Abstract Extracellular vesicles (EVs) demonstrate immense potential as naturally derived carriers of active therapeutics. To maximize their capacity, it is crucial to develop effective methods for manipulating cargo and ensuring scalability. To address this challenge, we propose that protein-free mRNA granule-like structures, named gene-encoded nanoparticle RNA for inducing superior EVs (GENRISE), can function as active translational sponge and as transient subcellular compartments. The overexpression of proteins in proximity to RNA assemblies stimulates parental cells to release excess exogenous proteins in induced superior EVs (iSEVs). The iSEV system enables the single-module–based enrichment of exogenous cargo in EVs with scalable manufacturing. By harnessing mass-produced iSEVs induced by GENRISEs, encoding an antigenic peptide, we have successfully demonstrated target-specific in vivo cancer immunotherapy. These findings suggest that the emerging iSEV platform shows considerable potential for biomedical applications by enabling the controlled production of cargo-specific EVs. __________________________________________________________________ mRNA granule-like GENRISE drives cells to mass-produce cargo protein–enriched iSEVs, enabling scalable therapeutic applications. INTRODUCTION Cells govern the overall organism by communicating with neighboring or distant cells. This crucial intercellular communication has drawn substantial attention from researchers across the fields of biological and clinical research, striving to decipher these messages and intervene in information transmission to prevent the spread of fatal diseases or aid in the regeneration of damaged tissues ([54]1–[55]3). Extracellular vesicles (EVs), among other carriers, have emerged as highly promising candidates for next-generation therapeutics and diagnostic tools. The utilization of EVs offers numerous advantages, including their ability to carry diverse signaling molecules within a single carrier, their ability to reflect the status of parental cells, and their abundance in nature ([56]4–[57]6). As a result, researchers have focused on engineering the capabilities of EVs to maximize their therapeutic potential. The strategy involves engineering EVs to equip them with organ-specific targeting or evasion capabilities ([58]7, [59]8). Furthermore, EVs can be designed to present antigens ([60]9) or adjuvants ([61]10), enhancing their potential as a vaccine platform. However, harnessing the full potential of EVs presents formidable challenges because of heterogeneity, lack of controllability, and low yield ([62]11, [63]12). Overcoming these obstacles remains a crucial area of focus for further advancement in the field. One of the approaches used in manipulating EV generation involves the induction of chemical and physical stresses through techniques such as ultrasonication, extrusion, detergent treatment, and freeze-thaw processes ([64]13). Such processes can produce a diverse array of cargoes, ranging from small chemicals to high–molecular-weight enzymes ([65]14, [66]15). Nevertheless, these techniques can often lead to cellular damage and adversely affect parental cells, ultimately yielding a population of EVs that might lack physiological relevance. Conversely, advances in genetic engineering, posttranslational modifications, and nucleic acid nanotechnology have considerably broadened the range of cargo options available ([67]16–[68]18). Although these approaches are less destructive, they often involve costly and highly intricate procedures, and the recovery of target EVs may be insufficient. In response to this challenge, we focused on harnessing the inherent properties of RNA and the cellular machinery to induce functional cell–derived vesicles. Within the intricate cellular machinery, RNA molecules perform diverse functions, ranging from regulating gene expression to facilitating molecular transfers, signaling receptors, and catalyzing reactions. The functional capacity of RNAs in natural systems is further expanded by intramolecular conformational changes and posttranscriptional modifications. Moreover, RNAs have been shown to form non–membrane-bound granules, such as stress granules and processing bodies, which assemble translationally inactive messenger ribonucleoprotein particles into unique supramolecular structures distinct from those in the cytoplasmic environment ([69]19). In contrast to the conventional approach of using RNA binding proteins to temporarily repress mRNA processing, we aimed to generate protein-free mRNA granule-like structures that could actively function as translational sponges and transient subcellular compartments. Inspired by the remarkable advancements in RNA technology, which have facilitated the development of self-assembled RNA materials ([70]20, [71]21) and enhanced accessibility of RNA-based therapeutics ([72]18, [73]22–[74]24), we designed and engineered RNA granule-mimicking mRNA condensates, termed gene-encoded nanoparticle RNA for inducing superior EVs (GENRISEs) ([75]Fig. 1A). We hypothesized that GENRISEs could be used to induce the generation of EVs through the highly localized translation activation ([76]Fig. 1B). The overexpressed proteins and peptides in the vicinity of mRNA assemblies are expected to trigger parental cells to release excess exogenous proteins within EVs, potentially giving rise to the synthesis of designer EVs. Fig. 1. Schematic overview of iSEV generation by the GENRISE. [77]Fig. 1. [78]Open in a new tab (A) Schematic illustration of the generation of GENRISE by in situ self-assembly during rolling circle transcription. The isothermal replication of multimeric mRNAs from template DNA, which encodes gene of interest (GOI) and internal ribosomal entry site (IRES), leads to the crowding of RNA strands, resulting in their self-packing. (B) The GENRISE composed of highly compact mRNA strands enables localized dissociation and sustained expression of cargo proteins upon delivery to the parental cells. The parental cells show localized overexpression of cargo proteins in the cytosol, especially near dissociating GENRISE. (C) Upon continuous and reversible translational stress caused by multiribosome recruiting units and protein overexpression, the parental cells shed iSEVs highly enriched with target proteins. RESULTS GENRISE drive the generation of induced EVs GENRISE are designed to contain long-repeat mRNAs, synthesized via isothermal enzymatic replication from template plasmid DNA ([79]Fig. 1A). The internal ribosomal entry site (IRES) sequence is placed upstream of the gene of interest, enabling cap-independent initiation of translation. The isothermal replication of multimeric mRNAs leads to the crowding of RNA strands, resulting in their self-packing. By using the compact structure of GENRISEs and mimicking the characteristics of natural RNA granules without the involvement of RNA binding proteins, we aimed to modulate the cellular translational machinery by inducing spatiotemporally controlled endoplasmic reticulum (ER) stress ([80]Fig. 1, B and C). Upon cellular uptake, GENRISEs are expected to recruit multiple cytosolic ribosomes ([81]Fig. 1C, step 1) and subsequently act as a transient translational sponge that facilitates sustained yet localized overexpression of target proteins by locally disentangling the mRNAs ([82]Fig. 1C, step 2). Localized translation at the