Mechanical regulation of extracellular-vesicle activity during tumour progression

Abstract

Extracellular vesicles (EVs) are naturally occurring membrane-bound vesicles secreted by cells. Functionalized with surface-targeting molecules and carrying signalling proteins and nucleic acids as cargo, EVs can rewire pathways and alter biological processes in recipient cells. Tumour-derived EVs play key roles in cancer progression, particularly in metastasis, by promoting tumour-cell invasion and the establishment of pre-metastatic niches. An evolving understanding of EVs in cancer highlights a complex intercellular communication network within and beyond the tumour microenvironment that involves cancer cells and non-cancerous cell types such as fibroblasts and endothelial cells. More recently, EVs have also been recognized for their role in modulating interactions between host and immune cells and in reprogramming the tumour immune microenvironment. In this Review, we discuss EV biogenesis and function in diverse mechanobiological and mechanoimmunological contexts, highlighting how mechanical cues influence EV targeting and activity. The intricate interplay between mechanical forces and EV dynamics contributes to tumour progression and links EVs to key disease hallmarks.

Extracellular vesicles (EVs)—heterogeneous membrane-enclosed structures released by cells — are central to advancing precision and personalised medicine. They offer a means for molecular profiling of disease across different stages^1–3. Initially regarded as cellular waste disposal mechanisms, EVs have since been recognised for their ability to selectively package functional biomolecules within their membranes, or on them, for local and distant signalling, complementing direct cell–cell communication^4,5 (Fig. 1a). EV biogenesis occurs both intracellularly within the endosomal pathway, generating exosomes, and at the plasma membrane, producing ectosomes^4–6. EVs regulate many physiological processes⁷ and are involved in multiple aspects of human disease^1,2,8. Their pathological roles have attracted particular interest in cancer, where an expanding body of evidence implicates EVs in several mechanisms underpinning cancer hallmarks⁹. EVs influence primary tumour progression, immune surveillance, treatment response, and the establishment of metastatic sites¹⁰.

Fig. 1 |. EVs as intercellular messengers.

a, At a systems level, cancer deploys an elaborate intercellular communication network within and beyond the TME between a swath of cell populations, involving not only cancer cells but also endothelial cells, fibroblasts, immune cells and subtypes thereof, to survive and thrive. (1) EVs from both cancer and non-cancer cell subpopulations contribute to pathological crosstalk¹⁰. (2) Physicochemical signals from the TME can regulate EV production. (3) Cells produce intraluminal vesicles (ILVs) through the invagination of the endosomal membrane when early endosomes undergo maturation to form MVEs, which may then be transported to lysosomes for breakdown or back to the plasma membrane for the secretion of ILVs into the surrounding medium. (4) Concurrently, plasma-membrane protrusions towards the extracellular milieu may undergo scission to release a separate set of EVs. (5) EVs can carry a variety of proteins, nucleic acids, lipids and glycans. (6) Considering high variability in the molecular cargo of EVs and a limited understanding of how cargo is sorted into vesicles destined for secretion^4,5, one way of characterizing EVs is on the basis of their size, namely small EVs (sEVs; includes both endosomal and plasma-membrane-derived EVs^6,219) for sizes smaller than 200 nm and large EVs (only plasma-membrane-derived EVs^6,219) for sizes larger than 200 nm. b, EVs are subject to stochastic forces and can be viewed as being guided by systematic cues toward the ‘target’, which can profoundly influence several cancer hallmarks by serving as agonists^{29,66,81,111,234} and antagonists^{136,139,202,219}, depending on the information content of the signalling cargo that they pack in their lumen or carry on their surface. Figure created in BioRender. Parihar, K. (2025) https://BioRender.com/y91c531.

Cellular function, whether normal or pathological, is shaped by physical interactions with the microenvironment¹¹. Tumorigenesis in solid cancers is accompanied by a profoundly remodelled microenvironment¹². Hallmarks of the tumour microenvironment (TME) include altered mechanical properties, such as a stiff extracellular matrix (ECM) and high interstitial fluid pressure (IFP)^13,14 (Supplementary Box 1). These mechanical cues enhance oncogenic signalling, promoting tumour growth, survival and invasion^15,16. In addition to ‘outside-in’ mechanosensing, cancer also exploits mechanical signalling associated with altered cell-surface mechanics, which encompass the plasma membrane, the underlying actin cytoskeleton and their connections. Notably, cancer stemness, invasion and metastatic potential strongly correlate with changes in cell mechanics^17–24.

This Review explores the reciprocal relationship between mechanical forces and EVs in the TME, highlighting how mechanical cues influence EV targeting of cancer hallmarks (Fig. 1b). We examine the mechanistic details of signalling proteins at multiple levels of mechanotransduction pathways and how cell mechanics regulate EV formation, cargo and secretion. We then discuss how secreted EVs contribute to ECM remodelling through fibroblast activation, stimulate mechanical signalling pathways, and induce suppressive signalling in activated T-cells from a mechanoimmunological perspective. We also describe the biophysical factors guiding EV transport through dense ECM and the vasculature to metastatic sites, which may facilitate further cancer-cell dissemination. Finally, we explore the potential of these insights to inform EV-based drug-delivery strategies, and outline future research directions into the interplay between EVs and the TME.

EV production and composition

In addition to altered mechanical properties (Supplementary Box 1), solid tumours are characterized by hypoxia, metabolic stress owing to nutrient deprivation and acidic extracellular conditions (pH 6.0–6.5). The TME influences EV bioactivity by altering EV abundance and, more importantly, the composition and type of EV cargo (Fig. 1a). These external chemical factors modulate EV production quantitatively and qualitatively^25–27, supporting the view that the TME itself plays a central role in EV-mediated crosstalk. Physicochemical signals from the TME likely regulate EV production in a complementary manner to promote tumour progression. To dissect the potential effects of mechanical cues within this complex framework, we discuss how mechanotransduction pathways (Supplementary Box 1) modulate EV formation, cargo and secretion.

Mechanical signals enhance exosome production.

ECM-based mechanical cues modulate EV-mediated cell–cell communication in cancer^28–31. Tumorigenic breast, liver and pancreatic cells increase exosome secretion when cultured on stiffer ECM, with EV cargo enhancing proliferation²⁹, recipient-cell motility²⁸ and pre-metastatic niche formation^30,31.

Mechanical signals sensed by membrane receptors and transduced to downstream biochemical pathways modulate exosome secretion at multiple stages (Table 1, Supplementary Box 2 and Fig. 2a). For instance, Src kinase, a proto-oncogene activated by mechanical stimuli^32,33 (Fig. 2a), localizes both to the plasma membrane and the cytoplasmic surface of late endosomal or late lysosomal (LE/Lys) compartments^34–36, a subset of multivesicular endosomes (MVEs) marked by Rab7 GTPase (ref.³⁷). Negative perturbation of LE/Lys-related proteins depletes active Src from focal adhesions (FAs; Supplementary Box 1), resulting in its accumulation in perinuclear LE/Lys compartments destined for lysosomal degradation^34,36. Given the involvement of Rab7-positive LE/Lys in exosome secretion^38,39 and the role of Src kinase in intraluminal vesicle (ILV) formation^40,41 (Fig. 2a), these findings suggest that mechanically induced Src kinase activation may enhance exosome production via LE/Lys compartments, with LE/Lys mediating Src trafficking to FAs.

Table 1 |. Mechanical cues sensed by membrane receptors can be transduced to activate downstream intracellular biochemical signalling, which can modulate exosome production.

Fig. 2a provides a detailed schematic of the different signalling pathways involved.

Steps in exosome production	Mechanotransduction-based regulation
MVE formation	Mechanically activated Src kinase32,33 can directly or indirectly activate ALIX to induce the formation of ILVs (ref.40,41).
Lysosome dysfunction	PTEN downregulation by outside-in mechanosignalling46 can lead to the sequestration of TFEB in the cytosol and to lysosomal impairment44.
MVE trafficking to plasma membrane	Rab8 activation downstream of FAK-Akt increases MVE transport to the plasma membrane for exosome secretion29. A subset of Rab-7-positive MVEs can undergo directed transport towards focal adhesions36,69. (#)
MVE exocytosis	A localized or global increase in intracellular Ca2+ owing to mechanical cues281–283 can mediate MVE–plasma-membrane fusion4,284,285. (#) The mechanosensitive activation of H1HR (ref.286) can instigate downstream signalling that induces MVE–plasma-membrane fusion287. (#)

^#Supplementary Box 2 provides detailed discussion.

