Tumor-derived extracellular vesicles: Bridging communication and next-generation theranostics

Abstract

Cancer remains one of the leading causes of mortality worldwide, and despite advancements in therapeutic strategies—including chemotherapy, radiotherapy, immunotherapy, surgery, hormone therapy, and targeted therapy—a definitive cure remains elusive. In recent years, tumor-derived extracellular vesicles (TD-EVs) have garnered attention due to their critical roles in tumorigenesis, angiogenesis, and metastasis. Generated via biogenesis pathways involving the endosomal sorting complex required for transport (ESCRT), TD-EVs facilitate diverse mechanisms that promote tumor growth and survival. These include the induction of epithelial–mesenchymal transition (EMT), stimulation of angiogenesis, suppression of natural killer (NK) and T cell activity, promotion of M2 macrophage polarization, and facilitation of metastasis. Beyond their tumor-promoting functions, TD-EVs also hold promise as diagnostic and therapeutic tools. For example, EV PD-L1 has emerged as a biomarker for the liquid biopsy, reflecting tumor immune evasion, while engineered TD-EVs loaded with therapeutic cargos such as siRNAs or chemotherapeutic agents have shown potential in targeted tumor delivery. Their presence in bodily fluids and selective enrichment of tumor-specific cargo position them as valuable candidates for liquid biopsy applications, enabling non-invasive monitoring of disease progression and treatment responses. Furthermore, engineered TD-EVs are being explored as delivery systems for chemotherapeutics, RNA interference molecules, and gene-editing tools. Despite these advances, different challenges hinder the clinical translation of TD-EV-based applications. These include the heterogeneity of EV populations, lack of standardized isolation and characterization protocols, and difficulty in distinguishing TD-EVs from normal EVs in complex biological samples. Key obstacles also include the pronounced heterogeneity of EV populations and the lack of standardized isolation and characterization protocols. This review explores the multifaceted roles of TD-EVs in cancer biology and their potential utility in diagnosis, prognosis, and therapeutic intervention.

1. Introduction

Cancer remains one of the most formidable health challenges worldwide, accounting for millions of deaths annually despite significant advances in diagnostics and therapeutics [1]. Conventional cancer therapies, including surgery, chemotherapy, and radiotherapy, have improved patient outcomes, yet their efficacy is often limited by tumor heterogeneity, drug resistance, and adverse side effects [2]. Targeted therapies and immunotherapies have heralded a new era in oncology, offering more personalized and effective treatment options. However, the development of resistance and the complexity of tumor microenvironment (TME) continue to pose significant obstacles [3]. Thus, there is a pressing need to explore novel mechanisms underlying cancer progression and to identify innovative therapeutic strategies.

In recent years, extracellular vesicles (EVs) have emerged as critical mediators of intercellular communication in both physiological and pathological contexts, including cancer [4]. EVs are heterogeneous, membrane-bound nanoparticles released by virtually all cell types and are broadly classified into exosomes (30–150 nm), microvesicles (100–1000 nm), and apoptotic bodies (>1000 nm), based on their size, biogenesis, and release pathways [5,6]. These vesicles carry a diverse cargo of proteins, lipids, and nucleic acids, reflecting the molecular signature of their cell of origin, and are capable of modulating recipient cell behavior at local and distant sites [7].

The role of EVs in cancer biology has garnered substantial attention, as accumulating evidence demonstrates their involvement in various aspects of tumor progression. Tumor-derived extracellular vesicles (TD-EVs), particularly exosomes, facilitate communication between tumor cells and the surrounding stroma, immune cells, and distant organs, thereby orchestrating a multitude of processes such as migration, proliferation, angiogenesis, immune modulation, and metastasis [8]. For instance, exosomes derived from highly metastatic melanoma cells have been shown to educate bone marrow progenitor cells toward a pro-vasculogenic phenotype, thus promoting pre-metastatic niche formation [9]. Similarly, breast cancer exosomes enriched in miR-105 disrupt endothelial barriers and enhance metastatic dissemination [10]. In addition to promoting tumor progression, EVs can also exert anti-tumor effects under certain conditions, highlighting their dual role in disease modulation [11].

Exosomes, a well-characterized subtype of EVs, are particularly notable for their ability to regulate key cancer phenotypes. They have been implicated in enhancing tumor cell migration and invasion through the transfer of oncogenic proteins and RNAs [12]. Exosomal integrins, for example, have been identified as determinants of organotropic metastasis, directing cancer cells to specific distant sites [13]. Moreover, exosomes can promote angiogenesis by delivering pro-angiogenic factors such as vascular endothelial growth factor (VEGF) and miR-210 to endothelial cells, thereby facilitating tumor vascularization and growth [14]. Conversely, some exosomal cargoes can suppress tumor growth and metastasis, either by delivering tumor-suppressive miRNAs or by modulating immune responses [15].

The immunomodulatory functions of exosomes are particularly intriguing, as they can influence macrophage polarization and the overall inflammatory milieu within the TME. TD-EVs have been shown to induce both pro-inflammatory (M1) and anti-inflammatory (M2) macrophage phenotypes, thereby shaping immune responses in a context-dependent manner [16]. For example, exosomal miR-21 from lung cancer cells promotes M2 polarization, facilitating tumor progression, whereas exosomal miR-155 can enhance M1 polarization and anti-tumor immunity [17]. This dynamic interplay underscores the complexity of exosome-mediated communication and its impact on cancer pathogenesis.

The past decade has witnessed a remarkable surge in exosome-based research, driven by the recognition of their potential as diagnostic biomarkers and therapeutic agents [18]. The unique molecular signatures carried by exosomes offer promising opportunities for minimally invasive cancer diagnosis and prognosis, while their natural biocompatibility and ability to deliver therapeutic payloads position them as attractive candidates for drug delivery systems [19]. Several preclinical studies and early-phase clinical trials are currently exploring exosome-based strategies for cancer therapy, including the use of engineered exosomes for targeted delivery of anti-cancer agents and immune modulators [20].

Here, we will provide a comprehensive overview of TD-EVs, with a particular focus on their roles as mediators of cell-cell communication and their therapeutic potential. The succeeding sections will discuss the biogenesis, classification, and molecular cargo of EVs; their functional roles in cancer progression, metastasis, and immune regulation; recent advances in exosome-based diagnostics and therapeutics; and the challenges and future directions in harnessing EVs for clinical applications.

2. Biogenesis and molecular composition of TD-EVs

EVs have different biogenesis pathways and contain essentials to help either signal or proliferate other cells such as proteins, RNAs, DNAs, lipids, and other metabolites. There are three main biogenesis pathways for EVs such as exosomes, ectosomes, and apoptotic bodies [21]. The exosomes use inward budding of the endosomal membrane to form multivesicular bodies (MVBs). These MVBs can either fuse with lysosomes or with the plasma membrane, which degrades or releases exosomes respectively. Exosomes are EVs that rely more extensively on both endosomal and independent pathways, both of which are used for intraluminal vesicles (ILVs) formation [22]. ILVs are then used to degrade, recycle, or exocytose protein, lipids, and nucleic acids. Ectosomes are another EV that buds out of the plasma membrane but unlike exosomes, they don’t heavily rely on the ESCRT system [22]. So, to facilitate plasma membrane budding, the membrane must be rearranged more rigorously. The budding of the plasma membrane requires the cell cytoskeleton to be disassembled, with the help of caspase-2. The ectosome is then released through actomyosin abscission [21]; though this process is still heavily explored, ESCRT also plays an important role in ectosome discharge [22]. The apoptotic bodies are simpler, as they are discharged when the cell undergoes apoptosis. They are > 1000 nm in size and contain DNA fragments, histones, and immature glyco-epitopes which serve as biological tags for cell recognition [21].

