1.1. PCa heterogeneity is a core therapeutic challenge

Globally, prostate cancer (PCa) represents a substantial health burden among men, ranking as the second most frequently diagnosed malignancy in males worldwide [1]. In the United States, PCa is projected to account for approximately 30% of all newly diagnosed cancers in males in 2026, with an estimated 333,830 new cases and 36,320 associated deaths [2]. Although the 5-year survival rate for localised PCa exceeds 95% when detected at an early stage, a considerable proportion of patients experience disease recurrence or treatment failure, ultimately progressing to metastatic PCa (mPCa). This advanced disease stage is associated with a markedly reduced 5-year relative survival of 30–40% [3,4].

This variability in clinical outcomes, in part, reflects the marked biological heterogeneity of PCa. PCa is typically multifocal, comprising spatially and morphologically distinct foci that may harbour divergent genetic mutations and exhibit variable degrees of aggressiveness [5]. High-throughput sequencing studies have shown that heterogeneity not only exists within individual primary tumours (intra-tumoural heterogeneity), but also between patients (inter-patient heterogeneity) and across metastatic lesions (inter-tumoural heterogeneity) [5,6,7,8]. Accumulating evidence indicates that tumour heterogeneity increases with disease progression and constitutes a critical determinant of patient prognosis and survival [9,10,11,12]. Such diversity has important clinical implications, as it contributes to variable treatment response, disease progression and the emergence of resistant phenotypes. However, current prognostic factors, such as PSA, Gleason scores and TNM staging, offer limited insight into the molecular heterogeneity that underpins therapeutic response [13]. This highlights an urgent need for novel biomarkers capable of improving diagnostic precision and enabling more accurate monitoring of PCa progression.

1.2. RT as a driver of PCa plasticity

Radiotherapy (RT) is a cornerstone treatment for PCa and is administered via external beam radiotherapy or brachytherapy [14]. Despite advances in RT techniques, approximately 20–40% of patients treated with RT experience biochemical recurrence and metastatic progression within 5 years [15]. Radioresistance in PCa is frequently associated with metastatic dissemination and development of castration-resistant prostate cancer (CRPC) with a median survival of less than 2 years [16]. While RT is intended to eradicate tumour cells, it may also impose selective pressure on heterogeneous tumour populations, favouring the survival and expansion of subclones with adaptive advantages [17,18]. This adaptive response is characterised by increased phenotypic plasticity [19] which enables PCa cells to reprogramme DNA repair pathways, metabolic processes, and cellular states in response to RT-induced stress [17]. Over time, these adaptations facilitate the emergence of radioresistant and more aggressive tumour phenotypes, ultimately driving disease recurrence, metastatic progression, and the transition to CRPC [20].

1.3. EVs as dynamic mediators of tumour-stroma crosstalk and adaptive signalling

The emergence of radioresistant phenotypes following RT is not solely driven by intrinsic tumour cell adaptations but is also due to therapy-induced communication between PCa cells and the tumour microenvironment (TME) [21]. Increasing evidence indicates that extracellular vesicles (EVs) are important mediators of this crosstalk. EVs are cell-released, lipid-bound, nano-sized vesicles which mediate bidirectional communication between cancer cells and stromal, immune, and endothelial components of the TME through the transfer of bioactive molecular cargo [22], including miRNAs, mRNAs, proteins, lipids and metabolites. Through transfer of this cargo, EVs can alter the phenotype and behaviour of recipient cells and thereby influence tumour progression, stromal remodelling, immune regulation and treatment response. Consequently, profiling dynamic changes in EV cargo following RT may provide emerging insight into the cellular and molecular mechanisms underpinning PCa plasticity and radioresistance.

1.4. Metabolomics: deciphering the functional output of cellular plasticity

Among the various types of cargo carried by EVs, metabolites are significant molecules of interest, as they represent the immediate functional state and biochemical activity of their origin cell and metabolic reprogramming is a recognised hallmark of cancer [23]. Metabolomics captures both the downstream output of genomic and proteomic regulation while also integrating upstream environmental inputs, thereby providing a direct readout of gene-environment interactions [24]. Unlike other static molecules, metabolic profiles are highly dynamic and respond rapidly to environmental influences, making metabolomics particularly well-suited for studying the RT adaptive responses and tumour plasticity [25,26]. Accordingly, analysis of EV-associated metabolites may provide insight into how tumours adapt to RT and may help identify biomarkers of response, early resistance and potential therapeutic vulnerabilities.

This review explores the emerging intersection between radiotherapy, EV biology and metabolic adaptation in PCa. Specifically, it considers how radiation-induced stress may reshape EV secretion and cargo composition, and how these changes may contribute to tumour plasticity, TME remodelling and the evolution of treatment-resistant phenotypes. In this context, particular emphasis is placed on addressing the role of EV cargo in facilitating metabolic reprogramming linked to tumour survival, heterogeneity and radioresistance. Further, this review evaluates the translational relevance of EV metabolomics as a minimally invasive approach for monitoring disease status and treatment response. By integrating current evidence from PCa and the broader literature, this review aims to define the conceptual framework of this field, highlight key biological and methodological challenges and identify priorities in establishing the clinical utility of EV metabolic profiling in precision oncology.