Fig. 2 |. Mechanosignalling can modulate the production of EVs.

a, Components of the mechanotransduction machinery (Supplementary Box 1) can act as links between the ECM properties and the formation and secretion of exosomes. (1) Src kinase activated by mechanical signals, whereby distal cytoplasmic Src, in addition to plasma-membrane localized EVs, is activated by signalling propagating through the cell via both actin filaments and microtubules^32,33, can enhance production of exosomes. (2) Phosphorylation by activated Src promotes the internalization of syndecans at the plasma membrane⁴⁰. At the level of ILV formation, the presence of phosphorylated syndecans enhances syntenin recruitment to endosomes, and syntenin phosphorylation by Src was observed to further increase endosomal budding⁴⁰. Syntenin binds to ALIX, which in turn recruits ESCRT III for the production of ILVs. (3) The activation of Akt downstream of focal adhesions phosphorylates the Guanine exchange factor Rabin8 which can then activate Rab8 GTPase, leading to an increase in the trafficking of MVEs towards the plasma membrane²⁹. (4) The loading of immunosuppressive cargo (PD-L1) onto exosomes is mediated by ERK (ref.⁵³). At the transcriptional level, (5) nuclear translocation of β-catenin instigates MYC-driven miR18a-mediated suppression of PTEN (ref.⁴⁶), which can result in reduced translocation of TFEB, leading to dysfunctional lysosome⁴⁴. (6) A YAP inhibitor attenuates an increase in exosome secretion on stiff ECM (ref.²⁸). (7) A subset of Rab7-positive MVEs can be directed to focal-adhesion sites, mediated by LAMTOR1/p18 subunit of LAMTOR complex and integrin-linked kinase (ILK; ref.³⁶). (8) Mechanosensitive ion channels can provide localized Ca²⁺ ion influx for docking of MVE to the plasma membrane. (9) Mechanically activated H1HR (GPCRs) can also mediate MVE exocytosis via activation of phospholipase C (PLC) and protein kinase C (PKC; ref.²⁸⁷). Supplementary Box 2 provides detailed discussion for (7–9). Connections in the pathways lacking thorough investigations showing clear mechanistic links are indicated by question marks. b, (1) Exosomes secreted by cells undergoing mesenchymal-type migration can act in autocrine fashion, contributing to the directionality and speed of migration^65–67. These EVs, along with others secreted in the retraction fibres (migrasomes²⁸⁸), can also provide paracrine signalling for mediating the migration of follower cells. Invadopodia are sites for exosome secretion that can further contribute to the growth of invadosomes^81–83. (2) The amoeboidal-type migration is accompanied by the release of ectosomes, owing to increased actomyosin contraction mediated by phosphorylation of myosin light chains (MLC; ref.^91,92). Metalloproteases on EVs can further aid in the migration by degradation of collagen in a dense ECM (ref.⁹³). c, Crowding-induced membrane deformation by glycocalyx in concert with (de)stability owing to actin (de)polymerization can lead to the release of ectosomes⁹⁷. Figure created in BioRender. Parihar, K. (2025) https://BioRender.com/s33s906.

At the transcriptional level, nuclear transduction of ECM-stiffness cues by yes-associated protein (YAP) has been implicated in exosome production^28,42 (Fig. 2a), although the precise mechanisms remain unclear. Additionally, cancer cells can impair lysosomal function to enhance exosome secretion^43,44 by suppressing the transcription factor EB (TFEB), a master regulator of lysosomal function⁴⁵. One study identified phosphatase and tensin homologue (PTEN) as a key regulator of TFEB dephosphorylation, enabling its nuclear translocation⁴⁴. PTEN deficiency in cholangiocarcinoma cells promoted proliferation and invasion by enhancing exosome secretion via TFEB-dependent lysosomal dysfunction⁴⁴. Another study showed that ECM stiffness transcriptionally activates MYC, upregulating miR-18a, which represses PTEN both directly, as well as indirectly via homeobox A9 (HOXA9)⁴⁶. These findings suggest that cancer cells exploit a mechanosensitive regulatory axis to impair lysosomal function, hijacking endosomal trafficking mechanisms and increasing exosome production (Fig. 2a).

Beyond increasing exosome secretion, mechanical signals also modulate exosomal cargo. In breast cancer cells, exosomes secreted under high ECM stiffness are enriched in thrombospondin-1, promoting recipient-cell migration2⁸. Given that thrombospondin-1 is a direct YAP-TEAD transcriptional target⁴⁷ (TEAD stands for ‘TEA domain’), YAP may regulate both exosome biogenesis and cargo sorting in response to mechanical stimuli. Exosomal thrombospondin-1 from breast cancer cells has been shown to disrupt endothelial intercellular junctions⁴⁸, suggesting that ECM stiffness may influence TME vasculature to facilitate metastasis^49,50.

A study investigating ECM stiffness and exosome production in cancer cells identified an oncogenic mechanosensitive signalling pathway that increases exosome secretion under stiff ECM conditions²⁹ (Fig. 2a). Proteomic analysis and mouse models indicated that exosomes from stiff ECM are enriched in Jagged1, a key Notch pathway activator⁵¹. Exosome-infusion experiments showed that tumour growth was enhanced when mice received exosomes derived from stiff ECM conditions²⁹. Notably, tumour-derived Jagged1 can help cancer cells evade immune surveillance by increasing macrophage secretion of the CD8⁺ T-cell-inhibitory factors CD14 and CD93 via Notch activation⁵². This suggests that ECM stiffness-driven Jagged1 enrichment in exosomes may promote tumour proliferation and immune evasion.

Exosomes from cancer cells can also suppress CD8⁺ T cells by directly interacting with programmed cell death protein 1 (PD-1). A recent study revealed that the recruitment of programmed death-ligand 1 (PD-L1) onto ILVs is mediated by phosphorylated hepatocyte growth-factor-regulated tyrosine kinase substrate (HRS; ref.⁵³), a key component of the endosomal sorting complex required for transport (ESCRT-0) involved in exosomal cargo sorting⁴. Signalling by the extracellular-signal-regulated kinase (ERK) phosphorylates HRS, enabling PD-L1 enrichment in exosomes⁵³. Given that ERK activation is potentiated by the chronically elevated focal-adhesion kinase (FAK) signalling axis in stiff ECM^54,55, these findings suggest multiple pathways through which mechanically altered TMEs may contribute to immune evasion.

Cell mechanics regulates EV production.

Cells respond to mechanical stimuli by altering their physical properties, such as area, shape, stiffness and tension⁵⁶. These changes are primarily modulated by cytoskeletal organization⁵⁷, which undergoes mechanosensing-based remodelling^58–61. In parallel, plasma-membrane mechanics influence mechanical signalling through alterations in tension and topology, underscoring their role in migration, cell polarity, endocytosis and exocytosis⁶². Cancer cells can exploit these mechanosensitive processes to modulate EV formation.

Cytoskeletal remodelling governs cell migration in response to external mechanical cues, such as ECM stiffness, confinement and substrate organization^63,64. Cancer cells exhibit phenotypic plasticity, switching between mesenchymal and amoeboidal migration to efficiently invade heterogeneous TMEs^63,64. EV secretion at protrusions further enhances cell motility.

In vitro and in vivo studies have indicated that exosome secretion contributes to directional mesenchymal-like motility in cancer cells^65–67 (Fig. 2b). Live-cell imaging of fibrosarcoma cells showed that exosomes are secreted at leading edges (lamellipodia) of migrating cells, and that inhibiting exosome production reduced migration speed and directional persistence. Exosome-bound fibronectin was identified as a key mediator of migration speed^65,66, although the molecular cargo responsible for exosome-mediated chemotaxis remains unidentified. In 3D collagen gels, exosome trails deposited by leader cells can facilitate directional migration of follower cells⁶⁷, which suggests collective migration behaviour⁶⁸.