ER has been shown to modulate membrane composition, cargo loading, and ER-endosome communication, processes that are tightly linked to the biogenesis of multivesicular bodies and EVs ([83]25–[84]27). Sustained translation near the ER can transiently activate unfolded protein response signaling, which remodels membrane dynamics to endosomal membranes, thereby facilitating intraluminal vesicle budding and exosome release ([85]28, [86]29). On the basis of these insights, we hypothesized that the ER stress caused by GENRISE-induced localized translation triggers the up-regulation of cellular regulatory mechanisms, ultimately leading to the export of locally overexpressed proteins into the extracellular space ([87]Fig. 1C, steps 3 and 4) ([88]30). The GENRISEs were synthesized as dispersed spherical particles, as revealed by atomic force microscopy (AFM) ([89]Fig. 2A and fig. S1). The self-assembled mRNA strands form GENRISEs with a hydrodynamic diameter of ~180 nm ([90]Fig. 2B). In transmission electron microscopy (TEM) analysis, continuous electron density was observed throughout the nanoparticulate structure, which highlights the high packing density of mRNA within these entities ([91]Fig. 2C). Using the densely packed mRNA structure, the introduction of an excess of foreign mRNA localized by GENRISEs resulted in the generation of vesicular structures enriched with the target protein [enhanced green fluorescence protein (EGFP)], followed by its expression in the cytosol ([92]Fig. 2D and fig. S3). Fig. 2. Synthetic profile of the GENRISE-based generation of iSEVs. [93]Fig. 2. [94]Open in a new tab (A and B) A packed spherical structure and size distribution of GENRISEs (100 to 200 nm), revealed by atomic force microscopy (AFM) image and dynamic light scattering analysis. D[h], hydrodynamic diameter. (C) Transmission electron microscopy (TEM) image and TEM-based selected-area electron diffraction (SAED) pattern of GENRISE. (D) HeLa cell generating iSEV–enhanced green fluorescence proteins (EGFPs) at 24 hours after the treatment of GENRISEs encoding EGFP, analyzed by 3D reconstructed confocal microscopy image [4′,6-diamidino-2-phenylindole (DAPI); blue] (also available in movie S1). (E) False-colored SEM images of the parental cells with or without treatment of GENRISEs; blue for the intact cell surface and orange for iSEVs shedding from the cells. (F) Sectioned TEM image of iSEV-shedding cells. The images shown here are also used in fig. S31A. G) The number of EVs isolated from HeLa cells at 24 hours after being left untreated (control) or treated with the GENRISEs (1.5 μg/ml). (H) The concentration of total EVs with different size range [0 to 300 (blue), 300 to 600 (green), and 600 to 900 (red) nm] from control cells (gray) or GENRISE-treated (navy). Inset numbers indicate an average fold increase compared to those in the control. (I) Fluorescence microscopy images of the isolated iSEV-EGFPs after staining with CellVue Claret (red). (J) TEM image of isolated iSEVs and cryo-TEM image of intact iSEV. The TEM image of cup-shaped iSEV shown here is also used in fig. S31C. (K) 3D reconstructed images of the hydrated iSEV. The images shown here are also used in fig. S11. dv, division. (L) Cell-type–independent synthesis of iSEVs revealed by false-colored SEM images of iSEV-releasing parental cells, including HeLa, MDA-MB-231, human dermal fibroblast (HDF), and PC-3 upon treatment of GENRISEs. For HDF cells, images are also presented in fig. S20 as the corresponding unprocessed black and white images. Data represent means ± SD. These vesicular structures enriched with the target protein, induced by GENRISEs, are referred to as induced superior EVs (iSEVs). Scanning electron microscopy (SEM) and TEM images of iSEV-releasing cells further revealed protruding vesicular structures originating from the parental cells, and these protruding structures displayed surface features similar to those of the iSEVs ([95]Fig. 2, E and F). To further validate the release of iSEVs from parental cells, osmium staining was performed for SEM-based energy-dispersive x-ray spectroscopy (EDX). For both untreated and GENRISE-treated cells, the shared elemental profile and the presence of phosphorus and osmium observed upon osmium tetroxide staining confirmed the presence of membranous components on the iSEVs, as osmium is known to bind to proteins and the head region of phospholipids (fig. S4) ([96]31). Furthermore, the temporary increase in stimulation facilitated by GENRISE allowed for scalable synthesis of iSEVs while ensuring that the viability and phenotype of the parental cells remained predominantly unaltered (fig. S5). Of note, a burst release of iSEVs was observed with an excess amount of GENRISEs treated (fig. S5C). To maximize the yield of iSEVs while maintaining cell viability, we used GENRISEs (1.5 μg/ml) for downstream applications. Notably, the morphological characteristics of iSEV-releasing cells are distinctly different from those of apoptotic cells, which exhibit an indistinguishable main cell body (fig. S6A). These findings provide strong evidence that the GENRISE-mediated generation of iSEVs from parental cells can be scaled on the basis of both the concentration of GENRISEs and the treatment duration, all without serious damage to the parental cells. Consistent with the conventional EV collection method, we used differential ultracentrifugation to isolate iSEVs from GENRISE-treated parental cells (fig. S7) ([97]32). In this study, we focused on conducting a comprehensive analysis of large iSEVs that can be likened to microvesicles or large EVs in general (fig. S8) ([98]33). In comparison to smaller iSEVs (with sizes ranging from 30 to 100 nm), larger iSEVs (with sizes ranging from 100 to 1000 nm) are anticipated to provide a larger internal cargo capacity, potentially facilitate improved receptor-ligand interactions due to their smaller curvature, and promote size-dependent preferential uptake by immune cells, thereby presenting potential advantages for immunotherapeutic applications ([99]34). The number of iSEVs isolated from parental cell cultures showed about fivefold increase without compromising cell viability compared to that of the control cells ([100]Fig. 2G and fig. S5B). In this case, both control and GENRISE-treated cells were maintained in the serum-deprived conditions, exacerbating the temporary ER stress ([101]35). However, the disparity in EV yield was more pronounced when the cells were cultured in serum-supplemented medium: ~4000-fold increase in the number of EVs with a size of exceeding 900 nm (fig. S6C). Notably, our strategy is applicable across a broad range of EV sizes, from 100 to 1000 nm. In particular, there was a substantial increase in the number of iSEVs released from parental cells, especially at the larger cutoff size (i.e., more than a 10-fold increase in the number of EVs with sizes exceeding 900 nm) compared to the numbers of EVs released from control cells in serum-deprived medium ([102]Fig. 2H). On the other hand, the mRNA-treated cells as a control did not show a significant increase in EV generation within the given concentration range, highlighting the superiority of GENRISEs over the endogenous expression of mRNA in the parental cells as achieved by transient transfection (fig. S9). The isolated iSEVs exhibited strong fluorescence, providing evidence that the GENRISEs induced