2.1. RT induces EV secretion and TME reshaping

EVs are increasingly recognised as critical mediators of cellular responses to RT. In return, RT has been shown to enhance EV secretion and to modulate EV cargo composition in both malignant and non-malignant cells [27,28,29,30].

For instance, exposure to X-rays increases the release of EVs from neuroblastoma cells, which are then taken up by non-irradiated cells, enhancing their viability and migration [27]. Similarly, gamma radiation significantly enhances EV release from both naïve and polarised macrophages [31]. These alterations are, in part, driven by DNA damage-induced activation of p53 signalling pathways, which promote the selective export of growth-regulatory proteins and other bioactive factors via EVs [32]. Importantly, EV-mediated responses to RT are not restricted to tumour cells. Stromal components, including cancer-associated fibroblasts (CAFs), also exhibit increased EV release following irradiation, suggesting that RT may also reshape TME dynamics via EV-mediated crosstalk [27,28]. Furthermore, RT enhances the uptake and internalisation of tumour-derived EVs by neighbouring recipient cells, thereby amplifying intercellular signalling within the irradiated microenvironment [33,34,35]. In summary, these findings indicate that RT-induced modulation of EV biogenesis, cargo loading, and intercellular transfer may promote adaptive responses that increase the radioresistant and metastatic potential of cancer cells; whether this applies in PCa remains to be confirmed.

2.2. Systematic reprogramming of PCa EV cargo post-RT

RT induces substantial alterations in the biomolecular cargo of EVs, which not only reflect acute cellular stress responses but may also represent adaptations in tumour phenotype.

The immune checkpoint molecule B7-H3 (CD276), a validated marker for distinguishing benign from aggressive PCa [36], has been shown to be enriched in EVs secreted by RT-induced senescent PCa cells. B7-H3 is implicated in the negative regulation of T-cell activation and effector function, thereby contributing to the suppression of anti-tumour immunity [37,38]. In parallel, RT has been reported to induce higher levels of PD-L1 in PCa-derived EVs compared with hormone therapy and chemotherapy [39], suggesting that RT may promote immune-evasive phenotypes. This has clear translational relevance, particularly given the ongoing interest in combining RT with immune checkpoint blockade in PCa. Beyond immune modulation, EV-mediated signalling may directly facilitate adaptive radioresistance. Prostatic stromal cells have been shown to promote radioresistance in PCa cells through EV-mediated transfer of interleukin-8 (IL-8). IL-8 engages CXCR1 and CXCR2 receptors on recipient tumour cells, activating AMP-activated protein kinase (AMPK)-dependent autophagy, which confers a survival advantage under RT-induced stress [40]. Such intercellular signalling mechanisms highlight how EVs can facilitate stress-driven phenotypic adaptation within TME. RT has also been found to upregulate the EV expression of cancer stem-cell (CSC)-associated gene markers (CD133, CD44, ALDH1 and GLUT1), which suggests RT may select for more aggressive, metabolically active CSC phenotypes [41].

2.3. EV-mediated RIBE

RT can induce off-target effects in non-irradiated cells via signal transduction from irradiated cells, a phenomenon termed as the radiation-induced bystander effect (RIBE) [42]. EVs have emerged as important mediators of this process owing to their capacity to facilitate intercellular communication and propagate stress signals. EV-mediated bystander signalling has been shown to trigger diverse biological responses following RT, including the activation of stress pathways, genomic instability, and alterations in cell proliferation and apoptosis [30,43,44,45,46]. This is supported by a study which showed that irradiated mice serum-derived EVs were found to contain mitochondrial DNA (mtDNA) which, upon transfer to recipient healthy cells, induced DNA damage [47]. Consistent with these findings, an in vivo study found that EVs derived from the bone marrow of irradiated mice induced DNA double-strand breaks (DSBs) in non-irradiated recipient mice [48]. These systemic alterations were accompanied by increased chromosomal aberrations and reductions of haematopoietic and immune cell populations. Another study found that EVs released from irradiated PCa cells significantly promoted epithelial–mesenchymal transition (EMT) in non-irradiated PCa cells, characterised by upregulation of mesenchymal markers and downregulation of epithelial markers [41]. This suggests RT-induced EVs promote EMT and metastatic ability in untreated PCa cells, propagating PCa plasticity across distinct tumour regions. Thus, EV-mediated RIBE may contribute to spatially heterogeneous responses to RT, which can influence tumour evolution, metastatic dissemination and overall treatment outcomes.

Taken together, these findings support a conceptual shift toward a more expansive and biologically realistic understanding of RT response in PCa: RT should be regarded not only as a cytotoxic modality, but also as a systems-level adaptation capable of reshaping intercellular communication within the tumour and its microenvironment (Figure 1). In this context, EVs have emerged as a plausible interface through which radiation-induced stress is distributed and potentially stabilised across heterogeneous cell populations. It is clear that increased EV release, altered cargo and RIBE are all consistent with adaptive reprogramming; however, what remains less clear is whether these changes are mechanistic drivers, amplifiers of pre-existing signals or merely molecular readouts of cellular stress. This distinction is fundamental to understanding whether and how such alterations meaningfully contribute to heterogeneity, microenvironment remodelling and long-term therapeutic resistance.

