Highly pure (> 98%) recombinant p17 protein (clade B isolate BH-10, amino acids 1-132, GenBank accession number M15654)^[27] was produced in lipopolysaccharide (LPS)-free and biologically active monomeric form, as previously described^[28]. Briefly, the coding sequence of p17 was cloned into the BamHI and EcoRI sites of the prokaryotic expression vector pGEX-2T (GE Healthcare / Cytiva, Marlborough, MA, USA), which ensures the production of large amounts of biologically active recombinant proteins. The glutathione S-transferase (GST)-linked p17 protein was expressed in BL21(DE3), an appropriate Escherichia coli strain for high-level protein production, by adding 0.8 mM isopropyl-β-D-thiogalactopyranoside (IPTG) for 4 h at room temperature, once the OD₆₀₀ reached a value of 0.6. The viral protein was purified by using Glutathione Sepharose 4B beads and cleaved from GST by thrombin. The p17 protein was further purified by reverse-phase Fast Protein Liquid Chromatography (FPLC). We analyzed protein fractions for purity by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and identified them by Western blot using monoclonal antibodies to p17, such as MBS-3 [recognizing amino acid(s) (aa) 9-22]. Protein fractions were quantified by bicinchoninic acid (BCA) assay. The absence of endotoxin contamination in the recombinant p17 batches (< 0.125 EU/mL) was assessed by Limulus Amebocyte Lysate (LAL) assay (Associates of Cape Cod, Inc., East Falmouth, MA, USA).

An integrated bioinformatic approach was employed to design p17-derived peptides and identify minimal functional epitopes with pro-angiogenic activity. Starting with the HIV-1 matrix protein p17 (clade B isolate BH-10), the sequences corresponding to fragments F2 (amino acids 17-36: EKIRLRPGGKKKYKLKHIVW) and F3 (amino acids 32-52: KHIVWASRELERFAVNPGLLE) were analyzed.

Electrostatic and physicochemical properties, including molecular charge and hydrophobicity, were assessed using online resources such as UniProtKB (http://www.uniprot.org/), BIOPEP (http://www.uwm.edu.pl/biochemia/index.php/en/biopep), and PeptideRanker (http://bioware.ucd.ie/~compass/biowareweb/Server_pages/peptideranker.php). Hydrophobic (Ile34, Val35, Trp36) and nonpolar residues (Pro23, Gly24, Gly25), considered less likely to be involved in receptor interactions, were excluded during peptide refinement.

Structural accessibility of selected residues was verified using the p17 crystal structures available in the Protein Data Bank (PDB entries 1TAM for monomeric p17 and 1HIW for trimeric p17) to ensure extracellular exposure. Additionally, phylogenetic analysis and sequence alignment of HIV-1 p17 F3 with human EPO, and with the corresponding HIV-2 p17 fragment, were performed to further shorten the F3 sequence.

All peptides were chemically synthesized to > 98% purity (Thermo Fisher Scientific, Waltham, MA, USA). We calculated theoretical solubility and stability parameters, including isoelectric point (pI) and net charge at pH 7.0, using the ExPASy “Compute pI/Mw” tool and the Peptide Property Calculator (Peptide 2.0 Inc., Chantilly, VA, USA).

Previously described p17-derived peptides and those identified through bioinformatic analysis and used for EV engineering in this study are listed in Table 1; their functional characterization is detailed in the RESULTS section.

A schematic representation of the amino acid sequence of the HIV-1 p17 matrix protein and its derived synthetic fragments is shown in Figure 1.

The coding sequence of canine CD63 was amplified by high-fidelity polymerase chain reaction (PCR) from a previously described plasmid^[23] using Q5® High-Fidelity DNA Polymerase (New England Biolabs, Ipswich, MA, USA). Primers were designed to introduce HindIII and XmaI restriction sites (forward: GGCAAGCTTCCATGGCGGTGGAAGG; reverse: AAAACCCGGGATGACTTCATAGCCACTTCGG). PCR amplification was performed according to the manufacturer’s instructions, and the resulting amplicon was purified using a PCR Clean-Up kit (Zymo Research, Irvine, CA, USA). The purified CD63 fragment was then subcloned into the pTagGFP-2N expression vector (Evrogen, Moscow, Russia).

