Tiny Messengers, Actionable Targets: sEV-Driven Epigenetic Changes in Cancer

Abstract

Small extracellular vesicles (sEVs) are pivotal mediators of intercellular epigenetic communication in cancer. Following MISEV2023, we use the size- and isolation-based term sEV throughout, because most primary studies cited here cannot resolve which biogenetic route generated the vesicles they analyzed. By selectively packaging and transferring noncoding RNAs (ncRNAs), DNA fragments, chromatin-modifying enzymes, and metabolic effectors, sEVs reprogram recipient-cell chromatin architecture without altering the underlying DNA sequence. Tumor-derived sEVs engage stromal, immune, and vascular compartments to drive malignant progression through mechanisms that include miRNA-directed suppression of DNA methyltransferases (DNMTs), lncRNA-scaffolded Polycomb Repressive Complex 2 (PRC2) recruitment, depositing H3K27me3, and oncometabolite-mediated inhibition of TET dioxygenases. This narrative review synthesizes mechanistic, preclinical, and translational evidence on sEV-driven epigenetic regulation in cancer, applies a four-level evidence hierarchy to calibrate mechanistic claims, and critically evaluates how distinct cargo classes–microRNAs (miRNAs), long noncoding RNAs (lncRNAs), circular RNAs (circRNAs), DNMTs, and histone-modifying enzymes–contribute to chromatin remodeling, aberrant DNA methylation, acquired therapy resistance, and immune evasion in recipient cells. We further examine sEV cargo signatures as minimally invasive liquid biopsy biomarkers and appraise engineered sEV platforms for the precision delivery of miRNA mimics, siRNAs, and small-molecule epigenetic inhibitors. Key methodological challenges, EV isolation standardization, MISEV2023 compliance, cargo stoichiometry at physiological concentrations, in vivo biodistribution, and the transition from post-transcriptional regulation to durable chromatin-state change are critically evaluated, and a translational roadmap is proposed to guide reproducible clinical implementation of sEV-mediated epigenetic cancer therapeutics.

1. Introduction

Research on extracellular vesicles (EVs) has emerged as a key area of academic interest owing to their central role in intercellular signaling across physiological and pathological contexts [1,2]. EVs are lipid bilayer-enclosed particles secreted by virtually all cell types and are operationally classified by size and biogenesis into small EVs (sEVs, 30–150 nm, operationally referred to as exosomes when strict biogenesis-based classification cannot be confirmed from the primary literature [3,4]), microvesicles (100–1000 nm), and apoptotic bodies (>1000 nm) [5,6]. Their cargo, selectively packaged proteins, RNAs, and lipids, travels via regulated intracellular sorting pathways and can modulate recipient-cell behavior and gene expression [7,8,9,10]. This conceptual framework positions sEVs as key mediators of intercellular epigenetic communication with direct consequences for immune regulation, tumor progression, and therapy resistance [11,12].

The biological significance of sEV-mediated epigenetic communication in cancer is underscored by three converging lines of evidence. First, tumor-derived sEVs carry a biased, functionally coherent epigenetic payload, not a random cross-section of cellular molecules, because selective cargo-sorting machinery (SYNCRIP, hnRNPA2B1, AGO2, tetraspanin checkpoints) enforces specific miRNA and lncRNA enrichment in secreted vesicles [13,14]. Second, recipient cells that internalize this cargo show measurable chromatin-state changes: altered CpG methylation at tumor suppressor loci, H3K27me3 deposition at pro-apoptotic promoters, and reprogrammed histone acetylation landscapes [15,16]. Third, these epigenetic changes are functionally consequential: they drive cisplatin resistance, T-cell exhaustion, cancer-associated fibroblast activation, and epithelial-to-mesenchymal transition (EMT), all of which are reversible when exosome secretion is pharmacologically blocked [17,18].

Despite this progress, three foundational mechanistic controversies remain unresolved and constrain clinical translation. First, can intact multi-subunit chromatin-modifying complexes (PRC2, NuRD) survive the intraluminal EV milieu and be delivered in catalytically competent form, or does epigenetic reprogramming occur primarily through ncRNA-templated nucleation of recipient-cell chromatin machinery? Second, do the sEV miRNA concentrations achievable in vivo, for which stoichiometric analyses suggest fewer than one copy per vesicle for most miRNAs, produce the phenotypic reprogramming reported in vitro at supraphysiological doses? Third, can engineered sEVs carrying CRISPR-dCas9 epigenome-editing complexes achieve sufficient tumor biodistribution and on-target chromatin reprogramming in vivo? This narrative review synthesizes mechanistic, preclinical, and translational evidence on sEV-driven epigenetic regulation in cancer, critically evaluates claims according to a four-level evidence hierarchy, and proposes a prioritized experimental roadmap for resolving these controversies and advancing clinical translation. Throughout, the term exosome is used operationally to denote sEVs of 30–150 nm isolated by differential ultracentrifugation or size-exclusion chromatography; strict biogenesis-based classification, as defined by MISEV2023 [3], is not always confirmable from the primary literature cited.

2. Search Strategy

We conducted a structured narrative literature search to compile mechanistic, pre-clinical, and clinical evidence on exosome-mediated epigenetic regulation in cancer. This review is explicitly a narrative synthesis and not a systematic review; it was not prospectively registered and does not include a PRISMA flow diagram [19]. Eligibility decisions were made by the authors based on mechanistic relevance and methodological quality, and selection bias toward studies supporting the conceptual framework presented cannot be excluded. Databases investigated include PubMed/MEDLINE, Web of Science, Scopus, and ClinicalTrials.gov for both ongoing and concluded exosome-associated cancer studies. Temporal scope: January 2000–April 2026. Language restriction: English. The principal Boolean string was as follows: (“exosome*” OR “extracellular vesicle*” OR “sEV” OR “small extracellular vesicle*”) AND (“epigenetic*” OR “DNA methylation” OR “histone” OR “chromatin” OR “non-coding RNA” OR “epitranscriptom*”) AND (“cancer” OR “tumor” OR “oncology*” OR “malignant*”). Inclusion criteria include the following: primary mechanistic studies (in vitro and in vivo) with EV characterization meeting MISEV2018/2023 minimum reporting standards [3,20], translational/clinical biomarker studies, and peer-reviewed reviews that directly address exosomal cargo and epigenetic outcomes. Exclusion criteria are as follows: studies lacking EV characterization, reports analyzing cell-free nucleic acids without explicit EV separation, and non-peer-reviewed material unless explicitly cited as preprints and labeled as provisional. Readers should interpret the mechanistic synthesis herein as reflecting the current weight of published evidence, subject to publication bias and the rapidly evolving nature of this field.

3. Exosome Biogenesis and Epigenetic Cargo Packaging

The biogenesis of exosomes is initiated by the invagination of the plasma membrane to form early sorting endosomes, which mature into late endosomes and ultimately give rise to multivesicular bodies (MVBs) [21]. The intraluminal vesicles contained within MVBs are selectively loaded with molecular cargo, including proteins, lipids, and nucleic acids, and are released as exosomes following fusion of MVBs with the plasma membrane [22]. Cargo sorting into exosomes is orchestrated by endosomal sorting complexes required for transport (ESCRT) as well as ESCRT-independent sphingolipid-mediated pathways, with RNA-binding proteins such as heterogeneous nuclear ribonucleoprotein A2B1 (hnRNPA2B1), YBX1, and AGO2 directing the selective packaging of miRNAs [23]. The epigenetic payload of sEVs is particularly consequential. Non-coding RNAs, including microRNAs (miRNAs), long noncoding RNAs (lncRNAs), and circular RNAs (circRNAs), represent the predominant epigenetic cargo and can regulate post-transcriptional gene silencing, chromatin remodeling, and transcription factor activity in recipient cells [22]. Beyond non-coding RNAs, exosomes also transport DNA methyltransferases (DNMT1, DNMT3A, DNMT3B) and histone-modifying enzymes that can reprogram the epigenome of recipient cells, a mechanism with profound relevance for cancer progression and drug resistance [21]. Although DNMTs, EZH2, and histone-modifying enzymes have been detected in exosomal proteomes, this detection does not establish that these proteins are present in catalytically competent form, that they are delivered intact to recipient cell cytoplasm or nucleus, or that they produce heritable chromatin changes in recipient cells. Intact multi-subunit complexes such as PRC2 (EZH2/SUZ12/EED) face particular challenges with respect to intraluminal stability, vesicle loading stoichiometry, and post-endocytic delivery. Direct mechanistic validation of enzymatic activity in recipient cells, not merely cargo detection, is required before these mechanisms can be stated as established. Claims in this review involving protein-mediated epigenetic reprogramming are qualified accordingly, as discussed in Section 8 and Section 12.

It is essential to note that the quantity and molecular makeup of sEVs serve as indicators of the physiological and pathological conditions of the source cell, thereby rendering their constituents valuable and informative candidates for diagnostic applications [21]. The RNA content within sEVs is distinctive, safeguarded against degradation by nucleases, proteases, and oxidative stressors, thereby facilitating effective transfer to target cells [24]. This protection makes exosomal cargo a concentrated source of information that can alter the function of neighboring and distant cells, carrying a wealth of information about transcriptomic and epitranscriptomic changes that occur during disease conditions [24]. The biogenesis routes and cargo-sorting machinery described above are summarized in Figure 1.

Figure 1. Exosome biogenesis and selective epigenetic cargo sorting. Schematic overview of multivesicular body (MVB) formation, selective loading of epigenetic regulators into intraluminal vesicles (ILVs), and release of mature exosomes. ESCRT-dependent (HRS, TSG101, CHMP4/VPS4) and ESCRT-independent pathways drive ILV biogenesis (Panel 1). RNA-binding proteins and sequence motifs mediate selective packaging of small RNAs (such as miR-21, miR-155), lncRNAs/circRNAs, and methylated DNA, while epigenetic enzymes (DNMT1/3A/3B, EZH2) are recruited as cargo (Panel 2). Two sorting checkpoints, ceramide-enriched lipid domains and tetraspanin-enriched microdomains (CD9/CD63/CD81), bias cargo selection (Panel 3). MVB–plasma membrane fusion releases exosomes bearing canonical markers (CD63, CD9, CD81, HSP70, TSG101) to modulate recipient-cell epigenomes (Panel 4).

Beyond the Endosomal Route: Heterogeneity of sEV Biogenesis and Non-Vesicular Confounders

The preceding account describes the endosomal route, in which intraluminal vesicles formed within multivesicular bodies are released as exosomes upon MVB–plasma membrane fusion. This is not the only route generating vesicles of 30–150 nm, and because no current isolation method separates vesicles by biogenesis, any preparation designated “sEV” is a mixture. MISEV2023 is explicit on this point, recommending operational rather than biogenetic nomenclature unless subcellular origin has been directly established [3]. We adopt that convention throughout and set out the principal alternative sources here, since sorting machinery differs between routes, and the selectivity arguments of Section 4 are route dependent.

Small ectosomes bud from the plasma membrane. Live intracellular tracking of tetraspanins showed that CD9 and a CD63 mutant stabilized at the plasma membrane are released more abundantly than wild-type CD63, leading to the conclusion that ectosomes are more prominent than exosomes in HeLa cells; LAMP1 and BSG/SLC3A2 were proposed as exosome- and ectosome-leaning discriminators, respectively [25]. Since CD9 and CD63 are the markers most often used to certify “exosome” preparations in the cancer epigenetics literature, a substantial share of published cargo may be ectosomal.

ARRDC1-mediated microvesicles (ARMMs). ARMMs bud directly from the plasma membrane through TSG101 recruitment by ARRDC1 in a VPS4-dependent manner and lack late-endosomal markers [26]. They are ESCRT-dependent yet non-endosomal, demonstrating that ESCRT involvement cannot be read as evidence of exosomal origin. Engineered ARMMs deliver p53 protein and CRISPR/Cas9 machinery to recipient cells [27], establishing them as competent carriers of the protein cargo discussed in Section 4.3.

Migrasomes. Migrating cells deposit retraction fibers from which migrasomes form and release their contents [28], a route mechanistically distinct from both MVB fusion and plasma-membrane budding and pertinent to the invasive populations frequently sampled in cancer EV studies.

Autophagy-coupled secretion. The LC3-conjugation machinery specifies loading of RNA-binding proteins and small RNAs into EVs [29]. This is directly germane here: the RNA-binding proteins governing selective ncRNA sorting operate within this route as well as the ESCRT-dependent one, so ncRNA enrichment cannot be attributed to endosomal sorting by default.

Non-vesicular nanoparticles. Asymmetric flow field-flow fractionation resolves sEV preparations into two exosome subpopulations and an abundant class of non-membranous exomeres with distinct nucleic acid, protein, and lipid profiles [30]; supermeres carry miR-1246, AGO2, and glycolytic enzymes and show greater in vivo uptake than sEVs [31]. Independently, most circulating miRNAs co-fractionate with Argonaute2 ribonucleoprotein complexes rather than vesicles [32], HDL particles deliver miRNA to recipient cells [33], and systematic reassessment has shown that several canonical “exosomal” constituents, including extracellular DNA and histones, are largely non-vesicular [14]. The implication is unavoidable: where epigenetic cargo is reported in a bulk ultracentrifugation pellet without density-gradient separation, nuclease- and protease-protection controls, or single-vesicle validation, it is attributable to the preparation rather than to vesicles. We have applied this criterion in assigning evidence levels throughout Section 4, Section 5, Section 6, Section 7, Section 8 and Section 9, and it accounts for several instances in which widely cited findings are placed at Level 1 rather than Level 2.

Capsid-mediated routes. Domesticated retroviral Gag proteins, including PEG10, form capsids that package their own mRNA and can be pseudotyped for delivery [34], indicating that extracellular RNA transfer is not exclusively vesicular.

Three practical measures follow: deposition of isolation and characterization metadata in EV-TRACK with EV-METRIC reporting [35]; inclusion of route-discriminating markers alongside tetraspanins [25]; and orthogonal separation of vesicular from non-vesicular carriers before any epigenetic effect is ascribed to sEVs. Throughout this review, “sEV” denotes the operationally defined preparation; where a cited study established endosomal origin, this is stated explicitly.

4. sEV Cargo Relevant to Epigenetic Regulation

A specialized subclass of sEVs mediates intercellular communication by transferring biologically active macromolecules, including RNAs, proteins, and DNA, between cells, thereby shaping tumor biology at both local and systemic levels [36,37]. In cancer, this communication is increasingly interpreted through an epigenetic lens. Epigenetic regulation, encompassing DNA methylation, histone and chromatin modification, and ncRNA-mediated gene-expression control, is fundamental to malignant transformation, progression, and treatment response [37,38]. A conceptual distinction is essential before describing individual cargo classes. In this review, the term epigenetic cargo refers strictly to molecules that produce heritable, chromatin-level alterations in recipient cells: covalent DNA modifications (5-methylcytosine; 5-hydroxymethylcytosine), post-translational histone modifications (acetylation, methylation, phosphorylation, ubiquitination), or chromatin-remodeling complex-mediated changes in nucleosome positioning [11,39]. sEV cargo that produces post-transcriptional silencing, for example, miRNA-mediated mRNA destabilization through RNA-induced silencing complex (RISC) loading without a concurrent chromatin-state change, is classified as post-transcriptional regulation and not as epigenetic regulation sensu stricto, even where it influences transcriptional outputs. This distinction is maintained throughout Section 4, Section 5 and Section 6. Cargo classes are described as epigenetic where direct evidence for chromatin-level changes in recipient cells exists [40]; where such evidence is absent or indirect, claims are explicitly framed as hypotheses or candidate mechanisms. Epigenetic control is organized around chromatin states and the enzymatic writing, erasing, and reading of chromatin marks, implemented by DNA methyltransferases (DNMTs), TET family dioxygenases, histone acetyltransferases (HATs), histone deacetylases (HDACs), and histone methylation enzymes [41,42,43,44]. In cancer, chromatin modifiers including KDM6A, KMT2D, and EZH2 are frequently mutated or overexpressed, reshaping tumor epigenomes and driving aggressive behavior and immune evasion [45,46], while in hematologic malignancies, epigenetic dysregulation is central to pathogenesis and represents a target for therapy [47]. Against this backdrop, sEV cargo classes can be organized into two functional categories: (i) direct epigenetic cargo; molecules that themselves constitute or transfer epigenetic information (methylated DNA fragments, chromatin-modifying enzymes); and (ii) indirect epigenetic modulators; molecules that alter the expression or activity of the host epigenetic machinery (ncRNAs targeting DNMT or HDAC mRNAs; metabolic intermediates that serve as obligate co-substrates for chromatin enzymes). Both categories are reviewed below, with mechanistic certainty calibrated to the available evidence.