Figure 1. RT stress reshapes EV secretion and cargo to modulate the TME in PCa. RT induces cellular stress in PCa by causing DNA damage, including double-strand breaks and increased production of ROS, which activates stress-response pathways including p53 signalling. These changes promote enhanced biogenesis and secretion of EVs via MVBs into the TME. Following RT, EVs exhibit altered cargo composition, including the upregulation of oncogenic and stress-associated molecules such as B7-H3, CD44, PD-L1, ALDH1, IL-8, GLUT1, CD133, and mitochondrial DNA. These EVs are taken up by recipient cells within the TME, including neighbouring cancer cells, fibroblasts, and immune cells, where they activate downstream signalling pathways such as IL-8/CXCR1/2 and AMPK and reprogramme recipient-cell metabolism. By transferring metabolic cargo between cells, EVs propagate metabolic plasticity across the tumour, favouring the selection and clonal expansion of radioresistant subpopulations. In this way, RT-induced EV signalling links altered cargo to metabolic adaptation and radioresistant subclonal evolution, ultimately contributing to tumour progression and radioresistance. (Created with BioRender.com). Abbreviations: ALDH1, aldehyde dehydrogenase 1; AMPK, AMP-activated protein kinase; B7-H3, B7 homolog 3; CD44, cluster of differentiation 44; CD133, cluster of differentiation 133; CXCR1/2, C-X-C motif chemokine receptor 1/2; DNA, deoxyribonucleic acid; EVs, extracellular vesicles; GLUT1, glucose transporter 1; IL-8, interleukin-8; MVBs, multivesicular bodies; PD-L1, programmed death-ligand 1; ROS, reactive oxygen species; RT, radiotherapy; TME, tumour microenvironment.

3.1. Metabolic hallmarks of PCa and their modulation by EVs

Metabolism adaptation in tumour cells arises from increased demands for energy production, biomass synthesis, cofactor generation and redox homeostasis. To sustain rapid proliferation and survival under environmental stress, cancer cells reprogramme central metabolic pathways to support these biosynthetic and bioenergetic requirements. While enhanced aerobic glycolysis (commonly known as the Warburg effect) is the most widely recognised metabolic alteration in cancer, PCa exhibits a distinct metabolic phenotype that diverges from this classical paradigm. Normal prostate epithelium is specialised for citrate production and secretion, a state maintained by high intracellular zinc concentrations that inhibit mitochondrial aconitase and truncate tricarboxylic acid (TCA) cycle activity [52,53,54]. During malignant transformation, this citrate-oriented metabolism is reprogrammed, leading to unique metabolic adaptations that distinguish PCa from other solid tumours. As a consequence, primary prostate tumours rely on oxidative phosphorylation (OXPHOS) and de novo lipogenesis rather than relying predominantly on glycolysis [55,56]. Enhanced fatty acid (FA) synthesis is evident early in PCa tumourigenesis and is strongly associated with disease progression and poor clinical outcome [57]. With disease progression, however, particularly under androgen deprivation, this metabolic configuration becomes increasingly plastic. In CRPC, glycolytic flux becomes markedly elevated [58]. This indicates that metabolic progression in PCa is best understood not as a linear shift from one dominant pathway to another, but as progressive expansion of metabolic adaptability.

AR signalling is a central regulator of the core metabolic pathways in PCa [59]. Beyond its established role in driving cellular proliferation, AR directly influences mitochondrial respiration, lipid biosynthesis and redox homeostasis to sustain tumour growth and progression. AR activation induces OXPHOS and de novo lipogenesis in parallel with cellular proliferative signalling [60,61,62]. In addition, AR promotes metabolic flux through the pentose phosphate pathway (PPP), partly via the mechanistic target of rapamycin (mTOR)-mediated upregulation of glucose-6-phosphate dehydrogenase (G6PD) [63,64]. Increased G6PD activity augments ribose-5-phosphate synthesis to support nucleotide production while elevating NADPH generation, thereby maintaining reductive biosynthesis and redox balance necessary for sustained lipogenesis and tumour cell survival [65]. AR is also associated with a downregulation of zinc transporter proteins, which reduces intracellular zinc accumulation and relieves inhibition of m-aconitase. This promotes increased TCA flux and ATP production in PCa compared with normal prostate epithelium [54,66], further enhancing bioenergetic efficiency and supporting malignant progression.

Within this framework, emerging evidence indicates that EVs may actively shape and maintain metabolic phenotypes in PCa. One study found that CAF-derived EVs reprogramme the metabolism of recipient PCa cells [67], characterised by decreased OXPHOS and a concomitant increase in glycolysis and glutamine-dependent reductive carboxylation, a metabolic feature so far described in other tumour models [68,69]. Another study in mice found that PCa EVs exhibit marked increases in enzymes involved in PCa metabolic pathways including glycolysis, amino acid metabolism, carbon and pyruvate metabolism [70]. Similarly, EVs released from acid-resistant PCa cells have been shown to be enriched in glycolytic enzymes. Upon uptake by recipient cells, these EVs enhanced glycolytic flux and increased cell proliferation, resulting in a shift towards a glycolysis-dependent metabolic phenotype, associated with advanced PCa [58,71]. Another study found that EVs derived from enzalutamide-resistant PCa cells exhibit increased expression of the transcriptional regulator Yes associated protein (YAP)-1 which drives cancer stemness and lipid metabolism [72]. Notably, treatment of androgen-sensitive LNCaP cells with these EVs not only promotes enzalutamide resistance but also enhances stem-like properties and lipid metabolic activity. These findings suggest that EVs secreted from metabolically or therapeutically adapted tumour subpopulations can disseminate metabolic traits to neighbouring cells, therefore reinforcing metabolic plasticity within the TME. Such EV-mediated metabolic crosstalk may contribute to disease progression, therapeutic resistance, and intra-tumoural heterogeneity in PCa.