Targeted exosome secretion at the leading edge of mesenchymal cells may involve LE/Lys adaptor MAPK and mTOR activator (LAMTOR) complex mediated LE/Lys trafficking to FAs (Fig. 2a, Supplementary Box 2). LAMTOR1/p18 subunit silencing reduced the peripheral density of LE/Lys within lamellipodia³⁶, whereas LAMTOR2/3, carried by Rab7-positive LE/Lys, targeted mature FAs and interacted with IQGAP1 to mediate FA turnover⁶⁹. IQGAP1, a regulator of exocytosis through the exocyst complex^70,71, may dissociate from FAs to initiate LE/Lys docking at the plasma membrane for exosome secretion. This mechanism suggests that a subset of LE/Lys trafficked to the leading edge of migrating cells simultaneously facilitates FA disassembly as well as ECM-carrying exosome secretion for the formation of future adhesions.

Metastatic cancer cells exhibit lower plasma-membrane tension than epithelial and low-invasive cancer cells⁷². Increased plasma-membrane tension was linked to reduced recruitment of curvature-inducing BAR-domain proteins, which are required for filopodia formation⁷². The inverted-BAR (I-BAR) domain-containing proteins MIM and IRSp53 drive ectosome secretion by promoting filopodia formation and subsequent scission⁷³. This process may be facilitated by external forces arising in the ECM or blood-capillaries⁷³. Because of the link between increased filopodia formation and malignancy^74,75, these findings suggest that ectosome production may be co-opted by cancer cells. Membrane-tension-mediated vesiculation also influences ILV formation (Box 1). In Supplementary Box 3, we describe biophysical mechanisms generating membrane curvature for EV production.

Box 1 |. Endosomal membrane tension regulates the formation of large EVs.

Core to the membrane’s role as a mechanotransducer is the generation of curvature, which depends on membrane tension and bending rigidity²⁹⁰ (Supplementary Box 3). At the organelle level, endosomal membrane tension can regulate the membrane-remodelling activity of the ESCRT-III protein complex, which is required for ILV formation^291,292. ESCRT-III polymerization on the endosomal membrane, nucleated by ESCRT-I/II and/or ALIX, drives membrane invagination, followed by Vps4-mediated depolymerization and ILV scission²⁹³. Treating cells with hypertonic solutions, which deflates endosomes and reduces endosomal membrane tension, increases ESCRT-III recruitment to the endosomal surface, triggering greater ILV formation²⁹¹. Further in vitro experiments using a reconstituted membrane model showed that the membrane binding and the polymerization rates of ESCRT-III subunits increase as membrane tension decreases²⁹¹. Similar reconstitution experiments showed that ESCRT-III function is more dependent on membrane tension than on the concentration of ESCRT-III components in solution²⁹². These findings underscore the role of endosomal membrane mechanics in driving exosome formation, which is consistent with the broader role of membrane tension in the regulation of the recruitment and self-assembly of curvature-inducing peripheral membrane proteins (Supplementary Box 3). Schematic illustration created in BioRender. Parihar, K. (2025) https://BioRender.com/s21g561.

Studies investigating how membrane tension at the endosomal compartment level is modulated remain limited, largely owing to the inherent difficulty of measuring organelle tension in the cellular environment. However, recent advances in mechanosensitive fluorescent flipper probes^294,295 offer promising tools to overcome these challenges and to address previously intractable questions. From a cancer mechanobiology perspective, given the disruption of plasma-membrane-tension homeostasis to promote invasion⁷², a compelling question is whether cancer cells can enhance exosome production through purely biophysical means by reducing endosomal membrane tension. In this view, endosomal compartments in invasive cancer cells may exhibit mechanical properties analogous to the observed softening of these cells at the tumour periphery^21,24, thereby facilitating exosome formation.

Cancer cells degrade ECM using actin-rich invadopodia, which concentrate ECM-degrading proteases to facilitate migration⁷⁶. Cortactin, a key regulator of invadopodia formation⁷⁷ and late endosomal trafficking^78–80, coordinates MVE docking at invadopodia for exosome secretion^81–83. Inhibiting exosome production reduced invadopodia formation and lifetime, whereas exogenous exosome addition increased both⁸¹. Although the exosomal cargo that promotes invadopodia formation remains unidentified, exosomal matrix metalloproteinases⁸⁴ (MMPs), particularly MT1-MMP, have been implicated^81–83. This evidence supports a reciprocal relationship between exosomes and invadopodia, wherein invadopodia provide docking sites for MVEs, while secreted exosomes stabilise invadopodia and enhance ECM degradation (Fig. 2b).

Amoeboidal cancer cells shed ectosomes, ranging from nanometre-sized microvesicles to micrometre-sized oncosomes (Fig. 2b). Oncosomes carry oncogenic cargo capable of reprogramming recipient stromal cells^85–88, and their shedding is enhanced by the loss of diaphanous-related formin-3 (DIAPH3)⁸⁹, which promotes amoeboidal behaviour⁹⁰. Smaller ectosomes are released at least partly via ADP-ribosylation factor 6 (ARF6)-mediated actomyosin contraction, induced through ERK or RhoA/ROCK signalling^91,92 (Fig. 2b). Amoeboidal cancer cell-derived ectosomes can carry proteolytically active MT1-MMP, which assists migration by degrading densely crosslinked collagen⁹³, expanding pores for invasion⁹⁴. Additionally, Ras association domain-containing protein 1 (RASSF1C) promotes amoeboid migration and RhoA/ROCK-mediated ectosome release, carrying cargo that induces an invasive stem-like phenotype in recipient cells⁹⁵. These findings highlight migratory protrusions as key sites for EV release with autocrine and paracrine functions.

Altered glycosylation is a hallmark of cancer cells⁹⁶. In fact, the overexpression of glycocalyx components such as heparan sulfate proteoglycans, mucins and hyaluronic acid correlates with cancer progression⁹⁶. A study showed that glycocalyx composition shapes membrane curvature, with high mucin density driving curvature formation⁹⁷ (Supplementary Box 3). Plasma-membrane tubules induced via the crowding of the glycocalyx and stabilizing forces from actin fibers can undergo vesiculation upon actin depolymerization, leading to the shedding of ectosomes⁹⁷ (Fig. 2c). Tumour cells with high mucin density secreted more ectosomes⁹⁷. Similarly, melanoma cells overexpressing hyaluronic acid synthase 3 secreted hyaluronic acid-coated ectosomes, which induced epithelial-to-mesenchymal transition and proliferation in recipient cells⁹⁸.

Conversely, glycoproteins also sense membrane curvature⁹⁹. Confocal imaging of nanostructure-induced plasma-membrane deformations and BAR-protein-mediated invaginations revealed that mucin 1 preferentially localizes to negatively curved membranes (protrusions) rather than to positively curved invaginations⁹⁹. Truncating the mucin 1 ectodomain alleviated this curvature preference and increased internalization via endocytosis⁹⁹. This may explain reports that exogenous heparinase, which cleaves heparan sulfate chains, increases exosome formation via the syndecan–syntenin–ALIX pathway¹⁰⁰. By trimming heparan sulfate chains on syndecans, heparinase may reduce steric and electrostatic repulsions at invaginations, promoting ILV formation. Collectively, these findings underscore the importance of glycocalyx composition in EV biogenesis and expand its interplay with mechanical signals^101,102 in tumour progression and invasion.

Secreted EVs in the TME

EVs influence the structural architecture and dynamics of the ECM by delivering membrane-bound MMPs. In this section, we explore how EVs can also contribute to the remodelling of the TME by activating stromal fibroblasts and can mediate mechanical signalling in recipient cells, their transport through the dense TME, and their role in promoting immune evasion via membrane-bound receptors.

EV-induced conversion of stromal fibroblasts into cancer-associated fibroblasts

Cancer-associated fibroblasts (CAFs) contribute to excessive ECM deposition and stiffening¹⁰³ (Supplementary Box 1) and support tumour progression through cytokine secretion and immune modulation^104,105. The TME consists of heterogeneous CAF subpopulations, including highly contractile myofibroblast-like CAFs, inflammatory secretome-producing CAFs, and antigen-presenting CAFs^104–106. Cancer cell-derived EVs play a key role in CAF activation (Fig. 3a).

