Stem cell secretome- and extracellular vesicle-based therapies for retinal pathologies

Table 1 Bioactive factors of the stem cell secretome and relevance to retinal homeostasis and disease modulation

Secretome component	Representative cargo or mediators	Main retinal targets	Potential biological role in retinal disease	Translational relevance	Ref.
Soluble growth and trophic factors	VEGF, HGF, IGF-1, FGF-2, BDNF, NGF, TGF-β, and related neurotrophic mediators	Photoreceptors, retinal ganglion cells, RPE, Müller cells, and vascular cells	Support neuronal survival, tissue repair, stress resistance, and retinal microenvironmental homeostasis	May contribute to neuroprotective and disease-modifying effects without direct cell replacement	[19,20]
Cytokines and chemokines	Anti-inflammatory and immunomodulatory cytokines; regulators of TNF-α, IL-1β, NF-κB, and related pathways	Microglia, Müller cells, astrocytes, and infiltrating immune cells	Modulate neuroinflammation, reduce glial activation, and limit secondary inflammatory injury	Relevant to retinal degeneration, diabetic retinopathy, uveitis-related inflammation, and ischemic retinal injury	[12,21]
EVs and exosome-enriched fractions	EVs carrying proteins, lipids, metabolites, messenger RNA, microRNAs, long non-coding RNAs, and other regulatory molecules	RPE, photoreceptors, retinal ganglion cells, microglia, and endothelial cells	Transfer complex molecular cargo and regulate multiple pathways simultaneously	Represent a central platform for acellular regenerative therapy and engineered retinal delivery	[9-11]
Regulatory RNAs	MicroRNAs, long non-coding RNAs, and other regulatory non-coding RNAs	Microglia, photoreceptors, RPE, and endothelial cells	Regulate inflammation, apoptosis, oxidative stress, angiogenesis, and cell-survival pathways	Candidate therapeutic cargoes for disease-specific or engineered EV products	[15-17]
Proteins and enzymes	Antioxidant, anti-apoptotic, extracellular matrix-related, and signaling proteins; pathway regulators such as NEDD4	RPE, retinal neurons, glial cells, and vascular cells	Influence oxidative stress, apoptosis, autophagy, necroptosis, extracellular matrix remodeling, and cellular signaling	May serve as potency markers or mechanistic mediators of therapeutic activity	[14,28]
Lipids and metabolites	Bioactive lipids, metabolic intermediates, and membrane-associated molecules	Retinal neurons, RPE, vascular cells, and immune cells	Support membrane signaling, metabolic regulation, intercellular communication, and adaptation to cellular stress	May influence EV biodistribution, cellular uptake, and biological activity	[9,10]
Engineered or preconditioned secretome products	Hypoxia-primed EVs, cargo-loaded EVs, surface-modified vesicles, and peptide-targeted EVs	Disease-specific retinal cell populations	Enhance cellular targeting, modify cargo composition, and increase therapeutic potency	Promising, but requiring standardization, biodistribution studies, potency testing, and long-term safety validation	[22,23,27]

VEGF: Vascular endothelial growth factor; HGF: Hepatocyte growth factor; IGF-1: Insulin-like growth factor-1; FGF-2: Fibroblast growth factor 2; BDNF: Brain-derived neurotrophic factor; NGF: Nerve growth factor; TGF-β: Transforming growth factor β; RPE: Retinal pigment epithelium; TNF-α: Tumor necrosis factor α; IL-1β: Interleukin-1β; NF-κB: Nuclear factor kappa B; EV: Extracellular vesicle; NEDD4: Neuron-expressed developmental downregulation 4.

