Science Harnessing semen-derived exosomes for noninvasive fundus drug delivery: A paradigm for exosome-based ocular fundus therapeutics - Translation: Scientists try treating eye cancer by farming pig semen to inject into the eyes of mice. Works pretty well!

https://www.science.org/doi/10.1126/sciadv.adw7275
Archive

JIANSONG ZHAO, YAXIN DENG, HONGBING LIU, MINGLI WEI, CHENXIAO CHU, XINXIN LIANG, XIAOSHUANG BI, HAIBING HE, JINGXIN GOU, XING TANG, AND YU ZHANG
SCIENCE ADVANCES
27 Mar 2026
Vol 12, Issue 13
DOI: 10.1126/sciadv.adw7275

Abstract​

Exosomes, despite their promise as drug carriers for crossing biological barriers, remain underexplored for noninvasive posterior ocular delivery. Here, we demonstrate that semen-derived exosomes (SEVs) penetrate ocular barriers effectively, owing to their epidermal growth factor expression, which mediates reversible tight-junction disruption. SEVs reach the posterior segment via dual corneal and conjunctival routes. Using this, we engineered FA-SEVs@CMG eye drops, where SEVs are modified with folic acid (FA) and loaded with a nanozyme system (CMG) composed of carbon dots, manganese dioxide, and glucose oxidase. This eye drop leverages SEVs’ excellent penetration ability and FA’s targeting effect to enhance drug delivery to retinoblastoma (RB) cells. Internalized CMG induces intense oxidative stress, disrupts the autophagy-apoptosis balance, and triggers RB cell self-destruction. In vivo, FA-SEVs@CMG effectively inhibits RB growth while preserving retinal function. This work establishes the first SEV-based platform for noninvasive posterior segment delivery, offering a transformative strategy for treating posterior ocular diseases.

INTRODUCTION​

Retinoblastoma (RB), the most prevalent intraocular malignancy in children, poses substantial therapeutic challenges due to its location within the retina and the delicate nature of ocular tissues (1). Current primary treatment modalities, including intravitreal injections, radiotherapy, cryotherapy, and systemic chemotherapy, often result in severe ocular structural damage and systemic toxicity. Moreover, patients at risk of extraocular spread frequently require enucleation, leading to permanent vision loss (2, 3). These limitations underscore the urgent need for noninvasive, targeted therapies that can effectively deliver drugs to the posterior segment of the eye while minimizing collateral damage. (4)
The posterior segment of the eye is protected by a complex array of biological barriers, including the blood-retinal barrier and the corneal epithelium, which severely restrict the penetration of therapeutic agents. However, the efficiency of noninvasive posterior segment drug delivery is substantially hindered by the complex and interconnected biological barriers of the anterior and posterior ocular segments (5). While nanoparticles, cell-penetrating peptides, and other carriers have been explored as potential delivery systems, their clinical translation has been hampered by low drug-loading capacity, poor selectivity, and systemic toxicity (6–8). In contrast, exosomes, nanoscale extracellular vesicles naturally secreted by cells, have emerged as promising candidates for drug delivery due to their biocompatibility, low immunogenicity, and ability to cross biological barriers (9). However, current research on exosome-based drug delivery has predominantly focused on injection-based methods, leaving the potential for noninvasive delivery largely unexplored (10).

Exosomes typically carry protein profiles reflective of their parent-cell characteristics (11). The distinct surface protein composition of exosomes derived from different cellular sources further determines their diverse functional properties (12). Notably, semen-derived exosomes (SEVs), which are present in semen and originate from the epididymis and prostate, have evolved to facilitate sperm penetration through the female reproductive tract (13, 14). This process mediated by specific proteins such as clusterin, prostaglandin D synthase (PTGDS), and acrosomal vesicle protein 1 (ACRV1) (15–17). These proteins enable SEVs to traverse formidable biological barriers, suggesting that SEVs may have inherent properties conducive to penetrating ocular barriers. This unique capability positions SEVs as a promising vehicle for noninvasive drug delivery to the posterior segment of the eye.

In parallel, the development of carbon dots (CDs) with peroxidase (POD)–like activity has opened promising avenues for localized cancer therapy. CDs can generate reactive oxygen species (ROS) in response to the elevated hydrogen peroxide (H2O2) levels characteristic of the tumor microenvironment (TME), offering a targeted approach to induce cancer cell death while sparing healthy tissues (18, 19). However, the therapeutic efficacy of CDs is often limited by the high glutathione (GSH; 1 to 10 × 10−3 M) levels and limited H2O2 concentrations (0.1 to 1 × 10−3 M) within the TME (20, 21). To overcome these challenges, we engineered a nanozyme system, CMG, composed of CDs with FeN3S structures, manganese dioxide (MnO2) nanozymes, and glucose oxidase (Gox). This system could enhance antitumor efficacy through sequential actions: MnO2consumed overexpressed GSH in TME (22), while Gox catalyzed glucose-to-H2O2 conversion (21), synergistically amplifying the POD-like activity of CDs for intensified chemodynamic therapy.

In this study, we demonstrate that SEVs exhibit superior posterior segment delivery efficiency compared to conventional cell-derived exosomes and cell-penetrating peptide-modified liposomes. This enhanced delivery is mediated by the transient opening of tight junctions (TJs).
in the epithelial barrier, facilitated by the unique protein composition of SEVs. Building on this finding, we developed an eye drop formulation, FA-SEVs@CMG, by modifying SEVs with folic acid (FA) and encapsulating the CMG nanozyme system. FA-SEVs@CMG leverages the excellent penetration ability enabled by SEVs and the targeting effect of FA to substantially enhance drug delivery to RB cells. Once internalized, the CMG system induces intense oxidative stress, shifting autophagy from a cell homeostasis protector to an amplifier of cell death and activating the extrinsic apoptotic pathway, ultimately leading to the self-destruction of RB cells. In vivo studies in RB model mice reveal that FA-SEVs@CMG effectively suppresses tumor growth while preserving retinal function, highlighting the potential of SEVs as a noninvasive, vision-preserving therapeutic strategy. Furthermore, the formulation enables real-time monitoring of tumor size through fluorescence imaging, offering a valuable tool for treatment assessment. By harnessing the unique properties of SEVs, this study not only advances the field of ocular drug delivery but also provides a blueprint for the development of noninvasive therapies for other posterior segment diseases (Fig. 1).


Fig. 1. Schematic illustration of the therapeutic role of FA-SEVs@CMG in RB.
SEVs could temporarily disrupt the tight junctions of the ocular barrier, facilitating the delivery of CMG to the RB tissue via both corneal and scleral pathways, thereby achieving an omnidirectional attack around RB. FA-SEVs@CMG generated substantial oxidative stress in response to the TME, which induced tumor cell death through a synergistic mechanism involving ferroptosis, apoptosis, and autophagy. Excessive oxidative stress induced autophagy, which activated the extrinsic apoptotic pathway (BID-Caspase-, further promoting apoptosis and inducing the self-destruction of RB cells.

RESULTS​

Synthesis of dual-functional CDs​

We synthesized CDs through a one-step solvothermal approach by reacting citric acid, thiourea, FeCl2, and boric acid in N,N′-dimethylformamide (DMF; Fig. 2A), aiming to create a dual-functional nanoplatform capable of both tracing SEVs (23) and exerting POD-like enzymatic activity (18, 19, 24). High-resolution transmission electron microscope (HRTEM) unveiled monodisperse CDs with a uniform diameter of 4.14 nm (Fig. 2B), exhibiting a distinct lattice spacing of 0.21 nm that aligns with the (100) crystallographic plane of graphitic carbon—a finding corroborated by x-ray diffraction analysis (fig. S1A). Elemental mapping (Fig. 2C) confirmed homogeneous codoping of N, S, B, and Fe, while Raman spectroscopy (I<em>D</em>/I<em>G</em> = 1.12; fig. S1B) revealed a crystalline and amorphous architecture, synergistically enhancing electron transfer for dual imaging-catalytic functionality. The Fourier transform infrared (FTIR) spectra (fig. S2) exhibited characteristic absorption peaks corresponding to O─H (3400 cm−1), C═O (1620 cm−1), C═C (1140 cm−1), and C─H (683 cm−1). Notably, vibrational modes associated with the boron-doping structure were distinctly observed at 1402 and 1087 cm−1, which correspond to B─O and B─C vibrations, respectively. Furthermore, the regions of 1000 to 1060 and 1300 to 1400 cm−1 were indicative of the stretching and bending vibrations of B─N bonds, while peaks in the 400- to 500-cm−1 range corresponded to the symmetric stretching of iron-sulfur bonds. The peaks observed at 570 to 600 cm−1 are attributed to C─S bonds. Collectively, these findings confirmed the formation of B/N/S/Fe codoped CDs.


Fig. 2. Synthesis of CDs and extraction of SEVs.
(A) Schematic illustration of the synthesis approach of the CDs. (B) TEM and HRTEM image of the CDs; the inset is the calculated average diameter (in nanometers) of particles. (C) The elemental atomic contents are estimated from the TEM energy dispersive X-ray spectroscopy mapping. (D) Schematic diagram of collection and isolation process of SEVs. (E) The TEM image of SEVs. The dynamic light scattering (DLS) (F) and nanoparticle-tracking analysis (NTA) (G) diameter of the SEVs (n = 3). (H) Immunoblotting analysis of exosome markers (CD9, CD63, and Alix), endoplasmic reticulum marker (calnexin), mitochondria marker (COX IV), and nucleus marker (histone H3) expressed in SEVs and supernatant + sperm cell.

Optical properties of CDs​

Building on the dual-functional design of CDs, we next validated their photophysical properties for real-time tracking. The CDs exhibited broad ultraviolet-visible (UV-Vis) absorption (450 to 600 nm; fig. S3A) with excitation-independent near-infrared (NIR) emission peaking at 621 nm (λex = 551 nm; fig. S3, B and C), a signature of homogeneous electronic transitions from a single emissive center. The CDs retained >80% of their initial fluorescence intensity after 96 hours of ambient storage (25°C; fig. S3D), demonstrating photostability that outperforms conventional NIR imaging agents (e.g., ICG: 50% decay in 24 hours).

