Phytochemical-Primed Extracellular Vesicles and Plant-Derived Exosome-Like Nanoparticles as Endogenous Nanomedicine Platforms Authors: Yaqin Hou, Liang Wu, Jing Wang, Bin Han, Xiaolei Yu Published: 2026 ABSTRACT AbstractThe mismatch between the low oral bioavailability of many phytochemicals and their broad systemic efficacy remains a central challenge for natural-product pharmacology and nanomedicine. Rather than relying solely on direct drug-target interactions at distant tissues, accumulating evidence suggests that phytochemicals can act at primary interfaces such as the intestinal epithelium, liver, tumor stroma, and immune microenvironment to alter EV biogenesis or cargo in selected models, while effects on tissue homing remain largely proposed. In this review, we discuss these findings from a vesicle-enabled nanomedicine perspective, highlighting how phytochemical-primed host EVs and plant-derived exosome-like nanoparticles (PELNs) may function as biogenic signal carriers that translate local chemical exposure into systemic therapeutic regulation. We summarize reported links between bioactive compounds and EV release, cargo loading, and membrane remodeling, while distinguishing directly demonstrated mechanisms from proposed ones. We further discuss how surface molecules including integrins and tetraspanins shape organotropism, blood-brain barrier transport, tumor pre-metastatic niche formation, and gut-liver or neuro-vascular communication. Special attention is given to the translational value of EVs and PELNs as natural nanocarriers for poorly soluble phytochemicals, as liquid-biopsy biomarkers of treatment response, and as quality-controllable platforms for traditional Chinese medicine (TCM)-derived interventions. Here, “vesicular information” refers to the measurable molecular cargo and surface features of a vesicle preparation. This emerging model does not yet establish that phytochemical-modified EVs explain the bioavailability paradox in all settings.Keywords: extracellular vesicles, plant-derived exosome-like nanoparticles, phytochemicals, nanomedicine, bioavailability, drug delivery Introduction The pharmacological landscape of natural products has long been dominated by a perplexing contradiction known as the bioavailability paradox. This phenomenon reflects a fundamental disconnect between the pharmacokinetic profile of a compound, often characterized by negligible systemic concentrations, and its pharmacodynamic reality, frequently manifested as profound therapeutic efficacy in distant tissues.1 A multitude of phytochemicals, including polyphenols such as curcumin and resveratrol, alkaloids like berberine, and diterpenoids such as triptolide, exhibit robust biological activity in clinical settings and in vivo models of systemic disease.2 These bioactive compounds demonstrate the capability to modulate neuroinflammation, suppress metastatic colonization in distant organs, and regulate metabolic homeostasis across complex physiological axes.3,4 However, classical pharmacokinetic evaluations consistently reveal that these compounds possess poor aqueous solubility, undergo rapid first-pass metabolism, and exhibit limited membrane permeability.5 For instance, curcumin, a hydrophobic polyphenol derived from Curcuma longa, is notoriously unstable in physiological pH and is rapidly conjugated into glucuronides and sulfates by the intestinal epithelium and liver, leaving free curcumin levels in the plasma at nanomolar or undetectable ranges.6–8 Similarly, berberine displays low oral bioavailability due to self-aggregation and P-glycoprotein-mediated efflux, yet it effectively regulates hepatic lipid metabolism and insulin sensitivity.2,9 This discrepancy challenges the classical pharmacological dogma, which predicates therapeutic action on the lock-and-key model—the requirement for the physical presence of a ligand at a receptor site in sufficient concentrations to elicit a response.10 The persistence of therapeutic effects in the absence of measurable plasma drug levels underscores a fundamental pharmacological enigma: How do compounds like curcumin or triptolide exert regulatory effects on distant targets, such as neuronal ceramide levels in the brain or pulmonary metastasis in the lung, despite their barely detectable concentrations in systemic circulation? Emerging evidence necessitates a shift from this linear perspective to a network-centric model. One testable interpretation is that, for some phytochemicals, an important part of their activity may begin at primary interaction sites including the intestinal epithelium, liver, and tumor stroma.11–13 In these tissues, the compound is present in sufficient concentrations to induce a cellular stress response. At these interfaces, cellular stress responses could in principle influence endogenous intercellular communication, including EV biogenesis and cargo sorting; direct compound-specific evidence is considered in the sections below. In this framework, the host organism is not a passive recipient of the drug but an active transducer that converts the phytochemical signal into a vesicular code. Compared with free small molecules, EVs can protect molecular cargo and show context-dependent barrier crossing and cellular uptake; however, these properties depend on vesicle source, preparation, and surface composition.14–17 Therefore, this review aims to consolidate emerging evidence for a vesicle-enabled nanomedicine model of phytochemical action, in which poorly bioavailable compounds reshape host-derived EVs or are themselves delivered through plant-derived exosome-like nanoparticles. We first summarize the molecular mechanisms by which bioactive compounds regulate EV biogenesis, release, and cargo sorting, with emphasis on Rab GTPases, calcium signaling, ESCRT-related trafficking, ceramide metabolism, and RNA-binding proteins. We then examine how phytochemical-primed EVs influence organotropism, barrier transport, and distal microenvironment remodeling, particularly in the gut-liver, tumor-immune, and neuro-vascular axes. Finally, we discuss how this framework can guide EV- and PELN-based drug delivery, diagnostic biomarker development, quality control, and the modernization of TCM-derived interventions for complex systemic diseases. In accordance with MISEV recommendations, we use “EV” as an umbrella term when biogenesis has not been demonstrated and reserve “exosome” for preparations