{"paper_id":"1a6dfb52-6f4c-4a7e-b678-a4b0015abb92","body_text":"TITLE \nIdentification of Novel Extracellular Vesicles Scaffold Proteins for Versatile Cargo \nEngineering \n \nAUTHORS \nTao Qiu1,*, Rui Hu1, Yuan Yi1, Wenqiang Lu1, Chuang Cui1, Shuiqin Niu1, Ke Xu1,* \n \nAFFILIATIONS  \n1. Vesicure Therapeutics, Biobay 3B, Suzhou Industrial Park, Suzhou, China \n*. Corresponding authors \nTao Qiu, E-mail: tao.qiu@vesicure.com \nKe Xu, E-mail: ke.xu@vesicure.com \n \nABSTRACT \nExtracellular vesicles (EVs)  are promising drug delivery platforms that have been \nengineered to carry various drug modalities. Key strategies for generating those \ntherapeutic EVs involve the direct fusion of protein of interest  (POI) to EVs scaffold \nproteins with inherently high EVs -sorting ability. In this work, we identified raftlin \n(RFTN1) and poliovirus receptor (PVR) as novel EVs scaffold proteins for loading EVs \nlumen and surface cargos respectively. Truncation studies revealed that the N-terminus \n15 residues from RFTN1 (RFTN1-N15) were sufficient for EVs engineering, as \ndemonstrated by distinct cargos including gene editing tools, cytosolic enzymes as well \nas type II transmembrane proteins. On the other hand, PVR efficiently displayed \nsecreted proteins including antibodies and serum albumins on EVs surface. Critically, \nRFTN1 and PVR-engineered EVs demonstrated consistent and efficient cargo delivery \nin vivo. In summary, the discovery of RFTN1 and PVR can potentially benefit EVs \nengineering for both fundamental research and clinical translation in the future.  \n \nKEY WORDS \nExtracellular vesicles, Extracellular vesicles engineering, Scaffold proteins \n \n \n \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.04.697584doi: bioRxiv preprint \n\n1. INTRODUCTION \nExtracellular vesicles (EVs) are naturally derived carriers that can be secreted by most \ncell types  (van Niel, D'Angelo, & Raposo, 2018) . Whereas EVs mediate cellular \ncommunications in physiological context, EVs are also undergoing intensive \ninvestigations for developing into novel therapeutic platform (Du et al., 2023; Kumar \net al., 2024). Their innate properties—including low immunogenicity, ability to cross \nphysiological barriers, and potential targeting capabilities —present attractive \nadvantages for delivering various drug modalities (Du et al., 2023; Kumar et al., 2024).  \nDepending on the purposes and properties of drug modalities, the molecules of interest \ncan be loaded either on EVs surface, or within EVs lumen (Ma et al., 2025; Yang, Xue, \nDuan, Mao, & Wan, 2024). Typically, EVs producer cells are genetically engineered to \noverexpress the protein of interest (POI) fused to an EV s scaffold  protein, thereby \nallowing EVs cargo enrichment (Ma et al., 2025; Yang et al., 2024). Therefore, the EVs \nscaffold proteins with high intrinsic EVs-sorting ability are the key element s in EVs \nengineering strategy.  \nTraditionally, EVs marker proteins such as CD9, CD63 and CD81 have been applied as \nscaffold proteins (Ma et al., 2025; Yang et al., 2024) . Over the years, several studies \nhave performed EVs proteomic analysis to identify novel EVs scaffold proteins with \nhigher efficiency, leading to the identification of PTGFRN and BASP1  (Dooley et al., \n2021), TSPAN2 and TSPAN3 (Zheng et al., 2023)  and PLXNA1 (Zhao et al., 2024) . \nHerein, we performed proteomic studies on HEK293 -derived EVs independently, and \nreported raftlin (RFTN1) and poliovirus receptor (PVR) as novel EVs scaffold proteins \nfor loading EVs lumen and surface cargos respectively.  Distinct cargos modalities \nincluding gene editing tools, cytosolic enzymes, type II transmembrane proteins and \nantibodies were tested for EVs loading and validated with desired activity both in vitro \nand in vivo. Therefore, we believe this study will potentially benefit EVs engineering \nfield. \n2. MATERIALS AND METHODS \n2.1 Cell culture and transfection \nSuspension-adapted HEK293 cell line (A23109, Quacell) was cultured in OPM-CD05 \nmedium (81075-001, OPM Biosciences) and maintained on orbital shaker at 90 RPM \nin a humidified incubator at 37℃ with 8% CO2 . MDA-MB-231 (CL-0150, Procell), \nSKOV3 (CL-0215, Procell) and B16F10 (CL-0319, Procell) were cultured in DMEM \nmedium (11965092, Gibco) supplemented with 10% FBS (A5669701, Gibco) and \nmaintained in a humidified incubator at 37 °C with 5% CO₂. Cells were passaged every \n2–3 days.  \n2.2 EVs purification from cell culture medium \nThe cell culture supernatant was first filtered through 0.45 μm filter units (SLHPR33RB, \nMillipore), followed by centrifugation (SW32Ti, Beckman Coulter) at 133,900 g at 4 ℃ \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.04.697584doi: bioRxiv preprint \n\nfor 60 min. The crude EVs pellet were resuspended in PBS, and further layered onto \n17.5% Optiprep /45% Optiprep gradient (D1556, Sigma), followed by centrifugation at \n150,000 g at 4 ℃ for 16 h. The extracellular vesicles appeared as a white layer between \nPBS/17.5% iodixanol, which were carefully pipetted out and washed with PBS by \ncentrifugation at 135,000 g at 4 ℃ for 3 h.  The refined EVs were finally resuspended \nin PBS (10010, Gibco), and stored in -80℃ freezer.  \n2.3 Western blot analysis \nThe cells or EVs samples were first lysed with RIPA lysis buffer (R0010, Solarbio) on \nice for 20 minutes. Then the total protein were quantified with MicroBCA protein assay \nkit (23235, Thermo scientific). After that, SDS-PAGE protein loading buffer (BL502A, \nBeyotime) was added into the samples and incubated at 95 °C for 10 min. Equal \nquantity of proteins for each sample  was loaded onto 4 –12% SurePAGE™, Bis-Tris \ngels ( M00653, GenScript). After electrophoresis (170 V , 35 min), the proteins were \ntransferred onto PVDF membrane ( ISEQ00010, Millipore). The membranes were \nblocked with QucikBlockTM Western blocking buffer ( P0252, Beyotime) for 1h at \nroom temperature (RT) before incubation with primary antibodies overnight at 4  °C. \nAfter extensive washing with TBST wash buffer (ST673, Beyotime), the membrane \nwas further incubated for 2h at RT with HRP conjugated secondary antibodies . \nFollowing extensive washes with TBST buffer , the memb ranes were incubated with \nPierce ECL Western Blotting Substrate (32209, Thermo scientific) and visualized with \nTannon 5200 imager (Tannon). \nThe following antibodies were used: CD9 (abcam, AB263019), CD81 (Cell Signaling \nTechnology, 56039S), Calnexin (abcam, ab22595), CD63 (Cell Signaling Technology, \n2897), Cre recombinase  (Cell Signaling Technology , 15036), TSG101 ( abcam, \nab125011), TMPRSS2 ( Cell Signaling Technology, 39665), 4-1BBL (Cell Signaling \nTechnology, 59127), Arginase1 ( Cell Signaling Technology , 93668 ), iNOS ( abcam, \nab178945), HRP Goat Anti -Rabbit IgG (H+L) (ABclonal, AS014), HRP Goat Anti -\nMouse IgG (H+L) (ABclonal, AS003). \n2.4 Electron Microscopy \nFor transmission electron microscopy, EVs were applied onto the glow -discharged \ncopper grid (200 mesh, coated with carbon film). To perform negative staining, 2% \nuranyl acetate were incubated with EVs at room temperature for 1 min, followed by \nquick wash with distilled water to remove excess stain. The grids were air-dried before \nbeing imaged under Tecnai G2 transmission electron microscope (Thermo FEI, 120 kV). \nFor cryo -electron microscopy, EVs were applied to glow -discharged copper grid \n(Cu300, Quantifoil, 212601), and cryo -frozen by liquid nitrogen with Vitrobot plunge \nfreezer (bolt time 4s, bolt force 0, wait time 30s). Images were acquired under cryo-EM \n(Thermo Glacios, 200 kV).  \n2.5 Nanoparticle tracking analysis (NTA) \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.04.697584doi: bioRxiv preprint \n\nTo evaluate the size distribution and concentration, EVs samples were freshly diluted \nat 1,000–50,000 fold in 0.22 -mm filtered PBS and analyzed immediately  with \nZetaView Nano Particle Tracking Analyzer (ParticleMetrix, PMX120-Z).  \n2.6 RT-qPCR \nTotal RNA from the cells plated on 96 well plates were first extracted and purified using \nthe RNAprep Pure Micro Kit (DP420, TIANGEN). The concentration and quality of \nisolated RNA were detected using a NanoDrop (ThermoFisher). Reverse transcription \nwas performed with HiScript Ill cDNA Synthesis Kit (R312 -01, Vazyme). Next, the \ncDNA was added into the 2xSYBR green qPCR mix (A0012 -R2, Ezbioscience), and \nquantitative PCR were analyzed with QuantStudio3 (Applied Biosystems).  \n2.7 Macrophage polarization assay \nIn polarization experiments, RAW264.7 cells were seeded at 20000 cells per well in 96-\nwell flat-bottom tissue culture plates and cultured in polarizing medium overnight. M1 \npolarization of RAW264.7 cells was induced by 100 ng/mL Lipopolysaccharides (tlrl-\n3pelps, Invivogen) and 2.5 ng/mL IFN -γ (575302, Biolegend) for 24 hours. M2 \npolarization of RAW264.7 cells was induced by 20 ng/mL IL -4 (574302, Biolegend) \nfor 24 hours.  \n2.8 B16F10 tumor xenograft model \nFemale, 8 weeks old C57BL6/J mice (GemPharmat) were implanted with 1E6 B16F10 \ncells/mice under the right fat pad region. When the average tumor volume reached \naround 60 mm 3, the mice were randomly grouped for different treatment conditions. \nIntratumoral injections and tumor volume measurement were performed every day for \nseven days consecutively. On last day, mice were sacrificed and tumors were excised \nout and imaged.  \n2.9 LPS induced acute lung injury model  \nTo establish acute lung injury model, 8 weeks old C57BL6/J mice (GemPharmat) were \nnasal instill ed with 50 μL LPS (tlrl -eklps, InvivoGen) at 5 mg/kg  dose. For EVs \ntreatment, EV s were adminstrated  by pulmonary nebulization using Micro Sprayer \nAerosolizer (Y655650918, YuYanbio), at 50 μL volume in PBS per mouse. At the end \nof the experiment, mice were euthanized, and lung, serum and bronchoalveolar lavage \nfluid (BALF) samples were collected.  \n2.10 Cytokines analysis in BALF samples \nThe BALF supernatants were collected, centrifuged at 1000 × g for 10 min at 4°C, and \ncytokine levels were measured by mouse cytometric bead array (CBA) Kit (BD \nBiosciences). Briefly, 50 μL of samples (BALF supernatants) or known concentrations \nof standard samples (0–5000 pg/mL) were added to a mixture of 50 μL each of capture \nantibody bead reagent and phycoerythrin (PE) -conjugated detection antibody. The \nmixture was then incubated for 2 h at room temperature in the dark and then washed to \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.04.697584doi: bioRxiv preprint \n\nremove unbound detection antibody. Data were acquired using a FACSCelesta \ncytometer and analyzed using CBA software FACP V3.0. \n2.11 Histological analysis \nTissue samples obtained from the C57BL6/J mice were first fixed in 4% \nparaformaldehyde at 4 ℃ overnigh t, then dehydrated in 30% (w/v) sucrose solution \n(A610498, Sangon Biotech) for 2 days. The dehydrated samples were then embedded \nin Tissue -Tek O.C.T Compound ( 4583, SAKURA) blocks and frozen overnight. \nSections acquired at 5 μm thick sections were fixed at 4% paraformaldehyde for 10 min \nbefore staining with hematoxylin and eosin ( C0105M, Beyotime). Imaging analysis \nwere performed under light microscope (MF43N, Mshot). \n2.12 EVs proteomics identification \nNanoflow LC -MS/MS analysis of tryptic peptides from EVs was conducted on a \nquadrupole Orbitrap mass spectrometer (Q Exactive HF -X, Thermo Fisher Scientific, \nBremen, Germany) coupled to an EASY nLC 1200 ultra-high pressure system (Thermo \nFisher Scientific) via a nano-electrospray ion source. 500 ng of peptides were loaded \non a 25 cm column (150 μm inner diameter, packed using ReproSil -Pur C18-AQ 1.9- \nµm silica beads; peptides were separated using a gradient from 8 to 12% B in 5 min, \nthen 12% to 30 % B in 33 min and stepped up to 40% in 7 min followed by a 15 min \nwash at 95% B at 600 nl per minute where solvent A was 0.1% formic acid in water and \nsolvent B was 80% ACN and 0.1% formic acid in water. The total duration of the run \nwas 60 min. Column temperature was kept at 60 °C using an in-house-developed oven. \nBriefly, the mass spectrometer was operated in “top -40” data -dependent mode, \ncollecting MS spectra in the Orbitrap mass analyzer (120,000 resolution, 350–1500 m/z \nrange) with an automatic gain control (AGC) target of 3E6 and a maximum ion injection \ntime of 80 ms. The most intense ions from the full scan were isolated with an isolation \nwidth of 1.6 m/z. Following higher -energy collisional dissociation (HCD) with a \nnormalized collision energy (NCE) of 27, MS/MS spectra were collected in the Orbitrap \n(15,000 resolution) with an AGC target of 5E4 and a maximum ion injection time of 45 \nms. Precursor dynamic exclusion was enabled with a duration of 16 s. \nFor data analysis, all raw files were analyzed using the Proteome Discoverer suite \n(version 2.4, Thermo Fisher Scientific). MS2 spectra were searched against the \nUniProtKB human proteome database containing both Swiss -Prot human reference \nprotein sequences . The Sequest HT search engine was used, and parameters were \nspecified as follows: fully tryptic specificity, maximum of two missed cleavages, \nminimum peptide length of 6, fixed carbamidomethylation of cysteine residues \n(+57.02146Da), variable modification s for oxidation of methionine residues \n(+15.99492Da), precursor mass tolerance of 15 ppm and a fragment mass tolerance of \n0.02Da for MS2 spectra collected in the Orbitrap. Percolator was used to filter peptide \nspectral matches and peptides to a false disco very rate (FDR) of less than 1%. After \nspectral assignment, peptides were assembled into proteins and were further filtered \nbased on the combined probabilities of their constituent peptides to a final FDR of 1%. \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.04.697584doi: bioRxiv preprint \n\nAs default, the top matching protein or ‘master protein’ is the protein with the largest \nnumber of unique peptides and with the smallest value in the percent peptide coverage \n(that is, the longest protein). Only unique and razor (that is, parsimonious) pep tides \nwere considered for quantification. \n2.13 ELISA \nTo determine IL15 and HSA concentration in engineered EVs, the IL15 ELISA kit \n(R&D systems, D1500) and HSA ELISA kit (Invitrogen, EHALB) were used according \nto manufacturer’s protocol. EVs were first permeabilized by incubating with lysis \nbuffer (PBS+0.3% TritonX-100) at room temperature for 30 min, before proceeding to \nELISA quantification.  \n2.14 Dye labeling of EVs and in vivo animal imaging \nFor fluorescent dye labeling of EVs, the lipophilic tracers DiO (Invitrogen, D275) and \nDiR (Invitrogen, D12731) were prepared according to manufacturers’ protocol. Briefly, \nEVs were incubated with dye (1:1000 dilution from stock) at room temperature for 30 \nmin, and were further washed twice with PBS by centrifuge at 135,000 g for 1 h.  \nTo analyze the distribution of EVs , the mice were I.V administered with dye-labeled \nEVs in 100 μl PBS through tail vein. At indicated time points, the mice were sacrificed \nand the internal organs (liver, spleen and tumours) were harvested and observed by an \nIVIS Spectrum (PerkinElmer, Waltham, MA, USA). The sampled blood was collected \ninto the 0.5 M EDTA -treated tubes and centrifuged at 1000×g for 10 min to get the \nplasma for EVs detection. For EVs quantification, the plasma samples were quantified \nfor fluorescent dye intensity by PHERAstar FSX plate reader (BMG labtech). \n2.15 Statistical analysis \nExperimental replicates  were defined in the figure legends for each experiment. \nStatistical analyses were performed in GraphPad Prism 7 using student’s t -test for \nexperiments with two groups  or one-way analysis of variance (ANOV A)  for \nexperiments with three or more groups. Values were expressed as mean ± standard \ndeviation (SD) or as mean ± standard error of the mean (SEM), as indicated in the figure \nlegends. Significance labeling and p values were presented in figures descriptions.  \n3. RESULTS \n3.1 Purification and proteomic characterization of HEK293-derived EVs. \nSuspension-adapted HEK293 cells grown in chemically -defined, serum-free medium \nwere used as the EVs parental cells. We reasoned that the quality of EVs preparations \ncritically affect the data fidelity of acquired EVs proteomics. Therefore, we first set out \nto develop an EVs purification process based on density gradient centrifugation. Crude \nEVs were refined and collected at the OptiPrep layer with density around 1.10 -1.12 \ng/mL (Figure 1A). Nanoparticle tracking (NTA) analysis revealed EVs mean diameter \nto be around 120 nm (Figure 1B). The typical cup-shaped morphology was observed \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.04.697584doi: bioRxiv preprint \n\nfor EVs under t ransmission electron microscopy (TEM) images (Figure 1C). \nAdditionally, minimum protein contaminants or cellular debris were observed, \nsuggesting EVs were acquired with high -purity (Figure 1C). Western blots (WB) \nanalysis further confirmed the purified EVs were positive for typical EVs markers CD9, \nCD63, CD81, TSG101 and negative for calnexin (Figure 1D). \nNext, to identify potential EVs scaffold proteins with high EVs enrichment ability, mass \nspectrometry analysis was performed for purified EVs as well as parental cells. We \nreasoned that EVs scaffold proteins should be up -regulated in EVs proteomics versus \ncellular proteomics, which represented the innate EVs sorting ability. H ence, \ndifferential expression analysis was further performed for EVs versus cellular \nproteomics (Figure 1E-F). Protein candidates were ranked based on the significance \nscore. Within this list, established EVs markers (CD9, CD63, CD81, TSG101) and \npreviously reported EVs scaffold proteins (PTGFRN, BASP1, MARCKS) appeared as \ntop candidates, confirming the good quality of proteomic data as well as the stringency \nof screening criteria.  \nBased on structural features, the top protein candidates could be further categorized as \nfollow: (1) single-pass transmembrane proteins, including type I and type II \ntransmembrane proteins; (2) tetraspanins; (3) multipass-transmembrane proteins; (4) \nmembrane-associated proteins, which were anchored to membranes through lipidation \nmodifications; (5) cytosolic proteins which typically had no association with \nmembranes.  \n3.2 Validation of RFTN1 as EVs scaffold proteins for loading luminal cargos. \nWe first set out to identify candidates that could be used to load cargos into EVs lumen. \nIn previous report, BASP1 and MARCKS were favorable scaffold proteins for loading \nEVs luminal cargos, both proteins associated with the inner leaflet of cellular \nmembranes through N-terminal myristoylation (Dooley et al., 2021). Similarly, the N-\nterminus octapeptide from Src kinases facilitated Cas9 protein encapsulation into EVs, \nwhich also depended on N -terminal myristoylation (Whitley et al., 2022) . We \nspeculated that these N-terminal lipidated proteins had unique advantages in both EVs \nenrichment and engineering without disturbing cargos’ function. Therefore, we focused \non screening candidate proteins that had N-terminus lipidation.  \nRFTN1, also known as raftlin, appeared to meet the criteria among the top candidates. \nRFTN1 was reported as a lipid raft–associated protein that played a critical role in the \norganization and function of membrane microdomains essential for immune receptor \nsignaling (Saeki, Miura, Aki, Kurosaki, & Yoshimura, 2003) . RFTN1 was \nmyristoylated on the second glycine residue, and palmitoylated on the third cysteine \nresidue (Saeki et al., 2003) . It also contained a stretch of charged residues on the N -\nterminus, with net charge s being +2 (Figure 2A). To test the feasibility of RFTN1 to \nengineer cargos into EVs lumen, Cre recombinase was selected and fused to the C -\nterminus of RFTN1 via a flexible glycine -serine peptide linker (Figure 2B). The viral \nfusogen VSVG was pseudotyped on EVs membrane to facilitate Cre release (Figure \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.04.697584doi: bioRxiv preprint \n\n2B). A Cre -loxP recombination assay was designed  such that upon successful Cre \ndelivery, the recipient cells would emit red fluorescence in nucleus in contrast to \nuntreated cells containing only membrane-bound green fluorescence (Figure 2C). Total \n1E10 EVs particles were added to 3E4 reporter cells, followed by fluorescence analysis \nat 48 hours post EVs treatment. As shown in Figure 2D, RFTN1 engineered EVs \nsuccessfully enriched Cre recombinase by WB analysis. Consequently, the Cre EVs  \ndelivered to recipient cells with high efficiency, recording around 70% recombination \nrate (Figure 2E). \nTo compare the efficiency between RFTN1 and other established scaffold proteins, \nBASP1 was similarly fused to Cre recombinase through GS linkers. Cre -loxP \nrecombination assay revealed that RFTN1 performed at similar efficiency as  BASP1 \nwhen added at same quantity (Supplementary Figure 1A). To test functional cargo other \nthan Cre recombinase, a constitutive active form of β-catenin (β-catenin ΔEX3) was \nselected and tested (Harada et al., 1999). RFTN1 and BASP1 were fused to β-catenin \nΔEX3 with GS linkers, and equal amount of EVs (1E10 particles) were added to \nTOPFLASH reporter cells for analysis 48 hours later.  Again, RFTN1-engineered EVs \nefficiently activated the β-catenin signaling pathway, at similar efficiency to BASP1 as \nwell as the small molecule lithium chloride (Supplementary Figure 1B). In summary, \nRFTN1 was validated as efficient scaffold proteins for engineering EVs luminal cargos. \n3.3 Identification of the minimum sequences from RFTN1 as EVs scaffold. \nAs reported previously, the N-terminus 10 residues from BASP1 and Src kinase were \nsufficient for EVs engineering and cargo loading  (Dooley et al., 2021; Whitley et al., \n2022). Therefore, we tested if minimum sequences could be derived from RFTN1 as \nwell. The N -terminus 5 residues (RFTN1 -N5), 10 residues (RFTN1 -N10) and 15 \nresidues (RFTN1 -N15) were fused to Cre recombinase and compared for loading \nefficiency with RFTN1 full length (RFTN1 -FL). Interestingly, RFTN1 N -terminus \ntruncates performed gradually better as the length increased, to the point where RFTN1-\nN15 performed at equal efficiency as RFTN1-FL (Figure 2D-E). Although RFTN1-N5 \ncontained the two putative N -lipidation mo difications, the results suggested that the \nstretch of charged residues were also indispensable for EVs sorting activity.  \nNext, mutagenesis study was performed to confirm that the two putative N -lipidation \non RFTN1 were indeed critical for its EVs sorting ability. A G2A mutation was designed \nto abolish the N-myristoylation, and a C3A mutation was designed to abolish the N -\npalmitoylation. As expected, either single mutation partially affected RFTN1 ’s EVs \nsorting activity, whereas combined double mutations almost completely abolished the \nactivity (Figure 2F-G).  \nIn summary, the N -terminus lipidation as well as charged residues were critical for \nRFTN1’s EVs sorting ability, and the N -terminus 15 aa were sufficient to be used for \nEVs engineering. \n3.4 RFTN1-N15 demonstrated versatility in EVs engineering and cargo loading. \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.04.697584doi: bioRxiv preprint \n\nSince RFTN1-N15 demonstrated sufficient efficiency in EVs engineering during proof-\nof-concept study using Cre -reporter system, we moved on to test on a wide range of \nfunctional cargos.  \nBoth Cre recombinase and β-catenin ΔEX3 were nucleus-localized cargos, suggesting \nthat RFTN1-N15’s EVs sorting signal was not mutual exclusive with cargos’ nucleus \nlocalization signal when engineered in fusion protein form. Subsequently, Cas9 protein \nas another typical nucleus cargo was tested for EVs encapsulation and delivery (Figure \n3A). The “Stoplight” Cas9 reporter assay was designed as described previously (de \nJong et al., 2020) . Briefly, the successful gene editing was evidenced by EGFP \nfluorescence due to the correction of frame-shifted EGFP expression cassette. RFTN1-\nN15 was directly fused to the N -terminus of C as9 protein, and the engineered EVs \n(1E10 particles) were added to 3E4 reporter cells. At 48 hours post treatment, prominent \ngene editing events was shown by the emerging of GFP positive cells, recording around \n30% efficiency by flow cytometry quantification (Figure 3B-C). \nNext, RFTN1-N15 was further tested for engineering cytoplasm-localized cargos. We \nhave previously designed minRISC -EVs platform as gene silencing tool, which \nessentially encapsulated EVs with AGO2 protein complexed with guide RNAs  (Tao \nQiu, 2025). Herein, RFTN1-N15 was fused to the N-terminus of AGO2, while keeping \nthe other elements same as previously described  (Figure 3D). More than 95% EGFP \nsilencing rate was observed (total 1E10 particles, 48 hours post treatment), suggesting \nRFTN1-N15 was capable in loading the gene silencing complex in EVs with high \nefficiency (Figure 3E-F). In another example, FCU1 was a chimeric protein consisted \nof yeast cytosine deaminase (CDase) and uracil phosphoribosyltransferase (UPRTase) \n(Erbs et al., 2000) . FCU1 efficiently catalyze d the direct conversion of 5 -FC, a \nrelatively nontoxic antifungal agent, into the toxic metabolites 5 -fluorouracil (5-FU) \nand 5-fluorouridine-5’monophosphate (5-FUMP), thus had been investigated for anti -\ntumor therapy (Erbs et al., 2000) . To examine if FCU1 could be loaded into EVs and \nexhibit function, RFTN1-N15 was fused to N -terminus of FCU1 (Figure 3G). In the \npresence of 5-FC and engineered EVs, around 50% cell death was observed, confirming \nthe efficacy of the strategy (Figure 3H-I). \nFinally, we wondered if RFTN1 -N15 could help to enrich type II transmembrane \nproteins on EVs as well. The transmembrane protease serine 2 (TMPRSS2)  was \nessential host cell factor  for aiding the cellular entry of t he severe acute respiratory \nsyndrome coronavirus 2 (SARS -CoV-2) (Koyou et al., 2025) . Engineered decoy EVs \nwith overexpressed angiotensin-converting enzyme 2 (ACE2) receptor and TMPRSS2 \nhad demonstrated neutralizing effect towards SARS-CoV-2, rendering the decoy EVs \nto be potential therapeutics (Cocozza et al., 2020). As a type II transmembrane protein, \nTMPRSS2 had its N-terminus in cytoplasm and EVs lumen side. Therefore, we directly \nfused RFTN1-N15 to TMPRSS2 N-terminus and analyzed protein level in EVs (Figure \n3J). Encouragingly, RFTN1 -N15-TMPRSS2 EVs indeed had elevated TMPRSS2  \ndeposition compared to TMPRSSE-WT EVs (Figure 3K). On the other hand, 4-1BBL \nwas a type II transmembrane glycoprotein that served as the ligand for the receptor 4 -\n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.04.697584doi: bioRxiv preprint \n\n1BB (Chin et al., 2018; Singh, Kim, Lee, Eom, & Choi, 2024). This interaction played \na key role in the immune system by promoting the activation, proliferation, and survival \nof T cells and NK cells, making it a significant target for immunotherapy  (Chin et al., \n2018; Singh et al., 2024) . EVs engineered to overexpress 4-1BBL had presented \nprominent anti -tumor effect in cancer therapy  (Semionatto et al., 2020) . Therefore, \nRFTN1-N15 was similarly fused to 4-1BBL N-terminus, and the results again \nconfirmed that RFTN1 -N15-4-1BBL EVs had increased 4-1BBL density versus 4-\n1BBL-WT EVs (Figure 3L).  \nIn summary, RFTN1-N15 demonstrated versatility in EVs engineering , allowing \nefficient cargo loading including nucleus -localized proteins, cytoplasmic proteins as \nwell as type II transmembrane proteins. \n3.5 Arginase1-loaded EVs demonstrated anti-inflammatory activity both in vitro and in \nvivo. \nIn macrophages, the enzyme arginase 1 (A RG1) and inducible nitric oxide synthase \n(iNOS) competed for the amino acid arginine which determined macrophage function \nand polarization (Chen et al., 2023). ARG1 broke down arginine into ornithine and urea \nleading to tissue repair and inflammation inhibition (M2 state), whereas iNOS produced \nnitric oxide to promote inflammation (M1 state). This polarization balance was crucial \nfor various immune processes, with dysregulation of these enzymes contributing to \ninflammatory diseases or impaired immunity against pathogens (Chen et al., 2023; Luo, \nZhao, Cheng, Su, & Wang, 2024; Orecchioni, Ghosheh, Pramod, & Ley, 2019) . We \nwere curious if EVs could be engineered to deliver these two enzymes to macrophages, \nin order to control the polarization fate of macrophages towards desired directions.  \nTo begin with, RFTN1-N15 was fused to the N-terminus of mouse ARG1 with a flexible \nGS linker (Figure 4A).  Since it was reported that ARG1 assembled as trimer for  \ncatalytic activity (Kanyo, Scolnick, Ash, & Christianson, 1996) , we reasoned that \nadding a trimerization motif to the RFTN1-N15-ARG1 fusion protein could potentially \nhelp to boost enzymatic activity. Hence, a foldon motif was also added to the C -\nterminus of the fusion protein (Meier, Guthe, Kiefhaber, & Grzesiek, 2004) . Both \nmARG1 and mARG1 -foldon EVs were generated, which presented similar size \ndistribution and morphology (Figure 4 A-C). On western blots, it was clear that \nmARG1-foldon had slightly increased protein weight than the non -foldon version as \nexpected (Figure 4D). To validate the effect of engineered EVs on macrophages, \nRAW264.7 cells were first polarized towards M1 phenotype by induction with \nlipopolysaccharide (LPS) and interferon-γ (IFN-γ). Next, EVs were added to those M1 \ncells, followed by qPCR analysis for gene expression changes at 48 hours post treatment. \nInterestingly, mARG1-foldon EVs , but not mARG1 EVs,  presented dose -dependent \nsuppression effect of pro-inflammatory cytokines including IL-1β and IL-12 (Figure \n4E). The observation indicated successful anti -inflammatory modulating effect  by \nARG1 EVs  as expected , and that the trimerization motif was indispensable for the \nengineering.  \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.04.697584doi: bioRxiv preprint \n\nTo confirm that mARG1-foldon EVs had consistent activity in vivo, acute lung injury \nmodel was established in C57BL/6J mice by a single dose of LPS nasal drip, followed \nby EVs treatment for two consecutive days through nebulization (Figure 4F). LPS \ntreatment induced high level of inflammatory cytokines in bronchoalveolar lavage fluid \n(BALF), whereas mARG1-foldon EVs treatment effectively reduced these cytokines \nlevel (Figure 4G). H ematoxylin and eosin (H&E) staining  for lung sections revealed \nthat LPS resulted in apparent alveolar edema and massive infiltration of lymphocytes \ninto alveolar regions, whereas mARG1-foldon EVs treatment significantly alleviated \nthese h istopathological changes (Figure 4H). Therefore, mARG1-foldon EVs  were \neffective in alleviating acute lung inflammation in vivo.  \n3.6 Inducible NOS-loaded EVs demonstrated pro -inflammatory activity both in vitro \nand in vivo. \nOn the other hand, to engineer iNOS EVs, RFTN1-N15 was fused to the N-terminus of \nmouse iNOS with a flexible GS linker (Figure 5A). The catalytically active iNOS was \nin homodimer form (Ghosh & Stuehr, 1995) , hence a dimer version was designed by \nadding GCN4 dimer motif (Harbury, Zhang, Kim, & Alber, 1993) to the C-terminus of \nfusion protein (Figure 5A). No obvious size or morphology differences were observed \nbetween iNOS-EVs and iNOS-GCN4 EVs, except for slightly increased protein size as \nexpected (Figure 5B -D). To determine EVs function in cell model, RAW264.7 cells \nwere polarized towards M2 phenotype by induction with IL-4, followed by treatment \nwith engineered EVs. Dose-dependent elevation of inflammatory cytokines expression \nwere observed for iNOS-GCN4 EVs treatment, whereas iNOS EVs had low activity \n(Figure 5E). In conclusion , efficient pro -inflammatory modulation effect by the \nengineered EVs were as expected, and that the dimerization motif was essential for the \nengineering.  \nFor validating the consistent efficacy in vivo, B16F10 melanoma xenograft model was \nestablished in C57BL6/J mice. The iNOS-GCN4 EVs were intratumorally injected for \nseven days consecutively (Figure 5F). Immune checkpoint inhibitors PD1 antibody was \nincluded as monotherapy, as well as in combination therapy with EVs to see if a \nsynergistic effect could be produced. B16F10 tumor expanded rapidly in PBS treated \ngroups, whereas iNOS-GCN4 EVs significantly inhibited tumor growth as measured \nby tumor size and tumor mass (Figure 5G-H). Notably, iNOS-GCN4 EVs monotherapy \nresulted in similar tumor inhibi tion efficacy to PD1 monotherapy , although the \ncombination therapy did not lead to greater inhibitory effect (Figure 5G-H). In summary, \niNOS-GCN4 EVs demonstrated anti-tumor efficacy in vivo.  \n3.7 Identification of type I transmembrane proteins for EVs surface display. \nWe next moved on to identify scaffold proteins for displaying cargos on EVs surface. \nParticularly, type I transmembrane proteins as exemplified by PTGFRN and PLXNA1 \nhad been validated for applicability, which also appeared as top enriched candidates in \nour EVs proteome (data not shown) . Structurally, type I transmembrane proteins had \nsignal peptide that directed the secretion of N -terminus to extracellular region, where \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.04.697584doi: bioRxiv preprint \n\nthe EVs surface cargos could be appropriately placed. Therefore, we focused on type I \ntransmembrane proteins as potential candidates. Since the length of PTGFRN (879 aa) \nand PLXNA1 ( 1896 aa) were not desirable for the ease of engineering , we set out to \nscreen for candidates that had shorter protein length. \nThe poliovirus receptor (PVR, 417 aa) was filtered out as potential candidate (Figure \n6A). As a type I transmembrane protein, PVR was initially identified as the receptor for \nthe human poliovirus (Mendelsohn, Wimmer, & Racaniello, 1989), and was later found \nto be essential in mediating c ell adhesion , as well as regulating immune response \n(Bowers, Readler, Sharma, & Excoffon, 2017) . Structurally, PVR  belonged to the \nimmunoglobulin (Ig) superfamily (Mendelsohn et al., 1989) , as  it contained several \nimmunoglobulin (Ig)-like domains in tandem at N-terminus (Figure 6A).  \n3.8 PVR displayed human serum albumin on EVs surface with high density, and \nextended EVs circulation half-life. \nTo analyze PVR’s ability for displaying EVs surface cargos, human serum albumin \n(HSA) was first selected for validation for following reasons: (1) HSA was secreted \nprotein, thus could be placed in between PVR ’s secretion signal and Ig -like domains \nfor EVs surface display; (2) HSA was popular platform for drug half-life extension, due \nto its interaction with the recycling receptor neonatal Fc receptor (FcRn)  (Sleep, \nCameron, & Evans, 2013). Upon entering the circulation, EVs were reported with short \nhalf-life due to the rapid clearance by liver and spleen (Imai et al., 2015; Lai et al., 2014; \nLiang et al., 2022; Matsumoto et al., 2020) . Therefore, we tested if displaying HSA \ndirectly on EVs surface could help to increase EVs retention time in circulation (Figure \n6B). A mutant form of human serum albumin ( HSA K573P) was selected for higher \naffinity with FcRn receptor (Andersen et al., 2014) . Surprisingly, the display of HSA \non EV surface by PVR were observed to be at extremely high-density, as the protein \ncorona-like structures were apparent to coat the EVs surface (Figure 6B). As a result, \nthe zeta -potential of EVs changed significantly, from averaging -40 mV of non -\nmodified EVs, to averaging -20 mV for HSA EVs (Figure 6C). Another interesting \nobservation was that the average yield of EVs also increased around 3- to 5-fold, when \nHSA was modified on EVs surface (Figure 6D).  \nThe HSA EVs were labeled with DiR fluorescent dye and intravenously injected in \nC57BL/6 mice. The quantity of remaining EVs in the plasma was analyzed at different \ntime points. As shown in Figure 6E, the amount of HSA EVs in the plasma was average \n2- to 3-fold higher to that of WT EVs at early time points (2 min and 5 min), suggesting \nthat HSA modification successfully extended EVs retention time as expected. \nNonetheless, beyond 15 min, the remaining EVs in the plasma were still too low to be \ndetected (Figure 6E). \n3.9 PVR displayed antibodies on EVs surface, and rendered EVs targeting specificity. \nTo further validate the functionality of PVR, we moved on to test additional EVs surface \ncargos. Displaying antibodies on EVs surface was classic strategy for altering EVs ’ \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.04.697584doi: bioRxiv preprint \n\ntargeting ability , and this strategy  was analyzed independently in our setting for \nefficiency. Atezolizumab and trastuzumab were approved antibody drugs which \ntargeted PD-L1 and HER2 receptor, respectively (Shah, Kelly, Liu, Choquette, & Spira, \n2018; Stebbing, Copson, & O'Reilly, 2000). Single-chain variable fragments (scFv) of \neither antibody were constructed and displayed on EVs surface, by direct fusion with \nPVR scaffold (Figure 7A). The EVs endocytosis analysis was subsequently performed \nin two cancer cell lines: MDA -MB-231 and SKOV3. SKOV3 expressed both PD -L1 \nand HER2 receptors on cell surface  (Ying et al., 2025) , whereas MDA -MB-231 only \nexpressed PD-L1 receptor with high abundancy on surface  (Mittendorf et al., 2014) . \nEqual number of  DiO dye-labeled EVs particles were added to cells, followed by \nfluorescent imaging and flow cytometry analysis at two hours post treatment. \nCompared to non-modified WT EVs, the αPD-L1 EVs had around 10-fold elevation of \nendocytosis rate in MDA -MB-231 cells, whereas the αHER2 EVs had no increase \n(Figure 7A-B). In contrast, both αPD-L1 and αHER2 EVs had around 4-fold higher \nendocytosis rate in SKOV3 cells  (Figure 7A -B). The results indicated that PVR \nsuccessfully displayed antibodies on EVs surface, and importantly rendered EVs with \ndesired targeting ability in vitro.  \nTo analyzed if the EVs targeting ability was consistent in vivo, MDA-MB-231 tumor \nxenograft model was established in nude mice. The αPD-L1 EVs, αHER2 EVs and WT \nEVs were labeled with DiR dye, followed by intravenously injection at equal particle \nnumber (Figure 7C). At four hours post treatment, the EVs distribution in mice were \nimaged and quantified by IVIS. Interestingly, the αPD-L1 EVs demonstrated more than \n50% higher accumulation in tumors when compared to both αHER2 EVs and WT EVs \n(Figure 7D-E). Therefore, the antibody-modified EVs retained targeting ability in vivo \nas well.  \n3.10 Comparison of PTGFRN and PVR for EVs cargo loading efficiency \nFinally, we compared the performance between PVR and the previously reported \nscaffold protein PTGFRN. Western blot  analysis for HSA-PTGFRN and HSA-PVR \nEVs revealed that the fusion protein sizes were as expected (Supplementary Figure 2A). \nELISA quantification revealed that PVR was able to display around 100 HSA molecules \nper EV , which was higher than PTGFRN of around 70 (Supplementary Figure 2B). \nFurthermore, interleukin-15 (IL15) was selected as another cargo which represented \nsecreted cytokines. The fusion proteins sizes presented on WB were also as expected \n(Supplementary Figure 2C). ELISA quantification for IL15 revealed that PVR was able \nto display around 4.3 IL15 molecules per EV , which was again higher than PTGFRN \nof around 1.9 (Supplementary Figure 2 D). We concluded that PVR had better \nperformance than PTGFRN in EVs engineering.  \n4. DISCUSSION \nEVs as novel therapeutic modality have presented promising therapeutic potential. \nEngineering EVs for cargo loading and additional gain of functions critically rely on \nEVs scaffold proteins  (Ma et al., 2025; Yang et al., 2024) . Several studies have \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.04.697584doi: bioRxiv preprint \n\nidentified novel EVs scaffold proteins based on EVs proteomics, with differences in \nEVs purification methodology and candidates sorting criteria (Dooley et al., 2021; Zhao \net al., 2024; Zheng et al., 2023) . Here, we utilized density gradient -based method for \nEVs purification, and performed differential expression -based proteomics analysis for \nidentifying novel EVs scaffolds being enriched from the rest of cellular proteins. As \nsuch, RFTN1 was identified as novel scaffold for EVs luminal loading, and PVR for \nEVs surface display. A variety of cargoes, including enzymes, gene editing tools, \ntransmembrane proteins, serum albumins and antibodies were successfully engineered \nonto the surface or into the lumen of EVs at biologically active levels.  \nMechanistically, we reported that RFTN1 relied on three signals for efficient EVs \nsorting ability: myristoylation on glycine 2, palmitoylation on cysteine 3 and a stretch \nof positively charged residues from lysine 7 to arginine 15. In previous reports, \nMARCKS relied on a single N -terminus lipidation  (Dooley et al., 2021) ; BASP1 \nutilized one lipidation anchor and four lysines at N-terminus (Dooley et al., 2021); Src \nkinase relied on multiple lipid anchors at N-terminus (Whitley et al., 2022). It appeared \nthat RFTN1 had a more balanced combination of amino acid lipidation and charges for \nmembrane association. Accordingly, a minimal sequence derived from the N-terminus \n15 amino acids from RFTN1, were found to be sufficient for EVs cargo engineering. \nWe believe RFTN1-N15 is quite versatile for use as EVs engineering tool.  \nDuring the EVs engineering process of metabolic enzymes including ARG1 and iNOS, \nwe realized that it was essential to consider protein  conformation for retaining \nenzymatic activity. Although a relatively long and flexible GS linker was used to fuse \nRFTN1-N15 with ARG1 or iNOS, the fusion protein delivered by EVs was not active \nunless the appropriate oligomerization motifs were further added. This observation \nshould be instructive for EVs engineering of enzyme cargos in the future.  \nType I transmembrane proteins including PTGFRN and PLXNA1 have been reported \nas EVs scaffolds for displaying surface cargos (Dooley et al., 2021; Zhao et al., 2024), \nyet both proteins have relatively long length, imposing significant difficulties in \nconstruct building and engineering. We reported another type I transmembrane proteins \nPVR, with significant shorter length while keeping efficient EVs sorting ability. \nInterestingly, both PVR and PTGFRN belonged to the immunoglobulin (Ig) \nsuperfamily (Dooley et al., 2021; Mendelsohn et al., 1989) . The observation suggests \nthat proteins from this superfamily may have advantageous structural features for EVs \nenrichment, which could deserve deeper investigation.  \nDeveloping EVs into therapeutics are currently facing several obstacles, with one of \nthese being short circulation half-life (Imai et al., 2015; Lai et al., 2014; Matsumoto et \nal., 2020). A previous report decorated EVs surface with serum albumin binding domain, \nsuch that the EVs would bind serum albumin upon entering the circulation, which  \neffectively extended EVs’ circulation half-life (Liang et al., 2022) . In our report, we \ndirectly displayed human serum albumins (HSA) on EVs surface, which strategy also \nprolonged EVs’ circulation half-life. Interestingly, HSA were found to be displayed on \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.04.697584doi: bioRxiv preprint \n\nEVs surface with high density, and that the serum albumin engineered EVs had higher \nproduction yield. Therefore, serum albumin-displayed EVs could potentially serve as a \nmodular platform with higher stability and lower production cost, which ultimately may \nbenefit the translation of EVs into therapeutics.  \n \nACKNOWLEDGMENTS  \nSome illustrative graphics were created with biogdp.com. This study was funded by \nVesicure Therapeutic. \n \nAUTHORSHIP CONTRIBUTION STATEMENT \nTao Qiu: Writing–original draft, Investigation, Data curation, Conceptualization. Rui \nHu: Investigation. Yuan Yi: Investigation. Wenqiang Lu: Investigation. Chuang Cui : \nInvestigation. Shuiqin Niu : Investigation. Ke Xu: Writing –review & editing, \nSupervision. \n \nDECLARATION OF COMPETING INTEREST \nThe authors declare that they have no known competing financial interests or personal \nrelationships that could have appeared to influence the work reported in this paper.\n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.04.697584doi: bioRxiv preprint \n\nREFERENCES \nAndersen, J. T., Dalhus, B., Viuff, D., Ravn, B. T., Gunnarsen, K. S., Plumridge, A., . . . \nCameron, J. (2014). Extending serum half-life of albumin by engineering neonatal \nFc receptor (FcRn) binding. J Biol Chem, 289 (19), 13492 -13502. \ndoi:10.1074/jbc.M114.549832 \nBowers, J. R., Readler, J. M., Sharma, P., & Excoffon, K. (2017). Poliovirus Receptor: More \nthan a simple viral receptor. Virus Res, 242, 1-6. doi:10.1016/j.virusres.2017.09.001 \nChen, S., Saeed, A., Liu, Q., Jiang, Q., Xu, H., Xiao, G. G., . . . Duo, Y. (2023). Macrophages \nin immunoregulation and therapeutics. Signal Transduct Target Ther, 8 (1), 207. \ndoi:10.1038/s41392-023-01452-1 \nChin, S. M., Kimberlin, C. R., Roe-Zurz, Z., Zhang, P., Xu, A., Liao-Chan, S., . . . Chaparro-\nRiggers, J. (2018). Structure of the 4 -1BB/4-1BBL complex and distinct binding \nand functional properties of utomilumab and urelumab. Nat Commun, 9(1), 4679. \ndoi:10.1038/s41467-018-07136-7 \nCocozza, F., Nevo, N., Piovesana, E., Lahaye, X., Buchrieser, J., Schwartz, O., . . . Martin -\nJaular, L. (2020). Extracellular vesicles containing ACE2 efficiently prevent infection \nby SARS-CoV-2 Spike protein-containing virus. J Extracell Vesicles, 10(2), e12050. \ndoi:10.1002/jev2.12050 \nde Jong, O. G., Murphy, D. E., Mager, I., Willms, E., Garcia-Guerra, A., Gitz-Francois, J. J., . . . \nVader, P. (2020). A CRISPR-Cas9-based reporter system for single-cell detection \nof extracellular vesicle-mediated functional transfer of RNA. Nat Commun, 11(1), \n1113. doi:10.1038/s41467-020-14977-8 \nDooley, K., McConnell, R. E., Xu, K., Lewis, N. D., Haupt, S., Youniss, M. R., . . . Williams, D. \nE. (2021). A versatile platform for generating engineered extracellular vesicles with \ndefined therapeutic properties. Mol Ther, 29 (5), 1729 -1743. \ndoi:10.1016/j.ymthe.2021.01.020 \nDu, S., Guan, Y., Xie, A., Yan, Z., Gao, S., Li, W., . . . Chen, T. (2023). Extracellular vesicles: a \nrising star for therapeutics and drug delivery. J Nanobiotechnology, 21 (1), 231. \ndoi:10.1186/s12951-023-01973-5 \nErbs, P., Regulier, E., Kintz, J., Leroy, P., Poitevin, Y., Exinger, F., . . . Mehtali, M. (2000). In \nvivo cancer gene therapy by adenovirus-mediated transfer of a bifunctional yeast \ncytosine deaminase/uracil phosphoribosyltransferase fusion gene. Cancer Res, \n60(14), 3813-3822.  \nGhosh, D. K., & Stuehr, D. J. (1995). Macrophage NO synthase: characterization of isolated \noxygenase and reductase domains reveals a head -to-head subunit interaction. \nBiochemistry, 34(3), 801-807. doi:10.1021/bi00003a013 \nHarada, N., Tamai, Y., Ishikawa, T., Sauer, B., Takaku, K., Oshima, M., & Taketo, M. M. \n(1999). Intestinal polyposis in mice with a dominant stable mutation of the beta-\ncatenin gene. EMBO J, 18(21), 5931-5942. doi:10.1093/emboj/18.21.5931 \nHarbury, P. B., Zhang, T., Kim, P. S., & Alber, T. (1993). A switch between two-, three-, and \nfour-stranded coiled coils in GCN4 leucine zipper mutants. Science, 262(5138), \n1401-1407. doi:10.1126/science.8248779 \nImai, T., Takahashi, Y., Nishikawa, M., Kato, K., Morishita, M., Yamashita, T., . . . Takakura, \nY. (2015). Macrophage -dependent clearance of systemically administered \nB16BL6-derived exosomes from the blood circulation in mice. J Extracell Vesicles, \n4, 26238. doi:10.3402/jev.v4.26238 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.04.697584doi: bioRxiv preprint \n\nKanyo, Z. F., Scolnick, L. R., Ash, D. E., & Christianson, D. W. (1996). Structure of a unique \nbinuclear manganese cluster in arginase. Nature, 383 (6600), 554 -557. \ndoi:10.1038/383554a0 \nKoyou, H. L., Salleh, M. N., Jelemie, C. S., Badrin, M. J. Q., Prastiyanto, M. E., & \nRamachandran, V. (2025). TMPRSS2: A Key Host Factor in SARS-CoV-2 Infection \nand Potential Therapeutic Target. Medeni Med J, 26 (4), 101 -109. \ndoi:10.4274/MMJ.galenos.2025.40460 \nKumar, M. A., Baba, S. K., Sadida, H. Q., Marzooqi, S. A., Jerobin, J., Altemani, F. H., . . . \nBhat, A. A. (2024). Extracellular vesicles as tools and targets in therapy for diseases. \nSignal Transduct Target Ther, 9(1), 27. doi:10.1038/s41392-024-01735-1 \nLai, C. P., Mardini, O., Ericsson, M., Prabhakar, S., Maguire, C., Chen, J. W., . . . Breakefield, \nX. O. (2014). Dynamic biodistribution of extracellular vesicles in vivo using a \nmultimodal imaging reporter. ACS Nano, 8(1), 483-494. doi:10.1021/nn404945r \nLiang, X., Niu, Z., Galli, V., Howe, N., Zhao, Y., Wiklander, O. P. B., . . . Andaloussi, S. E. \n(2022). Extracellular vesicles engineered to bind albumin demonstrate extended \ncirculation time and lymph node accumulation in mouse models. J Extracell \nVesicles, 11(7), e12248. doi:10.1002/jev2.12248 \nLuo, M., Zhao, F., Cheng, H., Su, M., & Wang, Y. (2024). Macrophage polarization: an \nimportant role in inflammatory diseases. Front Immunol, 15 , 1352946. \ndoi:10.3389/fimmu.2024.1352946 \nMa, Y., Dong, S., Grippin, A. J., Teng, L., Lee, A. S., Kim, B. Y. S., & Jiang, W. (2025). \nEngineering therapeutical extracellular vesicles for clinical translation. Trends \nBiotechnol, 43(1), 61-82. doi:10.1016/j.tibtech.2024.08.007 \nMatsumoto, A., Takahashi, Y., Chang, H. Y., Wu, Y. W., Yamamoto, A., Ishihama, Y., & \nTakakura, Y. (2020). Blood concentrations of small extracellular vesicles are \ndetermined by a balance between abundant secretion and rapid clearance. J \nExtracell Vesicles, 9(1), 1696517. doi:10.1080/20013078.2019.1696517 \nMeier, S., Guthe, S., Kiefhaber, T., & Grzesiek, S. (2004). Foldon, the natural trimerization \ndomain of T4 fibritin, dissociates into a monomeric A -state form containing a \nstable beta-hairpin: atomic details of trimer dissociation and local beta -hairpin \nstability from residual dipolar couplings. J Mol Biol, 344 (4), 1051 -1069. \ndoi:10.1016/j.jmb.2004.09.079 \nMendelsohn, C. L., Wimmer, E., & Racaniello, V. R. (1989). Cellular receptor for poliovirus: \nmolecular cloning, nucleotide sequence, and expression of a new member of the \nimmunoglobulin superfamily. Cell, 56 (5), 855 -865. doi:10.1016/0092 -\n8674(89)90690-9 \nMittendorf, E. A., Philips, A. V., Meric -Bernstam, F., Qiao, N., Wu, Y., Harrington, S., . . . \nAlatrash, G. (2014). PD -L1 expression in triple -negative breast cancer. Cancer \nImmunol Res, 2(4), 361-370. doi:10.1158/2326-6066.CIR-13-0127 \nOrecchioni, M., Ghosheh, Y., Pramod, A. B., & Ley, K. (2019). Macrophage Polarization: \nDifferent Gene Signatures in M1(LPS+) vs. Classically and M2(LPS -) vs. \nAlternatively Activated Macrophages. Front Immunol, 10 , 1084. \ndoi:10.3389/fimmu.2019.01084 \nSaeki, K., Miura, Y., Aki, D., Kurosaki, T., & Yoshimura, A. (2003). The B cell-specific major \nraft protein, Raftlin, is necessary for the integrity of lipid raft and BCR signal \ntransduction. EMBO J, 22(12), 3015-3026. doi:10.1093/emboj/cdg293 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.04.697584doi: bioRxiv preprint \n\nSemionatto, I. F., Palameta, S., Toscaro, J. M., Manrique -Rincon, A. J., Ruas, L. P., Paes \nLeme, A. F., & Bajgelman, M. C. (2020). Extracellular vesicles produced by \nimmunomodulatory cells harboring OX40 ligand and 4 -1BB ligand enhance \nantitumor immunity. Sci Rep, 10(1), 15160. doi:10.1038/s41598-020-72122-3 \nShah, N. J., Kelly, W. J., Liu, S. V., Choquette, K., & Spira, A. (2018). Product review on the \nAnti-PD-L1 antibody atezolizumab. Hum Vaccin Immunother, 14 (2), 269 -276. \ndoi:10.1080/21645515.2017.1403694 \nSingh, R., Kim, Y. H., Lee, S. J., Eom, H. S., & Choi, B. K. (2024). 4 -1BB immunotherapy: \nadvances and hurdles. Exp Mol Med, 56 (1), 32 -39. doi:10.1038/s12276 -023-\n01136-4 \nSleep, D., Cameron, J., & Evans, L. R. (2013). Albumin as a versatile platform for drug half-\nlife extension. Biochim Biophys Acta, 1830 (12), 5526 -5534. \ndoi:10.1016/j.bbagen.2013.04.023 \nStebbing, J., Copson, E., & O'Reilly, S. (2000). Herceptin (trastuzamab) in advanced breast \ncancer. Cancer Treat Rev, 26(4), 287-290. doi:10.1053/ctrv.2000.0182 \nTao Qiu, Y. Y., Rui Hu, Yuan Yi, Guowu Liu, Wenqiang Lu, Xin Zhou, Ke Xu. (2025). \nEncapsulating extracellular vesicles with a minimal RISC complex as novel gene \nsilencing tool. Extracellular Vesicle, Volume 6 . \ndoi:https://doi.org/10.1016/j.vesic.2025.100094 \nvan Niel, G., D'Angelo, G., & Raposo, G. (2018). Shedding light on the cell biology of \nextracellular vesicles. Nat Rev Mol Cell Biol, 19 (4), 213 -228. \ndoi:10.1038/nrm.2017.125 \nWhitley, J. A., Kim, S., Lou, L., Ye, C., Alsaidan, O. A., Sulejmani, E., . . . Cai, H. (2022). \nEncapsulating Cas9 into extracellular vesicles by protein myristoylation. J Extracell \nVesicles, 11(4), e12196. doi:10.1002/jev2.12196 \nYang, C., Xue, Y., Duan, Y., Mao, C., & Wan, M. (2024). Extracellular vesicles and their \nengineering strategies, delivery systems, and biomedical applications. J Control \nRelease, 365, 1089-1123. doi:10.1016/j.jconrel.2023.11.057 \nYing, F., Zhou, X., Chen, M., Huang, L., Gao, L., Zhao, Q., & Zhang, Y. (2025). Preclinical \nstudy of inetetamab combined with atezolizumab to synergistically inhibit HER2 \nand PD-L1 in the treatment of ovarian cancer. Mol Ther Oncol, 33 (1), 200938. \ndoi:10.1016/j.omton.2025.200938 \nZhao, H., Li, Z., Liu, D., Zhang, J., You, Z., Shao, Y., . . . Zhao, L. (2024). PlexinA1 (PLXNA1) \nas a novel scaffold protein for the engineering of extracellular vesicles. J Extracell \nVesicles, 13(11), e70012. doi:10.1002/jev2.70012 \nZheng, W., Radler, J., Sork, H., Niu, Z., Roudi, S., Bost, J. P., . . . El Andaloussi, S. (2023). \nIdentification of scaffold proteins for improved endogenous engineering of \nextracellular vesicles. Nat Commun, 14(1), 4734. doi:10.1038/s41467-023-40453-\n0 \n \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.04.697584doi: bioRxiv preprint \n\n \n \nFigure 1. Purification and proteomic characterization of HEK293-derived EVs.  \n(A) Schematic diagram showing the work flow for purification of HEK293 -derived EVs. (B) \nNanoparticle tracking analysis for purified EVs. (C) Representative transmission electron \nmicrographs of purified EVs. Scale bar: 100 nm. (D) Western blot analysis for cell lysates and \npurified EVs. (E) Differential expression analysis for EVs versus cells proteomics. N=3 . (F) \nV olcano plot showing the top candidates that were differently expressed for EVs versus cells. \n  \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.04.697584doi: bioRxiv preprint \n\n \n \nFigure 2. Validation of RFTN1 and derived minimum sequences as EVs scaffold \nproteins. \n(A) Sequence analysis for the N -terminus 15 amino acids from human RFTN1 protein. (B) \nSchematic diagram showing the EVs engineering of Cre recombinase mediated by RFTN1. (C) \nSchematic diagram of reporter assay designed for detecting Cre recombinase activity . (D) \nWestern blot analysis for engineered Cre EVs from different groups, focusing on the N-terminus \ntruncation of RFTN1. (E) Cre reporter assay analysis at 48 hours post EVs treatment. Scale bar: \n20 μm. (F) Western blot analysis for engineered Cre EVs from different groups, focusing on N-\nterminus mutation of RFTN1. (G) Cre reporter assay analysis at 48 hours post EVs treatment. \nScale bar: 20 μm. \n  \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.04.697584doi: bioRxiv preprint \n\n \n \nFigure 3. RFTN1-N15 demonstrated versatility in EVs engineering and cargo loading. \n(A) Schematic diagram showing the EVs engineering of C as9 mediated by RFTN1-N15. (B) \nCas9 reporter assay analysis at 48 hours post EVs treatment. Scale bar: 20 μm. (C) Cas9 editing \nefficiency analysis as quantified by EGFP positive cells ratio from flow cytometry analysis. \nN=2. Values are plotted as mean ± SD . (D) Schematic diagram showing the EVs engineering \nof the minimal RISC comlpex  mediated by RFTN1-N15. (E) EGFP silencing analysis at 48 \nhours post EVs treatment. Scale bar: 20 μm. ( F) EGFP silencing  efficiency analysis as \nquantified by EGFP mean fluorescence intensity from flow cytometry analysis. N=2.  Values \nare plotted as mean ± SD . (G) Schematic diagram showing the EVs engineering of FCU1 \nmediated by RFTN1-N15. (H) Cellular toxicity assay analysis at 48 hours post EVs treatment. \nScale bar: 100 μm. (I) Cell viability as quantified by CCK8 assay. N=2. Values are plotted as \nmean ± SD . (J) Schematic diagram showing the EVs engineering of type II transmembrane \nproteins mediated by RFTN1-N15. (K) Western blot analysis for engineered TMPRSS2 EVs. \n(L) Western blot analysis for engineered 4-1BBL EVs. \n  \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.04.697584doi: bioRxiv preprint \n\n \n \nFigure 4. Arginase1-loaded EVs demonstrated anti-inflammatory activity both in vitro \nand in vivo. \n(A) Schematic diagram showing the EVs engineering of mouse arginase1 protein mediated by \nRFTN1-N15 and foldon trimerization motif . (B) Nanoparticle tracking analysis for purified \nEVs. (C) Representative transmission electron micrographs of purified EVs. Scale bar: 100 nm. \n(D) Western blot analysis for purified EVs. (E) QPCR analysis for IL -1β and IL -12 mRNA \nchanges in M1 macrophages receiving different treatment groups. Four doses of EVs from total \n2.5E9 particles to 2E10 particles were tested. N=3. Values are plotted as mean ± SD . (F) \nSchematics of acute lung injury model in C57BL/6J mice. A single dose of LPS nasal drip was \nadministrated at day 0, followed by EVs administration for two consecutive days by nebulization. \nMice were euthanized at day 3 for analysis. (G) Pro-inflammatory cytokines analysis in BALF from \ndifferent groups. N=6. **: p value < 0.01;***: p value < 0.001;****: p value < 0.0001. Values are \nplotted as mean ± SD . (H) H&E analysis for lung sections from different groups. Scale bar: 100 \nμm.  \n  \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.04.697584doi: bioRxiv preprint \n\n \n \nFigure 5. Inducible NOS-loaded EVs demonstrated pro -inflammatory activity both in \nvitro and in vivo.  \n(A) Schematic diagram showing the EVs engineering of mouse iNOS protein mediated by \nRFTN1-N15 and GCN4 dimerization motif. (B) Nanoparticle tracking analysis for purified EVs. \n(C) Representative transmission electron micrographs of purified EVs. Scale bar: 100 nm. (D) \nWestern blot analysis for purified EVs. (E) QPCR analysis for IL-1β and IL-12 mRNA changes \nin M 2 macrophages receiving different treatment groups. Four doses of EVs from total 1.25E9 \nparticles to 1E10 particles were tested. N=3. Values are plotted as mean ± SD. (F) Schematics of \nB16F10 xenograft model in C57BL/6J mice. B16F10 cells were first implanted subcutaneously and \ngrew to appropriate size, followed by EVs intratumorally  injection for 7 consecutive days. Mice \nwere euthanized at day 7 for analysis. (G) Tumor mass measurement at end point for different \ntreatment groups. N=5. **: p<0.01; ***: p value < 0.001 . Values are plotted as mean ± SD . (H) \nImages of tumor at end point for different treatment groups. N=5. \n  \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.04.697584doi: bioRxiv preprint \n\n \n \nFigure 6. PVR displayed human serum albumin on EVs surface with high density, and \nextended EVs circulation half-life.  \n(A) Schematic diagram showing the domain organization of PVR. (B) Upper panel: schematic \ndiagram showing the EVs engineering of human serum albumin (HSA) through PVR scaffold. \nLower panel: TEM and cryo -EM analysis for EVs. Scale bar: 10 0 nm. (C) Zeta potential \nanalysis for WT  EVs versus HSA EVs. N=5. Values are plotted as mean ± SD . ( D) Yield \nanalysis for WT EVs versus HSA EVs. N=5. Values are plotted as mean ± SD . (E) The EVs \nwere intravenously injected in C57BL/6 mice, and the quantity of remaining EVs in the plasma \nwas analyzed at different time points. N=2. Values are plotted as mean ± SD. \n \n  \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.04.697584doi: bioRxiv preprint \n\n \n \nFigure 7.  PVR displayed antibodies on EVs surface, and rendered EVs targeting \nspecificity.  \n(A) Upper panel: s chematic diagram showing the EVs engineering of single-chain variable \nfragments (scFv) through PVR scaffold. Lower panel: endocytosis analysis for MDA-MB-231 \n(PD-L1+ HER2-) and SKOV3 (PD-L1+ HER2+) at two hours post EVs treatment. Scale bar: 10 \nμm. (B) Relative endocytosis efficiency quantified by mean fluorescence intensity from flow \ncytometry. N=2. Values are plotted as mean ± SD . (C) The nude mice were implanted with \nMDA-MB-231 xenograft, and imaged at four hours post EVs I.V . infusion. (D) The liver, spleen \nand tumor tissues were imaged at four hours post EVs I.V . infusion. (E) The amount of EVs in \neach tissue as quantified by total radiant intensity. Values are plotted as mean ± SD.  \n \n  \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.04.697584doi: bioRxiv preprint \n\n \n \nSupplementary Figure 1 . Comparison of BASP1 and RFTN1 in EVs engineering \nefficiency.  \n(A) Cre reporter assay analysis at 48 hours post EVs treatment.  BASP1-Cre and RFTN1-Cre \nEVs were added to reporter cells at equal quantity (1E10 EVs particles were added to 3E4 \nreporter cells). (B) TOPFLASH reporter assay analysis at 48 hours post EVs treatment. BASP1-\nβ-catenin (ΔEX3) and RFTN1-β-catenin (ΔEX3) EVs were added to reporter cells at equal \nquantity (1E10 EVs particles were added to 3E4 reporter cells). N=2. V alues are plotted as mean \n± SD. \n \n  \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.04.697584doi: bioRxiv preprint \n\n \n \nSupplementary Figure 2. Comparison of PTGFRN and PVR in EVs engineering \nefficiency. \n(A) Western blot analysis for purified EVs loaded with HSA by PTGFRN and PVR. (B) ELISA \nquantification analysis for HSA loading number per EV . N=2. Values are plotted as mean ± SD. \n(C) Western blot analysis for purified EVs loaded with IL15 by PTGFRN and PVR. (D) ELISA \nquantification analysis for IL15 loading number per EV . N=2. Values are plotted as mean ± SD. \n \n \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.04.697584doi: bioRxiv preprint","source_license":"CC-BY-4.0","license_restricted":false}