4.1. sEV ncRNA Cargo Classes as an Epigenetic Regulator

4.1.1. sEV miRNA Cargo and Epigenetic Coupling

miRNAs regulate gene expression primarily post-transcriptionally; however, in the context of sEV transfer, their functional consequences extend to chromatin-level alterations, warranting their inclusion as indirect epigenetic cargo. Epigenetic mechanisms and miRNA expression are bidirectionally coupled: promoter hypermethylation silences tumor-suppressive miRNAs, while miRNAs reciprocally target DNMT and polycomb-group mRNAs to alter DNA methylation and histone modification states [39,48]. Crucially, miRNA loading into sEVs is not passive but is governed by selective sorting machinery. The RNA-binding protein SYNCRIP controls miRNA sorting in hepatocyte-derived exosomes through sequence-specific recognition motifs, providing a molecular basis for the selective enrichment of individual miRNA species in secreted vesicles [39,40]. hnRNPA2B1 and AGO2 perform analogous sorting roles in other cell types [22]. This selectivity means that tumor-derived sEVs carry a biased, functionally coherent miRNA payload rather than a random sample of cellular miRNAs. Once delivered to recipient cells, sEV miRNAs engage the endogenous RISC machinery to suppress target mRNAs. Where targets include DNMT3A, DNMT3B, EZH2, or HDAC isoforms, the downstream consequence is a measurable epigenetic state change, the criterion distinguishing indirect epigenetic modulation from purely post-transcriptional regulation [11,38]. For example, miR-29 family members suppress DNMT3A and DNMT3B, producing genome-wide hypomethylation [49]. miR-21 activates STAT3 signaling that in turn recruits DNMT3A to silence effector immune-cell loci [16]. The evidence for miRNA-mediated DNMT suppression downstream of sEV transfer reaches Level 3 of the evidence hierarchy (measurable chromatin/methylation change in recipient cells), making it one of the better-supported indirect epigenetic mechanisms in the field. In hepatocellular carcinoma (HCC), sEV ncRNAs, including miRNAs, have been characterized as contributors to disease biology and candidate non-invasive biomarkers [50]. More broadly, tumor-derived sEVs propagate oncogenic signaling, modulate the tumor microenvironment (TME), and engage epigenetic regulatory programs linked to progression and therapy resistance [11,38]. sEV miRNAs are therefore both mechanistic drivers capable of triggering chromatin-state transitions in recipient cells and translational targets, warranting continued investigation as therapeutic and biomarker candidates across tumor types.

4.1.2. sEV lncRNA Cargo and Chromatin Regulation

Transcripts > 200 nucleotides with limited protein-coding potential regulate gene expression at epigenetic, transcriptional, and post-transcriptional levels, primarily through interactions with chromatin-modifying complexes, transcription factors, and other RNA species [51]. As scaffolds or guides for PRC2, DNMT, or KAT/HDAC family proteins, lncRNAs directly nucleate chromatin-state changes at specific genomic loci, qualifying them as direct epigenetic regulators when delivered to recipient cells via sEVs [51,52]. sEVs provide a biologically relevant vehicle for lncRNA transfer across cellular compartments. Dedicated reviews document the biological functions and translational significance of sEV lncRNAs in tumor progression and drug resistance [38,53]. Disease-specific evidence in colorectal cancer (CRC) is instructive: serum-exosome-protected lncRNA NNT-AS1 has been characterized as a candidate oncogenic biomarker acting through a defined miR-496/RAP2C axis [54]; differential profiling of tumor-tissue versus serum-exosome lncRNA patterns confirms selective lncRNA release into the circulation, reflecting complex extracellular regulation rather than passive leakage [55,56]. Computational ceRNA axis analyses in CRC further map the lncRNA/circRNA-miRNA-mRNA networks carried by circulating sEVs, providing mechanistic plausibility for their downstream gene-regulatory effects [57]. The epigenetic significance of sEVs’ lncRNA transfer is most directly demonstrated by HOTAIR. HOTAIR recruits the PRC2 complex (EZH2, SUZ12, EED) to silence tumor-suppressor loci via H3K27 trimethylation (H3K27me3) [15,58,59]. Evidence that cisplatin-resistant cells secrete HOTAIR-enriched sEVs and that HOTAIR transfer propagates H3K27me3-mediated silencing of pro-apoptotic loci (DAPK, PTEN) in cisplatin-sensitive recipient cells places this mechanism at Level 3–4 of the evidence hierarchy among the strongest available for any sEV lncRNA cargo. It is important to note, however, that the relative contributions of (i) lncRNA-templated PRC2 nucleation within recipient cells versus (ii) direct transfer of pre-assembled PRC2 complexes remain experimentally unresolved. Transfer of intact multi-subunit complexes requires orthogonal validation (proximity ligation, co-immunoprecipitation within isolated vesicles, single-vesicle proteomics) that has not yet been rigorously provided; this represents an open mechanistic question, not an established pathway. Beyond HOTAIR, sEV lncRNA transfer between tumor and stromal/immune compartments reshapes therapy-response epigenetic states. Transfer of cargo between cancer cells and tumor-associated macrophages (TAMs) modulates phenotypic reprogramming consistent with chromatin remodeling [60]. These observations support a model in which sEV-mediated lncRNA trafficking constitutes a direct epigenetic communication axis, while cautioning that chromatin-level readouts in recipient cells, not just lncRNA detection in EVs, are required to establish mechanistic completeness.

4.1.3. sEVs circRNA Cargo in Cancer

Covalently closed RNA structures generated by back-splicing are increasingly recognized as regulators of gene expression in cancer through two principal mechanisms: (i) competitive endogenous RNA (ceRNA) activity, sponging miRNAs away from their mRNA targets, and (ii) direct interactions with chromatin-regulatory proteins that alter transcriptional programs [61]. Their relevance spans hematological and solid malignancies [62,63]. Within the exosomal compartment, circRNAs have been identified as functional cargo in breast cancer, pancreatic cancer, and lung cancer contexts [51,64,65,66]. The best-characterized sEV circRNA with direct epigenetic consequences is circNSUN2: N6-methyladenosine (m6A) modification of circNSUN2 by the METTL3/METTL14 complex facilitates its YTHDC1-dependent selective loading into sEVs; following uptake by recipient cells, circNSUN2 stabilizes HMGA2, an architectural chromatin protein that remodels nucleosome positioning at metastasis-associated loci, thereby promoting hepatic metastatic colonization [13,67]. This example provides Level 3–4 evidence for sEV circRNA-mediated epigenetic remodeling. In pancreatic cancer, the circRNA cargo of irradiation-conditioned sEVs is reshuffled relative to untreated cells, consistent with therapy-induced reprogramming of sEV content [65]. More broadly, sEV ncRNA biology is synthesized through lncRNA/circRNA-miRNA-mRNA axis models in CRC and breast cancer [57,64]. While these network models provide mechanistic plausibility, the majority of cited circRNA axis evidence rests on in silico ceRNA interaction predictions and correlation analyses rather than chromatin-level functional validation in recipient cells. Cross-cancer generalization of specific circRNA mechanisms should therefore be treated as a testable hypothesis pending experimental confirmation of chromatin-state changes in relevant cancer-type-specific cellular and stromal contexts.

However, evidence for circRNA-specific epigenetic mechanisms in cancer-derived sEVs remains largely descriptive and cancer-type specific; mechanistic validation at the chromatin level is available for only a subset of reported circRNA-cancer associations. The mechanisms and clinical implications remain under active investigation, and cross-cancer extrapolation from in silico axis models requires experimental validation.

4.2. sEV DNA as Epigenetic Cargo

sEVs carry double-stranded DNA fragments, including genomic and mitochondrial sequences, as part of their molecular cargo repertoire [68,69]. Because cancer genomes are characterized by focal promoter hypermethylation at tumor-suppressor CpG islands and by global hypomethylation [70], DNA transferred via sEVs carries methylation imprints that can, in principle, influence epigenetic states in recipient cells, constituting direct epigenetic cargo in the strictest sense. Evidence that cancer cells transfer hypermethylated promoter fragments to neighboring cells via sEVs, leading to transcriptional silencing of recipient-cell tumor suppressors, has been documented [38,71]. The translational relevance of EV-associated DNA methylation is particularly well established in the liquid-biopsy context: 5-hydroxymethylcytosine (5hmC) signatures in circulating cell-free DNA and sEV DNA fragments serve as sensitive, tissue-of-origin-specific diagnostic and prognostic biomarkers across multiple cancer types [72]. In lung cancer, integrating circulating DNA methylation profiles with ncRNA measurements enhances diagnostic and therapeutic-prediction accuracy [73]. Single-vesicle nanoscale epigenetic profiling of CRC-derived exosomes using photo-induced force microscopy (PiFM) has successfully distinguished CpG island methylator phenotype (CIMP)-high from CIMP-negative vesicles, revealing intra-population epigenetic heterogeneity that requires single-vesicle resolution for clinically meaningful characterization [74]. An important mechanistic caveat applies: transfer of methylated DNA fragments into recipient cells does not automatically reprogram the recipient epigenome. For such fragments to alter gene expression, they would need to integrate into, or otherwise influence, the recipient chromatin landscape, a mechanism that has not been demonstrated in the context of sEV DNA delivery. The demonstrated translational utility of sEV DNA methylation, therefore, currently resides in its biomarker value rather than in functional epigenetic reprogramming of recipient cells.

4.3. sEV Protein Cargo and Chromatin-Modifying Enzymes

Beyond ncRNA and DNA, sEV cargo includes proteins, among which chromatin-modifying enzymes represent mechanistically consequential candidates. sEV DNMT1 transfer has been directly demonstrated in ovarian cancer, where recipient cells acquire elevated DNMT activity and cisplatin resistance, one of the clearest examples of protein-mediated epigenetic reprogramming via sEVs, with evidence reaching Level 3–4 [71]. Similarly, EZH2 has been detected in tumor-derived EVs, and its candidate delivery to recipient cells could deposit H3K27me3 marks at pro-metastatic or immune-evasion loci [75]. However, a fundamental unresolved question constrains how strongly these mechanisms can be stated: can single-subunit epigenetic enzymes (e.g., DNMT3A, EZH2 monomers) remain catalytically competent through the intraluminal EV milieu and be delivered in an enzymatically active form to recipient cells? For single-subunit enzymes, folding fidelity during vesicle biogenesis and uptake is the primary requirement, making their functional delivery biologically plausible. For multi-subunit complexes, including PRC2 (EZH2/SUZ12/EED/RBBP7, requiring 4+ subunits for full activity) and NuRD (6–8 subunits), co-packaging of the complete assembled complex in a single vesicle and its reassembly or maintenance of activity in the recipient cytoplasm have not been demonstrated [76]. Direct transfer of intact PRC2 or NuRD should therefore be regarded as a hypothesis pending proximity ligation assays, co-immunoprecipitation within isolated EVs, and single-vesicle proteomics that establish stoichiometry and activity status of transferred chromatin regulators. Exosome-mediated transfer of epigenetic protein cargo between tumor and stromal cells (cancer-associated fibroblasts, macrophages) modulates chemosensitivity and resistance through epigenetic events that include HDAC activity changes and altered DNA methylation landscapes [38,48]. This crosstalk is particularly relevant in the context of tumor-immune interactions: tumor-derived EVs carry mRNAs and proteins that regulate HDAC activity in T cells [77], potentially reshaping T-cell effector gene accessibility through histone modification. The protein cargo dimension of exosomal epigenetics remains comparatively under-characterized relative to ncRNA cargo, and systematic single-vesicle proteomics studies focused on epigenetic enzyme enrichment and activity represent a high-priority methodological gap.

4.4. Metabolic Cargo and Epigenetic Co-Substrate Transfer

The intersection of sEV-driven metabolite transfer and epigenetic regulation is mechanistically grounded in the obligate dependency of chromatin-modifying enzymes on specific metabolic intermediates. HATs require acetyl-CoA; DNA and histone methyltransferases depend on S-adenosylmethionine (SAM); and TET family dioxygenases utilize alpha-ketoglutarate (alpha-KG) as an essential cofactor for iterative 5-methylcytosine oxidation [7]. sEVs secreted by metabolically reprogrammed cancer cells can transfer these substrates or the enzymes that generate them to stromal and immune recipient cells, thereby altering the recipient-cell epigenetic landscape independently of genetic instruction. Oncometabolites provide the clearest mechanistic examples. Succinate and fumarate, accumulated through loss-of-function mutations in succinate dehydrogenase (SDH) and fumarate hydratase (FH), competitively inhibit alpha-KG-dependent TET dioxygenases and KDM demethylases, inducing global DNA hypermethylation and H3K methylation accumulation, a hypermethylated epigenetic phenotype demonstrated in paraganglioma and potentially applicable to other SDH/FH-mutant cancers [78]. 2-Hydroxyglutarate (2-HG), produced by gain-of-function IDH1/IDH2 mutations, is a competitive inhibitor of α-ketoglutarate-dependent TET dioxygenases and KDM histone demethylases, and this biochemistry is firmly established [79]. Its cell-non-autonomous consequences are also documented: tumor-derived 2-HG is taken up by T cells through membrane transporters and suppresses antitumor function [80,81]. Critically, however, the delivery route in these studies is soluble and transporter-mediated, not vesicular, and the reported T-cell effects are metabolic and signaling-level rather than chromatin-level. 2-HG has been detected in small EVs derived from IDH1-mutant cells, but the authors of that analysis note that further work is required to establish whether it is genuinely encapsulated and transferred to recipient cells [82]. Accordingly, EV-mediated oncometabolite transfer causing TET inhibition and a hypermethylation program in recipient stromal or immune cells reaches only Level 1 on the hierarchy applied here (cargo detected in EV preparations); the downstream steps remain hypothetical and are identified in Section 12 as a priority experiment. This assessment is maintained consistently in Section 6.6 and Section 7. The cell-non-autonomous dimension of oncometabolite transfer via EVs is biologically significant: epigenetic silencing of immune-recognition genes in tumor-infiltrating lymphocytes has been documented in hypoxic, metabolite-replete TMEs [78,83], and exosomal delivery of oncometabolite-laden cargo could extend this silencing program beyond the primary tumor niche. Metabolic regulators are best regarded as an emerging integrative cargo class whose contribution to recipient-cell epigenetic states converges with ncRNA-mediated and protein-mediated pathways during therapy response, metastasis, and microenvironmental adaptation [7,38,39,48,83].

