https://orcid.org/0000-0002-0068-8429
Abstract
Lackluster results from recently completed gene therapy clinical trials of VEGF-A delivered by viral vectors have heightened the need to develop alternative delivery strategies. This study aims to demonstrate the pre-clinical efficacy and safety of extracellular vesicles (EVs) loaded with VEGF-A mRNA for the treatment of ischaemic vascular disease.
After encapsulation of full-length VEGF-A mRNA into fibroblast-derived EVs via cellular nanoporation (CNP), collected VEGF-A EVs were delivered into mouse models of ischaemic injury. Target tissue delivery was verified by in situ analysis of protein and gene expression. Functional rescue was confirmed by in vivo imaging and histology. The safety of single and serial delivery was demonstrated using immune-based assays.
VEGF-A EVs were generated with high mRNA content using a CNP methodology. VEGF-A EV administration demonstrated expression of exogenous VEGF-A mRNA by in situ RNA hybridization and elevated protein expression by western blot, microscopy, and enzyme-linked immunosorbent assay. Mice treated with human VEGF-A EVs after femoral or coronary artery ligation exhibited heightened neovascularization in ischaemic tissues with increased arterial perfusion and improvement in left ventricular function, respectively. Serial delivery of VEGF-EVs in injured skin showed improved wound healing with repeat administration. Importantly, as compared with adeno-associated viral and lipid nanoparticle VEGF-A gene therapy modalities, murine VEGF-A EV delivery did not trigger innate or adaptive immune responses at the injection site or systemically.
This study demonstrated that VEGF-A EV therapy offers efficient, dose-dependent VEGF-A protein formation with low immunogenicity, resulting in new vessel formation in murine models of ischaemic vascular disease.
After encapsulation of full-length VEGF-A mRNA into extracellular vesicles (VEGF-A EVs) by cellular nanoporation, VEGF-A EVs were delivered by injection into the target tissue (e.g. myocardium, skeletal muscle, and skin for treatment of myocardial ischaemia, chronic limb ischaemia, and dermal wounds, respectively), VEGF-A EVs stimulated angiogenesis and improved tissue healing, resulting in functional recovery (e.g. improvement of ejection fraction in a murine model of myocardial infarction, femoral artery blood flow in a murine model of peripheral vascular disease, and wound healing in a skin injury model). Importantly, unlike VEGF-A therapy delivered by AAVs, VEGF-A EVs did not activate the innate or adaptive immune systems. EVs, extracellular vesicles; VEGF-A, vascular endothelial growth factor A; AAV, adeno-associated virus.
See the editorial comment for this article ‘Extracellular vesicles to mediate RNA therapies’, by J.-N. Boeckel and J. Xiao, https://doi.org/10.1093/eurheartj/ehaf046.
This is the first study that directly compares the efficacy and safety of gene therapy delivered using adeno-associated viral (AAV) vectors, lipid nanoparticles (LNPs), and extracellular vesicles (EVs). A head-to-head comparison was performed to evaluate the efficacy of VEGF mRNA gene therapy delivered by EVs vs. LNPs vs. AAV vectors to revascularize ischaemic limb muscle and heart tissue. Extracellular vesicles loaded with VEGF-A mRNA are more effective at improving lower limb perfusion and cardiac function in ischaemic mouse models as compared to VEGF gene therapy delivered by LNPs or AAV vectors. Importantly, this study demonstrated that increased efficacy may result from a higher immune tolerance of EVs as compared to other gene therapy vehicles, suggesting that EVs may have wider clinical applications for low-immunogenicity nucleic acid delivery than previously known. This study lays the foundation for continued development of VEGF-A EVs towards a first-in-human clinical trial.
Introduction
Despite being the foremost contributor to global morbidity and mortality, ischaemic vascular disease (IVD)—including myocardial infarction (MI), stroke, and peripheral vascular disease—has limited pharmacological revascularization options.1 For the past two decades, researchers have evaluated the potential of angiogenetic gene therapy, most notably vascular endothelial growth factor A (VEGF-A), to stimulate new vessel growth to rescue ischaemic tissue.2 Past randomized controlled trials for VEGF angiogenic therapy using direct administration of recombinant VEGF-A protein (i.e. VIVA trial) or manipulation of its gene expression via myocardial injection of VEGF-A plasmid (i.e. Euro inject One, VIF-CAD, and NORTHERN trials), VEGF-A DNA embedded in viral vectors (i.e. KAT, REVASC trials), and naked synthetic mRNA (i.e. EPICCURE trial) have shown limited efficacy in treating patients with ischaemic limb or heart tissue.3–10 The hurdles underlying poor clinical trial efficacy include the instability of injected recombinant VEGF-A proteins, poor transfection efficiency of plasmid DNA, immunogenicity of delivery vectors such as adeno-associated virus (AAV), and delayed pharmacokinetics. Some of these previous clinical efforts to deliver VEGF-A therapy in patients to treat ischaemic tissue are highlighted in Supplementary data online, Table S1.