Extraction and characterization of SEVs​

In this work, SEVs were isolated from porcine semen (Duroc), a cost-effective and abundant biological source, using differential ultracentrifugation (Fig. 2D). TEM imaging confirmed their classical exosomal morphology (Fig. 2E). Size distribution analyses by dynamic light scattering (DLS) [117.4 ± 23.2 nm; polydispersity index (PDI) = 0.221 ± 0.045] and nanoparticle-tracking analysis (NTA; 124.3 ± 16.8 nm) validated monodispersity (Fig. 2, F and G). Western blot analysis confirmed the expression of exosomal markers (CD63, CD9, and Alix) in SEVs while demonstrating nearly undetectable levels of endoplasmic reticulum (calnexin-negative), mitochondrial [cytochrome c oxidase subunit IV (COX IV)–negative], and nuclear (histone H3–negative). This indicated that the isolated SEVs were free from contamination by residual cells, such as sperm cells. Furthermore, exosomal markers remained absent in supernatant + sperm cell fractions postultracentrifugation, thereby validating the efficacy of our isolation protocol in obtaining pure SEVs from seminal fluid (Fig. 2H and fig. S4). These multimodal analyses collectively confirm the successful isolation of intact SEVs.

Validation of ocular permeability of SEVs​

Building on above advances, we integrated CDs into SEVs (SEVs-CDs), initiating exploration of SEVs’ potential for noninvasive posterior delivery. Y79 cell–derived exosomes (Y79Es-CDs), protamine [a type of cell-penetrating peptide from fish semen (25)] liposomes (ProLs-CDs), and free CDs were selected as control groups, and all groups were uniformly loaded with CDs at a concentration of 1 mg/ml. Despite similar carrier physicochemical profiles (hydrodynamic diameters: 110 to 125 nm; zeta potentials: −16 to −20 mV; Fig. 3A), SEVs-CDs exhibited superior transcorneal permeability. Franz diffusion assays (Fig. 3B) (26) demonstrated 14.0 ± 0.9% corneal penetration for SEVs-CDs at 10 hours, significantly exceeding ProLs (9.3 ± 0.5%), Y79Es (5.8 ± 0.7%), and CDs (2.0 ± 0.3%) (Fig. 3C). In addition, we constructed corneal epithelial cell monolayers (fig. S5) and investigated the Papp (apparent permeability coefficient) of different groups across the monolayers (fig. S6). Consistent with ex vivo rabbit corneal penetration assays, SEVs demonstrated the highest Papp value of 3.40 × 10−5 cm/s, approximately sevenfold greater than the CD-solution group (0.47 × 10−5 cm/s) and significantly higher than ProLs (2.33 × 10−5 cm/s) and Y79Es (1.13 × 10−5 cm/s). This disparity highlights that SEVs’ enhanced permeability arises from mechanisms beyond size and surface charge.

Fig. 3. Investigation of the ocular penetration capability of SEVs.
(A) Particle size and zeta potential profiles of CDs, ProLs-CDs, SEVs-CDs, and Y79Es-CDs. (B) Illustration of Franz transdermal diffusion with the rabbit cornea as the membrane. (C) Cumulative in vitro rabbit corneal penetration percentages of different groups. (D) Fluorescence confocal microscopy images of CD distribution in mouse ocular and retinal tissues following topical administration of distinct ophthalmic formulations at specified time points, accompanied by image quantification analyses (E) [4′,6-diamidino-2-phenylindole (DAPI): blue, CDs: red]. h, hours; a.u., arbitrary units. (F) Penetration of CDs in the whole mouse eyeball after applying different eye drops. Data were represented as means ± SD. P values in (B), (C), and (E) were calculated by one-way analysis of variance (ANOVA) with a Tukey post hoc test (*P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001) (n = 3).

Then, we conducted analysis of CD fluorescence distribution in mouse ocular tissues following topical administration (Fig. 3, D and E). Untreated eyes exhibited negligible background autofluorescence. Time-dependent fluorescence patterns revealed rapid posterior segment accumulation of SEVs-CDs, with retinal-choroidal signals peaking at 6 hours postinstillation before declining to baseline levels by 24 hours. In contrast, free CDs exhibited minimal fundus fluorescence at matched time points, while Y79Es-CDs and ProLs-CDs showed intermediate but substantially weaker signals compared to SEVs-CDs. Quantitative whole eyeballs penetration analysis demonstrated the marked superiority of SEVs-CDs, achieving peak ocular accumulation (4.8% penetration efficiency at 6 hours) that substantially exceeded all control formulations (Fig. 3F). These findings were further supported by analysis of rabbit ocular frozen sections (Fig. 4A). Detailed pharmacokinetic profiling of SEVs in rabbits revealed compartment-specific accumulation (Fig. 4, B to E). During the initial 6-hour accumulation phase, CD concentrations progressively increased in the retina and choroid. A subsequent redistribution phase showed decreased deposition in both the anterior (cornea, iris, and aqueous humor) and posterior (retina and choroid) segments, with increased vitreous accumulation at 12 hours. This confirms SEVs’ capacity for both rapid posterior segment accumulation and sustained intraocular drug redistribution. Collectively, multimodal analyses across mice and rabbit models establish SEVs as superior carriers for noninvasive posterior delivery, outperforming tumor-derived exosomes and ProL carriers in transocular barrier penetration and fundus biodistribution. Building upon prior studies demonstrating protamine’s capacity to transiently disrupt intercellular TJs and enhance paracellular transport (27), therefore, SEVs may also facilitate ocular penetration via similar paracellular pathways through TJ modulation.


Fig. 4. In vivo validation and mechanistic study of ocular permeability of SEVs.
(A) Confocal images of eyeballs collected from rabbits after applying different eye drops at different time points (DAPI: blue, CDs: red). Distribution of CDs in the cornea, iris, retina and choroid, lens, vitreum, and aqueous humor of rabbit eyes after applying SEVs-CDs eye drops for 1 (B), 6 (C), 12 (D), and 24 (E) hours. (F) Immunofluorescence imaging and quantitative analysis of ZO-1 distribution under different treatment conditions. (G) The Western blot analysis of (Ga) E-cadherin and (Gb) occludin in human corneal epithelial cell (HCEC) monolayers after incubation with different groups. Band intensities from Western blot analyses were quantified using ImageJ and normalized to the corresponding glyceraldehyde-3-phosphate dehydrogenase (GAPDH) levels. (H) Transepithelial electrical resistance (TEER) of HCECs across treatment groups monitored over a 23-hour period following 1-hour exposure and phosphate-buffered saline (PBS) washout. Data were represented as means ± SD. P values in (F), (Ga), (Gb), and (H) were calculated by one-way ANOVA with a Tukey post hoc test (*P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001) (n = 3).

Mechanistic study of SEVs enhanced ocular permeation​

To investigate whether SEVs enhance ocular penetration by modulating intercellular TJs to activate paracellular pathways, we investigated their interaction with corneal TJ architecture (28, 29). While both SEVs and ProLs reduced zonula occludens protein 1 (ZO-1), filamentous actin (F-actin), occludin, and epithelial cadherin (E-cadherin) expression (Fig. 4F and fig. S7) in human corneal epithelial cells (HCECs), critical distinctions emerged: SEVs preserved E-cadherin and occludin levels (Fig. 4, Ga and Gb, and fig. S, whereas ProLs caused significant TJ protein disruption (5, 30). In vivo investigations further confirmed the TJ modulation by SEVs, with treated rabbit corneas demonstrating ZO-1 signal loss at 6 hours posttreatment (fig. S9A), followed by obviously signal recovery at 24 hour after three washes with phosphate-buffered saline (PBS; fig. S9B). In contrast, ProLs induced irreversible TJ signal loss. The transepithelial electrical resistance (TEER) results were consistent with the previous findings: SEVs caused a 45% reversible TEER reduction (versus ProLs’ 27% irreversible loss) within 1 hour of exposure, with fully recovered to baseline levels within 23 hours post–PBS replacement (Fig. 4H). These data demonstrate SEVs’ unique capacity for reversible, nondestructive TJ remodeling, allowing for temporal control of paracellular permeability without compromising long-term epithelial integrity. This represents a critical safety advantage over ProLs, which caused TJ protein degradation.

Ocular trafficking of SEVs​

To confirm structural stability during ocular transit, we performed dual-label tracking using DiO-tagged SEVs-CDs. In both mouse and rabbit ocular models, dual fluorescence colocalization analysis at 6 hours post–topical administration (time-to-peak posterior segment accumulation) revealed high structural integrity of SEVs-CDs (Pearson’s correlation coefficient > 0.7) (figs. S10 and S11). By contrast, significantly reduced colocalization was observed at 9 hours, indicative of progressive vesicle disintegration and subsequent CD dispersion. These temporal dynamics demonstrate intact SEVs-CDs architecture during the critical therapeutic window preceding maximal retinal-choroidal drug deposition, enabling efficient posterior ocular delivery. This conclusion was further corroborated by bio-TEM imaging at 6 hours postadministration, which visualized structurally preserved FA-SEVs@CMG complexes within posterior ocular tissues (fig. S12).

Dual-pathway ocular delivery of SEVs​

Topical eye drop delivery to the posterior ocular segment primarily involves two pathways: the corneal route (cornea–aqueous humor–vitreous–retina) and the conjunctival route (conjunctiva-sclera-choroid-retina), both guarded by epithelial TJ barriers (31, 32). Compared to the conjunctival pathway, the corneal route requires traversing a longer anatomical distance to reach the retina. Consequently, effective posterior ocular delivery of eye drops might predominantly relies on the conjunctival route (5, 6). In our study, fluorescence tissue sections of rabbit and mouse eyes consistently revealed significantly higher fluorescence intensity in the conjunctiva and sclera compared to the cornea. Notably, ex vivo permeability assays through corneal versus conjunctival-scleral-choroidal-retinal demonstrated that during the 4-hour period of sustained posterior ocular drug concentration elevation, the cumulative penetration in the conjunctiva, sclera, choroid, and retina reached ~1.62-fold higher levels than that in the cornea (fig. S13). This provides direct experimental validation that the conjunctival-scleral-choroidal-retinal pathway serves as the predominant route for SEV-mediated intraocular delivery. However, rabbit ocular pharmacokinetic profiling also revealed corneal pathway participation in SEV permeation, with anterior segment deposition via this route detectable within 1 hour postadministration (Fig. 4B). This dual-route synergy accounts for the progressive central vitreous accumulation observed within 12 hours (Fig. 4D). Therefore, by concurrently exploiting distinct ocular routes, SEVs establish an “omnidirectional attack” strategy against fundus diseases, effectively overcoming anterior-posterior transport limitations. This strategy demonstrates potential for the noninvasive management of posterior segment diseases.