Fig. 3 |. EVs in the TME and vasculature.

a, EVs secreted by cancer cells induce the activation of stromal fibroblasts into CAFs, with CAF subtype being dependent on the EV cargo. CAFs play a critical role in tumorigenic ECM remodelling via the secretion of ECM components and via strain-stiffening (Supplementary Box 1). This indirect role of EVs further adds to the more direct involvement of EVs in ECM structuring through surface-bound proteases and by being an integral part of the matrix itself. b, The physical properties of EVs and their surface composition along with the physiological characteristics of the TME dictate the transport of EVs in the microenvironment. c, A mechanobiology-based framework of PD-L1 carrying cancer-cell-secreted EVs that initiate downstream signalling pathways through the interaction with PD-1 receptors on the T-cell surface to suppress cytotoxic T-cell activity. Delivering bioactive cargo to elicit a functional response in the recipient cell entails EVs first binding at the plasma membrane. The formation of a synapse-like structure involving mechanosensitive ICAM-1/LFA-1 and pMHC1/TCR bonds could assist in improving overall EV-binding avidity, owing to higher bond strength and the possibility of catch bonds, particularly considering that EVs are themselves interacting with a mechanically strained local TME and thus experience athermal forces. d, TEVs in the vasculature play an important role in the organotropic metastasis observed in solid cancers, wherein they instigate the formation of a pre-metastatic niche. The initial arrest and uptake of TEVs at distant tissue sites would depend on their surface profile and mechanical properties, as well as on hydrodynamic factors and the characteristics of target endothelial cells (Box 2). TEVs have also been suggested to mediate CTC arrest and vascular leakiness. Features currently lacking in in vivo studies that show a clear mechanistic involvement in TEV tissue targeting are indicated by question marks. Figure created in BioRender. Parihar, K. (2025) https://BioRender.com/m25d590.

Transforming growth factor-β (TGFβ) is a well-established mediator of fibroblast-to-myofibroblast conversion via SMAD-dependent signalling, which increases the expression of smooth muscle alpha-actin (αSMA) as well as actomyosin contractility¹⁰⁷. Cancer cells secrete small EVs (sEVs) enriched in TGFβ (sEV-TGFβ), which have been shown to induce myofibroblast differentiation in prostate cancer, gastric cancer and bladder cancer^108–110. Notably, sEV-TGFβ from prostate cancer cells was more potent than soluble TGFβ in activating CAFs, enhancing angiogenesis in vitro and tumour growth in vivo^111,112. The mechanism of sEV-TGFβ delivery likely depends on the cell type: prostate cancer sEVs predominantly carry active surface-bound TGFβ (ref.¹¹¹), whereas bladder cancer sEVs contain TGFβ within the vesicle lumen¹¹⁰. Depending on TGFβ localization, sEVs may activate fibroblasts via receptor–ligand interactions or endosomal signalling¹¹³. Outstanding questions include how sEV-TGFβ activity differs from ECM-bound TGFβ, what determines TGFβ localization in EVs, and how this affects function in recipient fibroblasts.

Beyond TGFβ, other EV cargo modulates CAF activation^88,114–118. Metastatic breast cancer cells activate fibroblasts to increase matrix deposition via transglutaminase-2-enriched ectosomes¹¹⁸. Hepatic stellate cells differentiate into tumour-promoting CAFs upon exposure to miR-21-enriched sEVs from hepatocellular carcinoma cells, which activate PDK1/Akt signalling by downregulating PTEN (ref.¹¹⁹). Similar EV-mediated CAF reprogramming by miRNA cargo has been observed in multiple solid cancers^120–125. Notably, the CAF phenotype induced by EVs may depend on whether the originating cancer cell is metastatic^126,127, highlighting the need for further investigation into how different EV cargo act synergistically to regulate CAF subtypes.

Gain-of-function mutant p53 (GOF mutp53) also drives fibroblast conversion into CAFs (ref.¹²⁸). Cancer cell-derived EVs serve as intermediaries in this process, with GOF mutp53 modulating podocalyxin sorting into sEVs, thereby regulating integrin trafficking in fibroblasts and increasing ECM deposition¹²⁹. Alternatively, GOF mutp53 is directly transferred to fibroblasts via sEVs (ref.¹³⁰). Heat shock protein 90 (HSP90) regulates the sorting of GOF mutp53 into sEVs, and GOF mutp53-expressing fibroblasts exhibit increased αSMA and fibronectin expression along with enhanced cytokine secretion¹³⁰. Increased ECM stiffness promotes HSP90-dependent stability and accumulation of GOF mutp53 through the mevalonate-RhoA-HDAC6 mechanotransduction axis¹³¹, which suggests that external mechanical cues regulate GOF mutp53 packaging in sEVs. Because both CAF activation and heterogeneity are influenced by mechanical signals^103,132,133, this could provide another mechanotransduction-based axis for ECM–cancer-cell–CAF crosstalk.

We have focused the discussion on CAF activation by cancer-cell-derived EVs. Yet CAF-derived EVs also influence cancer cells (Fig. 1a). FAK signalling in CAFs modulates miRNA composition in secreted sEVs, affecting cancer-cell migration¹³⁴. CAF-derived EVs also promote proliferation¹³⁵ and drug resistance in cancer cells¹³⁶. Reviews have discussed the emerging role of CAF-derived EVs in tumour progression^137,138. Given the role of mechanical cues in CAF function^103,132,133, mechanotransduction likely regulates CAF EV production, as indicated by findings that ECM stiffness enhances CAF-derived EV secretion and confers chemoresistance to cancer cells¹³⁹. These findings suggest a complex interplay between CAF activation, EV production and the mechanical properties of the TME.

Beyond CAFs, EV-mediated communication extends to other stromal cells¹⁰. Cancer-cell-derived EVs regulate tumour-associated macrophage (TAM) polarization, influencing cancer progression from initiation to metastasis^140,141. TAMs contribute to immunosuppression and ECM remodelling, both directly via matricellular protein production and matrix degradation, and indirectly by stimulating fibroblasts to promote fibrosis^142–145. TAM phenotype and function are also influenced by ECM stiffness and CAF interactions^146–149.

Mechanical cues from the TME regulate TAM function through EV production. In triple-negative breast cancer, mechanical strain increased pro-tumorigenic macrophages and myeloid-derived suppressor cells by enhancing EV secretion¹⁵⁰. Although extensive data highlight EV contributions to macrophage polarization, the role of functionally polarized TAMs by cancer-cell-derived EVs in ECM remodelling remains unclear, particularly because of TAM heterogeneity in the TME¹⁵¹. Taken together, these findings suggest that cancer-cell-derived EVs, aided by other tumour-derived factors, recruit pro-tumour macrophages and reprogram stromal fibroblasts, which leads to extensive ECM remodelling. This, in turn, influences the production and composition of EVs, establishing a feed-forward loop that perpetuates tumour progression.

Mechanical factors impact EV transport through the ECM.

Most studies on EVs have focused on chemical and physical cues influencing EV production, EV cargo, and the effects of the EVs on recipient cells. However, in the complex cellular environment of the TME, where EVs must traverse a highly remodelled ECM to reach target cells, understanding the role of ECM mechanics in EV transport is critical. Fig. 3b illustrates how mechanical properties of EVs and the ECM modulate EV dissemination within the TME.

Studies unravelling the interplay between ECM mechanics and EV mobility have shown that a stiff ECM with stress relaxation (viscoelastic) properties enhances EV transport in spatially confined environments¹⁵². Using EVs from mouse mesenchymal stromal cells and biomimetic alginate hydrogels, it has been shown that EV diffusion was more rapid in a stiff stress-relaxing matrix than in a soft stress-relaxing matrix, and that a stiff elastic matrix restricted EV movement¹⁵². These findings highlight the role of matrix viscoelasticity in regulating EV transport in confined spaces.

In the TME, increased tissue viscoelasticity has been observed in brain tumours¹⁵³ (relative to brain parenchyma) and in breast cancer¹⁵⁴ and liver cancer¹⁵⁵ (compared to benign lesions), suggesting a favourable microenvironment for EV transport. ECM collagen fibre alignment¹⁵⁶ also directs EV diffusion. Early-stage pre-metastatic breast cancer cells radially aligned ECM fibrils around tumour spheroids, enhancing the outward diffusion of sEVs towards co-cultured fibroblasts, thereby promoting CAF induction¹⁵⁷. CAFs further amplify ECM fibre alignment, facilitating cell migration^158,159. Notably, outward-aligned ECM collagen fibrils in human tumours correlate with poor survival in patients with breast cancer^160,161.