One recent strategy to overcome these delivery challenges is the development of therapeutic mRNAs encapsulated inside lipid nanoparticles (LNPs), as evidenced by the rapid clinical translation of mRNA-based COVID-19 vaccines.11 However, some reports of local inflammation at the injection site (i.e. COVID arm), systemic anaphylaxis, and organ-specific delayed hypersensitivity reactions (i.e. myocarditis) associated with LNPs carrying mRNA vaccines suggest that the core ingredients contained in these synthetic carriers—such as synthetic polymers, non-degradable inorganic materials, and polyethylene glycol—are immunogenic, presenting new challenges for their efficacy for non-vaccine clinical applications and raising safety concerns.12
As natural vehicles of the body that transport biomolecules and nucleic acids, extracellular vesicles (EVs), including exosomes and micro vesicles, have emerged as a promising, non-viral alternative delivery system for carrying gene therapy to target tissues.13,14 Here, we demonstrate that cellular nanoporation (CNP) enables large-scale loading of full-length VEGF-A mRNAs into secreted EVs, allowing for efficient manufacturing and storage for clinical translation. Delivery of VEGF-A mRNA loaded EVs (VEGF-A EVs) to mouse models of limb and myocardial ischaemia generated rapid, transient, and pulse-like expression of VEGF-A at the target site, resulting in improved functional recovery compared to controls treated with saline and animals treated with AAV and LNP VEGF-A gene therapies. Importantly, unlike AAV and LNP control groups, repeated in vivo administrations of VEGF-A EVs were not observed to cause irritation, allergic reactions, or other immune activation, thereby, demonstrating the potential of EV mRNA delivery to circumvent immunogenic hurdles associated with current clinical modalities of gene therapy (i.e. DNA/mRNA delivered with viral vectors or LNP encapsulation). Collectively, this study demonstrates that EVs loaded with VEGF-A may offer an effective and potentially safer path for delivering angiogenic gene therapy to ischaemic tissue.
Methods
Detailed methodologies and expanded materials are available in the Data Supplement. Raw data that support the findings of this study are available from the corresponding author upon reasonable request. The Institutional Animal Care and Ethical Use Committee of Shenzhen Bay Laboratories approved all animal procedures.
Cell culture and cellular nanoporation
Neonatal human primary dermal fibroblasts (nHDFs, PCS-201-010) were purchased from ATCC (American Type Culture Collection, Manassas, VA). Mouse embryonic fibroblasts (NIH/3T3 cell line) were obtained from the Cell Resource Center, Peking Union Medical College (Cat#1101MOU-PUMC000018). Human primary nHDFs and murine NIH/3T3 cells were used as donor cells for production of hVEGF-A EVs and mVEGF-A EVs and delivery into immunocompromised and immunocompetent mice, respectively. Cells were maintained in Dulbecco’s Modified Eagle’s Medium supplemented with 10% foetal bovine serum. For CNP, a single layer of fibroblasts was seeded on a 1 cm × 1 cm 3D CNP hardware surface as previously described.13 Human VEGF-A (VEGF165) cDNA (NM_001171626.1) plasmid with a GFP tag was purchased from Sino Biological (Cat# HG11066-ACG, Beijing, China). Murine VEGF-A (NM_001287057) cDNA plasmid pCMV3-mVEGF-A-GFP was purchased from Youbio Technologies (Cat# G138749, Shaanxi, China). Cellular nanoporation procedures were described previously.13,14 Briefly, plasmids pre-loaded in phosphate-buffered saline (PBS) buffer at a concentration of 500 ng/mL were injected into individual cells via a CNP device using a 100 V electric field with 5 pulses at 15 ms per pulse with a 0.1 s interval.
Collection and purification of extracellular vesicles
The cells were washed three times with PBS. Before CNP initiation, the culture medium was changed to serum-free medium. The EVs were collected from the cell culture supernatant after treatment with CNP for 48 h. The cell culture media was centrifuged at 200 × g for 5 min to remove cells and debris and then centrifuged at 2000 × g for 30 min. Amicon Ultra-4 Centrifugal Filter Unit, 10 kDa (Cat# 801024, EMD Millipore, Hayward, CA) was used to concentrate the cell culture media. The sample of EVs was purified using a total exosome isolation reagent (Cat# 4478360, Invitrogen, Waltham, MA). Total RNA from purified EVs was obtained using an RNA purification mini kit (Norgen Biotek, 55000) and a DNA removal kit (Norgen Biotek, 25720). A SuperScript III First-Strand Synthesis system (Invitrogen) was used to synthesize the first-strand complementary DNA, with random hexamers as primers. For detection of the DNA contamination in CNP EVs, the PCR program was set as, 94°C for 7 min, followed by 35 cycles of denaturation at 94°C for 30 s, annealing 55–58°C for 1 min, extension at 72°C for 30 s, followed by final extension at 72°C for 7 min. The primer sequence was shown in Supplementary data online, Table S2.
Statistical analysis
Data were analysed using GraphPad Prism software (version 8). All data are presented as mean ± SEM. Parametric vs. non-parametric statistical tests were applied based on the Shapiro–Wilk test for normality. For data that were normally distributed, the following tests were performed: an unpaired two-tailed Student’s t-test was used to compare the two groups. One-way analysis of variance (ANOVA) was performed for multiple groups comparison, followed by the Holm–Sidak multiple comparison test for post hoc testing. When multiple measurements were obtained in the same subject over time, a two-way repeated-measure ANOVA with Holm–Sidak post hoc test for multiple comparisons was applied. For data that were not normally distributed, Kruskal–Wallis analysis was performed with a Dunn post hoc test for multiple comparisons. Two-tailed, paired t-test was applied for matched data. Detailed statistical data and methods are summarized in Supplementary data online, Table S3. Significance was accepted when P < .05.