Table 2 Studies of cell-free secretome and extracellular vesicle therapies for retinal and optic nerve disorders

Ref.	Indication/model and evidence level	Product/source	Preparation and characterization	Route and dose (source-reported metric)	Comparator	Sample size and follow-up	Main endpoints and findings	Major limitations/design reporting
Sun et al[14]	Diabetic retinopathy; high-fat diet/STZ Sprague-Dawley rats and high-glucose RPE cells. Level 2; direct acellular evidence	Human umbilical-cord MSC-derived sEVs; NEDD4 identified as bioactive cargo	Conditioned medium; differential centrifugation followed by ultracentrifugation and 0.22-μm filtration. TEM, NTA and western blot; CD9, CD63, HSP70, Alix and TSG101 positive; Calnexin negative; mean diameter 171.3 ± 71.7 nm	Intravitreal; single dose of 1 × 106 particles in 3 μL at 12 weeks after STZ (particle-number dose)	PBS; HFL1-derived sEVs; normal control	Endpoint-dependent: Retinal thickness n = 6/group; biochemical/qPCR analyses n = 3/group. Animals evaluated 20 weeks after STZ, approximately 8 weeks after treatment	Retinal morphology/thickness, apoptosis, oxidative stress, NEDD4/PTEN/AKT/NRF2 signaling. MSC-sEVs reduced oxidative injury and apoptosis	Rodent metabolic model; single administration; relatively small and endpoint-dependent sample sizes; random allocation reported; blinding not clearly reported
Zhang et al[15]	Retinitis pigmentosa; rd10 mice, with complementary 661W/BV2 experiments. Level 2; direct acellular evidence	Human umbilical-cord MSC-derived EVs; miR-146a implicated in Nr4a3 regulation	EVs enriched by differential ultracentrifugation. CD63/CD81 characterization and particle-size analysis; 30-200-nm fraction predominated	Intravitreal; 1 × 1010 EVs in 1 μL/eye at P14 and P21 (particle-number dose)	Normal saline; WT mice	Outcome-specific group sizes; optomotor analysis included approximately n = 5-7/group. Functional/morphological assessment at P27-P28	Optomotor response, ERG, photoreceptor survival, retinal morphology, microglial/Müller-cell/macrophage activation, NF-κB and miR-146a/Nr4a3 signaling. EVs preserved photoreceptors and retinal function	Genetic mouse model; short developmental follow-up; endpoint-dependent sample sizes; no long-term assessment; blinding/randomization not consistently reported
Bian et al[24]	Inherited retinal degeneration; Royal College of Surgeons rats. Level 2; direct acellular evidence	Mouse neural progenitor cell-derived exosome preparation	Differential centrifugation plus sucrose-gradient density centrifugation. HSP70, CD63, CD9 and CD81 positive; TEM; NTA; approximately 50-180 nm, mean ~98 nm	Subretinal; 1 μL/eye containing approximately 1.09 × 109 vesicles (particle-number dose)	Vehicle and untreated controls	At least 3 biological samples/experiment; assessments through 28 days	OKR, scotopic ERG, outer nuclear layer preservation, photoreceptor apoptosis and microglial uptake/activation. Functional and structural preservation reported	Small experimental samples; rodent inherited-degeneration model; invasive subretinal delivery; follow-up ≤ 28 days; randomization/blinding incompletely described
Cui et al[25]	Optic-nerve crush in rats. Level 2; direct acellular evidence	Rat bone-marrow MSC-derived exosomes	Exosomes isolated from MSC-conditioned medium; TEM and western blot characterization; approximately 93-nm peak size; CD63/TSG101 positive and GM130 negative	Intravitreal; 3 × 109 vesicles in 5 μL, single administration	PBS/vehicle	NR for consolidated total sample size; approximately 30-day follow-up	RGC survival, retinal histology, apoptosis, inflammatory cytokines, Bcl-2/Bax/caspase-3 and PI3K/AKT signaling. Increased RGC survival and reduced inflammatory/apoptotic signaling	Acute traumatic optic-nerve model; sample size not consistently reported across endpoints; single administration; limited follow-up; randomization/blinding not clearly reported
