Engineered extracellular vesicles enriched with the miR‐214/199a cluster enhance the efficacy of chemotherapy in ovarian cancer

Abstract

Recurrent ovarian cancer (OC) remains a major cause of mortality due to chemoresistance and metastasis. Epigenetic aberrations, particularly dysregulated microRNA (miRNA) expression, contribute to disease progression and represent a promising therapeutic target. Here, we identify the miR‐214‐3p/miR‐199a‐5p cluster as a stage‐associated, tumor‐suppressive network that is lost in recurrent and chemoresistant OC but can be elevated using engineered small extracellular vesicles enriched with this miRNA cluster (m214‐sEVs). Using a clinically relevant mouse model that recapitulates spontaneous OC relapse following platinum‐based chemotherapy, we show that m214‐sEVs are internalized by OC cells and niche fibroblasts, leading to increased intracellular levels of this cluster and suppression of key chemoresistance‐associated pathways, including through downregulation of Toll‐like receptor 4 (TLR4), β‐catenin, and the soluble N‐ethylmaleimide‐sensitive factor attachment protein receptor (SNARE) protein YKT6. m214‐sEV treatment reprograms secondary tumor‐derived sEVs toward a less prometastatic cargo profile and decreases carboplatin resistance and cell migration. Enforced YKT6 overexpression abrogates these effects, establishing YKT6 as a key downstream effector. Collectively, these findings support engineered sEVs as a translatable strategy to overcome chemoresistance and disrupt pro‐tumorigenic EV signaling in recurrent OC.

Loss of the miR‐214/199a cluster is associated with recurrence in ovarian cancer. Engineered small extracellular vesicles (m214‐sEVs) elevate miR‐214‐3p/miR‐199a‐5p in tumor cells, suppress β‐catenin, TLR4, and YKT6 signaling, reprogram tumor‐derived sEV cargo, reduce chemoresistance and migration, and enhance carboplatin efficacy and survival in recurrent ovarian cancer.

Abbreviations

BLI

bioluminescence imaging

CEC

cerebral endothelial cell

CEC‐sEV

cerebral endothelial cell–derived small extracellular vesicle

CI

confidence interval

CPZ

chlorpromazine

cryo‐EM

cryo–electron microscopy

EMT

epithelial–mesenchymal transition

FBS

fetal bovine serum

GFP

green fluorescent protein)

Gluc

Gaussia luciferase

GO

Gene Ontology

H&E

hematoxylin and eosin

HGSOC

high‐grade serous ovarian cancer

HR

hazard ratio

IC₅₀

half‐maximal inhibitory concentration

IHC

immunohistochemistry

ISH

in situ hybridization

miRNA

microRNA

MMP

matrix metalloproteinase

MVB

multivesicular body

NTA

nanoparticle tracking analysis

OC

ovarian cancer

OCSC

ovarian cancer stem cell

PBS

phosphate‐buffered saline

qRT‐PCR

quantitative reverse transcription–polymerase chain reaction

RW

recurrence week

SEM

standard error of the mean

sEV

small extracellular vesicle

SNARE

soluble N‐ethylmaleimide–sensitive factor attachment protein receptor

TCGA

The Cancer Genome Atlas

TEM

transmission electron microscopy

TLR4

tToll‐like receptor 4

TME

tumor microenvironment

α‐SMA

alpha‐smooth muscle actin

1. Introduction

Ovarian cancer (OC) is the fourth leading cause of cancer‐related mortality among women, causing over 250 000 deaths annually worldwide, with its global burden projected to increase due to late‐stage diagnosis, therapeutic resistance, and frequent disease recurrence [1, 2, 3]. OC is a highly heterogeneous disease driven by diverse cellular phenotypes that contribute to its complexity and high recurrence rate [4, 5, 6]. Recurrent OC remains largely incurable, owing to resistance to standard therapies and aggressive metastatic behavior, with second‐line treatments rarely yielding durable responses [7, 8]. Compared to primary tumors, recurrent OC is enriched in ovarian cancer stem cells (OCSCs) and cancer‐associated fibroblasts, cell populations that actively promote therapeutic resistance and poor clinical outcomes [9, 10].

MicroRNA (miRNA) dysregulation is a hallmark of epigenetic dysfunction in cancer, coordinating gene regulatory networks that govern OC progression and treatment response [11, 12]. Aberrant miRNA expression contributes to tumor heterogeneity and exhibits cell‐type‐specific and stage‐dependent patterns aligned with disease advancement [13, 14]. Oncogenic miRNAs enhance tumor aggressiveness, whereas loss of tumor‐suppressive miRNAs is frequently associated with advanced disease and poor prognosis [15]. Although some studies report miR‐214 upregulation [16, 17], approximately 45% of patients with advanced OC and 60% of those with high‐grade serous ovarian carcinoma (HGSOC) display marked downregulation of miR‐214 and its cluster partner miR‐199a [18]. Functionally, miR‐214 and miR‐199a regulate key pathways central to chemoresistance, epithelial–mesenchymal transition (EMT), and cancer stemness. miR‐214 overexpression suppresses OC growth by targeting semaphorin signaling [19], while miR‐199a‐5p limits metastatic dissemination by targeting hypoxia‐inducible factor‐2α [20]. Consistent with these reports, our prior work demonstrated that loss of the miR‐214‐3p/199a‐5p cluster in OCSCs promotes tumor progression and chemoresistance [21, 22]. Collectively, these findings establish the miR‐214/199a cluster as a clinically relevant tumor‐suppressive network whose loss contributes to OC recurrence and therapeutic failure.

Small extracellular vesicles (sEVs) are membrane‐bound nanovesicles that mediate intercellular communication by transferring bioactive cargo, including miRNAs [23, 24]. Compared with synthetic delivery systems, sEVs offer enhanced biocompatibility, efficient cellular uptake, and intrinsic cell‐targeting properties [25, 26]. Within the tumor microenvironment (TME), sEVs regulate key processes such as chemoresistance, metastasis, and immune evasion [11, 27]. While tumor‐derived sEVs often promote malignancy [28, 29], sEVs released from nonmalignant cells, such as mesenchymal stromal cells [30], normal fibroblasts [31], and cerebral endothelial cells (CECs) [13, 32], can exhibit tumor‐suppressive effects. Emerging evidence further indicates that therapeutic sEVs can reprogram the release and molecular composition of secondary EVs from tumor and stromal cells, thereby amplifying their therapeutic impact [33].

Our previous work demonstrated that CEC‐derived sEVs (CEC‐sEVs) mitigate chemotherapy‐induced peripheral neuropathy and enhance the cytotoxicity of OC treatments by downregulating genes linked to neurotoxicity and tumor progression [13]. Building on these findings, the present study examines the clinical expression pattern of the miR‐214‐3p/miR‐199a‐5p cluster during OC progression and tests the hypothesis that engineered CEC‐sEVs enriched with this cluster (m214‐sEVs) can enhance chemotherapy efficacy and suppress recurrence in a clinically relevant mouse model of recurrent OC.

2. Materials and methods

2.1. Human subjects

Human OC samples were collected with approval from the Wayne State University Institutional Review Board (IRB‐20‐07‐2521) and the Karmanos Cancer Institute Institutional Review Board (IRB‐2013‐052). All samples were obtained following informed consent and subsequently de‐identified. Research involving human subjects was conducted in full compliance with institutional guidelines and actively reviewed by the respective IRBs. The study methodologies conformed to the standards set by the Declaration of Helsinki. The experiments were undertaken with the understanding and written consent of each subject.

2.2. The cancer genome atlas (TCGA) data analysis

Expression profiles of hsa‐miR‐214‐3p (MIMAT0000271) and hsa‐miR‐199a‐5p (MIMAT0000231) were obtained from TCGA ovarian serous cystadenocarcinoma dataset (TCGA‐OV) via the Genomic Data Commons portal (https://portal.gdc.cancer.gov) [34]. miRNA‐seq data generated by the British Columbia Genome Sciences Centre (BCGSC) pipeline were normalized as reads per million (RPM) and log₂‐transformed (log₂[value + 1]). Corresponding clinical data were downloaded from the same cohort, excluding samples without clinical information. A total of 487 patients with HGSOC were analyzed (Table S1).

For clinical correlation analysis, patients were stratified by histological grade (G1 + G2 vs G3 + G4), and differences in miRNA expression were compared using the Wilcoxon rank‐sum test. Statistical analyses were performed using R (v4.2.1) with the stats, car, deseq2, and ggplot2 packages [35].

