Extracellular vesicles conjugated with c(RGDyk) peptide targeting integrin αVβ3 repair optic nerve injury through YAP/TAZ and Smad2/3 signaling

Abstract

Extracellular vesicles (EVs) derived from mesenchymal stem cells have therapeutic potential for optic nerve injury. However, further investigations are needed to increase their efficacy. In this study, we tried to enhance targeting and recovery function of EVs using conjugation with integrin αVβ3 antagonist-c(RGDyk) peptide. The molecular mechanism of neuronal repair was investigated as a potential treatment for optic nerve injury. EVs were shown to restore effectively the abnormal regulations of neuronal markers in optic nerve injury models. Notably, the functionally optimized c(RGDyK)_EVs exhibited superior targeting capabilities and modulated neuroregeneration in cases of hypoxic damage and inflammation. Single-cell RNA-seq analysis of R28 cells revealed significant regulation of transcription of genes involved in retinal ganglion cell (RGC) regeneration, neuronal growth, and inflammation by c(RGDyK)_EVs via the Yap-Taz signaling pathway. This study highlighted the enhanced therapeutic potential of c(RGDyK)_EVs over naïve EVs in the context of optic nerve disease. The findings suggested that c(RGDyK)_EVs hold promise as an alternative therapy for optic nerve injury.

Graphical abstract

Graphical Abstract.

Significance statement.

Optic nerve injury often leads to permanent vision loss, and effective treatments are limited. This study shows that c(RGDyK)_EVs, extracellular vesicles modified with a targeting peptide, can regulate cellular responses to nerve damage. These vesicles modulate key signaling pathways associated with regeneration and inflammation to reduce damage, protect nerve cells, and promote tissue repair. This study enhances understanding of cell-free treatment strategies by revealing how c(RGDyK)_EVs support nerve recovery and highlights their potential as a safe and promising approach for the treatment of optic nerve injury.

Introduction

Optic nerve injury poses a significant challenge in patients with neurodegenerative diseases, often resulting in vision impairment or loss. Recent advances in regenerative medicine have revealed the therapeutic potential of mesenchymal stem cell (MSC)-derived extracellular vesicles (EVs) for optic nerve injury.¹ These EVs, derived from MSCs sourced from human placenta, have exhibited encouraging therapeutic properties in preclinical studies. However, it remains necessary to increase their efficacy for clinical use.

The complexity of optic nerve injury underscores the need for enhanced therapeutic interventions in regenerative medicine. Across a spectrum of incurable diseases, including optic neuropathy, there is a necessity to bolster the regenerative potential of biological therapeutics. Enhancing the efficacy of these agents is critical for advancing therapeutic approaches to otherwise untreatable conditions.² These nanosized vesicles possess inherent regenerative and modulatory properties, and a range of strategies have been proposed to enhance their functionality, each with its own advantages and limitations.

Surface modification of EVs involves altering their surface properties, enabling targeted delivery, and enhancing stability. This method facilitates targeted delivery to specific cell types or tissues, thereby enhancing therapeutic efficacy. However, it may impact biocompatibility, potentially affecting the natural properties of EVs, as well as increasing the complexity of production.^3,4 Cargo enrichment can be used to load EVs with specific therapeutic agents, such as proteins or nucleic acids, and thus, enhance their therapeutic potential. Their use allows tailored therapeutic cargo delivery, potentially improving the specificity and efficacy of treatment. However, the loading process may affect EVs stability or cargo release kinetics, posing challenges with regard to maintaining their integrity and functionality.^5,6

Recent studies have examined the genetic alteration of parent cells to produce engineered EVs with modified cargo or surface properties to provide precise control over cargo composition and surface properties, and thus, enhance therapeutic specificity. Genetic manipulation may affect EV yield or alter their inherent properties, necessitating thorough characterization and safety assessment.^7,8 Targeted delivery systems have been designed utilizing EV surface modification with specific moieties, such as peptides or antibodies, to improve their targeting capabilities. Enhancing the specificity of EV delivery to target tissues or cells may also alter their natural behavior or introduce concerns regarding immunogenicity.^9,10

This study was performed to enhance the therapeutic potential of MSC-derived EVs through functional optimization for treatment of optic nerve injury. Specifically, we investigated the impact of c(RGDyK) peptide-conjugated EVs [c(RGDyK)_EVs] derived from human placental MSCs (hMSCs) on retinal ganglion cells (RGCs) and macrophages in optic nerve injury models. Modifying EVs with integrin-related peptides, such as the RGD (Arg-Gly-Asp) sequence, provides a powerful means of enhancing their therapeutic potential in the context of nerve injury.^3,11,12

Methods

Preparation and culture of human mesenchymal stem cells

hMSCs were obtained from CHA Biotech (Seongnam, Republic of Korea), and were cultured as described previously.¹

Isolation of EVs

The collected medium was centrifuged at 1300 rpm for 3 min, followed by elimination of non-exosomal large particles, such as cells, cell debris, microvesicles, and apoptotic bodies, using a Vacuum Filter/Storage Bottle System with 0.22-μm pore membrane. EVs were isolated using a 500-kDa molecular weight cutoff filter in tangential flow filtration (TFF; Repligen, Waltham, MA, USA). The isolated EVs were then concentrated for further applications using an Amicon Ultra-15 centrifugal Filter Unit (Merck, Darmstadt, Germany). The rate of diafiltration was adjusted at 7. The concentrated EVs were further processed at 4 °C to minimize degradation. The Amicon Ultra-15 centrifugal filter was operated at 3000 g for 15 min, or until the desired final volume was reached. The purified and concentrated EVs were stored at −80 °C until further use in in vitro and in vivo experiments.