with appropriate evidence of endosomal origin.18 We use plant-derived EV-like nanoparticles (PELNs) for plant preparations unless their isolation and characterization justify a more specific designation.19 Within the framework of this review, endogenous mammalian EVs and PELNs represent complementary routes: phytochemicals may modify the release or cargo of host EVs at primary tissue interfaces, whereas PELNs may themselves carry plant-derived lipids, proteins, and RNAs to host cells. This relationship provides the conceptual bridge between the host-EV and plant-vesicle sections that follow. Most mechanistic studies discussed in this review are preclinical. Accordingly, direct compound-EV relationships are distinguished from proposed relay mechanisms and translational hypotheses. Throughout this review, we distinguish different levels of mechanistic support. We describe a mechanism as experimentally demonstrated when a defined phytochemical exposure is directly linked to a characterized EV preparation or subtype and a functional recipient-cell outcome within a coherent experimental framework. We use “proposed mechanism” when the pathway is biologically plausible but one or more causal steps remain untested. Where conclusions are derived by integrating findings across separate studies, they are presented explicitly as our interpretation of the available evidence rather than as experimentally established molecular mechanisms. From Poor Molecular Bioavailability to Vesicle-Enabled Nanomedicine Exosomes and other small EVs have transcended their initial characterization as cellular waste disposal units to be recognized as nanoscale mediators of paracrine and endocrine signaling. They function as signalosomes, carrying a diverse repertoire of lipids, proteins, mRNAs, and non-coding RNAs (miRNAs, lncRNAs, circRNAs) that reflect the metabolic and genetic state of the parent cell.20 In a nanomedicine context, these vesicles are not only biological messengers but also endogenous carriers with membrane stability, barrier-crossing capacity, and cell-type-dependent uptake. Phytochemical exposure can alter EV release and, in selected models, EV-associated cargo or membrane composition21–23 (Figure 1). In Huh7 hepatocellular carcinoma cells, resveratrol downregulated Rab27a and reduced exosome secretion; these changes were associated with reduced proliferation, migration, and epithelial–mesenchymal transition22 (Figure 1). Whether comparable effects occur in non-malignant hepatocytes or in vivo remains unknown. Similarly, in a lysosomal-storage-disease cell model, curcumin stimulates exosome/microvesicle release through a ceramide-dependent pathway, supporting clearance of intracellular lipid accumulation.23 Phytochemical-primed extracellular vesicle signaling as an endogenous nanomedicine model. Phytochemicals may interact with the gut interface (Left), alter intracellular stress pathways (eg, Nrf2/NF-kappaB), and modulate secretory regulators such as Rab27a in selected models (Middle). The framework proposes that these changes can yield EVs (Right) with altered cargo, such as miRNAs or proteins. These exosomes then function as biogenic signal carriers to transfer regulatory instructions to distant organs. This is a conceptual framework; arrows indicate proposed relationships and do not imply that every link has been established in vivo. Red cross symbols indicate inhibition or blockade of the indicated molecular target or regulatory step. This paradigm redefines the mechanism of action of traditional and plant-derived medicines by suggesting that the host organism can act as a biological nanofactory, converting short-lived phytochemical exposure into more stable vesicular signals. This is particularly relevant for TCM formulas, such as Danggui Buxue Tang (DBT) or Buyang Huanwu Decoction (BHD), which are often described as “tonifying Qi” or “activating blood”. Evidence is formula- and model-specific: BHD has been shown to regulate an exosomal MALAT1/YAP1/HIF-1α axis after cerebral ischemia, whereas DBT promotes reparative macrophage polarization in wound healing without directly establishing an EV cargo mechanism.24,25 Thus, the bioavailability most relevant to systemic efficacy may in some settings be the biodistribution, persistence, and target-cell uptake of modified EVs rather than the plasma concentration of free parent compounds. The interaction between phytochemicals and the host microenvironment is dynamic and bidirectional. Phytochemical and formula exposures can alter stress-response pathways such as Nrf2 and NF-κB in model-specific settings; whether these responses causally determine EV cargo in intestinal or microbial systems remains insufficiently tested26 (Figure 1). If these pathways alter EV machinery, they could influence cargo packaging into multivesicular bodies; however, this sequence remains to be demonstrated for most phytochemical and formula exposures. For example, berberine has low oral bioavailability and affects gut microbial, metabolic, and inflammatory pathways, but direct modulation of intestinal or microbial EV output has not been demonstrated.9 An intestine-to-liver EV relay is therefore a plausible hypothesis, but the cited literature does not yet identify a defined berberine-induced EV population, its cargo, and a causal hepatocyte response.20,27 If validated, such relays could help explain how local phytochemical exposure influences distant tissues, but this interpretation should not be generalized across compounds or organs.20,28 Phytochemical Regulation of Vesicle Biogenesis and Release The precise regulation of exosome biogenesis and secretion is governed by the Rab family of small GTPases, which orchestrate the intracellular trafficking of MVBs to the plasma membrane.29 Among the extensive Rab family, Rab27a and Rab27b have been identified as the master switches for the docking and fusion of MVBs.30 Recent research highlights that phytochemicals can selectively modulate these GTPases, thereby fine-tuning exosomal biogenesis and release. Resveratrol, a stilbenoid found in grapes, has been shown to exert its anti-metastatic effects in hepatocellular carcinoma by specifically downregulating the expression of Rab27a 22. In Huh7 cells, reduced Rab27a-associated exosome secretion was accompanied by lower proliferation, migration, and epithelial-mesenchymal transition, while resveratrol also altered exosomal content.22 