EVs have been extensively studied over the past few decades, but their therapeutic applications have emerged more recently. Understanding the history of EV discovery is essential for appreciating their biological importance and for guiding the development of new therapeutic strategies. The discovery of EVs dates back to 1946, when Chargaff and West (9) identified platelet-derived particles, termed ‘platelet dust’, in the blood. This was followed by the identification of cell-derived vesicles in mouse cartilage by Anderson in 1969 (10). In 1996, Raposo et al (11) demonstrated that B-cell-derived exosomes contain MHC class II molecules and can directly induce T-cell responses. Subsequent studies revealed that T cells release EVs carrying bioactive ligands, such as Fas and TNF-related apoptosis-inducing ligand, indicating a crucial role in immune regulation. EVs perform diverse immune functions by interacting with recipient cells via their surface proteins or by delivering bioactive cargo, thereby altering the functional properties of the recipient cells (12). Unlike classical signaling mechanisms based on direct cell contact or secreted factors, EVs provide a distinct mode of intercellular communication, with the ability to transfer proteins and RNA to directly modulate cellular functions (13). These findings have highlighted the crucial importance of EVs in cell signaling and suggested new strategies for disease diagnosis and treatment.

EVs are phospholipid bilayer-enclosed particles, which are released by diverse cell types under both physiological and pathological conditions and have been detected in tissues and bodily fluids, including serum, cerebrospinal fluid, saliva and urine (14). Cells package signaling molecules, including proteins, lipids and nucleic acids, into EVs. This packaging protects the molecules from degradation and enables them to evade immune surveillance, thereby facilitating local and long-distance intercellular communication (15).

Recent advances in EV isolation and characterization have led to the discovery of a growing number of EV subtypes. Most mammalian cells, which typically measure 10–100 µm in diameter, release a heterogeneous population of EVs, along with non-vesicular extracellular nanoparticles (NPs). The overlapping size ranges of EVs and NPs makes it challenging to differentiate between them. Well-characterized EVs types include exosomes, which are generated via the endocytic pathway and are released into the extracellular space when multivesicular endosomes fuse with the plasma membrane; microvesicles, which form through the direct outward budding and fission of the plasma membrane; and apoptotic bodies, which are released from cells during programmed cell death, specifically apoptosis (16). Among these, the term exosome is the most widely used and studied. However, due to a lack of well-defined, category-specific markers, exosomes and microvesicles can typically only be definitively distinguished by confirmation of their distinct biogenesis pathways, often using techniques such as cryo-transmission electron microscopy (17).

EVs are classified into six main types based on their biogenesis, size and molecular markers. Exosomes, 30–150 nm in diameter, are formed within multivesicular vesicles (MVBs) through the endosomal pathway and are released upon MVB fusion with the plasma membrane. The main molecular markers for exosomes are tetratrans proteins (such as CD9, CD63 and CD81), tumor susceptibility gene 101 protein, alix and heat shock protein (HSP)70/90 (18). Microvesicles, 100–1,000 nm in diameter, directly bud from the plasma membrane by outward blebbing, which is dependent on cytoskeletal rearrangements and calcium influx. The main molecular markers for microvesicles are annexin A1 and A2 (19). Apoptotic body, 500–4,000 nm in diameter, are released during apoptosis as fragmented portions of the dying cell. The main molecular markers for apoptotic bodies are annexin V, phosphatidylserine and caspases (20). Migrasomes, 0.5–3.0 µm in diameter, are a subtype of EVs that are released from the elongated membranous structure at the tail of cells during migration. First reported in 2015, they were named due to their close association with cell migration. The hallmark proteins of migrasomes are tetraspanin-4 and −7, which may help cells eliminate damaged organelles such as mitochondria, promoting the release of cellular contents (including protein, mRNA and miRNA) for absorption by recipient cells (21). Large oncosomes are atypically large EVs (1–10 µm in diameter) that arise from non-apoptotic membrane blebbing. This process can be triggered through silencing of diaphanous-related formin-3 and overexpression of oncoproteins such as protein kinase B, heparin binding-epidermal growth factor and caveolin-1 (22). Compared with the large EVs, the small ectosomes (≤100 nm) has a notably higher percentage of CD9- and CD81-positive particles (23). A summary of the detailed types and classifications of EVs is provided in Table I (18–23).