EV transport is also affected by EV-binding ECM components, including collagen¹⁶², laminins¹⁶³ and hyaluronic acid¹⁶⁴, as well as by interstitial fluid flow, which depends on how extracellular fluid interacts with the tumour’s deformable solid matrix^14,163–166. These studies underscore the mechanosensitive nature of EV transport, where specific ECM properties facilitate effective EV dissemination in an otherwise confined tumour environment. Interestingly, EVs escape from stiff stress-relaxing ECM hydrogels more efficiently than liposomes of similar size and lipid content¹⁵². This suggests that EV surface proteins may influence ECM polymer crosslinks to modify the local viscoelastic microenvironment and facilitate transport¹⁶⁷.

The mechanical properties of EVs themselves also impact their transport efficiency. Enhanced EV diffusion in stiff viscoelastic ECM is accompanied by EV deformation¹⁵². The aquaporin-1 (AQP1) water channel, packaged within the EV membrane, modulates EV deformability by regulating water flux. Depletion of AQP1 increases EV rigidity, thereby reducing EV diffusivity in confined spaces¹⁵². Similar rigidity-dependent transport behaviours have been observed for liposomes diffusing through biological hydrogels^168,169 and through multicellular spheroids of pancreatic ductal adenocarcinoma¹⁶⁹ (PDAC). These studies further showed that nanoparticle penetration in biomimetic TME models and in tumours in mice was optimal for particles with intermediate rigidity that transiently transform into ellipsoids. In contrast, stiffer nanoparticles lacked deformability, whereas excessively soft nanoparticles exhibited increased adhesion, impairing transport^168,169. Although EVs have more complex membrane and surface compositions than synthetic nanoparticles and have not been extensively studied in this context, they may similarly exhibit an optimal range of stiffness for efficient transport. Because of the reported variations in EV rigidity across subpopulations^170,171 and the influence of the malignant state of the parental cancer cell on EV stiffness^171–174, further research is needed to elucidate the determinants of EV rigidity and its impact on EV transport within the TME.

EVs as proxy mechanical messengers.

As with intratumoural genomic heterogeneity¹⁷⁵, tumours exhibit spatial heterogeneity in ECM architecture and mechanical properties^176–178. Within this heterogeneous microenvironment, EVs can transfer cargo that influences the mechanotransduction machinery of recipient cells, supporting the idea that EVs act as mediators of the mechanical response. For example, sEVs isolated from prostate cancer cells carry integrins α_vβ₆ and α_vβ₃, which they transfer to other cancer or benign cells, enhancing substrate adhesion and migration^179,180. Although sEVs are generally rich in cell-adhesion molecules¹⁸¹, large EVs from aggressive prostate cancer cells, enriched in α_v integrin, have been shown to transfer adhesion and invasion properties to less aggressive recipient cells by activating the FAK-Akt axis⁸⁷. Similarly, sEVs secreted by breast cancer cells in a stiff ECM, enriched in thrombospondin-1, promote FA formation and FAK phosphorylation in recipient cells²⁸. A comparison of sEVs isolated from tumour tissues and non-tumour-adjacent tissues from 10 patients with PDAC also revealed elevated thrombospondin-1 levels in tumour-derived sEVs¹⁸². Another study identified G-protein-coupled receptor 143 (GPR143) as a mediator of exosomal cargo sorting, including β₄, which stimulates the FAK/Src/ERK pathway in target cells, promoting migration¹⁸³. Additionally, Annexin A6 in CAF-derived sEVs modulates the β1 integrin-FAK-YAP signalling axis in recipient cancer cells by enhancing the surface expression of β₁ integrin. These CAF-secreted sEVs also contribute to cisplatin resistance in gastric cancer cells¹³⁶.

EVs carrying FN on their surface can induce FAK activation in target cells^184–187. FN fibril formation on EV surfaces may facilitate integrin clustering, as indicated by breast-cancer cell-secreted sEVs transporting FN crosslinked by transglutaminase-2, with implications for tumour growth and metastasis^184,188. FN bound to cancer-cell-derived EVs has also been shown to promote anchorage-independent growth by interacting with fibroblast integrins, contributing to the transformation of healthy cells¹⁸⁴. Notably, FN on plasma-derived EVs in breast cancer patient samples has been proposed as a biomarker for the disease¹⁸⁹.

Beyond their role as cargo carriers, matrix-bound EVs have been identified as integral components of the ECM (ref.¹⁹⁰). Label-free multiphoton microscopy of breast cancer tissue revealed enrichment of EVs in stromal regions with dense matrix fibres¹⁹¹. Recent findings suggest that caveolin-1 (Cav-1) regulates the sorting of tenascin-C (TnC), a key matrix adhesion molecule, into CAF-derived sEVs, which can then deposit TnC into the ECM, forming fibrillar networks that promote tumour invasion^192,193. Notably, the secretion of Cav-1-containing sEVs is enhanced under conditions of low pH and hypoxia^194,195, further illustrating the interplay between external physicochemical cues and EV production in a desmoplastic TME. Extensive proteomics analysis of patient tissue explants has revealed increased levels of TnC and of other matrix molecules in tumour-derived sEVs, compared with normal tissue-derived sEVs¹⁸².

Together, these studies support the concept that EVs mediate intercellular mechanical signalling through multiple mechanisms: transferring luminal cargo, facilitating interactions via surface-bound ECM molecules, and integrating into the ECM to promote adhesion and migration within the TME.

EV mechanobiology potentiates immune evasion.

Notable breakthroughs in harnessing antitumour immunity have positioned immunotherapy at the forefront of cancer treatment^196,197. Most immunotherapies approved by the United States Food and Drug Administration (FDA) involve monoclonal antibodies that block immune checkpoint inhibitors, such as cytotoxic T-lymphocyte-associated protein 4 and programmed cell death protein 1 (ref.¹⁹⁸). However, only a minority of patients achieve a durable response to checkpoint blockade therapies^199,200. Tumour-secreted small extracellular vesicles (sEVs) have been implicated in tumour resistance to checkpoint-inhibitor-based immunotherapies²⁰¹, with growing evidence indicating that sEVs carrying programmed death ligand 1 on their surface are critical to immune evasion^202–208. sEV PD-L1 can suppress cytotoxic CD8⁺ T cells by directly binding to PD-1 receptors. Loading of PD-L1 onto exosomes is enhanced by increased ERK signalling in response to stiff ECM (Fig. 2a). Although the molecular mechanisms underlying T-cell inhibition by sEV PD-L1 remain unclear, multiple lines of evidence suggest that interactions between sEVs and T cells may mimic key biomechanical aspects of the immunological synapse between T cells and antigen-presenting cells^209,210 (Fig. 3c).

Recent work has elucidated the mechanisms governing sEV adhesion to CD8⁺ T cells, which is essential for subsequent T-cell suppression²¹¹. PD-L1 co-localizes with intercellular adhesion molecule 1 (ICAM-1) on the surface of sEVs secreted by melanoma and lung and colon cancer cells stimulated with interferon-γ (ref.²¹¹). In vitro and in vivo studies using PD-L1/ICAM-1 knockdown and antibody-based inhibition have shown that receptor–ligand interactions between sEV ICAM-1 and the T-cell receptor (TCR) lymphocyte-function-associated antigen-1 (LFA-1) mediate exosome adhesion to T cells and are necessary for PD-L1-mediated suppression of CD8⁺ T-cell proliferation and cytotoxicity²¹¹. Biomechanically, because of the greater binding affinity of ICAM-1 for LFA-1 than that of PD-L1 for PD-1 (ref.²¹²), this suggests that cancer cells exploit bond-kinetic advantages to enhance inhibitory immune checkpoint signalling. ICAM-1 binding to LFA-1 is well established as a crucial component in immune-synapse formation, strengthening TCR interactions with peptide antigens presented on major histocompatibility complexes (pMHCs)^209,210. Because functional monomeric and dimeric soluble PD-L1 exist in the tumour microenvironment²¹³, future studies systematically comparing the T-cell suppression capabilities of sEV-bound PD-L1 with soluble PD-L1 are warranted. Nevertheless, in situ binding assays have shown that PD-L1–PD-1 interactions are more stable when PD-L1 is membrane-bound rather than soluble²¹². Because PD-L1 on EVs is likely to be multivalent, sEV-bound PD-L1 may be more effective as an immunosuppressant.