4.5. Epitranscriptomic Modifications as sEV Cargo

sEV RNAs carry covalent chemical modifications, collectively the epitranscriptome, which influence their biological function in recipient cells and serve as sorting signals that bias which RNAs are packaged into vesicles. N6-methyladenosine (m6A), the most prevalent internal modification on eukaryotic mRNA and lncRNAs, is installed co-transcriptionally by the METTL3/METTL14 complex and removed by the demethylases FTO and ALKBH5 [84]. In cancer, METTL3 is frequently amplified or overexpressed, and oncogenic m6A deposition reprograms RNA stability, translation efficiency, and alternative splicing to promote tumor progression [85]. The role of m6A as a cargo-sorting signal in exosome biogenesis is mechanistically supported: m6A modifications on miRNA precursors facilitate processing via YTHDF2-mediated DGCR8/DROSHA recruitment, altering the mature miRNA repertoire packaged into EVs [86]. The m6A-modified circRNA circNSUN2 is selectively exported in sEVs via YTHDC1-dependent recognition; following uptake by recipient cells, it stabilizes HMGA2 and promotes hepatic metastatic colonization, a pathway with Level 3–4 mechanistic evidence [13,67]. In gastric and colorectal cancer, tumor-derived EVs enriched for m6A-modified lncRNAs recruit YTHDF1 in recipient macrophages, reprogramming macrophage polarization toward an immunosuppressive, pro-tumorigenic phenotype, a mechanism bridging epitranscriptomic and epigenetic regulation within the TME [13]. METTL3 protein itself has been detected by single-vesicle mass spectrometry in colorectal- and gastric-cancer-derived EVs, raising the possibility that the m6A writer machinery is transferable and could directly reprogram the epitranscriptome of recipient cells [67]. This represents a mechanistically distinct pathway from ncRNA-mediated DNMT or HDAC modulation: rather than altering chromatin enzyme expression, transferred METTL3 would alter the RNA modification landscape and thereby change the post-transcriptional regulation environment of the recipient cell. Formal demonstration of catalytically active METTL3 delivery as opposed to mere protein detection requires activity assays in recipient cells and has not yet been provided; this is therefore a candidate mechanism. The epitranscriptomic dimension of exosomal biology constitutes a critical mechanistic frontier warranting dedicated investigation. Although m6A-modified RNA cargo is not yet represented as a row in Table 1 (the principal cargo classes, their epigenetic mechanisms, and associated functional outcomes are compared), due to the nascent state of direct mechanistic evidence, the epitranscriptomic layer constitutes a functionally important extension of the exosomal epigenetic cargo repertoire described therein, and its systematic incorporation into future summary frameworks is recommended as the field matures.

Table 1. Summary table comparing the major exosomal cargo classes by cancer type, epigenetic mechanism, and functional outcome.

Cargo Class	Cancer Type	Epigenetic Mechanism	Functional Outcome	References
miRNA	Breast, NSCLC (non-small cell lung cancer)	miR-21 suppresses PTEN/PDCD4; miR-29 family inhibits DNMT3A/3B mRNA, reducing CpG methylation; exosomal miR-148a targets DNMT1, promoting promoter demethylation; net effect: altered DNA methylation landscape and H3K27me3 at tumor suppressor loci.	Drug resistance (tamoxifen, cisplatin), EMT induction, cancer stem cell maintenance	[59]
lncRNA (HOTAIR)	CRC, HCC, ovarian cancer	HOTAIR scaffolds PRC2 (EZH2/SUZ12/EED) to gene promoters, depositing H3K27me3; sequesters miR-138-5p, relieving EZH2 suppression; promotes SIRT1-mediated deacetylation at pro-apoptotic loci.	Invasion, metastasis, cisplatin resistance	[15,58,61]
circRNA	Pancreatic ductal adenocarcinoma, Breast cancer (TNBC)	ceRNA sponging of tumor-suppressive miRNAs relieves post-transcriptional repression of chromatin modifiers; chromatin looping via interaction with RNA-binding proteins alters topological domains.	Therapy resistance, lineage plasticity	[68]
EV-DNA	Lung, CRC, pancreatic	Transfer of hypermethylated CpG island DNA from donor cancer cells to recipient cells silences tumor suppressor promoters in trans; 5-hydroxymethylcytosine (5hmC) content serves as cancer-state indicator.	TSG silencing in recipient cells; liquid biopsy biomarker	[71]
DNMT1/3A/3B proteins	Breast, gastric, ovarian	Exosomal DNMT1 transferred to cisplatin-sensitive cells methylates pro-apoptotic gene promoters (e.g., DAPK, RASSF1A), silencing them de novo; DNMT3A deposits new CpG methylation at immune-recognition loci contributing to immune exclusion.	Gene silencing, cisplatin resistance, immune evasion	[17]
Metabolites (acetyl-CoA, SAM)	Pan-cancer (IDH-mutant glioma/AML, SDH/FH-mutant tumors specifically noted)	Acetyl-CoA availability drives HAT activity and H3K27ac deposition at oncogene enhancers; SAM depletion reduces DNMT/HMT activity; alpha-KG is essential TET cofactor for 5mC oxidation; exosomal oncometabolites (2-HG, succinate, fumarate) competitively inhibit alpha-KG-dependent dioxygenases (TET, KDM), inducing hypermethylation.	Metabolites epigenetic rewiring, immune evasion via epigenetic silencing of immune effector genes	[87]
lncRNA/circRNA/miRNA (TMZ axis)	Glioblastoma	Exosomal MGMT mRNA transfer; miR-151a loss derepressing XRCC4; SBF2-AS1 and circWDR62 acting as ceRNAs on the miR-151a-3p/XRCC5 and miR-370-3p/MGMT axes. Post-transcriptional regulation intersecting the MGMT-methylation determinant; chromatin-state transfer not demonstrated.	Temozolomide resistance; serum-exosome MGMT/APNG as epigenetic liquid-biopsy surrogate	[88,89,90,91,92]
miRNA/non-coding Y RNA	AML, CLL, multiple myeloma	AML EV miR-150/miR-155 suppress c-MYB in HSPCs; CLL exosomal hY4 induces PD-L1 in monocytes via TLR7; stromal EVs depleted of miR-15a. miR-29b→DNMT3A/3B axis is transfection-based, not EV-delivered.	Niche subversion, impaired normal hematopoiesis, immune reprogramming, bortezomib resistance	[93,94,95,96,97,98,99,100]

Throughout this review, we distinguish, where the published evidence permits, four levels of mechanistic evidence for sEVs-mediated epigenetic effects that collectively constitute a mechanistically complete causal chain: level (i) demonstration of cargo enrichment within isolated EVs, defined as detection of a given cargo molecule in purified EV fractions meeting MISEV2018/2023 minimum characterization criteria; level (ii) confirmed delivery of cargo to recipient cells, confirmed by cargo internalization assays such as fluorescent labeling and reporter transfer; level (iii) measurable change in recipient-cell chromatin state or DNA methylation pattern, defined as quantifiable alterations in DNA methylation, histone modification, or chromatin accessibility attributable to the transferred cargo; and level (iv) causal contribution of the transferred cargo to a disease-relevant malignant phenotype, demonstrated through functional rescue, genetic complementation, or in vivo validation. Many cited studies provide compelling evidence at levels (i) and (ii), while evidence at levels (iii) and (iv) is available for a more limited subset of cargo-mechanism pairs. Where only levels (i)–(ii) evidence exists, mechanistic claims are qualified as “consistent with,” “suggestive of,” or “hypothesized to involve” epigenetic remodeling, rather than ‘demonstrating’ it. We further distinguish between indirect modulation of chromatin states via post-transcriptional regulation of chromatin modifier mRNAs, which is mechanistically plausible but not equivalent to direct epigenetic remodeling, and bona fide transfer of epigenetic information or enzymatic activity that produces heritable chromatin changes in recipient cells.

5. sEVs-Mediated Epigenetic Alterations in Recipient Cells

The preceding section catalogued the classes of epigenetic cargo carried by sEVs. This section describes how that cargo produces functional epigenetic changes in recipient cells, with attention to the level of mechanistic evidence available for each pathway. Four levels of evidence are used to calibrate claims: Level 1—cargo detected in purified sEV fractions; Level 2—cargo internalization by recipient cells confirmed; Level 3—measurable chromatin or methylation change in recipient cells attributable to the transferred cargo; Level 4—causal contribution to a malignant phenotype demonstrated through functional rescue, genetic complementation, or in vivo validation. Most studies reviewed below provide strong evidence at Levels 1–2, with growing Level 3 evidence, while Level 4 causal validation remains sparse outside of a subset of miRNA and lncRNA examples. Tumor-derived sEVs (TEXs) reprogram recipient stromal and immune cells through horizontal transfer of ncRNAs, proteins, and metabolic factors that converge on chromatin-modifying pathways [17,54,101,102]. TEXs silence tumor suppressors, activate oncogenes, and induce EMT in recipient cells via mechanisms that include the following: (i) miRNA-mediated suppression of DNMT and HDAC mRNAs, altering DNA methylation and histone acetylation patterns; (ii) lncRNA-scaffolded PRC2 recruitment leading to H3K27me3 deposition at pro-apoptotic and tumor-suppressor loci; (iii) direct DNMT enzyme transfer enabling de novo CpG methylation; and (iv) metabolite-mediated inhibition of TET dioxygenases, inducing hypermethylation [54,103,104,105,106]. These mechanisms, individually and in combination, underpin cancer-cell adaptation to the TME and contribute to resistance to chemotherapeutics including tamoxifen, cisplatin, and tyrosine-kinase inhibitors [54,104].

5.1. sEV miRNAs as Transferable Epigenetic Regulators

sEV miRNAs produce post-transcriptional silencing of target mRNAs in recipient cells via RISC loading. When targets include chromatin regulators DNMTs, EZH2, HDACs, or their upstream signaling effectors, the downstream effect is a measurable epigenetic state change. This section focuses on miRNA-cargo interactions supported by at least Level 3 evidence (chromatin or methylation readout in recipient cells). Mechanistically, evidence that exosomes lack key miRNA-biogenesis components and Argonaute proteins [107] confirms that sEVs function as passive miRNA carriers rather than active biogenesis compartments; internalized miRNAs must therefore engage the recipient cell’s pre-existing RISC machinery to exert their effects. This model is supported by multiple studies demonstrating that cell-line or patient-derived exosomal miRNA profiles can be recapitulated in recipient cells following sEV uptake and that the transcriptional consequences mirror those produced by miRNA mimics [25,107,108]. Cancer-associated fibroblast (CAF)-derived sEV miR-29b suppresses DNMT3B in recipient tumor cells, relieving methylation-mediated repression of target gene promoters and altering transcriptional profiles [49]. This represents a Level 3 mechanism: miRNA detection in EVs (Level 1), uptake confirmed (Level 2), and recipient-cell DNMT3B suppression with downstream promoter demethylation documented (Level 3). Similarly, sEV miR-21 targets PTEN and PDCD4, activating PI3K/AKT signaling that converges on epigenetic reprogramming through STAT3-mediated DNMT3A recruitment to immune-effector gene loci [61]. sEV miR-155 downregulates E-cadherin and upregulates mesenchymal markers in recipient non-cancer stem cells (CSCs), triggering EMT, a process accompanied by broad chromatin remodeling at epithelial-program loci [109]. Beyond individual tumor types, tumor- and stromal-derived sEV miRNAs couple distant tissue compartments and shift recipient cells toward phenotypes that favor tumor growth and metastasis [14,25,110]. The systemic dimension of sEV miRNA signaling reflected in circulating biofluid profiles from cancer patients supports models in which sEV-borne miRNAs coordinate epigenetic state transitions across the tumor, vasculature, immune infiltrate, and stroma. However, it is important to note that most studies demonstrate miRNA transfer and downstream phenotypic effects; the chromatin-level mechanism linking RISC-mediated mRNA silencing to durable epigenetic state change in recipient cells requires explicit validation (e.g., ChIP-seq or ATAC-seq) and is not uniformly established across the cited examples.

5.2. sEV lncRNAs and circRNAs as Chromatin Regulators

sEV lncRNAs and circRNAs exert epigenetic effects in recipient cells primarily through two non-mutually exclusive mechanisms: (i) scaffold-directed recruitment of PRC2 or other chromatin-modifying complexes to specific genomic loci and (ii) competitive endogenous RNA (ceRNA) activity, sponging tumor-suppressive miRNAs away from their DNMT or HDAC targets, thereby indirectly altering chromatin states. Both mechanisms are reviewed here with explicit evidence-level annotation. The clearest example of lncRNA-mediated epigenetic reprogramming via exosomes is HOTAIR. Delivered to recipient cells via cisplatin-resistant cell-derived sEVs, HOTAIR scaffolds the PRC2 complex to deposit H3K27me3 at DAPK, PTEN, and related pro-apoptotic loci, propagating cisplatin resistance across previously sensitive tumor cells [15,58,59]. Separately, HOTAIR sequesters miR-138-5p, relieving miR-138-5p-mediated suppression of EZH2 and SIRT1 and amplifying H3K27me3 deposition [111,112]. These convergent mechanisms constitute Level 3–4 evidence for sEV lncRNA-driven epigenetic reprogramming. In HCC, sEV ncRNAs, including lncRNAs, regulate pathophysiological processes in recipient cells and remodel the TME [113]. Pan-cancer reviews acknowledge that mechanistic details in recipient cells remain incompletely defined for most lncRNA cargo beyond HOTAIR [114]. In the ceRNA dimension, computational analyses in CRC and breast cancer map lncRNA/circRNA-miRNA-mRNA networks in circulating sEVs [57,64]; however, the majority of these axis models are based on in silico interaction predictions rather than chromatin-level functional validation in recipient cells, and cross-cancer generalization of specific lncRNA mechanisms should be treated as a hypothesis until experimental chromatin-state evidence is provided. For circRNAs, the mechanistically best-supported sEVs example with epigenetic consequences is m6A-modified circNSUN2, described in Section 4.5. Beyond this, sEV circRNA content is reshuffled in therapy-conditioned pancreatic cancer cells [65], suggesting a mechanism by which treatment pressure alters the epigenetic signaling capacity of secreted vesicles, though chromatin-level consequences in recipient cells remain to be characterized in most contexts.

5.3. sEV Proteins and Transcription Factors in Recipient Cells

sEVs transport mRNAs and proteins between cells, providing a mechanism to alter recipient-cell gene expression and signaling states independent of ncRNA-mediated pathways [6,114,115]. Transcription factors have been explicitly listed among the pathogenic components carried by cancer-derived EVs, and their paracrine delivery contributes to pre-metastatic niche formation and tumor progression [25,110]. The diverse molecular composition of sEVs reflecting the donor-cell state enables multifaceted reprogramming of recipient-cell transcriptional outputs [116]. In the epigenetic context, the most mechanistically specific protein-mediated pathway remains DNMT enzyme transfer (described in Section 4.3). For transcription factor cargo, the epigenetic consequence depends on whether the delivered transcription factor recruits chromatin-modifying complexes to its target loci, a plausible but incompletely characterized mechanism in most published studies. Where such recruitment has been demonstrated (e.g., STAT3-directed DNMT3A recruitment following sEV programmed death-ligand 1 (PD-L1)-mediated STAT3 activation [16,61]), the pathway qualifies as an indirect but mechanistically complete epigenetic reprogramming axis. In other cases, transcription factor delivery produces transcriptional changes through direct DNA binding without documented chromatin-state alterations, and these effects are therefore classified as signaling-mediated rather than epigenetic.

5.4. sEV DNA Transfer and Tumor Heterogeneity

Double-stranded genomic DNA has been detected in circulating sEVs from cancer patients, and cancer-derived vesicles can carry mutant oncogene sequences capable of transferring functional information to recipient cells [116,117]. In hematological cancers, exosome-mediated DNA transfer has been proposed to contribute to tumor heterogeneity by distributing oncogenic elements across subclonal populations [117]. In the epigenetic context, it is important to distinguish between two distinct claims: (i) that transferred DNA carries methylation imprints (a well-supported observation reviewed in Section 4.2); and (ii) that such transfer functionally reprograms recipient-cell chromatin (a hypothesis without rigorous experimental support to date). Most evidence positions sEV DNA primarily as a diagnostic substrate reflecting the epigenetic state of the donor cell rather than as a functional epigenetic reprogramming of recipient cells. Until recipient-cell chromatin integration or methylation-pattern alteration downstream of sEV DNA uptake is demonstrated, this mechanism should be characterized as a candidate pathway rather than an established one.