EVs are secreted by cells into their surrounding environment and exist as a variety of types, including exosomes, microvesicles and apoptotic bodies. As different types of EVs perform distinct biological functions, their study is inherently dependent on the isolation and purification of specific EVs subpopulations. This section summarizes the principles, advantages and disadvantages of various EVs isolation methods (Table II) (24–35).

The choice of isolation and purification methods directly determines the yield, purity and physicochemical properties of the obtained EVs. Consequently, selecting a method that is efficient, convenient and reliable based on the characteristics of the target EVs is essential. However, common isolation methods often fail to completely eliminate soluble contaminants and non-vesicular particles from cell culture supernatants or biological fluids. Therefore, it is necessary to develop novel isolation strategies that offer high specificity, purity and scalability for large-scale production. Established methods for EVs isolation can be broadly categorized into several categories.

Ultracentrifugation remains the predominant method for EV isolation, enabling the separation of vesicles based on size and density, often through sucrose or iodixanol density gradients (36). However, this technique can introduce undesirable artifacts, including EV aggregation and morphological alterations, which may affect vesicular integrity, composition and biological activity (24,37). A further limitation is that ultracentrifugation alone may not fully eliminate contamination from non-vesicular components. To mitigate these issues, methods combining ultracentrifugation with other techniques have been developed. For example, combining ultracentrifugation with cryo-electron microscopy and immunogold labeling has been shown to preserve EVs diversity and prevent aggregation (38). Similarly, ultracentrifugation can be combined with filtration techniques, including tangential flow filtration (TFF) and microfluidic filtration, to reduce contaminants and enhance separation efficiency (39,40).

Several complementary methods are available for EV isolation, including density gradient centrifugation, size exclusion chromatography (SEC), high-performance liquid chromatography and integrated or combination strategies. SEC is particularly effective at removing protein aggregates and lipoprotein particles; however, its recovery rate is limited, typically reducing the total particle number by approximately half (41,42). Furthermore, SEC is unsuitable for initial volume reduction during EV extraction, such as in the isolation of EVs from cell culture supernatants (43). In addition, TFF is a high-throughput method for the efficient and selective separation of EVs from biological fluids. It has been used as an early purification step following the removal of cellular debris via low-speed centrifugation and filtration through 0.22-µm membranes (44). Another promising method for the rapid isolation of EVs is protein organic solvent precipitation, which effectively removes soluble protein contaminants and yields EVs of higher purity compared with TFF, demonstrating potential for clinical translation (45).

The heterogeneity of EVs encompasses subpopulations characterized by specific surface markers, enabling their selective isolation. For example, HSPs commonly found on EVs derived from cancerous or infected cells enable targeted capture. Virucine peptides have been shown to exhibit high efficiency in the isolation of HSP-bearing EVs from diverse sources, including cell culture media, plasma and urine (46). Heparan sulfate proteoglycan, a cell surface receptor involved in various biological processes, can also be used to enhance EVs purification. For example, when conditioned media from 293T cells was mixed with heparin-coated agarose beads following ultracentrifugation, an EVs recovery rate of 60% was achieved (47). This method also reduces contamination while preserving EVs-associated proteins and other biomarkers (48).

TFF operates on the principle of parallel fluid flow across a membrane to reduce clogging. This technique is superior to traditional ultracentrifugation as it effectively maintains both high separation efficiency and the biological integrity of EVs. Tangential flow for analyte capture (TFAC) is a novel method built upon TFF. It enables the selective capture of target exosomes using functionalized membranes modified with specific antibodies, integrating both the capture and elution steps into a single process. This streamlined process minimizes impurities, enables the rapid processing of small clinical samples and yields high-purity exosomes. Due to its efficiency, scalability and preservation of EV bioactivity, TFAC is emerging as a promising method for EVs isolation (31).

These advancements address challenges in the the mass production of EVs for clinical translation and help the transition from qualitative research to translational applications, thereby opening new avenues for regenerative medicine, drug delivery and liquid biopsy (44).