Clinically, the metabolic dependence of PCa on AR signalling is shown by the rapid regression of PCa following androgen deprivation therapy (ADT) [73,74]. However, tumours can frequently develop mechanisms to restore AR activity or reprogramme core metabolic pathways, resulting in progression to CRPC [75,76,77,78,79,80]. The ability to reactivate AR-driven signalling and metabolism in the absence of androgens poses a significant clinical challenge and highlights the intrinsic metabolic plasticity of advanced PCa. While many CRPC tumours maintain reactivated AR signalling, some may undergo lineage plasticity, transdifferentiating into neuroendocrine prostate cancer (NEPC), which has an even worse prognosis than classic CRPC [20,81]. This represents an extreme manifestation of tumour plasticity, where metabolic pathways are fundamentally restructured. EVs may represent an important, though not yet fully defined, mechanism by which adaptive traits are propagated. Thus, the transition from hormone-sensitive disease to CRPC or NEPC should be understood not only as a cell-intrinsic evolutionary process but also as one that may be shaped by EV-mediated metabolic reprogramming.

3.2. Radiation-induced metabolic rewiring via EVs: A mechanism of adaptive survival

RT exerts profound oxidative and metabolic stress on tumour cells through DNA double-strand breaks, the generation of reactive oxygen species (ROS), and mitochondrial dysfunction [82,83]. While these effects underpin the cytotoxic efficacy of RT, irradiation can also impose selective pressure that favours the survival of tumour cells capable of adaptive metabolic reprogramming and enhanced DNA repair [84]. Emerging evidence suggests that EVs may function as amplifiers of these adaptive responses, facilitating the transfer of stress-associated metabolic signals and thereby reinforcing radioresistance (Figure 2).

Figure 2. EV metabolomics as a driver, amplifier, or readout of radioresistance. Three non-mutually-exclusive roles are proposed for EV metabolomics in RT-driven prostate cancer evolution, ordered from active/causal (top) to passive (bottom). As a driver, EV-delivered metabolites and metabolic regulators may actively induce a radioresistant phenotype in recipient cells. As an amplifier, EVs released from already-resistant subpopulations may propagate a pre-existing resistant phenotype across the tumour, reinforcing intratumoural heterogeneity. As a readout, the circulating EV metabolome may passively reflect tumour metabolic state and RT response, serving as a liquid-biopsy biomarker without a causal role. Distinguishing among these roles is a central open question for the field. (Created with BioRender.com.) Abbreviations: EV, extracellular vesicle; RT, radiotherapy.

3.2.4. EV-mediated mitochondrial transfer

Beyond soluble metabolites and regulatory RNAs, EVs and related vesicular structures can mediate the horizontal transfer of mitochondria and mitochondrial components, including mtDNA and respiratory-chain proteins, between cells. Recent evidence indicates that EVs can carry mitochondrial cargo and transfer functional mitochondria to recipient cells, which can restore or augment oxidative phosphorylation, rescue bioenergetics under metabolic stress, buffer reactive oxygen species and potentially support survival and proliferation [152,153]. This mechanism has been implicated in tumour progression and in resistance to chemotherapy and RT across other cancer types, including breast cancer [154] and NSCLC [155]. Such bioenergetic rescue may be especially relevant in PCa, where primary tumours depend substantially on oxidative phosphorylation (see Section 3.1), suggesting that EV-mediated mitochondrial transfer could contribute to metabolic plasticity and adaptive survival following RT. This metabolic role is distinct from, and complementary to, the transfer of EV-borne mtDNA as a genotoxic bystander signal discussed earlier (Section 2.3). Direct evidence for mitochondrial transfer in irradiated PCa, however, is currently lacking and represents an important avenue for future investigation.

Across glycolytic, lipid and amino acid metabolic domains, a consistent mechanistic theme emerges: RT acts not only as a cytotoxic modality but as a selective pressure that enriches for tumour cell populations capable of coordinated metabolic adaptation. Although presented pathway-by-pathway above, these programmes are not isolated; they converge on a shared set of effectors that together sustain the radioresistant phenotype (Figure 3). They reinforce redox control, as PPP-derived NADPH together with glutamine- and serine/glycine-derived glutathione buffers the reactive oxygen species generated by irradiation. They support DNA repair, with ribose-5-phosphate sustaining nucleotide synthesis while AR- and FASN-driven signalling enhances the repair of RT-induced damage. Through lipogenesis, they drive membrane remodelling, generating the phospholipids and bioactive lipid signalling molecules characteristic of radioresistant tumours. These activities are coordinated by overlapping regulatory nodes, most prominently AR signalling, a master regulator of glycolytic, lipogenic and redox metabolism in PCa, and HIF-1, which, under the hypoxic conditions prevalent in PCa, redirects metabolism towards glycolysis and the PPP to enable survival under hypoxia. Finally, this metabolic plasticity sustains stemness, supporting the glutamine-dependent, YAP-driven stem-like states associated with radioresistance. Mapped onto the classical radiobiological framework, these EV-propagated programmes act mainly on Repair (enhanced DNA-damage repair), intrinsic Radiosensitivity (ROS buffering), Reoxygenation and hypoxic survival, and Repopulation (via stem-like states), while EV-borne immune-checkpoint cargo (Section 2.2) may additionally impair Reactivation of anti-tumour immunity. EV-mediated regulation of key metabolic pathways and their roles in radioresistance are summarised in Table 1. This regulatory architecture makes it unlikely for radioresistance to be overcome by targeting any single metabolic pathway in isolation, and underscores the need for an integrative framework that accounts for resistance driven by coordinated metabolic remodelling.