Mutated CD63 constructs were generated using a two-step overlap extension mutagenic PCR strategy. Briefly, two separate PCR reactions were carried out to amplify the N-terminal and C-terminal portions of CD63 using primers introducing AgeI and KpnI restriction sites. PCR products were purified, mixed in equimolar amounts, and subjected to an extension PCR to obtain the full-length mutated CD63 sequence. Amplified products were purified, quantified by spectrophotometry (NanoDrop^TM, Thermo Fisher Scientific), digested with the appropriate restriction enzymes (New England Biolabs), and ligated into the pTagGFP-2N vector using T4 DNA Ligase (New England Biolabs).

Ligation products were transformed into chemically competent Escherichia coli XL1-Blue cells by heat shock and plated on Luria-Bertani agar supplemented with kanamycin (50 micrograms/mL; Sigma-Aldrich, St. Louis, MO, USA). Bacterial colonies were screened by PCR, and positive clones were verified by Sanger sequencing.

Recombinant constructs were generated to express CD63 fused to green fluorescent protein (GFP) and functionalized with p17-derived peptides. GFP served as a reporter to detect functionalized chimeric constructs by fluorescence-based techniques (flow cytometry and fluorescence microscopy) and as a target for immunological detection (Western blotting and dot blot). Indeed, the short p17-derived peptides could not be reliably used as detection targets themselves, and CD63 was equally unsuitable, being a natural, endogenous component of EVs; the GFP tag therefore provided the only reliable means of distinguishing the engineered constructs from endogenous vesicle components. The empty backbone vector encoding CD63 fused to GFP without any peptide insert (pTagGFP2N-CD63) was used to generate control EVs (EV-CD63-GFP). This construct served as an engineered baseline control to account for potential non-specific effects associated with genetic manipulation and GFP tag expression.

The second extracellular loop of CD63 was selected as the insertion site for peptide exposure, based on structural topology analysis [Figure 2]. Plasmids encoding CD63 fused to p17-derived oligopeptides were obtained by ligating annealed complementary oligonucleotides into the KpnI and AgeI restriction sites previously introduced into the mutated CD63 backbone. Briefly, forward and reverse oligonucleotides were mixed at equimolar concentrations, heated to 80 °C, and slowly cooled to room temperature to allow correct base pairing before ligation. Bacterial transformation, clone selection, and sequence verification were performed as described above. Oligonucleotide sequences encoding p17-derived peptides are reported in Table 1. As the F3 40-44 peptide failed to show biological activity in vitro (see the RESULTS section), the corresponding construct was generated, but EVs exposing this peptide were not produced.

MSCs were isolated from canine adipose tissue collected for therapeutic purposes. Surplus cells not required for clinical application were used for this study with the informed consent of the owner. The study protocol was reviewed and approved by the Comitato Universitario di Bioetica dell’Università degli Studi di Perugia (Bioethics Committee of the University of Perugia), protocol number 6/2026.

MSC isolation was performed according to a protocol previously described by Scattini et al.^[29] and further standardized for EV production and genetic manipulation^[23].

Briefly, adipose tissue samples were enzymatically digested with 0.1% Collagenase I (Worthington Biochemical Corp., Lakewood, NJ, USA) and plated to allow stromal cell adhesion. MSCs were expanded in Dulbecco’s Modified Eagle Medium- low glucose (DMEM-LG, Euroclone S.p.A., Milano, Italy) supplemented with 10% fetal bovine serum (FBS; Euroclone S.p.A., Milan, Italy) and 1% penicillin/streptomycin (Thermo Fisher Scientific, Waltham, MA, USA) in TPP tissue-culture flasks (TPP Techno Plastic Products AG, Trasadingen, Switzerland) and maintained at 37 °C in a humidified atmosphere with 5% CO₂. Culture medium was replaced every 2-3 days, and cells were used between passages 3 and 5. MSCs were seeded to reach approximately 60% confluence at the time of transfection. Cells were transfected using Lipofectamine^TM Stem Transfection Reagent (Thermo Fisher Scientific, Waltham, MA, USA) with plasmid DNA at a final amount of 250 ng/cm², complexed with the transfection reagent at a 1:4 (DNA: reagent) ratio in DMEM-LG. After 6 h of incubation, FBS was added to restore a final concentration of 10%. 24 h after transfection, the medium was replaced with serum-free DMEM-LG to reduce contamination by FBS EVs, and cells were incubated for an additional 48 h to allow the release of engineered EVs.