Flow cytometry is a key technique for analyzing and separating EVs. Following initial isolation by methods such as ultracentrifugation, it enables the quantification of EVs from various sources using fluorescent markers such as carboxyfluorescein succinimidyl ester and lipid-specific dyes. This technique is widely used to identify and characterize specific EVs subtypes. It employs antibody-fluorophore conjugates that target specific membrane antigens, allowing the detection and classification of individual particles based on their fluorescence characteristics, thereby enabling immunophenotyping and subgroup classification (41,49).

Immunological methods provide an alternative strategy for the isolation of EVs, particularly those derived from antigen-presenting cells (APCs), which release immunomodulatory EVs characterized by high levels of major histocompatibility complex II (MHCII) expression. Using immunomagnetic beads, exosomes can be efficiently enriched from cell-free supernatants, thereby accelerating the analysis of APC-derived EVs (42). In addition, microfluidic technology has been developed to capture CD41-positive exosomes on mica-coated surfaces using specific antibodies. Although this method allows the rapid and standardized isolation of EVs from various cell types, its clinical application is limited by its small-scale processing capacity. Furthermore, given the overlapping physical properties of EVs and viruses, a CD45-based immunodepletion approach has been developed to specifically remove HIV particles without affecting CD45-positive EVs (50). Finally, polymer precipitation represents a cost-effective and robust technique for large-scale EV isolation. Although its use in clinical settings is limited, it has been successfully applied to obtain substantial quantities of functional EVs from natural killer (NK) cells in vitro (51).

The type and quantity of EVs reflect the physiological and pathological states. A large body of research has focused on exploring the potential benefits and therapeutic applications of EVs in malignant tumors (Fig. 1).

EVs offer key advantages over alternative delivery systems due to their low immunogenicity and minimal cytotoxicity. As EVs can be isolated from endogenous cellular sources, they trigger negligible immune responses, making them highly suitable for drug and gene delivery in tissue repair and regenerative medicine (44). The encapsulation of therapeutic agents within EVs not only reduces systemic toxicity but also enhances bioavailability by minimizing the immune-mediated clearance of the cargo prior to it reaching target tissues. For example, in a mouse model of pancreatic cancer, small interfering RNA (siRNA) delivered via exosomes demonstrated greater efficacy and fewer side effects than was achieved using NP-based delivery systems (52). By contrast, numerous nanomedicines undergo rapid immune clearance, which limits their clinical effectiveness. Furthermore, the low immunogenicity and cytotoxicity of EVs enable large-scale use without typical toxic side effects. Studies have shown that artificial EVs injected into mice do not cause adverse reactions, highlighting their potential as therapeutic tools and providing a strong foundation for the clinical application of EV-based therapies (45).

EVs outperform liposomes in terms of drug delivery efficiency while maintaining low immunogenicity and toxicity. Although liposomes share a similar lipid bilayer structure with EVs and have been used as drug carriers for decades, they predominantly accumulate in liver and spleen macrophages, resulting in rapid clearance and reduced therapeutic efficacy. By contrast, EVs possess endogenous surface proteins that help them to evade immune detection, thereby extending their circulation time (53). For example, studies have shown that EVs loaded with porphyrins via electroporation, saponin treatment or dialysis exhibit higher loading efficiency compared with that of liposomes, highlighting the advantages of EVs as superior drug delivery carriers (54).

As aforementioned, EVs are membranous particles secreted by living cells into the extracellular space, where they play a key role in intercellular communication. Owing to their stability, broad tissue distribution and ability to cross biological barriers, EVs are emerging as promising drug delivery vehicles. They can encapsulate a diverse range of biological molecules, including microRNAs (miRNAs/miRs), siRNAs and recombinant proteins, that are otherwise difficult to deliver intracellularly without a carrier. Engineered EVs provide a targeted solution for the delivery of such molecules to specific sites (55,56).