For prognostic evaluation in the tumor‐free patient subset who were annotated as having no residual disease or recurrence at last follow‐up, survival analyses were performed in R using the survival, survminer, and ggplot2 packages [36]. Optimal expression cutoff points were determined by the surv_cutpoint function to dichotomize patients into high‐ and low‐expression groups. The proportional hazards assumption was tested before model fitting. Kaplan–Meier curves and log‐rank tests assessed associations between miRNA expression and overall survival, with P < 0.05 considered statistically significant.

2.3. Cell lines and culture conditions

Human CECs (ACBRI376, Cell Systems, Kirkland, WA, USA, no RRID available) were cultured and passaged in Complete Classic Medium (4Z0‐500, Cell Systems). To prepare conditioned medium for sEV isolation, the culture was switched to serum‐free medium (SF‐4Z0‐500, Cell Systems) for 48 h before collection [13, 32].

The cisplatin‐resistant human OC cell line A2780cis (RRID:CVCL_1942, Millipore‐Sigma, Burlington, MA, USA, 93 112 517) was maintained in RPMI‐1640 medium supplemented with 10% fetal bovine serum (FBS). To maintain cisplatin resistance, 1 μM cisplatin was added every two passages [37].

Human OCSC lines (no RRID available), including R182, R2615, and mCherry‐OCSC1‐F2, were established as previously described [10, 22, 38, 39, 40, 41] and cultured in RPMI‐1640 supplemented with 10% FBS, 1% MEM Non‐Essential Amino Acids (ThermoFisher, Waltham, MA, USA, 11 140 050), 1% HEPES (ThermoFisher, 15 630 080), and 1% sodium pyruvate (ThermoFisher, 11 360 070).

OVCAR3/luc cells were generated by transfecting OVCAR3 cells (RRID: CVCL_0465, ATCC HTB‐161, Manassas, VA, USA) with a luciferase expression vector (pLVX‐EF1α‐IRES‐Puro, Clontech, Mountain View, CA, USA), as previously described [13]. These cells were cultured in RPMI‐1640 with 10% FBS and selected with 1.2 μg·mL⁻¹ puromycin.

All human and mouse cell lines used in this study were authenticated within the past three years using short tandem repeat profiling. Cells were routinely tested for mycoplasma contamination via polymerase chain reaction (PCR)–based assay, and all cell lines used for the experiments reported in this manuscript were confirmed to be mycoplasma‐free.

2.4. Overexpression of miRNAs in CECs

Lentiviral constructs carrying the human miR‐214‐3p precursor (PMIR‐214‐PA‐1, System Biosciences, Palo Alto, CA, USA) or a scrambled control (PMIRH000PA‐1) were used to generate lentivirus using the Lenti‐X Packaging System (Takara, San Jose, CA, USA) [42]. CECs were infected with the viral supernatant, and stably transduced cells were selected using 1.1 μg·mL⁻¹ puromycin. Puromycin‐resistant CECs were expanded and used for downstream experiments [32, 42].

2.5. Overexpression of YKT6 in OCSCs

Retroviral constructs carrying green fluorescent protein (GFP) fused YKT6 (GFP‐YKT6, pMRXIP‐GFP‐YKT6) and the corresponding backbone control vector were used to generate retrovirus using pUMVC (Addgene #8449, Watertown, MA, USA) and pCMV‐VSV‐G (Addgene #8454) to co‐transfect into HEK293T cells. pMRXIP‐GFP‐YKT6 was a gift from Noboru Mizushima (Addgene plasmid #116946; http://n2t.net/addgene:116946; RRID:Addgene_116 946). pUMVC (Addgene plasmid #8449; http://n2t.net/addgene:8449; RRID:Addgene_8449) and pCMV‐VSV‐G (Addgene plasmid #8454; http://n2t.net/addgene:8454; RRID:Addgene_8454) were gifts from Bob Weinberg.

OCSCs were infected with the viral supernatant and selected using puromycin at a concentration of 2 μg·mL⁻¹. Stable, GFP‐positive single‐cell clones were isolated, expanded, and maintained for downstream experiments. Puromycin concentrations were optimized for each cell type to ensure effective selection with minimal cytotoxicity.

2.6. Isolation and characterization of sEVs

sEVs were isolated from conditioned medium using differential ultracentrifugation, as previously described [11, 13, 23, 42], and were characterized following the MISEV 2018 and 2023 guidelines [11, 13, 23, 42, 43, 44]. The concentration and size distribution of CEC‐sEVs were analyzed using nanoparticle tracking analysis (NTA) with a NanoSight NS300 (Malvern, Westborough, MA, USA). The ultrastructural morphology of sEVs was assessed via transmission electron microscopy (TEM) (JEM‐1400Flash, JEOL, Akishima, Tokyo, Japan) and cryo–electron microscopy (Cryo‐EM, Talos Arctica, ThermoFisher). The expression of sEV surface markers CD63, CD81, and CD9 was evaluated using western blotting and ExoView analysis, following our previously published protocols [11, 23].

2.7. Protein profile of sEVs and bioinformatics analysis

Total proteins were extracted from isolated sEVs and analyzed by mass spectrometry‐based proteomics, as described in our previous studies [45, 46]. Raw data were processed using Proteome Discoverer 2.4 (Thermo Scientific) and further analyzed with Scaffold software (Proteome Software, Inc., Portland, OR, USA). The raw protein identification results are provided in the Table S2.

Protein enrichment analysis was conducted using the Enrichr gene enrichment analysis tool [47], and Gene Ontology (GO) analysis was performed to identify top‐ranked molecular function (MF) categories. First, the enriched proteins are ranked by the spectrum number, and their IDs are converted to the corresponding gene IDs. The gene list was input into the Enrichr website, and the Ontologies database was then applied to perform analysis.

2.8. CEC‐sEV labeling and tracking

To visualize and track CEC‐derived sEVs in vivo, CECs were transfected with a plasmid encoding a Gluc‐Lactadherin fusion construct (a gift from Dr. Takahashi) [48], which directs Gaussia luciferase (Gluc) expression to sEVs via exosomal lactadherin.

For tracking sEV distribution in OC cells, CECs were also transfected with a plasmid encoding CD63‐EGFP (pEGFP‐CD63, a gift from Paul Luzio; Addgene plasmid #62964), enabling GFP labeling of sEVs via CD63 [13, 23, 42].

Labeled sEVs were administered in vivo 2 h before tissue collection. For detection, immunofluorescent staining was performed on 8 μm cryosections using an anti‐GFP antibody (1:500, Cat# 2555, RRID: AB_10692764, Cell Signaling, Danvers, MA, USA). To determine the subcellular localization of sEVs, immunogold labeling was conducted on 80 nm ultrathin sections using the same primary antibody, followed by 10 nm gold‐conjugated streptavidin [13, 23, 42].

2.9. Cell viability assays

The cytotoxic effects of sEVs on OC and OCSC cells were assessed using the MTT assay and the CellTox™ Green Cytotoxicity Assay (Promega, Madison, WI, USA, G8731), respectively [10, 13], as the latter is more sensitive for detecting cytotoxic responses in slowly proliferating OCSCs [9, 10].

For the MTT assay, OC cells were seeded in 96‐well plates and treated with sEVs and chemotherapy drugs for 72 h. After treatment, 0.5 mg·mL⁻¹ thiazolyl blue tetrazolium bromide (MTT; MilliporeSigma, Burlington, MA, USA, M2128) was added to each well and incubated for 4 h. Absorbance was measured using a microplate reader, and cell viability was calculated relative to untreated controls. The IC₅₀ was determined by linear regression analysis, representing the concentration of treatment required to inhibit 50% of cell viability.

For the CellTox assay, OCSCs were mixed with CellTox Green dye, seeded into 96‐well plates, and treated with sEVs and chemotherapy drugs. Fluorescence intensity, corresponding to the number of dead cells, was measured kinetically over 72 h using a Bio‐Rad plate reader.

2.10. Caspase‐Glo 3/7 assay

To assess apoptosis, 10 μg of protein extracted from treated OCSC‐R182 cells was diluted in a total volume of 50 μL and mixed with 50 μL of Caspase‐Glo 3/7 reagent (Promega, Madison, WI, USA). After a 1‐h incubation at room temperature, luminescence was measured using a TD 20/20 Luminometer (Turner Designs, Sunnyvale, CA, USA). Caspase 3/7 activity was calculated relative to untreated controls [9].