increased production and active packing of the target proteins. This observation was further reinforced by the colocalized fluorescence from both the encapsulated target protein, EGFP, and the lipophilic membrane staining reagent ([103]Fig. 2I and fig. S10). The enrichment of target proteins via the iSEV approach did not alter the morphological characteristics typically observed in EVs, as confirmed by TEM ([104]Fig. 2J). The formation of vesicles through phospholipid bilayers was further characterized using cryo-TEM, which revealed a distinct contrast with a thickness of ~6 nm, delineating the internal and external regions of the vesicular structure ([105]36). AFM imaging of hydrated and ruptured iSEV confirmed again the vesicular structure of the iSEVs ([106]Fig. 2K and fig. S11). Intriguingly, the surface charge of iSEVs was not significantly different from that of conventional EVs, suggesting that the lipid bilayer structure remained intact (fig. S12). Furthermore, we assessed the practical storage of iSEVs by tracking concentrations and modal sizes of total isolated iSEVs over a period of 3 weeks at 4°C, confirming the storage stability of the iSEVs (fig. S13). The serum stability of the iSEVs was also evaluated. The iSEVs showed a stable mean size with negligible distribution changes over a 24-hour time course in phosphate-buffered saline (PBS) at physiological temperature (fig. S14). In addition, the mean size of iSEVs did not differ significantly between the culture (10% serum) and blood-mimicking (50% serum) conditions, showing only a slightly broadened distribution. iSEV generation across a variety of cell lines To investigate the cell-type dependency of GENRISE-mediated iSEV synthesis, we evaluated nine different cell types, including primary human dermal fibroblasts (HDFs), human adipose-derived stem cells (hADSCs), human embryonic kidney (HEK) 293 (live cell recordings available in movies S9 and S10), a mouse immune cell line (RAW264.7), and various cancer cell lines (HeLa, MDA-MB-231, B16F10, A549, and PC-3). Cell viability was maintained at more than 90% at 24 hours after treatment with GENRISEs, and proliferation rate remained unchanged (fig. S15). Also, the isolated iSEVs from different cell types did not display distinct physical profiles (fig. S16). Moreover, pseudocolored SEM images of four different cell types treated with GENRISEs or left untreated revealed that the vesicle shedding response was not limited to a particular cell type ([107]Fig. 2L). Across various parental cell types, the number of iSEVs appeared to be a characteristic feature that was primarily dependent on the input GENRISEs. In addition, we observed significant fold increases in iSEV production across different cell lines (figs. S17 to S19). In particular, the non–small cell lung cancer cell line A549 produced a substantial number of iSEVs larger than 900 nm after treatment of GENRISEs, while untreated A549 cells shed a nondetectable population within the designated size range under the given conditions. This result highlights the potential of GENRISEs to scale up iSEV generation in diverse cell types, including cancer cells. The scalability of iSEV generation was further confirmed in HDFs, demonstrating the broad applicability of the GENRISE-based system (fig. S20). Localized translational stimulation inducing target protein–enriched iSEVs For an in-depth investigation of how protein overexpression affects iSEV generation, we used parental cells with a comparably limited cytosolic volume (RAW264.7) ([108]Fig. 3A and fig. S21A). Confocal microscopy analysis of these parental cells revealed distinct distribution of the expressed target proteins in the cytosol depending on the concentration of the GENRISEs. At lower concentrations of GENRISEs, the signal from the target protein was evenly distributed throughout the cytosol. Conversely, higher concentrations led to increased localization of the target proteins in iSEVs ([109]Fig. 3B). These results collectively provide substantial evidence supporting the hypothesis that localized protein overexpression induced by GENRISEs contributes to the synthesis of cargo-customized iSEVs. Fig. 3. In-depth analysis of cargo protein–enriched iSEV release upon treatment of the GENRISEs. [110]Fig. 3. [111]Open in a new tab (A) The distribution of cargo protein (EGFP) revealed by confocal microscopy images of the cells treated with different concentrations of GENRISEs, indicating the increased number of iSEVs encapsulating EGFP in response to GENRISE concentration. Cross-sectional EGFP intensity and the detailed analytical process are outlined in fig. S21 (A and B). (B) Peak fluorescence intensities from the iSEV in the GENRISE-treated cells, indicating increased packing of target proteins in iSEVs with higher concentrations of GENRISEs. au, arbitrary units. (C) Diagonal profiles of confocal images showing the internal vesicular structure containing cargo protein (EGFP) upon treatment of GENRISEs at 12 hours posttreatment (WGA-AF647, red; Hoechst33342, blue; EGFP, green). (D) The captured images from live recordings of HeLa cells treated with GENRISEs for EGFP over 32 hours (complete data available in movies S5 and S6). Red arrowheads indicate the shedding of iSEVs encapsulating EGFP cargo protein. (E) TEM-based 3D tomography analysis of iSEVs originating from HeLa cells upon treatment of GENRISEs, revealing proteins enclosed with membrane. Data are also available in movie S8. The images are also shown in fig. S25A, which provides details on 3D tomographic reconstruction. (F) Cargo encapsulation in iSEVs from HeLa cells treated with the GENRISEs encoding various target proteins (EGFP, DsRed-Express2, and TagBFP). (G) Preconditioning of HeLa cells with 100 pM AuNPs (20 nm) for the generation of AuNP-loaded iSEVs, revealed by dark-field imaging. (H) TEM image indicates that the GENRISE enabled generation of iSEVs incorporated with AuNPs. To investigate the underlying mechanism of iSEV production, we conducted an in-depth analysis of the cellular response following treatment with GENRISEs. At the early stage (3 hours posttreatment), we observed an increased number of internal membranous compartments, primarily in close proximity to the nucleus, by the imaging of GENRISE-treated cells (fig. S22A). Moreover, three-dimensionally (3D) reconstructed confocal images of iSEV-shedding cells revealed an abundance of vesicular structures encapsulating the target protein, both within and near the parental cells while retaining the integrity of the nuclear membrane ([112]Fig. 3C and fig. S22, B to E). The process of iSEV generation followed by target protein overexpression did not disturb cell viability, and real-time cell tracking indicated that mitosis was unaffected ([113]Fig. 3D and fig. S22F). In other words, the parental cells continue to divide, while the number of iSEVs enriched with overexpressed EGFP steadily increased over time (fig. S23 and movies S5 and S6). In coarse-grained molecular dynamics simulations, target protein molecules are also stably encapsulated by a self-assembled vesicle, as also experimentally captured. In this simulation, we have confirmed the self-assembly of stable protein-encapsulating iSEVs, implying the capability of these iSEVs to carry target proteins despite the limitations of simulation size and timescale (fig. S24 and movie S7) ([114]37). Cargo encapsulation was