Two primary strategies are used to direct EVs to cancer cells and enhance their accumulation at the tumor: i) Engineering the parent cells prior to EVs isolation and ii) directly modifying the membrane of purified EVs. Both approaches aim to improve the targeting specificity and efficacy of EVs-based delivery. For example, a fusion protein composed of apoptin and the EVs membrane protein CD9 was expressed in parent cells using light-responsive cleavable peptide linkers. This fusion protein was encapsulated within EVs, enabling the light-triggered release of apoptin (57). In another example, adipose-derived mesenchymal stem cells were transfected with a miR-122 plasmid to produce miR-122-enriched EVs. Intratumoral injection of these EVs increased the sensitivity of liver cancer cells to sorafenib (58).

As lipid-bound nanostructures, EVs naturally incorporate transmembrane proteins and surface glycans, including glycolipids. These native components act as natural anchoring sites for functionalizing EVs via genetic engineering or bioconjugation techniques. A wide range of bioactive molecules, including fluorescent probes, targeting peptides, therapeutic drugs, nanobodies and aptamers, can be displayed at these sites. For example, the GE11 peptide, which binds both epidermal growth factor receptor (EGFR) and platelet-derived growth factor receptor, was successfully expressed on the surface of engineered EVs derived from 293 cells, enabling targeted delivery to breast cancer cells with high levels of EGFR expression (59). In another approach, glycosylation, a common post-translational modification, was incorporated into a lysosome-associated membrane protein 2b-targeted peptide fusion protein. This modification improved the stability of the peptide and enhanced its expression in both parent cells and EVs, thereby increasing targeting efficiency toward neuroblastoma cells (60).

EVs are released by nearly all cell types and play a crucial role in disease progression and pathogenesis by transporting signaling molecules, including proteins, lipids, mRNAs, miRNAs and other bioactive substances. As they can be sampled by minimally invasive procedures, EVs show considerable potential as clinically useful biomarkers (61). Tumor development is influenced by dynamic interactions between tumor cells and immune cells within the TME. Tumor cell-derived EVs (TDEVs) contribute to the modulation of immune responses by transferring macromolecules to recipient cells, thereby acting as carriers of immune signals and contributing to immune surveillance (62). For example, TDEVs can promote immune evasion by suppressing the co-stimulation of T cells and dendritic cells (DCs), thereby conferring resistance to immune checkpoint inhibitor (ICI) therapies (63). In hepatocellular carcinoma, EVs secreted by hepatocytes deficient in the gluconeogenic enzyme fructose-1,6-bisphosphatase 1 have been shown to target infiltrating NK cells, resulting in their dysfunction and depletion, which facilitates immune escape and tumor progression through remodeling of the immune microenvironment (64,65).

Conversely, EVs can also support antitumor immunity. NK cell-derived exosomes can deliver cytotoxic agents such as perforin and granzymes, which directly induce the lysis of melanoma cells (66). In addition, EVs derived from MHCII-expressing DCs can induce antigen-specific CD4⁺ T-cell activation (67). Furthermore, it has been reported that bone marrow DC-derived EVs regulate allograft rejection in vivo, and the intravenous administration of donor DC-derived EVs has been shown to delay acute rejection in rats receiving a heart transplant (68).

