Mechanisms of tumor-derived extracellular vesicle-mediated immunometabolic reprogramming and immunotherapeutic resistance

Background

Tumor-derived extracellular vesicles (tEVs) are emerging as pivotal mediators of intercellular communication within the tumor microenvironment (TME). Beyond carrying oncogenic cargo, tEVs dynamically reprogram immune and metabolic networks that shape tumor progression and therapeutic response.

Main body

In this review, we delineate how tEV-mediated immunometabolic rewiring orchestrates resistance to immunotherapy. We summarize recent findings that demonstrate how tEVs remodel glucose, lipid, and amino acid metabolism in immune cells, which in turn generates immunosuppressive microenvironments. These tEV-driven changes give rise to “metabolic checkpoints,” a newly recognized layer of immune regulation parallel to classical immune checkpoints. We further discuss how these metabolic alterations contribute to both primary and acquired immunotherapeutic resistance. Finally, we outline potential therapeutic strategies—including targeting tEV biogenesis, blocking tEV–immune interactions, and modulating metabolic checkpoints—and highlight how artificial intelligence (AI)-integrated tEV profiling and liquid biopsy may enable patient stratification and precision intervention.

Conclusions

Understanding how tEV-mediated metabolic regulation shapes immune escape provides conceptual and translational opportunities. Integrating AI-driven analytics with tEV-based diagnostics may transform the prediction and reversal of immunotherapeutic resistance.

Highlights

∙ Tumor-derived extracellular vesicles (tEVs) play a critical role in orchestrating immune metabolic reprogramming within the tumor microenvironment (TME).

∙ We introduce the concept of “metabolic checkpoints”, offering novel mechanistic insights into tumor immune evasion and resistance to immunotherapy.

∙ Targeting metabolic checkpoints in combination with immune checkpoint inhibitors (ICIs) offers a promising strategy to overcome treatment resistance.

∙ Liquid biopsy platforms utilizing tEV-derived biomarkers, combined with artificial intelligence (AI), have transformative potential for real-time monitoring and personalized cancer immunotherapy.

∙ The integration of immunometabolism, extracellular vesicle biology, and computational analytics paves the way for a new frontier in precision oncology.

Innovation and highlights

Conceptual innovation

We introduce the concept of metabolic checkpoints as actionable targets that link extracellular vesicle biology with immune metabolism. This framework establishes a novel combinatorial axis for enhancing antitumor immunity and overcoming immune dysfunction.

Clinical translation potential

We systematically delineate how targeting tEV-mediated metabolic checkpoints can complement and potentiate immune checkpoint blockade, providing a rational strategy to circumvent therapeutic resistance and improve clinical outcomes.

Technological advancement

The integration of liquid biopsy technologies with AI enables high-throughput and longitudinal monitoring of tEV-derived signatures, thereby supporting real-time, personalized immunotherapy decision-making.

Interdisciplinary insight

By bridging tumor immunology, metabolic reprogramming, extracellular vesicle biology, and computational analytics, this review lays a conceptual and technological foundation for next-generation precision oncology and interdisciplinary collaboration.

Introduction

Immunotherapy has fundamentally transformed cancer treatment by reactivating antitumor immunity, offering durable responses in subsets of patients with advanced malignancies [1]. However, despite these promising results, the majority of patients ultimately develop therapeutic resistance, leading to disease progression and treatment failure [1]. A central mechanism underlying this resistance is the tumor microenvironment (TME), where tumor-immune cell interactions drive immune exhaustion, metabolic dysregulation, and immune evasion [2, 3].

Among the various mediators of tumor-immune crosstalk, tumor-derived extracellular vesicles (tEVs) have emerged as pivotal messengers orchestrating intercellular communication within the TME [4]. These nanoscale vesicles are released by virtually all tumor cells and carry a diverse cargo of nucleic acids, proteins, lipids, and metabolites, all of which reflect the state of their parental cells [5–8]. Acting as “signal shuttles,” tEVs deliver oncogenic molecules to recipient immune or stromal cells, thereby modulating antigen presentation, immune checkpoint expression, cytokine secretion, and, ultimately, the outcome of immune responses [9, 10].

Recent studies have shown that tEVs profoundly impact immune cell metabolism, reprogramming glucose, lipid, and amino acid utilization in macrophages, T cells, and dendritic cells [11–16]. By modulating metabolic enzymes or delivering regulatory microRNAs (miRNAs), tEVs influence key metabolic pathways such as oxidative phosphorylation, glycolysis, and fatty acid oxidation [13, 16, 17]. These metabolic perturbations collectively create an immunosuppressive microenvironment characterized by glucose depletion and lipid accumulation [13, 16].

These findings have led to the concept of “metabolic checkpoints”—metabolic bottlenecks that restrict immune activation either independently or in synergy with classical immune checkpoints, such as programmed cell death protein 1 (PD-1)/programmed death ligand 1 (PD-L1) and cytotoxic T-lymphocyte-associated protein 4 (CTLA-4) [18]. Unlike classical immune checkpoints that function through ligand-receptor interactions, metabolic checkpoints operate via metabolite accumulation or enzyme-driven suppression of immune signaling [18]. Recognizing tEVs as dynamic regulators of these metabolic checkpoints offers new insights into the mechanisms of immune resistance.

Understanding the role of tEV-mediated immunometabolic reprogramming in driving immunotherapeutic resistance is therefore crucial. This review focuses on: (1) the mechanisms by which tEVs reprogram immune metabolism; (2) their contribution to immunotherapeutic resistance; (3) emerging therapeutic strategies targeting tEVs or metabolic checkpoints; and (4) the future integration of artificial intelligence (AI)-guided tEV-omics for precision oncology.

Tumor-Derived extracellular vesicles

Biogenesis of tEVs

tEVs, secreted by tumor cells, serve as crucial molecular messengers mediating intercellular communication within the TME [19]. Based on their biosynthetic origin, release mechanisms, and size, tEVs are typically classified into three subtypes: 30–50 nm exosomes, 100–1000 nm microvesicles, and 1000–5000 nm apoptotic bodies [19].

The biogenesis of exosomes begins with endocytosis, which forms early endosomes. Subsequent invagination of the endosomal membrane, regulated by the endosomal sorting complex required for transport (ESCRT) machinery or neutral sphingomyelinase 2 (nSMase2), leads to the generation of intraluminal vesicles (ILVs) within multivesicular bodies (MVBs) [20]. Mature MVBs then either fuse with the plasma membrane to secrete exosomes or undergo degradation via lysosomal or autophagic pathways [21, 22]. Rab GTPases, particularly Rab27a/b, Rab11, and Rab35, play pivotal roles in coordinating vesicle trafficking and secretion [23].

In parallel, microvesicles bud directly from the plasma membrane through cytoskeletal remodeling and phospholipid reorganization [24]. Unlike exosomes and microvesicles, which are actively secreted by viable cells, apoptotic bodies are generated during the programmed cell death (apoptosis) process as the cell undergoes fragmentation [25]. Once released, tEVs interact with recipient cells via endocytosis, membrane fusion, or receptor-ligand interactions. These multifaceted interactions enable tEVs to deliver oncogenic cargo and modulate the behavior of target cells. Figure 1 schematically illustrates the biogenetic pathways of tEVs.

Fig. 1.