Proteomic analysis of SEVs​

To decipher the molecular basis of SEVs’ ocular penetration, we performed quantitative proteomic profiling across five different porcine-derived SEV groups (A: Hampshire, B: Berkshire, C: Yorkshire, D: Landrace, E: Duroc) (Fig. 5A and fig. S14) (15–17). Protein profiles of each group are provided in the data S1. While all groups expressed exosomal markers (CD9, CD63, and Alix) (Fig. 5B), ocular permeability assays revealed notable functional divergence: Group E showed obviously higher ocular permeability than group B in both mice and rabbits, with intermediate permeability in groups A, C, and D (Fig. 5, C and D).

Fig. 5. Proteomic analysis of SEVs and screening of their expressed proteins mediating ocular barrier penetration.
(A) Sample correlation heatmap. The bottom triangle (lift of the diagonal) uses red for positive correlations and blue for negative correlations, while the top triangle (right of the diagonal) displays numerical correlation coefficients. This analysis evaluates intersample correlations to reflect grouping patterns and reproducibility. (B) CD9, CD63, and Alix expression levels in groups A to E. Penetration of SEVs from groups A to E into mice (C) and rabbits (D) eyes 6 hours postadministration. Volcano plots of differentially expressed proteins for E versus B (E) and D versus C (F). The x axis represents log2 [fold change (FC)], with distance from zero indicating magnitude of change (right: up-regulated, left: down-regulated). The y axis represents –log10 (P value), with increasing values indicating higher significance. Blue and red dots represent up-regulated and down-regulated proteins, respectively, with deeper color indicating greater significance. Gray dots represent proteins with P ≥ 0.05. Differentially expressed proteins were identified using a threshold of |log2FC| > 1 and a false discovery rate < 0.05. (G) Proteomic screening and quantitative analysis of representative proteins in SEVs that facilitate sperm penetration of the reproductive barrier, such as PTGDS and ACRV1, along with proteins expressed in SEVs that are associated with regulatory pathways of tight junctions in the ocular barrier. (H) Heatmap visualization of correlation analysis between protein expression profiles in (G) and intraocular permeability metrics across experimental cohorts (groups A to E). The x axis represents the group or sample name. The y axis represents differentially expressed features (metabolites or proteins). The color gradient from blue to red indicates increasing abundance of the feature (metabolite or protein).

Screening of proteins associated with SEV-mediated ocular barrier penetration​

Differential expression analysis via volcano plots and clustering heatmap identified group-specific differentially expressed proteins (Fig. 5, E and F, and figs. S15 and S16). We specifically examined the expression of proteins associated with reproductive barrier penetration, such as PTGDS and ACRV1 (15–17), as well as those related to key receptors and pathways involved in ocular TJ regulation, including epidermal growth factor receptor (EGFR), Protein Kinase A (PKA) pathway, and Rho-associated protein kinase pathway (Fig. 5G) (33–36). Correlation analysis between intergroup expression levels and ocular permeability revealed a strong positive association with EGF (ligand of EGFR), which exhibited a high correlation to permeability while maintaining elevated expression across all groups (Fig. 5H). In contrast, the expression levels of PTGDS and ACRV1, which are proteins associated with reproductive barrier penetration, showed a weaker correlation with the intraocular penetration ability of SEVs. This may be due to the fundamentally distinct biological functions and compositions of the ocular and reproductive barriers. These findings prompted us to further investigate the role of exosomal EGF protein in modulating ocular barrier permeability.

Functional role of EGF protein in ocular barrier penetration​

Through incubation of EGF monoclonal antibodies with group E exosomes, Western blot analysis confirmed near-complete inhibition of EGF in SEVs (Fig. 6A). Subsequent evaluation of intraocular permeability revealed a marked reduction in ocular penetration rates in both mouse and rabbit eyes from the EGF-neutralized group compared to the noninhibited controls (Fig. 6, B and C). In cellular models, the EGF-inhibited group exhibited only an 8% decrease in TEER, significantly lower than the 45% reduction observed in the noninhibited group (Fig. 6D). These findings conclusively demonstrate that EGF serves as the key protein mediating SEV-enhanced ocular permeability. Furthermore, given the critical relationship between cytoskeletal structure and the phosphorylation level of myosin light chain 2 (pMLC), which directly regulates TJ redistribution (37, 38). We further validated the effects of human-derived EGF protein (structurally similar to porcine EGF) and SEVs on MLC phosphorylation and EGFR activation in corneal monolayer cell models. These experiments aimed to elucidate the role and mechanism of EGF in seminal exosome-mediated enhancement of intraocular permeability while exploring the potential of replacing porcine-derived EGF with human homologs.

Fig. 6. The mechanism by which EGF affects the ocular penetration of SEVs.
(A) Western blot validation of SEV-associated EGF inhibition by anti-EGF antibody. Penetration efficacy of SEVs in mouse (B) and rabbit (C) eyes pre– and post–EGF antibody inhibition. (D) Effect of SEVs on TEER in HECE monolayers pre– and post–EGF antibody inhibition. (E) Western blot analysis of EGFR–Src–MLC kinase (MLCK)–MLC pathway activation in HCEC monolayers treated with EGF, SEVs, or anti-EGF SEVs. TEER modulation in HCEC monolayers cotreated with EGF (F)/SEVs (G), and EGFR/Src/MLCK inhibitors (gefitinib, dasatinib, and ML-7). Induction of MLC phosphorylation in HCEC monolayers cotreated with EGF (H)/SEVs (I), and EGFR/Src/MLCK inhibitors (gefitinib, dasatinib, and ML-7). The values in (E), (H), and (I) represent the ratios of the target protein band intensities to those of the internal reference. Data were represented as means ± SD. P values in (B) and (C) were calculated by a paired t test (****P < 0.0001). P values in (D), (F), and (G) were calculated by one-way ANOVA with a Tukey post hoc test (*P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001) (n = 3).

Molecular mechanisms of SEVs and EGF ocular barrier penetration​

EGFR activation (phosphorylation) could initiate a phosphorylation cascade: This primary event may trigger Src protein phosphorylation (39, 40), subsequently inducing MLC kinase (MLCK) expression (41, 42) and ultimately leading to MLC phosphorylation (43). Using Western blot analyses, we confirmed that both human EGF protein and seminal exosomes induce phosphorylation of EGFR, Src, and MLC in corneal epithelial cell monolayers, along with elevated expression of MLCK. By comparison, the EGF-neutralized exosome groups exhibited markedly reduced phosphorylation levels across all four proteins (Fig. 6E). To further delineate the signaling cascade, we pretreated corneal cell monolayers with specific inhibitors: gefitinib (EGFR inhibitor) (44), dasatinib (Src inhibitor) (45), and ML-7 (MLCK inhibitor) (46). Subsequent treatment with human EGF or SEVs resulted in significantly attenuated effects on TEER modulation (Fig. 6, F and G) and diminished MLC phosphorylation levels (Fig. 6, H and I) compared to noninhibited controls. This systematic inhibition confirmed that seminal exosomes regulate TJ opening through EGF-dependent activation of the EGFR-Src-MLCK-MLC pathway, thereby enhancing ocular barrier permeability.

Potential risk assessment of SEVs and EGF ocular barrier penetration​

Phospho-Src (pSrc) analysis (fig. S17) in corneal epithelial monolayers following 12-hour incubation with blank medium posttreatment withdrawal demonstrated near-complete restoration to baseline levels, confirming transient and reversible EGFR pathway activation mediated by SEVs and EGF. Furthermore, in vitro angiogenesis assays in human umbilical cord endothelial cells (HUVECs) showed no statistically significant induction of tube formation (P > 0.05 versus controls) (fig. S1, thereby eliminating concerns regarding pathological neovascularization linked to fundus pathologies.

Validation of the role of EGFR in ocular barrier penetration by SEVs and EGF via EGFR knockout in HCECs​

To further confirm the critical role of the EGFR in the regulation of TJs in the ocular barrier by SEVs and EGF, we examined the effect of SEVs and EGF on the activation of the EGFR-Src-MLCK-MLC pathway in EGFR-knockout HCECs. Effective knockout of EGFR in corneal epithelial cells was verified by Western blot and quantitative polymerase chain reaction (PCR; fig. S19, A and B). As shown in the fig. S20, after EGFR knockout, neither SEVs nor EGF could effectively activate the phosphorylation levels of Src, MLCK, or MLC. This indicates that SEVs and EGF lose their ability to regulate TJs following EGFR knockout, further confirming that the EGFR-Src-MLCK-MLC pathway, initiated by EGFR activation, is responsible for the regulation of TJs in HCECs by SEVs and EGF protein.
In addition, we evaluated the effect of SEVs and EGF on TEER after EGFR knockout. We found that TEER in the EGFR-knockout group was significantly reduced compared to the nonknockout group, although it remained at a certain baseline level (fig. S21). After adding SEVs or EGF, TEER showed only minimal changes over 24 hours (fig. S22). Together, these results demonstrate that the reversible opening of TJs in ocular barrier cells is mediated through the direct interaction of SEVs with the EGFR and the subsequent reversible activation of the EGFR-Src-MLCK-MLC pathway, a process mediated by the EGF protein expressed in SEVs.

POD-like activity of CDs​

With establishment of efficient SEVs’ posterior delivery, we interrogated the therapeutic potential of their CD payloads through POD-like activity characterization. Using 3,3′,5,5′-tetramethylbenzidine (TMB) as the chromogenic substrate, CDs exhibited concentration-dependent hydroxyl radical (•OH) generation, influenced by both H2O2levels and CD dosage (fig. S23, A and B). Maximum catalytic efficiency was observed under physiological temperature (~37°C) and acidic pH (4.0 to 6.0), which mimics the pH conditions of the tumor TME (fig. S23, C and D). Michaelis-Menten kinetics revealed exceptional catalytic parameters: Vmax = 46.96 × 10−8 M/s (H2O2) and 26.62 × 10−8 M/s (TMB) with Km values of 161.57 mM and 49.04 μM, respectively (fig. S23 E and F), which substantially surpass those of conventional iron oxide nanozymes and natural horseradish peroxidase (table S1), positioning CDs as superior ROS generators under TME (47).
Iron-doped CDs have demonstrated excellent enzymatic activity and might attributed to iron-centered active structural motifs (19). To delineate the predominant contribution of Fe to the superior enzymatic activity of CDs, we synthesized iron-free CDs (N-CDs) using citric acid, thiourea, and boric acid as precursors under identical synthetic conditions to those used for CDs. HRTEM characterization revealed that the N-CDs have an average particle size of ~3.03 nm, exhibiting a lattice spacing of 0.21 nm (fig. S24A). Elemental analysis confirmed successful N, S, and B doping in the N-CD structure (fig. S24B).
Electron spin resonance experiments (fig. S24C) confirmed that the enzymatic activity of the N-CDs is weaker than that of CDs. Kinetic analysis of the enzymatic reactions for N-CDs (fig. S24, D and E) revealed catalytic parameters of Vmax = 25.30 × 10−8 M/s (H2O2) and 16.60 10−8 M/s (TMB), with corresponding Km values of 138.28 mM and 42.57 μM, respectively. These values are significantly lower than those observed for CDs [Vmax = 46.96 × 10−8 M/s (H2O2) and 26.62 × 10−8 M/s (TMB) with Km values of 161.57 mM and 49.04 μM, respectively], further substantiating that iron doping is critical for the enhanced enzymatic performance of CDs. Therefore, we further unveiled the iron-centered enzymatic active site architecture of CD nanozymes through systematic structural characterization.