Figure 3. Convergence of metabolic pathways on radioresistance in PCa.The glycolytic/pentose phosphate, lipid and amino acid metabolic programmes upregulated after radiotherapy are not independent but converge on a shared set of effectors that sustain the radioresistant phenotype. They are coordinated by two overlapping regulatory nodes: androgen receptor (AR) signalling, a master regulator of glycolytic, lipogenic and redox metabolism in PCa, and HIF-1, which, under hypoxic conditions, redirects metabolism towards glycolysis and the PPP. Through these nodes, the three pathways jointly drive redox control (PPP-, glutamine- and serine/glycine-derived NADPH and glutathione that buffer radiation-induced ROS), DNA repair (nucleotide supply and AR-/FASN-enhanced repair), membrane remodelling (phospholipid and lipid-signalling synthesis), cancer stem-like states (CSC/YAP-driven stemness) and survival under hypoxia, together producing a radioresistant phenotype and supporting the survival of resistant subclones. Abbreviations: ACC, acetyl-CoA carboxylase; ALDOA, aldolase A; AR, androgen receptor; CSC, cancer stem cell; FASN, fatty acid synthase; G6PD, glucose-6-phosphate dehydrogenase; GLS, glutaminase; GSH, glutathione; HIF-1, hypoxia-inducible factor 1; LDHA, lactate dehydrogenase A; NADPH, nicotinamide adenine dinucleotide phosphate; PCa, prostate cancer; PPP, pentose phosphate pathway; ROS, reactive oxygen species; SCD, stearoyl-CoA desaturase; YAP, Yes associated protein. (Created with BioRender.com).

Table 1. Summary of key metabolic pathways, their role in radioresistance, and EV-mediated regulation. 6 Rs of radiotherapy response: Repair (of sublethal DNA damage), Redistribution (within the cell cycle), Repopulation, Reoxygenation, intrinsic Radiosensitivity, and Reactivation of anti-tumour immunity. Abbreviations: AKT, protein kinase B; ALDH3A1, aldehyde dehydrogenase 3 family member A1; ALDOA, aldolase A; AR, androgen receptor; CAF, cancer-associated fibroblast; circMYC, circular MYC; CSC, cancer stem cell; EV, extracellular vesicle; GLUT1, glucose transporter 1; HMG-CoA, 3-hydroxy-3-methylglutaryl-coenzyme A; LINC01614, long intergenic non-protein coding RNA 1614; LMP1, latent membrane protein 1; MAPK, mitogen-activated protein kinase; MGMT, O6-methylguanine-DNA methyltransferase; NF-κB, nuclear factor kappa B; PI3K, phosphoinositide 3-kinase; RT, radiotherapy; SLC1A5, solute carrier family 1 member 5; SLC2A1, solute carrier family 2 member 1; SLC7A5, solute carrier family 7 member 5; SLC38A2, solute carrier family 38 member 2; SREBP1, sterol regulatory element-binding transcription factor 1; YAP, Yes associated protein 1. (please cite/mention Table 1 in the main text where it was first discussed)

Such a framework must also consider the intratumoural heterogeneity in the context of PCa. When the tumour is treated with RT, irradiation targets a spatially structured ecosystem in which subpopulations, for example, tumour cells and CSCs, stromal cells, differ substantially in their baseline metabolic state and adaptive capacity. Under these conditions, EVs assume their greatest potential significance: by packaging and transferring glycolytic enzymes, lipid metabolic regulators, amino acid transporters, and the transcriptional and epigenetic effectors that coordinate their expression, EVs released from radioresistant subpopulations may extend adaptive metabolic capacity beyond their cell of origin and disseminate resistance across the whole tumour ecosystem.

The EV-specific mechanistic evidence above is drawn predominantly from other tumour types and is largely correlative; direct evidence in irradiated PCa remains sparse. We must also acknowledge that the mechanistic studies cited derive predominantly from tumour types with distinct metabolic dependencies compared to PCa. Furthermore, the evidence is largely correlative, but not causal in nature. The additional intersection of EV-mediated metabolic signalling with tumour plasticity, including lineage switching and acquisition of stem-like features, further complicates this research field. Resolving these questions will require integrated metabolomics in tandem with other multi-omic profiling of EVs derived from irradiated PCa cells, combined with functional recipient-cell assays, in vivo validation and clinical outcome correlation, which remain largely missing from the current literature.

