supervision. Dae-Hyuk Kweon: Supervision. Man Kyu Shim: Project administration. Yoosoo Yang: Writing – review & editing, Writing – original draft, Supervision, Project administration, Funding acquisition, Conceptualization.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following is the Supplementary data to this article:

Data availability

Data will be made available on request.