Zhang et al[26]	Oxygen-induced retinopathy and laser-induced choroidal neovascularization in mice. Level 2; direct engineered-acellular evidence	Human MSC-derived exosomes engineered with miR-205	Differential ultracentrifugation; TEM, NTA and western blot; predominantly 100-150 nm; CD63, HSP70 and TSG101 positive. miR-205 loaded by electroporation	Intravitreal; administered at P12 in the OIR model and immediately after laser injury in the CNV model; injected particle-number/volume dose NR	PBS, unloaded exosomes and free miR-205	Principal in vivo analyses n = 5/group. OIR endpoint P17; CNV endpoint 7 days	Retinal/choroidal neovascularization, Evans-blue leakage, OCT, FFA and VEGFA/ANGPT2 signaling. Exo-miR-205 reduced pathological angiogenesis and leakage	Acute neovascularization models; small groups; very short follow-up; injected vesicle dose not clearly reported; authors noted lack of in vivo confirmation of uptake by retinal endothelial cells
Usategui-Martín et al[28]	Spontaneous degeneration of ex vivo porcine neuroretinal explants. Level 1; direct acellular secretome evidence	Human bone-marrow MSC secretome/conditioned co-culture medium; not a purified EV preparation	BMSCs characterized phenotypically; secretome consisted of medium obtained from NR-BMSC co-culture; no fractionation into soluble and EV components	Ex-vivo medium exposure; 0.75 mL conditioned medium plus 0.75 mL culture medium (conditioned-medium volume; not an isolated EV dose)	Fresh NR; monocultured NR; NR co-cultured with viable BMSCs	8 porcine eyes; 32 explants; n = 8 per experimental condition; 72 hours	Retinal morphology, RGC/photoreceptor markers, TUNEL, apoptosis, necroptosis, autophagy, oxidative-stress and inflammatory gene expression. Secretome attenuated several degeneration-associated changes	Ex vivo rather than in vivo model; short 72-hour exposure; secretome was not fractionated or compositionally standardized; cannot attribute effects specifically to EVs or soluble factors; selected analyses performed by a masked investigator
Zhang et al[29]	STZ-induced diabetic retinopathy in C57BL/6 mice and high-glucose microglial experiments. Level 2; direct acellular evidence	Mouse bone-marrow MSC-derived exosomes; let-7b-5p manipulated using agomir/antagomir strategies	BMSC-derived exosome preparation with vesicle morphology assessment; let-7b-5p cargo experimentally manipulated and quantified	Intravitreal; 1 μL of an exosome preparation reported at 5 μg/μL every 3 days for 6 weeks (protein-concentration metric)	Unmodified BMSC-Exo and let-7b-5p-modified exosome groups; control/PBS conditions	24 mice reported; 6 normal controls and 18 diabetic mice; n = 6 in diabetic treatment groups; 6-week intervention	Retinal histology/apoptosis, microglial activation, inflammatory cytokines and TLR4/ATF4 signaling. let-7b-5p-enriched exosomes reduced inflammatory and retinal injury markers	Single STZ mouse model; frequent repeated intravitreal injections; predominantly histological/molecular endpoints; no long-term post-treatment follow-up. Random allocation and blinded operators/assessors were reported
Chen et al[35]	Diabetic retinopathy; high-fat diet/STZ Sprague-Dawley rats plus cellular experiments. Level 2; direct acellular evidence	Human umbilical-cord MSC-derived sEVs carrying miR-22-3p	TEM and NTA; peak size approximately 159 nm and concentration ~2.9 × 1011 particles/mL; CD63, CD81 and TSG101 positive; Calnexin negative	Intravitreal; single dose of 1.5 × 109 sEVs in 5 μL after 12 weeks of diabetes (particle-number dose)	PBS-treated diabetic rats and normal controls	Consolidated animal NR; animals sacrificed 4 weeks after injection	Retinal morphology, blood-retinal-barrier leakage, microglial activation, inflammatory cytokines and NLRP3 inflammasome signaling. sEV-miR-22-3p attenuated inflammatory injury and barrier dysfunction	Rodent DR model; single administration; 4-week follow-up; no clinical functional endpoint; random allocation reported, blinding not clearly stated