Exosomes are nanocapsules enriched in MHCII molecules and secreted by B-lymphoblast-like cells. They play a crucial role in intercellular communication, particularly in processes such as tumor angiogenesis and cell differentiation. Advances in exosome production and purification have enhanced their potential as versatile platforms for drug delivery, antigen presentation and biologically targeted therapy. Consequently, immune cell-derived exosomes are now widely regarded as efficient and natural nanocarriers capable of trafficking to target cells and modulating immune responses within the TME (69). The application of immune cell-derived exosomes in immunotherapy and vaccine development has been explored under both physiological and pathological conditions (70). ICI therapy is a well-established approach in tumor immunotherapy, which significantly increases life expectancy in some patients. However, only a minority of patients achieve clinical benefit, highlighting the need for reliable predictive biomarkers (71). Programmed cell death protein 1 (PD-1)/programmed cell death-ligand 1 (PD-L1) inhibitors function by restoring the ability of suppressed T cells to recognize and kill tumor cells (71). Given the crucial role of EVs in tumor immune regulation, a combination of EVs and PD-L1 has been proposed as an effective biomarker for predicting patient response to ICI therapy. Notably, a clear dissociation has been observed between tissue-based and EV-based PD-L1 detection: while PD-L1 in tissue shows no association with survival, PD-L1 expression on circulating EVs reliably predicts survival in patients with non-small cell lung cancer (NSCLC). This dynamic, EV-based measurement offers a superior predictive model for the identification of patients who are likely to respond to ICI treatment and experience a survival benefit (72). In a study of 71 patients with metastatic melanoma (MM), circulating PD-L1⁺ and PD-1⁺ EVs were associated with responsiveness to ICI therapy. Circulating PD-1⁺ EVs, in particular, were implicated as a driving factor for resistance to anti-PD-1 therapy, suggesting a role in the stratification in patients with MM (73). These findings suggest that a comprehensive understanding of the biological roles of EVs within the TME is essential for effective cancer treatment and the development of effective cancer immunotherapies. Furthermore, the inhibition of EVs secretion pathways is currently being investigated as a therapeutic strategy (74). Combining targeted therapy that disrupt EV signaling with anti-PD-L1 therapy may potentially be more effective than anti-PD-L1 monotherapy

Historically, 2D cultures of primary cells and tumor cell lines have been widely used in cancer research and have provided valuable insights into tumor development and therapeutic mechanisms. However, 2D culture systems have inherent limitations, including an inability to recapitulate the genetic and phenotypic heterogeneity of tumors or accurately mimic the complex in vivo microenvironment (75). To more accurately model human tumor biology, 3D organoid culture systems have been developed. Organoids are self-organizing spheroidal structures derived from cancer stem cells or organ-specific progenitors with a 3D extracellular matrix. They exhibit self-renewal and differentiation capacity, while maintaining genetic and phenotypic stability over extended culture periods (76). By more closely recapitulating the structure and functional characteristics of primary tumor tissues, organoids have emerged as powerful tools in both basic and clinical cancer research. Their accessibility and experimental versatility make them particularly valuable for studying tumor immunity and mechanisms of drug resistance.

The limited predictive power of traditional in vitro models presents a major obstacle to the development of effective cancer immunotherapies. However, advances in 3D organoid-immune cell co-culture systems have enabled the establishment of immunocompetent tumor models that more accurately recapitulate the patient-specific TME. These models have transformed preclinical evaluation by enabling the functional screening of both immunotherapeutic and chemotherapeutic agents under conditions that mimic in vivo biology. Their application is exemplified by a study of prostate cancer in which the co-culture of organoids with cancer-associated fibroblasts (CAFs) demonstrated that CAF-derived neuregulin 1 promotes resistance to androgen therapy via HER3 activation in tumor cells (77). Similarly, Sebrell et al (78) showed that co-culturing monocyte-derived DCs with gastric cancer organoids enables patient-derived organoids (PDOs) to recruit chemokines and DCs, thereby participating in immune surveillance during Helicobacter pylori infection. In another study, Dijkstra et al (79) demonstrate that co-cultures of autologous tumor organoids and peripheral blood lymphocytes can be used to enrich tumor-reactive T cells from the peripheral blood of patients with mismatch repair-deficient colorectal cancer and NSCLC. This may provide a basis for individualized treatment of T cells in patients with cancer. Collectively, these advances in organoid-immune cell co-culture technology are crucial for advancing tumor immunotherapy, enabling more precise investigation of tumor-immune interactions and offering valuable insights to inform clinical treatment strategies.

Colorectal cancer organoids have been demonstrated to secrete significantly higher quantities of EVs than their conventional 2D-cultured counterparts (80). In colorectal cancer organoids, APC mutations that activate the Wnt pathway further promote EVs secretion in Matrigel-based cultures, a process that has been suggested to be facilitated by collagen, a key extracellular matrix component in matrigel (81). Another proposed mechanism involves the upregulation of specific transporters. For example, the ATP-binding cassette transporter G1 (ABCG1), a cholesterol efflux pump, is highly expressed in colon adenocarcinoma tumoroids with pronounced stemness features. Silencing ABCG1 inhibits EV release and results in the intracellular accumulation of vesicles (82).

