EVs are NPs derived from different cell activities with heterogeneity. According to Minimal Information for Studies of Extracellular Vesicles (MISEV) 2023, the EVs generated from multivesicular bodies (MVBs) are classified as exosomes and those derived from cell membranes are named ectosomes, e.g., microvesicles (MVs) and microparticles. Some EVs are related to a specific type of cellular process including apoptotic bodies from programmed cell death, migrasomes from cell migration, and oncosomes from tumor progression (30). Figure 1 illustrates the biogenesis of different types of EVs.

Furthermore, drugs and genetic intervention may contribute to the stimulation or suppression of EV release (31, 32). However, most isolation techniques are unable to enrich EVs of different biogenesis. The lack of universal biomarkers discourages definitive characterization of biogenesis-based subtypes. Therefore, the biogenesis-related terms such as exosomes and ectosomes are not encouraged to be applied unless they are specifically and carefully separated. However, the majority of the existing literature on “exosomes” and “ectosomes/microvesicles” refers to a broad population of EVs, rather than EVs originating from specific biogenesis pathways (30). In the article, the terms “EVs”, “exosomal”, and “exosome (EXO)” all refer to “extracellular vesicles” referred in MISEV2023 with no indication of their biogenesis.

Apart from the naïve EVs that possess an intrinsic ability to target different cell types, such as tumor cells, immune cells, and stem cells, a number of synthetically modified EVs have been developed to improve their biodistribution and targeting capabilities, boosting EV-based tumor immunotherapy (19, 33). Surface modifications through cellular machinery techniques facilitate tumor targeting and intercellular transformation (34). Cargo-loaded EVs are now a new trend in tumor immunotherapy, which can prevent the clearance of drugs in blood, limit the dose, and reduce side effects (35, 36). Pre-loading is performed before EV secretion or isolation. It enables the membrane integrity of EVs and allows for continuous and easy production of EVs since donor cells are preserved (37). Post-loading directly loads cargo into EVs, exhibiting a higher loading efficiency but potentially altering membrane integrity (38). Table 1 lists the common techniques for EV modification. Figure 2 illustrates the modification and isolation techniques of EVs.

EVs originating from tumor cells play an essential role in influencing their own growth through autocrine mechanisms and shaping the behavior of adjacent cancer cells through intercellular communication ( Figure 4 ). For instance, EVs derived from chronic myeloid leukemia cells contain TGFβ1, which promotes the growth of the producer cell through the activation of ERK, AKT, and anti-apoptotic pathways (41). Furthermore, impaired exosomal maturation and secretion due to the deficiency of vacuolar protein sorting protein 33b (VPS33B) significantly suppresses leukemogenesis (42). Tumor-derived EVs also act as promoting factors for adjacent cancer cells, exemplified by the transmission of oncogenic activity and increased proliferative capacity in glioma cells through the sharing of anti-epidermal growth factor receptor vII (EGFRvIII) via EVs (43). Additionally, these EVs are implicated in angiogenesis, a critical stage in tumor growth. They transport vascular endothelial growth factors (VEGFs) to endothelial cells, promoting the development of microvessels (44). CircRNA and mRNA in EVs also contribute to tumor progression in the TME (45–47). Exosomal circCMTM3 facilitates angiogenesis and tumorigenesis in hepatocellular carcinoma by regulating the miR-3619-5p/SOX9 pathway (45). let-7 g-5p derived from gastric cancer EVs drives M2 polarization in macrophages and contributes to the progression of gastric cancer (48). Moreover, the role of exosomal circRNAs in cancer chemotherapy resistance has been recognized (49). These circRNAs derived from drug-resistant cells are delivered to drug-sensitive cells and result in the resistance of one specific type of drug. For instance, the activation of the ciRS-122/miR-122/PKM2 axis promotes glycolysis and oxaliplatin resistance in colorectal cancer (50). The delivery of Circ-DNER induces the PTX resistance and cancer progression via the Circ-DNER/miR-139-5p/ITGB8 pathway in lung cancer (51).

Stromal cells, including those aiding lymphocyte formation and maturation, play essential roles in shaping the TME. Exosomes derived from normal bone marrow-derived mesenchymal stem cells (BM-MSCs) inhibit the growth of multiple myeloma (MM) cells through increased levels of the tumor suppressor miR-15a (52). Similarly, the high content of exosomal miR-16 downregulated the expression of VEGF in breast cancer cells, which suppresses its growth (44). Conversely, stimulation of cancer cells tends to shift stromal cell-derived EVs toward pro-tumor phenotypes. For instance, fibroblasts stimulated by hepatoma cells exhibit a significant upregulation of SPOCK1/testican-1 pathways, promoting the progression of hepatoma cells (53).

Stromal cell-derived EVs contribute to tumor progression through various mechanisms, including cell proliferation, angiogenesis, and metastasis. They transfer certain RNAs and proteins to stimulate cancer cell proliferation. For example, BMSC-derived non-coding RNA triggered by DNA damage (NORAD) enhances osteosarcoma growth and invasion. Human umbilical cord mesenchymal stem cells (hucMSCs) transmit miR-100-5p, promoting malignancy development. These EVs also participate in tumor angiogenesis, promoting vascular density and tumor growth (54–56). EVs participate in tumor angiogenesis via transporting VEGFs and MMPs. It has been noted that hBMSC-produced EVs contain high levels of VEGF and CRCX4 mRNA, which encourage tumor angiogenesis and development in vivo (57). In addition, high levels of matrix metalloproteinase 1 (MMP1) in oral leukoplakia EVs (OLK-EVs) and oral squamous cell carcinoma EVs (OSCC-EVs) have been reported to be relevant to angiogenesis (58). Additionally, MSC-derived EVs facilitate tumor migration, influencing factors such as integrin expression and mesenchymal–epithelial transition (MET). For example, miR-374a-5p-loaded EVs of gastric cancer-derived MSCs target HAPLN1 to increase the expression of integrins in gastric tumors and promote gastric cancer cell migration (59). Given that EVs are natural nanocarriers with remarkable biocompatibility, their potential as drug delivery platforms has been extensively investigated by numerous researchers (60). Dormancy-inducing EVs from hBMSCs contribute to the acquisition of chemoresistance in metastatic breast cancer cells (61).

Immune cells play crucial roles in the TME, influencing tumor proliferation dynamics in various ways. During the initial stages, tumor cells recruit and activate immune cells, fostering an inflammatory environment that inhibits tumor growth. However, as the tumors advance, some immune cells may experience exhaustion or remodeling, leading to dysfunction and immunosuppression in the TME (62–64). The progress is regulated by the comprehensive intercellular communication network, in which EVs are largely involved by transferring significant signal molecules.