Rab27a and Rab27b use effectors such as Slp4 and Slac2b to support MVB docking and fusion31 (Figure 2); whether resveratrol also alters this effector recruitment remains to be established. Whether phytochemicals can increase Rab-dependent EV secretion for aggregate clearance in neurodegenerative disease remains a hypothesis requiring direct testing.29 Molecular targets of phytochemicals in exosome biogenesis and release. The figure contrasts two mechanistic routes relevant to phytochemical studies: the ceramide-dependent pathway (Left), where curcumin promotes lipid-driven vesicle release in a defined glial model, and ESCRT/RNA-sorting machinery (Right), illustrated by hnRNPA2B1 but not yet directly regulated by most phytochemicals. Additionally, Rab GTPases and calcium signaling regulate MVB trafficking and release (Center/Bottom); direct phytochemical targeting of these regulators is model dependent. The final step of exosome release—the fusion of the MVB with the plasma membrane—is fundamentally dependent on intracellular calcium (Ca2+) transients.32 An increase in cytosolic calcium triggers the SNARE complex machinery, facilitating membrane fusion. Alkaloids, a diverse class of phytochemicals including berberine and tetrahydropalmatine, act as potent modulators of these intracellular calcium dynamics. For instance, isoquinoline alkaloids from Glaucium grandiflorum have been demonstrated to interact with calcium handling proteins such as SERCA and IP3 receptors, modulating the endoplasmic reticulum’s calcium stores33 (Figure 2). Small molecules that elevate intracellular calcium can increase EV release, as demonstrated in dendritic cells; whether individual alkaloids suppress or enhance EV release depends on the compound and model.34 Calcium ionophores are established EV-release stimuli, but extrapolating this principle to phytochemical enhancement of MSC EV yield requires direct validation.32 Artemisinin and its derivatives can induce oxidative and calcium-associated stress responses, but a direct link to a defined EV population and recipient-cell function has not yet been demonstrated.35 Although calcium signaling is a plausible route by which phytochemicals may influence EV release, direct evidence linking a defined compound, a defined EV subtype, and a functional recipient-cell outcome remains limited in most settings. Exosome biogenesis proceeds via two distinct but overlapping pathways: the ESCRT (Endosomal sorting complex required for transport)-dependent pathway and the ESCRT-independent, ceramide-mediated pathway.36 Selected phytochemical studies implicate one or more of these pathways, but direct pathway assignment remains model dependent. Curcumin, a polyphenol from turmeric, has been elucidated to stimulate exosome release primarily through the ceramide pathway23 (Figure 2). Mechanistic studies in glial cells reveal that curcumin treatment enhances the activity of serine palmitoyltransferase and ceramide synthase, leading to the accumulation of ceramides at the endosomal membrane. The unique cone-shaped structure of ceramide molecules induces spontaneous negative curvature of the lipid bilayer, facilitating the inward budding of intraluminal vesicles (ILVs) independent of the ESCRT machinery.37 In cell models of lysosomal lipid storage, including Niemann-Pick type C, curcumin-associated vesicle release has been linked to export of excess cholesterol and partial restoration of intracellular lipid homeostasis.38,39 The curcumin evidence above derives principally from a glial lipid-storage model; it should not be generalized to all EV populations or disease contexts without additional in vivo validation. In contrast, the ESCRT-dependent pathway involves a more intricate cascade of protein complexes, comprising the synergistic action of ESCRT-0, -I, -II, and -III, alongside accessory proteins such as ALIX and TSG101. While curcumin favors a lipid-driven mechanism, resveratrol has been reported to alter Rab27a-associated EV release in defined models; this should not be interpreted as direct modulation of the ESCRT complexes themselves. The seminal study by Ostrowski et al31 established the critical rate-limiting role of Rab27 isoforms in the transport of MVBs to the plasma membrane and their subsequent docking and fusion events. This discovery provides a molecular basis for testing how phytochemicals may alter exosome secretion through Rab-dependent, non-lipid pathways. More generally, ESCRT-associated and RNA-binding machineries participate in cargo sorting, but direct phytochemical regulation of these components should not be inferred from upstream signaling alone. Villarroya-Beltri et al40 elucidated that SUMOylation of specific RNA-binding proteins, such as hnRNPA2B1, serves as a decisive signal controlling the loading of miRNAs into exosomes (Figure 2). The ESCRT and RNA-sorting machinery provides a mechanistic framework for phytochemical studies, but compound-specific regulation of these proteins remains insufficiently demonstrated.41 Taken together, these mechanisms show how EV cargo can be regulated; whether phytochemicals act with comparable specificity remains compound- and model-dependent. Accordingly, proposed phytochemical effects on EV biogenesis or cargo sorting should be tested experimentally and should not be inferred solely from upstream signaling changes. Importantly, the direction of EV regulation is not uniform across compounds or models. Curcumin increased exosome/microvesicle release in C6 glial cells with impaired lipid trafficking through a ceramide-dependent mechanism,23 whereas resveratrol reduced Rab27a-associated exosome secretion in Huh7 hepatocellular carcinoma cells.22 In contrast to the Huh7 result, resveratrol stimulated the release of HMGB1-enriched EVs from lung cancer cells in an NSCLC model,42 and an exploratory human pharmacokinetic study of 16 healthy volunteers detected more plasma exosome-containing EVs 10 h after resveratrol intake.43 The human study identified resveratrol metabolites, but not free resveratrol, in an EV-containing fraction; because it did not establish endosomal origin, tissue delivery, or functional mediation, it supports an association rather than a causal vesicular mechanism. Together, these divergent observations argue against a class-wide assumption that a phytochemical consistently increases or suppresses EV secretion; the direction of release, cargo