5.5. Metabolic Reprogramming via sEV Cargo

sEV-mediated metabolic reprogramming of recipient cancer and stromal cells contributes to angiogenesis, metastasis, drug resistance, immunosuppression, and TME remodeling [104]. The mechanistic link between sEV metabolite cargo and recipient-cell epigenetics operates through the obligate metabolic dependency of chromatin enzymes described in Section 4.4: when tumor-derived sEVs deliver oncometabolites or metabolic reprogramming signals that deplete alpha-KG or elevate succinate/fumarate/2-HG in recipient cells, TET and KDM enzyme activity is inhibited, inducing global DNA hypermethylation and altered histone methylation states [7,78,87]. Hypoxia is a universal hallmark of solid tumors and a major driver of sEV biogenesis and cargo composition: hypoxia-conditioned sEVs carry oncogenic ncRNAs, mutant proteins, and metabolites that reprogram stromal cells and prime pre-metastatic niches [118]. Multiple TME stresses, hypoxia, inflammation, and nutrient deprivation, modulate ncRNA expression, and sEVs externalize those stress-adapted ncRNA programs to propagate adaptive states across cell populations [119,120]. Metabolic rewiring in pancreatic ductal adenocarcinoma (PDAC) is associated with mesenchymal plasticity and immune evasion through a convergent epigenetic-metabolic program [121], and mTOR-centered metabolic integration in immune cells provides a pathway-level rationale for why metabolically loaded sEV signals can exert immune-epigenetic consequences at the tissue scale [122]. Collectively, Section 5.1, Section 5.2, Section 5.3, Section 5.4 and Section 5.5 support a multifaceted model in which sEVs deliver ncRNAs, proteins, DNA, and metabolic factors that converge on chromatin-modifying pathways in recipient cells. Through indirect epigenetic modulation (ncRNA-mediated suppression of chromatin enzymes), direct enzyme transfer, and metabolic co-substrate manipulation, sEV-mediated communication promotes phenotypic plasticity, tumor progression, and therapy resistance. The mechanistic completeness of these pathways, particularly the transition from post-transcriptional regulation to durable chromatin-state change, varies across cargo classes and requires explicit experimental validation in each context, as summarized in Figure 2 above.

Figure 2. Mechanistic landscape of exosome-driven epigenetic rewiring in recipient cancer and stromal cells. Schematic overview illustrating both established and proposed mechanisms by which tumor-derived exosomes may transfer epigenetic regulators to recipient cells and reprogram gene expression. Panel (A) (established mechanism): Exosomal miRNAs (e.g., miR-21, miR-29, miR-148a) alter the expression or activity of DNA methyltransferases (DNMTs) and TET demethylases in recipient cells, shifting CpG methylation patterns at tumor suppressor loci; this pathway is supported by functional cargo delivery and recipient-cell methylation readout studies. Panel (B) (established mechanism): Exosome-delivered long noncoding RNAs (e.g., HOTAIR, MALAT1) scaffold PRC2 components (EZH2, SUZ12, EED) in recipient cells, promoting H3K27 trimethylation at target loci and facilitating transcriptional programs associated with EMT and invasion; this pathway is supported by recipient-cell chromatin immunoprecipitation evidence. Panel (C) (established mechanism): Metabolites carried by vesicles (acetyl-CoA, succinate, 2-HG) modulate chromatin-modifying enzyme cofactors, biasing histone acetylation states (HAT/HDAC balance) and favoring oncogenic transcriptional programs; supported by metabolomics and epigenomics data. Panel (D) (proposed model; not yet rigorously demonstrated): Transfer of DNMT3A protein via exosomes may deposit new CpG methylation marks on previously unmethylated promoters in recipient cells, contributing to immune evasion and stable gene repression. This pathway is mechanistically plausible for single-subunit enzyme transfer but has not been validated by orthogonal single-vesicle proteomics demonstrating enzymatically competent DNMT3A delivery; it is depicted as a hypothetical model pending rigorous experimental confirmation.

6. Functional Outcomes of sEV-Driven Regulatory Rewiring Across Cancers

The molecular mechanisms described in Section 4 and Section 5 manifest as tissue-scale phenotypic outcomes that define the biology of cancer progression and therapy resistance. This section maps the key emergent phenotypes, angiogenesis, immune rewiring, stromal activation, metabolic adaptation, and therapy resistance to the sEV-mediated epigenetic signaling axes that drive them. A critical interpretive caveat is maintained throughout: mechanistic principles that appear conserved across cancer types are distinguished from cancer-specific evidence that should not be generalized without independent validation in the relevant cellular, stromal, and genomic context. EV diversity further complicates cross-study comparison: some studies operationally analyze sEVs while referring to them as exosomes, and mechanistic conclusions about cargo-sorting or recipient-targeting specificity are contingent on the vesicle identity confirmed by each study [123,124].

A critical interpretive caveat applies throughout this section: while sEV-mediated epigenetic signaling is emerging as a pan-cancer phenomenon, the mechanistic details, including cargo selection specificity, recipient-cell uptake efficiency, stromal context-dependency, and chromatin-state prerequisites, can differ substantially across tumor types. The evidence reviewed here supports several mechanistic principles that appear to be conserved across malignancies: (i) ESCRT- and ceramide-dependent selective miRNA sorting into exosomes; (ii) lncRNA-scaffolded PRC2 recruitment leading to H3K27me3 deposition; (iii) oncometabolite-mediated TET/KDM inhibition inducing DNA hypermethylation; and (iv) sEVs PD-L1-driven T-cell exhaustion with DNMT3A-dependent effector gene silencing. In contrast, cancer-specific mechanisms, such as HOTAIR-EZH2 cisplatin resistance in ovarian cancer, miR-365-mediated gemcitabine resistance in PDAC, or miR-141-3p-driven JAK/STAT3 angiogenesis in ovarian cancer, should not be generalized to other tumor types without independent validation in the relevant cellular and stromal contexts. This distinction is maintained throughout the following sections.

6.1. sEV-Mediated Angiogenesis and Vascular Remodeling

Tumor-derived sEVs drive neovascularization by transferring pro-angiogenic ncRNAs to endothelial recipient cells, where they activate JAK/STAT3 and NF-κB signaling. In epithelial ovarian cancer, sEVs enriched for miR-141-3p activate these pathways in endothelial cells [125], constituting a Level 2–3 signaling-level rewiring event; whether sustained STAT3 activation produces durable H3K27ac changes at angiogenic gene enhancers, a prerequisite for classifying this as epigenetic reprogramming, has not yet been demonstrated by ChIP-seq in sEV-treated endothelial cells and remains an open question. In colorectal cancer, sEVs mediate coordinated angiogenic and immune-evasive outputs, illustrating multi-compartment regulatory rewiring rather than isolated single-cell effects [124].

6.2. sEV-Mediated Immune Rewiring in the TME

The best-characterized sEV-driven immune epigenetic mechanism is T-cell exhaustion mediated by membrane-anchored exosomal programmed death-ligand (PD-L1) and, mechanistically, exosomal PD-L1-mediated STAT3 activation, recruiting DNMT3A to silence T-cell effector loci, as detailed in Section 9.4. This represents Level 3–4 evidence for sEV-driven immune-epigenetic reprogramming with immediate clinical relevance: it provides a mechanistic rationale for combining exosome biogenesis inhibitors (e.g., GW4869, RAB27A knockout) with DNMT inhibitors and PD-1 blockade. Macrophage reprogramming provides a complementary axis: m6A-modified lncRNAs in tumor-derived EVs recruit YTHDF1 in recipient macrophages, shifting polarizations toward an immunosuppressive, pro-tumorigenic phenotype [67], while exosomal miRNAs promote M2 polarization, Treg expansion, and natural killer (NK) cell suppression [126].

6.3. sEV-Mediated Stromal Activation and ECM Remodeling

sEV-driven stromal rewiring produces CAF activation and ECM remodeling that support invasion, immune exclusion, and therapy resistance. CAFs are instructed by sEV cargo that activates TGF-β, Wnt, Notch, Hedgehog, Hippo, and PI3K/AKT/mTOR signaling, constituting upstream epigenetic remodeling triggers in the stromal compartment [122]. CAF-derived sEVs reciprocally transfer miR-21-5p and miR-29b to tumor cells, silencing PTEN and DNMT3B, respectively, and thereby alter the methylation-dependent transcriptional program in recipient cancer cells [127,128]. This bidirectional sEV crosstalk between tumor cells and stroma creates a self-reinforcing epigenetic ecosystem in which tumor-promoting signals are amplified across compartments [129]. The epigenetic dimension of CAF activation, TGF-β-induced H3K4me3 and H3K27ac changes at fibroblast activation gene enhancers, provides a chromatin-level readout framework for future studies of sEV-driven stromal reprogramming.

6.4. sEV-Mediated Metabolic and Stress Adaptation

Hypoxia, a universal hallmark of solid tumors, directly drives sEV biogenesis, alters cargo composition, and shapes the epigenetic signaling landscape of secreted vesicles [118]. This creates a feed-forward mechanism in which tumor microenvironmental stress amplifies sEVs’ epigenetic signaling to adapt both the tumor and its surrounding stroma. Multiple TME stresses, hypoxia, inflammation, and nutrient deprivation, modulate ncRNA expression through stress-responsive transcriptional programs, and sEVs externalize those ncRNA programs to propagate stress-adapted states across cell populations [119,120]. The epigenetic consequences in recipient cells are mediated through the pathways described in Section 4.4 and Section 5.5: oncometabolite-mediated TET/KDM inhibition, altered HAT/HDAC cofactor availability, and ncRNA-directed chromatin enzyme suppression. mTOR-center metabolic regulation in immune cells provides a pathway-level rationale for why metabolically loaded sEV signals exert immune-epigenetic consequences at the tissue scale [122].

6.5. sEV-Mediated Therapy Resistance and Adaptive Evolution

sEVs mediate non-genetic adaptive resistance by distributing resistance-associated epigenetic programs from drug-exposed minority subpopulations to the broader tumor. Drug-induced remodeling of sEV cargo composition provides the mechanistic route: 5-azacytidine (DNMT inhibitor) treatment alters the miRNA repertoire of leukemic cell-derived sEVs, reprogramming recipient-cell transcriptional profiles in ways that circumvent drug action [93,130]; EZH2 inhibitor (tazemetostat) treatment reshapes the sEV lncRNA landscape, enabling treated cells to re-establish H3K27me3 in adjacent untreated cells via vesicle-mediated transfer a form of epigenetic contagion that propagates resistance across tumor subpopulations [131]; and HDAC inhibitor exposure enriches vesicles for acetylated histone fragments and heat-shock proteins that activate oncogenic program in neighboring cells [132]. These observations support a therapeutic rationale for co-targeting EV biogenesis alongside epigenetic agents to interrupt intercellular propagation of adaptive resistance-associated chromatin states. Specific resistance mechanisms including drug efflux (miR-1246/Cav1/ABCB1 axis), DNA damage repair (miR-151a/XRCC4), EMT (miR-155/E-cadherin), and immunosuppressive remodeling (exosomal PD-L1/STAT3/DNMT3A) are detailed in Section 9.

6.6. sEV-Driven Epigenetic and Regulatory Rewiring in Glioblastoma

Glioblastoma (GBM) occupies a distinctive position: it is the tumor type in which functional EV-mediated transfer was first demonstrated, yet its epigenetic dimension remains less mechanistically closed than publication volume suggests. Glioma-derived vesicles were among the first shown to carry mRNA and miRNA that are translated and functional in recipient cells and detectable in-patient serum [133] and to transfer the constitutively active receptor variant EGFRvIII to EGFRvIII-negative cells, conferring MAPK/Akt activation and anchorage-independent growth [134]. These established Level 2–4 evidence for functional cargo transfer more than a decade before the epigenetic framing adopted here.

The most developed axis concerns temozolomide (TMZ) resistance, where epigenetic and vesicular determinants are unusually entangled. TMZ sensitivity is governed principally by promoter methylation of O⁶-methylguanine-DNA methyltransferase (MGMT), a chromatin-level property of the tumor genome, whereas sEVs modulate the same axis through post-transcriptional cargo. Reactive astrocyte-derived exosomes deliver MGMT mRNA and confer resistance [92]. Loss of exosomal miR-151a in resistant GBM derepresses XRCC4 and enhances non-homologous end-joining; restoring miR-151a via serum- or CSF-derived exosomes resensitizes recipient cells, satisfying Level 4 criteria through functional rescue [91]. The lncRNA SBF2-AS1, driven by ZEB1 and exported in exosomes, sequesters miR-151a-3p to relieve repression of XRCC5, and its serum-exosome abundance tracks poor TMZ response [90]; exosomal circWDR62 acts analogously through miR-370-3p/MGMT [89]. A conceptual caution is warranted: these mechanisms determine DNA-repair capacity and matter clinically, but the chromatin-level determinant—MGMT promoter methylation—is a property of the recipient genome rather than of transferred cargo. Under the framework of Section 4, they are forms of post-transcriptional regulation intersecting an epigenetic determinant, not epigenetic information transfer sensu stricto. The translational corollary is nonetheless substantial: microfluidic profiling of serum exosomes for MGMT and APNG transcripts distinguishes patients by tumor methylation status and tracks response longitudinally [88], establishing sEV cargo as a minimally invasive surrogate for a biomarker that otherwise requires tissue.

Glioma also supplies the field’s clearest oncometabolite epigenetics, though not through vesicles. Mutant IDH1 is sufficient to establish the glioma CpG island methylator phenotype in astrocytes [135,136], and the mechanism, competitive inhibition of α-ketoglutarate-dependent TET dioxygenases and KDM demethylases by 2-hydroxyglutarate, is biochemically established [79,137]. Whether tumor sEVs export 2-HG at concentrations sufficient to inhibit TET activity in recipient stromal or immune cells has not been demonstrated in any primary study; the relevant literature remains separate. We therefore state cell-non-autonomous oncometabolite transfer via sEVs as a testable hypothesis rather than an established pathway and identify it in Section 12 as a priority experiment. The same caution applies to H3K27M-mutant diffuse midline glioma, where the mutant histone inhibits PRC2 and creates EZH2 dependency [138,139] but no EV-mediated mechanism has been reported.

Where GBM sEVs do produce documented microenvironmental reprogramming is in the myeloid compartment: hypoxic glioma exosomes deliver miR-1246, which targets TERF2IP and drives M2 macrophage polarization through STAT3 and NF-κB signaling, with in vivo validation [140]. This reaches Level 4 for phenotype; chromatin-level readouts in polarized macrophages have not been reported and would convert a signaling observation into demonstrated epigenetic reprogramming. Therapeutically, the capacity of engineered EVs to cross the blood–brain barrier [141] makes glioma the most plausible near-term indication for sEV-delivered epigenetic agents, exemplified by exosomes loaded with PDGFRβ siRNA acting through the PI3K/Akt/EZH2 axis [23].

6.7. sEV-Driven Epigenetic and Regulatory Rewiring in Leukemia and Other Hematologic Malignancies

Hematologic malignancies are the archetypal epigenetically driven cancers: recurrent lesions in DNMT3A, TET2, IDH1/2, ASXL1, and EZH2 define disease subtypes, and hypomethylating agents are standard of care in acute myeloid leukemia (AML) and myelodysplastic syndromes. The comparative scarcity of chromatin-level readouts in the leukemic EV literature is therefore conspicuous.

The best-evidenced axis is reprogramming of the bone marrow niche by AML-derived EVs. AML exosomes induce DKK1 in stroma while downregulating the hematopoietic support factors CXCL12, KITL, and IGF1, converting the niche into a leukemia-permissive and hematopoiesis-hostile environment; Rab27a disruption delays progression, and DKK1 inhibition prolongs survival in vivo [100]. AML EVs additionally deliver miR-150 and miR-155 to hematopoietic stem and progenitor cells, suppressing c-MYB translation and impairing clonogenicity [99], and transfer coding and non-coding RNA to stroma with consequent alteration of growth-factor secretion [98]. These satisfy Level 4 criteria for niche subversion, though the readouts are transcriptional and secretory rather than chromatin-level.

Reciprocal stroma-to-tumor transfer is established in multiple myeloma: patient-derived bone-marrow stromal exosomes are depleted of tumor-suppressive miR-15a, enriched in oncogenic protein cargo, and promote tumor growth in vivo, whereas healthy-donor exosomes inhibit it [97]; stromal exosomes likewise confer bortezomib resistance [96]. In chronic lymphocytic leukemia, exosomal transfer of the non-coding Y RNA hY4 to monocytes triggers TLR7-dependent cytokine release and PD-L1 induction, and TLR7 inhibition attenuates disease in vivo [95], a Level 4 demonstration that a vesicle-delivered non-coding RNA reprograms an immune compartment, although durable chromatin remodeling at the CD274 locus has not been tested.