POD activity structural centers of the CDs​

X-ray photoelectron spectroscopy (XPS) revealed critical surface functional groups (fig. S25): In the N 1s spectrum, pyridinic N (397.88 eV), Fe─N (399.28 eV), pyrrolic N (399.98 eV), and graphitic N (401.18 eV) were observed. Both pyrrolic and pyridinic nitrogen atoms might coordinate with Fe to form active sites designated as Fe─Nx (48). The S 2p spectrum revealed Fe─S (161.88 eV), -C─S─C- (163.48 and 165.88 eV), and C─SOx (167.08, 168.68, and 169.88 eV). In the B 1s spectrum, B─B (186.28 eV), B─N (189.68 eV), B─O (191.48 eV), and B─C (192.88 eV) were existed. In the Fe 2p spectrum, Fe2+ 2p3/2 (710.08 eV), Fe2+ 2p1/2 (723.68 eV), Fe─Nx (713.48 and 728.68 eV), and Fe─S bonds (717.98 eV) were detected.
On the basis of XPS analyses confirming the existence of Fe─S and Fe─N bonds, we selected FeS2 and iron phthalocyanine (FePc) as reference compounds to further resolve the fine structural coordination of Fe with S/N, supplemented by classical references including Fe foil, FeO, and Fe2O3 for comprehensive spectral validation. Fe K-edge x-ray absorption near-edge structure (XANES) spectra positioned the absorption edge of CDs between FeS2 and FePc reference standards (Fig. 7A). Fourier-transformed k2-weighted extended x-ray absorption fine structure (EXAFS) analysis identified Fe─N (1.5 Å) and Fe─S (1.84 Å) coordination paths (Fig. 7B), providing conclusive evidence for the coexistence of Fe─S and Fe─N bonds. Furthermore, wavelet transform (WT) analysis of Fe K-edge EXAFS oscillations in k-space and R-space (Fig. 7C) provides additional support for this conclusion. The scattering path signals of CDs are localized near (5 and 1.6 Å−1), a region encompassing characteristic scattering paths of FePc and FeS2. The absence of discernible Fe─Fe scattering signals in the WT contour maps confirms the monoatomic dispersion of Fe and the lack of substantial Fe─Fe interactions. The k2-weighted EXAFS oscillations (Fig. 7D) of CDs also exhibit amplitude and periodicity similar to those of FeS2 and FePc, further reinforcing the aforementioned conclusion. Last, quantitative EXAFS fitting analysis (Fig. 7, E and F, and table S2) also confirms the presence of an N-and-S–coordinated Fe single-atom moiety. The Fe coordination number within the CDs was determined to be approximately 4, with an N:S coordination ratio of approximately 3:1. The average Fe─N and Fe─S bond lengths were found to be ~1.97 and 2.22 Å, respectively.


Fig. 7. Atomic structural characterization of CDs.
(A) XANES spectra of Fe K-edge in CDs. (B) Fourier-transformed (FT) k2-weighted χ(k) function of the EXAFS spectra of the CDs, FeS2, Fe foil, FePc, Fe2O3, and FeO. (C) WT contour plots of the k3-weighted EXAFS data of CDs and reference samples (Fe foil, FeS2, FeO, Fe2O3, and FePc). (D) FT k2-weighted EXAFS spectra of CDs, Fe foil, FeS2, FeO, Fe2O3, and FePc. (E) The corresponding EXAFS fitting curves of CDs at R-space. Inset: A model for the surface of CDs. Iron: Fe (brown), nitrogen: N (blue), sulfur: S (yellow), and carbon: C (gray). (F) Fitting curves of the EXAFS of CDs in the k-space.

Collectively, these structural insights demonstrate that CDs exhibit exceptional POD-like activity through their atomically engineered Fe-N3S-B/C coordination centers. The enzymatic superiority originates from dual synergistic mechanisms: Fe-N sites drive H2O2 activation, while S/B heteroatoms stabilize reactive intermediates (18, 19, 24).

Preparation and characterization of FA-SEVs@CMG​

To overcome the TME-imposed limitations of GSH-mediated ROS scavenging and insufficient H2O2 bioavailability that constrain the therapeutic efficacy of CDs’ POD-like activity, we engineered a TME-responsive nanosystem (CMG) integrating catalytic CDs, GSH-depleting MnO2 nanozymes (fig. S26), and H2O2-generating Gox. This nanosystem was encapsulated into SEVs and surface functionalized with FA to construct FA-SEVs@CMG, achieving 41.3% FA conjugation efficiency (Fig. 8A, visual appearance). TEM imaging (Fig. 8B) confirms effective CMG encapsulation within SEVs, preservation of intact membrane morphology, and homogeneous distribution of FA-SEVs@CMGs. DLS and zeta potential analyses (Fig. 8, C and D) revealed the stepwise assembly of FA-SEVs@CMG through controlled size evolution and surface charge modulation, ultimately exhibiting a hydrodynamic diameter of 142.2 ± 25.5 nm with a PDI of 0.241 ± 0.083. The observed particle size in TEM was marginally smaller than DLS measurements, consistent with inherent discrepancies between hydrated hydrodynamic diameters (DLS) and dehydrated morphological dimensions (TEM). Electrophoretic consistency across SEVs, SEVs@CMG, and FA-SEVs@CMG confirmed preserved membrane integrity (Fig. 8E). Stability studies demonstrated robust storage performance, with <10% variation in size and zeta potential after 14 days (fig. S27). FA-SEVs@CMG exhibited permeability comparable to that of unmodified SEVs across ex vivo rabbit corneal and scleral/choroidal/retinal tissues (fig. S2, indicating that the FA-SEVs@CMG construct does not compromise the ocular penetration capacity intrinsic to SEVs. These results validate the successful construction of FA-SEVs@CMG, merging synthetic nanozyme activity with SEV functionality for targeted tumor therapy.

Fig. 8. The preparation and intracellular anticancer therapy of FA-SEVs@CMG.
(A) The appearance of FA-SEVs@CMG under ambient light (left) and 350-nm laser irradiation (right). (B) The TEM images of MnO2, CMG, and FA-SEVs@CMG. (C) DLS size and (D) zeta potential of different materials. (E) SDS–polyacrylamide gel electrophoresis protein analysis of SEVs, SEVs@CMG, and FA-SEVs@CMG. (F) Confocal fluorescence imaging of free CDs, CMG, SEVs@CMG, and FA-SEVs@CMG in Y79 cells for 1 and 4 hours. Quantitative fluorescence results of CDs in each group are shown in the right figure. The nuclei were stained with DAPI (blue), while lysosomes were stained with LysoTracker Green. (G) Effects of temperature and endocytosis inhibitors on cellular uptake of different groups in Y79 cells. The noninhibited internalization of different group was used as positive control. (H) Confocal fluorescence imaging of ROS in Y79 cells by 2′,7′-Dichlorodihydrofluorescein diacetate (DCFH-DA). Quantitative results of ROS fluorescence intensity in each group are shown in the right figure. Data were represented as means ± SD. P values in (F) were calculated by a paired t test (***P < 0.001 and ****P < 0.0001). P values in (G) and (H) were calculated by one-way ANOVA with a Tukey post hoc test (****P < 0.0001) (n = 3).

In vitro cellular studies​

We next assessed the tumor-targeting efficacy and therapeutic potential of FA-SEVs@CMG in RB Y79 cells. CMG exhibited superior growth inhibition compared to individual components (CDs, MnO2, or Gox) (fig. S29A). While plain SEVs showed negligible cytotoxicity, encapsulation significantly enhanced therapeutic efficacy: FA-SEVs@CMG achieved an 82.7% cell viability inhibition against Y79 cells, outperforming SEVs@CMG (77.4%) and CMG alone (66.6%) at concentration of 50 μg/ml. All formulations preserved >85% viability in HECE, human lens epithelial cells (HLE-B3), and ARPE normal ocular cells (fig. S29, B to D) at the same concentration, confirming therapeutic selectivity.

Cellular uptake studies revealed dual enhancement mechanisms: FA-SEVs@CMG showed obviously higher Y79 internalization than SEVs@CMG at 4 hours (Fig. 8F and fig. S30), attributable to folate receptor–mediated active targeting. SEV encapsulation itself boosted CMG uptake versus nonencapsulated CMG, with time-dependent accumulation across all groups. Lysosomal colocalization analysis demonstrated almost entirely overlap between CD fluorescence (red) and lysotracker signals (green), confirming endolysosomal trafficking into acidic compartments—a microenvironment that potentiates CDs’ POD-like activity. This advantage was further highlighted by the verification that FA-SEVs@CMG also exhibited the highest enzymatic activity under mildly acidic pH conditions (4.0 to 6.0) (fig. S31), consistent with the acidic microenvironment of lysosomes.
We postulated that SEV encapsulation might enhance cellular uptake by modulating endocytic pathways in Y79 cells. To test this hypothesis, we systematically evaluated the internalization of different groups under distinct uptake inhibitors and temperature conditions (49) (Fig. 8G and fig. S32). All groups exhibited marked reductions in cellular uptake under low-temperature (energy inhibitor) incubation or when treated with chlorpromazine (clathrin inhibitor) or sodium azide (energy inhibitor), confirming clathrin-mediated endocytosis as the predominant entry route. Notably, SEVs@CMG and FA-SEVs@CMG displayed disproportionately attenuated uptake compared to non-SEVs’ encapsulated counterparts upon amiloride (macropinocytosis inhibitor) treatment, while no significant differences were observed across groups treated with other endocytic inhibitors (e.g., methyl-β-cyclodextrin for lipid raft-mediated pathways or filipin for caveolae-dependent pathways). This differential sensitivity demonstrates that SEV encapsulation reprograms the internalization mechanism in Y79 cells, transitioning from exclusive clathrin-mediated endocytosis to a hybrid pathway incorporating both clathrin-dependent trafficking and macropinocytosis. These mechanistic insights highlight how SEVs can strategically redirect cellular entry routes to optimize therapeutic delivery.