Results
Preparation and in vitro delivery of VEGF-A extracellular vesicles
To generate EVs loaded with VEGF-A mRNA, we employed a CNP approach by nano-transfecting nHDFs with VEGF-A-GFP fusion plasmid to generate nHDF-derived EVs encapsulated with full-length VEGF-A mRNA (see Supplementary data online, Figure S1A–C, Table S4).14 VEGF-A EVs generated by CNP exhibited structural homogeneity with control EVs collected from untreated cell culture medium with a size distribution peaking around 150 nm in diameter as measured by nanoparticle tracking analysis (NTA) (Figure 1A and B). No changes in charge density and membrane properties were noted in the VEGF-A EVs vs. control EVs by zeta potential and atomic force microscopy (see Supplementary data online, Figure S1D and E). The VEGF-A EV group, however, showed higher expression of exosome associated (i.e. CD9, CD63, and TSG101) and micro vesicle associated (ARF6) biomarkers, confirming that a higher number of EVs were secreted from CNP-treated nHDF donor cells (see Supplementary data online, Figure S1C), a finding that was also noted via NTA (Figure 1B). Importantly, RT-qPCR showed that CNP-secreted EVs contained more than 3000-fold higher VEGF-A mRNA than EVs secreted from non-transfected cells (Figure 1C), and gel agarose demonstrated large amounts of full-length transcribed VEGF-A mRNA encapsulated within the EV (see Supplementary data online, Figure S1F) without DNA template contamination (see Supplementary data online, Figure S2A and B). Single EV lumen encapsulated with VEGF-A mRNA was verified by total internal reflection fluorescence microscopy assay (see Supplementary data online, Figure S3A and B). VEGF-A EVs derived from both nHDFs and induced pluripotent stem cell-derived cardiomyocytes (iPSC-CMs) were successfully internalized by recipient cells (e.g. iPSC-CMs) (see Supplementary data online, Figure S4A–D).
Characterization of EVs encapsulated with full-length VEGF-A mRNA for in vitro VEGF-A delivery. (A) Transmission electron microscopy (TEM) of VEGF-A EVs collected from neonatal human dermal fibroblasts (nHDFs) using cellular nanoporation (CNP) vs. no CNP (culture medium alone without nanoporation). Scale bar 100 μm. (B) Nanoparticle tracking showing size distribution and concentration of VEGF-A EVs (i.e. EV number per cell) in CNP-treated vs. control. (C) Quantification by RT-qPCR of VEGF-A mRNA in CNP-treated vs. control at 24 h (n = 6 for all groups, ***P < .001). (D) Fluorescence images of serum-starved iPSC-CMs treated with VEGF-A-GFP CNP EVs vs. control after 48 h. Scale bar, 20 μm. (E) Fluorescence intensity analysis of cells treated with VEGF-A-GFP CNP EVs vs. control (n = 6 for all groups, ***P < .001). (F) Western blots of VEGF-A protein in CNP-treated iPSC-CMs vs. control (n = 6 for all groups, **P < .01). All data are obtained from three independent experiments and are presented as means ± SEM. Two-sided Student’s t-test was used for the comparison in (C, E, F)
To assess the therapeutical potential of CNP produced VEGF-A EVs to deliver VEGF-A mRNA in vitro, we treated human iPSC-CMs with VEGF-A-GFP EVs for 48 h in culture, and then tracked VEGF-A protein expression in recipient cardiomyocytes by co-localization of GFP under immunofluorescence microscopy (Figure 1D). Fluorescence intensity quantification confirmed elevated VEGF-A protein expression (Figure 1E). Successful mRNA delivery and protein translation in recipient cells were further assessed by western blot analysis, which demonstrated increased VEGF-A protein expression (Figure 1F).
In vivo VEGF-A mRNA expression and protein translation following VEGF-A extracellular vesicle delivery
To evaluate the kinetics of EV-mediated VEGF-A mRNA delivery and subsequent protein formation in vivo, we next delivered human VEGF-A EVs labelled with GFP (hVEGF-A EVs) and control EVs loaded with GFP mRNA only (GFP EVs) into the tibialis anterior (TA) muscle of immunocompetent mice via a 28-G needle. Twelve hours after delivery, we observed a significant increase of VEGF-A mRNA expression in hVEGF-A EV-treated mice vs. those treated with GFP EVs and saline as measured by RNAscope (Figure 2A). VEGF-A mRNA levels decreased at 48 h after injection, returning to baseline levels by 7 days (see Supplementary data online, Figure S5A and B). To assess in vivo translation of VEGF-A mRNA into protein, we performed immunofluorescence staining, western blot, and enzyme-linked immunosorbent assay (ELISA) at 48 h post-treatment and confirmed increased VEGF-A protein expression (Figure 2B–F) compared to the GFP EV and saline control groups. Increased VEGF-A protein expression persisted for 7 days (see Supplementary data online, Figure S5C) compared to GFP EV and saline control groups. Similar experiments were repeated using a murine MI model and confirmed the release of VEGF-A mRNA and VEGF-A protein generated from EV mRNA around the border zone and infarction area (see Supplementary data online, Figure S6A–C). Forty-eight hours after VEGF-A EV delivery, immunofluorescence staining of the TA muscle confirmed the presence of a higher number of CD31 positive cells (i.e. a marker for endothelial cells) in the hVEGF-A EV-treated group compared to GFP EV and saline groups, suggesting increased angiogenesis (Figure 2G and H). Collectively, these findings support that in vivo delivery of VEGF-A EVs results in increased VEGF-A protein formation and new vessel growth in recipient tissues.