Though T cell-derived EVs are always recognized as antitumor NPs exhibiting a tumor suppression effect, it has been reported that EVs from exhausted CD8⁺ T cells assist tumor progression indirectly by impairing the proliferation of normal CD8⁺ T cells. Incubation of exhausted CD8⁺ T cell-derived EVs with normal CD8⁺ T cells results in reduced proliferation and activity, leading to an increased percentage of exhausted CD8⁺ T cells (65). Microarray technology reveals differential lncRNA expression between exhausted and non-exhausted CD8⁺ T-cell exosomes, highlighting the upregulation of tumor-promoting genes like SUMF2 and CHCHD1, and the downregulation of tumor-suppressing genes like UBXN10 in lncRNA of exhausted CD8⁺ T cell-derived EVs, indicating their potential in promoting tumor growth (65–69). Tumor-associated lymphatic endothelial cells (LECs) secrete EVs rich in miR-142-5p, which upregulates expression of 2,3-dioxygenase (IDO), leading to the exhaustion of CD8⁺ T cells via ARID2–DNMT1–IFN-γ signaling (70). In some instances, even activated CD8⁺ T cells may induce tumor metastasis via activation-induced cell death (AICD), which is mainly modulated by factor-related apoptosis ligand (FasL) (71, 72). FasL in EVs may indirectly participate in the process. It is reported that activated CD8⁺ T cells may secrete EVs containing high levels of FasL, increasing the quantity of cellular FADD-like IL-1β-converting enzyme (FLICE) inhibitory proteins, activating the ERK and NF-kB pathways, subsequently upregulating the expression of MMP9 in B16 murine melanoma cells (73). Natural killer (NK) cell exhaustion and B-cell exhaustion are primarily induced by consistent exposure to antigen stimulation, virus infection, and chronic inflammation, among others. During the exhaustion process, the inhibitor receptors like TIGIT, LAG-3, TIM-3, and PD-1 increase on the surface of the cells accompanied by the reduced expression of antitumor molecules like INF-γ, TNF-α, PFP, and granzyme. (74, 75) These findings confirm the reprogramming of exhausted NK cells and B cells, yet alteration of their EV content and function is rarely explored. These exhausted immune cell-derived EVs, however, may play a role in tumor immune escape, which deserves further research (76, 77).

Major function macrophages can be categorized into three phenotypes, namely, M1 macrophages, M2 macrophages, and tumor-associated macrophages (TAMs), of which M2 macrophages and TAMs promote tumor growth (78). Research on M2-EVs and TAM-EVs focuses on their pro-tumor capacity induced by regulatory molecules. For instance, miR-193b-3p in M2-EVs targets TRIM62, promoting progression and glutamine uptake in pancreatic cancer (79). By targeting GRK6, miR-3917 in M2-EVs promotes tumor progression in a lung cancer model (80). Similar effects driven by RNAs from M2-EVs or TAM-EVs have been observed in colon cancer, EOC, and prostate cancer (81–84). Exosomal RNAs adjust tumor function in alternative ways. Renal cell carcinoma aggression can be driven by miRNA-21-5p in M2-EVs via PTEN/Akt signaling (85). M2 macrophage polarization-associated lncRNA (lncMMPA) facilitates hepatocellular carcinoma malignancy by polarizing M2 macrophages and activating the glycolysis pathway (86). HIF-1α-stabilizing lncRNA from TAM-EVs can also regulate aerobic glycolysis in breast cancer cells (87). Drug resistance in pancreatic adenocarcinoma can also be induced by miR-365 in TAM-EVs. MiR-4443 derived from M2-EVs plays a role in the differentiation of naïve T cells into Treg cells in malignant pleural effusion, facilitating lung tumor growth (88). The same effect is also observed in EOC induced by miR-29a-3p and miR-21-5p in TAM-EVs (89). Other than RNAs, proteins like Arginase-1 from TAM-EVs also assist in cancer proliferation (88).

EVs derived from Tregs, Bregs, M2 macrophages/TAMs, and myeloid-derived suppressor cells (MDSCs) are not the primary regulators, but they also transfer significant messengers, mostly RNAs, such as microRNAs, lncRNAs, or circRNAs. As listed in Table 2 , these EVs activate/silence different signaling axes, regulating key factors and then directly or indirectly boosting tumor growth or metastasis (90, 91). Treg-derived EVs contain several functional molecules that contribute to Treg suppressive activity. For example, exosomal CD73 converts adenosine-5-monophosphate to adenosine, activating the adenosine receptors on target cells’ surface, leading to immune modulation (92). The delivery of miR-146a-5p in Treg-derived EVs inhibit CD4⁺ T-cell growth (93). Moreover, miR-150-5p and miR-142-3p in Treg-derived EVs can modulate DCs’ cytokine constitution (94). It is reported that B-1a regulatory B cells (i27-Breg) can secrete EVs containing IL-27, suppressing and ameliorating uveitis (95). However, in the TME, IL-27 plays dual roles. It both supports CD4⁺ T cells’ proliferation and Th cells’ differentiation but promote tumor growth, invasion, and angiogenesis simultaneously (96). M2 macrophages and TAMs are all immune suppressive cells and their EVs serve a similar function. It is reported that M2 macrophage-derived EVs can foster tumor metastasis and increase vascular permeability in HCC via the delivery of miR-23a-3p, which targets phosphatase and tensin homolog (PTEN) and tight junction protein 1 (TJP1), promoting the secretion of GM-CSF, VEGF, G-CSF, MCP-1, and IL-4 from tumor cells, in turn facilitating M2 macrophage polarization (97). The cargoes found in MDSC-derived EVs have been shown to align with their role in mediating immune suppression by MDSCs (98). However, further in-depth research is necessary to assess the interactions between MDSC-derived EVs and other tumor-infiltrating immune cells, as well as their implications for cancer immunotherapy. Gaining a deeper understanding of the biological functions of MDSC-derived EVs will be crucial for their future therapeutic applications in cancer patients (99).

As EVs mimic the functions of their donor cells, immune cell-derived EVs often exhibit antitumor potential with varying mechanisms. In this section, we categorize EVs based on their sources, including T cells, NK cells, DCs, and macrophages, clustering EVs with similar functions and shared mechanisms.

T cells

EVs produced by CD8⁺ T cells play a crucial role in modulating the communication between immune and tumor cells, thereby influencing tumor development. The interaction between PD-1 on the T-cell membrane and its ligand PD-L1 on the tumor membrane is a well-recognized communication pathway between T cells and tumor cells. Studies have shown that exosomal PD-1 produced by activated CD8⁺ T cells can reduce immunological dysfunction caused by PD-L1 in triple-negative breast cancer (TNBC) patients (100). Strategies involving CD8⁺ T cells treated with EVs containing specific cargoes have been explored to enhance their anti-PD-L1 capacity (101, 102). Additionally, the reduction of exosomal PD-L1 has been associated with an improved antitumor capacity of CD8⁺ T cells (103). By controlling the miR-765/proteolipid protein 2 (PLP2) axis, exosomal miR-765 produced by CD45RO-CD8⁺ T cells prevents the growth of uterine corpus endometrial cancer (UCEC) that is induced by estrogen (104). Similarly, CD4⁺ T cells regulate CD8⁺ T cells’ function via secretion of functional EVs. Exosomal miR-25-3p, miR-155-5p, miR-215-5p, and miR-375 from CD4⁺ T cells are responsible for CD8⁺ T cells’ activation. Compared with IL-2 as antitumor preparations in clinic, CD4⁺ T cell-derived EVs will not stimulate Tregs, which may suggest a promising new avenue for cancer immunotherapy by fostering a CD8⁺ T cell-mediated antitumor response (105). Meanwhile, CD4⁺ T-EVs are crucial for the activation, proliferation, and antibody generation of B cells, which is how humoral immunity is regulated (106). According to a recent study, CD4⁺ T cells can modify macrophages for enhanced cancer immunotherapy based on a stimulator of interferon genes (STING) signaling pathway (107).