composition, isolation method, and biological consequence must be resolved in each model. Molecular Mechanisms Governing Cargo Sorting and Packaging The therapeutic potency of exosomes is intrinsically linked to their cargo, particularly the regulatory nucleic acids such as miRNAs and lncRNAs.44 The loading of these molecules into exosomes is a highly selective process, not a stochastic one, governed by RNA-binding proteins (RBPs) that recognize specific motifs (EXOmotifs) on the RNA sequences.45 Key players in this machinery include heterogeneous nuclear ribonucleoproteins, specifically hnRNPA2B1, and YBX1.44,45 Natural products can alter EV secretion and miRNA cargo in selected models; for example, bee venom changed exosomal miRNA profiles in Jurkat T cells.46 For example, the SUMOylation of hnRNPA2B1 is a critical prerequisite for its binding to miRNAs containing the GGAG motif and their subsequent sorting into exosomes45 (Figure 3). Resveratrol has been reported to alter exosomal miRNA signaling in neuronal models, but direct regulation of hnRNPA2B1 or YBX1 by resveratrol or quercetin has not been established.47 Mechanisms of phytochemical-induced cargo reprogramming in exosomes. This figure depicts how phytochemicals may influence distinct sorting machineries involved in exosomal cargo loading. The top-left region illustrates RNA-binding protein-mediated nucleic acid sorting, in which RNA-binding proteins such as hnRNPA2B1 and YBX1 represent general mechanisms of selective miRNA packaging; however, direct phytochemical regulation of these proteins remains insufficiently demonstrated. The top-right region shows metabolic enzyme and protein checkpoint loading, where shikonin reduces exosomal PKM2-associated signaling, whereas selective exclusion of PD-L1 by ginsenosides remains a proposed rather than demonstrated mechanism. The bottom-right region illustrates membrane lipid remodeling, in which curcumin-associated ceramide enrichment is linked to vesicle release, whereas lipid-raft-mediated exclusion of EGFR/HER2 by phytochemicals remains proposed. EV-associated non-coding RNAs contribute to disease-specific remodeling in settings including heart failure and colorectal cancer, but compound-specific regulation of YBX1-mediated sorting remains insufficiently demonstrated.48,49 Direct evidence that phytochemicals alter intracellular YBX1 sufficiently to shift exosomal RNA profiles remains limited. Furthermore, under conditions of oxidative stress, which are often targeted by antioxidant phytochemicals, the shuttling of redox-sensitive miRNAs (eg, miR-21, miR-155) into exosomes is tightly controlled by these RBPs.50 Antioxidant phytochemicals could influence EV RNA profiles indirectly through cellular redox state, but this proposed link requires direct RBP and cargo perturbation studies. EV-delivered miRNAs can produce marked phenotypic changes in recipient cells.49 The selective packaging of protein cargo into exosomes is a tightly regulated process determined by post-translational modifications, such as ubiquitination, and by interactions with chaperone complexes including Hsp70.41 This specificity is critical for the horizontal transfer of malignant phenotypes via exosome-mediated intercellular communication. A compelling instance of pharmacological intervention in this sorting machinery is observed with shikonin, a naphthoquinone that targets the glycolytic enzyme pyruvate kinase M2 (PKM2).51 In the context of non-small cell lung cancer, tumor cells do not merely secrete PKM2 passively. Rather, PKM2 can promote vesicle release by phosphorylating SNAP-23, a key SNARE protein involved in membrane fusion; combined with direct NSCLC evidence, these studies support a PKM2-associated mechanism for the effects of shikonin.51–53 The full sequence from PKM2 inhibition to SNAP-23-dependent EV release should nevertheless be interpreted across separate experimental studies rather than as a single universally demonstrated pathway. In NSCLC, shikonin reduced exosomal PKM2-associated signaling and enhanced cisplatin sensitivity (Figure 3). Peinado et al (2012) demonstrated that exosomal MET can educate bone-marrow progenitors and promote metastasis; reducing MET or Rab27A diminished these effects.54 Consequently, the most defensible conclusion is that shikonin suppresses PKM2-associated glycolytic and exosomal signaling and can enhance chemotherapy sensitivity in the tested NSCLC models. The presence of programmed death-ligand 1 (PD-L1) on the surface of tumor-derived exosomes is a potent mechanism of systemic immune suppression.36 Although ginseng-derived vesicles have immunomodulatory activity, direct evidence that purified ginsenosides selectively exclude PD-L1 from tumor exosomes remains insufficient.55 Natural compounds may influence heat-shock-protein-associated cargo, but direct evidence that a defined phytochemical alters Hsp70/90-client interactions to selectively exclude oncogenic proteins remains limited.21 The lipid bilayer of the exosome is not a passive envelope but a bioactive surface that dictates membrane fusion and receptor signaling. The selective incorporation of specific lipid species is regulated and can be phytochemical-responsive in defined models. As previously noted, curcumin treatment leads to a significant enrichment of ceramides within the exosomal membrane23 (Figure 3). Ceramide-rich domains can promote membrane curvature and non-conventional vesicle budding, but their effects on circulating EV stability and recipient-cell fusion remain context dependent.56 Because cholesterol-rich lipid rafts house signaling receptors such as EGFR and HER2, they are a plausible target for phytochemical modulation. CD63 and cholesterol metabolism can influence EV biogenesis and cargo organization, but direct demonstration that a phytochemical excludes EGFR or HER2 from EVs through lipid-raft disruption is still lacking.57 In liver disease, exosomes and lipid metabolism are closely linked, but direct evidence that berberine or another defined compound reprograms hepatocyte-EV FASN cargo remains limited.9,58 Thus, phytochemical-associated EV remodeling may involve nucleic acids, proteins, and lipids, although evidence strength is uneven across cargo classes and models. Representative phytochemicals and EV-related evidence, including EV biogenesis or secretion, altered cargo or pathways, target models, therapeutic outcomes, and key evidence