directly observed through 3D structural analysis of iSEV-shedding cells via transmission electron microtomography (TEMT; [115]Fig. 3E and fig. S25). The shared electron density between the intracellular and intravesicular spaces indicated cargo transmission, while the membrane structure surrounding the cargo exhibited complete enclosure of the iSEVs (movie S8). These findings provide evidence of the integrity and cargo-carrying capabilities of iSEVs as a general nanocargo carrier. Treatment with GENRISEs encoding three different target proteins provided evidence of the universal applicability of our system, irrespective of the specific target genes. As a demonstration of this versatility, GENRISEs encoding DsRed-Express2 (red), EGFP (green), and TagBFP (blue) successfully generated iSEVs encapsulating each of the target proteins ([116]Fig. 3F and fig. S21, B to D). By facilitating the large cargo capacity of iSEVs, we also sought to load extra cargo molecules, such as bulky nanoparticles. To assess the material potential of iSEVs, the parental cells were preloaded with gold nanoparticles (AuNPs; ~20 nm) as a representative example of bulky cargo ([117]Fig. 3G). With the validation of the loading of AuNPs, the preloaded cells were fed with the GENRISEs following treatment of AuNPs. The resulting iSEVs showed intact AuNPs loaded internally to the iSEVs ([118]Fig. 3H). Together, cargo availability not only is limited to internally overexpressed protein but also is applicable to externally introduced bulky nanostructures. Of note, the iSEV platform addresses a physical challenge in loading such bulky molecules to the conventional exosomes. To investigate the details in the cellular regulatory machinery at the transcriptome level, mRNA sequencing was conducted on HEK293 cell line after 24 hours of treatment with GENRISEs (a recording of the iSEV-shedding profile of HEK293 cells treated with GENRISEs for 34 hours is shown in movies S9 and S10). Principal component analysis ([119]Fig. 4A) and hierarchical clustering of differentially expressed mRNAs in control and GENRISE-treated cells ([120]Fig. 4B) revealed a distinct set of genes that were differentially regulated upon treatment with GENRISEs. Fig. 4. GENRISE induces ER stress to promote iSEV release. [121]Fig. 4. [122]Open in a new tab (A) Principal components analysis for the transcriptome of HEK293 cells treated with GENRISEs (navy) or left untreated (gray). PC, principal component. (B) Hierarchical clustering of differentially expressed mRNAs in HEK293 cells with or without treatment of GENRISEs. (C) Significantly up-regulated pathways associated with translation (R-HAS-72766.4), assessed by Reactome database. SRP, signal recognition particle. (D) Volcano plot indicating significantly up-regulated or down-regulated mRNA levels in GENRISE-treated cells compared to those in untreated cells. Entities associated with the regulation of ER stress [gene ontology (GO) term 0034976] are highlighted. (E) Chemical structure of ISRIB and its mechanism of action in impairing ER homeostasis. PKR, protein kinase R; PERK, protein kinase R-like endoplasmic reticulum kinase; GCN2, general control nonderepressible 2; HRI, heme-regulated inhibitor. (F) Total EV concentration changes in GENRISE-treated HeLa cells (1.5 μg/ml for 24 hours) with or without ISRIB treatment (1 μM). (G) Chemical structure of pantethine and its mechanism of action in inhibiting microvesicle budding. (H) Concentrations of total EV and (I) EVs with sizes larger than 300 nm from HeLa cells treated with GENRISE (1.5 μg/ml for 24 hours) with or without pantethine (200 nM) treatment. n.s., not significant; **P < 0.01 by unpaired t test. Specifically, a total of 826 mRNAs were differentially regulated, 487 up-regulated and 339 down-regulated, in GENRISE-treated cells compared to those in control cells ([123]Fig. 4, C and D). Pathways associated with translational machinery (R-has-72766.4) were found to be significantly up-regulated (P = 0.0435) according to Reactome-based pathway enrichment analysis ([124]Fig. 4C) ([125]38). Despite the transient nature of mRNA, which has a half-life of a few hours ([126]39, [127]40), translation initiation, elongation, and termination were significantly up-regulated at 24 hours posttreatment, indicating that GENRISEs had a long-lasting effect. A group of mRNAs associated with the regulation of ER stress [gene ontology term, GO0034976; ([128]41)] was also significantly up-regulated, providing further support for our hypothesis ([129]Fig. 4D). Furthermore, we performed experiments to determine whether ER stress increased the production of EVs. The parental cells were treated with ISRIB (an integrated stress response inhibitor; 1 μM). ISRIB is known to prevent cells from reestablishing ER homeostasis by blocking signaling through the Protein kinase R-like endoplasmic reticulum kinase (PERK) branch of the UPR but does not have global effects on translation, transcription, or mRNA stability in nonstressed cells ([130]Fig. 4E). As expected, ISRIB treatment reduced iSEV production by parental cells treated with GENRISEs but had a negligible effect on untreated cells ([131]Fig. 4F). On the other hand, when parental cells were pretreated with thapsigargin (an ER stress inducer via SERCA pump inhibition; 2.5 μM), iSEV production in GENRISE-treated cells was further enhanced compared with that in the thapsigargin-untreated cells (fig. S26). Additionally, we tested the effect of treatment with a microvesicle biogenesis inhibitor (d-pantethine) on iSEV production. Pantethine is a pantothenic acid (vitamin B5) derivative that plays a role in cholesterol metabolism. As cholesterol is important for maintaining the fluidity of the cell membrane, pantethine has been used as an inhibitor of microvesicle production ([132]Fig. 4G) ([133]42). After treatment with 200 nM pantethine, the parental cells treated with GENRISESs exhibited a significant reduction in the release of large iSEVs (>300 nm) ([134]Fig. 4, H and I). These findings illuminate the molecular mechanisms underlying the impact of GENRISEs on cellular processes, specifically the sustained effects on the translational machinery and ER stress regulation at the transcriptome level, which lead to iSEV generation. Surface engineering of iSEVs for extra functionality To explore the potential applications of iSEVs, we used metabolic engineering of parental cells for surface modification ([135]Fig. 5A and fig. S27, A to E). The substitution of l-methionine (l-met) with l-azidohomoalanine (AHA), an amino acid analog, enabled azido-modification of proteins within a few hours, facilitating additional engineering within a practical time frame in vitro ([136]43). Notably, this metabolic replacement of l-met with AHA exhibited negligible cytotoxicity across a broad concentration range (fig. S27C). Significantly, the successful modification of parental cells was directly translatable to iSEVs, enabling their functional engineering ([137]Fig. 5A). Azido-modified iSEVs exhibited no significant differences in size or morphology compared to unmodified iSEVs, as observed via TEM, indicating stable iSEV generation from modified parental cells ([138]Fig. 5B). Fig. 5. Surface decoration of iSEVs to provide extra functionality. [139]Fig. 5. [140]Open in a new tab (A) Azide modification of parental cells by metabolic engineering. Specificity of engineering