2.1. Distinct features of tumor vs. normal cell-derived EVs

A major difference between normal and TD-EVs is the specific labels on the surface of each cell type (Table 1). TD-EVs normally carry oncogenes and nucleic acids on their surface to promote tumorigenesis [23]. Additionally, the EVs of malignant and healthy cells have drastically different proportions of nucleic acids and proteins. TD-EVs also use paracrine cells-to-cell signaling as their major way to crosstalk between cells [23]. TD-EVs are more protected, making them very prominent in drug delivery treatments like chemotherapy [24]. TD-EVs, in contrast to EVs, use vast techniques for cell-to-cell communications, so to distinguish them, we need to look at their contents and their surface markers. For example, EVs can exhibit both paracrine signaling for intercellular communication. To distinguish between a TD-EV and a normal EV, we would need to look at its surface marker (Table 1). TD-EVs includes integrin heterodimers [25]. Integrins are essential for cancer cell metastasis as they provide for its mobility by binding to components of the extracellular matrix. Integrins also regulate proliferation by regulating some protein growth within the cancer cells [25]. TD-EVs also activate a lot of different signaling pathways compared to normal EVs. Some pathways are the PI3K-AKT, MAPK, Wnt, and cGAS/STING pathways [26]. Using the PI3K-AKT pathways, TD-EVs could initiate polarization in macrophages. The benefit of macrophage polarization will be discussed in more detail later, but in short, PI3K-AKT activates signal transducer and activator of transcription 3 (STAT3), which in turn transforms a macrophage into its M2 phenotype [27]. The MAPK pathway plays an important role in the polarization of macrophages, but it also phosphorylates the p38, c-Jun N-terminal kinases (JNK), and ERK proteins [28]. These phosphorylated proteins, in turn, can promote proliferation, angiogenesis, resistance to apoptosis, and activation of other proteins. In the case of p38, it is proven to activate angiogenesis, but more importantly can activate a bunch of downstream pathways like MSK1 and MSK2, which regulate and control gene expression; MNK1 and MNK2, which regulate protein synthesis; and MAPK-activated protein (MAPKAP) kinase 2 (MK2) and MK3 which changes protein function and gene expression [29,30]. JNKs are considered Pro-directed Ser/Thr protein kinases stated by the structure of CDK-2 with cyclin [31]. Thus, if JNKs can be considered Pro-directed Ser/Thr protein kinases, then the implications of JNKs can be extrapolated to be pro-tumorigenesis. Pro-directed Ser/Thr protein kinases are known for inducing cancer cell formation. For example, in the formation of breast cancer, Pro-directed Ser/Thr protein kinases enhance the transcription of the cyclin D1 genes, upstream and downstream, through transcription factors - one of which includes the use of JNKs [32]. The D1 cyclin protein is considered a rate-limiting factor in the cell cycle, particularly during the G1 phase of the cell cycle. Therefore, if the D1 protein is overexpressed, it can potentially act as an oncogene, leading to tumorigenesis in breast cancer [33]. If the D1 gene is knocked down, then it will prevent the Ha-Ras or c-Neu/HER2 pathway from inducing breast cancer [32]. Finally, the ERK pathway is well known for both exhibiting tumorigenesis and suppressing tumorigenesis [34]. The ERK 1/2 pathways play a prominent role in regulating cellular signaling and are widely recognized for their involvement in the apoptosis pathway. In short, ERK pathways promote intrinsic or extrinsic apoptosis pathways through the release of cytochrome-C within the mitochondria or caspase-8 activation [35]. However, the ERK pathway also suppresses tumorigenesis by inhibiting RasV12-induced senescence [36]. The Ras signaling pathway is a prominent membrane signaling protein that is considered an oncogene; once phosphorylated, it can activate the MEK pathway, which in turn can activate a range of transcription factors [37,38]. But the RasV12-induced senescence is important in that it prevents early oncogenic cells from being malignant. RasV12-induced senescence depends on ERK signaling, and by blocking ERK, the cell can bypass the senescence and become malignant [36,37]. The ERK pathway can also be activated by growth factors, cytokines, hormones, and heat or oxidative stresses through receptor tyrosine kinases (RTKs) or epidermal growth factor receptors (EGFRs) [38].

Table 1.

Major differences between EVs from normal and tumor.

	Normal EVs	Tumor EVs	Ref.
Origin	Non-cancerous cells (e. g., immune, epithelial)	Cancer cells (e.g., breast, lung, glioma)	[39–42]
Surface Markers	CD9, CD63, CD81, ALIX, TSG101	CD24, CD81, ALIX, tumor-specific proteins (e.g., EpCAM, HER2, PD-L1)	[39, 43–46]
Significant Proteins	HSP70, Rab proteins	HSP90, MMPs, EGFRvIII, TGF-β, survivin, integrins (e.g., ITGβ3)	[40, 46–50]
Enriched miRNAs	miR-16, let-7 family (homeostasis)	miR-21, miR-155, miR-210, miR-9, miR-1246, miR-23a (oncogenic)	[39,45, 51,52]
Functional Roles	Intercellular communication, homeostasis, immune modulation	Promote invasion, angiogenesis, immune evasion, metastasis	[52–56]
Immune Effects	Can enhance antigen presentation and immune tolerance	Induce Treg, inhibit NK and T cells, carry PD-L1, induce M2 macrophages	[56–59]
Effects on Recipient Cells	Maintain tissue repair, promote immune surveillance	Reprogram metabolism, promote TME, increase drug resistance	[60–62]
Diagnostic Value	Baseline physiological profile	Liquid biopsy marker: e. g., for glioblastoma (EGFRvIII+)	[56,63, 64]
Therapeutic Implications	Potential for regenerative therapies	Target for therapy, biomarker for progression, therapy resistance	[53,56, 58,65]

2.2. Biogenesis and selective cargo loading in EVs

A critical initial step in EV biogenesis is the selective loading of cargo, which occurs via two primary mechanisms: ESCRT-dependent and ESCRT-independent pathways (Fig. 1). The ESCRT-dependent pathway mainly facilitates the incorporation of proteins, whereas ESCRT-independent mechanisms enable the packaging of nucleic acids, lipids, and additional proteins [66]. Following selective cargo loading, ILVs form within MVBs. Exosome release is initiated when MVBs fuse with the plasma membrane, a process regulated by the activation of small Rab GTPases, including RAB27a/b, RAB11, RAB7, and RAB35. These GTPases orchestrate MVB trafficking and docking at the plasma membrane, resulting in the subsequent budding and secretion of exosomes [22,67]. The selective cargo loading in EVs plays a crucial role in each individual EV’s function. As the name suggests, selective cargo loading is the process by which materials are loaded into EVs. This mechanism is still being discovered, so the totality of this mechanism is not fully understood. Given the materials inside of these EVs, which are mostly membrane receptors, it is suspected that plasma membrane shedding or budding contributes to this mechanism [22]. Besides the membrane receptors, we have identified other mechanisms such as ESCRT-dependent and independent pathways including nucleic acid and lipid loading pathways (Fig. 1). Ubiquitin uses three enzymes, E1 (ubiquitin activating enzyme), E2 (ubiquitin conjugating enzyme), and E3 (ubiquitin-protein ligase). E1 first activates the ATP-dependent ubiquitin. Then E2 forms a complex with the ATP-dependent ubiquitin, and E3 then binds the ubiquitin to the proteins [68]. After the ubiquitin tags the proteins, the ESCRT mechanism uses the ubiquitinylated proteins to form ILVs. The ESCRT family consists of ESCRT-0, ESCRT-I, ESCRT-II, ESCRT-III, and the ESCRT-associated protein, all of which play a crucial role in the ESCRT mechanism (Fig. 1). ESCRT-0 first recognizes the ubiquitinylated proteins through ubiquitin-binding domains (UBDs) and drags them to the endosome membrane. Then, ESCRT-0 calls ESCRT-I by binding to its PSAP-like motifs through ESCRT-I’s ubiquitin’s E2 variant domains. Subsequently, ESCRT-I triggers ESCRT-II to induce membrane deformation [68]. ESCRT-II then calls over ESCRT-III for the scission and acceleration of endosomal membrane closing and vesicle budding. This process thus facilitates ILV formation [67].

Fig. 1.

Mechanisms of EV biogenesis via distinct cellular pathways. EV biogenesis occurs, through ESCRT-independent and ESCRT-dependent pathways, in healthy cells and in tumor cells; however, what differs is the quality, and quantity of output for healthy cells and tumor cells. For example, ILVs of the multivesicular bodies in healthy cells consist mainly of homeostasis microRNAs (i.e. let-7, miR-16, miR-24), markers CD9 and CD63, that mediate tissue homeostasis and regulate immune cells. Alternatively, tumor cells enrich EVs with oncogenic cargo (i.e. miR-21, miR-155, miR-210) and signaling proteins (e.g. phospho-STAT3 and β-catenin), that promote tumor development and cellular communication between tumor cells in the tumor microenvironment [69]. Tumor cells appear to excrete many orders of magnitude greater for EVs than healthy cells ones, mainly from apoptotic bodies, along with greater pro-tumorigenic signals. The dysregulation of EV-cargo and production demonstrates how TD-EVs have a significant role in reshaping the extracellular landscape for malignancy [70].