Biogenesis and molecular composition of tEVs. Tumor cells generate EVs through endocytic and plasma membrane–derived pathways. Early endosomes formed by endocytosis mature into multivesicular bodies (MVBs), where intraluminal vesicles (ILVs) arise via endosomal sorting complex required for transport (ESCRT)-dependent or neutral sphingomyelinase 2 (nSMase2)–mediated membrane invagination. MVBs either fuse with the plasma membrane to release exosomes or undergo lysosomal/autophagosomal degradation, a process regulated by Rab GTPases. In parallel, microvesicles bud directly from the plasma membrane. Once secreted, tEVs interact with recipient cells through endocytosis, membrane fusion, or receptor–ligand binding. Their cargoes include diverse nucleic acids, proteins, and lipids that mediate intercellular communication

Components of tEVs

tEVs carry a complex repertoire of nucleic acids, proteins, and lipids derived from their parental tumor cells (Fig. 1). The molecular composition of tEVs varies according to tumor type, microenvironmental stress, and therapeutic exposure [26].

Nucleic acids identified in tEVs include messenger RNAs (mRNAs), microRNAs (miRNAs), long non-coding RNAs (lncRNAs), circular RNAs (circRNAs), and DNA fragments [5]. Among these, miRNAs are particularly significant for their post-transcriptional regulatory roles in oncogenesis and immune modulation [5]. The lipid bilayer of tEVs is enriched with ceramides, cholesterol, sphingomyelin, phosphatidylserine, and saturated fatty acids [6]. Notably, phosphatidylserine exposure on the outer membrane serves as a marker for apoptotic and tumor-derived EVs and has been explored as a potential biomarker for cancer detection [7].

EVs can also carry various types of proteins, whose properties depend on the function of the associated cell type [8]. Table 1 summarizes the specific types of proteins contained in EVs and provides examples [7]. Functionally, these proteins regulate immune signaling, angiogenesis, and stromal remodeling [6]. Importantly, exosomal PD-L1 and oncogenic receptors can dampen cytotoxic T-cell function and fuel tumor progression [27, 28].

Table 1.

EV-carried protein cargoes

Proteins	Examples		Ref.
Tetraspanins	CD9, CD63, CD81	[29]
MVB-related proteins	TSG101, ALIX	[30]
HSP	HSP90 and HSP70	[7]
Growth factors and cytokines	TNF-α, VEGF, EGF, TNFR, TGF-β	[6]
Cell adhesion-related proteins	Integrins, ICAM-1	[29, 31]
Antigen presentation-related proteins	MHC class I/II	[29]
Signaling proteins	GTPases, RhoA, Ras-related proteins	[7]
Cytoskeleton components	Actins, Cofilin-1, Moesin, Myosin, Tubulins, Vimentin	[6]
Transcription and protein synthesis related proteins	Histone, Ribosomal proteins, Ubiquitin	[7]
Metabolic enzymes	FASN, PGK, ATP	[7]
Death receptors	FasL, TRAIL	[32, 33]
Iron transport proteins	Transferrin receptor	[29]

Multivesicular body (MVB), Heat shock proteins (HSP), Tumor susceptibility gene 101 (TSG101), ALG-2-interacting protein X (ALIX), Tumor necrosis factor-α (TNF-α), Vascular endothelial growth factor (VEGF), Epidermal growth factor (EGF), Tumor necrosis factor receptor (TNFR), Transforming growth factor-β (TGF-β), Intercellular adhesion molecule 1 (ICAM-1), Major histocompatibility complex (MHC), Ras homologous gene family member A (RhoA), Fatty acid synthase (FASN), Phosphoglycerate kinase (PGK), Adenosine triphosphate (ATP), Fas ligand (FasL), TNF-associated apoptosis-inducing ligand (TRAIL)

tEV-Mediated immunotherapeutic resistance through non-metabolic pathways

tEVs directly suppress immune surveillance through the surface display of inhibitory ligands, secretion of immunomodulatory cytokines, and modulation of receptor signaling. These mechanisms converge to blunt cytotoxic activity and promote resistance to immunotherapeutic agents (Fig. 2).

Fig. 2.

Immunosuppressive mechanisms mediated by tEVs. tEVs employ diverse mechanisms to orchestrate immunosuppression. They directly inhibit immune cell activity by delivering specific molecules: Fas-Fas ligand (Fas-FasL) suppresses CD4⁺ T cell activation; programmed cell death protein 1-programmed death ligand 1 (PD-1-PD-L1) inhibits CD8⁺ T cell proliferation; and natural killer cell group 2D (NKG2D) ligands induce receptor internalization to impair natural killer (NK) and T cell cytotoxicity. Furthermore, tEV-associated TNF-associated apoptosis-inducing ligand (TRAIL) activates nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB) via TRAIL-receptor2 (TRAIL-R2), promoting poliovirus receptor (PVR+) macrophages which in turn suppress NK cells through T cell immunoglobulin and TIM domains (TIGIT) interaction. Simultaneously, tEVs induce immunosuppressive cells. Surface heat shock protein 72 (HSP72) engages Toll-like receptor 2 (TLR2) on myeloid-derived suppressor cells (MDSCs), triggering the myeloid differentiation factor 88-interleukin-6-signal transducer and activator of transcription 3 (MyD88-IL-6-STAT3) pathway to enhance their function. Shipped miRNAs promote alternative (M2) macrophage polarization: miR-25-3p targets phosphatase and tensin homolog (PTEN) to activate phosphatidylinositol3-kinase/protein kinase B (PI3K/AKT), while miR-29a-3p, miR-222-3p, and miR-21-3p suppress suppressor of cytokine signaling (SOCS)1, SOCS3, and SOCS4/5 respectively to activate STAT6/3 signaling. Finally, tEV-delivered transforming growth factor-β (TGF-β) and IL-10 activate small mothers against decapentaplegic (SMAD)2/3 and STAT3 to drive regulatory T cell (Treg) differentiation and suppressive capacity

Expression of immunosuppressive molecules by tEVs

FasL, TRAIL, PD-L1

tEVs frequently express apoptosis-inducing ligands such as Fas ligand (FasL), TNF-associated apoptosis-inducing ligand (TRAIL), and PD-L1, which engage their respective receptors on immune cells to transmit inhibitory or apoptotic signals [34–36]. Blocking the FasL–Fas interaction abrogates this suppression, underscoring its direct role in immune evasion [34].

For instance, ovarian cancer–derived tEVs expressing major histocompatibility complex-I (MHC-I) and FasL inhibit Jurkat T-cell activation, downregulating CD3ζ and Janus kinase (JAK)3 expression after 48 h of co-culture [32]. Similarly, melanoma-derived tEVs bearing PD-L1 inhibit CD8⁺ T-cell proliferation, cytokine secretion, and cytotoxicity via PD-1–dependent signaling, leading to systemic immune escape [37]. TRAIL-enriched EVs from glucose-deprived tumor cells activate nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB) in macrophages via TRAIL-receptor2 (TRAIL-R2), polarizing them into a pro-inflammatory poliovirus receptor (PVR)⁺ phenotype that suppresses natural killer (NK) cells [33].

TGF-β

Transforming growth factor-β (TGF-β) within tEVs is a potent immunosuppressive mediator that reshapes myeloid and lymphoid differentiation. tEV-associated TGF-β drives monocyte differentiation into CD14⁺ HLA⁻DR⁻/low myeloid-derived suppressor cells (MDSCs), which lack antigen-presenting capability and express high levels of inhibitory molecules, including TGF-β itself [38]. Furthermore, elevated levels of TGF-β1 and PD-L1 in tEVs have been linked to resistance to human epidermal growth factor receptor 2 (HER2)-targeted therapies, as exosomes transfer resistance phenotypes to originally sensitive cells [39].

NKG2D ligands

As an activating immunoreceptor, natural killer cell group 2D (NKG2D) is expressed on the surface of NK cells, natural killer T (NKT) cells, and CD8⁺ T cells. It serves as a receptor for ligands that are induced by cellular stress, such as major histocompatibility complex class I chain-related protein A/B (MICA/B) and UL16-binding proteins (ULBPs) [40]. Tumor cells frequently downregulate these ligands to escape recognition. Intriguingly, tEVs carrying NKG2D ligands (NKG2DL) or TGF-β1 paradoxically suppress NK and CD8⁺ T cell activity by inducing NKG2D receptor internalization and degradation, thereby attenuating cytotoxicity [41]. Antibody blockade of ULBP1-5 or MICA/B restores NKG2D expression and effector function [42, 43].