Characterization of EVs

The quantity and size of particles were determined using a MONO ZetaView (PMX-120; Particle Metrix, Meerbusch, Germany) in 488 nm scatter mode. The EVs samples were diluted to 10⁷-10⁸ particles/mL using filtered phosphate-buffered saline (PBS; HyClone Laboratories, Logan, UT, USA). For all samples, the detailed parameters for accurate analysis were tuned with a sensitivity 75, shutter of 100, minimum trace length of 15, and cell temperature of 25 °C. Transmission electron microscopy (TEM; H-7600, 80 kV; Hitachi, Tokyo, Japan) was performed to examine the morphology of the EVs. Briefly, the EV solution was dried on a Formvar/copper grid with a 150-mesh carbon coating (FCF150-CU; Electron Microscopy Sciences, Hatfield, PA, USA). EVs were stained with 7% uranyl acetate or gadolinium acetate solution and dried for negative staining on the copper grid. The Formvar/copper grid was put on a grid box for TEM investigation after drying.

Conjugation of peptides to EVs

The process for conjugation of ligands to EVs analysis has been described in detail previously.³ Briefly, reactive dibenzocyclooctyne (DBCO) groups were incorporated into amine-containing molecules on EVs using a heterobifunctional crosslinker. Specifically, 3 μM dibenzocyclooctyne-sulfo-N-hydroxysuccinimidyl ester (DBCO-sulfo-NHS; Sigma, St. Louis, MO, USA) was added to 8 × 10¹⁰ EVs in PBS and allowed to react on a rotating mixer at room temperature (RT) for 4 h. Unconjugated DBCO-sulfo-NHS was removed by four washes in 100-kDa ultrafiltration tubes (Millipore, Billerica, MA, USA). The DBCO-conjugated EVs (DBCO-EVs) were then linked to azide-containing molecules via copper-free click chemistry. Cyclic [Arg-Gly-Asp-d-Tyr-Lys(azidoacetyl)] [c(RGDyK)], and Cyclic [HKHDNAQS-K(azidoacetyl)STRK-OH] (RKK12) peptides, were synthesized by Peptron (Daejeon, Republic of Korea). Cyanine 5.5 azide (Cy5.5 azide; Lumiprobe Co., Hallandale Beach, FL, USA) was also introduced onto peptide-conjugated EVs according to the manufacturer’s protocol. Briefly, 0.3 μM c(RGDyK) or RKK12 peptide was added to DBCO-EVs in PBS, and 0.3 μM Cy5.5 azide was added. The reaction was conducted on a rotator at 4 °C for 16 h. All reactions were performed at pH 7.4. The EVs were then floated on a 30% sucrose/D₂O cushion and centrifuged at 164 000 × g for 90 min using an SW32Ti rotor (Beckman Coulter, Fullerton, CA, USA) to remove unincorporated ligands. After washing with PBS at 164 000 × g for 60 min, the conjugated EVs were resuspended in PBS and stored at −80 °C prior to use.

EVs uptake assay

To confirm the successful conjugation of EVs and azide ligands, R28 cells were treated with Cy5.5-c(RGDyK)-conjugated EVs (100 μg/mL) on coverslips. After 1 or 24 h, the coverslips were mounted onto slide glasses, and images were captured by confocal microscopy for confirmation of fluorescence (LSM 880; Carl Zeiss, Jena, Germany).

Flow cytometry

To assess the tropism of c(RGDyK)_EVs toward R28 cells, 30 µg/mL of Cy5.5-labeled naive EVs or Cy5.5-labeled c(RGDyK)_EVs were incubated with R28 cells for 3 h. For blocking experiments, R28 cells were pre-treated with 40 µM free c(RGDyK) peptide for 30 min prior to EVs administration. Following incubation, cells were trypsinized, washed twice with PBS, and resuspended in FACS buffer. Untreated R28 cells were used as a negative control. The fluorescence intensity (Cy5.5) of over 10 000 cells per sample was measured using a CytoFLEX V5-B5-R3 flow cytometer (Beckman Coulter, Brea, CA, USA). Cy5.5-positive cells were presented by comparison with the control group.

Cell culture and EVs treatment

Immortalized R28 retinal precursor cells were cultured and treated with EVs as described previously.¹ A hypoxic environment was produced in cells by treatment with cobalt chloride (CoCl₂). Aliquots of 2 × 10⁵ R28 cells were seeded onto glass coverslips and treated with 200 µM CoCl₂ for 9 h. The hypoxia-damaged R28 cells were then treated with EVs 170-180 nm in diameter at a concentration of 24 µg/mL. After 24 h, the cells were harvested and prepared for analysis.

Inhibition of YAP/TAZ signaling in vitro

Aliquots of 2 × 10⁵ R28 cells were seeded onto glass coverslips and treated with 200 µM CoCl₂ and 1 µM of the TEA/TEF-domain (TEAD) transcription factor inhibitor TM2 (Tocris Bioscience, Bristol, UK) for 9 h. The hypoxia-damaged R28 cells were then incubated with EVs at a concentration of 24 µg/mL. After 24 h, the cells were fixed in 4% paraformaldehyde at RT for 20 min. The fixed cells were permeabilized with 0.2% Triton X‐100 in PBS, and then blocked with 2% BSA in PBS for 1 h. The cells were incubated for 12 h with primary antibodies (Supplementary Table S1—see online supplementary material) in PBS‐T at 4 °C. After washing three times with PBS containing Tween 20 (PBS‐T), the cells were counterstained with Alexa Fluor 555-conjugated goat anti‐rabbit IgG (1:1000; Santa Cruz Biotechnology, Santa Cruz, CA, USA) for 2 h. After additional washing with PBS‐T, the glass coverslips were mounted on slides using mounting solution. Fluorescence was detected by confocal microscopy (LSM 880; Carl Zeiss).