There are ubiquitin-like modifiers (Ubls), including small ubiquitin-like modifier (SUMO) and ubiquitin-like protein 3 (UBL3), which act as post-translational regulators that influence EV cargo selection. SUMOylation of RNA-binding proteins such as heterogeneous nuclear ribonucleoproteins (hnRNPs) facilitates the ESCRT-independent sorting of specific microRNAs (miRNAs) into exosomes by recognizing EXO-motifs in their sequences [68]. The ESCRT-independent pathways include nucleic acid and lipid pathways. The main mechanism for miRNA packaging is through RNA-binding proteins hnRNPs [68]. bnRNPs contain multiple RNA recognition motifs (RRMs) and other nucleic acid–binding domains that allow them to bind and regulate RNA cargo selection. A common hnRNP is called HnRNPA2B1, which regulates the sorting of specific miRNAS. They are first SUMOlyated and bind to EXO motifs, which regulate the sorting of miRNAs into exosomes [68]. However, research on the nucleic acid pathways remains limited, as several nucleic acid mechanisms — such as RNA-binding proteins, the KRAS status, and neutral sphingomyelinase 2 — are still not entirely understood. For lipid sorting pathways, there are many uncertainties on how lipids are sorted into exosomes, but research shows that lipid sorting complexes are independent of protein sorting. Lipid sorting appears to be more closely related to the size and quantity of lipids being sorted, as demonstrated by studies on the role of myelin proteolipid protein (PLP). Exosomes enriched in proteolipid protein (PLP), such as those derived from oligodendrocytes, exhibit higher lipid content, suggesting lipid-sorting mechanisms distinct from protein sorting [68,71].

In addition to ESCRT-mediated mechanisms, tetraspanins such as CD9, CD63, CD81, and CD82 promote ESCRT-independent ILV formation by organizing tetraspanin-enriched microdomains (TEMs) on endosomal membranes. They play a key role in sorting proteins into exosomes, including integrins, ICAM-1, IGFS-8, major histocompatibility complex class II proteins, and syndecan21. It also makes tetraspanins-enriched microdomains (TEM) by interacting with other tetraspanins and integrins to form ILVs [71]. The surface of exosomes can act as cargo with transmembrane proteins scattered around the surface such as flotillin 1 and 2, IL-6R, EGFR, T-cell receptor, chimeric antigen receptor (CAR), GPCR receptors, PD-L1, transforming growth factor (TGFB), and ADAM proteases [22]. These transmembrane proteins play an important role in scaffolding functions, which are used as anchorages in the extracellular membrane (ECM) [24]. Other major glycosylphosphatidylinositol (GIP) scaffolding anchors present are proteoglycans, glypican-1, DAF, and MAC-IP. In the inner phospholipid bilayer of the EVs membranes, we can find small GTPases that function in the biogenesis of EVs.

3. Functional roles of tumor-derived EVs

3.1. NK and T cell suppression or stimulation

Primarily, TD-EVs mediate communication between tumor cells and components of the tumor microenvironment, promoting tumor growth, angiogenesis, and immune modulation. (Fig. 2). Tumor growth is promoted by the TD-EVs and immune cell EVs to regulate antitumor immune responses [72]. Since TD-EVs can be used to regulate immune responses, this is especially helpful as a biomarker when looking at immune responses toward cancer cells. Although the exact mechanisms remain incompletely understood about the mechanisms in which they regulate immune responses; there are, however, a few proposed mechanisms like immune suppression or stimulation. For immune suppression, tumor cells can use TD-EVs to suppress NK cell cytotoxicity by transferring NKG2D ligands (MICA, MICB, and ULBPs) that downregulate the activating NKG2D receptor on NK cells. The NKG2D interacts with the ligands MIC-A, MIC-B, and UL-16 binding proteins so TD-EVs with NKG2D ligands decrease the expression of NKG2D on NK cells to avoid the immune response [72]. TD-EVs impair CD8⁺ T-cell function by interfering with the PI3K/AKT signaling pathway, leading to reduced survival and apoptosis of cytotoxic T cells. TD-EVs can also promote immune responses, but this mechanism is not as clear as immune response suppression. EVs derived from glioma do activate a tumor-specific T cell immune response in vivo. One possible explanation could be that EVs stimulate the upregulation of costimulatory receptors CD80, CD86, and MHC II. These receptors can provide a signal to the T cell CD28 receptor which activates the T cells [73].

Fig. 2.

Tumor-derived EVs promote diverse cancer progression processes via Communication. EVs derived from tumors function in important ways to impact multiple recipient cell types to develop immunosuppressive and pro-metastatic tumor microenvironments. The exosome cargo generally contains a mix of immunosuppressive molecules (such as miR-23a, miR-20a, miR-93, PD-L1, TGF-β, CD39/CD73, and MICA/B/ULBPs) that can engage and block the interaction of NK cell receptor (NKG2D) and suppress NK cell cytotoxic activity. cargo (STAT3, miR-361–3p and PI3K/AKT) can change macrophage-like cells to be more M2-like and continue the process of immune evasion and tumor progression. EV cargo can also influence epithelial cells to undergo EMT which is clearly important in cancer invasion and metastasis. In summary exosomes from tumors suppress T cell activation, induce angiogenesis, allow and promote metastasis, and taken together these findings aid in the development of a tumor-immune environment that provides robustness and aggressiveness of cancer.

3.2. Promotion of M2 macrophage polarization and angiogenesis

With M2 macrophages’ primary production consisting of anti-inflammatory factors like IL-10 and TGF-β, M2 macrophages promote immunosuppression, angiogenesis, and metastasis (Fig. 2). TD-EVs interact with macrophages within the TME, promoting their polarization toward the M2 phenotype and enhancing the secretion of anti-inflammatory cytokines such as IL-10 and TGF-β. They primarily promote macrophage polarization within the TME with miRNAs, IncRNAs, circRNAs, and proteins. The miRNAs regulate the macrophages through the PTEN/PI3K/AKT signaling pathway, which activates STAT3 - a crucial differentiation factor for macrophages to become M2 macrophages. Furthermore, miRNAs from colorectal cancer (CRC) can adjust the PTEN mutation to enhance M2 macrophage polarization. lncRNAs within TD-EVs can also induce M2 polarization by activating the STAT3 signaling pathway. For example, lncRNAs HMMR-AS1 can interact with miR-147a to decrease ARID3A degradation, which promotes M2 polarization of macrophages. lncRNA TP73-AS1, highly expressed in nasopharyngeal carcinoma cell-derived EVs, also promotes M2 macrophage polarization by binding to miR-342–3p. circRNAs also can polarize M2 macrophages by the same PTEN/PI3K/AKT pathway but within non-small cell lung cancer (NSCLC). Circ_0001142 from breast cancer cell’s EVs increases M2 polarization through the miR-361–3p/PIK3CB pathway [27]. Finally, EV-associated proteins such as CSF-1, MCP-1/CCL2, EMAP2/AIMP1, and LTA4H also contribute to promoting M2 macrophage polarization.

Angiogenesis is also promoted by TD-EVs. For example, in head and neck squamous cell carcinoma (HNSCC), In head and neck squamous cell carcinoma (HNSCC), EV-associated transforming growth factor-β (TGF-β) promotes angiogenesis by stimulating endothelial and stromal cells [24]. TGF and EVs can promote macrophage chemotaxis by facilitating angiogenesis through primary macrophages’ pro-angiogenesis factors. Also, a decreased expression of miR-TSGA10 supports angiogenesis. Likewise, EV-derived miR-494 and miR-142–3p from oral squamous cell carcinoma (OSCC) promote angiogenesis by activating endothelial nitric oxide synthase (eNOS) and TGFBR1 signaling pathways [24].

3.3. Metastasis and Pre-Metastatic Niche Formation

Along with the promotion of angiogenesis, TD-EVs also promote EMT (Fig. 3). EMT is the process by which epithelial cells lose their polarity properties, allowing them to become mesenchymal stem cells (MSCs). This can cause tumorigenesis, among other processes. One of the major causes for EMT is through EVs, and by extension, TD-EVs. The two major pathways for EVs to induce EMT are the Hippo and β-Catenin Signaling Pathways. Starting with the Hippo pathway, exosomes from the MSCs can activate the Hippo pathway to induce EMT in breast cancer cells. However, the Hippo pathway also regulates the activation of non-phosphorylated YAP/TAZ, which subsequently binds to TEA domain transcription factors (TEADs). TEAD activation promotes the transcription of mesenchymal marker genes, such as Myc, CTGF, Cyr61, AREG, and AXL transcription factors, RTK, and proteins (Fig. 3), while repressing epithelial cell markers. The β-Catenin Signaling Pathway is another prominent pathway that can trigger EMT. β-Catenin activates the SNAIL, SLUG, and TWIST transcription factors which contribute to the activation of EMT [74]. Also, β-Catenin activates the TCF/LEF transcription factors which further induces EMT for mesenchymal cells (Fig. 3). The major figure that triggers the β-Catenin Signaling Pathway is miR-301a, from hypoxic glioblastoma (GBM) cells, and attacks the transcription elongation factor A like 7 (TCEAL7). TCEAL7, in turn, activates the β-Catenin Signaling Pathway [74]. miR-301a also inhibits p63 and releases ZEB1/2, which also induces EMT. In lung adenocarcinoma cancer, miR-1260b serves as a suppressor of SOCS6 (suppressor of cytokine signaling 6) and a KIT proto-oncogene activator. KIT is known for inducing EMT through the KIT-MEK-ERK pathways.