Kinetics of in vivo VEGF-A EV mRNA delivery and protein formation in murine skeletal muscle. (A) RNAscope in situ hybridization of human VEGF-A mRNA 12 h after tibialis anterior (TA) injection of VEGF-A EVs vs. GFP EVs vs. saline. Contiguous sections stained with haematoxylin and eosin also shown. Scale bar 50 µm. (B) Immunofluorescence staining of VEGF-A-GFP protein by co-staining of GFP and VEGF-A protein 48 h after intramuscular delivery of VEGF-A-GFP EVs. Scale bar 100 µm. (C) Quantification of VEGF-A positive GFP signal in VEGF-A-GFP EV, GFP EV, and saline control group (n = 6 for each group, ***P < .001). (D) Fluorescence intensity analysis of VEGF-A protein expression in mice injected with VEGF-A-GFP EV vs. GFP EV, vs. saline (n = 6 for each group, *P < .05). (E) Western blots of VEGF-A protein in TA muscle of mice treated with VEGF-A-GFP EV vs. GFP EV vs. saline (n = 6 for each group, **P < .01). (F) ELISA of VEGF-A protein in TA muscle and serum. VEGF-A-GFP EV vs. GFP EV vs. saline (n = 6 for each group, ***P < .001). (G) Immunofluorescence staining of CD31 supporting neovascular formation after intramuscular delivery of VEGF-A-GFP EVs vs. GFP EV vs. saline in 48 h after delivery. Scale bar 100 µm. (H) Quantification of CD31 positive signal in mice treated with VEGF-A EV vs. GFP EV vs. saline (n = 6 each group, ***P < .001). All data are obtained from three independent experiments and are presented as means ± SEM. One-way ANOVA followed by Holm–Sidak multiple comparisons were used in (C–F, H)
In vivo therapeutic efficacy of VEGF-A extracellular vesicles in murine models of ischaemic injury
To further assess the therapeutic efficacy of VEGF-A EVs, we employed CNP utilizing the murine NIH/3T3 donor cell line to produce murine EVs encapsulated with VEGF-A mRNA (mVEGF-A EVs) for revascularization of ischaemic heart tissue after MI in immunocompetent mice. Male C57BL/6 mice were subjected to ligation of the left anterior descending (LAD) artery followed by intramyocardial delivery of either: (i) saline control, (ii) control GFP EVs, or (iii) mVEGF-A EVs into the peri-infarct area. Compared to the saline and GFP EV control groups, beginning at 4 weeks post-MI, the mVEGF-A EV-treated group was found to have a statistically significant, sustained improvement in fractional shortening by echocardiography (Figure 3A–D) and left ventricular function by cardiac magnetic resonance imaging (see Supplementary data online, Figure S7A–E). Histological analysis revealed a lower burden of fibrotic tissue and preserved myocardial wall thickness in mice treated with mVEGF-A EVs compared to saline and GFP EVs (Figure 3E–G). Systemic vessel perfusion with tomato-lectin dye demonstrated a higher vessel density in ischaemic heart tissue treated with mVEGF-A EVs, highlighting the capacity of mVEGF-A EVs to induce in vivo angiogenesis and revascularization (Figure 4A–C). Immunofluorescence microscopy for CD31 co-localized with and without SM22α (adult smooth muscle specific marker) as a measure of capillary and arteriole density, respectively, in the peri-infarct and remote regions of the myocardium revealed significantly higher neovascularization in the mVEGF-A EV treatment group compared to other groups (Figure 4D–I). Taken together, these results suggest that treatment with mVEGF-A EVs can attenuate cardiac fibrosis and significantly improve systolic recovery in a murine model of acute MI via revascularization of ischaemic tissue.
mVEGF-A EV ameliorates left ventricular dysfunction and promotes reverse left ventricular remodelling in immunocompetent mice post-LAD ligation. (A) Representative M-mode echocardiographic images show left ventricular (LV) wall motion of infarcted hearts. (B) Quantitative analyses of left ventricular end-diastolic dimension (LVEDD) (n = 6 for each group). (C) Quantitative analyses of left ventricular end-systolic dimension (LVESD) (n = 6 for each group). (D) Comparison of fractional shortening (n = 6 for each group). (E) Haematoxylin and eosin (right) and Masson’s trichrome (left) staining of left ventricular tissue of mice. Scale bar 1 mm. (F) Quantification of the per cent of fibrotic tissue by histology (n = 6 for each group) and infarct wall thickness (G) (n = 6 for each group). *P < .05, **P < .01, ***P < .001 mVEGF-A EV vs. saline and #P < .05, ##P < .01, ###P < .001 mVEGF-A EV vs. GFP-EV-treated group. All data are presented as means ± SEM. Two-way ANOVA followed by Holm–Sidak post hoc test was used for the comparisons in (B–D). One-way ANOVA followed by Holm–Sidak multiple comparisons was used in (F, G)
Evaluation of mVEGF-A EV induced angiogenesis in infarcted hearts of immunocompetent mice. (A) Lectin-perfused heart samples were collected from VEGF-A EV vs. GFP EV vs. saline-treated mice at 7 days after LAD surgery. Fluorescent lectin signal in LAD territory (dotted white trace) was observed to evaluate vessel formation. (B) Quantitative analyses of perfusion rate by lectin signal (n = 6 for each group). (C) Quantitative analyses of fluorescent lectin signal (n = 6 for each group). (D) Immunofluorescent staining of platelet endothelial cell adhesion molecule-1 (PECAM1/CD31) and cardiac troponin T (cTnT) to quantify capillary density. Scale bar 50 µm. (E) Quantitative analyses of CD31 positive capillary density in border zone and remote zone (F) (n = 6 for each group). (G) Immunofluorescent staining of smooth muscle 22 alpha (SM22α) to quantify arteriole density after 56 days post-surgery. Cardiac troponin T (cTnT) staining to distinguish myocardium from the arteriole media. Scale bar 100 µm. (H, I) Quantitative analyses of SM22α positive vessel density in border zone (H) and remote zone (I) (n = 6 for all groups). *P < .05, **P < .01, ***P < .001 mVEGF-A EV vs. saline and #P < .05, ##P < .01, ###P < .001 mVEGF-A EV vs. GFP-EV-treated group. All data are presented as means ± SEM. One-way ANOVA followed by Holm–Sidak multiple comparisons was used for the comparisons in (B, C, E, F, H, I)
To assess the therapeutic efficacy of human derived EVs for potential clinical applications, we next employed nDHFs as CNP donor cells to generate hVEGF-A EVs for delivery into immunodeficient mice with induced ischaemic injuries in the hindlimb and heart. GFP EVs and saline vehicle were used as control groups. For hindlimb ligation, hVEGF-A EVs, GFP EVs, and saline control were delivered into the gastrocnemius muscle of immunodeficient animals following ligation of the femoral artery and observed for limb revascularization over 14 days with Doppler imaging. Beginning at Day 4 post-ligation and up through Day 14 after injection, marked improvement in hindlimb perfusion was observed in the ligated limbs of animals treated with hVEGF-A EVs along with higher number of endothelial cells (e.g. CD31+ cells) and higher patent vessel density as compared to animals receiving GFP EVs or saline control (see Supplementary data online, Figure S8A–G). Delivery of hVEGF-A EVs into ischaemic heart muscle of immunodeficient animals similarly demonstrated improved functional recovery as compared to GFP EV control and saline control as evidenced by echocardiography beginning 4 weeks post-MI (see Supplementary data online, Figure S9A–D) and confirmed by cardiac magnetic resonance imaging (see Supplementary data online, Figure S10A–E). Importantly, animals treated with hVEGF-A EVs also exhibited less fibrotic tissue (see Supplementary data online, Figure S9E–G) and had higher capillary and arteriole density on histological analysis (see Supplementary data online, Figure S11A–F).