Other sEV surface-bound ligands may also contribute to immune evasion. PD-L1 co-expression with MHC class I on gastric-cancer-cell-derived exosomes has been reported to enhance T-cell suppression²⁰⁷. Although the underlying mechanism remains unknown, it may involve the co-engagement of PD-1 and TCR with their respective ligands. PD-1 engagement with PD-L1 in close proximity to TCR has been proposed as an essential step in PD-1-mediated inhibition, which is crucial for preventing excessive immune activation^214–217. Notably, TCR–pMHC interactions are mechanosensitive, with exogenous or endogenous forces inducing conformational changes that modulate bond kinetics and TCR activation²¹⁰. Because of the possibility of sEVs clustering at the T-cell membrane, as observed for nanoparticles²¹⁸, or ECM-scaffolded PD-L1-enriched sEVs at tumour peripheries that inhibit T-cell infiltration⁵³, further investigation into how immune-synapse-like structures form between T cells and sEVs and their downstream immunosuppressive signalling pathways is needed (Fig. 3c). Mechanistic insights into EV–T-cell interactions could also inform the broader physiological and pathological roles of EVs in adaptive immunity, such as antigen presentation by EVs carrying functional pMHC and costimulatory molecules secreted by B cells, dendritic cells and platelets^2,219.

EV mechanobiology in the vasculature

Clinical observations indicate that different cancer types non-randomly target and prime specific organ sites for metastasis, a process known as cancer organotropism²²⁰. Among the proposed mechanisms²²⁰, the primary tumour secretes bioactive factors to remodel potential metastatic sites — creating a favourable ‘soil’, referred to as the pre-metastatic niche²²¹—before circulating tumour cells (CTCs, or ‘seeds’) arrive to initiate metastatic outgrowth. As part of the pro-metastatic tumour secretome, tumour-derived extracellular vesicles (TEVs) play a key role in recruiting bone-marrow-derived and immune cells in circulation while reprogramming resident stromal cells to establish a permissive environment (inflammatory, immunosuppressive and fibrotic) for metastasis^222,223. Analogous to the role of EVs in activating CAFs in the TME, TEVs can remodel distant sites by reprogramming fibroblasts in host tissues¹⁰. For example, PDAC-derived EVs promote fibrotic ECM production by hepatic stellate cells, either indirectly by enhancing TGFβ secretion by Kupffer cells²²⁴ or directly by activating insulin-like growth-factor-1 signalling²²⁵, thereby facilitating liver metastasis. Notably, chemotherapy can alter the composition of cancer-cell-derived EVs, potentially influencing distant ECM remodelling. Breast cancer cells treated with paclitaxel secreted lung-tropic EVs that, upon intravenous injection into mice, induced fibronectin, collagen and MMP9 expression in lung tissue while increasing lung pliability, as measured by rheometry²²⁶.

An important consideration in the intravascular transport of both CTCs and TEVs is the hostile flow environment, which leads to rapid clearance. The half-life of EVs in circulation is only a few minutes²²⁷, whereas that of CTCs is a few hours²²⁸. The majority of CTCs and EVs perish owing to high shear stress, immune-cell attack, and in the case of CTCs, anoikis. Despite this, patients with cancer exhibit elevated EV levels in plasma, which suggests that primary tumours continuously release TEVs into circulation. Those TEVs that arrest at distant organ sites can then initiate local microenvironmental changes. Fig. 3d highlights the biophysical factors influencing TEV dissemination and the mechanisms by which TEVs mediate CTC adhesion in circulation and weaken endothelial intercellular adhesion to promote extravasation.

The biomechanics guiding TEV dissemination.

Alongside CTCs ‘seeding’ distant organ sites, the mechanical hypothesis posits that distinct circulation patterns and vessel architecture influence CTC arrest and metastasis²²⁹. A pivotal study showed a strong correlation between arterial blood flow and the frequency of metastasis in target organs²³⁰, and accumulating evidence over the past three decades suggests that the interplay between two central biomechanical factors — external fluid forces (hemodynamic or lymphatic) and cellular adhesion forces — regulates CTC arrest, survival and extravasation¹⁴. Several CTC-surface adhesion proteins (including CD146, CD44, CD24, ICAM1, mucin 1, and various integrins, such as α_vβ₃, α₆β₄ and β₁ with alternate α-subunits), have been implicated in this process^231–233.

Biodistribution studies using mouse models have shown that the organ-specific uptake of TEVs mirrors the organotropism of parental tumour cells³. Evidence suggests that the repertoire of TEV surface-adhesion proteins dictates their organ-specific homing behaviour²³³. TEVs enriched in α₆β₁ and α₆β₄ integrins preferentially interact with lung fibroblasts and epithelial cells, whereas α_vβ₅-carrying TEVs specifically bind Kupffer cells in the liver²³⁴. Although the mechanisms regulating integrin loading into EVs remain unclear, a recent study using breast cancer xenografts and mouse models of metastases suggests that G-protein-coupled receptor 143 (GPR143) controls the sorting of α_6, β₁ and β₄ in exosomes involved in lung-tropic metastasis¹⁸³. Additionally, TEV surface-bound glycoproteins contribute to pre-metastatic niche formation^225,235. For instance, pancreatic TEVs expressing CD44 reprogram hepatic stellate cells to induce a fibrotic liver pre-metastatic niche²²⁵. The TME itself may also enhance TEV recruitment to pre-metastatic-niche sites3⁰^,31. Exosomes produced by breast cancer cells under high ECM stiffness conditions display increased presentation of surface adhesion proteins, which notably improves vascular retention and biodistribution³⁰. Exosomal miRNA cargo tuned by ECM stiffness in hepatocellular carcinoma induce metabolic changes in lung fibroblasts and enhance vascular permeability during pre-metastatic niche formation in lungs³¹.

TEVs encounter the endothelial layer as the first barrier in their attempt to escape the vasculature and to contribute to the formation of a pre-metastatic niche. The similarities between the adhesion proteins that CTCs use to withstand vascular shear flow and those found on TEVs suggest that TEVs may exploit aspects of CTC intravascular arrest mechanisms. However, the influence of biomechanical factors (such as haemodynamic flow and TEV-endothelial-cell adhesion) on TEV dispersion and uptake remains largely unexplored.

Zebrafish models can be used to track EVs in circulation with high spatiotemporal resolution. Recent work using fluorescently labelled TEVs and endogenously produced EVs in zebrafish embryos revealed that EVs circulate at reduced velocity near vessel walls before undergoing rolling adhesion and subsequent arrest; or, in some cases, a sharp arrest without rolling, although this is likely dependent on the adhesive capacity of individual EVs^236,237. TEVs predominantly arrest in low-blood-flow regions^236,237, a finding supported by studies showing that red-blood-cell-derived EVs preferentially undergo endothelial uptake in low-shear-stress regions in both flow-chamber assays and animal models²³⁸. These observations highlight the role of vascular flow patterns in physiological and pathological EV uptake. Given the enrichment of adhesion proteins on EVs in the plasma of patients with cancer, it is plausible that the mechanical hypothesis of organotropic metastasis extends to TEVs and their role in the formation of a pre-metastatic niche.

The importance of biomechanical cues is further underscored by extensive research on synthetic nanoparticles, which shows that mechanical properties (including shape, size and rigidity) as well as external factors such as haemodynamic conditions and endothelial binding affinity influence systemic delivery and biodistribution (Box 2). For instance, larger and non-spherical nanoparticles exhibit greater margination²³⁹, and softer nanoparticles display longer circulation times owing to reduced phagocytic capture and enhanced escape from splenic filtration²⁴⁰. Notably, cryo-electron microscopy has revealed morphological diversity among EVs from bodily fluids and single-cell-line-derived EVs²⁴¹, although the functional implications remain unknown. Because of the heterogeneity in TEV subpopulations in terms of biophysical properties (particularly size and rigidity)^170–174, and the role of mechanical properties in nanoparticle biodistribution, further research is needed to determine how these physical factors influence TEV contributions to metastatic niches.

Box 2 |. A multiscale framework for capturing EV tissue targeting.