Fig. 3. Hippo signaling pathway versus β-catenin.

When the Hippo pathway is active, LATS1/2-mediated phosphorylation retains YAP/TAZ in the cytoplasm, preventing their nuclear translocation and downstream transcriptional activation. Active Hippo signaling also suppresses Wnt/β-catenin activity by inhibiting β-catenin stabilization and transcriptional cooperation. In contrast, when Hippo signaling is inactive, dephosphorylated YAP/TAZ translocate to the nucleus and bind TEAD transcription factors, while stabilized β-catenin associates with TCF/LEF to drive gene transcription and more importantly drives EMT. TEAD also transcribes for Myc, CTGF, Cyr61, AREG, and AXL transcription factors, RTK, and proteins which also drive EMT. Together, these coordinated transcriptional programs regulate proliferation, EMT, and tumor progression.

Since we have established the known ways for TD-EVs to promote angiogenesis and induce EMT, we can see how being able to trigger both of those can also promote metastasis. TD-EVs can directly interact with epithelial cells to transport EGFR and TGF to promote angiogenesis [27]. Angiogenesis is then able to provide support and nutrients for the tumor cells, as well as other benefits. By contrast, EMT facilitates tumor cell detachment from the epithelial cell/tissue, thereby contributing to cancer metastasis [75]. Also, since EMT produces the SNAIL1, SNAIL2, ZEB1, and ZEB2 transcription factors, the SNAIL transcription factor family is known to have anti-immune responses. For example, SNAIL in B16 melanoma cells can inhibit dendritic cell maturation and Treg-like CD4 + Foxp3 + cells [47].

3.4. EV-Driven Modulation of the Neural Niche in the PNS

Evidence suggests that neural regulation of cancer in the TME should be considered as a hallmark of cancer [76–78]. Low pH conditions in the TME facilitate the release and uptake of EVs [52,79]. Aside from this, EV-based communication in the neural microenvironment is also common due to the inability of the neural soma to supply the demands of anentire neuron, especially long-distance transport to the axon [54,80]. Moreover, it has been previously reported that EVs play an important role in cancer-nerve interactions [81–83],. More specifically, there are two main ways in which EVs partake in cancer-nerve crosstalk and facilitate tumorigenesis: TD-EV-delivered miRNAs induce nerve reprogramming and neuritogenesis (growth and development of dendrites and axon [59,84,85]. TD-EV-delivered axon guidance proteins promote sensory nerve neuritogenesis [62,86]. Studies on the therapeutic potential of nanovesicles in this field also exist. However, liposomes (artificial vesicles composed of lipid bilayers) have been the focus; EVs are yet to be investigated [87].

It has been found that TD-EV-delivered miRNAs (combination of increasing miR-21, miR-324, and a lack of miR-34a) induce nerve reprogramming and neuritogenesis, which facilitate tumorigenesis [84, 85]. Amit et al. showed that EVs from mutant p53 OSCC cells promote adrenergic neuritogenesis, which in turn promotes tumorigenesis. They found that these TD-EVs lacked miR-34a but presented neuritogenic miRNAs, such as miR-21 and miR-324, which resulted in the reprogramming and trans-differentiation of sensory neurons into adrenergic ones. This outcome was absent in EVs from wild type p53 tumors. Injecting mutant p53-derived EVs into wild-type tumors enriched adrenergic innervation and tumor growth. Interestingly, blocking sensory, but not adrenergic, nerves prevented this effect, indicating that adrenergic innervation arises from reprogrammed sensory neurons. These findings unveil an EV-mediated mechanism by which cancer manipulates neural plasticity to boost tumorigenesis [84,85,88].

Previous studies have also shown that TD-EV-delivered axon guidance proteins promote sensory nerve neuritogenesis, which facilitates tumorigenesis [86,89]. Vermeer et al. demonstrated that EVs from HNSCC promote the outgrowth of sensory nerves (β-III tubulin⁺/TRPV1⁺) in both clinical and mouse models. TD-EVs and serum-derived EVs significantly increased neurite outgrowth in PC12 cells, a phenomenon successfully replicated using EVs from the mEERL mouse model [62,86]. Notably, blocking EV release or eliminating TRPV1⁺ sensory nerves in vivo significantly reduced tumor growth, suggesting that EV-induced neuritogenesis is a driver of tumorigenesis [89,90]. While the axonal guidance protein EphrinB1 amplified neuritogenesis via Eph receptor signaling, it was neither necessary nor sufficient alone since blocking or deleting EphrinB1 did not significantly inhibit neurite outgrowth [86,89]. These findings indicate that EV-mediated sensory innervation may represent a tumorigenic mechanism across multiple cancer types.

A research team designed liposomes loaded with bupivacaine, a non-opioid selective sodium channel blocker, to disrupt cancer-nerve communication in TNBC [91,92]. By delivering these liposomes intravenously, researchers successfully impaired neuritogenesis and tumor-associated neural networks, which inhibited tumor growth, invasion, and metastasis [92]. Although the study targeted nerve–tumor crosstalk, a process heavily mediated by EVs, and despite liposomes and EVs sharing structural and functional similarities, the paper did not discuss EVs directly. This gap hints at the untapped potential of EVs as a therapeutic vehicle in this area. Compared to synthetic liposomes, EVs offer several advantages. As EVs are naturally occurring, they show promising safety and stability profiles [87,93]. Likewise, they do not cause cell death or inflammation upon repeated administration due to their biocompatibility and low immunogenicity [87,93]. In certain situations, some EVs could also cross the blood-brain barrier (BBB), as opposed to liposomes, also makes them better candidates for neuromodulatory strategies in cancer therapy [71,94]. Overall, the evidence supports the idea that EVs actively contribute to cancer-nerve interactions by driving neuritogenesis and remodeling neural networks of the TME [85,86,89]. These changes not only support tumor growth but may also influence invasion and metastasis. Despite compelling evidence for EV-mediated neural modulation, current therapeutic strategies have focused on synthetic vesicles such as liposomes. Given their natural origin, safety profile, and potential to target neural components of the TME, EVs represent a promising and underexplored tool for disrupting nerve-tumor communication, essential to lower tumorigenesis.

4. Diagnostic and prognostic Potential of TD-EVs

4.1. Application of TD-EVs in liquid biopsy

The main applications of EVs would be either as a diagnostic or a drug delivery vehicle (Fig. 4). With new in vivo studies demonstrating controllable production of bioactive EVs for targeted therapies such as treatment for myocardial infarctions and gut microbiota modulation, it shows promise that the same can be done in the field of cancer [95,96]. TD-EVs carry molecular signatures that mirror the genetic and proteomic landscape of their parent cells leading them to be a good contender for monitoring and diagnosing various diseases. Because of their presence in the blood, a liquid biopsy may be taken through either urine or blood, allowing for continuous monitoring of a patient [6]. The benefits of a liquid biopsy would be that the procedure is minimally invasive, easily accessible, and time efficient. EVs are shown to exhibit exceptional structural stability and are readily enriched in biofluids, with their circulating abundance closely correlating with tumor burden, thereby serving as a robust surrogate marker of disease progression. Thus, samples taken from blood samples should accurately reflect the status of the body. Similarly, urine reflects what type of waste is being excreted from the body providing data points for what is happening metabolically. The decision to use either blood or urine depends on what is being studied but both present good options for sample retrieval. Recently, it has been shown that TD-EVs mirror the composition of cancer cells leading them to be used as biomarkers [97]. These vesicles provide a minimally invasive means for diagnosis and prognosis due to their stability in bodily fluids and enrichment with tumor-specific biomarkers [98]. Their presence can correlate with tumor burden and disease stage, suggesting strong potential in early detection and monitoring disease progression [99]. An example of this was found when a TD-EVs was found carrying fibronectin, which when bound, promotes anchorage independent growth which is a hallmark of tumorgenesis [55]. Liquid biopsy technologies leverage biofluids such as blood, saliva, and urine to detect cancer-associated biomarkers without the need for invasive tissue sampling [100]. Among the most promising components of liquid biopsies are circulating EVs. Their abundance in plasma and content reflective of cancer cells make them ideal candidates for detecting malignancy [6,101]. Additionally, their prevalence could make them useful for monitoring of the tumor in term of malignancy [97]. This is particularly relevant in liquid tumors—such as leukemias, where malignant cells interface directly with the circulatory system—where EV-based assays can offer high sensitivity for real-time disease assessment [54]. Because of the lipid bilayer morphology of EVs, they can be stably stored indefinitely at low temperature (−80 °C), allowing for that taken sample to be further researched and provides a longer working time [102]. Because EVs can be easily extracted from a patient through minimally invasive testing with the extracted sample having higher stability, if proven to work as biomarkers, this could create a new technology used for diagnostic purposes.