EV isolation and characterization were performed in accordance with the Minimal Information for Studies of Extracellular Vesicles 2023 (MISEV2023) guidelines^[30]. Conditioned media from MSCs transfected with different constructs were collected 72 h after transfection and processed. Briefly, culture supernatants were centrifuged at 2,000 × g for 15 min at 4 °C to remove detached cells and cellular debris. The clarified supernatant was subsequently concentrated using Vivaspin® ultrafiltration units with a 100 kDa molecular-weight cut-off (Sartorius, Göttingen, Germany), centrifuging at 2,000 × g at 4 °C for 30 min, repeating centrifugation cycles as needed to reach a final concentration factor of approximately 100-150-fold. Concentrated samples were washed twice with sterile 0.1 micrometer-filtered phosphate-buffered saline (PBS; Thermo Fisher Scientific, Waltham, MA, USA) and subjected to a second ultrafiltration step.

All wound healing and tube formation assays were performed in three independent biological replicates (cell passages 5-7), with each condition tested in triplicate. Data are expressed as mean ± standard deviation (SD). Statistical significance among multiple treatment groups was evaluated using one-way analysis of variance (ANOVA) followed by Bonferroni’s post hoc test for pairwise comparisons. All statistical analyses were performed using GraphPad Prism software (version 8.0, GraphPad Software, La Jolla, CA, USA). In accordance with international reporting guidelines, exact P-values are reported throughout the text and figure legends. Differences were considered statistically significant at P < 0.05. Statistical significance with a calculated P-value lower than 0.0001 is explicitly indicated as ^****P < 0.0001.

The HIV-1 p17 F2 fragment (aa 17-36: EKIRLRPGGKKKYKLKHIVW) has been shown to be involved in angiogenic activity under stress conditions^[26]. To identify the smallest peptides derived from the F2 sequence capable of inducing angiogenesis, a structure-function analysis was performed. This analysis identified two motifs that are critical for receptor binding: an ELR-like motif (RLR aa 20-22), known to mediate CXCR1 interaction^[31], and a highly polybasic region (KKKYKLKHI aa 26-34) involved in binding to heparan sulfate proteoglycans (HSPGs) and CXCR1^[32].

Accordingly, the F2 fragment was refined by excluding non-essential hydrophobic/nonpolar residues (Ile34, Val35, Trp36; Pro23, Gly24, Gly25), which are less likely to participate in receptor interactions, and by focusing on the acidic (Glu) and basic (Arg, Lys) residues that typically drive such interactions. Based on these considerations, two sub-fragments were derived as potential candidates for triggering angiogenesis under stress conditions: F2 17-22 (EKIRLR) and F2 26-33 (KKKYKLKH) [Figure 1]. Structural analysis further confirmed that the proposed sub-fragments are exposed to the extracellular environment in both the monomeric and trimeric p17 protein structures (PDB entries 1TAM and 1HIW), supporting their potential for receptor interaction.

Previous studies on the HIV-1 p17 F3 fragment (aa 32-52: KHIVWASRELERFAVNPGLLE), which induces angiogenesis under both stress and non-stress conditions through molecular mimicry with EPO, led to the identification of two angiogenic fragments, F3 37-52 and F3 37-44 (an 8-amino-acid sequence ASRELERF)^[26] [Figure 1].

To identify the minimal functional sequence, a comparative analysis of HIV-1, HIV-2, and EPO sequences was performed, focusing on conserved residues and conservative amino acid substitutions with similar physicochemical properties.

The HIV-2 sequence was included in the analysis because the p17 proteins of both HIV-1 and HIV-2 contain a functional angiogenic motif within their respective F3 37-44 regions (ASRELERF for HIV-1 and AANELDRF for HIV-2), both of which promote angiogenesis in HUVECs despite sequence divergence^[26]. This comparative sequence alignment highlighted a conserved V/ELERx motif (aa 40-44) [Figure 1]. Consequently, the pentapeptide F3 40-44 (ELERF) was selected as a promising candidate motif for promoting angiogenesis. The C-terminal phenylalanine (Phe44) was initially retained because it is structurally conserved across viral variants^[26], suggesting a key functional role in mediating biological activity.

Finally, theoretical solubility profiling indicated that these shortened, hydrophilic peptides, characterized by a high proportion of charged residues (> 25%) and a low hydrophobic content (< 25%), exhibit good solubility in aqueous solutions, a critical feature for their potential application as angiogenic factors.

The biological activity of the identified p17-derived peptides was evaluated in vitro in HUVECs at 10 ng/mL using wound healing and tube formation assays.

In the wound healing assay, HUVECs starved under stress conditions for 16 h and treated with the reference peptide F2, as well as with its derived fragments F2 17-22 and F2 26-33, showed enhanced wound closure within 12 h of treatment. The full-length p17 protein, used as a positive control, induced a significant pro-migratory effect on HUVECs relative to untreated cells, comparable to that observed with F2 and its derived fragments. As expected, the scrambled peptide negative control did not promote wound closure [Figure 3A].