Notably, tEV-mediated effects can be context-dependent. For example, BCL2-associated athanogene 6 (BAG6)—a ligand for natural cytotoxicity receptor 30 (NKp30)—can suppress NK cytotoxicity when presented in soluble form on tEVs but can enhance activation when exosome-bound [44]. Likewise, tEVs may occasionally carry tumor antigens and heat shock proteins (HSP)90, eliciting protective immune responses under certain microenvironmental contexts [45]. This duality highlights the complex balance between immune activation and suppression mediated by tEVs.

Induction of immunosuppressive cells by tEVs

Regulatory T cells (Tregs)

T cells fall into cytolytic CD8⁺ and helper CD4⁺ subsets; the former destroy MHC-I-presenting tumor cells, while the latter orchestrate responses via MHC-II. Most intratumoral forkhead box P3 (FOXP3)⁺ Tregs differentiate from naïve CD4⁺ precursors and, through interleukin (IL)−10/TGF-β, suppress cytotoxic immunity to enable tumor escape [46]. Co-culture of patient-derived tEVs with primary T cells downregulates CD3ζ and JAK3 in activated CD8⁺ T cells, induces apoptosis via Fas/FasL, and promotes the differentiation of CD4⁺ T cells into CD4⁺CD25^highFOXP3⁺CD39⁺ Tregs expressing IL-10, TGF-β, and CTLA-4 [36].

tEVs also transfer CD73 to Tregs, enabling adenosine generation and immunosuppressive signaling even in CD73-negative cells [47]. Mechanistically, tEVs activate small mothers against decapentaplegic (SMAD)2/3 and signal transducer and activator of transcription (STAT)3 phosphorylation, reinforcing the TGF-β1 axis and enhancing IL-10 secretion [48]. Notably, these effects rely primarily on surface receptor signaling—including Ca²⁺ influx and adenosine A2A receptor (A2AR) activation—rather than tEV internalization [49].

Tumor-Associated macrophages (TAMs)

Macrophage polarization spans a broad spectrum of functional states, which are often categorized into simplified in vitro paradigms for study. In this context, the classically activated (M1) phenotype is recognized for its pro-inflammatory and microbicidal functions, while the alternatively activated (M2) phenotype is associated with anti-inflammatory responses and tissue repair [50]. M2 polarization is primarily governed by phosphatidylinositol3-kinase (PI3K)/protein kinase B (AKT) and JAK/STAT signaling [51, 52]. tEVs can modulate polarization via ncRNAs. Table 2 summarizes the key mechanisms by which tEV-carried ncRNAs promote M2 polarization, ultimately contributing to an immunosuppressive TME [53–60].

Table 2.

tEV-encapsulated ncRNAs promoting M2 polarization

Cancer Type	ncRNA Type	Molecular Target	Signaling Pathway Affected	Key Mechanisms & Downstream Effects	Ref.
Colorectal Cancer	miR-25-3p, miR-130b-3p, miR-425-5p	PTEN	PI3K/AKT	Inhibits PTEN 3’UTR to activate PI3K/AKT; promotes M2 phenotype (CD206, Arg-1, IL-10) and suppresses M1 markers (iNOS, TNF-α)	[56, 57]
NSCLC	circFARSA	PTEN	PI3K/AKT	Promotes PTEN ubiquitination/degradation, activating PI3K/AKT and upregulating M2 markers (Arg-1, CD206)	[55]
OSCC	miR-29a-3p	SOCS1	STAT6	Directly inhibits SOCS1, enhancing p-STAT6 and M2 markers (CD163, CD206, Arg-1, IL-10)	[54]
EOC	miR-222-3p	SOCS3	STAT3	Directly inhibits SOCS3, activating STAT3 to upregulate M2 markers (CD206, Arg-1, IL-10) and promote angiogenesis	[53]
Pancreatic Cancer*	miR-301a-3p	PTEN	PI3Kγ/AKT	Hypoxia-induced; targets PTEN to activate PI3Kγ/AKT/mTORC, driving M2 polarization (CD206, Arg-1, IL-10), metastasis and EMT	[59, 60]
EOC*	miR-21-3p, miR-125b-5p, miR-181d-5p	SOCS4/5	STAT3	Hypoxia-induced via HIF-1α/2α; targets SOCS4/5 to activate STAT3, inducing M2 polarization and tumor progression	[58]

*Indicates ncRNAs upregulated under hypoxic conditions

Phosphatase and tensin homolog (PTEN); Non-small cell lung cancer (NSCLC); Oral squamous cell carcinoma (OSCC); Epithelial ovarian cancer (EOC); Suppressor of cytokine signaling (SOCS); Signal transducer and activator of transcription (STAT); phosphatidylinositol3-kinase/protein kinase B (PI3K/AKT)

Myeloid-derived suppressor cells (MDSCs)

MDSCs, immature myeloid cells expanded in tumors and circulation, inhibit T-cell activation and support tumor progression [61, 62]. miR-210 and miR-494 in tEVs target C-X-C motif chemokine ligand 12 (CXCL12), phosphatase and tensin homolog (PTEN), and IL-16 to enhance MDSC recruitment [63, 64]. tEV surface HSP72 engages Toll-like receptor (TLR)2, activating the TLR2-myeloid differentiation factor 88 (MyD88)-IL-6-STAT3 axis, sustaining MDSC immunosuppressive functions [65]. Extracellular signal-regulated kinase (ERK) pathway activation further augments this effect [65].

Impairment of tumor antigenicity and antigen presentation by tEVs

Beyond immune modulation, tEVs promote immune evasion by reducing tumor antigen visibility and impairing antigen presentation. tEVs facilitate these processes by delivering miRNAs that suppress antigen presentation in dendritic cells (DCs). For instance, tEVs deliver miR-212-3p and miR-203 to DCs, suppressing regulatory factor X-associated protein-dependent MHC-II expression and TLR4 signaling, respectively, thereby impairing antigen presentation [66, 67].

tEV-Mediated immunometabolic reprogramming and metabolic checkpoints

Regulation of glucose metabolism

Glucose metabolism is a fundamental determinant of immune cell activation, differentiation, and effector function [68]. Pro-inflammatory lymphocytes and immunosuppressive subsets rely on distinct glycolytic programs, underscoring the context-dependent role of glucose flux in shaping antitumor immunity [68].

Numerous studies have shown that tEVs orchestrate a multifaceted reprogramming of glucose metabolism that subverts normal immune activity [11, 12, 16]. In cytotoxic lymphocytes, tEVs impair glycolytic flux, leading to metabolic exhaustion and functional suppression [11]; in contrast, within TAMs, they enhance glycolysis to sustain immunosuppressive polarization [12, 16].

Extracellular leucine rich repeat and fibronectin type III domain-containing 1-antisense RNA (ELFN1-AS1)-enriched tEVs promote M2 macrophage polarization through sequestering miR-4644 [16], which leads to the upregulation of the rate-limiting glycolytic enzyme pyruvate kinase M (PKM) and the glucose transporter GLUT1, as well as enhanced hexokinase 2 (HK2) expression via hypoxia-inducible factor-1α (HIF-1α) activation. This metabolic reprogramming toward glycolysis reinforces an immunosuppressive tumor microenvironment [16].