Murine VEGF-A extracellular vesicles elicited a minimal immunogenic response when delivered into heart tissue and skeletal muscle of immunocompetent mice
As previous modalities of VEGF gene therapy have been hampered by immunogenicity of the delivery vehicle,8,9,15–19 we next sought to assess the degree to which different modes of delivering VEGF EV mRNA therapy activates the immune system. We, therefore, subjected immunocompetent C57BL/6J mice to a hindlimb ischaemia (HLI) model via ligation of the femoral artery, and randomized the animals to one of the four following treatment groups post-surgery: (i) murine VEGF-A mRNA EVs generated from NIH/3T3 (mVEGF-A EVs), (ii) synthetic murine VEGF-A mRNA encapsulated in LNPs (mVEGF-A LNPs), (iii) AAV vector packaged with murine VEGF-A cDNA (mVEGF-A AAVs), and (iv) saline control. Although significant recovery of hindlimb perfusion was observed in all VEGF-treated groups compared to saline, mice treated with mVEGF-A AAVs and mVEGF-A LNPs exhibited slower recovery in limb perfusion compared to the mVEGF-A EV group after HLI (Figure 5A and B).
Evaluation of limb perfusion and immune responses after intramuscular delivery of mVEGF-A EVs in immunocompetent mice after femoral artery ligation. (A) Representative laser Doppler images for the mVEGF-A AAV, mVEGF-A LNP, mVEGF-A EV, and saline treatment groups. The arrows indicate the ischaemic limb on Day 14 in different groups. (B) Quantification of blood flow by laser Doppler imaging (n = 6 for each group). (C) Immunohistochemistry staining of F4/80 (a marker of macrophages) for tibialis anterior (TA) muscle 24 h after intramuscular injection of mVEGF-A AAVs, mVEGF-A LNPs, mVEGF-A EVs, and saline. Scale bar 100 µm at 20× and 40× magnification, respectively. (D) Quantitative analyses of infiltrating cell numbers on haematoxylin and eosin staining (n = 6 for each group). (E) Quantitative analysis of the proportion of neutrophils (n = 6 for each group), macrophages (n = 6 for each group), and lymphocytes (n = 6 per group) in muscle samples from the mice injected with a single dose injection of mVEGF-A AAVs, mVEGF-A LNPs, mVEGF-A EVs, or saline 24 h post-injection. (F) Analysis of serum cytokine/chemokine levels 24 h post-injection by using a bead-based multiplexed assay (n = 6 for each group). (G) Results from an - ELISpot assay for IFN-γ to measure T cell responses. (H) Quantification of ELISpot assay result by counting spots (n = 6 for each group). *P < .05, **P < .01, ***P < .001 saline vs. other groups; #P < .05, ##P < .01 mVEGF-A EV vs. mVEGF-A LNP; †P < .05, ††P < .01, †††P < .001 mVEGF-A EV vs. mVEGF-A AAV; ns, not significant. All data are presented as means ± SEM. Two-way ANOVA followed by Holm–Sidak multiple comparisons test was used for the comparisons in (B). One-way ANOVA followed by Holm–Sidak multiple comparisons test was used for the comparisons in (D, E, H)
We hypothesized that the observed differences in the speed of revascularization between treatment groups was due to the extent of immune activation for each gene therapy delivery modality, which in turn could affect the amount of VEGF-A gene therapy delivered to the target site. To assess whether a delivery vehicle elicited an immunogenic response, we explanted skeletal muscle from immunocompetent C57BL/6J animals 24 h after intramuscular delivery of mVEGF-A EVs, mVEGF-A LNPs, or mVEGF-A AAVs. Skeletal muscle injected with mVEGF-A LNPs or mVEGF-A AAVs demonstrated massive leukocyte infiltration via haematoxylin and eosin (H&E) staining (see Supplementary data online, Figure S12A), which was not observed in skeletal muscle injected with mVEGF-A EVs or saline control. Immunohistochemistry staining for F4/80 confirmed an increased number of macrophages around the injection site in mice receiving mVEGF-A AAVs and mVEGF-A LNPs (Figure 5C and D). Analysis of the infiltrate by flow cytometry also demonstrated a much higher proportion of macrophages and lymphocytes in the skeletal muscle of mice injected with mVEGF-A AAVs and mVEGF-A LNPs as compared to mice injected with saline or mVEGF-A EVs (Figure 5E). Cytokine panels taken from local skeletal muscle tissue 24 h after injection showed higher levels of inflammatory cytokines/chemokines, such as IL-18, IL-19, IL-33, and IL-23, in mice treated with mVEGF-A AAVs and higher levels of IL-18 and IL-33R in mVEGF-A LNPs than mice treated with mVEGF-A EVs or saline, which both exhibited minimal levels of inflammatory cytokines (Figure 5F and Supplementary data online, Figure S12B and C).