capability, suggested by conjugating the azide-modified HeLa cells with streptavidin-QD655 followed by treatment of DBCO-biotin. (B) Surface-decoration of iSEVs, enabled by premodification of parental cells with azide functionalities by metabolic engineering. TEM images for azide-modified iSEVs. Scale bars, 100 nm. (C) TEM images and fluorescence images of QD655-decorated iSEVs, achieved by the sandwiched conjugation of the azide-modified iSEVs with DBCO-biotin followed by an introduction of streptavidin-QD655. Magnified TEM images of QD655-iSEV revealed the quantum dot structure at the surface of iSEV. DIC, differential interference contrast. (D) TEM images of AuNP-iSEVs decorated by direct conjugation of azide-iSEVs with DBCO-modified AuNPs. Scale bars, 100 nm. The feasibility of direct modification of iSEVs’ surfaces underscores their broad capacity for surface engineering with diverse functional molecules tailored to specific purposes. First, the versatility of iSEVs’ available surface chemistry can be explored with quantum dot (QD) as functional inorganic particles. By applying sandwiched conjugation with Dibenzocyclooctyne (DBCO)-conjugated biotin and streptavidin-QD655, iSEVs were successfully decorated with QDs, as confirmed by TEM and fluorescence imaging ([141]Fig. 5C). Additionally, the successful modification of iSEVs with 20-nm-sized AuNPs further highlighted the extensive engineering capacity of iSEVs ([142]Fig. 5D). Collectively, these modifications demonstrate the adaptability of iSEVs in expanding their potential applications. Immunotherapeutic potential of iSEVs To evaluate the potential of iSEVs as immunotherapeutics, the ability to deliver endogenous proteins along with the overexpressed protein cargo of choice was assessed. First, parental cells stably expressing red fluorescence protein (DsRed-Express2) were treated with EGFP-encoding GENRISEs (fig. S28). The isolated iSEVs successfully cotransferred endogenous red fluorescent protein (RFP) and exogenously expressed EGFP via the GENRISEs to nonfluorescent recipient cells. The results unequivocally demonstrate the feasibility of iSEV as versatile delivery tools for both endogenous and exogenous proteins. Internal and external engineering capacities of iSEVs open a previously unidentified route to tailor EVs to various therapeutic applications. Particularly in this study, we sought to use iSEV system as cancer immunotherapeutics. To assess the biodistribution and splenic delivery of iSEVs, the membranes of iSEVs were fluorescently tagged with CellVue Claret (excitation max., 655 nm; emission max., 675 nm). At 1 hour after systemic administration, the spleen, liver, kidney, heart, and lung were harvested for In vivo imaging system (IVIS) imaging ([143]Fig. 6, A and B). The results indicated that the iSEVs were delivered to the spleen and liver. For further analysis, the internalization of iSEVs by splenic dendritic cells (DCs) and macrophages was evaluated by flow cytometry ([144]Fig. 6C and fig. S29A). Antigen-presenting cells (APCs; DCs) showed over 50% of the iSEV-positive population. In particular, by 6 hours after the injection of iSEVs, splenic DCs were substantially activated, as revealed by the up-regulation of CD40, CD80, CD86, major histocompatibility complex (MHC) class I, and MHC class II ([145]Fig. 6D). The dose was determined by administering a series of different doses and checking DC activation parameters (fig. S29B). The results suggest that iSEVs can serve as versatile delivery tools with surface modifications, capable of delivering selected cargoes, including antigens for immunotherapeutics. Fig. 6. In vivo melanoma therapy with iSEV-TRP2. [146]Fig. 6. [147]Open in a new tab (A and B) Biodistribution of fluorescence-tagged iSEVs [with CellVue Claret (excitation max, 655 nm; emission max, 675 nm)]. The spleen, liver, kidney, heart, and lung were harvested at 1 hour postinjection for imaging. (C) iSEV uptake to splenic immune cells after 1 hour of intravenous injection (15 μg of CellVue Claret-tagged iSEV per mouse), analyzed by flow cytometry. DCs, dendritic cells. The gating strategy is shown in fig. S29A. (D) Mean fluorescence intensities (MFIs) of CD40, CD80, and CD86 in splenic DCs were measured by flow cytometry 6 hours after 15 μg of iSEV injection. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001 by one-way ANOVA followed by Tukey post hoc test. Data represent means ± SD. (E and F) Schematic illustration of TRP2-iSEV–mediated immunotherapy. The iSEVs carrying both whole-tumor antigen and overexpressed TRP2 activate the immune response cascade in vivo to attack melanoma in a target-specific manner. (G) Schematic illustration of in vivo experimental setup for TRP2-iSEV–based melanoma therapy. sc, subcutaneous; ip, intraperitoneal. (H to K) C57BL/6 mice were inoculated with B16. The indicated drugs were injected at 3 day intervals starting 7 days after tumor injection. Tumors were dissected after treatment with PBS, control-iSEVs, or TRP2-iSEVs with or without anti–PD-L1 over 16 days postinjection. (H) B16 tumor growth curves are shown (significance determination by the log-rank test, means ± SEM. PBS versus TRP2-iSEV, P < 0.01; anti–PD-L1 versus anti–PD-L1 + TRP2-iSEV, P < 0.01; anti–PD-L1 + TPR2-iSEV versus anti–PD-L1 + control-iSEV, P = 0.03). (I) Dissected tumors and (J) their weights in each treatment group shown at 16 days posttreatment. (K) Production of interferon-γ (IFN-γ) and tumor necrosis factor–α (TNF-α) levels in the harvested splenic cell culture upon treatment of TRP2 antigenic peptides. **P < 0.01 by one-way ANOVA with Tukey post hoc test. Data represent means ± SD. TRP2-delivered iSEV for melanoma therapy To evaluate in vivo therapeutic efficacy of iSEVs, we designed GENRISEs encoding antigenic peptides for melanoma derived from tyrosinase-related protein 2 (TRP2) residues 180 to 188. GENRISE induce the continuous generation of TRP2 peptides, which leads to generating TRP2-enriched iSEVs ([148]Fig. 6E). Of note, a mouse melanoma cell line (B16F10) was used as a parental cell, so that the resulting iSEVs also bear whole-tumor antigens. Together, the system enables a large-scale synthesis of iSEVs for downstream applications in which iSEVs can deliver whole melanoma antigen with overexpressed antigenic peptides. Additionally, a comprehensive blood chemistry panel evaluating hepatic functions, kidney metabolites, and blood counts following iSEV treatment (25 μg per mouse) suggested that iSEVs had negligible systemic toxicity (fig. S30). With the safety of iSEVs, they were applied as cancer immunotherapeutics. In a melanoma-bearing mice model, iSEV-TRP2s and anti–programmed death-ligand 1 (PD-L1) antibodies (Abs) were administered, and the antitumor effects were assessed. At the injection site, iSEV-TRP2s are taken up by APCs. Notably, iSEVs carry the overexpressed TRP2 from GENRISEs as well as whole-tumor antigens from the parental melanoma cells. As shown in [149]Fig. 6F, upon processing and presentation of the antigens by APCs, T cells were activated to produce an antigen-specific immune response cascade and attack tumors in a target-specific manner. The coadministration of anti–PD-L1 Abs as a checkpoint inhibitor further increased the tumor elimination efficacy by T cells. iSEV-TRP2 significantly reduced the tumor size in the melanoma-bearing mice model (40 μg of iSEV per mouse), indicating