Breast cancer–derived tEVs attenuate AKT/mechanistic target of rapamycin complex 1 (mTORC1) phosphorylation in CD8⁺ T cells, suppressing glycolytic enzymes (enolase1 (ENO1), lactate dehydrogenase A (LDHA)), the glucose transporter GLUT1 and effector cytokines (interferon-γ (IFN-γ), Granzyme B), culminating in T-cell metabolic exhaustion [11, 69].

High-mobility-group protein B1 (HMGB1)-containing tEVs activate the TLR2/NF-κB axis in macrophages, upregulating HIF-1α and inducible nitric oxide synthase (iNOS), augmenting glycolysis and lactate release. Lactate-induced histone deacetylase (HDAC) inhibition then elevates PD-L1 expression, suppressing T-cell activity [12].

Collectively, these findings illustrate that tEVs reshape glucose metabolism across immune subsets to sustain a lactate-rich, hypoxic, and suppressive TME.

Regulation of lipid metabolism

Lipid metabolism governs membrane dynamics, signaling, and immune memory formation [70, 71]. Dysregulated lipid accumulation, however, drives T-cell exhaustion and macrophage dysfunction [72]. Studies have shown that tEVs exploit these vulnerabilities by reprogramming lipid synthesis and catabolism in immune cells [13, 14].

PD-L1-bearing tEVs engage PD-1 on T cells, thereby inducing DNA damage and activating the ataxia-telangiectasia mutated-cAMP response element-binding protein (ATM-CREB) signaling axis. Activated CREB further promotes the phosphorylation of STAT1/3. This coordinated signaling upregulates key cholesterol-synthesizing enzymes (3-hydroxy-3-methylglutaryl coenzyme A reductase (HMGCR), 3-hydroxy-3-methylglutaryl-Coenzyme A synthase 1 (HMGCS1), squalene monooxygenase (SQLE)) and cytosolic phospholipase A2α (cPLA2α). This promotes cholesterol and fatty-acid uptake, lipid droplet formation, and T-cell senescence, impairing cytotoxicity [13].

Spliced X-box binding protein 1 (XBP1s)-driven tEVs from stressed tumor cells enrich cholesterol cargo via upregulation of rate-limiting enzymes in the cholesterol biosynthesis pathway. These cholesterol-loaded tEVs are taken up by MDSCs, promoting their proliferation and immunosuppressive function, thereby suppressing anti-tumor CD8⁺ T cell responses [14].

Regulation of amino acid metabolism

Amino acids are vital for immune cell proliferation and effector functions, and their catabolism yields key immunoregulatory metabolites such as kynurenine [73, 74]. tEVs disrupt this balance by altering amino acid utilization or catabolic enzyme expression.

miR-193a-enriched tEVs from colorectal cancer suppress sirtuin 7 (SIRT7) in CD8⁺ T cells, enhancing branched-chain amino acid (BCAA) catabolism (via isovaleryl-CoA dehydrogenase (IVD) and BCAA transaminase 2 (BCAT2) succinylation), accumulating acetyl-CoA and fatty acids, and promoting PD-1/T-cell immunoglobulin and mucin-domain containing-3 (TIM-3) expression with diminished cytokine release [75–77].

miR-142-5p-positive tEVs from cutaneous squamous cell carcinoma (CSCC) suppress AT-rich interaction domain 2 (ARID2) in lymphatic endothelial cells (LECs), thereby reducing DNA methyltransferase (DNMT)1 recruitment to the IFN-γ promoter. This demethylates the locus, increases IFN-γ transcription, and up-regulates indoleamine 2,3-dioxygenase (IDO). Elevated IDO activity depletes tryptophan and raises kynurenine, promoting PD-1 expression and CD8⁺ T-cell apoptosis to establish immune privilege [15, 78].

Toward a novel concept: tEV-Defined metabolic checkpoints

The mechanisms outlined above reveal that tEVs selectively target metabolic control nodes—molecular “switches” that determine immune cell fate [11–16]. These hubs, termed metabolic checkpoints, represent the metabolic counterpart of canonical immune checkpoints [18]. When hijacked by tEVs, they impose metabolic paralysis, enforce immune suppression, and promote therapeutic resistance.

Conceptually, targeting these tEV-defined metabolic checkpoints offers a novel strategy to dismantle the immunosuppressive TME and restore immunotherapy sensitivity [13]. Key potential checkpoints summarized in Table 3 include regulators of glycolysis (HK2, LDHA, PKM) [16], lipid synthesis (HMGCR) [13], and amino acid catabolism (IDO) [15].

Table 3.

tEV-mediated potential metabolic checkpoints

Metabolic Pathway	tEV Cargo	Target Cell	Molecular Mechanism/Key Regulator	Functional Outcome	Proposed Metabolic Checkpoint
Glucose Metabolism	ELFN1-AS1	Macrophages	↑miR-4644 ↑ Glycolytic enzymes (HK2, PKM, GLUT1)	M2 Polarization, Immunosuppression	HK2, PKM, GLUT1
	Functional molecules	CD8⁺ T cells	↓ AKT/mTOCR1, ↓ Glycolytic enzymes (GLUT1, ENO1, LDHA)	Metabolic Exhaustion, Impaired Effector Function	GLUT1, ENO1, LDHA
Lipid Metabolism	PD-L1	CD4+/CD8⁺ T cells	↑ CREB/STAT ↑ Cholesterol synthases (HMGCR, HMGCS1) ↑ cPLA2α	Lipid Droplet Accumulation, T cell Senescence	HMGCR, HMGCS1, cPLA2α
	Cholesterol	MDSCs	↑ Rate-limiting enzymes in cholesterol biosynthesis	MDSC Expansion, Enhanced Immunosuppression	Rate-limiting enzymes in cholesterol biosynthesis
Amino Acid Metabolism	miR-193a	CD8⁺ T cells	↓ SIRT7 ↑ BCAA catabolism (IVD, BCAT2)	Suppressed CD8+T cell Activation, ↑ PD-1/TIM-3	SIRT7
	miR-142-5p	LECs	↓ ARID2 ↑ IDO, Tryptophan → Kynurenine	CD8⁺ T cell Suppression	IDO

Extracellular leucine rich repeat and fibronectin type III domain-containing 1-antisense RNA(ELFN1-AS1); Programmed death ligand 1(PD-L1); Myeloid-derived suppressor cells (MDSCs); Lymphatic endothelial cells (LECs); Hexokinase 2 (HK2); Pyruvate kinase M (PKM); Protein kinase B/Mechanistic target of rapamycin complex 1 (AKT/mTOCR1); Enolase1 (ENO1); Lactate dehydrogenase A (LDHA); cAMP response element-binding protein (CREB); Signal transducer and activator of transcription (STAT); 3-Hydroxy-3-methylglutaryl coenzyme A reductase (HMGCR); 3-Hydroxy-3-methylglutaryl-Coenzyme A synthase 1 (HMGCS1); Cytosolic phospholipase A2α (cPLA2α); Sirtuin 7 (SIRT7); Isovaleryl-CoA dehydrogenase (IVD); BCAA transaminase 2 (BCAT2); AT-rich interaction domain 2 (ARID2); Programmed cell death protein 1 (PD-1); T-cell immunoglobulin and mucin-domain containing-3 (TIM-3); Indoleamine 2,3-dioxygenase (IDO)

Emerging tumor therapeutic strategies

Targeting tEV biological processes

Inhibition of tEV production and secretion

tEVs are indispensable for tumor progression, angiogenesis, and metastasis. Thus, disrupting their production, secretion, or uptake offers a promising antitumor strategy.

• Rab27a inhibition: Silencing Rab27a, a GTPase critical for vesicle exocytosis, in cultured melanoma cells markedly reduces EV release [79].

• nSMase2 inhibition: The enzyme nSMase2 promotes ceramide-dependent exosome formation. nSMase2 inhibitors such as GW4869 directly reduce tEV secretion and metastatic potential [80].