Cancer vaccines activate the immune system against tumor cells. In the process, a high volume of high-quality tumor antigens are presented to DCs and activate them to promote CD8⁺ T cells and CD4⁺ T cells, in which EVs are mainly involved (140). EVs, mirroring the functions of their donor cells, can present exosomal tumor antigens to DCs, eliciting tumor-specific CD8⁺ T cells and CD4⁺ T cells (141–143). The inclusion of tumor neoantigens within EVs further contributes to the development of tumor vaccines (11). DCs-EVs as antigen presenters have also found application in immunotherapy, serving as agents of vaccines. Additionally, these EVs can activate immune cells through the regulation of cytokines (144). Figure 5 shows the basic mechanism of an EV-based cancer vaccine.

Tumor-derived EVs provide antigen to APCs. MHC class I and class II are found on the surface of tumor-derived EVs, which is responsible for antigen presentation (145). In lung cancer, EGFR, K-Ras, basigin, carcinoembryonic antigen-related cell adhesion molecule 6, claudin1, claudin3, and RAB family proteins are found to be differentially expressed (146). EVs derived from colorectal cancer can interact with DCs, significantly enhancing immune responses by lowering the antigen presentation threshold for activation at the mucosal level (147, 148). APC-derived EVs inherently rely on MHC compatibility, necessitating a precise match with the MHC haplotype. Conversely, EVs derived from tumor cells transcend this limitation, as they do not require MHC haplotype matching. This pivotal characteristic enables the development of anticancer vaccines that are cell-free and can be universally applied, eliminating the need for individualized engineering for each patient. Furthermore, these EVs harbor tumor antigens that transcend the confines of a single cancer type, hinting at their potential to confer protection against a diverse spectrum of cancers (149). It is noted that exosomal heat shock protein 70 (HSP70) can active DCs and monocytes to trigger immune response and stimulate NK cells to release granzyme B, inducing tumor apoptosis (150, 151). In conclusion, HSP70 can potentially act as an antigen on the surface of EVs to trigger antitumor responses. In certain stages of tumor progression or under proper stimulation, tumor-derived EVs can activate immune cells. For instance, it is reported that bladder cancer cell-derived EVs can boost CD8⁺ T-cell function via cytokine regulation (144). Tumor antigens on these EVs also efficiently activate immune responses (152). EVs derived from tumor cell exposure to methotrexate (MTX) impair the antitumor effect of neutrophils via internalization and degradation of PD-1 in the lysosomes (56). Similarly, antitumor drugs induced highly enclosed HSPs in EVs from human hepatocellular carcinoma cells, which induce inhibitory receptor CD94 and reduce activating receptors CD69, NKG2D, and NKp44, efficiently enhancing NK cell cytotoxicity and granzyme B production (153). However, their application relies on advanced isolation techniques capable of precisely extracting tumor-derived EVs from complex body fluid.

DC-derived EVs serve as the main source of exosomal antigen provider (154). As the most potent APC in vivo, DCs are responsible for antigen uptake and T-cell activation (155). Migration DCs (mDCs) in the TME encapsulate the tumor antigen via EVs and transfer them to resident DCs (rDCs) in draining lymph nodes, which is responsible for activating CD8⁺ and CD4⁺ T cells (156). rDC1, rDC2, and mDC1 are responsible for CD8⁺ T cells’ activation while only mDC2 can make CD4⁺ T cells work. The delivery of antigen also relies on the intercellular transferring of EVs. T cells cannot be activated if the EVs are unable to form (157). Nowadays, DC-derived EVs have been welcomed by many researchers as tumor vaccine carriers. Tumor neoantigen can be loaded in the nanovaccine delivery platform built via DC-derived EVs for individualized immunotherapies. The nanovaccine has demonstrated efficient cargo loading and sustained cargo delivery to the lymph nodes, leading to robust antigen-specific T-cell- and B-cell-mediated immune responses with excellent biosafety and biocompatibility. Notably, the delivery of the neoantigen-EV nanovaccine significantly inhibits tumor growth, extends survival time, delays tumor recurrence with long-term immunological memory, and eradicates lung metastasis in therapeutic, prophylactic, and metastatic B16F10 melanoma models, as well as in therapeutic MC-38 models. Furthermore, the EV-based nanovaccine exhibits a synergistic antitumor response that outperforms liposomal formulations, owing to the presence of EV proteins. Collectively, the research presents enhanced strategies for cell-free vaccines and highlights the potential of EV-based nanoplatforms in cancer immunotherapy and personalized nanotechnology. These findings pave the way for the rapid generation of individualized nanovaccines for clinical use (158). Immunogenic cell death (ICD) inducers are encapsulated within DC-derived EVs and used as a tumor vaccine against breast cancer cells. This approach demonstrates potent antitumor activity in both a mouse model and human breast cancer organoids by enhancing the activation of cDC1s in situ, thereby boosting subsequent tumor-reactive CD8⁺ T-cell responses (159).

Non-antigenic immune adjuvants are essential in cancer immunotherapy, as they enhance immunogenicity and promote antigen presentation, thereby improving the immune response against weakly immunogenic tumors. Previous research has shown that adjuvant-loaded EVs are more effective than administering free adjuvants. Leveraging these insights, as well as the unique properties of EVs, they have been investigated as nanocarriers for the targeted delivery of adjuvants (160, 161). Immune cell-derived EVs can serve as adjuvants themselves, influencing the proliferation and differentiation of immune cells (162). During infection, the amount of circulating phosphatidylserine⁺ (PS) EVs increases, actively modulating CD8⁺ T-cell responses and preferentially interacting with activated, but not naive, CD8⁺ T cells (163). The researchers either take advantage of the inherent homing ability of EVs or modify their surface for precise targeting (164, 165). Additionally, EVs’ capability to deliver adjuvants directly into the cytosol through membrane fusion exhibits significant potential for enhanced immune activation (166). Currently, CpG DNA (167), lipid adjuvants (168, 169), cytokine adjuvants (170–172), HSPs (153, 173), and Gram-negative bacterial outer membrane vesicles (OMVs) (174) are applied as adjuvants prioritizing activation of CD4⁺/CD8⁺ T cells. However, safety concerns persist because adjuvants can trigger inflammatory reactions, which may include fever, ulcers, or even potentially life-threatening cytokine storms.

ACT is a type of cancer treatment that genetically modifies T cells to detect and destroy cancer cells. This approach enhances or changes the intrinsic immune function of T cells, increasing their effectiveness in combating cancer (175). T cell receptor-engineered T (TCR-T) cell therapy and chimeric antigen receptor T (CAR-T) cell therapy are two major trends in ACT, involving genetically engineering T cells to express receptors that specifically target tumor antigens (176, 177). TCR-T cells possess receptors binding TSAs both on the tumor surface and inside tumor cells while CAR-T cells target cancer cells via membranal proteins and kill them without undergoing the antigen-presenting process (178). Traditionally, the engineered T cells are amplified in vitro and infused into the patient’s body, yet the effect is limited because of the reduced penetration ability and potential side effects like cytokine release syndrome (CRS), immune effector cell-associated neurotoxicity syndrome (ICANS), and secondary cancers, among others (179–181). Altering CAR-T cells with CAR-T cell-derived EVs may be a solution boosting its antitumor effect. First, CAR-T-derived EVs are stable particles with a limited lifespan and are unable to proliferate, which may reduce the side effects induced by CAR-T cells, especially CRS. Second, CAR-T-derived EVs are non-cell preparations with low antigenicity, making its application in third-party settings as an off-the-shelf product favorable. Third, CAR-T cell-derived EVs can penetrate the tumor barrier, which may be a solution towards solid tumors (182). CAR-T cell-derived EVs show great potential as direct agents in immunotherapy. These EVs equipped with EGFR and HER2-specific CARs demonstrate a strong capacity against EGFR+ and HER2+ tumor cells in xenograft models. Compared to CAR-T cells, CAR-T cell-derived EVs lack the expression of PD-1, making their antitumor activity resistant to suppression by recombinant PD-L1 treatment (183).