qualifications, are summarized in Table 1. Table 1Representative Phytochemicals and EV-Related Evidence Discussed in This ReviewCompound/ClassEV Biogenesis or SecretionAltered Cargo/PathwayTarget Tissue/ModelReported Therapeutic OutcomeEvidence Level/Key QualificationReferencesResveratrolReduced Rab27a-associated EV secretion in HCC; increased EV release in NSCLC and an exploratory human studyRab27a; lncRNA SNHG29; HMGB1-enriched EVs; resveratrol metabolitesLiver/HCC; lung/NSCLC; human plasmaSuppressed malignant phenotypes in HCC; reduced platelet activation in NSCLC; no therapeutic outcome established in the human studyPreclinical plus exploratory human evidence; direction is model dependent[22,42,43]CurcuminIncreased exosome/microvesicle releaseCeramide synthesisC6 glial lipid-storage modelReduced endolysosomal lipid accumulationCell model; not yet generalizable across EV types or diseases[23,38,39]ShikoninReduced EV secretion and exosomal PKM2PKM2-SNAP-23-associated glycolytic signalingLung/NSCLC cells and mouse modelSuppressed EV-mediated glycolytic cross-talk and increased cisplatin sensitivityPreclinical; requires replication across tumor types[51–53]BerberineEV-mediated gut-liver relay remains proposedDefined EV source and cargo remain undemonstratedMetabolic and liver-disease settingsHepatoprotective and lipid-lowering effects reported independently of EV causalityPreclinical/indirect; direct source-to-recipient validation is required[9,20,27]DBT / Astragalus-Angelica preparations / ginseng-derived nanoparticlesDBT macrophage-polarization effects; GDNP activityDBT EV cargo undefined; GDNP-associated PI3K/AKT/HIF-1α and TLR4/MyD88/MAPK signalingWound-healing and tumor-immune modelsImproved wound healing; Astragalus-Angelica antitumor immunomodulation reported independently of EV causalityPreclinical; direct EV evidence applies to the ginseng nanoparticle preparation, not all formula-level effects[25,59,60]TriptolideEngineered EV delivery carrierLoaded small moleculeOvary and melanoma modelsEnhanced antitumor activity; toxicity reduction was model dependentPreclinical carrier studies; ovarian TP-Exos retained liver/spleen toxicity, whereas engineered melanoma exosomes showed improved tolerability[61–64] Representative Phytochemicals and EV-Related Evidence Discussed in This Review Determinants of Organotropism and Tissue-Specific Homing One of the most transformative discoveries in exosome biology is the concept of organotropism, which refers to the intrinsic ability of exosomes to target specific organs based on their surface protein composition. Tumor-EV studies summarized in the cited literature indicate that distinct integrin expression patterns can influence biodistribution: exosomes enriched with integrins α6β4 and α6β1 preferentially home to the lung, while those carrying integrin αvβ5 target the liver65 (Figure 4). This targeted accumulation initiates the formation of a pre-metastatic niche by activating resident fibroblasts and immune cells (eg, Kupffer cells) before the arrival of tumor cells. Shikonin and resveratrol can modulate tumor-cell adhesion and metastatic signaling, but direct evidence that they reprogram EV integrin composition and thereby alter organotropism is not yet available.66 Thus, phytochemical disruption of EV-directed pre-metastatic niches remains a testable hypothesis rather than an established mechanism.65 Collectively, tumor-EV organotropism is established, whereas phytochemical reprogramming of this pathway remains to be demonstrated (Figure 4). Phytochemical intervention in exosome-mediated metastasis. (Left) Tumor-EV integrins govern organotropism; whether shikonin or resveratrol can reprogram this surface code remains to be tested. (Middle) The figure depicts the proposed possibility that bioactive compounds may influence EV uptake or barrier crossing. A gastrodin-specific EV mechanism for blood-brain barrier penetration remains hypothetical. (Right) Comparison of the pre-metastatic niche in untreated versus treated states. Triptolide and andrographolide can influence tumor and stromal pathways, but direct inhibition of EV-mediated MMP or VEGF delivery and pre-metastatic-niche formation remains unproven. Integrin-associated organotropism is established in tumor EV models, whereas phytochemical reprogramming of EV organotropism remains to be tested. Beyond integrins, the docking and uptake of exosomes by recipient cells are mediated by tetraspanins (CD63, CD81, CD9) and other surface receptors.67 Whether phytochemicals alter EV tetraspanin density or clustering sufficiently to redirect uptake remains to be established. For example, metabolic regulators can modulate CD63 levels, which in turn alters the intracellular trafficking and secretion of cholesterol-rich exosomes.57 In the context of the central nervous system, the ability of exosomes to traverse the BBB is a critical bottleneck for therapy. EVs are being explored as carriers capable of BBB transport, including for Chinese herbal medicines, but direct evidence that gastrodin modifies EV surface markers to enhance BBB transcytosis is lacking68 (Figure 4). Accordingly, a gastrodin-specific EV transcytosis mechanism should be treated as a hypothesis. Exosome loading can improve delivery of plant phytochemicals, as shown with black-bean extracts, whereas metabolite-driven remodeling of the EV protein corona remains hypothetical.69 The ultimate therapeutic consequence of disrupting organ-specific exosome targeting is the prevention of pre-metastatic niche formation. Tumor-derived exosomes are potent remodelers of the extracellular matrix, inducing vascular leakiness and suppressing local immunity at distant sites.65 Phytochemicals such as triptolide and andrographolide have been shown to inhibit the expression of matrix metalloproteinases (MMPs) and suppress the activation of cancer-associated fibroblasts70 (Figure 4). These extracellular-matrix-directed effects do not by themselves establish inhibition of exosome-mediated MMP or VEGF delivery. Paederia scandens-derived exosome-like nanoparticles have been used to deliver andrographolide in experimental ulcerative colitis; this carrier study should not be interpreted as evidence that andrographolide blocks tumor-derived EV release or liver pre-metastatic-niche formation.71 Separately, DBT attenuated vascular senescence through Nrf2/HO-1 signaling in a chronic intermittent hypoxia model, but that study did