The connection to the epigenetic machinery itself is where evidence thins, and we now state this directly. miR-29b directly targets DNMT3A and DNMT3B and indirectly targets DNMT1, producing global hypomethylation and re-expression of p15INK4b and ESR1 in AML cells [93], a Level 3 demonstration of recipient-cell methylation change, but achieved by transfection rather than vesicular delivery. We are not aware of a primary study reconstituting this axis with EV-delivered miR-29b; any inference that leukemic sEVs reprogram recipient methylomes by this route is an extrapolation and is flagged as such in Table 1. Similarly, IDH1/IDH2 mutations produce a TET2-disrupting hypermethylation phenotype in AML [94] paralleling G-CIMP in glioma, but EV-mediated oncometabolite transfer is again undemonstrated.

Finally, because epigenetic drugs are used clinically here, this setting offers the clearest illustration of therapy-induced cargo remodeling as a resistance mechanism, and the most direct rationale in any tumor type for co-targeting EV biogenesis alongside epigenetic therapy. This argues for incorporating serial sEV cargo profiling into trials of hypomethylating agents and EZH2 inhibitors.

7. sEV-Mediated DNA Methylation and Demethylation in Cancer

An expanding corpus of research characterizes exosomes not solely as inert byproducts of cellular metabolism but as dynamic facilitators of epigenetic reprogramming within the tumor microenvironment (TME). Fundamentally, DNA methylation, the covalent addition of a methyl group to the five-carbon position of cytosine, predominantly at CpG dinucleotides, is the most extensively studied epigenetic modification and is critically important for the initiation, proliferation, and metastasis of many tumors [11,52,142]. In the realm of oncology, irregularities in DNA methylation profiles, characterized by widespread hypomethylation and localized hypermethylation at the promoters of tumor suppressor genes, serve as defining features of neoplastic progression [142,143]. sEVs participate in the modulation of DNA methylation through at least two principal mechanisms. First, exosomes transport methylated DNA fragments directly from donor cancer cells to recipient cells, potentially transferring the methylation “imprint” of the originating tumor; evidence that cancer cells disseminate hypermethylated DNA segments via sEVs and that this leads to transcriptional silencing of recipient-cell tumor suppressors has been documented (Level 1–2 evidence; Level 3 functional methylation transfer in recipient cells not yet rigorously demonstrated) [11,50]. Second, exosomes deliver DNA methyltransferases (DNMTs) and regulatory RNAs that modulate the methylation machinery in recipient cells; exosomal DNMT1 transfer in ovarian cancer, where recipient cells acquire elevated DNMT activity and cisplatin resistance, represents one of the strongest examples of this pathway, with evidence reaching Level 3–4 [17,21].

A well-characterized indirect mechanism involves exosomal miRNAs that target DNMT mRNAs, inducing DNA demethylation in recipient cells. miR-29 family members suppress DNMT3A and DNMT3B, producing genome-wide hypomethylation (Level 3 evidence: methylation change in recipient cells documented) [49]. In the context of cancer-associated fibroblasts (CAFs), a critical stromal component of the TME, CAF-derived sEV miR-29b has been shown to suppress DNMT3B in recipient tumor cells, relieving methylation-mediated repression and altering transcriptional profiles (Level 3 evidence) [128]. This mechanism directly links sEV-mediated intercellular communication to epigenetic reconfiguration of the tumor transcriptome. Bovine milk-derived sEVs containing miR-148a and miR-29b demonstrate the same DNMT-targeting mechanism in a non-oncological context (Level 2–3 translational parallel) [144,145]; the mechanistic relevance to cancer lies in the shared DNMT-suppressive circuitry, not in the milk context itself. A mechanistically distinct axis of sEV-mediated DNA methylation control involves the TET family of dioxygenases, such as TET1, TET2, and TET3, which catalyze iterative oxidation of 5-methylcytosine (5mC) to 5-hydroxymethylcytosine (5hmC) and ultimately to unmodified cytosine, thereby executing active DNA demethylation [146]. In cancer, TET enzymes are frequently mutated or epigenetically silenced, producing global 5hmC depletion, a near-universal epigenetic hallmark of malignant transformation [147]. sEV cargo can modulate this axis bidirectionally. Mutant IDH1/IDH2 enzymes produce 2-hydroxyglutarate (2-HG) [148], which competitively inhibits TET activity within the producing cell [79]. Whether tumor-derived sEVs deliver 2-HG to recipient stromal or immune cells at concentrations sufficient to inhibit TET and establish a hypermethylation program has not been demonstrated (Level 1 evidence only: 2-HG detected in sEV preparations from IDH1-mutant cells [82]; recipient-cell TET inhibition and methylome profiling not reported) and is treated here as a hypothesis rather than an established mechanism, consistent with Section 6.6.

The translational relevance of this mechanism is underscored by studies demonstrating that 5hmC levels in circulating cell-free DNA and sEV-associated DNA fragments serve as sensitive, tissue-of-origin-specific diagnostic and prognostic biomarkers across multiple cancer types (Level 1–2 evidence for biomarker utility; functional epigenetic reprogramming of recipient cells by sEV-borne 5hmC-bearing DNA fragments has not been demonstrated; Level 3 not yet reached) [72,74]. Future studies should characterize how sEV cargo shapes the TET-5hmC landscape in recipient immune and stromal cells, where demethylation-associated transcriptional reactivation could determine the balance between immune activation and suppression within the TME.

8. sEV-Mediated Histone Modification in Cancer

Post-translational histone modifications, including acetylation, methylation, phosphorylation, ubiquitination, and sumoylation, constitute a critical layer of epigenetic regulation that controls chromatin structure and gene accessibility [11,52,145]. As established for DNA methylation in Section 7, evidence-level calibration is applied here: Level 1 refers to histone-modifying enzyme or ncRNA detection in sEV fractions; Level 2 to confirmed uptake by recipient cells; Level 3 to measurable histone mark changes in recipient cells attributable to sEV cargo; and Level 4 to causal in vivo or functional rescue validation. In the context of T-cell dysfunction in the TME, studies indicate that mRNAs and proteins in tumor-derived EVs regulate HDAC activity in T cells, with the potential to reshape effector gene chromatin accessibility through histone modification [75]. HDAC3 is required for T-cell maturation, and HDAC1/2 participate in proper thymic development; tumor-derived EVs that modulate HDAC expression or activity in immune cells could therefore remodel the immune landscape of the TME through histone modification-dependent mechanisms (Level 1–2 evidence; Level 3 chromatin readout in sEV-treated T cells not systematically demonstrated) [75].

The principal route by which sEVs influence histone modification is through the transfer of lncRNAs that recruit chromatin-modifying complexes to specific genomic loci. lncRNAs function as scaffolds or guides for the Polycomb Repressive Complex 2 (PRC2), comprising its catalytic component Enhancer of Zeste Homolog 2 (EZH2), SUZ12, and EED, which deposits H3K27 trimethylation (H3K27me3), a repressive histone mark [143]. Several lncRNAs carrying this epigenetic function are present in sEV cargo secreted by tumors [143]; their transfer to recipient cells constitutes a mechanistically direct route to heritable histone modification. In gastric cancer, HDACs repress tumor-suppressor lncRNAs and miRNAs, while histone methyltransferases and demethylases modulate oncogenic lncRNA expression, illustrating a bidirectional relationship in which histone modifications regulate sEV ncRNA content and sEV ncRNAs in turn modulate histone marks in recipient cells (Level 2–3 evidence for bidirectional crosstalk; causal Level 4 in vivo validation remains limited) [52]. The well-characterized example of EZH2-mediated H3K27me3 deposition at PTEN loci in the context of pre-metastatic niche formation (Level 3 evidence for H3K27me3 deposition at specific loci attributed to sEV-delivered cargo) provides mechanistic grounding for this class of histone-modifying sEV cargo [149].

As described in Section 4.1.2, the lncRNA-scaffolding model whereby sEV-delivered lncRNAs nucleate PRC2 assembly at specific loci in recipient cells is mechanistically well-supported by studies of HOTAIR and MALAT1 function in cancer cells (Level 3–4 evidence for HOTAIR/PRC2/H3K27me3 for clinical resistance implications: cisplatin resistance; see Section 9.6) [15,61]. In contrast, the direct transfer of assembled, catalytically competent PRC2 or NuRD complexes via sEVs remains unresolved. Intact multi-subunit complexes require the assembly of 3–7 protein subunits for catalytic activity [76], and whether such assemblies can survive the intraluminal EV milieu and be delivered in enzymatically competent form to recipient cells has not been rigorously demonstrated (Level 1 evidence only for EZH2 protein detection in EVs; Level 3 evidence for catalytically active complex delivery not yet achieved). The relative quantitative contributions of (i) lncRNA-templated PRC2 nucleation within recipient cells versus (ii) direct transfer of EZH2 monomer as single-subunit cargo are experimentally unresolved. Single-subunit epigenetic enzymes such as DNMT3A or EZH2 in isolation represent mechanistically more plausible EV cargo candidates, as their functional delivery requires only protein-folding fidelity rather than multi-protein complex reassembly in the recipient cytoplasm (Level 3–4 evidence for DNMT1 transfer in ovarian cancer [17]; Level 1–2 evidence only for EZH2 monomer transfer). Future studies employing proximity ligation assays, co-immunoprecipitation within isolated EVs, and single-vesicle proteomics should prioritize resolving this hierarchy before claims of PRC2 complex transfer are made.

9. sEV-Mediated Epigenetic Therapy Resistance

Resistance to chemotherapeutic agents and targeted therapeutic modalities continues to pose a major clinical challenge in oncology, with exosome-mediated epigenetic mechanisms increasingly recognized as critical drivers of therapeutic failure [12]. sEV cargo orchestrates resistance through interconnected pathways: drug efflux, DNA damage repair, metabolic reprogramming, apoptosis evasion, and maintenance of cancer stem cell properties. Among these cargo classes, sEV miRNAs are especially influential in modulating gene expression to diminish therapeutic efficacy [109]. An adaptive dimension of resistance involves drug induced remodeling of exosome secretion and cargo composition: exposure to the DNMT inhibitor, 5 azacytidine (5 AZA), alters the miRNA repertoire of leukemic cell derived exosomes and thereby reprograms recipient cell transcriptional profiles in ways that may circumvent drug action [93,130]. Likewise, treatment with EZH2 inhibitors (for example, tazemetostat) can provoke compensatory upregulation of alternative PRC2 subunits and reshape the sEV lncRNA landscape, enabling treated cells to re-establish H3K27me3 marks in adjacent, untreated cells via vesicle mediated transfer, a form of epigenetic contagion that propagates resistance across tumor subpopulations [131]. HDAC inhibitor exposure has similarly been reported to increase exosome secretion and to enrich vesicles for acetylated histone fragments and heat shock proteins that activate oncogenic programs in neighboring cells [132]. Collectively, these observations support a therapeutic rationale for co-targeting EV biogenesis (for example, with neutral sphingomyelinase inhibitors such as GW4869) alongside epigenetic agents to interrupt intercellular propagation of adaptive, resistance-associated chromatin states.

9.1. Drug Efflux and Metabolic Resistance

sEV miR-1246, secreted by ovarian cancer cells, inhibits Caveolin-1 (Cav1) and upregulates ABCB1 (P-glycoprotein) expression in recipient cells, conferring a drug-resistant phenotype by enhancing chemotherapeutic efflux (Level 3–4 evidence: phenotypic rescue via Cav1 restoration validated) [109,150]. Conversely, sEV-transmitted miR-128-3p downregulates MDR5 expression, decreasing oxaliplatin efflux and restoring chemosensitivity in oxaliplatin-resistant colorectal cancer (CRC) cells, illustrating the bidirectional regulatory potential of sEV miRNAs on transporter-mediated resistance (Level 3 evidence: MDR5 protein reduction in recipient cells and functional sensitivity assays documented) [151]. Tumor-associated macrophage (TAM)-derived exosomes transfer miR-365 to pancreatic ductal adenocarcinoma (PDAC) cells, increasing triphosphate nucleotide (NTP) pool levels that compete with phosphorylated gemcitabine for DNA incorporation and upregulating cytidine deaminase (CDA) to inactivate gemcitabine, generating resistance through a coordinated metabolic-epigenetic mechanism (Level 3–4 evidence: in vivo gemcitabine resistance phenotype validated with genetic rescue) [152].

9.2. DNA Damage Repair and Apoptosis Evasion

Exosomal miR-151a directly represses XRCC4, a core mediator of non-homologous end joining (NHEJ). In temozolomide (TMZ)-resistant glioblastoma multiforme (GBM) cells, loss of exosomal miR-151a derepresses XRCC4, enhancing NHEJ-mediated repair of TMZ-induced lesions and conferring resistance; resistant cells transmit this phenotype to sensitive recipients in a miR-151a loss-dependent manner [91]. Restoring miR-151a inhibits XRCC4-mediated repair and resensitizes recipient GBM cells to TMZ, validating exosomal miRNA as both a mechanistic driver and a therapeutic target (Level 4 evidence: cargo transfer, target suppression, and functional rescue all demonstrated) [91,109,153]. Cancer-associated adipocyte (CAA)- and CAF-derived sEV miR-21 iso miRNAs bind APAF1 in ovarian cancer cells, suppressing apoptosome formation and downstream activation of caspase-9 and caspase-3, thereby conferring paclitaxel resistance (Level 3 evidence: caspase pathway suppression documented in recipient cells; in vivo validation limited) [109].

9.3. Epithelial-Mesenchymal Transition and Cancer Stem Cells

Epigenetic differences between CSCs and non-CSCs, driven substantially by EMT, underlie the plastic, therapy-resistant CSC phenotype [109]. sEV miR-32-5p promotes multidrug resistance in hepatocellular carcinoma (HCC) by activating the PI3K/AKT pathway to drive EMT and angiogenesis (Level 3 evidence: pathway activation and EMT marker changes in recipient cells demonstrated) [109,154]. sEV miR-155, enriched in vesicles secreted by both CSCs and drug-resistant cells, downregulates E-cadherin and upregulates mesenchymal markers in recipient non-CSC cells, triggering EMT and conferring drug resistance (Level 3 evidence: chromatin-level changes at E-cadherin promoter not explicitly quantified; phenotypic EMT markers documented Level 3 partial) [109]. In PDAC, gemcitabine-resistant CSCs transfer miR-210-enriched exosomes to gemcitabine-sensitive cells, propagating chemoresistance across the tumor cell population (Level 2–3 evidence: miRNA transfer confirmed; methylation readout in recipient cells not yet provided) [109]. These mechanisms collectively exemplify intercellular “resistance contagion,” in which a minority of drug-resistant cells epigenetically reconfigure the broader tumor microenvironment.

9.4. Immunosuppressive Epigenetic Remodeling by Tumor-Derived sEVs

Resistance to anti-PD-1/PD-L1 immune checkpoint therapy remains a major unresolved clinical problem. Exosomal PD-L1, when membrane-anchored on the outer leaflet of tumor-derived sEVs, suppresses T-cell receptor signaling and induces T-cell exhaustion in tumor-draining lymph nodes and the TME, constituting a systemic immunosuppressive axis that actively antagonizes PD-1 blockade (Level 3–4 evidence) [18]. Mechanistically, sEV PD-L1-mediated STAT3 activation recruits DNMT3A to silence granzyme B, perforin, and IFN-γ loci in CD8+ T cells, establishing a heritable transcriptional silencing program that cannot be reversed by checkpoint antibody monotherapy alone (Level 3–4 evidence: locus-specific methylation in exhausted T cells demonstrated) [16]. In esophageal cancer, a complementary mechanism operates: exosomal PD-L1 activates the STAT3/miR-21/PTEN/AKT signaling axis while simultaneously suppressing CTL activity, a dual epigenetic and immunosuppressive effect (Level 3 evidence) [155,156]. Collectively, these findings provide a mechanistic rationale for combining exosome biogenesis inhibitors (see Section 10.1.1) with DNMT inhibitors, TET activators, and PD-1 blockade to address both the extrinsic immunosuppressive EV signal and the cell-intrinsic epigenetic silencing it induces. Tumor sEVs additionally transfer miRNAs that promote M2 macrophage polarization, expand Tregs, and suppress NK cell function, collectively amplifying immunosuppressive reprogramming across the TME [126]. This self-reinforcing epigenetic ecosystem involving CAFs, endothelial cells, myeloid-derived suppressor cells (MDSCs), and tumor cells is reviewed in Section 6.2 [129].