CAR-NK-EVs have also caught researchers’ attention in recent years. Compared with CAR-T cells, CAR-NK cells are less likely to induce life-threatening CRS. Furthermore, CAR-NK therapy is anticipated to be more cost-effective, as NK cells can be sourced from PBMCs, NK cell lines, and human pluripotent stem cells (hPSCs) (184). Moreover, NK EVs will not harm normal cells (118). It has been reported that NK EVs can induce the death of target cells through two major mechanisms: ligand–receptor interactions and plasma membrane fusion (185, 186). Recently, CAR-NK-EVs have been engineered to enhance antitumor therapy by targeting and disrupting the iron death defense mechanism. By modifying the transferrin receptor-binding peptide and expressing CAR on the surface of the EVs, these engineered vesicles can effectively cross the BBB and release therapeutic molecules precisely at the intended sites and times (187). However, the interaction between NK-EVs, other immune cells, and tumor cells is comprehensive and the mechanism has not been clearly delineated; thus, the application of CAR-NK-EVs is worth further exploration.

Given that EVs are natural nanocarriers with a remarkable biocompatibility, their potential as drug delivery platforms has been extensively investigated by numerous researchers.

EVs are recognized as safe vesicles. Native EVs in vivo are reported to undergo reduced hepatic clearance (36, 188). This helps reduce the administration dosage and decrease the potential side effects (34). EVs exhibit exceptional biocompatibility and reduced immunogenicity. In one study, EVs extracted from bovine milk show limited liver and kidney toxicity and no significant increase in histamine concentration (189). Furthermore, early-phase clinical trials have reported mild to moderate side effects of EV-based NP delivery platforms, supporting their safety in clinical applications (15, 190).

Their great penetration power brings EVs to almost everywhere in vivo. There are several barriers in the human body that protect some important organs or tissues from foreign bodies and maintain their normal function. Yet, the barriers prevent most of the drugs from entering these locations (191). The BBB is an intricate and highly selective barrier in the human body. It serves to safeguard the brain and maintain the stability of the central nervous system (CNS). This barrier is primarily composed of endothelial cells that form a tightly joined monolayer, covering the brain’s capillaries. What is worse, brain tumor cells tend to format the BTB (192). EVs can cross the barrier via several mechanisms like receptor-mediated transcytosis, lipid raft-mediated endocytosis, and micropinocytosis (21). Given this character, EVs are considered promising carriers targeting tumors in the brain. Ginseng-derived exosome-like nanoparticles (GENs), composed of phospholipids and various bioactive components, are currently being evaluated for their ability to stimulate antitumor immune responses in T cells and Tregs, with the aim of inhibiting tumor progression. Their enhanced targeting ability to the BBB and glioma shows a significant therapeutic effect, demonstrating strong efficacy in recruiting M1 macrophage expression within the TME. GENs are proved to be successful candidates for glioma therapeutics in both in vitro and in vivo studies, indicating excellent potential for inhibiting glioma progression and regulating TAMs (193). Fruit-derived EV-engineered structural droplet drugs (ESDDs) are created by programming the self-assembly of fruit-derived EVs at the interface of DOX@squalene-PBS, significantly enhancing their antitumor efficacy against glioblastoma. The blood–testis barrier is among the most restrictive blood–tissue barriers found in mammals. It separates the seminiferous epithelium into two distinct areas: the basal compartment and the apical (or adluminal) compartment (194). EVs have been applied in delivering drugs to testis. Sertoli cell-derived small extracellular vesicles (SC-sEVs) can cross the BTB and enter germ cells. By loading miR-24-3p inhibitors into these vesicles, the nano-drug SC-sEV@miR-24-3p inhibitor is created, which efficiently delivers the miR-24-3p inhibitor to germ cells. In a mouse model of gossypol-induced asthenozoospermia, treatment with the SC-sEV@miR-24-3p inhibitor significantly enhanced sperm motility, increased the success rate of in vitro fertilization, and improved blastocyst formation rates. As expected, it also increased the litter size in asthenozoospermia mice. These findings suggest that the SC-sEV@miR-24-3p inhibitor could be a promising clinical treatment for asthenospermia (195).

The potent targeting capacity of EVs boosts precision medicine. EVs can be passively accumulated in the TME via the enhanced permeability and retention (EPR) effect. The EPR effect refers to the phenomenon where NPs of appropriate sizes preferentially accumulate in tumor tissues compared to normal tissues, leading to an extended retention time of the NPs within the tumor area (196, 197). This phenomenon occurs because the abnormal blood vessels found in tumors enhance vascular permeability (198). NPs ranging from 20 to 200 nm in size can infiltrate the interstitial space due to the misaligned and defective endothelial cells (199). Additionally, the clearance of NPs from the TME is often delayed due to the limited lymphatic drainage present in these areas (200). Moreover, EVs can be actively targeted to tumor cells by the ligand–receptor connection. On the surface of tumor cells are tumor markers such as carcinoembryonic antigen (CEA) for colorectal cancer (201), carbohydrate antigen 125 for ovarian cancer (CA125) (202), and neuron-specific enolase (NSE) for neuroendocrine tumors (NET). Furthermore, taking advantage of the prosperity in the field of membrane-editing technology, the EVs’ membranes are engineered to enhance their targeting ability for precise cytotoxic effect and limited harm to normal cells. Engineered DCs-EVs, bound with membrane anchor lysosome-associated membrane glycoprotein 2b (Lamp2b) and brain-specific rabies viral glycoprotein (RVG), showcase the potential of EVs as targeted drug delivery systems. After tail vein injection of the EVs, knockdown of BACE1 mRNA and protein is demonstrated in the brains of mice (19). EVs can also be reprogrammed to promote their accumulation in the TME. It is reported that low pH, a significant feature in the TME, reprograms tumor EVs for enhanced homology via a glycolipid self-aggregation-based mechanism, which sheds light on the exploitation of environment-responding EVs (203). The engineered EVs derived from M1 macrophages are conjugated with dibenzocyclooctyne-modified antibodies targeting CD47 and SIRPα (aCD47 and aSIRPα) via a pH-sensitive linker. These EVs are designed to accumulate in the acidic TME and specifically target tumor cells by recognizing the interaction between aCD47 and CD47 on the tumor cell surface (204). Additionally, the magnetic field gradient is an alternate noninvasive technique to improve targeting efficacy. EVs derived from macrophages are loaded with drugs and iron oxide NPs, which spatially regulate the absorption of EVs and drugs by cancer cells in vitro (205). The self-targeting capacity of tumor-derived EVs can also be employed for precise targeting (206). EVs derived from tumor-repopulating cells (TRCs) isolated from three-dimensional fibrin gels enhance the efficiency of drug delivery. Compared to EVs derived from tumor cells cultured on conventional tissue-culture plastic, TRC-derived EVs, when intravenously injected into mice with tumor xenografts, show increased accumulation in tumor tissues, improved crossing of blood vessels, and deeper penetration into the tumor parenchyma. They are also preferentially taken up by highly tumorigenic TRCs. Additionally, the cytoskeleton-related protein cytospin-A plays a crucial role in regulating the softness of TRC-derived EVs. Modulating the mechanical properties of these EVs could improve the delivery efficiency of anticancer drugs (207). The EV preparation can be delivered to tumor sites via drainage tubes for some metastatic tumors like malignant pleural effusion (MPE) or malignant ascites, or injected directly into the superficial solid tumors like melanoma. For most tumors, intravenous injection is the most common administration method. Since tumor capillary permeability (~780 nm) is larger than that in normal tissue (5–8 nm), it is hard for tumor-derived EVs (100–1,000 nm) to reach normal tissues while they can enter tumor parenchyma easily (208–210). However, the potential oncogenesis risk of tumor cell-derived EVs limits their application as theriacal molecules and drug delivery, yet an attempt to apply tumor cell-derived EVs to deliver drugs is on trial (listed in Table 3 ).