4.1. EV metabolomics in liquid biopsy: capturing dynamic heterogeneity

A significant challenge in precision oncology is that clinically actionable tumour biology is neither spatially uniform nor temporally stable. Molecular profiling of tumours in patients has been shown to improve the selection of personalised cancer treatments, facilitate the detection of treatment resistance, and enable monitoring for tumour relapse [157,158]. Yet the utility of molecular profiling in PCa is complicated by the spatial and temporal heterogeneity that defines the disease. As a result, tissue biopsy, although still the gold standard for diagnosis and pathological assessment, provides only a limited and static representation of a biologically dynamic disease [159]. Moreover, tumour profiles evolve over time, meaning that clinical decisions based on historical biopsy data may be misleading [160,161]. Repeated tissue sampling is also constrained by procedural invasiveness, cost, patient burden, and feasibility.

In response to these limitations, liquid biopsy has emerged as a promising, non-invasive approach within precision medicine. Liquid biopsy involves the analysis of bodily fluids, including blood and urine, to detect molecular biomarkers for early cancer detection, surveillance, and monitoring for post-treatment recurrence [162,163], where tumour biology could be captured in a manner that is serial, systemic and ongoing. This enables continuous monitoring of cancer evolution and treatment response, supporting real-time clinical decision-making [164,165].

As discussed above, EV-derived metabolomics represents a promising yet underexplored analytical strategy for liquid biopsy. Biologically, this may be particularly relevant in the context of RT response: metabolic adaptation to RT can proceed through genetically and transcriptionally diverse mechanisms that converge on shared outputs, which means metabolomic profiling could, in principle, detect early radioresistant signals that may be invisible to mutation- or transcript-based approaches.

Moreover, the capacity of EV metabolomics to capture intratumoral heterogeneity is further enhanced by the composite nature of circulating EV populations. The aggregate EV metabolome, therefore, reflects a systems-level diversity that is inaccessible to single-site tissue biopsy. Longitudinal profiling of this signal could enable the early detection of metabolic shifts associated with the emergence of radioresistant clones prior to biochemical failure or clinical relapse, and may, with validation, offer a possible clinical window for therapeutic personalisation.

A complementary strategy for resolving this heterogeneity is single-EV analysis. Since bulk and aggregate measurements average signals across heterogeneous EV populations, they can obscure rare but uniquely informative subsets, such as tumour- or radioresistant-clone-derived vesicles. Single-EV technologies, including nano-flow cytometry (nFCM) [166,167,168,169,170,171,172,173], single-particle interferometric reflectance imaging (SP-IRIS) (e.g., SP-IRIS/ExoView) [174,175,176,177,178], super-resolution and total internal reflection fluorescence microscopy (TIRF) [179,180,181,182], microfluidic platforms and emerging single-EV proteomic approaches, can phenotype individual vesicles, co-localise multiple cargoes on the same EV and detect rare subpopulations [183]. In PCa, single-EV nFCM analysis was shown to have diagnostic utility in differentiating healthy patients from PCa, by measuring levels of the metalloreductase six-transmembrane epithelial antigen of the prostate 1 (STEAP1), with an AUC of 0.95 [166]. Another study showed that nFCM was able to separate intermediate-risk (Gleason score <8) from high-risk PCa (Gleason score of 8 or above) by measuring PSMA-positive prostate EV levels in plasma [167]. nFCM was also able to distinguish between prostate EVs of exocytic origin versus membrane origin (microvesicles) by co-expression of surface markers CD9, CD63 and CD81 [184]. In NPC, nFCM was able to determine the concentration of 5 surface proteins (LMP1, LMP2A, PD-L1, EGFR and EpCAM) on individual plasma EVs, and, in a combined panel, differentiate NPC from both healthy donors and nasopharyngitis with 96.3% and 83.1% accuracy, respectively, which was significantly superior to the traditional VCA-IgA assay [168]. Another study on intraductal papillary mucinous neoplasms (IPMN) showed that FCM could differentiate invasive from indolent IPMN with 82% specificity and 100% sensitivity by measuring mucin 5AC levels in plasma EVs [170]. In NSCLC, TIFR was able to distinguish good responders from poor responders to immunotherapy by quantifying single-EV PD-1/PD-L1 mRNA with an accuracy of 72.2% and could diagnose NSCLC with an accuracy of 93.2% [179]. In PCa, these approaches could potentially identify and enrich tumour-derived EV subsets associated with metabolic adaptation and radioresistance, improving the specificity of liquid biopsy signals and helping to detect the emergence of radioresistant clones anticipated by our heterogeneity framework. However, within the current literature, single-EV analysis is used mainly for surface marker and particle characterisation; its application to the EV metabolome is constrained by the low abundance of metabolites and limited throughput, cost and standardisation. As such, bulk and single-EV approaches are likely to remain complementary in the near future.