not examine neutrophil-derived EV homing.72 Thus, phytochemicals can influence tumor and stromal pathways relevant to EV-mediated disease, but EV-specific causality must be established in each model. Distant Regulation via Inter-Organ Communication Axes The gut-liver axis exemplifies the EV-mediated relay model of phytochemical nanomedicine. Berberine, an isoquinoline alkaloid widely used for metabolic disorders, exhibits poor oral bioavailability yet exerts potent hepatoprotective and lipid-lowering effects.9 One proposed mechanism involves a relay system mediated by exosomes. Berberine clearly affects gut microbial and host metabolic pathways, but direct evidence that it induces a defined intestinal-cell EV population is lacking. A defined EV cargo, source cell, portal-vein biodistribution profile, and hepatocyte response have not yet been demonstrated for berberine. Accordingly, suppression of hepatic lipogenesis or fibrosis through such vesicles should be regarded as a proposed mechanism rather than a demonstrated pathway (Figure 5). Berberine also alters gut microbiota and inflammatory signaling. Causal mediation by microbial EVs and Kupffer-cell uptake remains unproven.27 Berberine can modulate hepatic immune signaling, but attribution of this effect to gut-derived microbial vesicles remains unproven. Direct demonstration that a defined berberine-induced intestinal or microbial EV population transfers a defined cargo to hepatocytes remains an important outstanding requirement. Schematic of phytochemicals regulating distant organs via exosomal signaling axes. (Left) The berberine gut-liver EV relay is presented as a hypothesis requiring direct source, cargo, and recipient-cell validation. (Top) In the tumor-immune axis, shikonin affects exosomal PKM2-associated signaling in NSCLC, whereas DBT alters macrophage polarization in wound healing; a shared EV-mediated macrophage mechanism has not been established. (Right) In the neuro-vascular axis, BHD has been linked to a Cav1-associated exosomal MALAT1/YAP1/HIF-1α axis after cerebral ischemia, but vesicle crossing of the blood-brain barrier has not been established. The axes summarize preclinical observations and proposed EV-mediated relay mechanisms. In the tumor microenvironment (TME), the crosstalk between cancer cells and tumor-associated macrophages is a driver of progression. Cancer cells secrete exosomes that polarize macrophages towards an immunosuppressive M2 phenotype, which supports angiogenesis and tumor growth.73–75 Shikonin and triptolide have been proposed to influence this axis, but EV-specific evidence varies. Shikonin reduced exosomal PKM2-associated signaling and increased cisplatin sensitivity in NSCLC; whether this effect also prevents monocyte-to-M2 polarization was not directly tested53 (Figure 5). Conversely, TCM formulas like DBT have been shown to modulate macrophage polarization in the context of wound healing. Single-cell RNA sequencing reveals that DBT treatment promotes the transition from a pro-inflammatory M1 phenotype to a reparative M2 phenotype in a wound-healing model;25 whether macrophage-derived exosome cargo mediates this effect remains to be tested. A combination of Astragalus membranaceus and Angelica sinensis also showed immunomodulatory antitumor effects in lung cancer, although EV involvement was not tested.59 These studies illustrate that immune modulation and EV mediation must be evaluated separately. Earlier work showed that BHD enhances the angiogenic effects of MSCs after focal cerebral ischemia.76 A more recent study linked BHD-induced angiogenesis to a Cav1-associated exosomal MALAT1/YAP1/HIF-1α signaling axis in vivo and in vitro24 (Figure 5). This evidence supports an exosome-associated pathway, but it does not establish HIF-1α as vesicular cargo or prove that the relevant vesicles cross the BBB. Reviews have proposed EV-based delivery of Chinese herbal medicines across the BBB, but direct evidence that orally administered resveratrol or gastrodin induces intestinal or microbial EVs that reach the brain remains lacking.68 Therapeutic Nanomedicine Implications and Traditional Medicine Integration TCM utilizes complex multi-herb formulas believed to exert systemic regulatory effects, such as “tonifying Qi” and “activating blood”. The vesicle-enabled model offers a molecular framework for translating these concepts into measurable nanomedicine parameters, including vesicle source, cargo composition, organotropism, and functional potency. LWDH has direct exosome evidence, whereas EV mediation remains hypothetical for DBT. DBT, composed of Astragalus membranaceus and Angelica sinensis, has been reported to promote wound healing by modulating macrophage polarization.25 Whether DBT also changes macrophage-derived EV miRNA profiles and signals to fibroblasts remains to be determined25 (Figure 6). Similarly, LWDH alleviates IgA nephropathy by directly altering the function and secretion of mesangial cell-derived exosomes.77 By inhibiting the transfer of pathogenic signals from mesangial cells to podocytes, LWDH preserves the glomerular filtration barrier, explaining its kidney-protecting efficacy (Figure 6). Although EV-based regenerative strategies have been explored in wound healing, their interpretation remains constrained by vesicle-source heterogeneity and unresolved mechanisms.78 The traditional ‘guide drug’ (Yaoyin) concept may be used only as a conceptual analogy for tissue-selective delivery; no direct evidence currently demonstrates that specific herbs alter EV integrin composition or redirect EV organotropism. EVs and plant-derived exosome-like nanoparticles in nanomedicine-oriented traditional medicine applications. (Left) Exosomes as mediators: Host exosomes are modified by herbal formulas to transfer restorative signals (eg, for tissue regeneration and kidney protection). (Top right) Exosomes as active ingredients: Plant-derived nanoparticles such as GDNPs act as natural nanocarriers for small RNAs, lipids, and proteins. (Bottom right) Exosomes as carriers: Therapeutic exosomes can be loaded with phytochemicals such as triptolide to reduce toxicity and improve targeted delivery. The “Guide drug (Yaoyin)/Effect: Altered integrins” element in the left panel represents a conceptual hypothesis for EV targeting; it should not be interpreted as evidence that a specific herb or formula alters EV integrin composition or redirects