that GENRISE-induced overexpression of the antigenic peptide led to an antigen-specific immune response in vivo ([150]Fig. 6, G to J). Notably, the whole-tumor antigen effect was not detectable (in the control iSEV group), whereas overexpression of TRP2 in iSEVs showed significant target specificity. When combined with checkpoint blockade therapy by anti–PD-L1 Abs, iSEV-TRP2 synergistically outperformed control iSEVs, leading to a significant tumor regression. Notably, the treatment was effective without additional adjuvants, suggesting that iSEV-mediated therapy includes intrinsic adjuvants. The administration of iSEV-TRP2 to mice elicited TRP2 peptide-specific production of interferon-γ and tumor necrosis factor–α in the splenocytes, whereas control iSEVs showed a negligible effect ([151]Fig. 6K). Therefore, these data demonstrated that iSEVs can increase the anticancer effect of immune checkpoint inhibitors through the elicitation of antigen-specific immunity. DISCUSSION On the basis of previous developments in long-term and stable gene regulation by highly tangled genomic polymers [e.g., pre-siRNAs; ([152]23, [153]44)], we sought to translate RNA nanotechnology to engineered EVs. Thus, we developed an interwoven booster RNA platform that mimics RNA granules to amplify the generation of iSEVs. Like natural RNA granules that dynamically remodel the translational profile, GENRISEs, which are higher-order RNA assemblies, provide a depot for the local stimulation of protein synthesis upon delivery. In turn, the single RNA module serves concurrently both to assign a specific payload to EVs and to manipulate the scale of synthesis of EVs. The current approach requires a transfection reagent for efficient intracellular uptake, potentially restricting its applicability across diverse biological or clinical contexts. The organ- and cell-specific delivery strategies may expand the applicability of GENRISE toward therapeutic contexts, enabling controlled, tissue-specific EV production while minimizing off-target effects ([154]45). The GENRISEs successfully facilitated the increased release of iSEVs with enriched target proteins and the induction of ER stress. On the other hand, current methods of achieving the scalable synthesis of EVs are based on the induction of chemical or physical damage, including Ca^2+ influx and actin polymerization inhibition, which can cause irreversible consequences for parental cells ([155]14, [156]46, [157]47). In addition, further elaborate genetic engineering techniques or complicated chemical modifications are often necessary to manipulate the cargo. However, the local and sustained exposure of mRNAs from GENRISEs inside the parental cells enables the generation of large numbers of cargo-controlled iSEVs. In comparison to their artificial counterparts (i.e., liposomes), iSEVs essentially have all the beneficial features of naturally shed EVs with enriched target proteins. In addition to its inherent biocompatibility, the importance of our system should be further emphasized for a facile synthesis within one doubling time, allowing minimal changes in parental cells in producing the offspring iSEVs. Notably, RNA determines the yield of iSEVs and assigns cargo with minimal phenotypical changes in the parental cells. The GENRISE-based approach achieves the accelerated generation of cargo-encapsulated iSEVs independently of parental cell types for all tested mammalian cell lines and primary cells, suggesting the broad applicability of the system. Nevertheless, the chemical and physical properties (i.e., size, surface roughness, internal contents, etc.) are assigned for each type of iSEV depending on the parental cell type as a signature. Furthermore, we demonstrated that the GENRISEs boost the generation of large EVs (100 to 1000 nm) by up to three orders of magnitude, as well as the generation of small EVs (30 to 100 nm). To date, large-scale synthesis of EVs has focused mainly on small EVs due to their natural abundance. However, large EVs have distinct advantages, as their cargo capacity is higher by several folds to three orders of magnitude in terms of internal volume, they provide a better representation of surface proteins owing to their smaller curvature, and they can be obtained by a relatively simple isolation process ([158]48). More importantly, the immunotherapeutic efficacy of in vivo administered iSEV-TRP2 suggests the practical translation of iSEV-based immunotherapies. With an antigen specificity, iSEVs successfully impeded tumor growth. As patient-to-patient and tumor-to-tumor variations require identifying unique antigen profiles case by case, we envision that the iSEV platform would help advance personalized vaccine formulations ([159]49). In particular, the iSEV system allows cell-type–independent mass production of EVs predominated by specific cargoes, suggesting that future studies can be applied to generate patient cell-derived iSEVs enriched with neoantigens for personalized vaccines. In this case, the scaled-up synthesis of iSEVs with overexpressed target neoantigens is straightforward, and the membrane surface of the particles can be easily modified to introduce additional functionalities [e.g., targeting molecule ([160]50, [161]51) or adjuvant ([162]52)] to further improve therapeutic outcomes. Thus, we propose using iSEVs as a next-generation personalized drug delivery vector, empowered by a GENRISE-based approach. MATERIALS AND METHODS Generation of iSEVs The GENRISEs for generating iSEVs were synthesized with the protocol modified from previously reported ([163]24). Briefly, plasmid DNA bearing sequences for T7 promoter, IRES, and gene of interest (EGFP, DsRed-Express2, TagBFP, or TRP2 peptide) in order was purchased from Integrated DNA Technologies. In vitro transcription reaction was carried out with 5 nM pDNA, 2 mM ribonucleotide triphosphates (New England BioLabs), 2× reaction buffer (80 mM tris-HCl, 12 mM MgCl[2], 2 mM dithiothreitol, and 4 mM spermidine; pH 7.9 at 25°C), ribonuclease inhibitor (0.5 U/μl), and T7 RNA polymerase (50 U/1 ml) diluted in nuclease-free water for 20 hours at 37°C. The final product was purified with Zeba Spin Desalting Column (Thermo Fisher Scientific). The GENRISEs were mixed with TransIT-X2 Dynamic Delivery System (Mirus Bio) at 1:2.15 (w/v) ratio in Opti-MEMI reduced serum medium (Gibco), followed by 15 min of incubation at room temperature. The mixture was diluted in Opti-MEMI to obtain desired concentration and treated to Dulbecco’s phosphate buffered saline (DPBS)-washed cells. Cell culture HeLa, HEK293, A549, PC-3, B16F10, and RAW264.7 were distributed from Korean Cell Line Bank. MDA-MB-231 cell line was purchased from American Type Culture Collection. HeLa, HEK293, A549, primary hADSCs (provided by CHA hospital), and HDF (provided by Kangwon University) were grown in Dulbecco’s modified Eagle’s medium (Gibco), and MDA-MB-231, B16F10, RAW264.7, and PC-3 were grown in RPMI 1640 (Gibco). Both media were supplemented with 10% fetal bovine serum (FBS; Gibco) and 1× antibiotic-antimycotic (Gibco) at 37°C in a humidified atmosphere supplemented with 5% CO[2]. The cells were passaged routinely to maintain exponential growth. At 24 hours before the transfection (at ~80% confluency), the cells were trypsinized, diluted with fresh medium, and transferred to culture plates (15,000 to 20,000 cells/cm^2 depending on cell