In contrast, the candidate peptide ELERF (F3 40-44), derived from F3, failed to exhibit pro-migratory activity in wound healing assays under stress conditions. Conversely, the peptides F3, F3 37-52, and F3 37-44, used as positive controls, significantly promoted endothelial repair and complete wound closure, similar to p17, F2, and F2-derived fragments [Figure 3A]. The angiogenic activity of the F2-derived (F2 17-22 and F2 26-33) and F3-derived peptides was then evaluated using tube formation assays in HUVECs cultured under stress conditions for 16 h. F2-derived peptides, similar to F2, significantly induced tubular structure formation after 8 h of treatment compared with untreated control cells. In contrast, F3 40-44 was unable to promote tube formation, like the scrambled peptide, whereas F3, F3 37-52, and F3 37-44 exhibited angiogenic activity comparable to that of the full-length p17 protein [Figure 3B].

Consistent with these findings, wound healing and tube formation assays were also performed under standard culture conditions to further assess the lack of pro-migratory and angiogenic activity of the F3-derived fragment F3 40-44. Under these conditions, F3 40-44 did not exhibit any detectable activity [Figure 4], showing results comparable to untreated cells and cells treated with the scrambled peptide.

While the pentapeptide F3 40-44 (ELERF) was initially selected based on its conservation across HIV-1, HIV-2, and EPO, its subsequent lack of in vitro activity prompted us to evaluate the adjacent pentapeptide F3 39-43 (RELER). We hypothesized that the new peptide would benefit from both the introduction of the basic arginine 39 (Arg39) and the removal of the bulky hydrophobic phenylalanine 44 (Phe44)-despite the conservation of Phe44 in HIV-1 and HIV-2-to generate a palindromic sequence. This sequence adjustment allowed us to test whether shifting the sequence by a single residue, while maintaining a 5-amino-acid length, could restore the pro-angiogenic activity. Treatment with 10 ng/mL F3 39-43 resulted in a significant increase in endothelial cell migration, leading to complete wound closure, and in the formation of tubular structures. These effects were observed under both starved [Figure 5A and B] and standard culture conditions [Figure 5C and D].

The activity of F3 39-43 was comparable to that of the reference peptide F3 and significantly higher than that observed in untreated cells or in cells treated with the scrambled peptide used as a negative control.

Canine MSCs exhibited a typical fibroblast-like morphology and strong adherence to plastic throughout the culture period [Figure 6A and B]. Cells expanded efficiently under standard conditions and maintained a homogeneous morphology within the passage range used for EV production (P3-P5). No evident cytotoxic effects or major morphological alterations were observed following transfection with CD63-based constructs [Figure 6C].

Replacement of the culture medium with serum-free DMEM-LG 24 h after transfection did not visibly affect cell morphology and enabled the collection of EV-enriched conditioned media after an additional 48 h. Fluorescence microscopy confirmed the expression of CD63-GFP and peptide-functionalized variants in transfected MSCs, showing a localization consistent with vesicular compartments [Figure 6C].

EVs were successfully isolated from conditioned media by ultrafiltration. NTA showed that both untransfected MSC-EVs (UT) and engineered EVs displayed size distributions consistent with small EV populations. Most particles were detected within the 80-150 nm range, with comparable modal diameters among untransfected EVs, EV-CD63-GFP, and EVs functionalized with p17-derived peptides [Figure 7A]. Quantitative NTA revealed that overall EV yields were not substantially affected by MSC transfection or by expression of CD63-based fusion constructs. Similar particle concentrations were obtained from untransfected MSCs, MSCs expressing CD63-GFP, and MSCs engineered to express CD63-peptide constructs. NTA revealed particle concentrations in the range of 10¹⁰-10¹¹ particles/mL across all concentrated samples, corresponding to 2-7 × 10⁸ particles/mL in the MSC-conditioned media, with an average yield of 0.8-1.4 × 10⁹ particles per million cells [Figure 7B], supporting the suitability of the obtained EV preparations for downstream analyses.

Western blot analysis confirmed enrichment of EV-associated markers (Alix) and depletion of cellular contaminants (Mitofilin, a mitochondrial marker, and GAPDH, a cytosolic marker) in the EV fraction compared with the corresponding cell lysates, consistent with a bona fide EV preparation largely free of cellular debris. GFP detection in cell lysates confirmed expression of the CD63-GFP fusion constructs, while its presence in the EV fraction - restricted here to representative populations (CD63-mut, F2, F3) - further validated the successful sorting of the engineered constructs into EVs [Figure 7C].