Efforts to overcome the limitations of conventional drug delivery have extensively explored nanotechnology-based approaches, particularly drug-loaded nanoparticles²³⁹. The transport of nanoparticles in the vasculature is governed by hydrodynamic, microscopic and molecular interactions across multiple length and time scales^296,297. These multiscale interactions can be broadly categorized into hydrodynamic forces in blood flow, subcellular binding and uptake, and molecular-scale processes such as antibody engineering and receptor–ligand interactions. Computational models of nanoparticle targeting facilitate the optimization of in vivo experiments by narrowing down the vast design space of nanoparticle properties influencing tissue delivery²⁹⁷.

At the macroscale, nanoparticle interactions with red blood cells dictate the ability of the former to reach the endothelium (margination), a process that is influenced by blood flow rates, viscosity (determined by haematocrit) and vessel diameter²⁹⁸. The physical properties of nanoparticles (including shape, size, rigidity and density) also strongly impact margination dynamics^{298–300,240}. At the endothelial interface, nanoparticle adhesion depends on shear forces from blood flow, on exclusion forces from the glycocalyx and on the strength of receptor–ligand interactions^301,302. Computational models integrating continuum membrane mechanics with coarse-grained molecular-scale representations of nanoparticles and receptors have shown that nanoparticle avidity is governed by ligand density on the nanoparticle surface, by receptor abundance, by membrane bending properties and by entropic contributions from receptor motion (diffusion and flexure) and membrane undulations^{218,303–305}. Schematic illustration created in BioRender. Parihar, K. (2025) https://BioRender.com/w64x991.

Multiscale pharmacokinetic models are advancing beyond traditional nanoparticle-biodistribution estimates based on compartmental analysis^{296,303,306,307}. These models integrate mechanistic interactions of nanoparticles within the vasculature, incorporating physiological factors, specific and non-specific uptake mechanisms, membrane mechanics and the nanoparticle characteristics outlined in the schematic. In the context of TEVs, such integrative models offer a framework for exploring the physicochemical characteristics of EV subpopulations — both known and yet to be identified — that mediate organotropic metastasis.

Yet key questions remain: How do variations in the quantity and cargo of TEVs in response to mechanical and chemical cues from the tumour microenvironment affect TEV biodistribution? How do catch/slip bond characteristics (such as those mediated by integrins in TEV tissue targeting²³⁴) influence TEV binding and uptake in different vascular environments? Because of the role of cytokine binding to glycosaminoglycan side chains on TEVs in lung tropism²⁵³, and owing to the entropic contributions, on nanoparticle uptake, of binding sites on tethered polymers²¹⁸, what is the effect of glycan-related factors (such as glycan length, flexibility and cytokine distribution) on TEV targeting of chemokine-enriched endothelial cells?

These questions can be addressed using multiscale pharmacokinetic models, which can predict how system-level parameters influence TEV distribution and thus provide insights into the role of EVs in cancer dissemination. However, these models must be iteratively refined and validated through experimental data, to ensure that they are both informative and challenged by future experiments. Moreover, mechanistic understanding gained from such models could inform the rational design of EV-based drug-delivery systems.

Advances in imaging strategies for in vivo TEV tracking will further clarify their role in the priming of the pre-metastatic niche. Additionally, computational models developed to study the intravascular behaviour of nanoparticles could provide new insights into the interplay between biomechanical and biochemical parameters that govern TEV dissemination and organotropism, which are factors that are challenging to dissect experimentally (Box 2). Future research delineating the dynamic interactions between vascular architecture, haemodynamic patterns and TEV surface-adhesion proteins will be crucial to understanding the mechanisms that enable TEVs to target specific organs and to prime pre-metastatic niches.

TEVs potentiate the arrest and extravasation of CTCs.

CTCs adhere to endothelial cells through integrin-based binding, with integrin ligands found on the endothelial surface or within ECM deposits associated with the endothelium²³². For example, fibronectin deposits on the luminal side of hepatic blood vessels promote the talin-1-dependent integrin adhesion of CTCs, facilitating colorectal cancer metastasis to the liver²⁴². Notably, these deposits were observed with greater frequency in mouse models with primary colon tumours as compared to control mice²⁴², suggesting that fibronectin accumulation may be influenced by the secretome of primary tumours.

A study using a 3D human liver-on-a-chip model showed that breast-cancer-cell-derived EVs increase fibronectin deposition on the liver vasculature, thereby enhancing CTC adhesion to endothelium^243,244. Specifically, vesicular TGFβ1 induced fibronectin production in liver sinusoidal endothelial cells, and plasma-derived EVs from patients with liver metastases from triple-negative breast cancer were enriched in TGFβ1 (ref.²⁴⁴). Liver-chip models treated with EVs from patients with triple-negative breast cancer with liver metastases showed greater adhesion of breast cancer cells than cells treated with EVs from patients without metastases or from healthy individuals²⁴⁴. Similarly, pancreatic-cancer-cell adhesion was substantially enhanced following liver-chip treatment with their respective EVs²⁴⁴. Although PDAC, which frequently metastasises to the liver and leverages TEVs for the formation of liver pre-metastatic niches^224,225,245, has not been conclusively shown to rely on TEVs for fibronectin deposition on liver endothelium, this mechanism remains plausible. Although in vivo studies are needed to confirm TEV-induced endothelial modifications at metastatic sites, these findings reinforce the role of TEVs in preparing distant organs for metastasis.

Increased vascular permeability is a defining feature of pre-metastatic niche sites, and tumour-secreted factors have been implicated in vascular-barrier disruption²²¹. A study using mouse models showed that melanoma-derived EVs promote vascular leakiness at distant organ sites²⁴⁶, a finding that has since been supported by further research. Specific TEV cargo taken up by endothelial cells can disrupt tight junctions, facilitating vessel permeability. For example, miR-25–3p in colorectal-cancer-derived EVs downregulates zonula occludens-1 (ZO-1) in endothelial cells, increasing vascular permeability and promoting colorectal cancer metastasis to the liver and lungs in mice²⁴⁷. Similarly, miR-105, miR-638 and miR-27b-3p, enriched in EVs from metastatic breast cancer, from hepatocellular carcinoma and from colorectal cancer cells, respectively, have been shown to enhance vascular permeability by suppressing vascular endothelial cadherin and ZO-1 in recipient endothelial cells^248–250. Because EVs can carry multiple miRNAs, it is likely that several miRNAs act in concert to promote endothelial barrier breakdown via downregulation of cadherins, ZO proteins and other junction-associated proteins²⁵¹.

Chemokine production by cancer cells also contributes to extravasation²³². For example, in addition to recruiting inflammatory monocytes, colon-cancer-derived CC-chemokine ligand 2 (CCL2) interacts with CC-chemokine receptor 2 (CCR2) on endothelial cells to increase vascular permeability in the lung, thereby promoting metastasis²⁵². TEVs can transport CCL2 and other cytokines secreted in the primary TME, as these cytokines selectively bind glycosaminoglycan side chains of proteoglycans (such as syndecans, CD44 and heparan sulfate proteoglycans) on the EV surface²⁵³. Studies using mouse models further indicated that breast cancer TEVs enriched in CCL2 contribute to lung-tropic metastasis²⁵³. These findings suggest that the primary TME optimally loads TEVs to simultaneously increase vascular permeability and promote an inflammatory environment at specific organ sites, thereby priming them for metastasis.

Outlook

Understanding the production and composition of different EV subtypes is essential for developing treatment strategies that target the pathological roles of EVs. In cancer, where progression from initiation to metastasis is driven by oncogenic mutations alongside mechanical and chemical cues from the remodelled TME, EVs contribute to multiple cancer hallmarks. This contribution must be considered in the context of how external signals influence EV generation in both cancer and resident stromal cells. Notably, among isogenic cells, only premalignant and tumorigenic cells show increased exosome secretion on stiff ECM (ref.²⁹). Advancing a mechanotransduction-based framework could also provide critical insights into the role of EVs in therapy resistance. For instance, the contribution of CAFs to chemotherapy resistance in PDAC (ref.²⁵⁴) has been linked to cooperativity between stiff ECM and CAF-derived EV secretion¹³⁹, and post-chemotherapy breast cancer cells produce lung-tropic EVs that prime distant tissue for metastasis²²⁶.