• Natural/repurposed inhibitors: Compounds such as chloramidine suppress tEV biogenesis and enhance chemotherapy sensitivity by increasing intracellular drug accumulation [81].

Inhibition of tEV cellular uptake

tEV internalization by recipient cells depends on heparan sulfate proteoglycans (HSPGs) and endocytic machinery [82].

• In glioblastoma models, soluble heparin competitively inhibits HSPG-mediated tEV uptake by monocytes, thereby preventing their immunosuppressive reprogramming [83].

• In breast cancer, heparin treatment blocks tEV internalization and mitigates pro-tumorigenic effects [84].

• Reserpine also impairs tEV uptake by stromal cells, disrupting pre-metastatic niche formation and metastasis [85].

A key limitation remains the lack of tumor specificity. Future therapeutic strategies should prioritize cargo-specific targeting—selectively inhibiting oncogenic or immunosuppressive tEV subtypes while sparing physiological vesicle communication [86].

Targeting metabolic checkpoints in combination with immune checkpoint inhibitors (ICIs)

The transformative impact of ICIs in oncology has been tempered by the persistent challenge of both primary and acquired resistance. A considerable proportion of patients fail to achieve durable responses or eventually relapse, highlighting the limitations of targeting PD-1/PD-L1 or CTLA-4 pathways in isolation [1]. These observations emphasize the need for a paradigm shift—from focusing solely on receptor–ligand inhibition to a broader, systems-level understanding of the TME.

As discussed in Sect. 3, tEVs orchestrate immunosuppression not only through surface immune-inhibitory ligands but also by reprogramming immune metabolism. This tEV-driven metabolic remodeling establishes a “metabolic barrier” that renders effector T cells bioenergetically exhausted and unresponsive to ICI-mediated reinvigoration.

This conceptual framework supports a synergistic dual-target strategy: simultaneous blockade of tEV-defined metabolic checkpoints and immune checkpoints. Effective antitumor immunity requires two conditions:

• Metabolically competent immune cells capable of sustained effector function.

• The removal of inhibitory signals restraining their activation.

While ICIs address the latter, they fail to rescue metabolic dysfunction established by tEVs. Targeting metabolic checkpoints aims to restore the former, rebuilding a reservoir of metabolically healthy T cells that can then fully respond to PD-1/PD-L1 blockade. This complementary interaction—metabolic repair plus signal release—offers a rational next-generation immunotherapy framework. Below, we summarize three representative tEV-defined metabolic checkpoints that illustrate the translational potential of this dual-target paradigm (Fig. 3).

• The miR-193a/SIRT7 Axis in Colorectal Cancer

  In colorectal cancer, tEVs enriched in miR-193a suppress the metabolic regulator SIRT7, a nicotinamide adenine dinucleotide (NAD⁺)-dependent deacetylase. This repression enhances BCAA catabolism through succinylation and activation of IVD and BCAT2, leading to excessive acetyl-CoA and fatty acid accumulation. Consequently, CD8⁺ T cells undergo metabolic collapse, cytokine suppression, and upregulation of exhaustion markers PD-1 and TIM-3 [75–77].

Fig. 3.

Overcoming immunosuppression via metabolic checkpoint inhibition combined with ICIs. Representative strategies for reversing tEV-induced immune dysfunction. (1) In colorectal cancer, tEV-derived miR-193a downregulates sirtuin 7 (SIRT7) in T cells, promoting aberrant branched-chain amino acid catabolism and upregulating programmed cell death protein 1 (PD-1)/T-cell immunoglobulin and mucin-domain containing-3 (TIM-3). Combining SIRT7 activators with immune checkpoint inhibitors (ICIs) may restore T-cell metabolic fitness. (2) In cutaneous squamous cell carcinoma, miR-142-5p delivered by tEVs silences AT-rich interaction domain 2 (ARID2) in LECs, leading to indoleamine 2,3-dioxygenase (IDO) upregulation, tryptophan depletion, and kynurenine accumulation. Dual blockade with IDO inhibitors + ICIs can reinvigorate T cells. (3) In melanoma and breast cancer, tEVs present PD-L1 on their surface, which engages with PD-1 on T cells and triggers a DNA damage response. This DNA damage leads to the activation of cAMP response element-binding protein (CREB), which in turn further potentiates the phosphorylation and activation of signal transducer and activator of transcription (STAT)1/3 signaling. Collectively, the coordinated action of CREB and STAT signaling orchestrates a pathological reprogramming of lipid metabolism, ultimately driving T cells into a state of senescence. Combining statins with ICIs may prevent lipid-driven exhaustion and reinforce antitumor immunity

Combinatorial hypothesis

Activating SIRT7 or inhibiting miR-193a could restore metabolic fitness and cytokine production. When combined with PD-1 blockade, metabolically rejuvenated T cells would respond more robustly, achieving synergistic tumor control.

• The miR-142-5p/ARID2/IDO Axis in CSCC

  CSCC-derived exosomes deliver miR-142-5p to LECs. By targeting ARID2, miR-142-5p inhibits DNMT1-mediated methylation of the IFN-γ promoter, leading to enhanced IFN-γ expression. This upregulates IDO activity, accelerating tryptophan-to-kynurenine conversion, which in turn drives CD8⁺ T cell exhaustion [15, 78].

Combinatorial hypothesis

Dual targeting of IDO and PD-1/PD-L1 could reverse metabolic and signaling suppression simultaneously. IDO inhibition restores tryptophan availability and T-cell viability, while ICIs release checkpoint inhibition—together revitalizing cytotoxic immunity.

• The PD-L1–Lipid Biosynthesis Axis in Melanoma and Breast Cancer

  In melanoma and breast cancer, tEVs carrying PD-L1 activate ATM–CREB and STAT1/3 signaling upon PD-1 engagement. This cascade transcriptionally upregulates lipid biosynthesis enzymes (HMGCR, ACAT, cPLA2α), enhancing cholesterol and fatty acid accumulation. The resulting lipid overload promotes T-cell senescence distinct from exhaustion, as evidenced by the lack of CTLA-4 upregulation and impaired cytotoxicity [13].

Combinatorial hypothesis

Combining statins (e.g., simvastatin)—which inhibit HMGCR—with anti-PD-L1/CTLA-4 antibodies could normalize T-cell lipid metabolism while simultaneously removing inhibitory signaling. This dual correction of metabolic and immune dysfunction could overcome ICI resistance in lipid-rich tumors.

Summary

The convergence of immune and metabolic checkpoints represents a fertile ground for therapeutic innovation. Rational combinatorial strategies targeting tEV-mediated metabolic checkpoints and canonical immune checkpoints may unlock durable responses in patients who currently fail ICI monotherapy. Rigorous preclinical and clinical validation will be essential to transform these conceptual synergies into clinical reality.

Exosomes in tumor liquid biopsy and clinical translation

Overview and diagnostic potential of exosomes in liquid biopsy

Liquid biopsy is a diagnostic strategy that analyzes circulating biomarkers in bodily fluids, overcoming limitations of traditional tissue biopsy—such as invasive procedures, tumor heterogeneity, and temporal sampling constraints. This method primarily utilizes non-invasive/minimally invasive sample sources like plasma, serum, saliva, and urine, but can also be extended to more invasive samples, such as cerebrospinal fluid via lumbar puncture [87–89]. Among all circulating biomarker carriers, exosomes have emerged as the most promising due to their cell-specific origin, molecular diversity, and remarkable stability in circulation.

Exosomes can be isolated through various techniques, including ultracentrifugation (the gold standard), polymer-based precipitation, immunoaffinity capture, size-exclusion chromatography, and microfluidic chip technologies (Fig. 4). Each method offers unique advantages depending on the downstream application, as summarized in Table 4 [90, 91].