Single-cell RNA sequencing

Library construction was performed using a 10X Chromium Single Cell 3′ Reagent Kit v3.1 (10X Genomics, Pleasanton, CA, USA). Samples were sequenced using the Illumina NovaSeq 6000 platform (Illumina, San Diego, CA, USA) and preliminary sequencing results were converted to FASTQ files with Cell Ranger. We followed the 10× Genomics standard sequencing protocol by trimming the barcode and unique molecular identifier (UMI) end to 26 bp, and the mRNA end to 98 bp. The FASTQ files were, then, aligned to the human reference genome (GRCh38). Subsequently, we used Cell Ranger for preliminary data analysis and generated a file containing a barcode table, a gene table, and a gene expression matrix. We used WinSeurat v2.1 (Ebiogen Inc., Seoul, Republic of Korea) based on Seurat v3 for quality control (QC), analysis, and exploration of single-cell RNA-seq data.^13,14

Single-cell RNA data analysis

Data analysis was carried out with R v4.0.3 (R Development Core Team, 2011) and Seurat v4.3.0.¹⁵ We identified 5214 cells (CoCl₂), 6652 cells (RGDyK), and 5714 cells (RKK12), and QC was conducted on UMI count, gene count, and mitochondrial gene count before downstream analysis. Following QC, 4640 cells (CoCl₂), 6104 cells (RGDyK), and 5242 cells (RKK12) with UMI counts ≥ 1000, gene counts ≥ 200, and mitochondrial transcripts ≤ 20% were included in data analysis.

Differentially expressed gene analysis

Sets of differentially expressed genes (DEGs) from different sample comparisons—i.e., CoCl₂ vs c(RGDyK), CoCl₂ vs RKK12, RKK12 vs c(RGDyK), CoCl₂ vs c(RGDyK) & RKK12, c(RDGyK) vs CoCl₂ & RKK12, and RKK12 vs CoCl₂ & c(RGDyK)—were identified using the FindMarkers function, which calculated the average log₂-fold change in expression of each gene.

Gene ontology analysis

Functional analysis of DEG results was performed with clusterProfiler v4.11.0¹⁶ in R, which supports the statistical analysis and visualization of functional profiles with Gene Ontology (GO)^17,18 based on DEGs. We visualized the results of GO analysis using VlnPlot, DotPlot, RidgePlot, GOBubble, and pheatmap in R and compared expression profiles between samples.

Animals

Six-week-old female Sprague–Dawley (SD) rats (Koatech, Gyeonggi-do, Republic of Korea) were housed in a standard animal facility at a constant temperature of 21 °C with food and water provided. All animal experiments were conducted in accordance with protocols approved by the Institutional Animal Care and Use Committee of Bundang CHA Medical Center (IACUC No. 220101).

Optic nerve compression and lipopolysaccharide-induced optic nerve inflammation models

Optic nerve compression (ONC) models were generated in the left eye (oculus sinister, OS), and the right eye (oculus dexter, OD) was used as a control. The animals were anesthetized by intraperitoneal injection of Zoletil (Virbac, Hamilton, Newzealand) and Rompun (Bayer AG, Leverkusen, Germany). A lateral canthotomy and conjunctival incision were made after topical application of 0.5% proparacaine hydrochloride. Subsequently, the tissues surrounding the optic nerve were carefully dissected to expose the optic nerve without damaging the adjacent blood supply. The ONC models were constructed using Calison DSEK graft forceps, 3 1/4″, non-self-closing (Ambler Surgical, Exton, PA, USA) to apply mild compression to the nerve 2 mm behind the globe for 5 s.^19,20

The optic nerve sheaths were opened longitudinally, and microinjection was performed with a 30 G needle attached to a Hamilton syringe (Hamilton, Reno, NV, USA). The needle was inserted into the optic nerve 2 mm posterior to the globe, and 1 μL of 4.5 μg/μL Salmonella typhimurium lipopolysaccharide (LPS; Sigma) in saline, was injected over approximately 10 s.²¹ After thoroughly suturing the canthal site, subtenon injections of 60 μL of nai¨ve_EVs or c(RGDyK)_EVs (5 μg/μL) were performed into the nasal aspect of the eyeball. The animals were euthanized after 1 or 4 weeks, and the tissue was collected for analysis.

Immunoblot analysis

Total proteins from cells or tissues were extracted using PRO-PREP buffer (iNtRON Biotechnology, Gyeonggi-do, Republic of Korea). Protein concentration was determined by bicinchoninic acid assay (BCA; Thermo Fisher Scientific). Target proteins were separated by SDS-PAGE and transferred onto PVDF membranes (GE Healthcare, Chicago, IL, USA), which were then incubated overnight at 4 °C with primary antibodies (Supplementary Table S1—see online supplementary material). After a series of washes, the membranes were incubated overnight at 4 °C with horseradish peroxidase-conjugated anti-rabbit or mouse secondary antibody at 1:10 000 dilution (GeneTex, Irvine, CA, USA). Target bands were were detected on an ImageQuant LAS 4000 (GE Healthcare).

Immunostaining of retinal wholemounts

To prepare retinal whole-mount specimens, both normal and injured eyes were immersion-fixed overnight in 4% paraformaldehyde (PFA), after which the retina was isolated. The immunofluorescence process was performed as described previously.² The retinas were incubated with primary antibodies overnight at 4 °C (Supplementary Table S1—see online supplementary material). Subsequently, counting boxes (0.26 mm²) were then designated along quadrant radii of 1.5 mm, 2.5 mm, and 3.5 mm from the optic nerve head for fluorescence quantification,²² and images were captured with a confocal microscope (LSM 880; Carl Zeiss).

Histological analysis

Optic nerves were fixed with 4% PFA, dehydrated through a graded ethanol series, and embedded in paraffin. Paraffin blocks were sectioned at a thickness of 5 μm and stained with hematoxylin and eosin (H&E) according to standard protocols. In both the ONC and optic nerve inflammation (ONI) models, optic nerve thickness was measured at 1000 μm intervals distally from the globe-facing end of the optic nerve.²³ For each section, nerve diameter was quantified using ImageJ/ZEN 3.1 software by averaging three independent measurements taken across the transverse section. Images were acquired using an Axioscan slide scanner (Carl Zeiss) and processed using ZEN 3.1 (Zeiss). Inflammatory cell infiltration in the ONI model was evaluated using a semi-quantitative scoring system as follows: (1) occasional infiltration of single leukocytes, (2) focal infiltration of clustered leukocytes, and (3) coalescing or widespread leukocyte infiltration throughout the nerve tissue.²⁴

Statistical analysis

All results are presented as the mean ± standard error of the mean (SEM). Data was analyzed using GraphPad Prism 9 (GraphPad Software, La Jolla, CA, USA). The criteria for statistical significance in data analyses are shown in figure legends.