Electron microscopy further confirmed the identity of the isolated particles. Both SEM and TEM analyses revealed round-shaped vesicles consistent with the typical morphology of EVs, free of major aggregates or cellular debris, with sizes in agreement with the range determined by NTA [Figure 7D].

Flow cytometry analysis was performed to assess EV-associated GFP signal across conditions. Due to the application of a fluorescence threshold on the FITC channel, the acquisition was biased toward GFP-positive events, while only a fraction of non-fluorescent particles was detected. As a result, the percentage of positive events could not accurately reflect the true proportion of fluorescent EVs and was therefore not used for quantitative comparisons [Figure 8A]. Under identical dilution and acquisition settings for all samples, the relative number of GFP-positive EVs was instead used for comparative analysis. As shown in Figure 8B, CD63-GFP and F3-tagged CD63 samples exhibited the highest number of GFP-positive EVs. Notably, the F3 39-43 variant also showed relatively high levels of GFP-positive EVs, approaching those observed for the CD63-GFP control. In contrast, F2 constructs displayed generally lower counts, indicating reduced or variable labeling efficiency, a finding also confirmed by Western blotting and dot blot results on EVs carrying the F2 peptide [Figure 8C and D]. Details on the gating strategy, instrument calibration, and Triton X-100 controls are provided in Supplementary Figure 1.

Western blotting analysis confirmed the presence of peptide-tagged CD63 in the EVs derived from transfected MSCs [Figure 8C]. The CD63-mut construct (CD63-GFP lacking peptide insertion) appeared as a single signal at approximately 75 kDa, whereas the peptide-tagged constructs appeared as two distinct signals, one at 75 kDa and one at approximately 40-45 kDa, with the relative proportion of the two bands varying between samples. Positivity to Alix, Syntenin (SDCBP), and CD81 confirmed the identity of the EVs, while the GFP dot blot confirmed the presence of GFP-tagged, engineered EVs, consistent with the flow cytometry results [Figure 8A], and allowed quantification of the GFP signal used in Figure 8D. The relative GFP abundance in EV samples was calculated as the ratio between the GFP signal obtained by dot blot and the Alix signal obtained via Western blotting [Figure 8D].

The angiogenic potential of EVs functionalized with p17-derived peptides (EV-F2, EV-F2 17-22, EV-F2 26-33, EV-F3, EV-F3 37-52, EV-F3 37-44, and EV-F3 39-43) was assessed in HUVECs using a tube formation assay. Native EVs (nEVs) and EV-CD63-GFP were used as negative controls.

As shown in Figure 9, functionalization of EVs with F2 and F3 fragments significantly enhanced the formation of capillary-like structures compared with the nEV and EV-CD63-GFP control groups (^****P < 0.0001). Notably, both shorter sub-fragments derived from F2 (EV-F2 17-22 and EV-F2 26-33) failed to induce a significant angiogenic response. In contrast, all F3-derived sub-fragments (EV-F3 37-52, EV-F3 37-44, and EV-F3 39-43) retained strong pro-angiogenic activity, comparable to that of the parental F3 fragment [Figure 9].

In this proof-of-concept study, our primary objective was to demonstrate in vitro the technological feasibility of displaying engineered functional peptides on the EV surface to drive angiogenesis.

While the long-term goal of this platform is its translational application as a therapeutic nanocarrier, a limitation of the present study is that it remains confined to an in vitro setting. Several intermediate steps will therefore be required before in vivo application can be considered. Mechanistic studies addressing receptor engagement, binding kinetics, and internalization dynamics of the EV-displayed peptides, as discussed below, will be essential to substantiate the proposed mode of action. In addition, more complex, physiologically relevant models - such as 3D co-culture or ex vivo tissue explant systems - could help bridge the gap between simplified 2D endothelial assays and the multicellular, dynamic environment of native tissue. Indeed, the wound healing (scratch) and tube formation assays employed here, while widely used and informative, remain surrogate, single-cell-type readouts of angiogenesis; they do not capture the contribution of pericytes, immune cells, or extracellular matrix remodeling that shape the angiogenic response in vivo, and the present results should be interpreted within this constraint.

Moreover, the functional characterization of the engineered EVs was restricted to HUVECs; potential off-target effects on other cell populations present in the wound microenvironment, such as fibroblasts or resident immune cells, were not investigated and should be addressed in future safety-oriented studies.