Fig. 4.

Integration of liquid biopsy and artificial intelligence for precision oncology Current and emerging applications of artificial intelligence (AI)-integrated exosome analysis in oncology. At present, exosomes isolated from various body fluids are profiled for molecular contents (nucleic acids, proteins, lipids), which are analyzed via AI algorithms to enable biomarker discovery, early cancer detection, and optimization of exosome-based therapeutics Future directions include the development of multimodal AI systems integrating exosomal multi-omics with clinical and imaging data to support intelligent diagnostic and prognostic platforms. AI-driven microfluidic devices will facilitate real-time exosome monitoring and standardized, large-scale analyses, ultimately advancing precision medicine

Table 4.

Advantages and disadvantages of exosome isolation methods

Isolation Method	Key Advantages	Key Disadvantages
Ultracentrifugation	- Mature and widely adopted, considered the “gold standard” - Suitable for large-volume samples - Relatively low cost	- Time-consuming and cumbersome protocol - High-speed centrifugation may damage exosome structure and bioactivity - Moderate purity, with co-precipitation of contaminants like protein aggregates
Polymer Precipitation	- Simple to operate, requires no special equipment - Suitable for both small and large sample volumes - High yield	- Low purity, prone to co-precipitate contaminants like lipoproteins and proteins - Residual polymers may interfere with downstream analysis
Immunoaffinity Capture	- High purity and specificity - Capable of isolating exosomes of specific origin or subpopulations - Ideal for diagnostic applications	- High cost (dependent on specific antibodies) - Low yield and processing volume - Elution conditions may compromise the native structure of exosomes
Size-Exclusion Chromatography	- Well-preserves the native structure and biological function of exosomes - High purity and good reproducibility - Rapid procedure	- Relatively high device/column cost - May require an additional concentration step for low-concentration samples
Microfluidics	- Very low sample consumption - Efficient, rapid, and amenable to automation and integrated detection - Potential for high-throughput analysis	- Technology not yet standardized, limiting clinical application - Limited isolation volume may affect downstream omics analysis - Device fabrication can be complex

Exosomal microRNAs (ex-miRs) have demonstrated exceptional diagnostic potential, showing high sensitivity and specificity in various cancers [92]. Their strengths include:

• Broad availability across multiple body fluids [87–89];

• Significant tumor-specific upregulation [93];

• Nuclease resistance via lipid bilayer protection [94];

• Compatibility with quantitative detection platforms such as quantitative real-time polymerase chain reaction (qPCR) and next-generation sequencing (NGS) [95, 96].

Furthermore, immune-related proteins, particularly those immune checkpoint molecules, hold prognostic significance and may predict responsiveness to immunotherapy [97].

Exosome-Based tumor diagnostics

Researchers have identified unique exosomal miRNA expression profiles across different cancer types with diagnostic utility. For example, oncogenic species like miR-21-5p and miR-221-3p are elevated in hepatocellular carcinoma (HCC) serum [98], while miR-155 and miR-375 are upregulated in the blood of breast and gastric cancer patients, respectively [99, 100]. Notably, diagnostic markers from other bodily fluids enable “local liquid biopsy,” such as upregulated salivary miR-486-5p in head and neck squamous cell carcinoma (HNSCC) [101] and reduced cerebrospinal fluid miR-200c in primary central nervous system lymphoma [102]. The biomarker spectrum also extends beyond miRNAs to include molecules like urinary lncRNA urothelial carcinoma-associated 1 (UCA1) in bladder cancer [103] and serum glycoprotein-1 (GP1) protein in pancreatic cancer [104], revealing the broad potential of multi-component exosomal assays. We summarize more exosomal diagnostic biomarkers in Table 5.

Table 5.

Exosomes as tumor diagnosis and prognosis biomarkers

miRNAs/LncRNAs	Cancer Types/Source	Diagnosis/Prognosis	Expression	Ref.
miR-223-3p, miR-21-5p, miR-221-3p, and miR-10b-5p	HCC/Serum	diagnosis	↑	[98]
miR-92b-5p and miR-155	Breast cancer/Serum	diagnosis	↑	[100, 105]
miR-486-5p and miR-10b-5p	HNSCC/Saliva	diagnosis	miR-486-5p↑ miR-10b-5p↓	[101]
miR-23a, miR-486, and miR-320a	Colorectal cancer/Plasma	diagnosis	miR-320a↑ miR-23a and miR-486↓	[106]
miR-92b-5p and miR-200b-3p	NSCLC/Serum	diagnosis	miR-200b-3p↑ miR-92b-5p↓	[107]
miR-31, miR-192, and miR-375	Gastric cancer/Plasma	diagnosis	↑	[99]
miR-200c and miR-141	PCNSL/Cerebrospinal fluid	diagnosis	↓	[102]
miR-200b-3p, miR-182-5p, and miR-629-5p	Lung adenocarcinoma/Pleural effusion	diagnosis	↑	[108]
UCA1	Bladder cancer/Urine	diagnosis	↑	[103]
GP1	Pancreatic cancer/Serum	diagnosis	↑	[104]
miR-151a-3p, miR-106b-5p, miR-183-5p, and miR-452-5p	HCC/Serum	poor prognosis	↑	[109]
miR-3960	TNBC/Serum	poor prognosis	↑	[110]
miR-200a/c-3p	Cholangiocarcinoma/Bile and Serum	poor prognosis	↑	[111]
miR-181a-5p and miR-574-5p	NSCLC/Serum	poor prognosis	↑	[112]
miR-92a-3p and miR-221-3p	Colorectal cancer/Plasma	poor prognosis	↑	[113]
miR-23b-3p	ESCC/Serum	poor prognosis	↓	[114]
miR-1307-5p	OSCC/Saliva	poor prognosis	↑	[115]
lncRNA MALAT1, PCAT-1, and SPRY4-IT1	Bladder cancer/Urine	poor prognosis	↑	[116]
PD-L1	NSCLC, melanoma/Blood	poor prognosis	↑	[37, 97]

Hepatocellular carcinoma (HCC); Head and neck squamous cell carcinoma (HNSCC); Non-small cell lung cancer (NSCLC); Primary central nervous system lymphoma (PCNSL); Triple-negative breast cancer (TNBC); Esophageal squamous cell carcinoma (ESCC); Oral squamous cell carcinoma (OSCC)

Exosome-Based tumor prognosis assessment

Beyond diagnosis, exosomal cargoes have proven to be powerful indicators for tumor prognosis and therapeutic response. The upregulation of specific exosomal miRNAs, such as miR-183-5p in HCC and miR-1307-5p in oral squamous cell carcinoma (OSCC), is strongly correlated with aggressive tumor phenotypes and poorer patient survival [109, 115]. Crucially, these molecules often mechanistically underlie treatment resistance; for instance, elevated exosomal miR-3960 in triple-negative breast cancer (TNBC) is associated with cisplatin resistance [110], while downregulation of miR-23b-3p in esophageal squamous cell carcinoma (ESCC) suggests resistance to chemoradiotherapy [114]. In the era of immunotherapy, exosomal PD-L1 has emerged as a particularly significant prognostic factor. High levels of circulating exosomal PD-L1 prior to treatment predict diminished response rates and shorter progression-free survival in melanoma and non-small cell lung cancer (NSCLC) patients undergoing immune checkpoint blockade therapy [37, 97], highlighting its potential in guiding patient stratification. We summarize more exosomal prognostic biomarkers in Table 5. Table 6 presents clinical studies on exosomes as diagnostic and prognostic biomarkers.

Table 6.