Results

Characterization of peptide-conjugated EVs

EVs were isolated from the conditioned culture medium of hMSCs using a TFF system. We conducted a two-step reaction to conjugate EVs with the RKK12 peptide targeting the integrin α2β1 receptor or the c(RGDyK) peptide, an inhibitor of integrin αVβ3 (Figure 1A). To analyze changes in the characteristics of peptide-conjugated EVs, we analyzed the expression of markers, such as CD63, CD9, CD81, and TSG101, and the size of the EVs (Figure 1B and C). TEM analysis showed that all EVs were 100-150 nm in diameter (Figure 1D). The levels of integrin expression in retinal precursor R28 cells were analyzed. The results confirmed that intracellular integrin avβ3 expression was higher in R28 cells than in CCD-986SK cells (Figure 1E), and cellular uptake of conjugated EVs was analyzed (Figure 1F). To evaluate the internalization level of c(RGDyK)_EVs in cells, flow cytometry (FACS) analysis was performed. As a result, cells treated with c(RGDyK)_EVs showed an increase in Cy5.5-positive intensity compared to the control group. Moreover, the addition of c(RGDyK) peptide effectively inhibited this increase (Figure 1G). The zeta potential of c(RGDyK)_EVs showed a significantly less negative charge compared with Naive_EVs. This reduction in the magnitude of the negative surface charge indicates that peptide conjugation altered the surface properties of the EVs (Figure 1H).

Figure 1.

Conjugation and characterization of EVs. (A) Schematic diagram of conjugating c(RGDyK), RKK12 peptides, and Cyanine-5.5 dye to EVs by a two-step. (B) Expression of exosomal marker, CD63, CD9, CD81, and TSG101 from MSCs and Naïve or conjugated EVs. (C) Size distributions of Naïve and conjugated EVs based on Nano particle Tracking Analysis (NTA). (D) Transmission electron micrograph (TEM) images of Naïve and conjugated EVs. Scale bar, 100 nm. (E) Verification of the expression of integrin family proteins in R28 and CCD-986SK cells. (F) Uptake analysis of Cy5.5-labeld EVs in R28 cells. Cells were incubated with 100 μg/mL of Cy5.5-labeled EVs. Confocal microscopy demonstrated the presence of Cy5.5-labeled EVs in cells after 1 or 24 h of incubation. Magnification × 400. (G) Flow cytometric analyses of R28 cells (control) or Cy5.5_Naïve EVs, or c(RGDyK)_EVs. Cy5.5-c(RGDyK)-EVs were added with free c(RGDyK) peptide. (H) Zeta potential measurements were performed after peptide conjugation. (*P < .05, **P < .01). All figures are original and were created by the authors.

c(RGDyK)_EVs showed a greater effect on cellular recovery following hypoxic damage

To investigate their effects on recovery of cells damaged by hypoxic conditions, R28 cells were exposed to CoCl₂ for 9 h, and then, treated with the two types of peptide-conjugated EVs. c(RGDyK)_EVs significantly increased the expression levels of the RGC markers Brn-3a and Rbpms (1.25- and 1.26-fold, respectively), compared with the CoCl₂ group, in contrast to RKK12_EVs (Figure 2A). In addition, the expression levels of neuroregeneration-associated proteins, including Neurofilament (Nf) and Neuronal nuclear protein (NeuN), were decreased under hypoxic conditions. Nf levels showed significant recovery after treatment with c(RGDyK)_EVs or RKK12_EVs (1.17- and 1.18-fold, respectively). In contrast, NLR family pyrin domain containing 3 (Nlrp3) inflammasomes were rapidly increased by CoCl₂ treatment, which was only reduced significantly by c(RGDyK)_EVs (Figure 2A).

Figure 2.

Hypoxic damaged R28 cells treated with conjugated EVs and single-cell RNA sequence analysis. (A) Brn-3a, Rbpms, Nf, NeuN, and Nlrp3 expression levels were presented from R28 lysates with CoCl₂.Two types of conjugated EVs were treated in CoCl₂-exposed R28 cells. After 24 h, a western blot analysis was performed. Data were expressed as the percentage (means ± SEM). Statistical significance was determined using a nonparametric statistical test, followed by the Mann–Whitney U test (**P < .01, ***P < .005, ****P < .001 vs the control; #P < .05, ##P < .01 vs CoCl₂). (B) Uniform manifold approximation and projection (UMAP) representation visualized the seven subpopulations in the three groups [CoCl₂, CoCl₂+c(RGDyK)_EVs and CoCl₂+RKK12_EVs]. (C) Heatmap representing the alteration in genes following treatment with CoCl₂ in R28 cells. (D) Distribution of expression of target genes in each cluster within the integrated group. The (E) violin plot and (F) dot plot display the expression of the target genes across three groups. (Ga1-4) Gene ontology (GO) representing the biological process (BP), cellular component (CC) and molecular functional (MF) expression of altered genes across three groups (*P < .05).