Mechanistic investigation of FA-SEVs@CMG response in the TME in vitro​

Subsequently, to evaluate the TME-responsive behavior of FA-SEVs@CMG, we first validated its catalytic activity in vitro. CDs and MnO2 components displayed catalytic H2O2 scavenging capacity (fig. S33A), while MnO2-containing groups efficiently generated O2 and depleted GSH without intergroup variation (fig. S33, B and C). Gox exhibited enhanced glucose consumption (fig. S33D), with CMG, SEVs@CMG, and FA-SEVs@CMG showing amplified metabolic activity. This amplification arose from a self-sustaining cycle: Gox-mediated glucose-to-H2O2 conversion fueled MnO2/CD catalysis, which, in turn, accelerated glucose depletion through ROS cycling. Furthermore, the CMG, SEVs@CMG, and FA-SEVs@CMG groups demonstrated lower H2O2 generation levels compared to the Gox-only group but significantly higher than blank controls (fig. S33E). This observation indicated that the FA-SEVs@CMG system produced a greater quantity of H2O2 than it consumes, further supported by the decreased total oxygen concentration involved in both H2O2 generation and consumption within the system (fig. S33F). Notably, the cellular assays further revealed significant TME remodeling in Y79 cells (fig. S34, A to C). SEV encapsulation and FA targeting synergistically enhanced intracellular accumulation of therapeutic components, driving profound GSH/glucose depletion and dynamic H2O2 regulation where catalytic consumption by MnO2/CDs outweighed the production by Gox.

We further investigated the system’s •OH production kinetics under TME-mimetic conditions (fig. S35). While CDs alone exhibited limited POD-like activity to generate •OH from glucose/H2O2, increasing glucose concentrations markedly amplified radical yields. Integration of MnO2 enhanced •OH production. •OH levels displayed biphasic dependence on GSH concentrations. At low-to-moderate GSH levels, MnO2-mediated GSH oxidation liberated Mn2+, Mn2+-Fenton catalysis, and POD-like activity cooperatively amplified •OH accumulation. However, excessive GSH competitively consumed H2O2, reducing the substrate available for enzymatic and Fenton reactions and thus decreasing •OH accumulation. This dynamic interplay highlights how CDs’ nanozyme activity and Mn2+-mediated redox catalysis collaboratively exploit TME components to maximize therapeutic oxidative stress.

In conclusion, as illustrated in fig. S36, the CMG framework operates through coordinated TME remodeling: MnO2 depletes GSH/H2O2 to yield O2 and Mn2+, while O2 sustains Gox activity to amplify H2O2 flux. This peroxide pool fuels both CD-mediated POD-like catalysis and Mn2+-driven Fenton chemistry, generating cytotoxic •OH bursts. FA-SEVs@CMG amplifies this cascade through dual engineering: SEVs enhance cellular internalization, while FA confers tumor-selective targeting. This strategy effectively enhances the local POD activity of CDs, creating a TME-responsive, self-amplifying therapeutic loop while sparing healthy tissues. The system’s ability to hijack tumor metabolic pathways for selective ROS amplification highlights its potential as a precise nanotherapeutic agent for cancer.

Anticancer mechanism investigation of the FA-SEVs@CMG in vitro​

The ROS-induced ferroptosis and mitochondrial membrane potential loss signaling axis​

To further investigate the anticancer mechanisms of FA-SEVs@CMG at the cellular level, we systematically dissected oxidative stress cascades and their downstream consequences in Y79 cells. ROS-specific probes targeting •OH revealed a progressive enhancement of oxidative stress across sequential formulations (CDs < CM < CMG < SEVs@CMG < FA-SEVs@CMG), with FA-SEVs@CMG inducing stronger intracellular ROS levels (Fig. 8H and fig. S37). This oxidative surge drove lipid peroxidation (LPO) accumulation (fig. S3, a hallmark of ferroptosis, which was further corroborated by glutathione peroxidase 4 (GPX4) down-regulation in Western blot analysis (Fig. 9A). As ferroptosis may induce mitochondrial damage, we further assessed mitochondrial membrane potential (MMP) in tumor cells after treatment in each group. Notably, FA-SEVs@CMG–treated cells exhibited catastrophic mitochondrial depolarization (fig. S39), establishing ROS-mediated MMP loss as a pivotal event in ferroptosis execution.

Fig. 9. Anticancer mechanism investigation of FA-SEVs@CMG.
(A) Western blot results of LC3I/II, GPX4, and Caspase-9 proteins in Y79 cells after treatment by different groups. (B) Caspase-3 and (C) Caspase-7 protein activity in Y79 cells with and without treated by 3-methyladenine (3-MA). (D) Flow cytometry of annexin V/propidium iodide (PI)–stained Y79 cells treated by different groups treated. (E) Flow cytometry of annexin V/PI–stained Y79 cells treated by different groups treated with 3-MA. (F) Quantitative results of annexin V/PI–stained Y79 cells by flow cytometry treated in different groups without or with 3-MA. Western blot results of Caspase-9, Bcl-2, Caspase-9, and BID proteins in Y79 cells treated by different groups (G) without or (H) with 3-MA. Data were represented as means ± SD. P values in (B), (C), and (F) were calculated by a paired t test (*P < 0.05 and ****P < 0.0001) (n = 3)

MMP loss triggers concurrent up-regulation of autophagy and apoptosis​

Furthermore, the loss of MMP may trigger caspase-9 activation, subsequently activating caspases-3/7, a key family of proteins involved in the intrinsic (mitochondrial) apoptotic pathway. Results in Fig. 9 (A to C) demonstrated the progressively enhanced and sequential activation of caspase-9 and caspases-3/7 across treatment groups, linking oxidative stress to intrinsic apoptosis. Flow cytometric analysis demonstrated a formulation-dependent apoptotic progression, with apoptotic rates ascending sequentially across treatment groups: 13.8% in CD-treated cells, 25.4% in CM group, 67.2% in CMG group, 75.3% in SEVs@CMG group, and 85.6% in FA-SEVs@CMG group (Fig. 9D). This apoptotic hierarchy directly correlated with the severity of MMP collapse observed in fig. S39, establishing mitochondrial dysfunction as the central regulator of cell fate. We hypothesize that ROS-mediated MMP collapse activated the caspase-9 to caspase-3/7 axis, which is a signature of intrinsic apoptosis. SEVs enhanced this cascade through improved cargo delivery, while FA targeting amplified spatial precision, culminating in near-complete apoptotic execution in FA-SEVs@CMG–treated cells.
Beyond apoptotic activation, MMP collapse triggered mitochondria autophagic response. Confocal laser scanning microscopy (CLSM) imaging revealed autophagosome-lysosome coalescence around mitochondria (fig. S40), while monodansylcadaverine (MDC) fluorescence intensity (figs. S41 and S42) confirmed autophagic vacuole accumulation proportional to MMP loss. Crucially, LC3-II/LC3-I ratio elevation (Fig. 9A), peaking in FA-SEVs@CMG groups, demonstrated unabated autophagic flux, a hallmark of mitophagy attempting to clear ROS-damaged organelles. Paradoxically, this self-repair process exceeded reparative thresholds, transitioning from cytoprotection to cytotoxicity.

Stress-intensity–dependent autophagy-apoptosis cross-talk​

Building on the biphasic nature of autophagy in cancer (50, 51), we dissected its context-dependent duality using 3-methyladenine (3-MA) (autophagy inhibitor)–mediated inhibition (52). In low-stress microenvironments (CD/CM groups), autophagy blockade increased apoptosis (Fig. 9, E and F), revealing its survival-promoting role. Conversely, under conditions of high oxidative stress (CMG, SEVs@CMG, and FA-SEVs@CMG), 3-MA suppressed apoptosis and reduced caspase-3/7 activity (Fig. 9, B and C), indicating that autophagy potentiates apoptosis in a stress-threshold–dependent manner. Our findings reveal a redox-regulated duality in autophagy that pivots on cellular stress intensity. Below the critical ROS threshold observed in CD and CM groups, autophagy functions as a homeostatic protector to suppress apoptosis. However, surpassing the ROS tipping point (CMG, SEVs@CMG, and FA-SEVs@CMG), autophagy undergoes a lethal metamorphosis, transitioning into a death amplifier that directly promotes apoptosis.
This paradigm was rigorously validated through ROS scavenging with N-acetylcysteine (NAC) (ROS inhibitor) (53), which abolished both autophagic flux (fig. S42) and apoptotic signals (figs. S43 and S44), further confirming ROS as the master regulator of this binary switch. Ultrastructural evidence from bio-TEM imaging captured the consequences exceeding the threshold: FA-SEVs@CMG–treated cells exhibited rampant autophagosome proliferation (fig. S45) alongside nuclear pyknosis and fragmentation, hallmarks of irreversible cell death. CLSM imaging further demonstrated coordinated escalation of autophagy and apoptosis in CMG, SEVs@CMG, and FA-SEVs@CMG groups (fig. S46), contrasting sharply with the muted responses in CD and CM groups.

Regulation of autophagy and apoptosis interactions by FA-SEVs@CMG​

Building on the established ROS-autophagy-apoptosis axis, we dissected the molecular circuitry through which FA-SEVs@CMG converts autophagic flux into apoptotic lethality. Autophagy-mediated apoptosis might proceed via three nonmutually exclusive mechanisms (51): (i) selective degradation of apoptosis inhibitors [e.g., B-cell lymphoma 2 (Bcl-2), a typical antiapoptotic protein that promotes mitochondrial outer membrane permeabilization (50, 54, 55)], (ii) activation of intrinsic/extrinsic caspase cascades (caspase-9/-, and (iii) lysosomal failure–driven mitochondrial permeabilization. Given sustained lysosomal activity in treatment groups (fig. S40), we focused on the first two pathways. Western blotting revealed autophagy-dependent apoptotic priming: Progressive caspase-8/-9 activation paralleled Bcl-2 suppression across CMG, SEVs@CMG, and FA-SEVs@CMG groups (Fig. 9G). Crucially, 3-MA–mediated autophagy inhibition abolished caspase-8 induction (Fig. 9H) while leaving caspase-9/Bcl-2 profiles intact, pinpointing autophagy specifically to extrinsic apoptosis pathway activation.