2.11. Transwell migration assay

OVCAR3 and OCSC1‐F2 cells were resuspended in serum‐free RPMI‐1640 medium and seeded into Transwell inserts (24‐well format) at a density of 2.5 × 10⁴ cells in 0.5 mL per insert. The lower chambers were filled with 0.75 mL of complete RPMI‐1640 medium containing 10% FBS as a chemoattractant [13].

sEVs (3 × 10⁸ particles·mL⁻¹) and chemotherapy drugs (Carboplatin, 100 μm) were added to the upper chambers and incubated for 24 h. Migrated OVCAR3 cells were detected using CellTracker Red CMTPX (ThermoFisher, C34552), while migrated OCSC1‐F2 cells were identified by mCherry fluorescence.

2.12. Animals and the in vivo OC model

All animal procedures were approved by the Henry Ford Hospital Institutional Animal Care and Use Committee (IACUC) under the following protocols: #1634 (April 16, 2018–April 15, 2021), #1278 (16 Feb 2021–15 Feb 2024), and #1303 (14 July 2021–13 July 2024).

Female BALB/c nude mice (8 weeks old) were used to establish an intraperitoneal OC xenograft model. Mice were injected i.p. with 5 × 10⁶ OVCAR3/luc cells one week before treatment initiation [13]. Tumor‐bearing mice were randomly assigned to receive carboplatin, sEVs, a combination of carboplatin and sEVs, or phosphate‐buffered saline (PBS) as a placebo control. Seven mice with recurrent OC were used for functional and therapeutic studies, and three to four mice were used for tumor tissue collection for labeling, characterization, and secondary EV isolation.

Tumor burden was monitored weekly using bioluminescence imaging (BLI) with the IVIS Spectrum 200 system (Caliper Life Sciences, Hopkinton, MA, USA) [13]. At the endpoint, the mice were sacrificed, the tumors were dissected, and the wet tumor weight was measured.

2.13. In situ hybridization (ISH) of miRNAs

To detect miRNAs in formalin‐fixed, paraffin‐embedded (FFPE) human OC tumor sections, the miRNAscope assay (ACD Bio‐Techne, Newark, CA, USA) was performed according to the manufacturer's protocol. Predesigned probes targeting hsa‐miR‐214‐3p (SR‐hsa‐miR‐214‐3p‐s1), hsa‐miR‐199a‐5p (SR‐hsa‐miR‐199a‐5p‐s1), a positive control probe for U6 snRNA (SR‐RNU6‐s1), and a scramble control probe (SR‐Scramble‐s1) were used. Tissue sections were pretreated with the RNAscope Target Retrieval Kit (322 000, ACD Bio‐Techne), followed by hydrogen peroxide (H₂O₂) and protease treatment (Pretreat Reagents 322 381). Sections were then incubated overnight with the probes. Signal development was performed using the miRNAscope HD Detection Reagents Red (324510). Stained slides were visualized and imaged using a light microscope.

To detect miRNAs in frozen xenograft OC tumor sections from nude mice, FISH was performed based on our published protocols and prior study [49]. Sections were fixed with 1‐ethyl‐3‐(3‐dimethylaminopropyl) carbodiimide (EDC) to prevent miRNA degradation. miRCURY LNA™ microRNA detection probes (Exiqon, Woburn, MA, USA) targeting hsa‐miR‐214‐3p, hsa‐miR‐199a‐5p, U6 snRNA, and a scramble control were used. Fluorescent signals were amplified using TSA PLUS Fluorescence Kits (PerkinElmer, Shelton, CT, USA), and images were acquired using a confocal microscope.

2.14. Isolation of sEVs from the OC tumor

The t‐sEVs were isolated according to the published protocol with some modifications [50, 51]. OC xenografts and tumor nodules were dissected from nude mice and rinsed thoroughly with sterile saline to remove blood and debris. Tumor tissues were then minced into small fragments (~1 mm³) on ice using a sterile razor blade and rinsed again with sterile PBS.

Approximately 0.1 g of tumor tissue was transferred into a 100‐mm Petri dish containing 15 mL of OCSC growth medium supplemented with exosome‐depleted FBS (Exo‐FBS, System Biosciences, Palo Alto, CA, USA). After 24 h of incubation, the supernatant was collected and processed for sEV isolation as described above.

2.15. Quantitative reverse transcription‐polymerase chain reaction (qRT‐PCR)

Total RNA from sEVs and OC xenografts was extracted using the miRNeasy Mini Kit (Qiagen, Germantown, MD, USA, Cat# 217084). RNA was reverse transcribed using the TaqMan MicroRNA Reverse Transcription Kit (ThermoFisher, Cat# 4427975), followed by amplification with TaqMan™ MicroRNA Assays specific for: miR‐214‐3p (MIMAT0000271, Assay ID: 002306), miR‐199a‐5p (MIMAT0000231, Assay ID: 000498), miR‐15b‐5p (MIMAT0000417, Assay ID: 000390), miR‐16‐2‐3p (MIMAT0004518, Assay ID: 002171). U6 snRNA (MIM180692, Assay ID: 001973) was used as the endogenous control. qRT‐PCR was performed using the following thermal cycling conditions: 95 °C for 20 s (initial denaturation), followed by 40 cycles of 95 °C for 1 s and 60 °C for 20 s. Relative miRNA expression levels were calculated using the 2^−ΔΔCt method, normalized to U6 [13, 42].

2.16. Western blotting

Proteins were extracted from sEVs, OC cells, or xenograft tumor tissues, and analyzed by Western blotting as previously described in our published protocols [13, 42]. Briefly, 5–10 μg proteins were loaded for SDS/PAGE. Following protein transfer (Bio‐Rad Trans‐Blot Turbo, Hercules, CA, USA), membranes were blocked in 0.2% i‐Block solution (ThermoFisher, T2015) for 1 h at room temperature. Primary antibodies were incubated with membranes overnight at 4 °C, followed by incubation with secondary antibodies for 2 h at RT. Protein bands were visualized using ECL western blotting substrates (ThermoFisher, 34 577), and band intensity was quantified using AlphaView SA software (version 3.4.0, Bio‐Techne, Minneapolis, MN, USA). The original, uncropped images of Western blots are presented in Table S3. Densitometric analyses were normalized to β‐actin for cell and tissue lysates and to CD63 for sEV lysates, as indicated. The antibodies used are listed in Table 1.

Table 1.

Antibodies used for Western blots.

Name	Supplier, catalog #	RRID	Dilution
Primary antibodies
CD63	Santa CruZ, sc‐5275	RRID:AB_627877	1:500
CD81	Abcam, ab109201	RRID:AB_10866464	1:1000
CD9	Abcam, ab223052	RRID:AB_2922392	1:1000
Calnexin	Biolegend, 699 041	–	1:1000
TLR4	Novus, NBP2‐27149	RRID:AB_2936296	1:1000
β‐catenin	Santa Cruz, SC‐1496	RRID:AB_1563968	1:1000
Caspase3, cleaved	Cell Signaling, 9661	RRID:AB_2341188	1:1000
MMP2	MilliporeSigma, MAB13405	RRID:AB_94120	1:1000
MMP9	Novus, NBP2‐41233	RRID:AB_3302960	1:1000
YKT6	Abclonal, A17083	RRID:AB_2772927	1:1000
Integrin β1	Santa Cruz, sc‐8978	RRID:AB_2130101	1:1000
β‐Actin	Abcam, ab6276	RRID:AB_2223210	1:5000
Secondary antibodies
Goat anti‐mouse IgG‐HRP	Santa Cruz, sc‐2005	RRID:AB_631736	1:2000
Goat anti‐rabbit IgG‐HRP	Santa Cruz, sc‐2004	RRID:AB_631746	1:2000

2.17. Immunohistochemistry (IHC) and semiquantitative assessment

OC tissue specimens were fixed in 10% neutral‐buffered formalin, embedded in paraffin, and sectioned at 6 μm thickness. Sections were deparaffinized in xylene and rehydrated through a graded ethanol series. Antigen retrieval was performed by heating sections in 10 mm citrate buffer (pH 6.0) using a microwave or pressure cooker. Endogenous peroxidase activity was quenched with 3% hydrogen peroxide, followed by blocking with 5% normal serum to reduce nonspecific binding.