Clinical trials of exosomes as biomarkers

Title	Status	Condition/Purpose	Prospective Outcome Measures	Dates	ID
Serum Exosomal Long Noncoding RNAs as Potential Biomarkers for Lung Cancer Diagnosis	Completed	Lung cancer/ Diagnosis	Enrollment number: 1000 The expression levels of serum exosome long non-coding RNA and tumor biomarkers such as CEA, NSE, SCC, and CYFR2A-1	Start: 2017.1.1 Completion: 2020.12.31	NCT03830619
Role of Exosomes in Pancreatic Cancer Progression	Recruiting	Pancreatic cancer/ Diagnosis	Enrollment number: 60 Characterize the extracellular vesicle/exosome content of patients with pancreatic carcinoma.	Start: 2022.5.22 Completion: 2027.4.30	NCT06777030
Construction of Microfluidic Exosome Chip for Diagnosis of Lung Metastasis of Osteosarcoma	Completed	Osteosarcoma, lung metastases/Diagnosis and Prognosis	Enrollment number: 60 The association of disease recurrence with plasma levels of exosome and their subgroups	Start: 2020.10.1 Completion: 2022.9.30	NCT05101655
Plasma Exosome RNA to Diagnose Prostate Cancer	Enrolling by invitation	Prostate cancer/ Diagnosis	Enrollment number: 1200 (1)The diagnostic efficacy of plasma exo-RNA panel alone in diagnosing Pca (2)Diagnostic efficacy of plasma exo-RNA panel in diagnosing csPCa, insignPCa, and its comparison with other imaging methods or blood tests	Start: 2021.3.12 Completion: 2025.7.31	NCT06604130
Stomach Cancer Exosome-based Detection	Completed	Gastric cancer/ Diagnosis	Enrollment number: 809 (1)Sensitivity (2)Specificity (3)Proportion of correct predictions (true positives and true negatives) among the total cases (i.e., accuracy)	Start: 2023.3.15 Completion: 2024.6.15	NCT06342427
A Prospective Study of Predicting Prognosis and Recurrence of Thyroid Cancer Via New Biomarkers, Urinary Exosomal Thyroglobulin and Galectin-3	Active, not recruiting	Thyroid cancer/ Prognosis	Enrollment number: 74 The correlation of outcome (including recurrence, lymph nodes metastasis) together with unknown/fresh biomarkers: Thyroglobulin, galectin-3, Calprotectin A8, Calprotectin A9, TKT, Annexin II, Afamin, Keratin 8, Keratin 9, Angiopoietin-1, and TIMP	Start: 2018.8.3 Completion: 2023.7.31	NCT03488134
A Companion Diagnostic Study to Develop Circulating Exosomes as Predictive Biomarkers for the Response to Immunotherapy in Renal Cell Carcinoma	Recruiting	Renal cell carcinoma/ Prognosis	Enrollment number: 100 The correlation between the circulating exosome levels and the tumor responsiveness	Start: 2023.1.1 Completion: 2025.12.31	NCT05705583

Artificial intelligence in liquid biopsy applications

AI technologies are revolutionizing exosome research by decoding complex biomolecular signatures [117–119], as illustrated in Fig. 4.

• Machine learning (ML) enables high-dimensional biomarker mining, optimizing miRNA panels for early detection. For example, Pu et al. identified a diagnostic miRNA panel for pancreatic cancer via ML, improving accuracy over conventional assays [119].

• Generative and reinforcement learning optimize exosome-drug delivery systems by simulating uptake and tuning cargo ratios [117].

• AI-integrated Raman spectroscopy (SERS) enables single-assay detection of multiple cancers with high accuracy [118].

However, current models rely largely on unimodal exosomal data, which ignores critical multi-omics and clinical variables, thereby limiting prediction accuracy. As shown in Fig. 4, future work will focus on multimodal AI and AI-driven microfluidics.

• Multimodal AI platforms integrating exosomal multi-omics, imaging, and clinical metadata for real-time, precision oncology [117];

• AI-driven microfluidics enabling continuous, real-time exosome monitoring for dynamic cancer management [120].

Current challenges

Despite compelling evidence supporting the critical role of tEVs in orchestrating immunometabolic reprogramming and mediating resistance to immunotherapy, several challenges hinder their mechanistic study and clinical application as biomarkers and therapeutic targets.

A major challenge lies in the functional duality and pleiotropy of tEVs. The effects of specific tEV components are often context-dependent and paradoxical. For instance, colorectal cancer-derived tEVs carrying DNA have been linked to reduced liver metastasis in one study [121], yet in another context, they contribute to the establishment of an immunosuppressive pre-metastatic niche in the liver [122]. This underscores the complexity and microenvironment-dependent nature of tEVs, making their net effect difficult to predict and therapeutically modulate.

Another significant challenge is the heterogeneity of tEVs. Rather than a uniform population, tEVs comprise heterogeneous subsets (e.g., exosomes, microvesicles) originating from diverse tumor cell subpopulations and intracellular compartments. As a result, there is considerable variability in their size, molecular cargo (proteins, nucleic acids, lipids), and functional properties [19, 26]. This diversity complicates the precise molecular characterization of tEVs and hinders the identification of the most pathologically relevant subpopulations for targeted therapies.

From a technical and translational perspective, the absence of standardized and scalable protocols for tEV isolation, purification, and characterization remains a critical bottleneck. Current isolation methods, such as ultracentrifugation, often face limitations in yield, purity, and reproducibility [123]. The lack of harmonized standards across laboratories impedes direct comparison of research findings and hinders the clinical validation of tEV-derived biomarkers.

Discussion and perspective

Tumor microenvironment (TME) is now widely recognized as a dynamic ecosystem, where immune cell metabolic reprogramming plays a critical role in shaping anti-tumor immune responses [124]. Within this complex network, tumor-derived extracellular vesicles (tEVs) have emerged as key modulators of immunometabolic remodeling, bridging the tumor’s intrinsic metabolic demands with extrinsic immune suppression [11–14, 16]. By delivering bioactive molecules—including proteins, nucleic acids, and metabolites—tEVs significantly alter the metabolic landscape of surrounding immune cells, thereby facilitating immune evasion and contributing to resistance to immune checkpoint inhibitors (ICIs) [125].

A key conceptual advancement presented in this review is the introduction of “metabolic checkpoints”—metabolic nodes or pathways that serve as central regulatory hubs for immune cell function and fate. In contrast to classical immune checkpoints, which primarily function through ligand-receptor interactions, metabolic checkpoints integrate multiple signals, including intracellular metabolic flux and extracellular signaling [18]. tEVs modulate these checkpoints—for instance, by suppressing glycolysis or reprogramming lipid metabolism in CD8⁺ T cells—creating an immunosuppressive microenvironment that dampens cytotoxic responses and actively promotes immune tolerance [11, 13].

Targeting these metabolic checkpoints in combination with ICIs represents a promising therapeutic strategy to overcome resistance to immunotherapy. A preclinical study suggest that metabolic modulators can synergize with ICIs to restore effector T-cell function [13]. However, the clinical translation of these findings requires careful patient stratification, real-time monitoring of metabolic states, and a deeper understanding of the TME’s complexity.

In this context, liquid biopsy platforms utilizing tEV-derived biomarkers offer a transformative diagnostic approach [126]. Unlike traditional tissue biopsies, which provide static snapshots of tumor status, tEV signatures from peripheral blood reflect the dynamic and evolving immunometabolic landscape in real time [127]. When combined with artificial intelligence (AI)-driven analytics, these non-invasive strategies hold the potential to greatly enhance precision oncology by enabling adaptive treatment decisions based on the tumor’s dynamic trajectory [120].

Despite these promising advances, several challenges remain. The inherent heterogeneity of tEVs—originating from diverse tumor regions and metastatic niches—complicates their analysis [19]. While current studies, often relying on single-omics approaches like proteomics, have begun to link tEVs to clinical outcomes [128], broader efforts are hampered by a lack of standardized isolation protocols [19]. Therefore, critical next steps must focus on standardizing tEV isolation, achieving comprehensive multi-omics profiling, and integrating spatial metabolic mapping with clinical data to fully decipher the role of tEVs in cancer progression.