types). Collection and purification of iSEVs At 12, 24, or 48 hours after the treatment of GENRISEs, the medium was collected in sterile 15-ml conical tubes. The cells were trypsinized and diluted with the medium collected to achieve maximal recovery. The collected solution containing the trypsinized cells with iSEVs was centrifuged at 300g for 10 min at 4°C to remove the cells. The supernatant was centrifuged at 2000g for 10 min at 4°C to remove dead cells along with cell debris, as previously reported ([164]32). Then, the supernatant was centrifuged at 10,000g for 45 min at 4°C to collect large iSEVs (>100 nm) as a pellet. The iSEVs were resuspended in DPBS for downstream analysis. Concentrations and size distribution were measured by nanoparticle tracking analysis (NTA; Malvern Panalytical, NS300). For total iSEV (both small and large iSEVs) analysis, supernatant from 2000g centrifugation step was used. Characterization of iSEVs and iSEV-shedding cells Live-cell imaging microscopes (Eclipse T1, Nikon and DeltaVision system, GE Healthcare) were used to image untreated or GENRISE-treated cells in real time. Field-emission scanning electron microscope (Regulus8230, Hitachi), transmission electron microscope (JEM-2100F, JEOL), cryo-TEM (Tencai F20 G2, FEI), and AFM were used to obtain high-resolution images of the cells and iSEVs. For SEM imaging, the cells or iSEVs were fixed with glutaraldehyde, followed by the secondary fixation with osmium tetroxide (OsO[4]). Then, different concentrations of ethanol (25, 50, 75, 90, and 100%) were applied to achieve gradual dehydration. The dehydrated samples were then freeze-dried before analysis. The SEM-based EDX was also used to analyze the elemental compositions of the cells and iSEVs. The TEM sampling was carried out by the Advanced Analysis Center at the Korea Institute of Science and Technology (KIST) as previously reported. The TEM was used to characterize the iSEVs operating at an accelerated voltage of 200 kV. The samples were deposited onto Lacey Formvar/carbon-coated copper grids and then air-dried at room temperature. Different sampling conditions and the resulting images of iSEVs are outlined in fig. S31. For AFM analysis, iSEVs were scanned in noncontact mode using PPP-NCHR (Nanosensors) or BL-AC40TS (Olympus) probes. For the imaging in air, iSEVs were deposited on freshly cleaved mica (Ted Pella) and left to be adsorbed at 4°C for 30 min. The sample was rinsed and dried using nitrogen gas. For the imaging in liquid, iSEVs were deposited on 10% 3-aminopropyltriethoxysilane–coated mica. After the adsorption of samples for 30 min, DPBS was added to the liquid cell. Color enhancement and cross-sectional analysis for AFM images were carried out with XEI software (Park Systems). A fluorescence microscope (Eclipse Ti-U, Nikon) and confocal laser microscope (LSM800 with Airyscan, ZEISS) were used to visualize cells and iSEVs. All images were processed with ImageJ software provided by National Institutes of Health. Concentrations and size distribution of iSEVs or GENRISEs were assessed by NTA. For storage stability, iSEVs were resuspended in PBS, kept at 4°C, and then were subject to NTA-mediated size and concentration measurements. Zeta potential and size distribution measurement for iSEVs were carried out by dynamic light scattering. For investigation of iSEV-shedding cells, HeLa cells seeded in 35-mm confocal dishes were treated with GENRISEs (1.5 μg/ml) for EGFP. After 3 hours, the cells were washed once with PBS and analyzed with confocal microscopy. For visualizing plasma membrane and membranous intracellular compartments, cells were stained with wheat germ agglutinin conjugated with Alexa Flour 647 (Invitrogen) according to the manufacturer’s protocol. The cells were then fixed with 4% paraformaldehyde, which was followed by Hoechst33342 (final concentration of 2 μg/ml, Invitrogen) or 4′,6-diamidino-2-phenylindole (DAPI; final concentration of 300 nM, Invitrogen) staining for staining the nucleus before imaging with the confocal microscope. 3D reconstruction of images and diagonal fluorescence profiles were obtained by ImageJ. For cell viability and cytotoxicity measurement, the cells were treated with GENRISEs at concentrations indicated for each data point. After 24 hours of treatment, the cells were trypsinized and stained with the LIVE/DEAD Viability/Cytotoxicity Kit (Invitrogen; Calcein-AM and ethidium homodimer-1 for live and dead cell staining) following the manufacturer’s protocol. Then, the cells were analyzed by cytometry (NucleoCounter NC-3000, ChemoMetec). For the comparison of the transfection efficiency of GENRISEs with conventional mRNA, the HeLa cells were treated with GENRISE-EGFP or commercially available mRNA-EGFP (VectorBuilder Inc., Chicago, IL 60609, USA) complexed with TransIT-X2 reagent. The particle number quantification of mRNA-TransIT-X2 and GENRISE-TransIT-X2 was performed with NTA. After 24 hours of treatment, the cells were trypsinized and analyzed by cytometry (NucleoCounter NC-3000, ChemoMetec). RNA sequencing for iSEV-shedding cells GENRISE-EGFP (1.5 μg/ml) were treated to HEK293 cells seeded in six-well plates at 18,000 cells/cm^2. Untreated and treated cells were washed twice with PBS, trypsinized, and pelleted down at 300g by centrifugation. Total RNA was isolated using the RNeasy Mini Kit (QIAGEN), and concentrations, A260/280 and A260/230, were measured by spectrophotometer (NanoDrop 2000, Thermo Fisher Scientific). Two biological replicates were set for each sample, and all samples were sent to Novogene Co. Ltd. for sequencing by Illumina HiSeq platform. Quality control was performed on the clean reads, and mapping was performed with reference to the GRCh38/hg38 genome. Reactome pathway analysis (Reactome database release 81) was performed to classify genes into various categories and biological processes ([165]38). 3D construction of iSEVs with protein by TEMT Experimental data for TEMT were obtained by JEM-1400 (JEOL Ltd., Tokyo, Japan) operating at 120 kV. The tilt series of TEM images were recorded with a 1-s exposure time using a Veleta CCD camera (Olympus-SiS, Münster, Germany). In total, 55 images were acquired at tilting angles between −60° and +48°, with an increment of 2°. The magnification was ×60k, corresponding to a pixel size of 1.18 nm. Tilting, refocusing, and repositioning were carried out after every individual tilt increase. Alignment and reconstruction of tilt series were performed in IMOD software. 