Most studies investigating the role of EVs in tumour progression have relied on 2D cell-culture models. However, dimensionality is an essential physical parameter that affects mechanotransduction²⁵⁵. 3D environments better recapitulate in vivo conditions, as indicated by the requirement for biomimetic 3D models (such as spheroids, organoids and tunable 3D hydrogels) to capture biological behaviours that do not emerge in 2D cultures^255,256. In the context of EV production, recent studies report that 3D conditions increase EV secretion per cell and alter EV cargo^257–260, producing a composition that is more similar to that of patient-plasma-derived EVs^257,259,260. The ECM composition of 3D cultures also influences EV secretion^257,260. Further work is needed to disentangle the biochemical and biophysical cues regulating EV production in 3D environments.

Although most research focuses on the biological effects of EVs on recipient cells, the factors determining EV-uptake mechanisms and cargo fate remain poorly characterized, particularly given the biophysical, biochemical and functional heterogeneity of EVs. Moreover, cell mechanics play a crucial role in the initial docking of EVs onto target cells (Box 2), and the microenvironment strongly influences cellular mechanics⁵⁶. More work is required to elucidate how interactions between EVs, target cells and surrounding ECM modulate EV adhesion and uptake efficiency. For example, EV-mediated CAF activation is enhanced under high ECM stiffness conditions that mimic the breast tumour periphery²⁶¹. Understanding the context and mechanisms underlying EV function could extend their relevance to personalized oncology well beyond their role as biomarkers.

Lymphatic flow is a key component of circulation. Plasma leakage from blood vessels into tissue is transported through the interstitium and drained into lymphatic vessels. Lymph nodes act as filters for lymphatic vessels carrying tissue-derived signals (such as soluble antigens, cytokines and EVs) from upstream peripheral tissues, facilitating adaptive immune responses and immune homeostasis. Primary tumour growth is accompanied by substantial remodelling of tumour-draining lymph nodes (TDLNs)²⁶². The presence of cancer cells in TDLNs is a major negative prognostic indicator for many solid tumours²⁶². Although the factors driving TDLN remodelling for metastasis remain largely unknown, TEVs have emerged as key contributors^262–265. TEVs are enriched in lymph exudate relative to paired plasma in patients with melanoma^266,267, likely owing to convection driven by net interstitial flow from blood capillaries to lymphatics in the TME¹⁶³. TEV interactions with lymphatic endothelial cells (LECs), similar to those with vascular endothelium, are at least partly mediated by surface-bound integrins²⁶⁵. TEVs can induce the upregulation of ICAM-1 on LECs, which arriving tumour cells can then use for adhesion²⁶⁴. TEVs also induce transcriptional changes associated with ECM remodelling in TDLNs (ref.²⁶³), although the functional implications for the formation of a pre-metastatic niche remain unclear. TEVs further contribute to immune suppression in TDLNs, either by directly presenting programmed death-ligand 1 or by upregulating PD-L1 on stromal cells^268,269. These findings highlight the need for further investigations into the mechanobiological factors underlying TEV-mediated TDLN remodelling.

The deformability of EVs is a key mechanical determinant influencing their interactions with the microenvironment, their dispersion through vascular and lymphatic systems, and their adhesion to target cells (Box 2) and uptake by them. Although how parental cells regulate the mechanical properties of different EV subpopulations remains unclear, several studies have reported links between parental cell mechanics and EV mechanics. For example, breast cancer cells with increasing invasiveness show decreasing bending rigidity, and EVs secreted by high-invasive phenotype cells have lower bending rigidity than those from low-invasive cells^173,174. Because of the role of mechanical cues in regulating cancer-cell and stromal-cell properties, these findings suggest that EV mechanics may also be influenced by mechanical signals, potentially via the modulation of the protein-to-lipid ratio in EV membranes²⁷⁰.

EV cargo provides a rich source of biological information for liquid biopsies (Fig. 4, Supplementary Box 4). Analytical tools for profiling tumour-derived EVs and EVs from the TME offer a window into tumour physiology, which presents opportunities for improving cancer diagnosis, prognosis, monitoring and treatment guidance. Although most EV-based liquid-biopsy assays remain in preclinical stages, some are under clinical evaluation for various cancers^271,272. Research into EV-based liquid-biopsy tests is also ongoing for neurodegenerative diseases, infectious diseases, immunological diseases and cardiovascular diseases^1,2,272.

Fig. 4 |. EV-centric precision medicine in oncology.

EVs isolated from patient samples collected non-invasively or minimally invasively provide biomarker information for the molecular profiling of tumours. This information can support tumour detection and treatment planning (Supplementary Box 4). EVs can also serve as drug-delivery vehicles, which would require the careful selection of the EV source, the optimization of large-scale manufacturing and separation processes, the choice of an appropriate drug-loading strategy, and the establishment of quality-control methodologies (Supplementary Box 5). Although EVs offer advantageous properties, limitations distinct from those of established and FDA-approved diagnostic (circulating tumour cells (CTCs) and circulating tumour DNA (ctDNA)) and therapeutic (liposomes) systems hinder clinical translation^272,274,289. Figure created in BioRender. Parihar, K. (2025) https://BioRender.com/s38u927.

EVs are also being explored as drug-delivery vehicles (Fig. 4, Supplementary Box 5), with research progressing rapidly over the past decade²⁷³ and spanning preclinical investigations into treatments for multiple pathological conditions, including cancer (by far the most studied), cardiovascular diseases, inflammatory diseases and neurological diseases²⁷³, with some progressing to early-phase clinical trials²⁷⁴. An emerging body of evidence suggests that the mechanical environment influences the production of therapeutic EVs (Supplementary Box 5), which highlights the need to elucidate fundamental mechanisms governing mechanics-mediated EV biogenesis and cargo sorting to optimise EV production.

The dense ECM network in solid tumours can physically impede drug delivery by increasing interstitial-fluid pressure and limiting vascular access. Similarly, ECM components can hinder T-cell infiltration and thus affect immunotherapy efficacy. Engineered EVs could counteract fibrotic tumour stroma by leveraging the same mechanisms through which EVs promote invasion, ECM degradation and CAF activation. For example, a study demonstrated that EVs targeting myofibroblast-like hepatic stellate cells improved liver function2⁷⁵, suggesting potential applications in reducing fibrosis in hepatocellular carcinoma. In preclinical models, combining anti-PD-L1 therapy with therapeutic EVs carrying ECM-degrading proteins enhanced anti-tumour immune responses^276–278.

The biomechanical properties of therapeutic EVs and their ECM-degrading proteases on their surface influence the ability of EVs to navigate the TME. Soft deformable drug-carrying EVs exhibit enhanced extravasation, tumour accumulation and penetration²⁷⁹. Numerous design strategies have been proposed to improve the tumour targeting and infiltration of therapeutic EVs²⁸⁰. Although bioengineering approaches for modifying EVs have shown promise in preclinical models, further research is needed to elucidate the mechanisms governing EV organotropism, EV transport through dense TME, and selective EV uptake by specific cell populations. A deeper understanding of these processes will be crucial for optimising EVs as drug-delivery carriers for clinical use.

Citation diversity statement

Recent work in several scientific fields has identified biases in citation practices such that papers from women and other minority scholars are under-cited relative to the number of such papers in the field^308–316. Here we sought to proactively consider choosing references that reflect the diversity of the field in thought, form of contribution, gender, race, ethnicity and other factors. First, we obtained the predicted gender of the first and last author of each reference by using databases that store the probability of a first name being carried by a woman^312,317. By this measure and excluding self-citations to the first and last authors of the article, our references contain 11.32% woman(first)/woman(last), 21.88% man/woman, 21.93% woman/man and 44.87% man/man. This method is limited in that the names, pronouns and social-media profiles used to construct the databases may not, in every case, be indicative of gender identity and in that it cannot account for intersex, non-binary or transgender people. Second, we obtained the predicted racial/ethnic category of the first and last author of each reference by databases that store the probability of a first and last name being carried by an author of colour^318,319. By this measure (and excluding self-citations), our references contain 30.14% author of colour(first)/author of colour(last), 12.84% white author/author of colour, 24.01% author of colour/white author, and 33.02% white author/white author. This method is limited in that names and Florida Voter Data to make the predictions may not be indicative of racial/ethnic identity, and in that it cannot account for Indigenous and mixed-race authors or for those who may face differential biases owing to the ambiguous racialization or ethnicization of their names. We look forward to future work that could help us to better understand how to support equitable practices in science.