Sections were incubated overnight at 4 °C with primary antibodies against CD44 (Abcam, Waltham, MA, USA, ab51037, RRID:AB_514798; 1:100), vimentin (Abcam, ab92547, RRID:AB_10562134, 1:100), or α‐smooth muscle actin (α‐SMA, Dako, Santa Clara, CA, USA, m0851, RRID:AB_2223500, 1:100). After washing, sections were incubated with appropriate HRP‐conjugated secondary antibodies (previously listed; 1:200), and signal detection was performed using DAB substrate, followed by hematoxylin counterstaining.

Stained sections were imaged and digitized under 20× or 40× objectives (BX40; Olympus Optical, Tokyo, Japan). Immunoreactivity was assessed independently using a semiquantitative, criterion‐based scoring system adapted from a published method [52]. Scores were assigned based on the percentage of positively stained cells as follows: 0 (none), 1 (<5%), 2 (5–20%), 3 (20–35%), 4 (35–50%), and 5 (>50%).

2.18. Statistical analysis

All experiments were performed with at least three independent biological replicates unless otherwise specified by the indicated n values. Statistical analyses were conducted using GraphPad Prism v8.2.1 (GraphPad Software, Boston, MA, USA). The following statistical tests were applied as appropriate: one‐way ANOVA with Tukey's multiple comparisons test for comparisons among more than two groups; two‐way repeated‐measures ANOVA for longitudinal comparisons of OC tumor growth between treatment groups; unpaired two‐tailed Student's t‐test for comparisons between two groups; and the Wilcoxon rank‐sum test for miRNA expression analyses in patient samples. Kaplan–Meier survival curves were compared using the log‐rank test, with hazard ratios (HRs) and 95% confidence intervals (CIs) estimated using the Cox proportional hazards model. Data are presented as mean ± standard error of the mean (SEM), and a P value <0.05 was considered statistically significant.

3. Results

3.1. Loss of the miR‐214‐3p/miR‐199a‐5p cluster correlates with chemoresistance and recurrence in OC

Our studies, together with others, have identified miR‐214‐3p and its cluster partner miR‐199a‐5p as frequently dysregulated and often reduced in advanced OC, where their loss has been linked to cancer stemness, EMT, and chemoresistance [19, 20, 21]. Analysis of TCGA dataset confirmed reductions in both miRNAs in HGSOC patients with grade 3/4 tumors (n = 425) compared to those with grade 1/2 disease (n = 62), with a more pronounced and statistically significant decrease in miR‐199a‐5p (Fig. 1A). Although the reduction in miR‐214 did not reach statistical significance, prognostic survival analysis of the tumor‐free HGSOC patient subset, defined as patients without residual disease or recurrence at last follow‐up, showed that higher expression levels of these miRNAs were strongly associated with prolonged tumor‐free survival (Fig. 1B, Table S4). Furthermore, ISH of recurrent HGSOC specimens demonstrated minimal or undetectable expression of both miRNAs in tumor epithelial cells, but robust expression in adjacent stromal compartments (Fig. S1), indicating a cell‐type‐specific loss of this tumor‐suppressive miRNA cluster during OC progression and at recurrence.

Fig. 1.

Tumor recurrence is associated with downregulation of miR‐214‐3p and miR‐199a‐5p in OC. (A) Expression levels of miR‐214‐3p and miR‐199a‐5p in HGSOC samples from the TCGA‐OV dataset stratified by histological grade (G1+G2, n = 62; G3+G4, n = 425; detailed demographics provided in Table S1). Group comparisons were performed using the Wilcoxon rank‐sum test. (B) Kaplan–Meier survival analysis of overall survival of HGSOC patients based on miRNA expression levels. Tumor‐free patients were dichotomized into high‐ (miR‐214, n = 31; miR‐199a, n = 28) and low‐expression groups (miR‐214, n = 63; miR‐199a, n = 67) using the median expression cut‐off. Solid lines represent the Kaplan–Meier survival estimates, and dashed lines indicate the 95% CIs. Survival differences between groups were evaluated using the log‐rank test, and the HR and 95% CI were estimated using the Cox proportional hazards model and are displayed in the plot (Details provided in Table S4). (C) Representative BLI of mice bearing intraperitoneal OVCAR3/luc xenografts treated with carboplatin (60 mg·kg⁻¹) at Weeks 1 (W1) and 2 (W2) post implantation. Tumor recurrence was first detected and designated as recurrence week 0 (RW0) and progressed through recurrence Weeks 4 (RW4) and 8 (RW8). The color scale reflects photon flux intensity (photons/s/cm²). Images are representative of two independent experiments (n = 20 mice per experiment) with comparable tumor growth and recurrence patterns. (D) qRT‐PCR analysis of miR‐214‐3p and miR‐199a‐5p expression in primary tumors (W2) compared to recurrent tumors (RW4). Data are presented as mean ± SEM (n = 3/group). *P < 0.05 versus primary group, determined by one‐way ANOVA.

To determine whether this pattern is recapitulated in animal models, mice bearing OVCAR3/luc xenografts [13] were treated with carboplatin (60 mg·kg⁻¹, tail‐vein on Days 7 and 14 post implantation). This dosing schedule has been widely used in dose–response animal studies in ovarian and other cancers [53, 54, 55, 56, 57] (Fig. 1C). Six weeks after treatment, tumors were undetectable by BLI (total flux <5 × 10⁵ p·s⁻¹·cm⁻²). However, approximately 40% of mice developed recurrent tumors (flux >1 × 10⁶ p·s⁻¹·cm⁻²) within eight weeks post treatment (designated as recurrent week 0, RW0; Fig. 1C), leading to 100% mortality by recurrent week 8 (RW8). Recurrent lesions exhibited markedly reduced miR‐214‐3p and miR‐199a‐5p expression compared with primary xenografts (Fig. 1D). In addition, and consistent with our previous findings [10], recurrent tumors exhibited significantly higher levels of CD44 and vimentin compared with primary tumors, indicating enrichment of stemness and EMT‐associated features, while α‐SMA expression remained unchanged (Fig. S2). Together, these findings demonstrate that miR‐214‐3p and miR‐199a‐5p are downregulated in advanced and recurrent OC in both patients and animal models.

3.2. Engineered miR‐214‐3p/miR‐199a‐5p‐enriched sEVs enhance the therapeutic efficacy of carboplatin in recurrent OC

Building on our previous findings that CEC‐derived sEVs enhance chemotherapy efficacy in animal models [13], we investigated whether engineered sEVs enriched with miR‐214‐3p/miR‐199a‐5p (m214‐sEVs) could similarly improve chemotherapy response in recurrent OC. We generated m214‐sEVs by transducing human CECs with a lentiviral vector encoding miR‐214‐3p as previously described [42]. The transduced CECs maintained typical endothelial morphology and ZO‐1 expression, indicating no detectable alteration in cell phenotype (Fig. 2A). sEVs were isolated from conditioned media and characterized following Minimal Information for Studies of Extracellular Vesicles (MISEV) 2018/2023 guidelines [43, 44]. TEM, cryo‐EM, and NTA confirmed comparable sEV morphology and size (Fig. 2B–D), while western blot and ExoView verified comparable tetraspanin enrichment (CD63, CD81, CD9) and absence of the endoplasmic reticulum marker calnexin cross m214‐sEVs, sEVs from scramble‐transfected CECs (scra‐sEVs), and naïve CEC‐sEVs (Fig. 2E,F, Fig. S3). These findings indicate that lentiviral transduction did not alter vesicle morphology or yield [42]. However, qRT‐PCR analysis showed an ~11‐fold increase of miR‐214‐3p and ~6‐fold increase of miR‐199a‐5p in m214‐sEVs compared with scra‐sEVs and CEC‐sEVs, with no change in unrelated miRNAs (miR‐15b‐5p, miR‐16‐2) (Fig. 2G). Proteomic analysis showed that the overall protein composition of m214‐sEVs was comparable to that of naïve CEC‐sEVs (Table S2). The most enriched GO terms included cadherin binding and RNA binding processes, which are functions associated with sEV‐recipient cell interactions and cargo sorting, respectively (Fig. S4, Table S5). These results confirm that m214‐sEVs selectively enrich therapeutic miRNAs without altering the core protein cargo.

Fig. 2.