3D visualization of the final volumes was carried out using Amira 6.0 visualization software (Thermo Fisher Scientific, MA, USA). A nonlinear anisotropic diffusion filter was used to reduce the noise and segmentation of the reconstruction of the iSEV layer. Molecular simulations for the synthesis of iSEVs All simulations and analyses were performed using the GROMACS-2018 simulation package ([166]53–[167]55). The coordinates of green fluorescence protein (GFP) were downloaded from the Protein Data Bank (PDB code: 1GFL) ([168]56). Potential parameters for GFP and 1,2-dimyristoyl-sn-glycero-3-phosphocholine (DMPC) were taken directly from the MARTINI coarse-grained (CG) force field, which lumps a few (three or four) heavy atoms into each CG bead ([169]57–[170]59). Two GFP molecules and 1526 DMPC lipids were randomly distributed and then solvated with ~33,300 water molecules (representing ~133,200 real waters) in the periodic box of size 18 nm per side. To achieve electroneutrality, 12 counterions (Na^+) were added. A temperature of 310 K and a pressure of 1 bar were maintained by applying a velocity-rescale thermostat ([171]60) and Parrinello-Rahman barostat ([172]61) in an NPT ensemble with an isotropic pressure coupling. Lennard-Jones and electrostatic potentials were smoothly shifted to zero between 0.9 and 1.2 nm and between 0 and 1.2 nm, respectively. The LINear Constraint Solver (LINCS) algorithm was used to constraint the bond lengths ([173]62). Simulations were performed for 700 ns with a time step of 20 fs on computational facilities supported by the National Institute of Supercomputing and Networking/KIST Information with supercomputing resources including technical support (KSC-2020-CRE-0095). Inhibitor/inducer assays For ER stress and microvesicle (large EV) inhibitor assays, integrated stress response inhibitor (ISRIB, Selleckchem) and pantethine (Selleckchem) were used. For ER stress inducer assays, ER stress inducer (Thapsigargin, Merch) was used. ISRIB (1 μM) or pantethine (200 nM) or Thapsigargin (2.5 μM) was treated to HeLa cells in full growth medium in a 24-well plate. At 2 hours after the initial inhibitor or inducer treatment, GENRISE (1.5 μg/ml) was treated with refreshing of inhibitors or inducer. After 24 hours, supernatant was collected, and iSEVs were isolated by differential ultracentrifugation, followed by NTA. Cargo analysis For cargo transfer analysis, B16F10 stably expressing RFP was treated with GENRISE-EGFP (1.5 μg/ml). After 24 hours, iSEVs expressing both RFP and EGFP were isolated. iSEV-RFP-EGFP were treated to nonfluorescent B16F10 cells to evaluate cargo transfer. Six hours after the treatment, cells were washed twice with DPBS, fixed with 4% paraformaldehyde, stained with DAPI, and visualized with a fluorescence microscope. Cross-sectional fluorescence intensity profiles were obtained by ImageJ. For total protein content analysis, a micro BCA protein assay kit (Thermo Fisher Scientific) was used following the manufacturer’s protocol. For Western blotting analysis, cells or iSEVs were homogenized in lysis buffer [50 mM tris (pH 7.4), 150 mM NaCl, 0.5% Triton X-100, and 1 mM EDTA] supplemented with protease inhibitors (Roche). The lysates were then centrifuged at 13,500g for 30 min, and the supernatants were recovered. About 10 μg of protein extracts were resolved by SDS–polyacrylamide gel electrophoresis and analyzed by Western blotting using primary Abs against GM130 (B-10, Santa Cruz Biotechnology), TSG101 (2B7G8, Proteintech), Alix (1H9D9, Proteintech), CD63 ([174]EPR21151, Abcam), and annexin A1 (polyclonal, Proteintech). This was followed by incubation with a corresponding immunoglobulin G horseradish peroxidase–conjugated secondary Ab (Thermo Fisher Scientific) and detection using enhanced chemiluminescence (Bio-Rad). Animal experiments Balb/c (6 weeks old, female) and C57BL/6 (6 weeks old, female) mice were purchased from (Orient Bio Inc., South Korea) and Hyochang Science (Daegu, Korea), respectively. This study was conducted in accordance with the animal protection law and the Institutional Animal Care and Use Committee regulations of Asan Medical Center. The Committee on the Ethics of Animal Experiments of Asan Medical Center approved the protocol (mouse protocol numbers 2019-12-316, 2021-12-316, and 2022-12-296). In vivo uptake Balb/c mice were injected intravenously with PBS or CellVueClaret (by the CellVue Claret Far Red Fluorescent Cell Linker Mini Kit for General Membrane Labeling)–labeled iSEVs (15 μg resuspended in 100 μl of PBS). One hour after injection, the spleen, liver, kidney, heart, and lung were harvested. The organs were imaged together by the IVIS (PerkinElmer) system. The average region of interest was determined for each tissue type through the same threshold setting. For splenocyte uptake analysis in [175]Fig. 6C, the spleen of C57BL/6 mice was harvested at 1 hour after intravenous administration of CellVueClaret-labelled iSEVs (15 μg). The splenocytes were stained with anti-CD11b (M1/70), anti-CD19 (6D5), anti–TCR-b (H57-597), MHC class I (H2Kd, 28-8-6), and MHC class II (I-A/I-E, M5/114.15.2), anti-CD317(129C1), anti-NK1.1 (PK136), and anti-CD11c (N418) Abs on ice for 20 min. The cells were then washed and resuspended in PBS and analyzed using NovoCyte (ACEA Biosciences Inc.) and NovoExpress software. The gating strategy is shown in fig. S29A. Splenic DC activation analysis For splenocyte activation analysis in [176]Fig. 6D, spleen of C57BL/6 mice was harvested at 6 hours after injection 15 μg of iSEVs (7.5, 30, and 60 μg data shown in fig. S29B). Mice injected with lipopolysaccharide at 0.1 mg/kg were used as control group. Spleens were digested and homogenized with deoxyribonuclease and collagenase IV containing digestion buffer for 20 min at room temperature. Subsequently, 100-nm nylon mesh filter out undigested and aggregated tissues. Cells were washed with PBS, and the pellets were resuspended with 5 ml of Histopaque-1077 (Sigma-Aldrich) medium, with the upper layer suspended with 5 ml of the same Histopaque-1077. An additional 1 ml of FBS was added to the layer above Histopaque-1077. After centrifugation at 1700 rpm for 10 min, a fraction exceeding 1.077 g/cm^3 cells was harvested as the leukocytes. The leukocytes were stained fluorescence-labeled monoclonal Abs for 30 min. Anti-CD45R/B220 (RA3-6B2), anti-CD3 (17A2), anti-CD49b (DX5), anti-Gr1 (RB68C5), and anti–TER-119 (TER-119) were added as lineage markers. Splenic DCs were defined as a lineage-CD11c^+ cells. In vivo toxicity test Balb/c nude mice without tumor mass were administered with PBS or iSEV-EGFP (15 μg or 25 μg, intravenously, fig. S30). Twenty-four hours after injection (n = 3 per group), the animals were euthanized, and blood was collected from the heart. The blood sample was separated into EDTA tubes and serum separator tubes (SST) and then was used for the complete blood count (CBC) test and liver blood tests, respectively. CBC test was performed by Advia 2120i automated analyzer (Siemens). The serum contents of blood urea nitrogen, creatinine, total protein, total bilirubin, aspartate aminotransferase, and alanine aminotransferase levels were performed by Hitachi Clinical analyzer 7180 (Hitachi). In vivo therapeutic efficacy For melanoma therapy model ([177]Fig. 6, G to K), C57BL/6 mice were subcutaneously injected with B16F10 (1 × 10^6/100 μl). The mice were randomly divided into groups of PBS, control iSEV, and TRP2-iSEV (for melanoma), with or without anti–PD-L1 Abs (10 mg/kg). Each group was subcutaneously injected with indicated reagent for every 3 days from day 7 of tumor injection. Tumor volume was measured every 3 days and calculated using the formula V = 0.5 (L/w^2), where L is the longest dimension and w is the shortest dimension. Tumor was excised at day 16 for weight measurement. Statistical analysis Data in this study represent the mean values of independent measurements. Error bars indicate means ± SDs of each experiment as indicated. Statistical analysis was performed with unpaired two-tailed Student’s t test and one-way or two-way analysis of variance (ANOVA) with post hoc test as indicated. Statistical significance was assigned for P < 0.05, P < 0.01, P < 0.001, or P < 0.0001 (confidence levels of 95, 99, 99.9, and 99.99%, respectively). Acknowledgments