To assess the adaptive immune system response to mVEGF-A EVs as compared with mVEGF-A LNPs and mVEGF-A AAVs, we collected blood and spleen samples from animals at 7 days post-treatment, a timepoint chosen to detect changes in T cell or B cell proliferation in response to antigen. In contrast to the mVEGF-A EV group, blood collected from mice injected with mVEGF-A AAVs and mVEGF-A LNPs contained a significant number of IFN-γ producing T cells as detected by an ELISpot (enzyme-linked immunospot) assay (Figure 5G and H). Immunofluorescence staining and fluorescence-activated cell sorting of the spleen further demonstrated a significant number of cytotoxic CD8 T cells secreting IFN-γ specifically in the mVEGF-A AAV group, suggesting activation of T cells in response to AAV (see Supplementary data online, Figure S13A and B). Finally, higher IgG and IgM antibody concentrations and elevated neutralization titres were found in the serum of mVEGF-A AAV and mVEGF-A LNP groups, representing an activated humoral immune response, as compared to animals receiving mVEGF-A EVs (see Supplementary data online, Figure S14A–C). In total, these results indicate that mVEGF-A AAVs elicit a high degree of both innate and adaptive immunogenicity when delivered in vivo; whereas mVEGF-A LNPs elicit a strong innate immune response and lower adaptive immune response. By comparison, however, in vivo delivery of mVEGF-A EVs is characterized by minimal activation of both the innate and adaptive immune systems. Similarly, injection of mVEGF-A EVs into the myocardium of immunocompetent mice also demonstrated minimal activation of the innate and adaptive immune response (see Supplementary data online, Figure S15A–H).
Low immunogenicity of mVEGF-A extracellular vesicles allows for serial in vivo delivery for improved treatment of wound healing and angiogenesis as compared with adeno-associated virus and lipid nanoparticles
Because traditional AAV gene therapies have traditionally been limited to one injection per patient lifetime due to antibody formation against the AAV capsid,17–19 we next sought to test the feasibility of serial in vivo injections of mVEGF-A EVs for treating ischaemic injury, and whether repeated injections would trigger an immune response. Because MI and HLI serial treatment models involve repeated deep tissue surgery, which may confound results, we elected to utilize a skin wound healing model where serial gene therapy delivery could be accomplished via superficial injection. We created a full thickness wound by sterile biopsy punch (8 mm size) on the dorsal skin of immunocompetent C57BL/6J mice. Over the subsequent 3 weeks, the mice were treated with five serial intradermal injections of one of the following: (i) saline control, (ii) mVEGF-A AAVs, (iii) mVEGF-A LNPs, and (iv) mVEGF-A EVs (Figure 6A). We recorded the skin wound area by microscopic visualization at 0, 4, 7, 10, 14, and 18 days post-injury. Assessment of photographic wound size revealed that mice treated with mVEGF-A EVs exhibited significant improvement in wound healing as early as 4 days post-injury and fully recovered by Day 18 (Figure 6B). The rate of wound closure was accelerated in mice treated with mVEGF-A EVs as evidenced by a significant decrease in wound size beginning at Day 4 post-injury compared to the saline, mVEGF-A AAV, and mVEGF-A LNP groups. Notably, the mVEGF-A AAV and mVEGF-A LNP-treated groups demonstrated no significant improvements in wound area as compared to saline control at Day 18 post-injury (Figure 6C). H&E and Masson’s trichrome staining of skin tissue, harvested at Day 18 post-injury, confirmed that mVEGF-A EV-treated mice exhibited more granulation tissue and had thicker dermal skin layers with a flat epidermal surface and shorter wound length than the other groups (Figure 6D and E). Abundant capillaries were observed in the subcutaneous layer of mVEGF-A EV-treated animals, highlighting effective and improved angiogenesis induced by repeated intradermal administration as compared to the saline control, mVEGF-A AAV, and mVEGF-A LNP cohorts (Figure 6F and G).