Looking ahead, the intersection of immunometabolism, extracellular vesicle biology, and computational innovation marks a new frontier in cancer research. Harnessing the regulatory potential of metabolic checkpoints, leveraging tEV-based biomarkers, and applying AI-enhanced analytics may ultimately overcome immunotherapy resistance and enable durable clinical responses in cancer patients.

Conclusions

In conclusion, this review consolidates the multifaceted roles of tumor-derived extracellular vesicles (tEVs) as central orchestrators of the immunosuppressive tumor microenvironment (TME). We have highlighted two synergistic mechanisms by which tEVs contribute to immunotherapy resistance: (1) through reprogramming core immunometabolic pathways—glucose, lipid, and amino acid metabolism—tEVs establish metabolic checkpoints that disrupt the function of effector immune cells [11–16]; and (2) by employing direct non-metabolic mechanisms, such as the presentation of immune-inhibitory ligands, induction of immunosuppressive cell populations, and impairment of antigen presentation [129].

Targeting these tEV-mediated processes offers a promising therapeutic avenue. We propose a dual-targeting strategy: combining tEV-defined metabolic checkpoint inhibitors with immune checkpoint inhibitors (ICIs) can synergistically restore anti-tumor immunity by rectifying intrinsic metabolic impairments and relieving extrinsic co-inhibitory signals, the feasibility and efficacy of which have been supported by the existing studies [13]. Additionally, strategies aimed at inhibiting tEV biogenesis, secretion, or uptake present complementary approaches to disrupting this pervasive form of intercellular communication [79–81, 83–85].

Translationally, integrating liquid biopsy platforms based on tEV-derived biomarkers with artificial intelligence (AI) may hold significant potential for advancing personalized cancer immunotherapy. This integrated approach could enable dynamic, non-invasive monitoring of the immunometabolic state, thereby paving the way for adaptive treatment strategies.

By bridging tumor immunology, metabolic regulation, extracellular vesicle biology, and computational innovation, this field is poised to redefine the future of cancer immunotherapy. Continued interdisciplinary efforts will be vital for translating these conceptual advances into clinically actionable strategies, particularly for patients with refractory or relapsing cancers.

Abbreviations

tEVs

Tumor-derived extracellular vesicles

TME

Tumor microenvironment

AI

Artificial intelligence

miRNAs

MicroRNAs

PD-1

Programmed cell death protein 1

PD-L1

Programmed death ligand 1

CTLA-4

Cytotoxic T-lymphocyte-associated protein 4

ESCRT

Endosomal sorting complex required for transport

nSMase2

Neutral sphingomyelinase 2

ILVs

Intraluminal vesicles

MVBs

Multivesicular bodies

mRNAs

Messenger RNAs

lncRNAs

Long non-coding RNAs

circRNAs

Circular RNAs

FasL

Fas ligand

TRAIL

TNF-associated apoptosis-inducing ligand

MHC-I

Major histocompatibility complex-I

JAK

Janus kinase

NF-κB

Nuclear factor kappa-light-chain-enhancer of activated B cells

TRAIL-R2

TNF-associated apoptosis-inducing ligands-receptor 2

PVR

Poliovirus receptor

NK

Natural killer

TGF-β

Transforming growth factor-β

MDSC

Myeloid-derived suppressor cell

HER2

Human epidermal growth factor receptor 2

NKG2D

Natural killer cell group 2D

NKT

Natural killer T

MICA/B

Major histocompatibility complex class I chain-related protein A/B

ULBPs

UL16-binding proteins

NKG2DL

NKG2D ligands

BAG6

BCL2-associated athanogene 6

NKp30

Natural cytotoxicity receptor 30

HSP

Heat shock protein

Tregs

Regulatory T cells

FOXP3

Forkhead box P3

TSG101

Tumor susceptibility gene 101

IL

Interleukin

SMAD

Small mothers against decapentaplegic

STAT

Signal transducer and activator of transcription

A2AR

Adenosine A2A receptor

TAMs

Tumor-associated macrophages

M1 macrophage

Classical macrophage

M2 macrophage

Alternative macrophage

PI3K

Phosphatidylinositol3-kinase

AKT

Protein kinase B

CXCL12

C-X-C motif chemokine ligand 12

PTEN

Phosphatase and tensin homolog

TLR

Toll-like receptor

MyD88

Myeloid differentiation factor 88

ERK

Extracellular signal-regulated kinase

DCs

Dendritic cells

ELFN1-AS1

Extracellular leucine rich repeat and fibronectin type III domain-containing 1-antisense RNA

PKM

Pyruvate kinase M

HK2

Hexokinase 2

HIF-1α

Hypoxia-inducible factor-1α

mTORC1

Mechanistic target of rapamycin complex 1

ENO1

Enolase1

LDHA

Lactate dehydrogenase A

IFN-γ

Interferon-γ

HMGB1

High-mobility-group protein B1

iNOS

Inducible nitric oxide synthase

HDAC

Histone deacetylase

ATM

Ataxia-telangiectasia mutated

CREB

cAMP response element-binding protein

HMGCR

3-Hydroxy-3-methylglutaryl coenzyme A reductase

HMGCS1

3-Hydroxy-3-methylglutaryl-Coenzyme A synthase 1

SQLE

Squalene monooxygenase

cPLA2α

Cytosolic phospholipase A2α

XBP1s

Spliced X-box binding protein 1

SIRT7

Sirtuin 7

BCAA

Branched-chain amino acid

IVD

Isovaleryl-CoA dehydrogenase

BCAT2

BCAA transaminase 2

TIM-3

T-cell immunoglobulin and mucin-domain containing-3

CSCC

Cutaneous squamous cell carcinoma

ARID2

AT-rich interaction domain 2

LECs

Lymphatic endothelial cells

DNMT

DNA methyltransferase

IDO

Indoleamine 2,3-dioxygenase

HSPGs

Heparan sulfate proteoglycans

ICIs

Immune checkpoint inhibitors

NAD⁺

Nicotinamide adenine dinucleotide

ex-miRs

Exosomal miRNAs

qPCR

Quantitative real-time polymerase chain reaction

NGS

Next-generation sequencing

UCA1

Urothelial carcinoma-associated 1

GP1

Glycoprotein-1

HCC

Hepatocellular carcinoma

HNSCC

Head and neck squamous cell carcinoma

OSCC

Oral squamous cell carcinoma

TNBC

Triple-negative breast cancer

ESCC

Esophageal squamous cell carcinoma

NSCLC

Non-small cell lung cancer

ML

Machine learning

SERS

Surface-enhanced Raman spectroscopy

TIGIT

T cell immunoglobulin and TIM domains

SOCS

Suppressor of cytokine signaling

ALIX

ALG-2-interacting protein X

TNF-α

Tumor necrosis factor-α

VEGF

Vascular endothelial growth factor

EGF

Epidermal growth factor

TNFR

Tumor necrosis factor receptor

ICAM-1

Intercellular adhesion molecule-1

RhoA

Ras homologous gene family member A

FASN

Fatty acid synthase

PGK

Phosphoglycerate kinase

ATP

Adenosine triphosphate

EOC

Epithelial ovarian cancer

Authors’ contributions

HYZ drafted the manuscript and performed the revisions. RQZ and XYZ created the figures. SQS compiled the tables. HZ critically revised the manuscript. All authors contributed to the writing of this manuscript. All authors read and approved the final version of the manuscript.

Funding

This research was funded by the National Natural Science Foundation of China (grant number 82373176).

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.