9.5. Radiotherapy Resistance and Epigenetic Bystander Effects via sEVs

sEVs have been implicated in radiation-induced bystander effects (RIBE), whereby irradiated cells transmit epigenetic signals to non-irradiated neighboring cells via sEV miRNA and lncRNA cargo, resulting in radiotherapy resistance and DNA-damage responses in distant cells (Level 2–3 evidence: cargo transfer confirmed; epigenetic readout specific chromatin marks in bystander cells documented in a subset of studies; Level 4 in vivo therapeutic rescue limited) [157]. This mechanism is particularly relevant in heterogeneous tumors, where radioresistant minority subpopulations can epigenetically confer resistance to the broader tumor population through sEV-mediated communication, thwarting the intended anti-tumor effect of ionizing radiation.

9.6. Cisplatin and Platinum-Based Resistance via sEVs Epigenetic Crosstalk

Platinum-based chemotherapy (cisplatin, oxaliplatin, carboplatin) forms the therapeutic backbone across multiple major cancers, and cisplatin resistance remains a leading cause of treatment failure in ovarian, bladder, head-and-neck, and non-small cell lung cancers (NSCLC). sEV-mediated epigenetic resistance to platinum-based agents operates through several converging mechanisms. CAF-derived exosomes enriched for miR-21-5p silence PTEN and PDCD4 in epithelial ovarian cancer cells, activating AKT/PI3K signaling and conferring cisplatin resistance through inhibition of caspase-3-dependent apoptosis (Level 3 evidence: PTEN/PDCD4 suppression and AKT activation in recipient cells confirmed; apoptotic rescue demonstrated in vitro) [127]. Cisplatin-resistant ovarian cancer cells overexpress lncRNA HOTAIR, which sequesters miR-138-5p and thereby relieves miR-138-5p-mediated suppression of EZH2 and SIRT1; the resulting increase in EZH2 catalytic activity deposits H3K27me3 at pro-apoptotic loci, including DAPK and PTEN, propagating cisplatin resistance across previously sensitive tumor cells (Level 3–4 evidence: H3K27me3 ChIP-seq at specific loci confirmed; in vitro and in vivo functional rescue by HOTAIR knockdown demonstrated) [111,112]. This mechanism provides a pharmacological rationale for combining EZH2 inhibitors (e.g., tazemetostat, GSK126) with platinum-based chemotherapy in HOTAIR-high tumors [111,112].

10. Translational Landscape

10.1. Therapeutic Strategies Targeting sEV-Mediated Epigenetic Pathways

The clinical advancement of engineered exosome therapies requires rigorous characterization of pharmacokinetic (PK) and pharmacodynamic (PD) parameters that remain poorly defined for most EV-based modalities. Electroporation is the most widely studied active loading method for nucleic acid cargo, achieving loading efficiencies for miRNAs of approximately 10–25% per loading cycle, with significant batch-to-batch variability driven by RNA aggregation on the exosomal surface and disruption of vesicle integrity at high field strengths [158]. Co-incubation (passive loading) achieves substantially lower efficiencies (~2–5%) but preserves membrane integrity and is more scalable. Novel membrane-active loading approaches, including saponin-mediated transient permeabilization and click chemistry-based surface conjugation, are emerging as alternatives with improved efficiency and reproducibility [159]. In terms of pharmacokinetics, intravenously administered exosomes demonstrate rapid hepatic and splenic clearance (half-life: 15–30 min in murine models), with preferential accumulation in Kupffer cells and sinusoidal endothelium rather than tumor tissue in the absence of surface targeting modifications [160]. Surface decoration with tumor-homing peptides (such as iRGD, GE11, anti-EGFR nanobodies) significantly improves tumor biodistribution but adds manufacturing complexity and potential immunogenicity. Regarding safety, the inherent miRNA pleiotropy, individual miRNAs regulate hundreds of target mRNAs, which raises the risk of off-target epigenetic effects in non-tumor tissues; systematic off-target profiling using RNA sequencing and ATAC-seq in tissues of accumulation (liver, spleen, lung) should be standard in preclinical IND-enabling packages for any exosome-miRNA therapeutic candidate [161]. The immunogenicity of allogeneic exosome preparations, particularly those derived from non-autologous mesenchymal stem cells (MSCs), may trigger complement activation or adaptive immune responses that accelerate clearance and reduce therapeutic efficacy, necessitating immunological profiling as part of quality control [162]. The varied and sophisticated functions of exosomes in the epigenetic control of cancer have disclosed a vast range of treatment possibilities. These can be broadly categorized into (a) inhibition of exosome biogenesis and secretion, (b) engineering of exosomes as drug delivery platforms, (c) loading of tumor-suppressive miRNAs, siRNAs, and epigenetic reprogramming agents, and (d) exosome-based cancer immunotherapy. These four strategies are shown schematically in Figure 3.

Figure 3. Therapeutic targeting proposed strategies for exosome-mediated epigenetic pathways. Diagram illustrating proposed intervention points to modulate exosome-driven epigenetic communication between tumor and recipient cells, where exosome cargo transfers (miRNAs, lncRNAs, DNA, and proteins) are shown influencing epigenetic reprogramming in recipient cells. Panel (a): Biogenesis inhibition: blocking exosome production/release using agents (e.g., GW4869) or genetic approaches (e.g., RAB27A targeting) to limit downstream effects; Panel (b): Cargo interception: neutralizing oncogenic exosomal contents (e.g., miRNAs) via antisense or RNA interference; Panel (c): Engineered delivery: using modified exosomes to deliver therapeutic cargos (tumor-suppressive miRNAs, epigenetic drugs) to reprogram aberrant states; Panel (d): Combination immunotherapy: integrating exosome-targeting with immune checkpoint blockade to enhance anti-tumor responses.

10.1.1. Inhibition of Exosome Biogenesis and Release

Given that cancer cells secrete exosomes at markedly elevated levels compared to normal cells, pharmacological inhibition of exosome biogenesis and release represents an attractive strategy for blocking the intercellular epigenetic communication that sustains tumor progression [23]. The neutral sphingomyelinase 2 (nSMase2) inhibitor GW4869 impedes both the biosynthesis and release of exosomes in multiple cancer cell types, including ovarian cancer and prostate cancer, and has been shown to enhance sensitivity to trastuzumab in HER2-overexpressing breast cancer cells and to reverse paclitaxel resistance conferred by exosomal PD-L1 in esophageal cancer [8]. GW4869 has additionally been shown to enhance the efficacy of PD-L1 checkpoint blockade by stimulating CTL activity in a mouse melanoma model, highlighting the potential for combining exosome inhibitors with immune checkpoint therapies [23]. RAB GTPases, which regulate exosome biogenesis, represent additional pharmacological targets. Tipifarnib, a pharmacological substance aimed at RAB27A, effectively diminishes the exosome count and the expression of PD-L1 across various cancer cell lines, thereby enhancing susceptibility to sunitinib [163]. Knockout of RAB27A in B-cell lymphoma enhances sensitivity to chemo-immunotherapy regimens, validating the utility of targeting exosome secretory machinery as a means of overcoming resistance [163].

10.1.2. Engineered sEV Delivery Vehicles for Epigenetic Cargo

Exosomes exhibit intrinsic properties that make them favorable vehicle properties; however, the unmodified particle is described as a container; they do not establish that native vesicles carry therapeutic cargo at usable concentrations, which is the subject of the following section: they display low toxicity and immunogenicity, excellent biodegradability, efficient tissue-targeting capabilities, the ability to cross physiological barriers, including the blood–brain barrier, and a double-membrane structure that protects encapsulated cargo from enzymatic degradation [23]. These properties are not confined to neural applications: the capacity of exosomes to traverse restrictive anatomical barriers, including the corneal epithelium, blood–retinal barrier, and conjunctival membrane, further validates their universal potential as delivery platforms for epigenetic cargo in compartmentalized disease sites where conventional nanoparticles fail to achieve therapeutic concentrations [164]. These properties position engineered exosomes as a compelling alternative to synthetic nanoparticle delivery systems, particularly for the delivery of nucleic acid-based epigenetic therapeutics that are vulnerable to degradation in the systemic circulation [23]. A range of techniques has been established for the incorporation of therapeutic agents into exosomes, encompassing co-incubation, electroporation, sonication, extrusion, and endogenous loading achieved via genetic alteration of donor cells [23]. Electroporation has demonstrated particular utility for loading miRNAs and siRNAs; expanded natural killer cell (eNK-EXO)-derived exosomes loaded with cisplatin via electroporation enhance its cytotoxic effect against drug-resistant ovarian cancer (OC) cells, and umbilical cord blood-derived macrophage exosomes loaded with cisplatin exhibit greater potency against cisplatin-resistant ovarian cancer cells than free cisplatin [165,166]. BMSC-derived exosomes loaded with doxorubicin via extrusion show enhanced efficacy against osteosarcoma with fewer side effects than the free drug, while exosome-delivered cisplatin evades endosomal trapping through clathrin-independent endocytosis, achieving more uniform cytosolic distribution [167].

Why Engineering Is Necessary Rather than Optional

Four constraints make native sEVs unsuitable as carriers of therapeutic epigenetic instruction, each mapping to a specific intervention.

The first is stoichiometric. Absolute quantification across several sEV sources established that there is, on average, far less than one molecule of any given miRNA per vesicle, even for the most abundant species [168]. An effect requiring engagement of RISC or of chromatin-modifying machinery cannot be achieved by a carrier whose modal cargo content is zero copies; active loading is therefore a precondition for function, not an enhancement of it.

The second is compositional. Native vesicles carry the donor cell’s full repertoire, including oncogenic species; the same vesicles proposed as vehicles transfer transforming receptors and pro-tumorigenic miRNA when unmodified (Section 5 and Section 6.6). Producer-cell engineering that enriches a defined cargo while depleting endogenous species is what makes the payload definable, which is in turn a regulatory prerequisite for a potency assay.

The third is pharmacokinetic. Systemically administered EVs distribute predominantly to the liver and spleen with a circulating half-life of tens of minutes, and biodistribution is governed by cell source, route, and surface composition [160,169]. Without surface modification, the epigenetic agent is delivered principally to the reticuloendothelial system.

The fourth is specificity. The pleiotropy that makes miRNA attractive as a network-level regulator makes it hazardous therapeutically: a single species engages hundreds of transcripts, so off-target chromatin consequences in accumulation organs are expected rather than exceptional [161]. This argues for a shift in cargo class from miRNA mimics toward locus-specific epigenome editors, whose specificity derives from a programmable guide rather than seed-match promiscuity.

Methods for Engineering sEVs and Their Epigenetic Cargo

Strategies fall into three families, distinguished by whether modification occurs in the producer cell, on the vesicle surface, or within the isolated vesicle.

Producer-cell engineering exploits native sorting machinery to load cargo during biogenesis, preserving membrane integrity. The EXOtic system combines a booster that increases vesicle production with an intraluminal RNA-packaging device and a cytosolic delivery helper promoting endosomal escape, achieving functional mRNA delivery in vivo [170]. The EXPLOR platform uses a blue-light-reversible CIBN–CRY2 interaction to dock soluble protein cargo to the luminal face of CD9 during budding and release it afterwards [171], directly applicable to loading DNMT, TET, or dCas9-effector proteins, whose delivery in catalytically competent form is identified in Section 4.3 and Section 12 as the field’s principal unresolved question. This route is the most scalable and GMP-compatible but affords the least control over copy number per vesicle.

Surface display and targeting. Fusion of targeting peptides to vesicle-resident proteins, most commonly Lamp2b, redirects biodistribution: RVG-Lamp2b exosomes cross the blood–brain barrier to deliver siRNA to neurons and microglia after systemic injection [141]; GE11-displayed exosomes deliver let-7a to EGFR-positive tumors [172]; iRGD-Lamp2b vesicles achieve integrin-mediated targeting [173]. Post-isolation modification avoids producer-cell manipulation; the CD63-binding anchor peptide CP05 permits capture and surface loading irrespective of vesicle origin [174], and copper-free click chemistry allows covalent conjugation without membrane disruption. Producer-cell CD47 overexpression provides a complementary “don’t-eat-me” signal that reduces monocyte clearance [175].

Exogenous loading into isolated vesicles. Electroporation is most widely used for nucleic acids, with efficiencies of roughly 10–25% per cycle but appreciable batch variability arising from surface RNA aggregation and membrane disruption at high field strength [158]. Passive co-incubation preserves integrity at substantially lower efficiency. Sonication, extrusion, freeze–thaw cycling, saponin permeabilization, and pH-gradient loading occupy intermediate positions on the efficiency–integrity trade-off [159]. Because membrane integrity determines both uptake and immunogenicity, the loading method should be reported as a formal experimental variable.

Programmable epigenome editors. The most consequential development is packaging of ribonucleoprotein rather than nucleic acid. NanoMEDIC uses chemically induced dimerization to package Cas9 protein with sgRNA, achieving efficient editing with only transient editor exposure [176]; Gag–Cas9 fusion vesicles deliver ribonucleoprotein to primary human immune cells without introducing DNA [177]. Neither has yet delivered a catalytically dead Cas9 fused to DNMT3A, TET1 or p300, but the packaging chemistry is agnostic to the effector domain, making this the most tractable route to locus-specific epigenetic reprogramming in vivo. Transient exposure is a particular advantage here, since durable editor expression risks progressive off-target methylation.

Clinical status. The furthest-advanced platform, combining producer-cell CD47 modification with electroporated cargo, is iExosomes carrying KRAS^G12D-targeting siRNA [175], which suppressed tumor growth and prolonged survival in pancreatic cancer models and has completed first-in-human evaluation (NCT03608631), where it was well tolerated without dose-limiting toxicity and showed target engagement with increased intratumoral CD8⁺ T-cell infiltration [178]. No engineered sEV carrying a defined epigenetic effector has entered clinical evaluation; this remains the field’s principal translational gap. Representative engineered platforms, their cargo classes, and reported effects are summarized in Table 2.

Table 2. Engineered sEV platforms, their epigenetic and gene-regulatory cargo, and reported effects. Evidence levels follow the hierarchy defined in Section 4.