The progressive decline in Bcl-2 expression across treatment groups mirrors the escalating mitochondrial stress, consistent with the concomitant increase in caspase-9 expression. This phenomenon synergizes with autophagy-mediated activation of the extrinsic apoptotic pathway, as evidenced by BH3-interacting domain death agonist (BID)’s pivotal role as a molecular hub connecting oxidative stress to caspase-8 execution. Autophagy inhibition via 3-MA markedly attenuated BID expression in CMG, SEVs@CMG, and FA-SEVs@CMG groups (Fig. 9, G and H), establishing a causal link between autophagic flux and BID–caspase-8 axis activation.

FA-SEVs@CMG anticancer mechanism summary​

In summary, FA-SEVs@CMG orchestrates RB cell self-destruction via a trimodal mechanism driven by TME-amplified oxidative stress (fig. S47). The CMG system, activated by TME, generates localized ROS surge that simultaneously ignites ferroptosis (via LPO accumulation), while further triggering MMP collapse to activate the intrinsic apoptosis (through caspase-9 activation), and autophagic overload. Crucially, the synergistic interplay of SEV-enhanced cellular internalization and FA-mediated targeting drives an oxidative surge that catastrophically overwhelms cytoprotective autophagy thresholds.

This redox escalation forcibly repurposes autophagic flux into a pro–death signaling hub, hyperactivating the extrinsic apoptotic cascade through BID-dependent caspase-8 proteolytic cleavage, effectively converting cellular self-preservation machinery into a driver of cell death. This creates a self-propagating death cycle where ROS begets autophagy, autophagy fuels apoptosis, and apoptotic mitochondrial damage further escalates ROS, a synthetic lethal triad that dismantles cellular homeostasis. On the basis of aforementioned anticancer mechanism and the ability of SEVs to efficiently deliver to the posterior segment via dual intraocular routes, FA-SEVs@CMG demonstrates substantial potential as a noninvasive therapeutic paradigm for RB.

Pharmacodynamic studies​

To evaluate the in vivo antitumor efficacy of FA-SEVs@CMG, we established an orthotopic xenograft mouse model via retinal injection of Fluc/green fluorescent protein (GFP)–Rb-1 cells (n = 8; Fig. 10A). Successful model validation was confirmed by grossly visible posterior segment opacity in tumor-bearing eyes (fig. S4. Bioluminescence imaging revealed distinct therapeutic contrast between experimental groups. Mice treated with FA-SEVs@CMG or SEVs@CMG eye drops exhibited significantly reduced intraocular signals (Fig. 10, B and Ba), particularly in the FA-SEVs@CMG group, indicating potent tumor growth inhibition. In contrast, CD and CMG groups showed elevated bioluminescence due to poor ocular barrier penetration without SEV encapsulation. Longitudinal analysis at day 30 demonstrated severe neovascularization and intraocular opacity in untreated mice, while FA-SEVs@CMG– and SEVs@CMG–treated eyes maintained near-physiological clarity (Fig. 10C). After 30 days of treatment, histopathological analysis of hematoxylin and eosin (H&E)–stained sections (Fig. 10D) from ocular tumor tissues further confirmed the aforementioned conclusions. Quantitative assessment of tumor areas (Fig. 10, Da) revealed significant differences in tumor suppression rates across experimental groups: The FA-SEVs@CMG group demonstrated efficacy with only 2.35 ± 0.38% residual tumor mass relative to the control group, followed by SEVs@CMG (10.78 ± 2.94%). In contrast, the CMG and free CD groups maintained substantially higher tumor burdens at 74.76 ± 3.96 and 93.16 ± 2.52% of control levels, respectively.

Fig. 10. In vivo pharmacodynamic study of FA-SEVs@CMG in RB mice.
(A) Schematic illustration of establishing RB and the design of animal experiments. (B) Representative in vivo bioluminescence images and average bioluminescence signal quantitative results (Ba) of mice with Y79 RB after different treatments within 30 days. d, days. (C) Representative images and (D) H&E staining slices of mouse eyes with RB after 30 days of treatment by different groups. (Da) Quantitative results of tumor area in (D). (E) Representative electroretinography (ERG) wave responses of RB mice after different treatments under scotopic conditions. The dashed line indicates the a- and b-wave values of the positive control group. (F) Representative immunofluorescence results of P62, [terminal deoxynucleotidyl transferase–mediated deoxyuridine triphosphate nick end labeling (TUNEL)], and Ki-67 in RB tissues after different treatments. Data were represented as means ± SD. P values in (Ba) and (Da) were calculated by one-way ANOVA with a Tukey post hoc test (n = 3) (*P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001).

To further elucidate the therapeutic mechanisms, we performed intratumoral drug accumulation (expressed as a percentage of the total administered dose) within 24 hours post–initial administration following successful tumor model establishment (fig. S49). At 6 hours postadministration, the FA-SEVs@CMG group demonstrated a higher tumor retention rate (2.39 ± 0.07% of administered dose), surpassing the SEVs@CMG group (1.30 ± 0.09%) and markedly exceeding both the CMG group (0.26 ± 0.03%) and free CDs group (0.18 ± 0.02%). Subsequently, intratumoral drug levels progressively declined, with negligible ocular accumulation detectable by 24 hours postadministration. These findings collectively demonstrate a positive correlation between drug penetration efficiency and ultimate tumor suppression efficacy. The surface engineering of SEVs combined with FA-mediated active targeting synergistically enhanced FA-SEVs@CMG’s tumor-specific accumulation, thereby establishing its optimal therapeutic superiority through both spatial biodistribution control and sustained pharmacodynamic effects. Notably, histopathology (Fig. 10D) confirmed spatial tumor confinement to the lens-retina interface in SEVs@CMG/FA-SEVs@CMG groups, contrasting with extraocular dissemination and ocular enlargement in CD and CMG groups. Considering the ocular delivery mechanism of SEVs, the encapsulation of SEVs enhances the delivery of SEVs@CMG and FA-SEVs@CMG via both corneal and conjunctival routes. This dual approach allows for an omnidirectional attack on the intraocular tumors, preventing their spread (6). FA functionalization further refines specificity by directing payloads to folate receptor–dense tumor microdomains, amplifying localized oxidative stress while sparing healthy tissues.
In addition, the SEVs@CMG and FA-SEVs@CMG groups demonstrated accelerated increase in body weight and a greater extension of survival time (fig. S50). Electroretinography (ERG) revealed near-complete ablation of retinal function in untreated tumor-bearing mice, characterized by suppressed dark-adapted a- and b-wave amplitudes (Fig. 10E), whereas FA-SEVs@CMG– and SEVs@CMG-treated mice retained retinal responses indistinguishable from healthy controls, confirming functional preservation. Immunofluorescence analysis (Fig. 10F) further demonstrated that SEVs@CMG and SEVs@CMG/FA significantly suppressed tumor proliferation (Ki-67) while elevating apoptosis [terminal deoxynucleotidyl transferase–mediated deoxyuridine triphosphate nick end labeling (TUNEL)] and autophagy (P62), with progressive enhancement across treatment groups consistent with cellular-level findings. Overall, FA-SEVs@CMG integrates the TME-responsive properties of CMG, the tumor-targeting capability of FA, and most importantly, the efficient dual corneal/conjunctival penetration and enhanced tumor tissue uptake facilitated by SEVs as a delivery vehicle. This combination makes it possible to use an omnidirectional attack strategy for treating RB, preserving retinal function, and preventing metastasis, thereby redefining noninvasive RB therapy with direct translational relevance for pediatric oncology.

Intraocular tumor fluorescence tracing​

Intraocular fluorescence tracing identified 2 hours postadministration as the optimal time point for assessing tumor targeting, with fluorescence intensity decline plateauing across groups (fig. S51).

SEVs@CMG and FA-SEVs@CMG exhibited comparable intraocular fluorescence at this time point, achieving obviously higher accumulation than CD/CMG groups—a direct consequence of SEVs’ ocular barrier penetration. Longitudinal monitoring over 30 days revealed persistently weak fluorescence in CD/CMG groups (fig. S52, A and B), whereas SEVs@CMG maintained stable fluorescence intensity independent of tumor progression. Notably, FA-SEVs@CMG demonstrated fluorescence-bioluminescence correlation (fig. S52C), attributable to its dual functionalization: (i) SEV-driven omnidirectional-ocular therapeutic distribution ensured comprehensive tumor coverage, and (ii) FA-mediated tumor targeting enabled fluorescence tracking of tumor volume. While SEVs@CMG penetrated posterior segments, its lack of targeting specificity precluded progression monitoring. These results establish FA-SEVs@CMG’s unique capacity for theranostic integration.

Safety evaluation​

We first conducted nucleic acid testing for African swine fever virus and porcine reproductive and respiratory syndrome virus in 15 randomly selected batches of porcine semen, and no pathogens were detected in any batch (table S3). Then, we investigated the intraocular metabolic profile of CDs, as their potential accumulation in the eye could raise safety concerns. On the basis of the metabolic behavior of SEVs-CDs in rabbit and mouse eyes, both species exhibited near-complete intraocular metabolism of CDs within 24 hours (Figs. 3Dand 4E). Similarly, FA-SEVs@CMG demonstrated nearly undetectable intratumoral CD accumulation in orthotopic ocular tumor models at 24 hours postadministration (fig. S49). To further evaluate potential long-term accumulation, we administered FA-SEVs@CMG eye drops to rabbit and mouse eyes consecutively for 14 days. Quantitative analysis of CDs in ocular tissues 24 hours after the final dose, combined with fluorescence imaging of ocular sections, revealed no significant accumulation of CDs even under prolonged administration (fig. S53). In addition, histopathological assessment of mouse eyes and major organs (e.g., liver and kidneys) 30 days posttreatment demonstrated no structural abnormalities in H&E-stained sections (fig. S54), indicating that FA-SEVs@CMG induces no detectable tissue damage. In vivo optical coherence tomography (OCT) imaging further demonstrated no detectable effects on posterior ocular structures (e.g., retina and choroid) following 30-day administration across all treatment groups (fig. S55). Furthermore, hematological parameters in healthy mice remained within physiological ranges (fig. S56). These results collectively confirm the safety of FA-SEVs@CMG in terms of systemic metabolism and biocompatibility.
Rabbit ocular safety studies demonstrated that long-term use of FA-SEVs@CMG induced negligible corneal irritation, with stable intraocular pressure (IOP) (figs. S57 and S5. Critically, SEV-based eye drops did not activate intraocular immune responses, as evidenced by undetectable levels of CD80 and CD4+markers (fig. S59). Proinflammatory cytokines [interleukin-1β (IL-1β), IL-6, tumor necrosis factor–α (TNF-α), and transforming growth factor–β (TGF-β)] in aqueous humor remained at baseline levels, further confirming the immunocompatibility of the formulation (fig. S60). Comprehensive biocompatibility assessments confirmed the excellent safety profile of FA-SEVs@CMG eye drops.