Moreover, surface modification of EVs through cellular machinery techniques facilitates tumor targeting and intercellular transformation (34). For example, modified NK-EVs can be obtained by exogenous and endogenous alterations. Exogenous modifications involve loading drugs like sorafenib or cisplatin onto NK-EVs, which enhance apoptosis in TNBC and reactivate NK cell functions against drug-resistant ovarian cancer, respectively (211, 212). Engineered NK-EVs with small interfering RNA (siRNA) and hydrophobic photosensitizer Ce6 present cytotoxic effects towards tumor cells via reactive oxygen species (ROS) and conscripted immune cells (122). Endogenous modifications achieved enriched specific cargo through lentiviral transduction into the parent cells. Lentiviral transduced NK92MI cells enrich BCL-2 siRNAs in EVs, enhancing their intrinsic apoptosis in breast cancer cells (121). Though NK cells share similar functions with T cells, rare attention has been paid to CAR-NK-derived EVs, which may be endogenous modification methods of the next generation.

M1-EVs, when engineered, demonstrate enhanced abilities to inhibit tumor growth and regulate the immunosuppressive TME. Engineered M1-EVs deliver RSL3 as a ferroptosis inducer, which disrupts redox equilibrium to increase the oxidative stress-triggered robust ferroptosis of tumor cells (213). Docetaxel-loaded M1-EXOs polarize naïve M0 macrophages toward the M1 phenotype as opposed to the M2 phenotype by using mitochondrial function (214). M1-EVs can also load therapeutic agents, like catalases, DNA damage repair inhibitors, and anti-PD-L1, which simultaneously target tumor hypoxia, cancer DNA damage, and T-cell function (215). In addition, EVs derived from macrophages are potent drug delivery systems. Engineered M1-EVs have been created by functionalizing the membrane with chemical excitation source CPPO and photosensitizer Ce6, as well as encapsulating the hydrophobic prodrug AQ4N. These modified M1-EVs penetrate the BBB, induce M2-to-M1 polarization, and increase hydrogen peroxide (H₂O₂) levels. The reaction between H₂O₂ and CPPO activates Ce6, generating large amounts of oxygen species to achieve chemiexcited photodynamic therapy (CDT). AQ4N also converts into toxic AQ4 in the hypoxic TME, inducing apoptosis of glioblastoma multiforme (GBM) (216). AS1411 aptamer-modified macrophage exosomes are also utilized to coat the sonosensitizer indocyanine green, enhancing the sonodynamic therapy of glioblastoma (217). Furthermore, the modified macrophages’ EVs demonstrate strong antitumor activity, indicating the significant potential of macrophages as sources of EVs (218).

The excellent feature of inflammatory chemotaxis makes neutrophil-derived EVs (NE-EVs) a remarkable drug for targeting tumors. NE-EVs are loaded with DOX for targeted glioma therapy, which penetrates the BBB and reacts to inflammation. Additionally, NEs-EVs/DOX intravenous infusion effectively slows tumor growth and lengthens survival in a mouse model of glioma (219). In another study, NE-EVs are decorated with superparamagnetic iron oxide nanoparticles (SPIONs) to improve the tumor-targeting capacity, enhancing the antitumor effect of DOX. These EVs induce tumor apoptosis without affecting normal cells, exhibiting superiority in targeting and efficacy compared to normal NE-EVs (220).

As mentioned above, strategies applying EVs as therapeutic agents have been explored in-depth. However, the widespread clinical application requires large-scale synthesis of EVs. Notably, scaling out and scaling up can optimize the cell culture process and thus contribute to the vast production of EVs.

The scale-out of the culture system refers to the cultivation of more cells in a limited space. For instance, the Integra CELLine Culture System increases EV yield to 10.06 ± 0.97 mg/mL compared with 0.78 ± 0.14 mg/mL by traditional culture (221). This technology is designed for scalable production of EVs from adherent cell lines, like cancerous bladder cells (222). Regretfully, this platform is not suitable for stromal cells since it maintains cells at high densities for prolonged periods of time. Another strategy for vast production is the automatic release of EVs in hollow fiber bioreactors. In this system, cells are grown on the surface of semi-permeable fibers and release EVs to flowing supplemented media. However, previous research reported the occurrence of cell differentiation and increased cell density after 6 weeks of culture. This phenomenon leads to the low EV yield of a single cell (223, 224). Optimizing collection frequency may solve such challenges. In particular, a combination of hollow fiber bioreactor and size exclusion chromatography/tangential flow filtration enables the production and enrichment or purification of clinical-grade EVs at a moderately vast scale (225). In addition, the application of hyperflasks reduces manual operations during cell culture and media harvest and promotes the production of EVs, yet compared with 2D flasks, they are not economical (222). The CellBIND^® surface is pretreated with oxygen-containing functional groups and carries a net negative surface charge. Bioreactors are commonly utilized for large-scale production due to their dynamic monitoring systems, which are advantageous for GMP processes (226). Xeno-free cell culture significantly improves EV production by reducing cell doubling time, increasing EV yield, and achieving up to 97% removal of contaminating proteins. Specifically, a 10% pooled human platelet lysate (HPL)-based, EV-depleted medium effectively supports the production of human MSC-derived exosomes while maintaining their characteristic surface markers, morphology, viability, and in vitro differentiation potential (226, 227).

Meanwhile, the scale-up strategy increases EV yield by cultivating cells on microcarriers in stirred tank bioreactors. This 3D culture method outperforms the traditional 2D culture method in the aspect of silencing siRNA loading (228). Additionally, physical stimulation, including hypoxia (229), low pH (162, 230), heat shock (231), or ultrasound (232), can improve the production of EVs. Serum deprivation is another stimulation to boost EV yield that prevents the contamination of serum-derived EVs and particles (233, 234). Of note, although EVs produced by stimulation are often reported to display similar physical characteristics with those produced without stimulation, their protein and RNA contents may be different, which could undermine the process of EV preparations (235).

In addition to optimizing the cellular culture, ideal cellular sources can also facilitate the large-scale production of EVs. One potential source could be embryonic stem cells (ESCs). ESC exhibits the capability of almost unlimited self-renewal and offers sufficient EV sources for clinical application (236). According to previous research, EVs secreted by ESCs display satisfactory antitumor properties. Human red blood cells (RBCs) are another ideal cellular source. Specifically, group O-RBCs could be utilized as universal donors for the vast production of EVs. This is because group O-RBCs are devoid of DNA and are available in blood banks (237).