Characterization of engineered CEC‐derived sEVs enriched with miR‐214‐3p. (A) Representative fluorescence images of naïve CECs (with phase image) and CECs transduced with either miR‐214‐3p (m214‐CECs) or scrambled control (scra‐CECs), showing robust GFP expression and preserved endothelial morphology, as indicated by the tight junction marker ZO1. (B, C) Representative TEM with enlarged insets of individual sEVs and cryo‐EM images show isolated sEVs from CECs (CEC‐sEVs), m214‐CECs (m214‐sEVs), and scra‐CECs (scra‐sEVs). (D) NTA results show comparable size distributions of CEC‐sEVs, m214‐sEVs, and scra‐sEVs. (E) Western blot results demonstrate the presence of tetraspanin markers (CD63, CD81, CD9) and the absence of the ER marker calnexin, indicating the purity of vesicle preparations. (F) ExoView‐based quantification of CD63⁺, CD81⁺, and CD9⁺ particle counts across EV preparations. (G) qRT‐PCR analysis of miR‐214‐3p, miR‐199a‐5p, miR‐15b‐3p, and miR‐16‐2 in the different EV samples (n = 3/group). *P < 0.05 vs. scra‐sEVs, assessed by one‐way ANOVA. Images and quantitative data in (A–F) are representative of three independent experiments showing comparable results. Error bars in F indicate SEM. Scale bars in A, 20 μm; B, 200 nm; C, 50 nm.

We next evaluated the therapeutic efficacy of m214‐sEVs in vivo. As our prior studies showed that CEC‐sEVs alone do not alter OC progression [13], we assessed the efficacy of m214‐sEVs in a recurrent OC model in combination with chemotherapy. Mice bearing recurrent OVCAR3/luc tumors were randomized to receive PBS, carboplatin alone, carboplatin plus scra‐sEVs, or carboplatin plus m214‐sEVs (Fig. 3A). Carboplatin (60 mg·kg⁻¹) was given on RW1 and RW2, and sEVs (1×10¹¹ particles, intraperitoneally, i.p.) every other day for 6 weeks. The PBS‐treated group exhibited rapid tumor progression, with all animals reaching experimental endpoint by RW6 (Fig. 3B–D). Treatment with carboplatin alone or in combination with scra‐sEVs significantly decreased tumor growth compared to PBS control, but still, all animals in these groups reached the endpoint by RW10 (Fig. 3B–D). Strikingly, mice treated with the combination of carboplatin and m214‐sEVs showed disease regression with tumor burden significantly less than the other treated groups by the RW1 timepoint (Fig. 3B,C). More importantly, the combination of carboplatin and m214‐sEVs significantly improved overall survival, with all animals surviving beyond 200 days post recurrence (Fig. 3D). These results demonstrate that m214‐sEVs significantly enhance the efficacy of carboplatin, providing sustained therapeutic benefit.

Fig. 3.

m214‐sEVs enhance carboplatin therapeutic efficacy and prolong survival in OC models. (A) Schematic illustrating the treatment schedule for the survival study. Mice bearing intraperitoneal OVCAR3/luc tumors were treated with carboplatin (Carbo, 60 mg·kg⁻¹, black arrows), and/or sEVs (scra‐sEVs or m214‐sEVs, blue arrows) beginning at the time of tumor recurrence (RW1). (B) Representative BLI of mice show tumor burden across the treatment groups (n = 7/group). (C) Quantification of BLI signal over time demonstrates reduced tumor progression in the Carbo + m214‐sEV group. *P < 0.05 vs. PBS; #P < 0.05 vs. Carbo alone, assessed by two‐way ANOVA with Tukey's post hoc test. (D) Kaplan–Meier survival analysis reveals significantly prolonged survival in mice treated with Carbo + m214‐sEVs compared to other groups. Log‐rank (Mantel–Cox) test results: P < 0.05 for Carbo vs. PBS and Carbo + scra‐sEVs vs. PBS; P < 0.001 for Carbo + m214‐sEVs vs. Carbo and Carbo + m214‐sEVs vs. PBS; non‐significant for Carbo vs. Carbo + scra‐sEVs. (E) Schematic of a second treatment regimen using low‐dose carboplatin. Mice bearing intraperitoneal OVCAR3/luc tumors received carboplatin (Carbo, 30 mg·kg⁻¹; black arrows) and/or sEVs (scra‐sEVs or m214‐sEVs; blue arrows) starting one week post implantation (W1) and were sacrificed at week 6 (W6) for endpoint analysis. (F) Representative BLI of mice from each treatment group (n=7/group). (G) Quantification of BLI signal over time showed differences in tumor progression among groups. *P < 0.05 vs. PBS; #P < 0.05 vs. Carbo 30 alone; $P < 0.05 vs. Carbo + scra‐sEVs; assessed by two‐way ANOVA with Tukey's post hoc test. (H) Endpoint quantification of tumor mass at W6. *P < 0.05 vs. PBS; determined by one‐way ANOVA with multiple comparisons. Error bars in C, G and H indicate SEM.

Although carboplatin‐based chemotherapy is effective for treating primary OC [13, 58], its dose‐dependent toxicity limits long‐term use [59]. To test whether m214‐sEVs allow chemotherapy dose reduction, mice bearing primary OVCAR3/luc xenografts were treated with low‐dose carboplatin (30 mg·kg⁻¹) plus m214‐sEVs or scra‐sEVs (Fig. 3E). Carboplatin monotherapy suppressed tumor growth, and the addition of scra‐sEVs conferred a modest additional benefit. By comparison, combining m214‐sEVs with low‐dose carboplatin produced the greatest tumor suppression (approximately 45% improvement at Week 6; Fig. 3F–H). Collectively, these findings show that engineered m214‐sEVs deliver tumor‐suppressive miRNAs that potentiate the antitumor activity of carboplatin in both recurrent and primary OC, enabling dose reduction while achieving sustained therapeutic benefit.

Because advanced HGSOC frequently develops resistance to platinum and taxane chemotherapy [60, 61], we evaluated whether m214‐sEVs could restore chemosensitivity in resistant OC cells. In cisplatin‐resistant A2780cis cells, m214‐sEVs markedly reduced the cisplatin half‐maximal inhibitory concentration (IC₅₀) at all tested doses (3×10⁷–3×10⁹ particles·mL⁻¹), whereas scra‐sEVs or naïve CEC‐sEVs had no effect (Fig. S5A). Comparable efficacy at 3×10⁸ particles·mL⁻¹ established this as the optimal in vitro dose. In patient‐derived paclitaxel‐resistant OCSCs (R182 and R2615) [9, 10], m214‐sEVs, but not scra‐sEVs, significantly enhanced paclitaxel‐induced cytotoxicity without affecting cell viability alone (Fig. S5B,C). Together, these results show that m214‐sEVs effectively sensitize chemoresistant OC and OCSC models to both platinum and taxane chemotherapy, supporting their use as a therapeutic adjunct for treatment‐resistant OC.

3.3. m214‐sEVs elevate miR‐214‐3p/miR‐199a‐5p and reduce toll‐like receptor 4 (TLR4) and β‐catenin in OC

To determine how m214‐sEVs sensitize OC to chemotherapy, we first examined their tumor uptake in vivo. Mice bearing recurrent OVCAR3/luc tumor at Week 4 post relapse (Fig. 4A) were injected i.p. with GFP‐labeled m214‐sEVs (GFP‐m214‐sEVs), which were readily detected within tumor tissues by confocal microscopy and immunogold EM, confirming internalization by both tumor cells and adjacent fibroblasts (Fig. 4B–E). GFP signal was not detectable in tumors from mice treated with unlabeled m214‐sEVs, confirming the specificity of the GFP signal (Fig. 4B). In mice bearing mCherry‐labeled OCSC1‐F2 tumors, biodistribution imaging using Gluc‐tagged m214‐sEVs (Gluc‐m214‐sEVs) further revealed selective accumulation in mCherry⁺ tumor nodules with minimal uptake in other organs (Fig. 4F,G). In vitro, blocking clathrin‐mediated endocytosis with chlorpromazine (CPZ) at the minimal cytotoxic dose (5 μg·mL⁻¹) [62] abolished the sensitizing effect of m214‐sEVs, indicating that clathrin‐dependent uptake is required for their function (Fig. 4H). These results show that m214‐sEVs are internalized by OC tumors and that this is required for their function.

Fig. 4.