organotropism. The three panels illustrate distinct concepts and should not be interpreted as evidence that all TCM effects are EV mediated. In addition to modulating host exosomes, plants contain PELNs that can interact with mammalian cells and tissues in preclinical models. These PELNs, present in fresh plant juices and decoctions (eg, ginseng, ginger, and grape), contain bioactive lipids, proteins, and small RNAs that can function in mammalian cells.28,79,80 From the perspective of nanomedicine, PELNs are attractive because they combine edible-source biocompatibility, intrinsic membrane protection, oral stability, and natural interactions with intestinal epithelial cells, macrophages, and gut microbiota. Ginseng-derived preparations have shown model-specific effects, including protection against liver fibrosis and gut dysbiosis and improved diabetic wound healing through macrophage polarization and endothelial restoration60,81 (Figure 6). These PELNs represent a direct delivery system in which the “phytochemical” is partly embodied by the vesicle itself. Their lipid composition, often enriched in phosphatidic acid, facilitates uptake by gut bacteria and intestinal cells, influencing the gut microbiome and host immunity.80 This offers a dual therapeutic strategy: using free phytochemicals to modulate endogenous host EVs and administering PELNs as exogenous natural nanocarriers of plant-derived regulatory molecules. Because plant vesicle preparations are heterogeneous and isolation workflows vary, their nomenclature, purity, and cross-species mechanisms remain active areas of debate. The biocompatibility and targeting capabilities of EVs make them promising candidates for drug delivery. Poorly soluble or toxicity-limited phytochemicals, such as triptolide, curcumin, and paclitaxel, can be loaded into host-derived EVs, engineered exosomes, or PELN-inspired carriers to improve aqueous dispersibility, protect labile cargo from degradation, and enhance biodistribution.61 Triptolide-loaded EV systems have shown model-dependent benefits: ovarian-cancer TP-Exos improved tumor inhibition but retained liver and spleen toxicity, whereas cRGD-engineered exosomes improved melanoma targeting, survival, and systemic tolerability.62,63 These vesicles protect the cargo and facilitate target-cell entry through receptor-mediated uptake, endocytosis, or membrane fusion. Therapeutic precision can be further improved by modulating surface ligands, exploiting natural organotropism, or combining phytochemical priming with ex vivo EV engineering.64 This strategy also offers a molecular interpretation of the TCM principle of “jun-chen-zuo-shi” (monarch, minister, assistant, and guide), where the EV or PELN can function as the assistant/guide carrier that improves delivery of the active phytochemical or vesicular cargo (Figure 6). The major features, strengths, and translational constraints of host EVs, PELNs, engineered exosomes, and liposomes are compared in Table 2. Table 2Comparison of Nanomedicine Platforms Discussed in This ReviewPlatformSourceCargo ControlTargetingMain StrengthKey ConstraintReferencesHost EVsMammalian donor cellsEndogenous and variableContext dependentBiological compatibilityHeterogeneity and source safety[14–18,20,82,83]PELNsPlant preparationsEndogenous and heterogeneousOften gut focusedEdible-source appealNomenclature, purity, mechanism[19,28,55,60,79–81]Engineered exosomesModified donor EVsPartly controllableLigand engineering possibleCustomizable deliveryManufacturing and characterization[14–16,61–64,83,84]LiposomesSynthetic lipidsHighly controllableLigands can be addedDefined composition and scalabilityClearance and biological mimicry[5,8,83,84] Comparison of Nanomedicine Platforms Discussed in This Review Methodological Standards, Quality Control, and Future Directions To rigorously validate the vesicle-enabled nanomedicine model, precise in vivo tracking of EV transfer is essential. Traditional lipophilic dyes (eg, PKH, DiR) are limited by non-specific labeling and short half-lives. The Cre-LoxP system provides a powerful tool for genetically tracing EV transfer by enabling the specific identification of recipient cells that have internalized EVs from a defined donor population.85,86 By utilizing reporter mouse lines (eg, Rosa26-LoxP-Stop-LoxP-GFP) and injecting exosomes containing Cre recombinase (or Cre mRNA) derived from phytochemical-treated cells, researchers can permanently label recipient cells that have internalized the exosomes. This technique allows for the mapping of functional inter-organ communication axes, such as the gut-liver or tumor-lung pathways.85 Lipidomics can characterize spatial and temporal lipid distributions across organelles; complementary EV-specific metabolic or bio-orthogonal labeling methods will be needed to track newly synthesized vesicles during phytochemical exposure.87 Engineered EV platforms are being developed for advanced cargo-loading and gene-editing applications; whether berberine enhances EV uptake or nuclear entry in such systems remains untested.84 Clinical translation relies on the ability to monitor exosome modulation in patients. Liquid biopsy, defined as the non-invasive analysis of disease-relevant analytes in biofluids such as blood, urine, or saliva, can include EVs alongside circulating nucleic acids, cells, proteins, and metabolites.88 Specific EV-associated miRNAs and proteins are being investigated as treatment-response biomarkers in oncology, but this evidence does not yet validate them for phytochemical interventions.89 Rab27a is mechanistically relevant in inflammatory lipid disease, but neither circulating Rab27a nor an EV miRNA shift has been clinically validated as a response biomarker for BHD or berberine.90 Adhering to the minimal information for studies of extracellular vesicles (MISEV) guidelines for isolation and characterization is crucial to ensure the reproducibility and specificity of these biomarkers.91,92 The clinical evidence base remains early. ClinicalTrials.gov lists NCT01294072, an early study of plant exosomes combined with curcumin, but its registry status is unknown and no posted efficacy result establishes a therapeutic benefit. This registry record does not establish phytochemical-primed host EVs or PELNs as approved medicines.93 Two additional cell-derived EV trial records illustrate the exploratory and uncertain nature of early clinical development, but