4.2. Clinical biomarker application of EV metabolomics

The translational potential of EV metabolomics in PCa should span three important domains: primary diagnosis and risk stratification, monitoring of therapeutic response and early detection of treatment resistance. The available evidence provides meaningful proof-of-concept across several of these domains. An early proof-of-concept study investigating the diagnostic potential of urinary EV lipid profiles demonstrated that nine lipids were significantly different between PCa patients and healthy controls. Using EV lipid analysis, the groups could be distinguished with 93% sensitivity and 100% specificity [185]. However, the study cohort was small, including only 15 patients and 13 controls. Therefore, larger cohort studies are required to validate these findings. Metabolomic analysis of urinary EVs has also shown promise in differentiating PCa from benign prostatic hyperplasia (BPH). One study reported that EV metabolites reflect metabolic adaptations characteristic of PCa, including changes in phosphatidylcholines, acyl carnitines, citrate and kynurenine [186]. Notably, acyl carnitine, an intermediate in lipid metabolism, was elevated in urinary EVs, suggesting increased β-oxidation of FAs which is a recognised feature of PCa metabolism reprogramming [187]. Furthermore, 3beta-hydroxyandrost-5-en-17-one-3-sulphate, a key precursor in androgen synthesis, was elevated in PCa-derived EVs compared to BPH EVs, further supporting the potential of EV metabolites as non-invasive diagnostic biomarkers [186].

Additional metabolomic studies have identified alterations in other metabolic pathways. One investigation reported reduced glucuronate levels in urinary EVs from PCa patients prior to prostatectomy compared with healthy controls and post-prostatectomy samples [188]. Importantly, these changes were detected in urinary EVs but not in whole urine, showcasing the potential of EV-based analyses to reveal tumour-specific metabolic signatures that may otherwise remain undetected. Another study found that EVs from highly metastatic PC-3 cell lines contain higher levels of n-9 oleic acid and monounsaturated fats than less metastatic LNCaP cell lines [134,135,136]. These lipids are associated with enhanced activity of the enzyme stearoyl-CoA desaturase-1, an enzyme linked to tumour invasiveness. Conversely, PC-3 EVs exhibited lower saturated fatty acids and polyunsaturated fatty acids levels than LNCaP EVs, indicating alterations in lipid signalling pathways associated with PCa aggression [137,138,139,140]. Beyond tumour-derived EVs, stromal cells within the tumour microenvironment may also contribute to metabolic support. PCa-associated fibroblasts have been shown to release EVs containing metabolites such as lipids and TCA-cycle intermediates, which can be utilised by cancer cells to sustain tumour growth under nutrient-deprived conditions [67]. Targeting this metabolic crosstalk between stromal cells and tumour cells may therefore represent a novel therapeutic strategy, as has already been explored through EV-associated miRNAs in other cancer types [189].

Despite growing interest in EV biology, most studies investigating EV-mediated responses to RT have focused on transcriptomic and proteomic cargo, with comparatively little attention given to the EV metabolome. To date, no studies have specifically investigated whether EV-derived metabolites can differentiate RT responses in PCa. Evidence from other tumour types, while requiring cautious extrapolation, is nonetheless instructive regarding the feasibility of this proposed approach. One study on small-cell lung cancer EVs found that RT led to decreased levels of EV metabolites involved in nucleotide metabolism and folate metabolism [190]. Similarly, metabolomic profiling of EVs from rectal cancer patients identified multiple metabolite differences between good and poor responders to RT, including significantly elevated glucose levels in the EVs from poor responders [100]. Increased glucose availability and upregulated glycolysis have been associated with radioresistant phenotype through enhanced DNA repair capacity [86,87]. Conversely, not all studies have reported consistent findings. In head and neck cancers, EV metabolomic analysis detected no significant metabolic alterations following RT, raising questions regarding the reliability of EV metabolomics for assessing RT response [191]. Such inconsistencies represent the challenges of reproducibility and specificity that must be resolved before EV metabolomics can be considered a clinically viable biomarker. A summary of EV metabolomic studies with clinical relevance is provided in Table 2. For EV metabolomics to move beyond exploratory relevance towards clinical utility in PCa, several prerequisites must be met:

• Standardised, reproducible EV isolation and characterisation, reported in line with the Minimal Information for Studies of Extracellular Vesicles (MISEV2023) [192] and the open-source knowledge base EV-TRACK, which centralises data of EV separation and characterisation [193].

• Longitudinal sampling before, during and after RT to capture treatment-induced dynamics

• Link to matched clinical outcomes such as biochemical recurrence, local control and survival.

• Robust metabolite normalisation, for example, to EV particle number, EV protein or sample volume [188].

• Analytical and clinical validation in independent, external cohorts.

Table 2. Overview of EV metabolomic studies with clinical relevance. Abbreviations: BPH, benign prostatic hyperplasia; DNA, deoxyribonucleic acid; dTTP, deoxythymidine triphosphate; EV, extracellular vesicle; MUFA, monounsaturated fatty acid; PCa, prostate cancer; PPP, pentose phosphate pathway; PUFA, polyunsaturated fatty acid; RT, radiotherapy; SFA, saturated fatty acid; TCA, tricarboxylic acid; TME, tumour microenvironment.

Until these conditions are satisfied, EV metabolomics in PCa should be regarded as a promising but pre-clinical, hypothesis-generating approach rather than a near-term clinical tool.

4.3. Therapeutic implications: targeting EV-mediated metabolic crosstalk

The recognition that EVs participate in metabolic adaptation following RT opens potential therapeutic avenues. If the EV-mediated transfer of metabolic enzymes or substrates contributes to radioresistant phenotypes, then targeting these metabolic dependencies may enhance radiosensitivity. The strategies can focus on two domains: pharmacological disruption of EV biogenesis and secretion and exploitation of EVs as precision delivery vehicles.