Repeated intradermal administrations of mVEGF-A EVs promote cutaneous wound healing and angiogenesis in immunocompetent mice. (A) A schematic representation and timeline of five repeated cutaneous injections at Days 0, 4, 7, 10, and 14. Skin tissue was collected at 18 days. (B) Photos of the wound healing process visualized under a microscope at 0, 4, 7, 10, 14, and 18 days. (C) Wound closure rate was measured by ImageJ after receiving different treatments (n = 6 for each group). (D) Representative haematoxylin and eosin (H&E) and Masson’s trichrome staining of whole wound sections injected with mVEGF-A AAV, mVEGF-A LNPs, mVEGF-A EVs, and saline on Day 18. Scale bar 1 mm. (E) Quantification of wound length by measurement of histological slides on Day 18 (n = 6 for each group). (F) Immunofluorescent staining of CD31 for angiogenesis evaluation at Day 18. Scale bar 200 µm. (G) Fluorescence intensity analysis of CD31 staining at Day 18 (n = 6 for each group). *P < .05, **P < .01, ***P < .001 saline vs. other groups; #P < .05, ##P < .01, ###P < .001 mVEGF-A EV vs. mVEGF-A LNP; †P < .05, ††P < .01, †††P < .001 mVEGF-A EV vs. mVEGF-A AAV; ns, not significant. All data are presented as means ± SEM. Two-way ANOVA followed by Holm–Sidak multiple comparisons test was used for the comparisons in (C). One-way ANOVA followed by Holm–Sidak multiple comparisons test was used for the comparisons in (E, G)
To assess the immunogenicity of serial administration of mVEGF-A injected agents, we evaluated CD8 cytotoxic T cell activation in immunocompetent C57/Bl6 mice using immunofluorescence staining of major histocompatibility class I (MHC1). We found overexpression MHC1, a marker of immune activation, in skin tissue harvested from the mVEGF-A AAV-treated group at Day 18 when compared with mVEGF-A LNP and mVEGF-A EV-treated groups (Figure 7A and B). FACs of the skin and spleen showed significantly increased number of CD8 T cells including those producing IFNγ in mVEGF-A AAV group at Day 18 (Figure 7C), confirming heightened activation of adaptive immune system with serial dosing as compared to the mVEGF-A LNP and mVEGF-A EV groups. With respect to the innate immune system, an increase in neutrophils and macrophages was found in both mVEGF-A AAV and mVEGF-A LNP groups (Figure 7D) but was absent in the mVEGF-A EV group. Taken together, these findings suggest that unlike mVEGF-A AAV delivery, delivery of either mVEGF-A LNPs or mVEGF-A EVs can be administered serially in vivo without heightened CD8 T cell immune activation. However, serial delivery of mVEGF-A LNPs was found to trigger innate inflammation with each successive dose as compared with mVEGF-A EVs, which were largely indistinguishable from saline control in terms of immune activation. Because innate and adaptive immune cell-mediated destruction of mVEGF-A AAVs and mVEGF-A LNPs would compromise the therapeutic payload delivery for these cohorts, we conclude that the low immunogenicity of mVEGF-A EVs contributes to the improved angiogenic effects observed in this treatment group.
Comparison of immunogenic response after serial cutaneous injections of mVEGF-A AAV, mVEGF-A LNPs, mVEGF-A EVs, and saline in immunocompetent mice. (A) Immunofluorescent staining of major histocompatibility complex 1 (MHC class 1), a marker of immune activation, in skin samples. Scale bar 50 µm. (B) Fluorescence intensity analysis of MHC class 1 staining (n = 6 for each group). (C) Skin samples were harvested. Digested cells were then analysed by flow cytometry to evaluate T cell response for CD4 positive cells percentage and CD8 positive cells percentage (n = 6 for each group) and IFNγ positive cells in CD8+ cells (n = 6 for each group). (D) Cells from skin samples were analysed by flow cytometry for neutrophil cell percentage (n = 6 for each group), and macrophage cell percentage (n = 6 for each group). *P < .05, **P < .01, ***P < .001 saline vs. other groups; #P < .05, ##P < .01 mVEGF-A EV vs. mVEGF-A LNP; †P < .05, †††P < .001 mVEGF-A EV vs. mVEGF-A AAV; ns, not significant. All data are presented as means ± SEM. One-way ANOVA followed by Holm–Sidak multiple comparisons test was used for the comparisons in (B–D)
Discussion
Over the past three decades, significant advancements have been achieved in the surgical management of acute ischaemic injuries affecting the heart and lower limb. Recent studies, however, have called into question whether surgery should be used as the first line of therapy in all patients suffering from IVD.20–22 Among patients with stable coronary artery disease with moderate to severe ischaemia and even in those with left ventricular dysfunction, surgical revascularization has not reduced death or hospitalization for heart failure as compared to optimal medical therapy.20,21 Similarly, limited evidence supports the superiority of peripheral limb revascularization over medical management in patients with non-critical limb ischaemia.22 Surgical revascularization may not even be an option for some patients who have significant co-morbidities23 or who are fraile.24 Moreover, patients with diffuse epicardial disease and/or microvascular disease may derive little benefit from surgery because the benefits of surgical revascularization are greatest in those patients with focal lesions in large, epicardial arteries.
Therapeutic angiogenesis, including the injection of cells, proteins, and nucleic acids, has recently emerged as an alternative strategy for surgical intervention. In recent years, cell-free approaches have gained momentum given the lack of clear benefit in clinical trials using cellular therapy,25 and studies showing poor retention and survival of injected cells.26,27 In this study, we conduct the first head-to-head comparison of VEGF-A AAVs versus VEGF-A LNPs and VEGF-A EVs. Our findings indicate that cell-free, VEGF-A mRNA encapsulated into EVs can rescue ischaemic tissue with improved efficacy and minimal immunogenicity. Most notably, both acute and serial therapeutic delivery of VEGF-A EVs in vivo improved efficacy without significant immune activation of innate or adaptive immune system, whereas VEGF-A AAVs and VEGF-A LNPs were both found to trigger immune responses (Structured Graphical Abstract).
Previous efforts to load EVs with full-length mRNAs have failed due to low efficiency of standard bulk electroporation (BEP) methods.28 To circumvent the hurdles of BEP, our lab recently pioneered a unique mRNA loading technique via a CNP process that allows intact endogenous mRNAs to be encapsulated into EVs secreted from donor cells.13,14 Expanding on these previous studies, we demonstrate that the CNP platform can load over 3000-fold higher VEGF-A mRNAs into the EV lumen compared to natural EVs which are secreted by donor cells. This, in turn, allows for therapeutic nucleic acid delivery to form large amounts of VEGF-A protein at sites of ischaemic injury in vivo to facilitate revascularization. Intramuscular delivery of VEGF-A EVs labelled with GFP was found to result in heightened GFP positive VEGF protein beginning as early as 12 h after delivery with a peak of protein expression at 48 h, which subsequently tapered off over 7 days. Because sustained expression of VEGF-A for >4 weeks has been previously shown to produce angioma-like structures that do not contribute to blood flow and is a risk for tumour formation,29 the short half-life of VEGF-A mRNA and resulting protein offers a safety profile that may be preferred for clinical translation. To the best of our knowledge, this is the first study to characterize the pharmacokinetics of VEGF-A EVs and VEGF-A protein formation in vivo.