Dose-response and pharmacokinetic characterization, together with assessment of EV stability and reproducibility across production batches, and robustness of the engineering strategy across different MSC biological backgrounds, will also be necessary before translational application can be considered.

Closely related to this, the manufacturing scalability of this EV-engineering approach has not been addressed in the present study; translating peptide-displaying EVs from small-scale laboratory production to a large-scale, GMP-compliant process suitable for clinical development represents an additional, non-trivial technical hurdle.

Moreover, in vivo validation will require the prior identification of a disease-relevant setting - including the specific pathological condition and animal model best suited to test this platform. Extending the platform in this direction could represent a valuable line of future work.

The rational design of EV-based delivery platforms relies on translating known bioactive motifs onto a nanovesicular scaffold. We previously demonstrated that angiogenesis induced by the p17-derived fragments F2 and F3 37-52 requires distinct receptors - CXCR1, CXCR2, and the βc receptor^[24,26]. Consequently, each engineered construct may exhibit unique surface docking, entry kinetics, and downstream crosstalk. Although the interaction and internalization of MSC-derived EVs by endothelial cells are well-established phenomena^[38-40] and our functional readouts confirm that EV-displayed peptides successfully trigger angiogenic responses, a current limitation of this study is the lack of precise molecular mechanisms and receptor-mediated dynamics for the EV-displayed constructs.

In particular, receptor-blocking or neutralization experiments, which previously confirmed the involvement of CXCR1, CXCR2, and the βc receptor for the free peptide counterparts, were not performed for the EV-displayed constructs, and therefore cannot yet formally establish that the same receptors mediate the angiogenic activity observed in the engineered EV format.

Mapping the spatio-temporal dynamics of EV internalization and determining whether these interactions occur exclusively at the plasma membrane or dynamically persist within the endosomal network represent an intriguing area for future investigation, building on a fully characterized baseline of the binding dynamics and signaling of the soluble, free peptide counterparts, which is currently under active investigation.

Future studies may also explore alternative scaffold proteins, optimized linker sequences, or different insertion sites to further improve peptide accessibility and functional density on the EV surface. In addition, while the present in vitro results are promising, future in vivo translation will need to account for the complexity of the wound microenvironment. In particular, the formation of a biomolecular corona and the presence of proteolytic enzymes may hinder peptide bioavailability and receptor recognition, warranting further investigation for clinical tissue-repair applications^[41].

Direct molar equivalence between free peptides and EV-displayed constructs was not applied in our bioassays, as concentrations were chosen based on optimal biological responsiveness. Owing to the high surface density of displayed peptides on the vesicle membrane, restricted spatial orientation, and increased local avidity, EV-bound ligands may engage target-cell receptors differently than freely soluble, monovalent counterparts^[42-45]. In addition, the absolute surface density of the displayed peptides and the fraction of EVs effectively carrying the engineered constructs were not quantitatively determined. Therefore, differences in biological activity among EV formulations cannot be unequivocally separated from potential differences in peptide copy number or display efficiency. Quantitative studies using advanced biophysical tools will be needed to precisely map peptide-per-vesicle stoichiometry, binding kinetics, spatial stoichiometry, and dose-response dynamics of functionalized EVs.

The cross-species experimental design, combining canine MSCs, previously optimized for EV production and engineering, with human HUVECs as a standardized model for angiogenic assessment, introduces a degree of biological heterogeneity that cannot be fully controlled. While EV-mediated intercellular communication is thought to be largely conserved across species, species-specific differences in receptor structure, signaling components, or membrane composition cannot be excluded and may have influenced the magnitude of the responses observed. Results obtained in this setting will therefore need to be confirmed using matched-species models before any translational extrapolation.

Finally, using GFP as a reporter and detection tag, while methodologically necessary to track and quantify the engineered constructs, introduces a protein sequence that is itself potentially immunogenic and is not part of the intended therapeutic cargo. Future iterations of this platform aimed at translational applications should consider alternative, non-immunogenic detection strategies, or removal of the reporter tag once construct validation has been achieved.

In conclusion, this study advances the understanding of structure-function relationships governing peptide activity in EV-based delivery systems and highlights the critical role of peptide structure, orientation, and accessibility in determining functional outcomes. Taken together, these observations support the use of p17-derived peptides as a source of unconventional bioactive sequences and provide generalizable design principles for future EV engineering strategies targeting angiogenesis-related applications.