Sun et al[37]	Diabetic retinal neurodegeneration in db/db mice, with additional STZ/genetic mechanistic experiments. Level 2; direct engineered-acellular evidence	MSC-derived EVs engineered for photoreceptor targeting with MH42 and enriched with USP25	EVs isolated and characterized by TEM, particle analysis and EV-marker profiling; engineered by cargo enrichment and surface targeting	Intravitreal; 1 μL containing 106, 107, or 108 particles; 107 particles were selected for repeated administration every 4 weeks (particle-number dose)	Unmodified MEVs, MEV-USP25, targeted and scrambled-control preparations, and disease controls	Animal sample size varied by endpoint; several quantified experiments used n = 3/group. Treatment followed to approximately 24 weeks	ERG, retinal morphology, ONL/photoreceptor survival, TUNEL, opsin expression and USP25/CRYAA-related ubiquitination. Targeted USP25 delivery improved structural and functional retinal outcomes	Complex engineered product; mouse models only; multiple comparator constructs and endpoint-specific sample sizes; repeated injections; no clinical validation. Data collection/analysis reported as blinded
Guo et al[38]	MNU-induced retinal degeneration and Pde6βrd1/rd1 mice. Level 2; direct engineered-acellular evidence	Photoreceptor-derived EVs (non-stem-cell source) engineered with MH42 targeting peptide and Cul7 siRNA	Ultracentrifugation; TEM/NTA and EV-marker analysis; CD9, CD63, Alix and TSG101 positive; Calnexin negative. siCul7 loaded by electroporation	Intravitreal; 1 μL at 1 × 107 particles/μL (particle-number concentration)	Targeted control-siRNA EVs and non-targeted siCul7-EVs	Sample size reported separately for each experiment/figure; no overall number of mice provided; 2-week follow-up	Photoreceptor targeting, retinal structure/function, Cul7/GPX4 signaling and ferroptosis. Targeted siCul7 delivery reduced photoreceptor degeneration	Non-stem-cell EV source and therefore supportive mainly as an engineered-delivery proof of concept; two mouse models; short 2-week follow-up; endpoint-dependent sample sizes
Tang et al[39]	Dry-AMD-related oxidative injury; ARPE-19 cells and NaIO3-induced retinal degeneration in rats. Level 1-2; direct acellular evidence	Adipose-derived MSC exosomes	MSC exosome preparation characterized in the original study; detailed characterization parameters should be interpreted according to the authors’ reported “exosome” terminology	Intravitreal in the in vivo model; exact administered dose NR	Oxidant/NaIO3 controls; mechanistic Nrf2 inhibition with ML385	Consolidated animal sample size and exact follow-up NR	LDH, ROS, SOD, RPE/photoreceptor morphology, Bcl-2/Bax and Nrf2/Keap1/HO-1 signaling. Exosome pretreatment reduced oxidative retinal damage	Oxidative/toxin model rather than spontaneous human AMD; pre-administration design; exact dose, consolidated sample size, and follow-up NR; no human evidence
Zhang et al[41]	Large and long-standing idiopathic macular holes; uncontrolled human case series. Level 3; exploratory human evidence, with a direct acellular MSC-Exo subgroup	Human umbilical-cord MSC-derived exosome preparation; 5/7 patients received MSC-Exos and 2/7 received viable MSCs	MSC-Exos isolated from cultured MSC supernatants using sequential ultracentrifugation	Intravitreal administration following pars plana vitrectomy and internal limiting membrane peeling, with intraocular tamponade; exact MSC-Exo dose NR	No untreated concurrent control; viable-MSC recipients did not constitute a randomized comparator	Seven patients overall; five MSC-Exo recipients and two viable-MSC recipients; exact follow-up NR	Six of seven macular holes closed and BCVA improved in five patients overall. One patient receiving a higher MSC-Exo dose developed an inflammatory reaction. Aggregate outcomes cannot be attributed exclusively to the MSC-Exo subgroup	Very small, uncontrolled, mixed-intervention case series; concomitant PPV, ILM peeling, and tamponade; two patients received viable MSCs; aggregate outcomes prevent attribution of independent efficacy to MSC-Exos; no randomized confirmatory study