In our comparison of VEGF-A EVs with other clinical delivery modalities such as VEGF-A AAVs and VEGF-A LNPs, we found that VEGF-A EVs resulted in improved revascularization and functional recovery. Differences in resulting efficacy were most notable with serial VEGF-A EV injections, which suggested that the improved efficacy likely results from differences in immunogenicity between EVs and other delivery methods. Our findings are consistent with previous studies that have shown that humans injected with AAV vectors develop T and B cell responses against the AAV capsid and even its cargo, leading to (i) T cell-mediated rejection of transduced cells,30 (ii) B cell production of AAV-neutralizing antibodies that limit efficacy,31–33 and (iii) death from uncontrollable immune activation.34,35 The LNP immunogenicity observed in our study has also been observed in a number of pre-clinical and clinical studies,36–40 although in our head-to-head comparison, AAVs were found to elicit much stronger adaptive immune activation than LNPs.
Specifically, we demonstrate that mRNA EVs produce minimal immune activation of the innate and adaptive immune system as evidenced by the low level of immune cell infiltration and cytokine production post-injection following in vivo delivery. In fact, the levels of immune activation following mRNA EV treatment were comparable to control (i.e. saline injection). While other studies have implicated EVs in playing a dual role in immune suppression and immune regulation, the immunogenicity or immune privilege of these naturally occurring transporters appears to be dependent on the source of the EVs as well the cargo the EVs are carrying.41 While a recent study reported that xenogeneic delivery of human EVs into mice can elicit a low level of immunogenicity,42 other studies have shown minimal immune activation if engineered EVs and their cargo are derived from the same species.28,41,43 Consistent with these reports, our findings indicated that VEGF-A mRNAs derived from murine dermal fibroblasts and encapsulated into murine EVs injected into murine models of injury elicited minimal immune response. Notably, we observed that EVs, serving as biocompatible delivery vehicles, exhibited a reduced macrophage response and lower levels of inflammatory cytokines such as IL-18 and IL-19. Additionally, there was a noticeable attenuation in T cell immune responses, further supporting the potential of EV mRNA to circumvent immunogenic challenges of AAV and LNP delivery modalities.
For the purposes of therapeutic revascularization, maintaining a local concentration of angiogenic factors is highly influenced by the choice of the vector system and the mode of delivery. Viral vectors, such as AAV, can induce robust yet transient gene expression but would be neutralized by pre-existing antibodies in patients who have been previously exposed to the AAV capsid as part of a common cold infection or previous AAV gene therapy treatment.44,45 Non-viral vectors such as LNP suffer the challenge of eliciting inflammatory infiltrate to enter the site of delivery, which results in accelerated clearance of the delivery vehicle and therapeutic cargo.36–40 Serial delivery of EV mRNA is a potential method to maintain thresholds of pro-angiogenic factors at sites of ischaemic injury without activating a rapid immune response which would destroy exogenously delivered VEGF-A nucleic acids. Indeed, we observed that multiple administrations of VEGF-A EVs resulted in accelerated wound healing, fostering the rapid formation of new capillaries at dermal sites of injury. We anticipate that serial delivery of EV mRNA can result in beneficial therapeutic responses for other clinical indications where repeated dosing of AAV and LNP are not possible due to accelerated immunogenic clearance with repeated delivery.
Future challenges in the clinical application of VEGF-A EVs involve enhancing its targeting to areas of ischaemia and optimizing delivery routes. For this study, all EVs were injected locally into sites of injury due to inefficient targeting from intravenous or systemic administration. For targeted delivery of VEGF-A EVs to the ischaemic myocardium in human subjects, catheter-based trans-endocardial injection may reduce the likelihood of systemic toxicity but may be inconvenient for repeat administration. Alternatively, intravascular delivery may facilitate serial treatment, but an inability to deliver agents passed arterial occlusions may hinder gene delivery to ischaemic tissue.46 Moreover, intravascular delivery poses increased risk of systemic circulation, which may result in off-target effects. Future endeavours will need to focus on enhancing precision delivery of VEGF-A gene therapy to sites of ischaemia via targeting peptide modification of the EV surface13 or via antibody conjugation to EV surface proteins.13,47 These methods will enrich VEGF-A delivery to sites where it is beneficial, while minimizing delivery to other organs. Safety studies will need to be conducted to confirm off-target delivery to sites, such as the liver or kidneys, does not result in unintended toxicity.
In summary, our findings indicate that VEGF-A EV mRNA delivery offers a novel low-immunogenicity alternative to AAVs and LNPs for angiogenic gene therapy to treat ischaemic injury. Our study offers proof of concept for EV-mediated encapsulation and delivery of VEGF-A mRNA to treat MI, limb ischaemia, and dermal wound healing as compared to other nucleic acid delivery modalities which have been hampered by immune clearance. We believe that the future development of the EV mRNA platform may be able to circumvent immunogenic hurdles, which have prevented other clinical stage VEGF-A assets from yielding efficacious results.
Supplementary data
Supplementary data are available at European Heart Journal online.
Declarations
Disclosure of Interest
A.S.L. and L.J.L. are shareholders in Spot Biosystems, Ltd.
Data Availability
The data supporting the findings of this study are available from the corresponding authors upon reasonable request.
Funding
All authors declare no funding for this contribution.
Ethical Approval
Animal work was approved by the Institutional Animal Care and Use Committee (IACUC) of Shenzhen Bay Laboratory (no. D2021-107) or partner laboratories.
Pre-registered Clinical Trial Number
None supplied.
References
Author notes
Yi You and Yu Tian contributed equally to the study.