Platform (Producer Cell)	Engineering Strategy	Cargo	Model	Reported Effect	Level	References
RVG–Lamp2b exosomes (dendritic cell)	Producer-cell surface display of neurotropic peptide	siRNA (GAPDH, BACE1)	Mouse brain, systemic dosing	BBB crossing; knockdown in neurons, microglia, oligodendrocytes	4	[141]
GE11–exosomes (HEK293)	Surface display of EGFR-binding peptide	let-7a miRNA	RAG2−/− breast cancer xenograft	Tumor growth suppression via RAS/HMGA2	4	[172]
iRGD–Lamp2b exosomes (immature DC)	Surface display of integrin-binding peptide	Doxorubicin (targeting-platform exemplar; cargo class extensible to ncRNA	Murine breast cancer	Targeted accumulation; reduced systemic toxicity	4	[173]
EXOtic devices (HEK293T)	Booster + RNA-packaging device + cytosolic delivery helper	mRNA (catalase); extensible to epigenetic effectors	Mouse brain, Parkinson’s model	Functional mRNA delivery; reduced neuroinflammation	4	[170]
EXPLOR (HEK293T)	Optogenetic CIBN–CRY2 luminal docking to CD9	Soluble proteins; applicable to DNMT/TET/dCas9	In vitro and in vivo	Intracellular delivery of intact protein cargo	2–3	[171]
ARMMs (HEK293T)	ARRDC1-driven plasma-membrane budding	p53 protein; Cas9/sgRNA	Recipient cell lines	Protein and editor delivery via a non-endosomal route	3	[27]
NanoMEDIC (HEK293T)	FRB/FKBP chemically induced dimerization + VSV-G	Cas9 RNP + sgRNA; extensible to dCas9 editors	DMD iPSC-myotubes; mice	Efficient exon skipping with transient editor exposure	4	[176]
Gag–Cas9 enveloped vesicles	Gag fusion + envelope pseudotyping	Cas9 RNP ± transgene	Primary human immune cells	Genome editing without DNA delivery	4	[177]
CP05-anchored exosomes (any origin)	Post-isolation CD63-binding anchor peptide	Surface-loaded peptides and oligonucleotides	mdx mice	Vesicle capture and targeted surface loading	3–4	[174]
iExosomes (MSC)	Producer-cell CD47 overexpression + electroporation	KRASG12D siRNA	PDAC models; Phase I NCT03608631	Preclinical tumor suppression and survival benefit; tolerated in humans with target engagement and increased CD8+ TILs	4	[175,178]
BMSC/HEK293T exosomes	Electroporation, co-incubation or endogenous loading	Tumor-suppressive miRNA (miR-22-3p, miR-34a, miR-126, miR-563)	Melanoma, NSCLC, OSCC	EMT inhibition; proliferation suppression	3	[179,180,181,182,183]
Glioma-derived exosomes	Electroporation	PDGFRβ siRNA acting via PI3K/Akt/EZH2	Glioma	Proliferation suppression through an EZH2-linked axis	3	[23]

10.1.3. Delivery of Tumor-Suppressive miRNAs and Epigenetic Reprogramming Agents

The delivery of tumor-suppressive miRNAs via engineered exosomes represents one of the most actively investigated therapeutic strategies in cancer epigenetics. Exosomes derived from bone marrow stem cells, which are encapsulated with miR-22-3p, impede the proliferation of human melanoma cells by specifically targeting galectin-1 (LGALS1) and subsequently inhibiting EMT [179]. NSCLC serum-derived exosomes loaded with miR-126 suppress tumor growth by reducing integrin α-6 expression, while lung cancer-derived exosomes carrying miR-563 inhibit cancer cell proliferation, invasion, and apoptosis with acceptable tissue-level safety profiles [180,181]. HEK293T-derived exosomes carrying miR-34a suppress oral squamous carcinoma cell proliferation by reducing SATB2 expression, and exosomes loaded with miR-317b-5b suppress lung cancer proliferation and induce apoptosis [182,183]. Exosomal siRNA delivery has also demonstrated strong therapeutic potential. Glioma cell-derived exosomes loaded with PDGFRβ siRNA suppress glioma cell proliferation via the PI3K/Akt/EZH2 pathway and display no appreciable toxicity with high targeting specificity [23]. NSCLC-derived exosomes loaded with PD-L1 siRNA via electrostatic interaction induce apoptosis with enhanced PD-L1 downregulation and low toxicity compared to controls, while BMSC-derived exosomes loaded with galectin-9 siRNA suppress pancreatic cancer proliferation by enhancing CTL activity and inhibiting Treg function through M1 macrophage polarization [184,185]. Of particular significance, MSC-derived exosomes carrying KrasG12D siRNA are currently being evaluated in a Phase I clinical trial (NCT03608631) in patients with metastatic pancreatic cancer harboring the KrasG12D mutation, with overall survival and progression-free survival as primary outcomes [175].

10.1.4. sEV-Based Cancer Immunotherapy

The immunomodulatory characteristics of exosomes have been strategically employed in the formulation of exosome-derived cancer vaccines and immune-enhancing therapeutic interventions. Exosomes originating from dendritic cells (DCs), especially those activated by interferon-gamma (IFN-γ-Dex), facilitate NK cell-mediated anti-tumor immune responses via the activation of the NKp30 signaling pathway. A Phase II clinical trial of IFN-γ-Dex in 22 patients with NSCLC demonstrated disease stabilization in 64% of participants and partial response to platinum-based chemotherapy in 36% following treatment cessation, validating the therapeutic potential of exosome-based immunotherapy [186]. Chimeric exosomal tumor vaccines constructed from antigen-presenting cell (APC)-tumor chimeric cells are currently under evaluation in a Phase I trial in bladder cancer patients (NCT05559177), measuring clinical response rate, overall survival, and safety. A Phase I trial of plant exosome-encapsulated curcumin (NCT01294072) in 35 colon cancer patients is further evaluating the safety, tolerability, and immune-modulatory effects of this natural epigenetic agent delivered via the exosomal platform. Curcumin is generally regarded as a deterrent of DNMT enzymatic processes and a facilitator of histone modifications, and the incorporation of this compound within exosomes is expected to ameliorate its historically poor bioavailability while augmenting its epigenetic impacts in neoplastic tissues.

10.2. Translational Implications Across Cancers

As of early 2026, more than 100 clinical trials evaluating EV-based diagnostics or therapeutic platforms are registered on ClinicalTrials.gov, reflecting the rapid maturation of this field toward clinical application (ClinicalTrials.gov; accessed April 2026). Among those with explicit epigenetic endpoints, the MSC-KrasG12D siRNA exosome trial (NCT03608631, Phase I) stands as the most advanced therapeutic application, with safety and preliminary efficacy data establishing proof of concept for nucleic acid delivery via vesicle platforms in pancreatic cancer [175]. Regulatory agencies, including the FDA and EMA, are increasingly engaging with EV-based companion diagnostic concepts; however, jurisdiction-specific guidance documents tailored to EV-derived biomarker assays remain absent, and the classification of EV products varies substantially across regulatory regions, from biologics in the United States to Advanced Therapy Medicinal Products (ATMPs) in Europe, creating significant barriers to multi-regional trial design and commercialization [187], and harmonization of pre-analytical variables (sample type, storage, isolation method) across trial sites remains an unmet need [188]. Registered trials with an explicit epigenetic or cargo-profiling component are listed in Table 3.

Table 3. Extensive clinical studies demonstrate exosome-mediated epigenetic approaches for oncological treatment, in conjunction with associated biomarkers and diagnostic practices. Recruitment statuses verified at ClinicalTrials.gov accessed (4 May 2026).

NCT Identifier	Source of Exosomes/Cargo Class	Cancer Type	Phase	Status (As Cited)	Epigenetic Target
Therapeutic Clinical Trials
NCT03608631	MSC-derived exosomes + KrasG12D siRNA	Metastatic pancreatic ductal adenocarcinoma (PDAC) cancer	I	Recruiting	KrasG12D oncogene silencing via RNA interference; siRNA-mediated post-transcriptional gene silencing with downstream epigenetic consequences for MAPK/ERK transcriptional programs
NCT01294072	Plant exosome-encapsulated curcumin (oral delivery)	Colon cancer	I	Unknown status	DNMT inhibition (curcumin is a DNMT enzymatic inhibitor); HAT activation; general epigenetic reprogramming toward tumor-suppressive states
NCT05559177	APC-tumor chimeric exosome cancer vaccine	Bladder cancer	I	Unknown status	Immune epigenetic reprogramming via antigen-presenting cell activation; T cell epigenetic rejuvenation
NCT06536712	MSC-derived exosome therapy	Rectal cancer	I	Not yet recruiting	Not specified; general MSC-derived EV cargo (likely anti-inflammatory ncRNAs and growth factors). Biomarker trials
Biomarker and Diagnostic Trials
NCT06388967	Case-control/exosomal miRNAs	Pancreatic ductal adenocarcinoma (PDAC)	Observational	Recruiting	Exosomal miRNA signatures, lncRNA panels, or cfDNA methylation profiles as specified by the respective trial protocols
NCT06342427	Case-control/exosomal miRNAs	Gastric cancer (Stomach Cancer)	Observational	Completed
NCT07553754	Case-control/plasma exosomal RNAs	Prostate cancer	Observational	Recruiting
NCT05705583	Case-control/circulating exosomes as companion diagnostic biomarker	Renal cell carcinoma	Observational	Unknown status
NCT04939324	Molecular profiling pulmonary and peripheral vein exosomes	Early-stage non-small-cell lung cancer (NSCLC)	Not Applicable	Unknown status
NCT01840306	Exosomal and RNAs and proteins	Breast cancer	Observational	Completed
NCT07558447	Analysis of blood extracellular vesicles (lipids and microRNAs)	Cancer patients receiving chemotherapy	Not Applicable	Recruiting
NCT06991452	Body fluids exosomes	Colorectal cancer	Observational	Not yet recruiting

10.2.1. sEV Biomarkers for Cancer Diagnosis and Prognosis

The diagnostic and prognostic utility of exosomal content in cancer, particularly exosomal non-coding RNAs, is one of the most rapidly advancing areas of translational oncology [189]. Because exosomes are shed into all body fluids, including blood, urine, saliva, and bile, and because their molecular cargo reflects the genomic and epigenomic state of their cell of origin, they constitute an accessible, non-invasive source of cancer biomarkers that can be profiled through liquid biopsy approaches [190]. In clinical validation studies, plasma exosomal miR-320d, miR-4479, and miR-6763-5p are significantly decreased in patients with epithelial cancers compared to healthy controls, while serum exosomal miR-122-5p, let-7d-5p, and miR-425-5p reliably distinguish HCC patients from healthy donors [23]. The expression levels of exosomal miR-21, miR-155, miR-182, and miR-373 in the serum of breast cancer patients are markedly elevated compared to healthy controls, and urinary exosomal lncRNAs distinguish bladder cancer patients from controls with high sensitivity by RNA sequencing [23]. The biomarker utility of EV-associated miRNA signatures extends beyond saliva to additional non-invasive biofluids. In the ophthalmological setting, EV-miRNA profiles in tears have been identified as sensitive, non-invasive biomarkers for corneal disorders, a precedent that demonstrates the feasibility of tissue-of-origin-specific EV-miRNA profiling across diverse biological fluids and supports the broader application of this approach to cancer liquid biopsy workflows [191]. For prognostication, low serum exosomal miR-134 predicts shorter overall and relapse-free survival in gastric cancer patients, while high levels of circulating exosomal LINC00963 predict poor prognosis in lung cancer [23]. In cholangiocarcinoma, high bile exosomal miR-200a-3p and miR-200c-3p and serum exosomal miR-200c-3p levels predict poor prognosis, and low levels of circulating exosomal lncRNA-GC1 predict a favorable response to adjuvant chemotherapy in gastric cancer patients [23]. The collective findings elucidate a foundational structure for panels of exosomal epigenetic biomarkers that have the potential to inform clinical decision-making across a diverse array of cancer types. A particularly innovative advance is the nanoscale epigenetic profiling of individual exosomes derived from colorectal cancer cell lines using photo-induced force microscopy (PiFM), which has successfully distinguished CpG island methylator phenotype (CIMP)-high from CIMP-negative exosomes at the single-vesicle level [74]. This technique reveals heterogeneity among individual exosomes, suggesting the existence of distinct subpopulations with unique epigenetic profiles, a finding that may require single-vesicle resolution for clinically meaningful diagnostic precision [74].

10.2.2. sEV miRNAs as Predictors of Therapeutic Response

Together with their diagnostic and prognostic value, the examination of exosomal miRNAs is slowly being recognized as a promising predictive biomarker for evaluating therapeutic response. In a research investigation involving patients with triple-negative breast cancer (TNBC) who were subjected to neoadjuvant chemotherapy, microarray-based miRNA profiling discerned 16 distinctively expressed exosomal miRNAs differentiating between individuals achieving pathological complete response (pCR) and those not achieving pCR [192]. A combined signature of four miRNAs (miR-4448, miR-2392, miR-2467-3p, and miR-4800-3p) discriminated between pCR and non-pCR patients with an area under the curve (AUC) of 0.7652, establishing exosomal miRNA panels as viable tools for predicting chemotherapy sensitivity in aggressive breast cancer subtypes [192]. Exosomal miR-7-5p enhances the therapeutic efficacy of everolimus by inhibiting the MNK/eIF4E signaling axis, suggesting that high levels of this circulating exosomal miRNA could predict a favorable response to mTOR inhibitor-based regimens [23]. In NSCLC, plasma exosomal lnc-SNAPC5-3:4 is significantly upregulated when anlotinib treatment is effective, positioning it as a pharmacodynamic biomarker for anti-angiogenic therapy response [23]. These empirical findings suggest the potential for real-time epigenetic surveillance of therapeutic responses through the utilization of minimally invasive liquid biopsy techniques. The corresponding analytical workflow is outlined in Figure 4.

Figure 4. Liquid biopsy biomarker pipeline model for extracellular vesicle (EV)-based precision oncology. Panel (a): Schematic of minimally invasive collection sites (blood, urine, saliva, pleural effusion, ascites, cerebrospinal fluid) used to obtain patient-derived EVs; emphasizes routine, low-burden sampling across cancers stages. Panel (b): Overview of common isolation strategies (differential ultracentrifugation, size-exclusion chromatography, precipitation polymers, immunoaffinity capture) and the target EV size range (≈30–150 nm), highlighting trade-offs between yield and purity. Panel (c): Workflow for molecular characterization of EV cargo, including small RNA (miRNA) and lncRNA profiling, cell-free DNA and methylation assays, proteomics (DDA/DIA, targeted MS), and metabolomics/lipidomics; illustrates downstream assays used to generate high-dimensional biomarker data. Panel (d): Data-processing and machine-learning pipeline for quality control, normalization, feature extraction, multi-omics integration, predictive modeling, and validation; depicts how computational models translate EV signatures into diagnostic, prognostic, and therapeutic predictions. Panel (e): Illustration of translational endpoints showing applications in diagnosis (early detection, tumor classification), prognosis (risk stratification, survival modeling), treatment prediction (therapy selection, resistance mechanisms, companion biomarkers), and disease monitoring (minimal residual disease, recurrence surveillance, longitudinal response tracking).

11. Challenges and Limitations

Despite the remarkable mechanistic progress described in preceding sections, the translation of sEV-mediated epigenetic cancer therapies into clinical practice faces a formidable constellation of biological, technical, and regulatory challenges that must be candidly acknowledged. First, as described in the search strategy, this review was not prospectively registered and does not include a PRISMA flow diagram; eligibility decisions were made by the authors based on mechanistic relevance, introducing the possibility of selection bias toward studies that support the conceptual framework presented. The sEV epigenetics literature is dominated by positive mechanistic findings, and publication bias toward reported effects is likely. The rapidly evolving nature of this field means that studies published after April 2026 are not represented. Readers should interpret the mechanistic synthesis herein as reflecting the current weight of published evidence rather than a stable consensus.

Isolation methodology remains the most fundamental barrier to reproducible discovery. Differential ultracentrifugation, size-exclusion chromatography, precipitation-based kits, and immunoaffinity capture each enrich distinct sEV subpopulations with different purity profiles, size distributions, and cargo compositions, and head-to-head comparisons reveal substantial method-dependent variability in yield and molecular content [193]. The MISEV2023 consensus specifies that primary reports should disclose particle counts, protein content, at least one transmembrane marker (CD9, CD63, CD81), one cytosolic marker (TSG101, ALIX, syntenin-1), and appropriate negative markers (calnexin, GM130) to exclude intracellular contaminants [3,20]. Compliance with these criteria across the exosome-epigenetics literature remains uneven, particularly in studies that infer functional cargo transfer from bulk EV preparations without orthogonal single-vesicle validation. Future mechanistic claims should be benchmarked against MISEV2023 reporting, and meta-analytic syntheses should pre-specify MISEV adherence as a minimum quality criterion [3,20].