Single-cell RNA-seq revealed different effects of c(RGDyK) and RKK12_EVs on R28 cells damaged by hypoxia

Single-cell RNA-seq analysis of R28 cells was performed under three conditions: CoCl₂, CoCl₂+c(RGDyK); c(RGDyK) and CoCl₂+RKK12; RKK12. The results of single-cell RNA-seq were prefiltered using the CellRanger pipeline, yielding estimates of 5214, 6652, and 5714 cells from CoCl₂, c(RGDyK), and RKK12 groups, respectively. We visualized single-cell populations from CoCl₂, c(RGDyK), and RKK12 groups within a total of seven clusters by creating UMAP plots to characterize their heterogeneity (Figure 2B). The analysis revealed 53 DEGs after treatment of R28 cells with CoCl_2, the expression of which differed significantly by at least 1.5-fold (27 upregulated genes and 26 downregulated genes) with an adjusted P-value < .05 (Figure 2C). Seurat was used to identify the DEGs within each cluster, and the target genes of interest in each cluster were visualized on a stacked violin plot (Figure 2D). The samples from the three groups exhibited different expression patterns. Specifically, the anti-apoptosis-related gene, Egr1, and angiogenesis-related gene, Ankrd1, were significantly enriched within the upregulated genes, and the proapoptotic gene, Oscp1, was downregulated in the c(RGDyK) group (Figure 2E); a dot plot showed that Oscp1 and Csnk11a1 were enriched in the CoCl₂ group (Figure 2F). After treatment of R28 cells with CoCl_2, followed by subsequent treatments with c(RGDyK)_EVs or RKK12_EVs, GO analysis was performed to elucidate the biological functions of the target genes showing altered expression. The target genes were related to biological processes, cellular components, and molecular functions on GO analysis (Figure 2G). Cyclic (RGDyK)_EVs enhance ribosomal biosynthesis and mitochondrial function while inhibiting RNA splicing and stress-associated transcriptional activity, suggesting improved protein synthesis and energy metabolism under damage by CoCl₂. Conversely, RKK12_EVs downregulated mitochondrial respiration and ribosomal components but strongly upregulated immune and antiviral response pathways, indicating a transition to immune activation. Based on the results of in vitro and single-cell RNA-seq analyses of two types of peptide-conjugated EVs, the c(RGDyK)_EVs were used as surface-conjugated EVs (SC_EVs) for optic nerve injury models in this study.

Effects of SC_EVs on recovery of retinal RGCs in optic nerve injury animal models

The SC_EVs were injected into the subtenon of two optic nerve injury animal models, i.e., ONC and ONI rat models (Figure 3A). To determine the persistence of injected EVs, naïve EVs and SC_EVs labeled with Cy5.5 were injected. The labeled EVs were clearly observed for seven days post-injection, indicating that they settle around the optic nerve and influence its recovery (Figure 3B).

Figure 3.

Surface conjugated EVs promote RGC and axon regeneration in Optic nerve compression (ONC) & Optic nerve inflammation (ONI) models. (A) The in vivo experimental scheme for EVs injection and analysis. (B) The distribution of EVs was assessed seven days post-injection. The EVs were labeled with Cy-5.5. (C) Representative confocal microscopy-based fluorescence images of Brn-3a and Ankrd1 (n = 3/each group) staining of sham, EVs and SC_EVs groups in ONC model (scale bar: 100 μm). (D) In ONI model, the images of Brn-3a and Iba-1 positive cells were presented (n = 3/each group). Three zones were imaged and calculated from each retina flat mount (scale bar: 100 μm). (E) Quantitative analysis of optic nerve thickness in ONC and ONI models at 1- and 4-week post-injection. Optic nerve diameter was measured at 1000 μm intervals distally from the globe-facing end of the optic nerve. For each section, three independent diameter measurements were obtained using ZEN 3.1 software and averaged for analysis (*P < .05, ***P < .005 vs the age-matched sham; ###P < .005, ####P < .001 vs Normal). (F) H&E stain of an injured area (white dot line) on the optic nerve in ONC model (scale bar: 200 μm). Iba-1 and Gfap fluorescence quantification were measured in a box with 7690 mm² including the injured area (scale bar: 100 μm). (G) Representative H&E-stained sections from the ONI model evaluated using a semi-quantitative inflammatory scoring system: score 1, occasional infiltration of single leukocytes; score 2, focal infiltration of clustered leukocytes; and score 3, coalescing or widespread leukocyte infiltration. In parallel, ICAM-1 and F4/80-positive areas were quantified (scale bar: 100 μm). Data were expressed as the percentage (means ± SEM). Statistical significance was determined using a nonparametric statistical test, followed by the Mann–Whitney U test (*P < .05, **P < .01, ***P < .005 vs the age-matched sham; #P < .05, ##P < .01, ###P < .005 vs age-matched EVs; +P < .05, +++P < .005 vs zone of sham of 1 week).

Recovery of RGCs in the rat retina was investigated by counting the numbers of retinal RGCs positive for Brn-3a expression at 1 and 4 weeks after treatment with EVs in the ONC and ONI groups. The retina was divided into zones 1, 2, and 3, and Brn-3a-positive cells were counted in each region. We found that the numbers of Brn-3a-positive cells were significantly increased in all zones in the EV and SC_EV groups at 1 week in the ONC model (Figure 3C). After 1 week, the recovery of Brn-3a-positive cells was significantly increased in the SC_EVs groups compared to the EV group (P < .05). At 4 weeks, however, significant recovery of Bnr-3a-expressing cells was observed only in zones 1 and 3 in the SC_EV group (Figure 3C). In addition, Ankrd1 expression was observed around Brn-3a-positive cells. We also found that SC_EVs promoted Ankrd1 expression up until 4 weeks after injection (Figure 3C). At 1 week in the ONI model, recovery of Brn-3a-positive cells was observed only in zone 1 in the SC_EV group (Figure 3D). In response to LPS-induced inflammation, the expression of the microglia marker, Iba-1, was increased in both EV and SC_EV groups at 1 week after injection (Figure 3D).

Neuroprotective effects of SC_EVs against optic nerve damage in the ONC and ONI models

Histological examination revealed morphological changes in the optic nerve at 1 and 4 weeks after EV injection in both ONC and ONI models. Optic nerve thickness was measured every 1000 μm distally from the end of the optic nerve tissue on the side of the eyeball on H&E-stained sections. In ONC models, the optic nerve thickness in the sham group was significantly decreased, compared to normal controls, at 4 weeks. After SC_EV injection, the optic nerve thickness showed significant recovery by 1.18-fold at 4 weeks. In contrast, SC_EV injection did not significantly restore the optic nerve thickness in the ONI model, compared to normal controls, at 4 weeks after injection (Figure 3E).