DISCUSSION​

Current therapeutic interventions for posterior segment diseases such as RB face dual challenges of collateral damage to ocular structures and insufficient efficacy in noninvasive delivery systems. While exosome-based carriers hold potential for safe drug delivery, their application in noninvasive posterior segment delivery remains underexplored. Inspired by the natural role of SEVs in facilitating sperm penetration through the female reproductive barrier, we investigated SEVs as a biomimetic system for noninvasive fundus delivery. Our study confirms that SEVs can reversibly open TJs of the ocular barrier, enabling dual corneal/conjunctival penetration without structural disruption. However, the expression in SEVs of proteins known to facilitate sperm penetration through the female reproductive barrier (e.g., PTGDS and ACRV1) is not the primary drivers of their ocular barrier penetration, likely due to the fundamental differences between the ocular and reproductive barriers. Further investigation reveals that the reversible opening of ocular barrier TJs by SEVs is primarily mediated by the expression of the EGF protein. This mechanism involves the direct activation of the EGFR on ocular barrier cells, leading to the reversible activation of the EGFR-Src-MLCK-MLC signaling axis. These findings provide key mechanistic insights for the broader application of SEVs in ocular drug delivery systems.

By engineering an eye drop formulation, FA-SEVs@CMG, we validated SEVs’ therapeutic potential for posterior segment diseases (RB). The CMG core integrates NIR fluorescent CDs with POD-mimetic Fe-N3S-B/C active sites, MnO2nanozymes, and Gox. Within the TME, MnO2 reacts with GSH and H2O2 to generate Mn2+ and O2. The liberated O2 fuels Gox-mediated glucose oxidation, amplifying H2O2 levels and enhancing POD activity of CDs. This cascade synergistically elevates ROS via Mn2+-driven Fenton reactions, inducing profound oxidative stress. Encapsulation of CMG within SEVs enhances tumor cellular uptake through clathrin-mediated endocytosis and macropinocytosis. FA functionalization further enhances tumor targeting, where the oxidative surge surpasses the cytoprotective autophagy threshold in tumor cells, shifting autophagy from a cell homeostasis protector to a driver of cell death. Subsequently, autophagy activates the extrinsic apoptotic pathway via BID–caspase-8 cascade, ultimately driving tumor self-destruction.

In orthotopic RB models, FA-SEVs@CMG leveraged SEVs’ dual corneal/scleral penetration routes to achieve omnidirectional-ocular therapeutic coverage, confining tumors within the fundus and preventing extraocular dissemination. This strategy preserved retinal function, as evidenced by intact ERG responses, while fluorescence imaging confirmed tumor progression monitoring via FA-SEVs@CMG’s targeting specificity.

In summary, we report a posterior segment delivery strategy using SEVs that addresses a critical gap in noninvasive and efficient posterior eye segment delivery of exosome-based therapeutics. Leveraging their dual-pathway posterior segment delivery mechanism and enhanced tumor cell uptake efficiency, SEV-based eye drops encapsulating a nanozyme system effectively induced RB cell self-destruction, substantially suppressing rapid tumor progression and extraocular extension while preserving vision in RB mice model. This groundbreaking study on SEVs marks a paradigm shift in posterior ocular disease therapeutics. Unlike existing delivery methods such as microparticles or microneedles, the SEV-based eye drop platform avoids potential ocular structural damage and systemic toxicity. Critically, whereas conventional microparticles or microneedles exhibit unidirectional diffusion within the eye, the dual-pathway delivery via both corneal and conjunctival routes with SEV eye drops offers distinct advantages for the noninvasive treatment of fundus diseases, such as ocular fundus tumors.

In addition, we used healthy Duroc boars aged 10 to 18 months to collect fresh porcine semen. Because of the limited semen yield from individual boars, the porcine semen used in the experiments consisted of pooled semen from multiple boars raised under identical housing conditions. Although different semen batches were used in the experiments, the consistent intraocular penetration efficacy of SEVs (derived from standardized Duroc boars: 10 to 18 months old, healthy, and maintained under identical husbandry conditions) across distinct experimental groups of mice or rabbits validated our preliminary control over SEV source standardization during early-stage studies.

Under regulatory and ethical frameworks, large-scale production and clinical translation for SEVs still face challenges, including controlling batch-to-batch variation, meeting guanosine 5′-monophosphate (GMP) manufacturing standards, mitigating immunogenicity, and addressing pathogen contamination risks. These issues urgently require stringent standardization. Current clinical exosome research primarily relies on differential centrifugation for isolation, a method validated in this study for effective SEV purification. Assessing the feasibility of large-scale SEV manufacturing under GMP conditions is warranted. Furthermore, before large-scale production and clinical studies, standardized breeding or sourcing protocols that ensure uniform animal species, age, and husbandry conditions while meeting GMP standards establish SEV source standardization.

While rabbit ocular studies demonstrated negligible SEV immunogenicity, rigorous clinical monitoring must track long-term immune responses and antibody-mediated therapeutic interference. To address pathogen risks, we propose a tripartite biosafety strategy: First, implementing source control through breeding clean-grade boars in specific pathogen–free facilities, thereby eliminating potential viral sources at the origin; second, enforcing process surveillance via routine multiplex PCR screening of porcine semen for prevalent viruses, complemented by prophylactic antiviral vaccination protocols and natural antiviral feed additives to suppress viral carriage in swine populations; third, applying stringent pathogen inactivation protocols involving optimized detergent or enzymatic treatment to purified exosomes, ensuring consistent production of contaminant-free vesicles.

Notably, our exploratory studies revealed that human EGF shares functional and mechanistic parallels with SEVs in enhancing ocular penetration. This discovery establishes a foundation for developing EGF-based nanocarriers (e.g., liposomes and nanoparticles) to replicate SEVs’ delivery mechanisms. Alongside advancing SEVs through standardized manufacturing protocols and species compatibility evaluations, developing these bioinspired nanocarriers could serve as an alternative strategy for clinical translation. This two-pronged approach, focusing on optimizing natural exosomes while developing synthetic carriers mimicking their functions, may accelerate the translational of our research. Furthermore, we also aim to investigate their broader therapeutic potential in vision-threatening retinal pathologies, including neovascular age-related macular degeneration and proliferative diabetic retinopathy.

MATERIALS AND METHODS​

Materials​

Citric acid, thiourea, ferrous chloride, boric acid, TMB, H2O2, glucose, GSH, gefitinib, dasatinib, and ML-7 were purchased from Aladdin Co. Ltd. (Shanghai, China). Cationic chitosan, potassium permanganate (KMnO4), and poly(allylamine hydrochloride) (PAH) were obtained from Macklin Biochemical Technology Co. Ltd. (Shanghai, China). Porcine semen was provided by Huaxia Subu Biotechnology Co. Ltd. Distearoylphosphatidylethanolamine-poly(ethylene glycol)2000-folic acid (DSPE-PEG2000-FA) was supplied by Lipoid GmbH (Ludwigshafen, Germany). Dulbecco’s modified Eagle’s medium (DMEM), DMEM/Nutrient Mixture F-12 (DMEM/F12), RPMI 1640, fetal bovine serum (FBS), antibiotics (Penicillin-Streptomycin), Cell Counting Kit-8 (CCK-, NAC, 3-MA, 4′,6-diamidino-2-phenylindole (DAPI), 1,1′-dioctadecyl-3,3,3′,3′-tetramethyindotricarbocyanine iodide, fluorescein isothiocyanate (FITC), 2′,7′-Dichlorodihydrofluorescein diacetate (DCFH-DA), LysoTracker Green, MitoTracker Red, protamine, NaN3, mβ-CD, amiloride, chlorpromazine, filipin, and human EGF were acquired from Dalian Meilun Biotechnology Co. Ltd. (Dalian, China). The porcine-derived anti-EGF antibody was purchased from Lianzu Biotechnology Co. Ltd. (Shanghai, China, catalog no. LZ-KH1290). TNF-α, IL-6, IL-1β, and TGF-β enzyme-linked immunosorbent assay (ELISA) kits were bought from Elabscience Biotechnology Co. Ltd. (Wuhan, China). H2O2 assay kit, GSH analysis kit, Actin-Tracker Red-594, Rhodamine 123, Lipid Peroxidation malondialdehyde (MDA) Assay Kit, Glucose Assay Kit with O-toluidine, Autophagy Staining Assay Kit with MDC, Caspase-3 activity kit, Caspase-7 activity kit, bicinchoninic acid (BCA) protein assay kit, and Annexin V–FITC/Propidium Iodide (PI) Apoptosis Detection Kit were obtained from Beyotime Biotechnology (Shanghai, China). The other reagents were of analytical or chromatographic purity grade.

Cells and animals​

HCECs (Shanghai EK-Bioscience Biotechnology Co. Ltd., no. CC-Y1835), HLE-B3 (Shanghai EK-Bioscience Biotechnology Co. Ltd., no. CC-Y1830), and HUVECs (Shanghai EK-Bioscience Biotechnology Co. Ltd., no. CC-Y1285) were cultivated in DMEM supplemented with 10% FBS and 1% antibiotics. Human retinal pigment epithelial cells (ARPE-19, Shanghai EK-Bioscience Biotechnology Co. Ltd., no. CC-Y1051) were cultivated in DMEM/F12 medium supplemented with 10% FBS and 1% antibiotics. Human RB cells (Y79, Shanghai EK-Bioscience Biotechnology Co. Ltd., no. CC-Y1545) and Fluc/GFP–Rb-1 cells were cultured in RPMI 1640 medium supplemented with 10% FBS and 1% antibiotics. All cells were validated using short tandem repeat analysis by Shanghai EK-Bioscience Biotechnology Co. Ltd. and maintained in a humidified incubator at 37°C with 5% CO2. Nude mice (6- to 8-week-old) and Kunming mice (6- to 8-week-old), as well as Japanese White rabbits (5- to 7-month-old), were used in this study and were obtained exclusively from Liaoning Changsheng Biotechnology Co. Ltd. (Benxi, China). All animals were maintained at 22° ± 2°C on a 12-hour light-dark cycle with access to food and water ad libitum. Mice were anesthetized via intraperitoneal injection of a ketamine-xylazine mixture (100 mg/kg–10 mg/kg), whereas rabbits underwent intramuscular administration of the same agents at adjusted doses (40 mg/kg–5 mg/kg), respectively. Furthermore, all animal experiments were conducted in accordance with the guidelines approved by the Animal Care and Use Committee of Shenyang Pharmaceutical University (Shenyang, China) (approval number: SYPU-IACUC-2024-1221-401).