For example, EVs released by PCa cells have been shown to impair anti-tumour immune responses by upregulating PD-1 and TIM-3 expression in CD8⁺ T cells, thereby reducing their cytotoxic ability. Pharmacological inhibition of EV secretion using the neutral sphingomyelinase inhibitor GW4869 rejuvenated T cell function and subsequently suppressed tumour growth both in vitro and in vivo [194]. Similarly, other studies have reported that EV-associated PD-L1 contributes to PCa progression, and that inhibition of EV biogenesis suppresses tumour growth [195]. EV secretion has also been implicated in therapy resistance. GW4869 and dimethyl amiloride have demonstrated efficacy in reducing cell viability in enzalutamide-resistant PCa cells characterised by elevated EV secretion, with syntaxin-6 knockdown producing concordant reductions in EV release and increased apoptosis, suggesting that EV hypersecretion may itself constitute a targetable vulnerability in therapy-resistant disease [196]. Additional pharmacological strategies targeting EV production have been explored. The natural bacterial metabolite manumycin-A (MA), which impacts RAS/RAF/ERK1/2 by targeting farnesyltransferases (FTases) inhibited EV secretion in CRPC cells. Additionally, tipifarnib, another FTase inhibitor, was also found to disrupt RAS-mediated ERK activation and the endosomal sorting complex required for transport (ESCRT) which is critical in EV biogenesis, cargo sorting and secretion [197], leading to significantly decreased EV secretion in PCa cells [198]. Kholia et al. found that the nuclear translocation of peptidylarginine deiminase (PAD)2 and PAD4 was associated with EV secretion in PCa cells, and the pharmacological inhibition of PAD activity using Cl-amidine reduced EV secretion and increased the sensitivity of PCa cells to chemotherapy [199]. This was supported by another study on PC-3 cells which found that treatment of PC-3 cells with 5-fluorouracil (5-FU) in combination with EV biogenesis inhibitor Cl-amidine significantly enhanced PCa cell apoptosis, resulting in a 62% decrease in viable PC-3 cells compared with treatment with 5-FU alone [200]. Collectively, targeting EV biogenesis or secretion may represent a promising strategy to enhance the efficacy of existing therapies such as chemotherapy or RT.

On the other hand, EVs themselves can be exploited as therapeutic delivery vehicles. Engineered EVs are being explored as carriers for small molecules, RNA therapeutics and metabolic inhibitors, taking advantage of their inherent biocompatibility, stability in circulation and tumour-targeting capabilities. This approach may enable selective delivery of radiosensitising agents to metabolically active tumour subclones. For instance, EVs loaded with the glycolysis inhibitor FX-11, which targets LDH, were shown to significantly inhibit cellular energy metabolism in breast cancer cells and enhance the anti-tumour effects of sonodynamic therapy [201]. Another study found that aptamer-engineered EVs delivering SIRT6-targeting siRNA achieved selective PCa cell targeting and resulted in the inhibition of proliferation and metastasis [202]. Similarly, EVs derived from PCa cells have been used to deliver paclitaxel, achieving efficient intracellular uptake via the endocytic pathway and increased cytotoxicity [203]. In the RT context, this platform could enable the targeted delivery of radiosensitising metabolic inhibitors to the precise tumour subpopulations, such as hypoxic clusters or androgen-independent variants where metabolic radioresistance is most pronounced, while minimising systemic toxicity.

Despite this therapeutic promise, several challenges must be addressed before EV-based strategies can be implemented clinically. The foremost is EV heterogeneity: secreted EVs comprise diverse subpopulations (exosomes, microvesicles and apoptotic bodies) that differ in biogenesis, cargo and function, and preparations are frequently contaminated by co-isolated non-vesicular particles and protein aggregates, which confounds both mechanistic interpretation and reproducible therapeutic formulation. A second challenge is the lack of standardisation: different isolation methods (ultracentrifugation, size-exclusion chromatography, precipitation and immunoaffinity capture) yield EV populations of differing purity and composition, while pre-analytical variables: sample type, collection and anticoagulation protocol, platelet contamination and storage conditions, strongly influence the results [22]. For therapeutic deployment specifically, additional challenges include scalable, reproducible and Good Manufacturing Practices-compliant EV production, efficient and consistent cargo loading, control of biodistribution and off-target uptake, potential immunogenicity, and the absence of standardised potency and quality-control assays. These challenges may be mitigated by adopting harmonised reporting and methodological standards (MISEV2023 [192], reported via EV-TRACK [193]), orthogonal isolation strategies and reference materials, single-EV characterisation to resolve functionally relevant subpopulations and engineering approaches that improve EV targeting and cargo stability, followed by validation in well-controlled preclinical models and, ultimately, prospective multicentre clinical studies within clear regulatory frameworks. Addressing these issues in parallel with the biomarker-development priorities outlined in Section 4.2 will be essential to realise the clinical potential of EV-based approaches in PCa.

Ethics Statement

Not applicable.

Consent for Publication

Not applicable.

Funding

This work is supported by the St George Hospital Cancer Research Trust Fund.

Competing Interests

The authors declare no conflicts of interest.

Author Contributions

K.N. and Q.W. conceived the structure of the manuscript. K.Y. and Q.W. wrote the manuscript and drew the figures. K.N. provided guidance throughout the preparation of this manuscript. R.G., J.B. and K.N. reviewed and revised the manuscript.