We evaluated the effects of EVs on optic nerve recovery in the ONC and ONI models. In ONC models, treatment with EVs or SC_EVs induced significant recovery of the injured area by 0.65- and 0.7-fold at 1 week after injection (Figure 3F). This reflects a 35% and 30% reduction in the injured area compared to sham. The level of Gfap expression, the major protein of the astrocyte cytoskeleton, was significantly increased by 1.74-fold at 4 weeks after SC_EV injection. In addition, the level of Iba1 expression in the optic nerve, which was markedly increased by 6.46-fold in the ONC model, was reduced in the SC_EV treatment group compared to sham controls at 1 week after injection but was significantly increased at 4 weeks (Figure 3F). The inflammation scores in the ONI models were significantly decreased by 0.67- and 0.5-fold, respectively, following SC_EV treatment, compared with an age-matched sham group (Figure 3G). This means that the inflammation levels were reduced by 67% and 50% following SC_EVs treatment, respectively. ICAM-1 expression was reduced by 0.54-fold at 1 week but markedly increased by 17.65-fold at 4 weeks after injection of SC_EVs in ONI model. In addition, injection of EVs or SC-EVs downregulated F4/80 expression by 0.45- and 0.5-fold, respectively, at 4 weeks (Figure 3G).

SC_EVs promoted neuronal repair and reduced inflammation of retina and optic nerve in optic nerve injury

Immunoblotting analysis was performed to assess protein expression in both the retina and optic nerve tissues of two distinct optic nerve injury models. In the ONC model, there were significant increases in the expression levels of the retinal hypoxia-related proteins Hif1α (1.16-fold) and Vegf (1.50-fold), neuronal-associated proteins Nf (2.52-fold), Gap43 (1.51-fold), and Syntaxin-12 (2.91-fold), RGC marker Brn-3a (3.20-fold), and mitochondrial homeostasis-related proteins Mitofusin-2 (Mfn2; 1.45-fold) and Peroxiredoxin 2 (Prdx2; 1.51-fold) at 1 or 4 weeks after SC_EV injection compared with the sham group (Figure 4A). Moreover, the levels of Hif1α (1.35-fold), Vegf (1.68-fold), Nf (2.49-fold), Gap43 (1.50-fold), Syntaxin-12 (2.47-fold), and Brn-3a (2.09-fold) expression were significantly increased after SC_EV injection compared with EV injection. In addition, there was a significant increase in Vegf expression at 4 weeks in the optic nerve in the ONC model (2.61-fold; Figure 4B).

Figure 4.

Analysis of ONC and ONI in vivo models. Changes in target proteins were assessed by immunoblot analyses of rat (A) retina and (B) optic nerve extracts in ONC model. The samples were analyzed at 1 and 4 weeks after injection with optic nerve injury. Changes in levels of target proteins were presented in (C) retina and (D) optic nerve of ONI model. Quantified expression values of target proteins of OS and OD tissues were normalized to β-actin then, the values of OS were divided OD. The results are expressed as the mean ± standard error of the mean (SEM) of the independent retina and optic nerve analyses and are expressed as fold changes compared to the control (*P < .05, **<0.01, ***<0.001 vs the age-matched sham; #P < .05, ##<0.01 vs EVs). OD, oculus dexter; OS, oculus sinister.

In the ONI model, the expression levels of the retinal inflammation-related proteins Tnfα (0.30-fold), Nlrp3 (0.24-fold), IL-6 (0.26-fold), and IL-17 (0.19-fold) were significantly decreased at 1 or 4 weeks after SC_EV injection in comparison with the sham group. Nlrp3 and IL-17 expression showed significant reductions (0.24- and 0.52-fold, respectively) after SC_EV injection compared with EV injection. In addition, significant increases in the expression levels of the neuronal-associated proteins BDNF (1.42-fold at 1 week), Nf (1.21-fold at 1 week, 1.29-fold at 4 weeks), syntaxin-12 (2.01-fold at 1 week, 1.34-fold at 4 weeks), Brn-3a (1.25-fold at 1 week), and Prdx2 (1.37-fold at 1 week) were observed after SC_EV injection compared with the sham group (Figure 4C). In the optic nerve of ONI model, significant increases were observed in expression of Vegf (2.05-fold), Gap43 (3.22-fold), and syntaxin-12 (1.35-fold) compared with the sham group. In addition, the levels of Nlrp3 and IL-6 expression were also decreased by 0.58- and 0.68-fold after SC_EV injection compared with the sham group. Moreover, Vegf and Gap43 expression were significantly increased by 1.92- and 3.30-fold, respectively, after SC_EV injection, compared with EV injection. These results suggested that treatment with SC_EVs facilitated hypoxia-, neuronal, mitochondrial regulation in the ONC model and reduction of inflammation in the ONI model (Figure 4D).

Identification of target genes and functional classification of SC_EVs under in vitro hypoxic conditions

Figure 5A shows a heat map of the up- and downregulation of genes in CoCl₂ and CoCl₂ + SC_EV treatment groups. Network analysis was also performed to elucidate the interactions among target genes, such as Ankrd1, Egr1, Ccn1, Ccn2, and Serphine1 (Figure 5B). CoCl₂ treatment strongly activates stress and inflammation-related genes, and SC_EVs treatment significantly reverses these changes, suggesting a protective or recovery effect against damage caused by CoCl₂.

Figure 5.

Cluster analysis of differentially expressed genes (DEGs) by SC_EVs in CoCl₂-treated R28 cells. Identification of DEGs in treated groups (CoCl₂ and CoCl₂ + SC_EVs). (A) Heatmap of DEGs (P ≤ .05 and log2FC ≥ 1) in lysates of two groups analyzed by hierarchical clustering. High expression is shown in red; low expression is shown in green. (B) Identification of protein–protein interaction of DEGs in experimental groups of R28 cells. (C) The heatmap represented the functional expression of target genes.