Purification strategies have always been a major problem affecting the process of the clinical application of EVs (238). The selection of purification strategies has a direct impact on clinical effects. On the one hand, it is significant to eliminate contaminations in EV preparation including the unexpected EVs, liposomes, proteins, and RNAs. For instance, some EVs are extracted from culture medium of cancer cells in vitro for their well-known tumor-targeting capacity, yet the pro-tumor EV subtypes may harness the therapeutic effect. What is worse, it is noted that tumor-derived EVs play a critical role in modulating the TME, promoting tumor growth, metastasis, immune evasion, and even drug resistance via various means (239, 240). On the other hand, some so-called “contaminations” ought to be kept in EV preparation. One reason is that it is impossible to isolate the single-component EVs or one designated subtype of EVs from complex EV groups of body fluid or culture medium (241). One major obstacle in the field of EV isolation is the large scale of co-isolated lipoproteins sharing similar characteristics such as density, size, and component (242, 243). It is noted that the application of density-gradient ultracentrifugation for EV purification results in the co-isolation of LDL and HDL due to the similar density, and a SEC-based isolation strategy will lead to contamination of chylomicrons (244). Even the purification strategy combining the SEC purification step and the differential ultracentrifugation cannot avoid the presence of LDL in plasma EV preparations since the LDL particles have a higher concentration by several orders of magnitude than EVs in human plasma (245). Ultracentrifugation, the main applied isolation method, is also impaired by LDL and HDL residues due to the sedimentation rate difference of EVs and HDL/LDL (246). Though there is a new technique applying the styrene-maleic acid (SMA) copolymer to selectively break down lipoproteins, it has not been widely applied and its adverse effect on SMA remains unknown (247). Another reason is that the so-called “contamination” may facilitate efficacy in some cases. In a study, researchers discovered that the presence or absence of EVs did not impact the ability of human mesenchymal stem cell (hMSC)-conditioned medium to promote angiogenesis and wound healing in vitro while much smaller soluble factors like VEGF play a more important role in the progress. However, when applied in far higher concentrations than those presented in conditioned medium, the hMSC-derived EVs also exhibit wound-healing capacity (248). The research can be seen as a warning, urging scientists to reconsider the complex relation between EV and non-EV factors and the potential experimental hazards when conducting experiments regarding EV bioactivity. In a word, it is both unlikely and unnecessary for researchers to achieve “complete purification”, namely, isolating the single-ingredient EV during preparation. For the clinical application of EVs, we can focus on isolating EV groups including specific subtypes of EV and non-EV factors. The EV groups should satisfy clinical needs, avoid potential adverse effects, and entail reduced purification cost.

Different isolation methods lead to different outcomes. Despite starting with the same source, different processing procedures may lead to the various mixtures of co-isolates and EV subsets. In the study, higher pERK/ERK ratios are observed after stimulation of SEC-EVs than after stimulation of UC-EVs, which indicates the significance of optimizing isolation strategies in the clinic (249). Another study reports that different isolation methods leave different contaminations in HEK293T-derived EV preparation. These media component contaminations result in the suspicious error that HEK293T-derived EVs possess anti-inflammatory bioactivity (250). These studies reveal that different isolation methods applied in different research studies influence the validity and accuracy of the result and even may lead to false outcomes.

Standard purification strategy matters in clinical settings. The impact of isolation methods has been mentioned above. The heterogeneity of isolation methods hinders the quantification and comparison of the results (251). It is noted that 80% of studies simply conduct ultracentrifugation to isolate the EVs (242, 243). Ultracentrifugation has been recognized as the gold standard for EV isolation and is efficient in enriching EV fractions, allowing for the collection of additional larger EV components (252). However, it has restricted the processing volume due to the limited thin loading zone (253). Furthermore, the requirement of expensive equipment and well-trained technicians restricts its wide application. In addition, prolonged exposure to ultracentrifugal forces can negatively impact the structure and biological function of isolated EVs, making them less suitable for downstream applications like EV-based functional studies and drug development (254). According to MISV2018, you cannot obtain both high yield and high purity in EV production (18, 255). Thus, we ought to establish a standard purification strategy, which is a series of isolation technique combinations catering different EV sources (culture medium or body fluid). Moreover, such a strategy entails guidelines that will evaluate existing and new methods and purify EVs in order to confirm their usage in different scenarios. The strategy aims at stipulating standard, economic, and efficient purification methods for mass production and clinic use.

Another noteworthy problem preventing EV preparations from widespread clinic application is storage. Although numerous studies have examined the properties of, and the roles played by, EVs in vivo and their potential in immunoregulation, drug delivery, and biomonitoring, they keep the EVs or raw materials containing EVs (e.g., culture medium, body fluid, and extracts) for a short period, ignoring the changes in the active ingredient in EVs during long-term storage (256, 257). According to the research, even if the EVs are carefully enriched, isolated, and purified, improper preservation strategies may still result in EV fusion or crack, protein aggregation, or degradation, leading to failure in EV preparation (256, 258). Like other biopharmaceuticals, the application of EVs will involve a series of procedures including transfer, storage, and disposal, requiring the exploration of economic storage strategy and the definition of quality period (256, 258, 259).

In light of this lack of standard procedures for EV storage and regarding research, current studies have not reached a consensus (260). The application of cryoprotective agents (CPAs) like trehalose (261, 262) or dimethyl sulfoxide (263, 264) is suggested by some authors. Researchers have discovered that adding 25 mM trehalose to the isolation and storage buffer for pancreatic beta-cell exosome-like vesicles narrows the particle size distribution and increases the number of individual particles per microgram of protein. In macrophage immune assays, beta-cell EVs stored in trehalose consistently show higher TNF-alpha cytokine secretion stimulation indexes, indicating better preservation of biological activity (261). Furthermore, polyacrylamide gel electrophoresis (PAGE) analysis demonstrated that both proteins and RNA within EVs are preserved after lyophilization when trehalose is present. Lyophilization has minimal impact on the pharmacokinetics of Gaussia luciferase (gLuc)-labeled EVs following intravenous injection into mice (262). Cryopreservation of platelets with DMSO leads to the release of platelet microvesicles (PMVs) and a significant increase in thrombin generation and procoagulant activity (TG-PCA) compared to liquid-stored platelets (LSPs) (263). The morphology of EVs cryopreserved using DMSO is similar to SEM images of fresh EVs. Although the sizes and shapes of a certain percentage of EVs are preserved, DMSO is unable to maintain the morphology of all vesicles in the sample (264). Another widely applied method for long-term preservation is cryopreservation including freezing and lyophilization (262, 265). The EVs are usually recommended to be preserved at −80℃ for long-term preservation and at 4℃ for temporary storage (266). A series of studies discuss the effect of different temperatures or the speed of freezing or thawing. Despite encouraging findings, some results remain conflicting, and comprehensive studies that compare different storage strategies simultaneously are still lacking. Additionally, most studies have analyzed samples after relatively short time periods (e.g., hours, days, or occasionally weeks). Those that have examined samples after longer preservation periods have primarily focused on the storage of biofluids rather than isolated EVs (265, 267–272). Lyophilization makes its progress in vacuum to protect the easily oxidized components. Moreover, lyophilization reduces the water in EV samples, which enhances their stability and reduces the risk of contamination, facilitating transportation and extending storage duration (273). Lyophilization without a cryoprotectant results in the aggregation of exosomes derived from B16BL6 melanoma cells, while adding trehalose, a cryoprotectant, prevents this aggregation. PAGE analysis reveals that trehalose protects the proteins and RNA of the exosomes during lyophilization. The procedure has little effect on the pharmacokinetics of Gaussia luciferase (gLuc)-labeled exosomes after intravenous injection into mice. Additionally, lyophilized exosomes retain the activity of loaded gLuc and immunostimulatory CpG DNA for approximately 4 weeks, even when stored at 25°C (262). The spray-dry technique is another technique preserving EVs. When an EV solution is atomized in a drying chamber, the moisture quickly evaporates once in contact with hot air and leaves dry powder. During this process, atomization pressure and outlet temperature are factors that influence the stability of EVs. Compared to lyophilization, spray drying is a continuous process that can achieve one-step formation, making it more economical and suitable for large-scale production (274). However, like lipid nanoparticles (LNPs), shear stress, liquid interface expansion, and stress caused by thermal dehydration during the collection process may harness the EV membrane (275, 276). In recent years, some researchers try to preserve EVs in gelatin methacryloyl hydrogel (GelMA) (277). Since the irregular Brownian transport of EVs is the cause of membrane fusion and the inactivation of its contents (278–281), GelMA can encapsulate EVs, limiting their random movement and reducing their aggregation, thus improving their stability. The good biocompatibility, the ability to be administered without affecting EV activity, and the well-established preparation techniques that enable synthesis or commercial availability highlight its potential for clinical application (277).