Biodistribution and intratumoral uptake of m214‐sEVs increase miR‐214‐3p/miR‐199a‐5p levels and suppress target gene expression in recurrent OC. (A) Schematic illustrating the experimental timeline for sEV uptake. Mice bearing relapsed OC at relapse week 4 (RW4) were intraperitoneally injected with either GFP‐labeled m214‐sEVs or unlabeled m214‐sEVs and sacrificed 2 h post injection. (B) Immunofluorescence images of relapsed OC nodules showed strong GFP signal in tumors from mice treated with GFP‐m214‐sEVs, while no GFP signal was detected in the unlabeled control group. (C) Representative TEM image of a relapsed OC nodule, with tumor cells outlined by red dashed lines. (D) Higher‐magnification view of the boxed region in (C) revealed immunogold‐labeled GFP particles (black arrows) localized to both the nucleus (*) and cytoplasm of tumor cells. (E) Adjacent stromal region from the same nodule showed immunogold‐labeled particles (black arrows) within fibroblast cytoplasm, closely associated with collagen fibers (#). (F) Schematic illustrating the experimental design for Gluc‐m214‐sEV biodistribution analysis. Mice bearing mCherry⁺ OCSC1‐F2 xenografts were intraperitoneally injected with Gluc–tagged m214‐sEVs (Gluc‐m214‐sEVs) to enable tracking of sEV biodistribution. Tissues were harvested 2 h post injection. (G) Ex vivo BLI demonstrated Gluc signal derived from Gluc‐m214‐sEVs, demonstrating preferential accumulation of sEVs within OC tumor tissues (identified by mCherry expression), with minimal signal detected in non‐tumor organs, including liver, ovary, and intestine. (H) Real‐time CellTox Green cytotoxicity assay demonstrated the effect of pretreatment with CPZ prior to paclitaxel and/or m214‐sEVs in OCSC‐F182 cells. *P < 0.05, assessed by one‐way ANOVA with Tukey's post hoc test; n = 3 wells/group. (I) Schematic illustrating the treatment timeline for miRNA and protein analyses. Mice bearing relapsed OC were treated with carboplatin (Carbo, black arrows) and/or sEVs (blue arrows), and tumors were collected at relapse week 4 (RW4). (J) qRT‐PCR analysis of tumor tissues showed significantly elevated levels of miR‐214‐3p and miR‐199a‐5p in the combination treatment group (m214‐sEVs + Carbo) compared to all other groups (n = 3/group). *P < 0.05, one‐way ANOVA with Tukey's post hoc test. (K) Representative hematoxylin and eosin (H&E) staining and fluorescent ISH images of relapsed OC nodules revealed increased miR‐214‐3p expression (red signal) in OC cells following combination treatment with m214‐sEVs and carboplatin. (L) Representative western blots and (M) corresponding quantitative analysis of tumor lysates showed reduced expression of TLR4 and β‐catenin, and increased levels of cleaved caspase‐3 in mice receiving the combination treatment with m214‐sEVs and carboplatin (n = 3/group). *P < 0.05, assessed by one‐way ANOVA with Tukey's post hoc test. Images and quantitative data are representative of three independent experiments showing comparable results. Scale bars in B, 50 μm; C, 6 μm; D–E, 100 nm; G, 10 mm; K, 20/100 μm.

We next investigated whether m214‐sEVs modulate miR‐214‐3p and miR‐199a‐5p levels in recurrent OC tumors. qRT‐PCR of recurrent OVCAR3/luc tumors collected at Week 4 post treatment (Fig. 4I) showed that carboplatin + m214‐sEVs significantly increased miR‐214‐3p and miR‐199a‐5p expression compared with carboplatin alone or with scra‐sEVs (Fig. 4J). Fluorescent ISH (FISH) confirmed restoration of miR‐214‐3p levels within tumor tissues (Fig. 4K). These findings demonstrate that m214‐sEVs increase miR‐214‐3p/miR‐199a‐5p expression in OC following chemotherapy.

TLR4 and β‐catenin are validated targets of miR‐214‐3p [63, 64] and key mediators of chemoresistance [21, 22, 65]. Our previous studies demonstrated that miR‐199a‐5p regulates chemoresistance through the TLR4 pathway [21] and that β‐catenin is selectively upregulated in chemoresistant OCSCs [65]. We next analyzed their expression in recurrent OC. Western blot analysis of recurrent OC tumors isolated from mice carrying recurrent OVCAR3/luc tumor after 4 weeks of treatment (Fig. 4I) showed that carboplatin alone elevated both proteins, whereas the combination with m214‐sEVs markedly reduced their levels and increased cleaved caspase‐3, indicating enhanced apoptosis (Fig. 4L,M). Similarly, in paclitaxel‐resistant OCSC‐R182 cells, m214‐sEVs suppressed paclitaxel (20 μm)–induced upregulation of TLR4 and β‐catenin while enhancing caspase‐3/7 activity (Fig. S5D–F). To confirm that the therapeutic effect depends on encapsulated vesicle cargo, the protease protection assay showed that proteinase K or RNase treatment in the presence of detergent Triton X‐100 (which disrupts vesicle membranes and abolishes sEV protection) [66] abolished m214‐sEV cytotoxicity, whereas enzyme treatment alone did not (Fig. S6), demonstrating that encapsulated miRNAs, not surface molecules, mediate the effect.

Together, these data show that m214‐sEVs deliver functional miR‐214‐3p/miR‐199a‐5p to recurrent OC, downregulate their targets TLR4 and β‐catenin, and enhance apoptosis and chemosensitivity to both platinum and taxane therapies.

3.4. m214‐sEV and carboplatin combination therapy reprograms secondary tumor‐derived sEVs toward a less pro‐tumorigenic phenotype

Further ultrastructural analysis of immunogold‐labeled recurrent OVCAR3/luc tumors from mice at four weeks post relapse (Fig. 5A), treated with GFP‐m214‐sEVs in combination with carboplatin revealed abundant GFP‐positive signals within intraluminal vesicles of multivesicular bodies (MVBs) in OC cells (Fig. 5B). These findings indicate the presence of internalized m214‐sEVs within the MVB compartment, suggesting the involvement of sEV biogenesis in treated OC cells. In addition, GFP signals were also detected on vesicular structures in the adjacent extracellular space (Fig. 5C), implying that m214‐sEVs may interact with neighboring tumor niche cells. Primary tumor cell‐derived sEVs (t‐sEVs) are known to promote malignancy [67, 68]; however, emerging evidence [69, 70], including the concept proposed by Askenase [33], suggests that secondary EVs released by the tumor and its niche cells that internalize exogenous EVs can substantially shape the overall biological outcomes. To test this, we harvested tumor tissues from recurrent OC‐bearing mice 4 weeks after the initiation of the combination treatment and cultured them ex vivo for 24 h to collect secondary t‐sEVs (Fig. 5D). Using these t‐sEVs (3 × 10⁸ particles·mL⁻¹), we then treated mCherry‐labeled chemoresistant OCSC1‐F2 cells [10]. Secondary t‐sEVs from PBS‐ or carboplatin‐only–treated mice (PBS‐t‐sEVs and carbo‐t‐sEVs) significantly increased the IC₅₀ of carboplatin, suggesting that these t‐sEVs promote chemoresistance. By comparison, t‐sEVs isolated from mice that received the combination treatment (m214 + carbo‐t‐sEVs) significantly reduced the cisplatin IC₅₀, indicating a loss of their pro‐resistance function (Fig. 5E). Western blot analysis revealed that Carbo‐t‐sEVs contained elevated levels of integrin β1 and matrix metalloproteinase 9 (MMP9), both of which are associated with extracellular matrix remodeling, stemness, invasion, and chemoresistance [71, 72]. In contrast to these groups, m214+Carbo‐t‐sEVs showed significantly reduced levels of integrin β1 and MMP9 compared to both PBS‐t‐sEVs and Carbo‐t‐sEVs (Fig. 5F,G). Notably, MMP2 levels remained unchanged across all treatment groups. Consistent with the protein data, transwell assays showed that PBS‐t‐sEVs and Carbo‐t‐sEVs enhanced the migratory capacity of OCSC‐F2 cells, whereas m214+Carbo‐t‐sEVs had no such effect, indicating a loss of pro‐metastatic potential (Fig. 5H). Collectively, our findings suggest that secondary t‐sEVs generated from m214‐sEV‐recipient tumor and niche cells may transition from a pro‐tumorigenic to a less tumor‐promoting or even an antitumorigenic phenotype, thereby amplifying the therapeutic impact of m214‐sEVs.