they should not be used to characterize the entire EV therapeutic pipeline. For example, the Phase I ExoFlo study for medically refractory ulcerative colitis (NCT05176366) was terminated, and a Phase I/II bone-marrow MSC-EV study for acute respiratory distress syndrome (NCT05127122) was withdrawn at registry review. These registry records demonstrate continuing exploratory activity, but they do not establish efficacy and should not be extrapolated to plant-derived or phytochemical-primed preparations.94,95 From a regulatory-development perspective, an EV therapeutic requires product-specific evidence of identity, source and donor control, manufacturing consistency, purity and contaminant control, particle and cargo characterization, potency, sterility/endotoxin status, biodistribution, and repeated-dose toxicology. The appropriate regulatory pathway is product and jurisdiction specific; therefore, clinical translation should proceed through controlled investigational development rather than through claims of class-wide efficacy. These general EV principles are particularly relevant to PELNs and phytochemical-primed EVs, but dedicated PELN-specific regulatory guidance and validated potency standards remain limited.18,96 Despite the immense potential, several nanomedicine-specific challenges remain. Differences in marker distribution and cargo composition illustrate that common tetraspanins do not define functionally uniform EV populations.97 Acorus calamus extract has been associated with reduced Rab27a/nSMase2 expression and lower EV secretion, but causal loss-of-function and cargo-specific perturbation remain necessary to distinguish direct phytochemical effects from indirect EV-mediated effects.98 Scalable production of standardized EV therapeutics, phytochemical-primed EVs, or PELNs will also require quality-by-design workflows, including source control, isolation method reporting, particle-to-protein ratio, vesicle size distribution, marker profiling, cargo potency assays, sterility/endotoxin testing, storage stability, and batch-to-batch comparability.18,83,96 For clinical translation, future studies should prioritize quantitative biodistribution, pharmacokinetic comparison of free compound versus vesicular cargo, safety assessment after repeated dosing, and harmonization with MISEV-aligned characterization standards. Deciphering the molecular codes that connect phytochemical stress responses to EV cargo sorting and tissue homing will not only clarify the systemic efficacy of traditional medicines, but also support a new generation of cell-free, naturally derived nanomedicine platforms for complex systemic diseases. Several limitations require explicit emphasis. First, EV heterogeneity means that a preparation may contain multiple vesicle populations with distinct origins and functions. Second, isolation-method-dependent recovery and co-isolation of non-vesicular particles can change apparent cargo and activity. Third, the field lacks broadly validated, mechanism-linked potency assays that permit comparison across batches and laboratories. Fourth, clinical validation remains limited, particularly for phytochemical-primed EVs and PELNs. Finally, reproducibility requires transparent reporting of source material, pre-analytical handling, isolation, characterization, dose normalization, storage, and functional assays. These issues should be addressed alongside particle-to-protein measurements, marker panels, sterility/endotoxin testing, biodistribution, and batch-to-batch comparability.18,83,96 Conclusion Phytochemical-primed EVs and PELNs provide a useful nanomedicine framework for reconsidering how plant-derived interventions may influence intercellular communication. The review identifies two complementary routes: phytochemicals may alter the release or cargo of endogenous host EVs at primary tissue interfaces, whereas PELNs may themselves deliver plant-derived molecular constituents to host cells. Direct preclinical support is strongest for a limited number of compound-model pairs, including resveratrol-associated changes in Rab27a and EV secretion in hepatocellular carcinoma, curcumin-associated ceramide-dependent vesicle release in a glial lipid-storage model, and shikonin-associated reduction of exosomal PKM2 in non-small-cell lung cancer. In contrast, several proposed gut-organ, neurovascular, organotropism, and formula-level EV mechanisms remain associative or require direct source-to-recipient validation. Future studies should therefore combine rigorous EV isolation and characterization with quantitative biodistribution, donor-cell and cargo perturbation, recipient-cell rescue, and comparison of free phytochemical, host-EV, and PELN formulations in the same disease model. For translation, product identity, source control, manufacturing consistency, purity, potency, sterility/endotoxin status, repeated-dose safety, and regulatory-grade clinical evaluation will be decisive. These requirements do not diminish the promise of the field; rather, they define the experiments and development standards needed to determine when phytochemical-associated vesicle strategies can become reproducible, scalable, and clinically credible nanomedicines. Funding Statement This research was funded by the City-school Cooperation Project of Nanchong (No. 20SXQT0047) and Shenzhen Yechenghai charity Foundation (YCH2024-42). Use of Artificial Intelligence Tools During the preparation of this manuscript, the authors would like to acknowledge the use of DeepSeek v4 for language polishing and refinement. The tool was employed solely to assist in improving grammar, spelling, clarity, and sentence flow. In addition, some visual elements in the figures were generated with assistance from ChatGPT. The authors have thoroughly reviewed and refined all content generated by the tool and assume full responsibility for the final version of the manuscript. Data Sharing Statement Data sharing not applicable to this article as no datasets were generated or analyzed during the current study. Author Contributions All authors made a significant contribution to the work reported, whether in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work. Disclosure The authors declare no competing interests in this work. References Associated Data Data Availability Statement Data sharing not applicable to this article as no datasets were generated or analyzed during the current study.