In conclusion, there are six key points in EV storage. (1) Freeze–thaw reduction. It is reported that the freeze–thaw cycle decreases EV yield and increases particle size due to membrane fusion and protein loss. EV transportation should follow the principle of minimizing freeze–thaw cycles. If the EVs being transported are stored under frozen conditions, it is recommended to use sufficient dry ice for transportation. When transporting freshly isolated EVs, it is advised to use adequate ice packs and deliver them to the destination as quickly as possible (282, 283). (2) Application of screw caps and rubber seals to reduce the impact of freeze-drying during the storage of EVs. The rubber seals enhance the tightness of screw caps, separating EVs from contaminants from the air and reducing oxidation (284, 285). (3) Utilizing low-adsorption materials to store EVs. It reduces loss of EVs and assists in keeping the key molecules on the EVs’ surface (286). (4) To prevent ice crystal formation and reduce low-temperature precipitation, seal the container with sealing film, aliquot the EV samples, and quickly freeze them in liquid nitrogen. Store at −80°C or below, and thaw at 37°C (287). (5) Purified EVs can be temporarily stored at 4°C, but should not be kept for more than 48 h. Furthermore, it is recommended to prioritize storing unextracted samples at −80°C because untreated samples are better stored at −80°C compared to purified EVs (282, 283, 288). (6) Storage conditions and duration have a significant impact on EVs. A higher storage temperature and a longer storage time contribute to fewer EVs remaining in the sample (260).

Recently, applying EVs as a biomarker and as a diagnostic and prognostic predictor has been another promising clinical application (289). There are several reasons for EVs to be an excellent disease indicator. First, EVs are intercellular vesicles transmitting bioactive molecules regulating cell development, differentiation, and function, which can serve as sensitive and specific biomarkers (290). Second, secreted by most cells in vivo, EVs exhibit a high concentration in most bodily fluids, which means they can be easily captured and tested, especially large EVs (291). Third, accumulating studies have revealed the relation between changed EV content and certain diseases (292). For example, it is reported that the combination of EV TDP-43 levels and EV 3R/4R tau ratios can assist in diagnosing frontotemporal dementia (FTD), FTD spectrum disorders, and amyotrophic lateral sclerosis (ALS). EV tau ratios are low in progressive supranuclear palsy (PSP) and high in behavioral variant frontotemporal dementia (bvFTD) with tau pathology. EVs TDP-43 levels are elevated in ALS and in bvFTD with TDP-43 pathology. Both markers effectively discriminate between diagnostic groups, achieving area under the curve values greater than 0.9. They also differentiate between TDP-43 and tau pathology in bvFTD. Additionally, both markers strongly correlate with neurodegeneration, as well as with clinical and neuropsychological indicators of disease severity (293). Through selective reaction monitoring/multi-reaction monitoring (SRM/MRM), EphA2 on urinary EVs presents significant expression differences between patients with bladder cancer/non-malignant hematuria and healthy controls. The subsequent research also shows that EVs-EphA2, which promotes the proliferation, invasion, and migration of bladder cancer cells, exhibits strong diagnostic performance, with a sensitivity of 61.1% and a specificity of 97.2% (294). In addition, EVs may have some prognostic value. It is reported that tRNA-derived small RNAs (tsRNAs) are specifically enriched in salivary EVs of ESCC patients with high sensitivity (90.50%) and specificity (94.20%). According to the bi-signature Risk Score for Prognosis (RSP), patients with a high RSP have significantly shorter overall survival (OS) (HR 4.95, 95% CI 2.90–8.46) and progression-free survival (PFS) (HR 3.69, 95% CI 2.24–6.10) compared to those with a low RSP. Moreover, adjuvant therapy is found to improve OS (HR 0.47, 95% CI 0.29–0.77) and PFS (HR 0.36, 95% CI 0.21–0.62) only in patients with a high RSP, but not in those with a low RSP (295). Another study indicates that EV-derived B7-H3 and B7-H4 emerge as noninvasive predictors of survival in patients with metastatic NSCLC treated with ICIs, functioning independently from their expression in tumor tissue. Interestingly, an increase in PD-L1⁺ EVs is primarily linked to disease progression, while a rise in B7-H3⁺ EVs seems to be associated with a positive response to ICI treatment (296). Moreover, EVs can guide drug application. For example, it is reported that small extracellular vesicles (sEVs) carry multiple inhibitory immune checkpoint proteins, creating a potentially targetable adaptive mechanism that suppresses antitumor immunity. It sheds light on the role of sEVs in tumor drug resistance mechanisms, which is helpful in improving the patient response rate of immune checkpoint blockade therapy (297). Another comprehensive review reveals the role of EVs in assisting drug resistance and the possible methods that target EVs to overcome tumor drug resistance (298). However, the application of EVs as biomarkers is still in its infancy and there are some practical problems that remain to be solved.