Evidence levels were used as a review-specific descriptive framework: (1) Level 1, in vitro/ex vivo mechanistic evidence; (2) Level 2, in vivo animal evidence; (3) Level 3, early uncontrolled human feasibility/efficacy evidence; and (4) Level 4, controlled clinical efficacy evidence. “Direct acellular evidence” indicates administration or testing of a secretome, conditioned-medium, extracellular vesicle (EV), small EV, or engineered-EV preparation rather than viable-cell transplantation. This classification is conceptual and is not intended as a validated clinical grading system. Dose metrics are reproduced in the units reported by the primary source. Particle-number doses (“particles” or “vesicles”, as reported), protein-mass or protein-concentration doses (μg or μg/μL), and conditioned-medium volumes are analytically distinct and should not be directly compared. Particle counts are not assumed to represent purified EV numbers unless supported by the source methodology. No particle-to-protein or other cross-unit conversion was performed. NR indicates information not reported or not sufficiently specified in the primary source. STZ: Streptozotocin; RPE: Retinal pigment epithelium; MSC: Mesenchymal stem cell; sEVs: Small extracellular vesicle; NEDD4: Neuron-expressed developmental downregulation 4; TEM: Transmission electron microscopy; NTA: Nanoparticle tracking analysis; HSP70: Heat shock protein 70; TSG101: Tumor susceptibility gene 101; PBS: Phosphate buffered saline; qPCR: Quantitative polymerase chain reaction; AKT: Protein kinase B; NRF2: Nuclear factor erythroid 2-related factor 2; ERG: Electroretinography; NF-κB: Nuclear factor kappa B; OKR: Optokinetic response; RGC: Retinal ganglion cell; PI3K: Phosphatidyl-inositol-3-kinase; OIR: Oxygen-induced retinopathy; CNV: Choroidal neovascularization; OCT: Optical coherence tomography; FFA: Fluorescein angiography; VEGFA: Vascular endothelial growth factor A; ANGPT2: Angiopoietin 2; BMSC: Bone marrow mesenchymal stem cell; EV: Extracellular vesicle; BMSC-Exo: Bone marrow mesenchymal stem cell-derived exosome; TLR4: Toll-like receptor 4; ATF4: Activating transcription factor 4; DR: Diabetic retinopathy; ONL: Outer nuclear layer; siRNA: Small interfering RNA; GPX4: Glutathione peroxidase 4; AMD: Age-related macular degeneration; LDH: Lactate dehydrogenase; ROS: Reactive oxygen species; SOD: Superoxide dismutase; HO-1: Heme oxygenase-1; MSC-Exo: Mesenchymal stem cell-derived exosome; BCVA: Best-corrected visual acuity; PPV: Pars plana vitrectomy; ILM: Internal limiting membrane.