Fig. 5.

m214‐sEVs change secondary t‐sEV composition and abrogate its pro‐metastatic effects in OC. (A) Schematic illustrating the in vivo treatment regimen and timeline for tissue collection. (B, C) Representative TEM images of relapsed OC tissue showed immunogold‐labeled GFP‐m214‐sEVs localized within MVBs (B) and in the extracellular space (C). (D) Schematic of the experimental design for secondary t‐sEV isolation. t‐sEVs were isolated from minced tumor tissues of mice treated under four different conditions (n = 3/group). (E) Quantification of carboplatin IC₅₀ values in mCherry⁺ OCSC‐F2 cells following treatment with various t‐sEVs preparations. *P < 0.05, ***P < 0.001, assessed by one‐way ANOVA with Tukey's post hoc test. (F, G) Western blot analysis (F) and corresponding quantification (G) of t‐sEV cargo proteins, including Integrin β1, MMP2, MMP9, CD63, and CD9. Quantitative data in (G) were normalized to CD63. *P < 0.05 vs PBS‐t‐sEVs; #P < 0.05 vs carbo‐t‐sEVs, assessed by one‐way ANOVA with Tukey's post hoc test. (H) Quantitative results of transwell migration assays in OCSC‐F2 cells treated with t‐sEV. *P < 0.05 vs control, assessed by one‐way ANOVA with Tukey's post hoc test. Images in A and data shown in (E–H) are representative of three independent experiments. Error bars in E, G and H indicate SEM. Scale bars in B and C, 100 nm.

3.5. YKT6 downregulation mediates the antimigratory and antitumor effects of m214‐sEV and carboplatin combination therapy in recurrent OC

We found that treatment with m214‐sEVs in combination with carboplatin significantly reduced the expression of YKT6, a high‐confidence predicted target of miR‐214‐3p identified by TargetScan [73](weighted context++ score = −0.06) (Fig. 6A), in ovarian tumors isolated from mice bearing recurrent OVCAR3/luc xenografts after four weeks of treatment (Fig. 6B,C). Consistently, proteomic profiling of OVCAR3 cells treated with m214‐sEVs revealed a downregulation of YKT6 (Table S2). YKT6 is a member of the vesicular soluble N‐ethylmaleimide‐sensitive factor attachment protein receptor (v‐SNARE) family [74], and promotes metastasis in lung and breast cancers by driving EMT and enhancing MMP activity [75, 76].

Fig. 6.

Functional role of YKT6 in cell motility and sEV cargo in OC. (A) Schematic illustrating the predicted miR‐214‐3p binding site within the 3′UTR of YKT6 mRNA. (B) Schematic of the treatment timeline showing administration of carboplatin (Carbo) and/or sEV, followed by tissue collection at relapse week 4 (RW4). (C) Representative western blot and corresponding quantification of YKT6 protein levels in recurrent OC xenografts. *P < 0.05 vs PBS, #P < 0.05 vs carbo, assessed by one‐way ANOVA with Tukey's post hoc test. (D) Representative images and quantification of transwell migration assays (E) comparing YKT6‐overexpressing (YKT6‐F2) and control vector‐transduced (vector‐F2) cells. (F) Quantification of sEV secretion per cell, measured by NTA, in YKT6‐F2, vector‐F2, and F2 parental cells following the indicated treatments. (G) Quantification of mean and mode sizes of t‐sEVs from different cell types. (H) Representative TEM images and corresponding NTA profiles showing t‐sEV morphology and size distribution. (I) Representative Western blot and (J) quantification of protein cargo in three types of t‐sEVs. *P < 0.05, ***P < 0.001, ****P < 0.0001, ns = not significant; assessed by one‐way ANOVA with Tukey's post hoc test. N = 3/group as indicated. Images and quantitative data are representative of three independent experiments showing comparable results. Scale bars in D, 100 μm; H, 100 nm.

To test its role in therapy response, we generated YKT6‐overexpressing OCSC1‐F2 cells (YKT6‐F2) via retroviral transduction (Addgene #116946, pMRXIP‐GFP‐YKT6, Watertown, MA, USA). Western blot analysis confirmed successful YKT6 overexpression in YKT6‐F2 cells compared to control vector‐transduced F2 cells (vector‐F2) (Fig. S7A,B). Migration assays showed that treatment of vector‐F2 cells with the combination of m214‐sEVs and carboplatin significantly suppressed cell migration compared to either scra‐sEVs plus carboplatin or carboplatin alone (Fig. 6D,E). Importantly, this combination treatment did not reduce the migration of YKT6‐F2 cells, indicating that YKT6 overexpression abrogates the antimigratory effect (Fig. 6D,E). Notably, no significant differences in migration were observed between vector‐F2 and YKT6‐F2 cells under PBS, monotherapies (carboplatin, scra‐sEVs, m214‐sEVs), or the scra‐sEV plus carboplatin treatment (Fig. 6D,E). These findings suggest that the inhibitory effect of m214‐sEV and carboplatin combination treatment on F2 cell migration is specific and YKT6‐dependent. Together, these in vitro results support a role for YKT6 downregulation in mediating the therapeutic activity of m214–sEV–based combination treatment of recurrent OC.

Given YKT6's established role in EV biogenesis and cargo loading [77, 78], we next examined whether YKT6 overexpression influences t‐sEV production and content. NTA and TEM results revealed no significant differences in particle concentration or overall vesicle morphology among YKT6‐F2, vector‐F2, and parental OCSC1‐F2 cells (Fig. 6F–H, Fig. S7C). However, sEVs from YKT6‐F2 cells (YKT6‐t‐sEVs) exhibited significantly increased particle size (Fig. 6G) and were enriched with pro‐metastatic proteins, including integrin β1, MMP2, and MMP9, compared to control vector‐derived sEVs (Fig. 6I,J). Because YKT6 is a SNARE protein that regulates vesicular trafficking pathways influencing key cellular processes such as adhesion [79, 80], our in vitro findings, together with the observed in vivo downregulation of YKT6 by the m214‐sEV and carboplatin combination therapy, support a model in which YKT6 suppression enhances the therapeutic efficacy by limiting the production of pro‐tumorigenic t‐sEVs in recurrent OC.

4. Discussion

Recurrent OC remains incurable, primarily due to the development of chemoresistance and metastatic progression [7, 8, 81, 82]. Clinically, patients with advanced‐stage OC frequently exhibit marked downregulation of the miR‐214‐3p/miR‐199a‐5p cluster, and higher expression levels of this cluster are correlated with improved tumor‐free survival (Fig. 1). In this study, we demonstrate that engineered m214‐sEVs, enriched with miR‐214‐3p and miR‐199a‐5p, enhance chemosensitivity and suppress tumor growth in preclinical models of recurrent OC. Mechanistically, these therapeutic effects are mediated, at least in part, through coordinated downregulation of chemoresistance‐associated genes such as TLR4 and β‐catenin, and through inhibition of the SNARE protein YKT6 (Fig. 7A,B). Together, our findings establish a therapeutic paradigm in which engineered sEVs elevate clinically lost tumor‐suppressive miRNAs to reprogram chemoresistance‐associated gene networks and remodel the TME in recurrent OC.

Fig. 7.

Schematic illustrating the proposed molecular and cellular mechanisms by which m214‐sEVs exert therapeutic effects in recurrent OC. (A) Following clathrin‐mediated endocytosis, m214‐sEVs deliver miR‐214‐3p and miR‐199a‐5p into OC cells, resulting in the downregulation of chemoresistance‐associated genes, TLR4 and β‐catenin. This suppression inhibits antiapoptotic signaling pathways and elevates sensitivity to chemotherapy. (B) m214‐sEVs also downregulate YKT6, a SNARE protein involved in MVB trafficking, leading to altered biogenesis and protein cargo composition of secondary tumor‐derived sEVs (t‐sEVs), contributing to reduced tumor invasiveness and recurrence. Created in BioRender (BioRender.com/2fgsdlc).