Beyond methodological standardization, a persistent and under-reported caveat across the exosome-epigenetics literature is the disconnect between the EV doses used in mechanistic in vitro studies and the physiological exosome concentrations measured in patient biofluids. Multiple primary reports of miRNA-mediated phenotype reprogramming use vesicle preparations at 10⁹–10¹¹ particles/mL applied to 10⁵–10⁶ recipient cells [168], concentrations that exceed plasma levels by two to four orders of magnitude, and stoichiometric analyses indicate that most circulating exosomal miRNAs are present at less than one copy per vesicle [168], and heterogeneity in small RNA content across individual vesicles is substantial [194]. The functional epigenetic effects reported in recipient cells, DNMT suppression, H3K27me3 redistribution, and chromatin remodeling require engagement of the RISC or chromatin-modifying machinery at levels that demand either highly efficient uptake of many vesicles or a proximity-dependent delivery mechanism that has not been demonstrated for most cargo-mechanism pairs. Future mechanistic claims about sEV-driven epigenetic reprogramming should be benchmarked against absolute miRNA copy quantification per vesicle and EV doses calibrated to physiologically attainable concentrations in the recipient compartment. We call on the field to adopt stoichiometric reporting as a standard requirement in primary mechanistic publications.

Population-level bulk EV preparations mask substantial cargo heterogeneity at the single-vesicle level [194]. The demonstration by PiFM nanoscopy that CRC-derived exosomes segregate into CIMP-high and CIMP-negative epigenetic subpopulations [74] illustrates that clinically relevant epigenetic information may reside in minority vesicle subpopulations that are diluted or obscured in bulk assays. This heterogeneity has direct consequences for both mechanistic interpretation and biomarker development: a cargo molecule detected at low abundance in a bulk preparation may be present at high copy number in a specific subpopulation with potent recipient-cell activity or conversely may represent background contamination. Advances in single-vesicle analysis nanoparticle tracking with fluorescence, nanoscale flow cytometry, and single-vesicle proteomics are urgently needed to resolve this ambiguity.

The vast majority of exosomal miRNA and lncRNA biomarker candidates reported to date have been validated only at the analytical level, demonstrating that an assay consistently measures the intended target in small, single-institution discovery cohorts. Clinical validity, defined as prospectively demonstrated prediction of a clinical outcome with a pre-specified decision threshold, has been achieved for very few exosomal epigenetic markers [193]. Furthermore, variability in EV-miRNA quantification arising from differences in isolation method, RNA extraction protocol, normalization strategy, and detection platform introduces significant inter-study variability that undermines biomarker reproducibility [191,193]. Regulatory agencies require demonstration of both analytical and clinical validity before approving companion diagnostic biomarkers; this gap represents the primary bottleneck for clinical translation of liquid biopsy applications.

Current biological yields from cell culture are insufficient to support the multi-dose systemic administration regimens envisioned for therapeutic sEV applications, and ultracentrifugation, the reference isolation method, is not readily scalable to GMP grade [22,195]. Emerging hollow-fiber bioreactor platforms and tangential flow filtration approaches offer potential solutions but have not yet been validated in GMP settings with demonstrated lot-to-lot consistency. Without scalable, validated production processes, any sEV-based epigenetic therapeutic candidate will stall in early Phase I regardless of mechanistic proof of concept [195].

The absence of unified global regulatory frameworks creates a substantial translational barrier. EV therapeutics are classified as biologics subject to Investigational New Drug (IND) requirements in the United States, as Advanced Therapy Medicinal Products (ATMPs) in Europe, and under source-based categorization in Asian jurisdictions, with no single international standard governing engineered EVs containing epigenetic cargo [187]. This jurisdictional divergence substantially increases development costs, complicates multi-regional clinical trial design, and necessitates parallel regulatory strategies that delay commercialization. Addressing this requires international consortium-driven harmonization analogous to the MISEV initiative but extended to analytical and regulatory domains.

As noted in Section 8, the role of sEVs in direct histone modification transfer remains specifically controversial [11]. Whether intact multi-subunit chromatin-modifying complexes can survive the intraluminal EV milieu and be delivered to recipient cell nuclei in catalytically competent form has not been rigorously demonstrated. Single-subunit enzymes (DNMT3A, EZH2 monomer) are mechanistically more plausible candidates than intact PRC2 or NuRD complexes, and the field should prioritize proximity ligation assays, co-immunoprecipitation within isolated EVs, and single-vesicle proteomics to establish stoichiometry and activity status of transferred chromatin regulators before histone complex transfer is stated as an established mechanism [76].

Finally, the inherent pleiotropy of exosomal miRNAs, which individual species typically regulate hundreds of target mRNAs simultaneously, raises concerns about off-target epigenetic effects and unintended phenotypic consequences when delivered systemically [23]. Systematic off-target profiling using RNA sequencing and ATAC-seq in tissues of accumulation (liver, spleen, lung) should be considered a standard component of preclinical IND-enabling packages for any exosome-miRNA therapeutic candidate [161]. Addressing these challenges through rigorous experimental design, harmonized reporting standards, and transparent pre-clinical validation frameworks will be essential to realizing the translational promise of sEV-mediated epigenetic cancer medicine. Table 4 pairs each barrier discussed above with its mitigation strategy and with the metric by which mitigation would be judged successful.

Table 4. Challenges to sEV-mediated epigenetic cancer therapy and corresponding mitigation strategies.

Domain	Challenge	Mitigation Strategy	Success Metric	References
Vesicle identity	Preparations designated “sEV” contain exosomes, ectosomes, ARMMs and migrasome-derived vesicles with distinct sorting machinery	Operational MISEV2023 nomenclature; route-discriminating markers (LAMP1 vs. BSG/SLC3A2) alongside tetraspanins	Full MISEV2023 panel plus subtype markers reported	[3,20,25,26,28]
Cargo attribution	Exomeres, supermeres, AGO2 complexes, and HDL co-isolate and carry ncRNA	Density-gradient or asymmetric flow field-flow fractionation; nuclease and protease protection assays	Cargo shown in vesicle-containing fractions and nuclease-protected	[14,30,31,32,33,178]
Isolation reproducibility	Method-dependent variation in yield, purity, and cargo	Pre-specified orthogonal isolation; harmonized pre-analytics; EV-TRACK deposition	EV-METRIC score reported; inter-laboratory concordance	[3,35,193]
Cargo stoichiometry	Fewer than one miRNA copy per vesicle versus supraphysiological in vitro doses	Absolute copy quantification by digital PCR; dosing calibrated to attainable concentrations	Copies per vesicle reported; dose within twofold of plasma-attainable	[168,194]
Population heterogeneity	Informative cargo may reside in minority subpopulations masked in bulk assays	Single-vesicle analysis: PiFM nanoscopy, nanoscale flow cytometry, single-vesicle proteomics	Subpopulation-resolved cargo distribution reported	[74,194]
Enzyme functionality	Detection of DNMT/EZH2 does not establish catalytically competent delivery; PRC2 and NuRD assembly unproven	Proximity ligation and co-immunoprecipitation within vesicles; activity assays in recipient cells	Catalytic activity, not detection, demonstrated	[76,196]
Therapeutic loading	Native loading is sub-stoichiometric and variable	Active loading (electroporation, sonication, saponin) or producer-cell engineering (EXOtic, EXPLOR, ARMMs)	Defined copies per vesicle with lot-to-lot consistency	[27,158,159,170,171]
Pharmacokinetics	Rapid hepatic and splenic clearance; low tumor accumulation	Targeting ligands (RVG, GE11, iRGD, CP05); CD47 overexpression	Tumor-to-liver ratio and circulating half-life reported	[141,160,169,172,173,174,175]
Off-target effects	miRNA pleiotropy produces unintended chromatin effects in accumulation organs	RNA-seq and ATAC-seq in liver, spleen, and lung; substitution of programmable editors for miRNA mimics	Off-target chromatin profiling in IND-enabling package	[161,176,177]
Immunogenicity	Allogeneic and MSC preparations may trigger complement activation and accelerated clearance	Autologous or well-characterized immortalized producer lines; immunological release testing	Anti-drug antibody and complement assays in repeat-dose toxicology	[162]
Manufacturing	Ultracentrifugation is not GMP-scalable; lot-to-lot variability	Hollow-fiber bioreactors and tangential flow filtration; defined potency assay	Lot-to-lot consistency demonstrated at clinical scale	[22,195]
Regulatory	Divergent classification (FDA IND, EMA ATMP, source-based frameworks in Asia)	Early parallel scientific advice; consortium-led harmonization analogous to MISEV	Harmonized characterization package accepted across regions	[187]

12. Discussion and Future Directions

The evidence synthesized in this review establishes sEV-mediated epigenetic communication as a mechanistically grounded process in Section 4, Section 5, Section 6, Section 7, Section 8 and Section 9, which have systematically calibrated mechanistic claims against the four-level evidence hierarchy; this discussion directly addresses the three foundational controversies identified in the Introduction.

First, can intact, catalytically competent multi-subunit chromatin-modifying complexes survive the intraluminal sEV milieu and be functionally delivered to recipient cells? Single-subunit enzymes (DNMT3A, EZH2 monomer) are mechanistically plausible sEV cargo because their functional delivery requires only protein folding fidelity. By contrast, PRC2 (EZH2/SUZ12/EED/RBBP7; 4+ subunits) and NuRD (6–8 subunits) require complete assembly for catalytic activity [76]; co-packaging of a complete assembled complex in a single vesicle and its maintenance of activity in the recipient cytoplasm have not been demonstrated. Future studies should prioritize proximity ligation assays and single-vesicle mass spectrometry specifically designed to assess the stoichiometry and enzymatic activity status of transferred chromatin regulators, not just protein detection, in bulk EV fractions [3,196].

Second, do the sEV miRNA concentrations achievable in vivo produce the phenotypic reprogramming reported in vitro? Stoichiometric analyses indicate that most circulating sEV miRNAs are present at less than one copy per vesicle [168], yet multiple mechanistic studies use vesicle preparations at 10⁹–10¹¹ particles/mL applied to 10⁵–10⁶ recipient cells, concentrations that exceed plasma levels by two to four orders of magnitude. Until mechanistic claims are benchmarked against absolute miRNA copy quantification per vesicle and against EV doses calibrated to physiologically attainable concentrations, their translational relevance remains provisional. We call on the field to adopt stoichiometric reporting as a standard requirement in primary mechanistic publications.

Third, can engineered sEVs carrying CRISPR-dCas9 epigenome editing complexes achieve sufficient tumor biodistribution and on-target chromatin reprogramming in vivo? Proof-of-concept studies demonstrating sEV delivery of catalytically inactive Cas9 fused to DNMT3A or TET1 effector domains should be prioritized as the convergence point of the CRISPR and sEV delivery fields [159,160]. The engineering of hybrid exosome-liposome nanoparticles to improve drug-loading capacity and circulation half-life addresses two persistent barriers to clinical translation and merits accelerated preclinical evaluation. The m6A epitranscriptome represents a critically underexplored dimension: METTL3-containing sEVs and m6A-modified circRNAs (e.g., circNSUN2) carry functional epigenetic instructions extending into chromatin remodeling in recipient cells [67,86], and a systematic characterization of how m6A modification status affects cargo sorting, uptake efficiency, and epigenetic consequences is a high-priority experimental agenda.

On the translational front, AI-driven multi-omics integration of sEV miRNA profiles, cfDNA methylation signatures, and proteomics from liquid biopsies holds genuine promise for early cancer detection and real-time treatment monitoring [74,189], but only if the pre-analytical standardization problems described in Section 11 are resolved first. Harmonization of isolation methods, extraction protocols, normalization strategies, and detection platforms across clinical trial sites is a prerequisite, not a downstream refinement for any clinically actionable multi-analyte panel. The regulatory landscape is evolving but has not kept pace: neither the FDA nor the EMA has issued guidance documents specific to EV-derived companion diagnostics, and EV therapeutics are classified differently across jurisdictions (biologics/IND in the USA; ATMPs in the EU; source-based categories in Asian jurisdictions), creating significant multi-regional trial design barriers [187]. Addressing this requires international consortium-driven harmonization analogous to the MISEV initiative but extended to analytical and regulatory domains. The concrete validation priorities that will determine whether sEV-mediated epigenetic cancer therapeutics can be translated into the clinic are as follows: (i) dose-relevant functional assays using physiologically calibrated EV concentrations; (ii) single-vesicle cargo characterization by proximity ligation assay and single-vesicle mass spectrometry; (iii) orthogonal EV purification combining immunoaffinity capture and size-exclusion chromatography; (iv) recipient-cell chromatin readouts (ATAC-seq, ChIP-seq, WGBS) paired directly with phenotypic assays; and (v) in vivo biodistribution studies with co-registered epigenetic endpoint measurements in tumor, immune, and stromal compartments. These experiments are within near-term technical reach. Achieving them will determine whether sEV-mediated epigenetic reprogramming can be harnessed to deliver a fundamentally new class of precision oncology agents, one that targets not just individual driver mutations but the intercellular epigenetic ecosystem that sustains malignant progression.

Additionally, a key conceptual caveat in pan-cancer interpretation is that exosomal cargo selection, recipient cell uptake, stromal and immune context, and chromatin state dependency are all highly tumor-type specific. The cross-cancer mechanistic framework proposed in this review, wherein exosome-mediated transfer of miRNAs, lncRNAs, DNMTs, and metabolites broadly reprograms recipient-cell epigenomes, should be understood as a conceptual scaffold that identifies shared biological principles, not as evidence that identical mechanisms operate across all tumor types. Future mechanistic studies should be stratified by cancer type, stromal composition, and immune context to delineate which elements of this framework are conserved and which are tumor specific.

13. Conclusions

This review establishes a coherent mechanistic framework: tumor-derived sEVs (operationally termed exosomes, where biogenesis cannot be confirmed per MISEV2023 guidelines) function not as passive carriers of molecular debris but as active orchestrators of epigenetic reprogramming across the TME. Through the coordinated transfer of miRNAs, lncRNAs, circRNAs, DNA methyltransferases, histone-modifying enzymes, oncometabolites, and m6A-modified RNAs, sEVs reprogram recipient-cell chromatin states, silence tumor suppressors, activate drug efflux programs, exhaust cytotoxic immune cells, and propagate therapy resistance across heterogeneous tumor populations, all without altering the recipient genome. Systematic application of the four-level evidence hierarchy throughout this review reveals that the field has moved well beyond cargo detection (Level 1–2) for several cargo-mechanism pairs, with Level 3–4 evidence now available for exosomal miR-29/DNMT3B suppression, HOTAIR/PRC2/H3K27me3 cisplatin resistance, exosomal DNMT1 transfer in ovarian cancer, and PD-L1/STAT3/DNMT3A T-cell exhaustion. These represent the most clinically actionable epigenetic mechanisms for therapeutic targeting.

Three foundational mechanistic questions must be resolved to advance clinical translation. First, can intact multi-subunit chromatin-modifying complexes survive intraluminal sEV conditions and be delivered in a catalytically competent form to recipient cells, or does epigenetic reprogramming occur primarily through ncRNA-templated nucleation of recipient-cell chromatin machinery? Current evidence supports only Level 1–2 for multi-subunit complex transfer; this mechanistic gap must be closed by single-vesicle proteomics and proximity ligation assays before PRC2 or NuRD complex transfer can be stated as established. Second, do the sEV miRNA concentrations achievable in vivo stoichiometric analyses indicate that sub-copy-per-vesicle for most miRNAs produces the phenotypic reprogramming reported in vitro at supraphysiological doses? Addressing this requires the stoichiometric reporting standards and physiologically calibrated dose experiments outlined in Section 11. Third, can engineered sEVs delivering CRISPR-dCas9 epigenome-editing complexes achieve sufficient tumor biodistribution and on-target chromatin reprogramming in vivo? Proof-of-concept studies with DNMT3A-fused or TET1-fused dCas9 delivered via sEVs represent the highest-priority translational agenda for the coming decade.

Until these questions are answered, sEV-mediated epigenetic cancer therapeutics should be framed as a mechanistically compelling, preclinically grounded translational hypothesis, one with early clinical signals and a clear experimental roadmap rather than as an emerging therapeutic class approaching clinical readiness. The convergence of single-vesicle analytical platforms, AI-driven multi-omics integration, and precision EV engineering places these validation experiments within near-term technical reach, creating an unprecedented opportunity to deliver a new class of precision oncology agents that target not just individual driver mutations but the intercellular epigenetic ecosystem that sustains malignant progression.