Table 3 Conceptual advantages and current limitations of secretome- and extracellular vesicle-based therapies compared with conventional cell-based approaches

Domain	Potential benefits of secretome- and EV-based therapy	Current limitations	Practical implication	Ref.
Therapeutic concept	Acellular approach based on paracrine signaling rather than direct cell replacement	Does not replace lost photoreceptors, RPE, or retinal architecture	More realistic as a disease-modifying or adjunctive strategy than as a complete regenerative cure	[3,7,12]
Safety profile	Avoids administration of viable proliferative cells	Long-term ocular safety, immunogenicity, toxicity, and off-target effects remain insufficiently defined	May reduce some risks associated with cell transplantation, but still requires rigorous preclinical and clinical safety testing	[3,41]
Immunological risk	Potentially lower risk of immune rejection compared with living donor cells	EV cargo may still trigger inflammatory or immune responses depending on its source and biological context	Source, purity, and inflammatory profile must be carefully characterized	[8,41]
Tumorigenic and differentiation risk	Avoids uncontrolled proliferation, ectopic differentiation, and tumor formation associated with viable cell grafts	Bioactive cargo may still have unintended effects on proliferation, angiogenesis, or fibrosis	Functional potency assays are required in addition to particle counting and marker analysis	[3,4,11]
Biological activity	Can deliver proteins, lipids, metabolites, mRNA, microRNAs, and other regulatory molecules	Cargo composition is heterogeneous and highly dependent on donor-cell source, culture conditions, and isolation methods	Batch-to-batch reproducibility is a central translational requirement	[9,10,12]
Mechanistic breadth	May simultaneously modulate inflammation, oxidative stress, apoptosis, angiogenesis, glial activation, and neuronal survival	Multifactorial effects make it difficult to identify the active components and dominant mechanisms	Mechanism-specific validation is required for each disease model	[7,12,17]
Manufacturing and storage	Potentially easier to isolate, store, standardize, and administer than living-cell products	Scalable production, purification, storage stability, and quality control have not yet been standardized	GMP-compatible workflows and validated release criteria are required	[11,49,51]
Engineering potential	EVs can be loaded or modified with selected RNAs, proteins, peptides, or surface ligands	Engineering may alter biodistribution, potency, safety, and regulatory classification	Engineered products require separate validation from native secretome or EV preparations	[23,27,48]
Retinal delivery	Small vesicular carriers may protect their cargo and facilitate intraocular delivery	Retinal penetration, cell-specific uptake, optimal route, dose, and clearance remain unresolved	Intravitreal, subretinal, suprachoroidal, and biomaterial-assisted delivery approaches should be compared systematically	[42,43,45]
Clinical evidence	Strong mechanistic rationale and encouraging preclinical findings across several retinal disease models	Human evidence remains limited and heterogeneous and does not demonstrate durable visual restoration	Clinical translation should proceed cautiously using disease-specific structural and functional endpoints	[13,41]
Disease-stage suitability	May be useful when viable retinal cells remain responsive to trophic, anti-inflammatory, or vascular-stabilizing signals	Benefits are likely to be limited in advanced disease with extensive photoreceptor or RPE loss	These therapies may be better positioned for early or intermediate disease or for combination treatment	[1,3,4]
Combination potential	May complement gene therapy, anti-VEGF therapy, neuroprotective drugs, biomaterials, or cell-replacement strategies	Optimal combinations, timing, sequencing, and patient selection remain unknown	Future studies should define whether each product is primarily neuroprotective, immunomodulatory, vascular-stabilizing, or delivery-enhancing	[3,23,46]

EV: Extracellular vesicle; RPE: Retinal pigment epithelium; VEGF: Vascular endothelial growth factor; GMP: Good manufacturing practice.

Table 4 Comparison of the present review with recent relevant reviews

Ref.	Scope	Primary emphasis	Main distinction from the present review
Manai et al[2]	EVs in degenerative retinal diseases	Retinal EV biology, pathophysiology, biomarkers, and therapeutic potential	The present review additionally separates soluble secretome, native EV, exosome-enriched, and engineered-EV product classes and links study-level heterogeneity to translational maturity
Yang and Peng[12]	Human stem-cell-derived EVs in retinal repair and regeneration	Molecular mechanisms and regenerative potential	The present review additionally emphasizes product taxonomy, direct vs indirect evidence, study-level methodological comparison, and translational readiness
Tseng et al[13]	Cell-free EV therapy across ocular diseases	Broad ocular applications, cellular sources, preclinical evidence, and clinical translation	The present review provides a retina/optic-nerve-focused synthesis with detailed study-level evidence mapping and an explicit evidence-maturity framework
Guan et al[27]	Engineered exosome therapies for ocular vascular disease	EV engineering, targeting, and ocular neovascular disease	The present review includes non-engineered secretome/EV products and both neurodegenerative and vascular retinal disorders in addition to engineered platforms
Present review	Stem-cell secretome- and EV-based therapies in retinal and optic-nerve disorders	Product classes, disease-specific evidence, delivery, manufacturing, quality control, and translation	Integrates evidence directness, product-class distinctions, study-level heterogeneity, evidence maturity, and disease-by-product mapping within a single retinal-focused framework