Fig. 4.

Theranostic potential of EVs in cancer. EVs are emerging as pivotal mediators in oncology, enabling targeted drug delivery across tissue barriers, serving as vehicles for RNA-based therapies and immunomodulation, and providing diagnostic biomarkers through their molecular signatures in liquid biopsies. Additionally, EVs hold potential as cancer vaccines by presenting tumor-specific antigens to activate the immune response.

4.2. EVs as biomarkers for early detection and monitoring

The diagnostic utility of EVs lies in their cargo, especially proteins, lipids, and RNAs, which function as biomarkers (Fig. 4). These include cancer-type-specific miRNAs and protein signatures that can be isolated and quantified [98]. Technologies such as ELISA, Western blotting, and mass spectrometry allow for bulk detection, while newer single-EV analysis platforms, like Multiplexed Analysis of Single Extracellular Vesicles (MASEV), enhance sensitivity and specificity [99]. Notably, surface proteins on EVs can be directly probed, making them amenable to multiplexed diagnostic approaches. EVs are naturally present in nearly all bodily fluids, but they are particularly enriched in blood plasma, making this biofluid a practical medium for EV-based diagnostics [6]. However, not all EVs in circulation originate from tumor cells and so, enrichment techniques or selective marker analysis is required to distinguish TD-EVs from those secreted by healthy tissues [103]. In the context of hematologic malignancies, such as leukemias or lymphomas, this barrier is lower due to the direct shedding of tumor EVs into the bloodstream [54].

For EVs to be used effectively in diagnosis, especially via liquid biopsy, the need to isolate TD-EVs is particularly important (Fig. 4). This is more feasible in liquid tumors where malignant cells are shed directly into circulation. Solid tumors present a greater challenge due to the dilution of TD-EVs in a vast background of vesicles from healthy tissues. Techniques targeting tumor-specific surface proteins or nucleic acid cargo can help address this specificity concern [101]. EVs have demonstrated high potential as diagnostic tools, particularly in detecting early-stage cancers. By profiling EV-associated biomarkers, such as tumor-specific miRNAs and proteins, clinicians can achieve minimally invasive early detection and monitor therapeutic response [99]. These vesicles offer temporal snapshots of tumor evolution, helping in dynamic treatment planning and minimizing the need for repeated tissue biopsies. This works by capturing the circulating EVs and then investigating their cargo. It was found in a non-small cell lung cancer study, the miRNA cargo retrieved from the EVs showed that there is an association between miRNA concentration and immunotherapeutic response in patients [104]. Circulating EVs can also be used as biomarkers for drug response. Specifically, the expression of long-non-coding RNA, miR-21–5p, and miR-486–3p are associated with drug resistance because of miR-21’s ability to target the apoptotic protease activating factor 1 preventing programmed cell death from occurring [104]. Surface expression of proteins on EVs also offers extra insight into cell-cell communication such as with the case of programmed death protein 1 ligand (PD-L1), a protein that when bound, blocks tumor cells from being detected by the immune system, allowing for their proliferation. It has been demonstrated in previous research that PD-L1 has been expressed on the surface of small EVs [105]. It was found that the PD-L1 on the sEVs did in fact inhibit T cell activation and cytokine production of activated T cells despite slightly enhancing the apoptotic abilities of an already activated T cell. These findings are consistent with PD-L1 working to progress tumor growth and evade death.

In terms of finding new biomarkers, recent advancements have been made in technology allowing for discovery. While not a biomarker specific technology, a localized surface plasmon resonance biosensor with self-assembly gold nanoislands developed by Thakur et al. was successfully able to distinguish EVs from microvesicles (MVs), differentiating the EVs from the MVs isolated from A-549 cells, SH-SY5Y cells, blood serum, and urine from a lung cancer mouse model [106]. This technology serves to ameliorate finding EVs which can then be further researched to see if they can be utilized as biomarkers. A detection method specifically for EVs has been created utilizing biotinylated antibody-functionalized TiN (BAF-TiN) to quantitatively detect EVs within the 30–200 nm range. This biosensor allows for high performance label-free sensing with a detection limit in the thousandths showing promise for detecting exosomal biomarkers [106–108]. Utilizing glioma’s association with glycolytic reprogramming, researchers were able to induce a state of hypoxia to investigate the changes in monocarboxylate transporter 1 (MCT 1) and cluster of differentiation 147 (CD147) which push out lactate for energy production. This hypoxia enhanced the release of exosomes which were found to have high amounts of MCT1 and CD147 making them candidates for biomarkers to track glioma’s metabolic reprogramming which in turn tracks malignancy [107,108]. The study of energy usage by EVs has also been vastly helpful in understanding their proliferation in the TME. Recently, it has been found that exosomes can produce adenosine triphosphate (ATP) [109]. Exosomes do this by incorporating some mitochondrial enzymes or parts of the electron transport chains with it then membrane thus, the exosomes could have some respiratory chain proteins. Accumulation of ATP and lactate, a byproduct of glycolysis which can be turned into energy depending on the state of the body, may provide the energy needed for exosome transfer into cancer cells [110].

The cargo carried by TDEVs as well as the presence of TDEVs themselves are used as biomarkers and trackers of cancer. In the case of glioblastoma, there is promise in tracking TDEVs as many are released from glioblastoma allowing researchers to track the progress of glioblastoma in a patient and use the presence of TDEVs as a method to diagnose glioblastoma early [111]. Xu et al., 2021 were successfully able to track the progression of glioma using a TiO₂-CTF enhanced-gold nanoislands biosensor to the BIGH3 receptor where the presence of exosomes was tracked in response to temozolomide, a possible treatment. Through this method, temozolomide was shown to have an anti-cancer effect [112]. In a similar study, Qu et al., 2021 were able to isolate and inspect the cargo contents of exosomes from oral tongue squamous cancer in nodal and non-nodal phenotypes finding 43 proteins in common and 136 proteins collectively from either type of cancer [113]. These proteins can be further studied to see how well they would function as biomarkers.

4.3. EV cargo across cancer types

Despite their usefulness, not all exosomes are disease specific. Many participate in physiological processes such as tissue repair, immune regulation, and intercellular communication [6]. In cancer diagnostics, it is essential to identify and isolate exosome subsets that originate specifically from tumor cells. Without this specificity, bulk exosome analysis may yield false positives or lack diagnostic precision [98]. Depending on the type of cancer, small EVs have great potential for diagnostic purposes. As previously stated, PD-L1 has been found on the surface of TD-EVs. In lung cancer, it was found that there was a correlation between the higher expression of PD-L1 on small TD-EVs from cancerous tissue versus healthy tissue [114]. However, the same cannot be said for pancreatic cancer where it was found that the difference between small EV PD-L1 expression from cancer tissue and non-cancer tissue was insignificant [114]. Still, TDEVs are being increasingly explored as early detection biomarkers as they show much promise in many cancer types. Their cargo, particularly miRNAs and proteins, can signal tumor presence before clinical symptoms manifest. In one study, long-noncoding RNA obtained from breast cancer patients EVs was able to correctly distinguish healthy patients from cancer patients [115]. Moreover, changes in EV concentration and composition over time reflect disease progression and response to therapy, positioning EVs as powerful tools for both baseline diagnosis and ongoing patient monitoring [54].

Beyond diagnostics, EVs play pivotal roles in mediating intercellular communication by transferring functional proteins, lipids, and RNA molecules between cells [6](Table 2). In cancer, they facilitate processes like immune evasion, angiogenesis, and metastasis. Therapeutically, this communication pathway can be hijacked as engineered EVs can deliver drugs, siRNAs, or CRISPR components directly to target cells, offering a novel platform for precision medicine [98]. It is essential to note that TD-EVs are directly secreted by malignant cells. These vesicles exhibit distinct compositions compared to those produced by stromal or immune cells within the TME. Consequently, their isolation and analysis can provide a more accurate molecular fingerprint of the cancer, enabling precise diagnostic and therapeutic interventions [101].

Table 2.

Comparison of major EV isolation methods and their implications for biomarker discovery and therapeutic applications.