Genes showing alterations in expression were categorized based on their functions (e.g., angiogenesis, apoptosis, neurogenesis, and inflammation) and subsequently organized into a heat map for visualization (Figure 5C). Expression of angiogenesis-related genes activated by CoCl₂ decreased after treatment with SC_EVs. Certain genes, such as CCN family member 1 (Ccn1), Ccn2, Ankyrin repeat domain 1(Ankrd1), and Tgfb2, showed significant expression changes between the two groups. This suggests that SC_EVs may inhibit or regulate the angiogenic gene expression triggered by CoCl₂-induced stress. Additionally, CoCl₂ activates apoptotic pathways. However, treatment with SC_EVs led to a reduction in the expression levels of apoptosis-related genes. Notably, genes such as Fam162a, Cryab, Id1, and Oscp1 showed marked differences in expression between the groups. These findings indicate that SC_EVs may exert a protective effect by mitigating CoCl-induced apoptosis. In the CoCl₂-treated group, most neurogenesis-related genes were downregulated. SC_EVs treatment led to increased expression of genes such as Hmgb2l1, Egr1, and Jun, suggesting that SC_EVs may partially restore or preserve neurogenesis-related gene expression. Furthermore, CoCl₂ suppressed the expression of inflammatory genes. In the SC_EVs group, some of these suppressed genes showed partial recovery, indicating that SC_EVs may help repair or enhance inflammatory gene expression. In particular, Hmgb2l1 has further increased expression after SC_EVs treatment, suggesting that certain inflammatory responses may be selectively increased by SC_EVs, contributing to the regulatory or balanced effect on inflammation (Figure 5C).

Regulation of target proteins expression of SC_EVs in optic nerve injury

Based on the results of single-cell RNA-seq analysis, we selected Ankrd1 and Ccn2 as genes associated with angiogenesis, Oscp1 as associated with apoptosis, Egr1 as associated with neurogenesis, and Csnk1a1 as associated with inflammation, and confirmed their expression in the retina and optic nerve of ONC model rats. In the ONC model, retinal Oscp1 expression was significantly decreased by 0.53-fold at 4 weeks after injection and retinal Csnk1a and Ankrd1 expression were increased by 1.56- and 1.89-fold, respectively, at 1 week in the SC_EV group compared with the sham group (Figure 6A). Moreover, in the optic nerve of ONC model rats, Oscp1 expression was significantly decreased (0.43-fold at 4 weeks), while significant increases were observed in Csnk1a1 (1.29-fold at 1 week, 1.89-fold at 4 weeks), Ankrd1 (1.66-fold at 1 week), and Ccn2 (1.92-fold at 4 weeks) in the SC_EV group compared with age-matched sham controls (Figure 6B).

Figure 6.

Validation of target genes obtained via single-cell RNA sequencing in vivo/vitro models. Levels of interested proteins in (A) the retina and (B) optic nerve of ONC were presented. During recovery process by SC_EVs, protein levels involved in signal pathway were confirmed in (C) retina of ONC models and (D) R28 cells (*P < .05, **P < .05 vs the age-matched sham; #P < .05, ##P < .01 vs EVs). At 9 h prior to SC_EVs treatment, R28 cells were exposed to 200 µM of CoCl₂ and 1 µM of TM2 TEAD inhibitor. Then, by (D) immunoblot and (E) immunofluorescence were performed for expression of proteins. The results are expressed as the mean ± standard error of the mean (SEM; scale bar: 10 μm). Statistical significance was determined using a nonparametric statistical test, followed by the Mann–Whitney U test (*P < .05, **P < .01 vs control; #P < .05, ##P < .01vs CoCl₂; ++P < .01 vs CoCl₂+SC_EVs).

The Wnt3a-Yap-Taz signaling pathway is involved in the SC_EV-mediated recovery process

Using single-cell RNA-seq, we analyzed the expression of target genes related to signaling pathways in the retina of the ONC model. Regulation of the YAP-TAZ-TEAD1 cascade pathway was analyzed (Figure 6C). At 1 week following injection of SC_EVs, Yap levels were recovered by 2.25- and 1.57-fold in the EV and SC_EV treatment groups, respectively, compared to sham controls. Taz and Tead1 expression levels were increased by 2.16- and 1.17-fold, respectively, 1 week after SC_EV injection. Phosphorylated smad2 and smad3 was significantly increased 1.3- and 2.46-fold, respectively, compared to sham, 4 weeks after SC_EVs injection. Conversely, Wnt3a level was downregulated by 0.59- and 0.79-fold at 1 and 4 weeks after SC_EV injection, respectively (Figure 6C), and were also confirmed in vitro using R28 cells and the TEAD inhibitor TM2. In hypoxia-damaged R28 cells, Yap and Ankrd1 levels were recovered by 1.23- and 1.42-fold in the SC_EV treatment group compared with CoCl₂-exposed group, while SC_EVs had no effect on the expression of either Taz or Tead1 (Figure 6D). We found that the Yap and Taz protein levels were altered by CoCl₂ exposure, and the relations between their protein expression and localization after hypoxic damage or TM2 treatment were investigated. The expression of Yap in the CoCl₂-exposed group seemed to be increased in the nucleus but was not clear and was not affected by treatment with EVs or TM2. A portion of Taz protein was translocated to the cytosol (white arrowhead) by CoCl₂ treatment. SC_EVs ameliorated this change in localization in hypoxia-damaged R28 cells without TM2 treatment (Figure 6E). Treatment of hypoxia-damaged R28 cells with SC-EVs upregulated the expression of Ankrd1, but this effect was not seen in the presence of TM2 (Figure 6E).