The source and isolation techniques of EVs determine their application. Currently, EVs as noninvasive biomarkers are usually extracted from blood and urine (299); CSF and saliva are also included in some studies (300, 301). Blood is the most abundant EV source among bodily fluids, yet its complex components and viscosity bring up significant challenges for EV isolation (233); thus, kits are not recommend for blood–EV isolation since co-precipitated proteins and liposomes may contaminate the EVs (302–304). Size exclusion chromatography (SEC) and ultracentrifugation are effective methods for separating protein from EV components with a higher yield and purity compared to other methods, yet there are still residual lipoprotein particles. Furthermore, EVs separated by ultracentrifugation tend to have lower yields, which may not meet the minimal testing dose (305). It is recommended that a combination of different methods can be applied to remove liposomes and proteins when isolating blood EVs (306). However, an improper isolation method may damage the target protein, leading to a false result (307). Urine, as another widely applied bodily fluid for disease detection, is considered the most suitable biological specimen for research on urinary system-related tumors due to its direct connection with the urinary system (308). Unlike blood, urine samples can be collected in larger volumes with a lower protein content. Given the characteristics of urine samples, ultrafiltration is commonly used to concentrate the urine, removing soluble contaminating proteins and obtaining concentrated EV-containing solution. Subsequently, methods such as SEC or precipitation are employed to isolate relatively pure EVs (309). CSF is recognized as the best bodily fluid reflecting brain and spinal cord conditions, and CSF EVs are helpful in maintaining healthy nervous system function (310). CSF can be collected via brain puncture and lumbar puncture. Brain puncture can collect a large volume of samples but may be contaminated by blood. Lumbar puncture collects 20 to 25 mL of sample at a time; it is usually free of blood and can ensure a replicable strategy (311). It is important to exclude blood from CSF samples. Thus, centrifugation of the lumbar CSF sample before freezing and storage is recommended to eliminate any potential cellular contamination that could interfere with EV research. The sampling position, method, and the subsequent processing strategy should be identical when making effective comparison in medical research or clinical settings (312). As to CSF EV isolation, filtration and SEC are recommend techniques since they work well on small-volume samples (313). Saliva has its advantages of being easily collected, noninvasive, and safe (314). Saliva EVs not only reflect the health condition of the adjacent tissue of salivary gland (295, 315), but also can be a biomarker for other tumor diseases like breast cancer, pancreatic cancer, and lung cancer (316–319). However, saliva is primarily secreted by the sublingual, submandibular, and parotid glands, and the composition of saliva varies depending on the sampling location and time (320). Thus, when collecting saliva EVs, it is necessary to either standardize the sampling location or collect saliva from all locations (320). Moreover, the composition of saliva also changes at different times of the day; thus, sample timing is also important (321). Notably, eating, drinking, smoking, and vigorous exercise are prohibited within 1 h before sampling to ensure that saliva EVs are not contaminated (322).

It is noted that race may have an impact on EV content. This should be taken into consideration when determining the threshold. For example, exosomal miR-1304-3p has been identified as the most upregulated mRNA in African American breast cancer patients, showing a significant difference when compared to Caucasian American patients (323). However, limited attention is paid to the field, partly because of the difficulty in sample acquisition and insignificance in content difference.

Attempts to apply EVs in preclinical and clinical settings have emerged recently with satisfactory results (324). The application mainly focuses on utilizing EVs as diagnosis and prognosis biomarkers and modifying EVs as therapeutic agents (325–327).

Compared with healthy people, cancer patients tend to present high levels of EV production (328). Their correlation with tumor not only helps with early diagnosis but also indicates potential targets for tumor immunotherapy (329). It is noted that the early detection of tumors is crucial for effective cancer immunotherapy (330). For instance, miRNA-200-5p, miRNA-378a, miRNA-139-5p, and miRNA-379 are proven to be noninvasive sensitive biomarkers in the diagnosis and screening of lung cancer (331). It is also found that postoperative blood samples from pancreatic ductal adenocarcinoma (PDAC) patients showed reduced levels of exosomal Sox2ot expression, which plays a role in tumor development and may serve as a valuable prognostic marker for pancreatic cancer (332). Understanding how EVs package their cargo, release it into bodily fluids, and maintain stable levels in health and disease is crucial for developing powerful biomarkers to monitor disease onset and progression (324, 333).

Besides biomarkers, EVs can also be therapeutic agents. EVs derived from DCs originate from either immature or mature DCs activated by cytokines, such as recombinant interferon-γ. The injection dose typically ranges from 8.5 × 10¹¹ to 4.0 × 10¹³ EVs containing MHC class II molecules. To stimulate an immune response in cancer patients, these DC-derived EVs, loaded with tumor antigens, are administered subcutaneously. Clinical trials using these EVs have been conducted for melanoma and non-small cell lung cancer, showing similar safety outcomes. However, in the case of non-small cell lung cancer, MAGE-specific T-cell responses have been observed (334). These tumor antigens, such as carcinoembryonic antigen, can be obtained directly from cancer patients by harvesting ascites-derived EVs. A phase I trial demonstrated that this method is safe and well-tolerated. Moreover, in the group receiving ascites-derived EVs combined with granulocyte–macrophage colony-stimulating factor, a tumor-specific cytotoxic T lymphocyte response is observed (335). Manipulating EV content and production also aids immunotherapy. In a study, researchers discovered that EVs extracted from malignant ascites contained the MET oncogene, which enhanced tumor invasiveness. Furthermore, inhibiting the secretion of these EVs suppressed tumor progression (336).

Co-delivery of chemotherapy drugs or siRNAs also boosts tumor immunotherapy. EVs have been extensively utilized in the treatment of various malignancies (337). It is reported that repeated injection will not harm to body (328). Moreover, their immune compatibility and biocompatibility make them well-suited for therapeutic applications (338). Two clinical trials (NCT01854866 and NCT02657460) are investigating the application of chemotherapy drugs to treat patients with malignant pleural effusion. In the preclinical study and the NCT01854866 trial, MTX and cisplatin are used as anticancer drugs, respectively. Results from the preclinical study show a higher survival rate in MTX application compared with cisplatin (339). KrasG12D siRNA has been explored as a potential anticancer therapy for patients with metastatic pancreatic cancer. In another study, mesenchymal stromal cell-derived EVs have been proposed as a therapeutic approach, as outlined in clinical trial NCT03608631. In an orthotopic pancreatic tumor model, targeting tumor-initiating cells (TICs) with MSC membrane-derived nanovesicles, known as “nano-ghosts,” loaded with a CXCR3 antagonist improved treatment efficacy and delayed tumor recurrence when combined with gemcitabine. Since MSC-derived nano-ghosts may preferentially home to tumors and specifically target the TIC population, we propose utilizing them as “Trojan horses”. A promising approach for overcoming treatment resistance, particularly in desmoplastic cancers such as pancreatic adenocarcinomas, is nano-ghost-based therapy (340).

Insults to donor cells can produce EVs with distinct features. Generally speaking, tumor cell-derived EVs inhibit tumor immunotherapy and promote tumor growth and metastasis. However, there are some exceptions (341, 342). UV-treated tumor cells produce EVs devoid of HSPs but enriched with abundant genomic and mitochondrial DNA fragments. These EVs can serve as vaccines stimulating DCs via cGAS/STING signaling. This finding suggests a new tumor cell-free vaccine strategy with promising potential for clinical applications (343). Radiation-treated cell-released EVs show the capacity to suppress murine brain metastasis via the reprograming of the TME by inhibiting the MAPK pathway (342).

To date, the heterogeneity of EVs may have an impact on their clinical application. The heterogeneity of EVs likely reflects their size, composition, functional effects on recipient cells, and cellular source. Size heterogeneity may be induced by the uneven invagination of the limiting membrane during formation and flaws in isolation techniques (344, 345). Even after purification, EVs are observed to exhibit variable abundance of cargoes, including mRNAs and proteins (244, 346). Subsets of EVs may function differently in one set of the isolated EVs: one set may promote cell survival while another induces cell apoptosis (347). The heterogeneity of EVs may influence the results of EV studies, yielding false-positive or false-negative results. Furthermore, advanced techniques in EV purification and classification ought to be developed to distinguish the subtypes of EVs with different functions.