We employed a clinically relevant OC mouse model that recapitulates relapse following first‐line platinum‐based chemotherapy (Fig. 1). Unlike models that rely on prolonged drug exposure or artificial induction of resistance, this model enables spontaneous OC recurrence and captures the complexity of intratumoral heterogeneity, including OCSCs, stromal fibroblasts, and immune cell populations [10, 83, 84, 85, 86]. Importantly, this model mirrors the stage‐dependent dysregulation of miR‐214‐3p/miR‐199a‐5p observed in patients, in which downregulation of these tumor‐suppressive miRNAs parallels disease progression (Fig. 1). ISH revealed consistent downregulation of the miR‐214‐3p/miR‐199a‐5p cluster in recurrent OC tumors and human HGSOC, while adjacent stromal cells retained expression of this cluster, highlighting a cell‐type–specific and progression‐linked pattern of miRNA loss. Treatment with m214‐sEVs in combination with carboplatin resulted in sustained tumor suppression for more than 200 days, whereas monotherapies did not block tumor recurrence (Fig. 3). In vitro, m214‐sEVs enhanced chemosensitivity across multiple chemoresistant and recurrent OC cell lines. Moreover, internalization of m214‐sEVs by tumor cells increased intracellular levels of the miR‐214‐3p/miR‐199a‐5p cluster (Fig. 4). Together, these findings indicate that miRNA dysregulation in OC is both stage‐specific and context‐dependent, and that targeted restoration of clinically lost tumor‐suppressive miRNAs via m214‐sEVs can effectively overcome chemoresistance and suppress tumor recurrence.

sEVs are increasingly recognized as promising therapeutic vehicles due to their biocompatibility and ability to deliver functional cargo to recipient cells [25, 26, 87, 88]. However, the intracellular mechanisms by which EV cargo exerts its effects on recipient cells remain incompletely understood. We selected CEC‐sEVs based on our previous findings that these vesicles enhance chemotherapy efficacy. In this study, m214‐sEVs, but not scra‐sEVs, elevated miR‐214‐3p/miR‐199a‐5p levels in tumor tissues, indicating that selective enrichment of therapeutic miRNA cargo is required to achieve targeted biological effects. Notably, m214‐sEVs suppressed the expression of known miR‐214‐3p/miR‐199a‐5p target genes, TLR4 [22, 89], β‐catenin [90, 91], and YKT6 [75, 92], all of which are key regulators of tumor progression, metastasis, and chemoresistance (Fig. 4). A gain‐of‐function experiment revealed that YKT6 overexpression in OCSCs abrogated the antimigratory effects of the m214‐sEV and carboplatin combination treatment, implicating YKT6 as a critical functional target of the therapy (Fig. 6). Moreover, suppression of TLR4 and β‐catenin sensitized OCSCs to paclitaxel (Fig. S5), although the precise mechanistic link between individual miRNA targets and responsiveness to different chemotherapeutic agents remains to be defined.

Compared to primary tumor‐derived sEVs (t‐sEVs), the contribution of secondary t‐sEVs in modulating tumor progression remains understudied [33]. The present study showed that secondary tEVs, generated by tumor and niche cells that internalized m214‐sEVs following combination treatment with m214‐sEVs and carboplatin, shift primary t‐sEVs from a pro‐tumorigenic to an antitumorigenic state, as evidenced by a marked reduction in cargo proteins such as integrin β1 and MMP9 (Fig. 5). In contrast, MMP2 levels were unchanged, consistent with its more constitutive role in matrix remodeling, whereas MMP9 is more dynamically regulated and closely associated with invasion and metastasis [93, 94, 95]. Consistent with the established role of YKT6 in endosomal membrane cycling and cargo sorting [77, 78], our data suggest that YKT6 plays a key role in determining the protein composition of t‐sEVs. Specifically, t‐sEVs derived from YKT6‐overexpressing OCSCs were enriched in pro‐metastatic proteins and could override the antimigratory effects of m214‐sEV and carboplatin combination treatment (Fig. 6). Thus, YKT6 suppression not only limits migration, but also curtails the release of pro‐metastatic secondary t‐sEVs that can shape the TME, extending the therapeutic benefit of m214‐sEVs beyond direct tumor cell targeting. Notably, selective enrichment of MMP9, but not MMP2, in t‐sEVs following YKT6 overexpression further suggest that m214‐sEV‐mediated YKT6 suppression selectively alters EV cargo sorting rather than globally suppressing MMPs. This supports the emerging concept proposed by Askenase [33] that secondary EVs, in concert with m214‐sEVs, collectively shape the biological outcome of the combination intervention.

Furthermore, m214‐sEVs were taken up by adjacent stromal fibroblasts, indicating that the combination treatment may also influence the TME by modulating nonmalignant niche cells (Fig. 4). Given that secondary EV‐mediated signaling plays a critical role in regulating immune cell responses, particularly macrophage polarization [33, 69, 70, 96, 97], we speculate that immunomodulatory mechanisms may contribute to the therapeutic efficacy of m214‐sEVs in combination with carboplatin. This potential effect on the immune compartment warrants further investigation to fully elucidate the role of EV‐mediated crosstalk in the antitumor activity of the combination therapy.

Together, these findings provide compelling evidence that m214‐sEVs, when combined with carboplatin, represent a promising therapeutic strategy to overcome chemoresistance and prevent recurrence in OC. By delivering tumor‐suppressive miRNAs and modulating both tumor‐intrinsic pathways and the composition of secondary tumor‐derived sEVs, this combination therapy targets multiple mechanisms of disease progression. Importantly, the ability of m214‐sEVs to alter the TME, including potential effects on immune modulation, further enhances their therapeutic relevance. Given their biocompatibility, targeted delivery capabilities, and scalable production, engineered sEVs represent a clinically translatable platform [98, 99, 100, 101] that may improve outcomes for patients with recurrent OC, a population with limited treatment options. Finally, cell‐type–specific miRNA dysregulation and differential intracellular handling of m214‐sEVs may underlie compartment‐specific therapeutic effects, representing an important direction for future investigation. The present study provides proof‐of‐concept that precision elevation of clinically depleted miRNA networks through engineered sEVs offers a next‐generation therapeutic modality for recurrent OC.

This study has several limitations. The m214‐sEV dose was selected based on prior data [13, 42]; future work will define dose–response relationships and optimal therapeutic windows. While ultrastructural data suggest m214‐sEV recycling through MVBs, distinguishing between secondary tEVs generated via MVB recycling and those released through alternative pathways requires further investigation. Such studies will also help determine whether miR‐214/miR‐199a are incorporated into secondary EVs to mediate downstream biological effects.

5. Conclusions

This study shows that engineered sEVs with elevated miR‐214/199a cluster enhance chemotherapy efficacy in OC by modulating tumor progression, chemoresistance, and EV cargo composition. Using a clinically relevant recurrence model, we demonstrate that m214‐sEVs reprogram secondary tumor‐derived sEVs toward a less prometastatic phenotype. These findings support engineered sEVs as a promising and translatable approach to improve therapeutic responses in advanced OC.

Conflict of interest

The authors declare no conflict of interest.

Author contributions

ZGZ and YZ conceived and designed the project; WW, AA, YQ, MW, AF, YL, MM, AK, and XSL acquired the data. WW, AA, YQ, AK, and YZ analyzed and interpreted the data; GM, MC conceptualized the study and guided the data analysis. WW, AA, GM, MC, ZGZ, and YZ wrote the paper. All authors have been involved in the final approval of the manuscript.

Supporting information

Fig. S1. In situ detection of miR‐214‐3p and miR‐199a‐5p reveals stromal enrichment in recurrent HGSOC.

Fig. S2. Immunohistochemical characterization of primary and recurrent ovarian cancer tissues.

Fig. S3. Single EV profiling using the ExoView system.

Fig. S4. Top 20 enriched pathways associated with proteins in naïve CEC‐sEVs and m214‐sEVs.

Fig. S5. m214‐sEVs sensitize chemoresistant OC cells to cisplatin and paclitaxel.

Fig. S6. Vesicle integrity is required for m214‐sEV–mediated sensitization to cisplatin.

Fig. S7. YKT6 overexpression in OCSC1‐F2 cells and TEM overview of derived t‐sEVs.

Table S1. Patient demographics of the study population stratified by OC grade.

Table S2. Proteomic analysis results of sEVs.

Table S3. Uncropped western blot images.

Table S4. Cox proportional hazards model for survival analysis.

Table S5. Protein enrichment analysis of m214‐sEVs.

Data accessibility

The data that support the findings of this study are available upon request from the corresponding authors.