EV: Extracellular vesicle.

Table 5 Review-specific evidence-maturity framework

Level	Evidence category	Interpretation in this review	Principal limitations
Level 1	In-vitro, ex-vivo, or organoid mechanistic evidence	Biological plausibility and mechanistic proof-of-concept	Does not establish in-vivo exposure, biodistribution, safety, or therapeutic efficacy
Level 2	In-vivo animal efficacy evidence	Preclinical proof-of-concept in an intact biological system	Species/model differences, heterogeneous products and doses, and limited prediction of human efficacy
Level 3	Early uncontrolled human feasibility or safety evidence	Preliminary human feasibility, safety, or efficacy signal	Small samples, lack of concurrent controls, confounding/co-interventions, and inability to establish efficacy
Level 4	Controlled human clinical efficacy evidence	Comparative clinical evidence capable of testing efficacy	Requires adequate power, standardized products, appropriate controls, clinically meaningful endpoints, and external validity

This is a review-specific conceptual framework. It is not a validated grading system, a formal risk-of-bias instrument, or a substitute for GRADE. The level reflects translational maturity and evidence directness rather than overall methodological quality.

Table 6 Disease-by-product-class evidence map for retinal secretome- and extracellular vesicle-based applications

Disease/application	Direct acellular evidence	Supportive/indirect evidence	Highest maturity	Interpretation
Diabetic retinopathy	Native MSC-derived sEVs[14,35,36]; engineered MSC-derived EVs[37]	Viable-MSC evidence[30]	Level 2	Among the broadest preclinical application areas; no controlled human retinal evidence
Inherited retinal degeneration/retinitis pigmentosa	MSC-derived EVs[15]; neural progenitor cell-derived EV preparations[24]	Paracrine/miRNA evidence[16]; non-stem-cell photoreceptor EV engineering[38]	Level 2	Multiple animal studies, with substantial product-, model-, and dose-level heterogeneity
Optic-nerve injury	MSC-derived EV/exosome preparation[25]	Viable-cell neuroprotective evidence[32]	Level 2	Preclinical evidence only
Dry AMD/oxidative RPE injury	MSC-derived exosome preparation[39]	iPSC-RPE complement evidence[40] provides disease context rather than EV therapeutic evidence	Level 1-2	Preclinical evidence primarily from experimental oxidative/toxic injury models
Neovascular retinal disease	Engineered MSC-derived exosome/EV approaches[26,33]		Level 2	Preclinical anti-angiogenic proof-of-concept, particularly for engineered products
Large/refractory macular hole	MSC-derived exosome subgroup in an uncontrolled surgical case series[41]	Concomitant PPV, ILM peeling, tamponade, and viable-MSC cases	Level 3	Only direct human retinal signal in the reviewed corpus; independent EV efficacy cannot be established
Other retinal/optic-nerve applications	Sparse or heterogeneous direct evidence	Viable-cell, non-stem-cell EV, or disease-context studies	Level 1-2 or insufficient	Least mature areas; disease-specific validation remains necessary

MSC: Mesenchymal stem cell; sEVs: Small extracellular vesicles; EV: Extracellular vesicle; miRNA: MicroRNA; AMD: Age-related macular degeneration; iPSC: Induced pluripotent stem cell; RPE: Retinal pigment epithelium; PPV: Pars plana vitrectomy; ILM: Internal limiting membrane.