Isolation Method	Principle	Advantages	Limitations	Impact on Biomarker Discovery	Impact on Therapeutic Applications	Typical Use Context	Ref.
Differential Ultracentrifugation (UC)	Size- and density-based sedimentation at sequential centrifugal forces	Widely used; low cost; compatible with large volumes	Low purity; co-isolation of protein aggregates and lipoproteins; time-consuming	May introduce contaminants that confound downstream omics analysis; limited reproducibility	Potential vesicle deformation under high g-forces can reduce integrity for therapeutic loading	Gold-standard laboratory method for proof-of-concept studies	[116–119]
Density Gradient (DG)-UC	Separation by buoyant density using sucrose or iodixanol gradients	Higher purity than UC; better subtype resolution	Labor-intensive; low throughput	Enhances proteomic and miRNA biomarker specificity	Improved preservation of intact EVs; limited scalability for GMP production	Preclinical biomarker validation	[119–121]
Size-Exclusion Chromatography (SEC)	Separation by column-based molecular size filtration	Gentle on vesicles; preserves biological activity; reproducible	Lower yield; dilution of sample; may miss smallest EVs	High purity facilitates reproducible biomarker profiling	Ideal for functional assays requiring intact EVs; scalable for clinical-grade prep	Diagnostic biobanking; EV functional studies	[65,119, 122,123]
Affinity Capture (e.g., CD63-, CD9-, CD81-based)	Immunoaffinity binding to surface markers	High specificity; can isolate TD-EVs	Expensive; biased toward known surface markers; low yield	Enables identification of tumor- or cell-type-specific EV markers	Yields highly defined vesicles for precision delivery; limited scalability	Targeted liquid biopsy or engineered EV studies	[65,118, 124]
Precipitation (e.g., PEG-based kits)	Polymer-induced EV aggregation and precipitation	Simple; rapid; suitable for clinical throughput	High contamination with non-EV proteins; inconsistent reproducibility	Useful for preliminary screening but may obscure quantitative biomarker signatures	Suboptimal for therapeutic use due to impurities and poor cargo stability	High-throughput exploratory biomarker screens	[65,117, 118]
Microfluidics-based Isolation	Microchannel filtration, electrophoresis, or acoustic trapping	High precision; integrates isolation and detection; small sample volumes	Requires specialized equipment; limited large-volume capacity	Enables single-vesicle analysis and multiplexed biomarker profiling	Potential for on-chip EV engineering or drug loading; still early-stage	Emerging liquid biopsy and point-of-care platforms	[65,125, 126]
Tangential Flow Filtration (TFF)	Pressure-driven filtration using semipermeable membranes	Scalable, gentle, GMP-compatible	Expensive instrumentation; may cause partial loss of small EVs	Suitable for large-scale EV production and consistent biomarker recovery	Preserves vesicle integrity for clinical-grade therapeutic applications	GMP manufacturing of therapeutic EVs	[127–129]

5. Therapeutic applications of TD-EVs

5.1. Targeting tumor-derived EVs

Tumor-derived EVs play a key role in tumor progression, immune suppression, and metastasis. Inhibiting the biogenesis or secretion of these vesicles has emerged as a strategy to disrupt intercellular communication network [130]. Small molecule inhibitors targeting components of the ESCRT machinery, neutral sphingomyelinase (nSMase), or Rab GTPases can reduce EV release from tumor cells [54]. Such inhibition may limit tumor-supportive signaling and sensitize tumors to conventional therapies. Another approach prevents EV uptake by recipient cells, thereby halting downstream signaling cascades that facilitate tumor growth or immune evasion. Strategies to block this uptake include competitive inhibition with synthetic ligands, masking surface adhesion proteins on EVs, or using antibodies against known uptake mediators [131]. Blocking EV-cell interactions could mitigate the systemic effects of tumor-derived EVs without affecting EV production ^132,133.

The natural biocompatibility and stability of EVs make them advantageous vehicles for targeted therapeutic delivery. EVs can be engineered to carry chemotherapeutic drugs, RNA molecules (e.g. siRNA, miRNA), or gene-editing tools such as CRISPR/Cas9 [132]. These engineered vesicles can deliver payloads across biological barriers while evading immune detection, increasing drug accumulation at the tumor site and reducing systemic toxicity. TD-EVs are isolated using techniques including but not limited to ultracentrifugation and size exclusion chromatography. After the isolation process, drug loading may begin utilizing endogenous or exogenous loading methods. Endogenous loading is best for small hydrophobic drugs as they can passively pass through the EV membrane during incubation, keeping the properties of both the drug and EV unchanged. Exogenous methods are used for larger or hydrophilic drugs to enter the EV. However, methods used to introduce these drugs such as sonication can change the physical properties of the EV and vesicle fusion also pose problems. Once induced, extra steps can be taken for specialized targeting [133]. To improve specificity, EVs can be modified with tumor-targeting ligands, peptides, or antibodies on their surface. For example, the addition of RGD peptides or tumor-specific monoclonal antibodies enhances EV binding to cancer cells [134]. Another strategy involves engineering donor cells to overexpress membrane proteins that naturally home to tumor tissues. These design enhancements ensure preferential uptake by malignant cells and minimize off-target effects.

5.2. EV-based drug delivery systems

One specific area of concern is the issue of passing through the BBB. Because of their small size, exosomes can pass through negating this issue. This makes them a good vehicle for brain diseases such as glioma. Using a microfluidics device called Exo-load, Thakur et al., 2020 were able to successfully load anticancer drugs doxorubicin (DOX) and paclitaxel (PTX) into EVs derived from glioma cells thus incorporating the glioma cell (glioblastoma) model system. Although the success rate in loading was not optimal with 20 % and 8 % observed for DOX and PTX respectively, it does prove that this method is possible. Additionally, the researchers found that exosomes derived from autologous tissue had a much better targeting than cell line derived exosomes suggesting the combination of drug-loaded patient derived TDEVs and the Exo-load method could be used together to optimize results [135]. Another example of this applies to Alzheimer’s Disease where engineered exosomes were able to successfully pass through the blood brain barrier and target the amyloid-β precursor protein (APP) expressing neuron cells. This blocks the pathogenesis of Alzheimer’s disease and even seemed to trigger autophagy in those same cells improving cognitive abilities [136].

5.3. Application of AI/ML in EV based theranostics

Artificial intelligence (AI) and machine learning (ML) are being applied to various biomedical research for optimized sorting, predictions, and possible treatments, including in the context of EVs (Table 3) [137,138]. An AI based ML algorithm was trained using models like support vector machines and decision trees to successfully identify the initiating parent cell (IPC) of nano-vesicles which would allow for easier identification of a sample of EVs [139,140]. Though still in its early stages of development, a quantum machine learning (QML) electrokinetic mining system was able to successfully differentiate between nano-vesicles and exosomes based solely on electrokinectics such as zeta potential to bypass traditional more labor-intensive methods such as transmission electron microscopy which was able to accurately identify the different types of vesicles and where they had originated from. Additionally, this system successfully distinguished nanoparticles and exosomes in plasma samples from colorectal cancer patients and experimental mice, despite being trained on a limited dataset, indicating its potential for broader exploratory applications [141].

Table 3.

Biomedical studies employing the application of AI/ML.