Furthermore, the 3D architecture of organoids supports improved cell polarization and spatial asymmetry. Organoids derived from the LIM1863 colon carcinoma line have been shown to secrete two spatially distinct EV subtypes. Apically released EVs carry epithelial cell adhesion molecule (EpCAM) and are uniquely enriched in CD63, mucin 13, sucrase-isomaltase, dipeptidyl peptidase IV and prominin 1. By contrast, basolaterally released EVs contain the A33 glycoprotein and are associated with early endosome antigen 1, ADP-ribosylation factor and clathrin. These findings demonstrate that EpCAM-positive and A33-positive EVs are selectively sorted and secreted from the apical and basolateral membranes, respectively (83). This spatial organization indicates that distinct EV populations carrying different markers and cargo proteins are released from the apical and basal sides of cells in 3D culture (84).

EVs mediate crucial intercellular communication within the TME by facilitating autocrine and paracrine signaling between tumor cells and stromal components (85). Elevated EV levels have been observed in numerous malignancies, where they transport molecular cargo essential for tumor progression, and serve as potential diagnostic and prognostic biomarkers (86). Also, in tumor-stromal interactions, EVs coordinate complex intercellular crosstalk, which has led to exploration of their applications in drug delivery and targeted therapies (87).

Organoids, through their 3D structure, recapitulate the structural and physiological complexity of in vivo tissues more accurately than do 2D monolayer cultures, and have emerged as powerful systems for tumor biology, regenerative medicine and EV research (75). The development of organoid technology has enabled the functional evaluation of EVs in physiologically relevant 3D models, offers alternative sources for EV production and the potential for advancing our understanding of human diseases (88).

Several studies illustrate the promise of this approach. For example, in an in vitro aging model, mouse intrahepatic bile duct-derived organoids co-cultured with exosomes from human placental mesenchymal stem cells demonstrated the ability to delay cellular senescence (89). In another study, Taha et al (90) revealed that matrix metalloproteinase 3 (MMP3) regulates tumor growth and EV integrity; MMP3 knockout disrupted EV structure and function, reduced tumor organoid proliferation, and suppressed tumor progression, highlighting the protumorigenic role of MMP3 in the maintenance of organoid and EV integrity. In addition, the similarity of small RNA profiles of EVs from 3D cervical cancer spheroids to in vivo-derived EVs is greater than that of 2D cultures, underscoring the physiological relevance of 3D models in EV research (91).

Multiple studies have demonstrated that EVs loaded with specific miRNAs can regulate immune responses, chemotherapy resistance and metastatic processes in various cancers. For example, esophageal adenocarcinoma (EAC)-derived EVs promote the formation of multilayered gastric organoids through the delivery of miR-25 and miR-210, establishing a novel co-culture model of EAC-EVs with 3D organoids (92). Similarly, ovarian cancer-derived EVs from malignant ascites and plasma enhance tumor aggressiveness and organoid growth via the transfer of miR-1246 and miR-1290 (93). In addition, in a study of organoids derived from patients with pancreatic ductal adenocarcinoma, the EVs miRNA profiles of the organoids revealed significantly upregulated levels of miR-21 and miR-195, matching those in patient plasma. This further validates tumor-derived organoids as physiologically relevant models for the analysis of EVs cargo (94).

Collectively, these findings indicate that tumor-derived organoids serve as robust platform for analyzing the molecular cargo of tumor cell-derived EVs. Emerging strategies integrating EVs with organoid models are enhancing our ability to recapitulate and evaluate dynamic disease processes. EV-loaded organoid systems provide synergistic advantages for simulating complex tumor-matrix interactions within the TME, and studies have leveraged EV-organoid co-culture platforms to elucidate TME-mediated mechanisms of tumor progression and therapeutic resistance (95,96).

Overall, tumor organoids are a valuable platform for studying EVs biology, while EVs enhance the physiological relevance of organoids in cancer research. Consequently, the integration of organoid and EVs technologies to generate organoid-derived EVs (OEVs) holds considerable promise across multiple research domains (Fig. 2).