Discussion

There is a great deal of interest in functionally enhanced EVs in the field of regenerative medicine. Surface modification of EVs involves altering their surface properties, enabling targeted delivery, and enhancing stability.⁴ Enhanced targeting and specificity via integrins, which are cell-surface receptors around the optic nerve, play crucial roles in cell–cell and cell−extracellular matrix (ECM) interactions. Integrin-targeting peptides, such as RGDyK, possess affinity for specific integrin subtypes expressed on the surface of various cells, including those involved in hypoxic or ischemic injury.²⁵ Conjugating EVs with these peptides allows enhanced targeting and specificity, directing them toward cells relevant to optic nerve regeneration or protection, such as RGCs.³ Modified EVs carrying these peptides may adhere more efficiently to the surfaces of target cells, thus, enhancing their uptake and allowing the delivery of cargo molecules, such as neuroprotective miRNAs, growth factors, anti-inflammatory agents, or neurotrophic factors. RGDyK-conjugated EVs could enhance the targeted delivery of these molecules, and thus, increase therapeutic efficacy, promote neuronal survival, and improve optic nerve repair.²⁶

In ONC injury, microglia may initially respond to the mechanical insult by becoming activated and releasing proinflammatory factors. This activation of microglia may contribute to the secondary inflammatory response, exacerbating tissue damage, and inflammation over time.²⁷ In addition, compression-induced ischemia and tissue damage may lead to recruitment of peripheral immune cells, such as macrophages, further amplifying the inflammatory cascade.² In contrast, in LPS-induced ONI, microglia are directly activated by binding of LPS to Toll-like receptor 4 (TLR4), initiating a primary inflammatory response.²⁸ This activation leads to the rapid production and release of proinflammatory cytokines and chemokines, contributing to the acute inflammatory process.²⁹ Peripheral immune cells may also be recruited to the site of inflammation in response to released cytokines, further amplifying the inflammatory response. These processes are known to be ameliorated by the anti-inflammatory actions of miRNAs associated with EVs.^30,31

Interestingly, certain integrin subtypes, including αvβ3, are involved in the activation of latent TGF-β complexes and the extracellular matrix (ECM). Dysregulation of integrin-mediated TGF-β activation has been implicated in various pathological conditions, including fibrosis, cancer, and inflammation. Specifically, αvβ3 integrins have been associated with promoting TGF-β-driven fibrotic responses and facilitating tumor progression through interactions with the TGF-β pathway.³¹

YAP/TAZ activity has been reported to modulate Smad phosphorylation in response to TGF-β stimulation,³² with activation enhancing Smad2/3 phosphorylation and nuclear translocation, thereby increasing TGF-β signaling output, whereas inhibition suppresses Smad phosphorylation and attenuates signaling. Additionally, YAP/TAZ signaling influences angiogenesis by regulating microvascular endothelial cell behavior and network formation in response to VEGF.³³

Regarding the Hippo-YAP/TAZ pathway, accumulating evidence indicates crosstalk between YAP/TAZ and Wnt signaling, including β-catenin and Wnt receptors, with CSNK1A1 potentially regulating this interaction.³⁴ Because YAP/TAZ cannot directly bind DNA, they require TEAD transcription factors to regulate Hippo-responsive gene expression.³⁵ Transcriptional targets of the TEAD-YAP/TAZ complex are involved in cell proliferation, cell survival, immune evasion, and stemness.³⁶ Direct targeting of YAP/TAZ with small molecules has been challenging; however, the TEAD inhibitor TM2 exhibits strong antiproliferative effects alone or combined with MEK inhibitors in YAP-dependent cancer cells.³⁷ However, TEAD inhibition or YAP/TAZ knockdown temporarily inhibits cell cycle progression,³⁸ which may explain the results of intracellular immunofluorescence image analysis after TM2 treatment in the present study.

Recent preclinical studies have demonstrated that MSC-derived EVs can exert sustained neuroprotective and regenerative effects in optic nerve and retinal injury models. For example, EVs derived from human embryonic stem cell–derived MSCs, administered via tail-vein injection, significantly improved RGC survival, prevented retinal nerve fiber layer (RNFL) thinning, and enhanced GAP43-positive axon counts in a mouse optic nerve crush model over a 2-month period.³⁹ EV-based therapies provide long-term structural and functional protection of retinal neurons and demonstrate translational potential due to their safety, low immunogenicity, and ease of administration.

Building upon these studies, we used c(RGDyK)-modified EVs to enhance targeted delivery and uptake in injured retinal and optic nerve tissue. Given the established survival and axonal regeneration of RGCs months and the preservation of RGCs following EVs treatment up to 2 months post-injury, our results suggest that peptide-modified EVs may further improve structural recovery and functional outcomes.

In the present study, we evaluated the effects of c(RGDyK)_EVs, administered via subtenon injection, on optic nerve injury over a 4-week period. While significant improvements were observed, the long-term durability of these effects remains to be determined. Future studies with extended follow-up will be essential to assess sustained efficacy, late-phase neuroprotection, and functional recovery of RGCs, providing a more comprehensive evaluation of the translational potential of peptide-modified EVs for chronic optic nerve and retinal injuries.

Conclusion

In conclusion, SC_EVs modulate cellular responses to injury, including the YAP/TAZ signaling pathway in ONC models. Furthermore, they can attenuate fibrotic responses, enhance tissue repair, and exert neuroprotective effects by regulating both the YAP/TAZ and Smad2/3 signaling pathways.

Author contributions

Mira Park (Data curation [equal], Investigation [lead], Methodology [lead], Writing—original draft [equal], Writing—review & editing [supporting]), Hyun Ah Shin (formal analysis [equal], Investigation [equal], Methodology [equal]), Hey Jin Lee (Investigation [equal], Methodology [equal]), Jung Hyun Moon (Methodology [equal]), Jun Yong Kim (Resources [equal]), Won-Kyu Rhim (Resources [equal]), Dong Keun Han (Supervision [equal]), and Helen Lew (Conceptualization [lead], Funding acquisition [lead], Project administration [lead], Writing—original draft [lead], Writing—review & editing [lead])

Supplementary material

Supplementary material is available at Stem Cells Translational Medicine online.

Funding

This research was supported by grants from the National Research Foundation of Korea (NRF): 2021R1A2C2010523.

Conflicts of interest

All authors of this paper declare no potential conflicts of interest.

Data Availability

The data that supports the findings of this study are available from the corresponding author upon reasonable request.