Study	AI/ML models	Dataset	Ref.
Quantum machine learning-based electrokinetic mining for the identification of nanoparticles and exosomes with minimal training data	Classical and quantum ML models	Electrokinetic data of green synthesized nanoparticles, and EVs from various sources (blood serum of mouse; healthy and colorectal cancer patients)	[141]
Clinical data classification with noisy intermediate scale quantum computers.	QML algorithms for data classification on IBM quantum hardware: a quantum distance classifier (qDS) and a simplified quantum-kernel support vector machine (sqKSVM)	Three open-access clinical datasets: Pediatric bone marrow transplant dataset, Wisconsin breast cancer dataset, and Heart failure dataset.	[142]
Machine learning-assisted assessment of extracellular vesicles can monitor cellular rejection after heart transplant.	Random forest regressor	Prospective longitudinal cohort study with 24 heart transplant recipients. 285 blood samples were analyzed for EV surface antigens.	[143]
Deep Learning Promotes Profiling of Multiple miRNAs in Single Extracellular Vesicles for Cancer Diagnosis.	Deep learning (DL) was successfully used for sorting TIRF-mediated single-EV multi-miRNAs	20 patients: 5 healthy 5 cervical cancer 6 breast cancer Around 10,000 individual EV images	[144]
Fluorescence Analysis of Circulating Exosomes for Breast Cancer Diagnosis Using a Sensor Array and Deep Learning.	DL was used to change unordered fluorescence spectra into useable feature maps distinguishing healthy vs breast cancer donors	5000 exosomes/ML 2 × 105 cells/ML	[145]
Label-Free Identification of Exosomes using Raman Spectroscopy and Machine Learning	Integration of surface-enhanced Raman spectroscopy with artificial neural networks (ANN), a DL model, to identify cellular origins of EVs	Exosomes from HT-1080, HEC-1-A, THP-1, HeLa, and MEF cell lines were used	[146]
High Resolution Imaging and Analysis of Extracellular Vesicles Using Mass Spectral Imaging and Machine Learning.	Used ML with secondary mass ion spectroscopy to analyze chemical changes in EVs	Trained using toroidal topology and 36 hexagonal neurons	[147]
Accurate label free classification of cancerous extracellular vesicles using nanoaperture optical tweezers and deep learning.	Used nanoaperture optical tweezers (NOTs) and deep learning for cancerous EV classification	Created hybrid optical tweezers with deep learning abilities pre- and post-processing stages prepare the data for analysis which trains, validates, and tests their system Training size: 3267 Validation size: 408 Test size: 409	[148]
Single test-based diagnosis of multiple cancer types using Exosome-SERS-AI for early-stage cancers.	Used AI to analyze surface-enhanced Raman spectroscopy profiles of exosomes	520 test samples were used	[149]
Deep Learning-Based Classification of NSCLC-Derived Extracellular Vesicles Using AFM Nanomechanical Signatures.	Combined atomic force microscopy with deep learning for classification of NSCLC EVs	Dataset was 239 sets of AFM images Training set: 1222 Validation set: 306 Test set: 384	[150]
Machine learning-based analysis identifies and validates serum exosomal proteomic signatures for the diagnosis of colorectal cancer.	Used proteomics and machine learning to identify proteins for diagnosis in colorectal cancer	Discovery set: 25 cases of colorectal cancer and 5 healthy controls Expansion cohort: Training set: 338 cases Test and validation set: 328	[151]
ChatExosome: An Artificial Intelligence (AI) Agent Based on Deep Learning of Exosomes Spectroscopy for Hepatocellular Carcinoma (HCC) Diagnosis.	Created a chatbot to aid clinical spectroscopic analysis and diagnosis for exosomes from hepatocellular carcinoma	Dataset: 165 clinical plasma samples and 400 spectral data points collected for each sample Of the plasma samples, 40 are healthy controls and 125 are cancer patients Of the cancer patients, 61 are AFP+ and 65 are AFP−	[152]
Evaluation of machine learning and deep learning models for the classification of a single extracellular vesicles spectral library.	Combine ML and SERS for analysis of generated fingerprints from EVs as potential biomarkers	Uses data from data library integrated by SERS spectra from single EVs derived from culture media or blood plasma for cancer detection	[153]

5.4. Role of electric field stimulation in optimization of EV production

Low-level electric treatment (ET), commonly used in transdermal drug delivery methods like iontophoresis, has been shown to induce endocytosis and enhance the cellular uptake of macromolecules [154]. Given that exosomes originate from MVBs, which are part of the endocytic pathway, researchers hypothesized that ET might also stimulate the secretion of EVs, including exosomes [41,154,155]. The researchers investigated this possibility using both cancerous and non-cancerous murine cell lines. They observed that ET led to a marked increase in the quantity of EVs secreted by both cell types, along with a general rise in EV-associated protein content. Furthermore, the EVs retained representative EV markers (CD9, CD81, and HSP70) and were functionally competent, as evidenced by comparable uptake rates between treated and untreated conditions [141,154,156]. These results suggest that low-level ET could serve as a non-invasive strategy to increase EV production in vitro, potentially facilitating scalable EV isolation for diagnostic and therapeutic applications or modeling studies.

5.5. CAR-T cell-derived EVs as potential cancer therapy

CAR-T cell immunotherapy is a cancer treatment that involves engineering T cells to target specific cancer cells [157]. The first report of the successful use of this technique was in 2010, where CAR-engineered T cells effectively targeted CD19 molecules in patients with B-cell lymphoma [158]. Ever since, immunotherapy has garnered increasing rapport and interest in modern oncology [159]. Despite CAR-T cell therapy’s tumor-targeting capabilities, it has limitations in terms of safety and efficacy. Cytokine release syndrome (CRS) and “on-target off-tumor” response are key safety concerns [157]. CRS involves chronic cytokine release, leading to cytokine storms that trigger hyperactive immune responses and subsequently jeopardize vital organs [160,161]. Moreover, “on-target off-tumor” toxicity results from the direct attack on nonpathogenic cells that share the targeted antigen, leading to damage of healthy tissue [162]. Concerning efficacy, the issues rely on accessibility barriers due to CAR-T cells’ large size and inability to fine-tune for chemoresistance and cargo purposes. Here is where exosomes offer a compelling alternative. A proposed approach for CAR-T cell-derived exosome therapy involves collecting T cells from the peripheral blood of cancer patients, followed by the insertion of CARs into the T cells using either viral or non-viral methods. Then, the CAR-engineered T cells are expanded ex vivo. The resulting CAR-T cells release exosomes into the culture media, which are then isolated and subsequently administered to the patient [157].

CAR-T cell-derived exosomes excel in both areas where CAR-T cells lack safety and efficacy. The safety of the exosomes over the CAR-T cells relies on their cell-free nature [159]. Unlike CAR-T cells, exosomes cannot replicate, which reduces long-term risks, as uncontrolled proliferation of CAR-T cells can increase the chance of CRS [163]. Similarly, exosomes do not produce cytokines on their own; they carry those from the parent CAR-T cells, further reducing the risk of CRS. Moreover, exosomes are minimally immunogenic [164]. In terms of the “on-target off-tumor” response, exosomes do not eliminate the risk since they still carry CAR receptors; however, they represent a safer alternative due to their lower cytotoxicity and inability to replicate [159].

In terms of efficacy, the nanoscale size of exosomes allows them to reach more targets than the larger CAR-T cells [42,165]. For instance, exosomes may be able to penetrate solid tumor masses, which has been a historic barrier for CAR-T cells [163,165]. Nevertheless, further studies are necessary to prove the clinical effectiveness of exosomes diffusing through solid tumors. Furthermore, exosomes can cross biological barriers that CAR-T cells cannot, such as the BBB, expanding their potential as therapeutics for brain tumors [42,157,159]. When it comes to fine-tuning, exosomes can be efficiently loaded with cargo after isolation because of their cell-free nature. This allows experimenters to add additional therapeutic agents (proapoptotic proteins, checkpoint inhibitors, siRNA, and mRNA), which may boost therapeutic efficacy and help bypass tumor resistance [157,159]. However, efforts to overcome resistance to chemotherapy have not been successful, which may be attributed to the complex heterogeneity of cancer [159].

CAR-T cell-derived exosomes address major limitations of traditional CAR-T therapy by offering improved safety and therapeutic potential. Their cell-free nature reduces risks like CRS, and their nanoscale size allows for better tumor penetration and crossing of barriers like the BBB. Additionally, their capacity to be loaded with specialized cargo makes them more clinically relevant. While current studies highlight the potential of mitigating CRS and on-target/off-tumor toxicity, clinical efficacy of these approaches remains to be validated, and significant challenges persist in standardizing manufacturing processes and meeting release-criteria requirements for clinical translation.

5.6. Clinical trials and translational research

The translational potential of EVs is currently being investigated in numerous clinical trials (Table 4). These include studies using EVs as biomarkers for early cancer detection, prognostic evaluation, or therapy monitoring. Additionally, trials are underway investigating engineered EVs for targeted drug delivery in cancers such as glioblastoma and pancreatic adenocarcinoma. The results of these early-phase trials will be critical in determining the viability of EVs in clinical oncology. Despite their promise, EV-based technologies face significant hurdles before full clinical integration. These obstacles include the lack of standardized protocols for EV isolation, purification, and quantification [24,166]. Variability in EV yield and content across methods can affect reproducibility and regulatory approval [167]. Furthermore, scaling up EV production for therapeutic use demands consistent quality control and validated manufacturing pipelines compliant with Good Manufacturing Practices (GMP) [168,169].

Table 4.

Clinical progress of EVs related to cancer diagnostic and therapeutic studies.

*NCT#	Title	Condition	Purpose	Sponsor
NCT06278064	Exosome-based liquid biopsies for upper gastrointestinal cancers diagnosis	Esophageal Cancer, gastric cancer	Quantify protein expression in exosome for use as early cancer diagnosis	Beijing Friendship Hospital
NCT05625529	ExoLuminate study for early detection of pancreatic cancer	Pancreatic ductal adenocarcinoma (PDAC)	Use exosomes to identify protein biomarkers	Biological Dynamics
NCT03262311	Pimo Study: Extracellular vesicle-based liquid biopsy to detect hypoxia in tumours	Cancer	Use EVs to track the progress of the therapy patients undergo	Institute of Cancer Research, United Kingdom
NCT06104930	Plasma extracellular vesicles in meningioma patients (MOLI)	Meningioma	Use EVs in plasma to observe change under radiotherapy and identity potential biomarkers	University Hospital Heidelberg Rion Inc
NCT01344109	Breast cancer: exosome diagnostic and prognostic biosignature	Breast Cancer	Explore exosome-based biomarkers for diagnosis & prognosis	Leo W. Jenkins Cancer Center
NCT01779583	Gastric cancer: exosome biomarker prognostic tool	Gastric Cancer	Identify exosome-derived biomarkers for diagnosis, prognosis, and prediction	Hospital Miguel Servet
NCT06342414	Differential diagnosis of Intrahepatic cholangiocarcinoma (ICC) vs HCC via exosomal biomarkers	ICC, HCC	Enhance differential diagnosis between two primary liver cancers via exosome analysis	City of Hope Medical Center
NCT06777030	Exosome role in pancreatic cancer progression	Pancreatic Cancer	Characterize EV content in fluids of pancreatic cancer patients, understanding progression role	RCCS Azienda Ospedaliero-Universitaria di Bologna
NCT04529915	Early lung cancer detection via exosomal markers	Lung Cancer	Assess feasibility of exosome markers for early lung cancer detection	N/A
NCT03608631	iExosomes in treating participants with metastatic pancreas cancer with KRASG12D mutation	Pancreatic ductal adenocarcinoma (metastatic, KRAS G12D)	Use MSC-derived exosomes to deliver KRASG12D siRNA (iExosomes) as targeted therapy	M.D. Anderson Cancer Center
NCT01294072	Study investigating the ability of plant exosomes to deliver curcumin to normal and colon cancer tissue	Colon cancer	Test plant-derived exosomes for delivering curcumin to colon tissue/tumors	University of Louisville Hospital
NCT04939324	Molecular profiling of exosomes in tumor-draining vein of early-staged lung cancer (ExOnSite-Pro)	NSCLC (early stage)	Compare pulmonary veins vs peripheral blood EVs to develop relapse biomarkers	University Hospital (CHU) Limoges (France)
NCT03874559	Exosomes in rectal cancer	Locally advanced rectal cancer	Longitudinal exosomal biomarker profiling during neoadjuvant chemoradiation; correlate with pathological response	University of Kansas Medical Center
NCT05955521	Exosome as the prognostic and predictive biomarker in early breast cancer (EBC) Patients	Early breast cancer (TNBC and HER2 +)	Serial ctDNA + exosome assessment to predict neoadjuvant chemotherapy response	Samsung Medical Center
NCT04053855	Evaluation of urinary exosomes presence from clear cell renal cell carcinoma	Clear cell renal cell carcinoma	Develop urine EV detection (e.g., CD9/CA9 +) as a noninvasive liquid biopsy	CHU Saint Étienne (France)
NCT05854030	Serum exosomal miRNA predicting the therapeutic efficiency in lung squamous carcinoma	Lung squamous cell carcinoma (advanced)	Identify serum exosomal miRNA signatures predictive of chemo-immunotherapy response	Tianjin Medical University Cancer Institute & Hospital
NCT02147418	Exosome testing as a screening modality for HPV-positive oropharyngeal squamous cell carcinoma	Oropharyngeal squamous cell carcinoma	Evaluate blood/saliva EV testing as screening modality for HPV-related OPSCC	New Mexico Cancer Research Alliance
NCT01159288	Trial of a vaccination with tumor antigen-loaded dendritic cell-derived exosomes (Dex2)	Non-small cell lung cancer	Use dendritic-cell–derived exosomes loaded with tumor antigens as a therapeutic vaccine	Gustave Roussy Cancer Campus (France)
NCT04720599	Clinical evaluation of ExoDx prostate (IntelliScore) in men presenting for initial prostate biopsy	Urological cancer (risk assessment prior to biopsy)	Prospectively confirm performance of urine exosome gene-expression test (EPI) in initial biopsy setting	Exosome Diagnostics, Inc.

The trials focus on using EVs as either a tool to diagnose and identify biomarkers or a drug delivery vehicle. The diseases of focus range from early-detection/screening cohorts (e.g., upper GI, PDAC, lung) while others involve advanced/metastatic disease (e.g., KRAS-mutant metastatic PDAC, NSCLC, rectal cancer during neoadjuvant therapy). Feasibility, safety, and regulatory restrictions are the biggest challenges these diagnostic and delivery tools face. Because of the heterogenic nature of EVs, there is a lack of standardized protocol making the experiments difficult to reproduce and criteria hard to specify causing the product to be more difficult to manufacture [170]. As for safety and regulations, because this technology is so new, there hasn’t been a lot of time available for studies on long-term effects to be released. Advancements are still being made in the field, as therapeutic companies, such as Kimera Exosomes, have been FDA approved for use in clinical trials where the safety and effectiveness of MSCs EVs will be studied for treatment of COVID-19 symptoms [171]. Additionally, the newly discovered method of disguising immuno-evasive ligands on TDEVs to remove phagocytosis block can be applied to patient specific TDEVs showing that large general progress is being made in the field [172].

6. Current challenges and future directions

One of the primary bottlenecks in EV research is the development of efficient, reproducible, and scalable isolation and purification techniques. Current methods, such as ultracentrifugation, size exclusion chromatography, and immunoaffinity capture, vary widely in yield and purity [103]. These inconsistencies not only compromise downstream analyses but also complicate comparisons across studies. Improved isolation technologies that ensure specificity for TD-EVs while maintaining biological activity are urgently needed for clinical translation. EV populations are inherently heterogeneous, encompassing a range of sizes, compositions, and cellular origins [173]. This biological variability poses a significant analytical challenge, particularly when distinguishing EV subtypes that are disease-relevant from those that are physiologically normal. This heterogeneity also affects functional studies, making it difficult to attribute specific effects to EV subpopulations. Advances in single-EV analysis and high-throughput profiling are expected to address this issue by enabling fine-scale characterization ^176,177.

Because of their role in cell-cell communication, EVs can have conflicting effects on cancer progression with some being found to promote tumor growth and others, suppression. An example of this would be the difference response in PD-L1 expression in different cancer tissues. In most cancer tissues (e.g., breast, lung, head and neck, etc.), the expression of PD-L1 on EVs can be used as a biomarker for tumor growth as it has been found to suppress T-cell function, but no significant difference was found for pancreatic cancer. For pancreatic cancer, this seems to be due to heterogeneity as some samples show positive results while others show nothing proving PD-L1 to not be an ideal biomarker in this case [59]. Another is MSC exosomes showing to help in tumor growth and angiogenesis. In an in vivo experiment, it was found that mice whose tumors were co-implanted with MSC exosomes expressed higher levels of pro-tumorigenic proteins such as VEGF and MDM2, a negative regulator of tumor suppressor p53 [174,175]. This shows that EVs cannot be categorized as tumor suppressors or promoters. This means more time, effort, and money must be used to research EVs depending on what the ultimate use is and the biological context they appear in, providing another challenge [174]. As previously discussed, lack of standardization in EV research continues to hinder reproducibility and clinical application. Variables such as sample collection, storage conditions, and analysis workflows significantly influence results. The implementation of universally accepted protocols, such as those proposed by the International Society for Extracellular Vesicles (ISEV), is critical for ensuring consistency [176]. Moreover, curated databases cataloging EV-associated biomarkers and cargo (e.g., Vesiclepedia, EV-TRACK) must be continually updated and validated to facilitate data sharing and integration across studies. The potential for EVs to revolutionize personalized medicine lies in their ability to reflect the dynamic molecular landscape of an individual’s tumor. TD-EV profiling could enable stratification of patients based on disease subtype, treatment responses, or resistance mechanisms. With continued development, EVs could serve as real-time monitors of disease evolution, guiding precision therapies tailored to each patient’s tumor biology, and ushering in a new era of individualized oncology care [177–180].

7. Conclusion

TD-EVs represent a transformative component in the landscape of cancer biology, diagnosis, and treatment. Their cargo, reflective of the genetic and proteomic profile of tumor cells, positions them as powerful biomarkers for early detection, prognosis, and therapeutic monitoring. At the same time, EVs actively contribute to disease progression through intercellular communication, immune modulation, and metastasis facilitation. Despite their dual role as both mediators and potential disruptors of cancer pathology, recent advances suggest that EVs can be repurposed as therapeutic delivery systems and diagnostic tools. Engineered EVs offer targeted drug delivery with high specificity, while EV-based liquid biopsy promises a minimally invasive alternative to conventional diagnostic methods. Looking forward, integrating EV research into clinical oncology requires overcoming critical barriers such as standardization, scalability, and subtype resolution. Nevertheless, with ongoing innovation and growing clinical interest, EVs are poised to become integral components of personalized and precision medicine and thereby transforming both cancer care and biomedical research.

Data availability

No data was used for the research described in the article.