Instrumentation and characterization​

Morphological information was obtained by a TEM (JEM-2100, JEOL, Tokyo, Japan). Raman spectrum was recorded using a laser confocal micro-Raman spectroscopy (InVia Reflex, Renishaw, London, Britain). FTIR spectroscopy was collected using a Nexus 870 FTIR spectrometer. The XPS was obtained with an ESCALAB 250 XPS using Al Kα radiation (1486.6 eV). Near-edge x-ray absorption fine structure was carried out at the Catalysis and Surface Science Endstation at the BL11U beamline in the National Synchrotron Radiation Laboratory in Hefei, China. UV-Vis spectra were measured using a UV-1800PC spectrophotometer (Shanghai Meipuda Instrument Co. Ltd., China). The fluorescence spectra were obtained on a Hitachi FL-7100 (Hitachi High Technologies Corporation, Tokyo, Japan). The hydrodynamic diameters were measured using DLS. Zeta potential and DLS analysis were performed on a Zetasizer analyzer (Nano ZS90, Malvern Instruments Ltd.). NTA was performed to characterize the particle concentration. Fluorescence microscopy images were recorded using CLSM (Leica TCS SP8X).

Methods​

Synthesis of cationic MnO2, CDs, and N-CDs​

The synthesis of MnO2 nanoparticles was conducted according to a reported procedure with some modifications (56). First, PAH (50 mg in 2 ml of deionized water) was used to reduce KMnO4 (100 mg in 10 ml of deionized water) to MnO2. The mixture was centrifuged at 50,000 rpm for 30 min. Subsequently, the collected nanoparticles were resuspended in a cationic chitosan aqueous (2 mg/ml) solution and centrifuged again at 50,000 rpm for 30 min. The final product was resuspended in purified water to obtain cationic MnO2.
The CDs were synthesized by the solvothermal method. Briefly, citric acid (2.0 g) and thiourea (4.0 g) were used as carbon and nitrogen sources, respectively, and were mixed with boric acid (0.3 g) and ferrous chloride (0.3 g) in 20 ml of DMF. The mixture was sonicated and then transferred to a 30-ml polytetrafluoroethylene-lined autoclave, heated at 200°C for 6 hours. The product was mixed with methanol (60 ml), centrifuged at 12,000 rpm for 10 min, washed three times with methanol to remove DMF, and dried under vacuum to obtain solid CDs.

The N-CDs were synthesized under conditions identical to those used for undoped CDs, except for the omission of ferrous chloride. All other reaction parameters, including temperature, duration, and precursor ratios, were rigorously maintained to ensure a controlled comparison of doping effects. The synthesis procedure of N-CDs remained identical to that of CDs in all experimental parameters, with the sole exception of excluding ferrous chloride addition during the preparation process.

POD-like activity of CDs and N-CDs​

The POD activity was evaluated from the recipes mentioned elsewhere (19). The typical POD-like activity assays were determined in the presence of TMB (as the substrate) and H2O2 in buffer solution. The absorbance of the blue-colored product (at 652 nm) was measured at a particular reaction time using a microplate reader spectrophotometer. The control of pH conditions was implemented on the basis of methodologies established in previous studies (57). The standard steady-state kinetic tests were conducted at 37°C in reaction buffer solution (0.1 M HAc-NaAc, pH 4.5). The equal volume of CD or N-CD solution (10 μl of 20 μg/ml) was allowed to react in the presence of TMB and H2O2 substrates. By fixing the working concentration of H2O2 (192 mM) and varying the working concentration of TMB (0 to 350 μM), the reaction kinetics of CDs or N-CDs were recorded. Similarly, by altering the working concentration of H2O2 (0 to 900 mM) while maintaining a constant TMB concentration (250 μM), the POD-like kinetics of CDs or N-CDs were analyzed. The characteristic Michaelis-Menten constant (Km) and velocity (Vmax) were calculated using the saturation curve in Origin 2021.

Isolation of SEVs​

The fresh porcine semen from Duroc boars (the certification of species origin is shown in fig. S61) aged 10 to 18 months was diluted five times with PBS to its original volume and then subjected to centrifugation at 300g for 10 min to remove proteins and other impurities. The supernatant was subsequently centrifuged at 2000g for 10 min to eliminate residual cells and debris, followed by a centrifugation step at 15,000g for 30 min to clear apoptotic bodies and cellular fragments. After filtration through a 0.22-μm membrane, the supernatant underwent further centrifugation at 126,000g for 1.5 hours. The resulting pellet was resuspended in PBS (pH 7.4) to obtain an extracellular vesicle suspension. Following quantification with a BCA protein assay kit (Beyotime Biotechnology, catalog no. P0009), the extracellular vesicle suspension was diluted to a concentration of 2 mg/ml and stored at −80°C for future use.

Cumulative permeability of SEVs using Franz diffusion​

The fresh corneas of rabbits were dissected and placed between the donor and receptor chambers of Franz diffusion. Each Franz diffusion cell was placed on a thermostatic stirrer. Different samples were then added into the donor chamber (a 1-ml sample solution in donor chamber), and PBS was added into the receptor chamber (a 5-ml receptor solution in receptor chamber). A 0.2-ml receptor solution was collected, and a 0.2-ml fresh receptor solution was replenished at the 0.5-, 1-, 2-, 4-, and 10-hour time points. The concentration of penetrated CDs was measured according to the fluorescence signal to calculate the cumulative permeability.

The penetration of SEVs after topical ophthalmic delivery​

After administration, the mice were euthanized, and their eyes were collected at different time points. The eyes were then homogenized using a tissue homogenizer, and the concentration of CDs in the resulting lysate was quantified. To evaluate intraocular distribution in mice, mice were euthanized at 1, 3, and 6 hours postadministration. The eyes were then dehydrated overnight in 30% sucrose and fixed in Davidson’s solution for 0.5 hours. DAPI-stained cryosections were observed under an inverted fluorescence microscope. To determine the permeability of SEVs in rabbit eyes, rabbits from different groups were euthanized, and their eyes were collected at various time points. The eyes were cleaned and dissected into cornea, aqueous humor, iris, lens, vitreous body, and retina. Each part was weighed and lysed. In addition, eyes collected at different time points were dehydrated overnight in 30% sucrose and fixed in Davidson’s solution for 0.5 hours. DAPI-stained cryosections were observed under an inverted microscope.

Immunocytochemistry analysis of TJ proteins in HCECs​

HCECs were seeded in 12-well plates at density of 2.0 × 104 cells per well and incubated 24 hours before coculturing with samples. PBS was used as the negative control. After treatment, the cells were rinsed with PBS, followed by fixation for 15 min and block for 30 min at room temperature. Subsequently, for F-actin staining, the cells were incubated with 300 μl of Actin-Tracker Red-594 (1:300; Beyotime Biotechnology, catalog no. C2205S) in the dark at room temperature for 15 min. For ZO-1, occludin, and E-cadherin staining, the cells were incubated with anti–ZO-1 (1:100; Beyotime Biotechnology, China, catalog no. AF8394), anti-occludin (1:500; Abcam, USA, no. Ab216327), and anti-E-cadherin (1:500; Abcam, USA, no. Ab216327) primary antibody at 4°C overnight. After washing away unbound primary antibodies, the cells were further incubated with FITC-conjugated secondary antibody (1:1000; Abways, catalog no. AB0121) in the dark at room temperature for 1 hour. DAPI was used for cell nuclear counterstaining. After the staining, CLSM images were captured.

In vitro simulating corneal epithelial barrier using a Transwell membrane system​

HCEC was first seeded in the upper chamber of the system at a density of 6 × 104cells per well to simulate the corneal epithelial barrier. The cell culture surfaces were coated with polycarbonate (6.5 mm; Corning) to promote cell adhesion and TJ formation. While the TEER value of HCECs remained stable, SEVs-CDs (with a protein concentration of 1 mg/ml in SEVs), Y79Es-CDs (with a protein concentration of 1 mg/ml in Y79Es), ProLs-CDs (with a protamine concentration of 1 mg/ml), CD solution, or EGF solution (50 μg/ml) was added to the insert chamber for 1 hour. Subsequently, each formulation was rinsed three times with PBS, and fresh medium was added. The TEER value was monitored at 1, 3, 6, 9, 12, and 24 hours. The following formula was used to calculate TEER (in ohm centimeters squared):

TEER (ohm·cm2) = (R1 − R0) × 0.33, where R1 is the TEER value of inserts with cells and R0 is the TEER value of inserts without cells. The insert membrane area in the plate was 0.33 cm2. To evaluate permeability of different groups, the apparent permeability coefficient (Papp; in centimeters per second) was calculated as previously reported (25).

To investigate the roles of SEVs and EGF (50 μg/ml) in activating corneal barrier–associated pathways, inhibitor pretreatment experiments were performed. Corneal epithelial monolayers were incubated with gefitinib (1 μM), dasatinib (50 nM), or ML-7 (1 μM) for 4 hours, followed by three washes with fresh culture medium to remove residual inhibitors. SEVs, SEVs (anti-EGF) [the SEVs were incubated with EGF antibody (5 μg/ml) at 37°C under gentle shaking at 15 rpm for 2 hour], or EGF was then applied to the treated monolayers. TEER was monitored using a standardized protocol as previously described.

For molecular pathway analysis, cells from each experimental group were lysed, and target proteins were assessed by Western blotting. Key signaling nodes [e.g., EGFR/phosphorylated EGFR (pEGFR), Src/pSrc, MLCK, and MLC/pMLC] were analyzed to confirm pathway-specific inhibition on barrier integrity; Anti-EGFR (1:1000; Abcam, USA, no. Ab32077), anti-pEGFR (1:1000; Abcam, USA, no. Ab134005), anti-Src (1:1000; Abcam, USA, no. Ab109381), anti-pSrc (1:1000; Abcam, USA, no. Ab4816), anti-MLCK (1:1000; Abcam, USA, no. Ab314185), anti-MLC (1:1000; Abcam, USA, no. Ab92721), anti-pMLC (1:1000; Cell Signaling Technology, USA, no. 3672S), anti–glyceraldehyde-3-phosphate dehydrogenase (GAPDH) (1:1000; Abcam, USA, no. Ab8245).

EGFR knockout in HCEC cells was performed using CRISPR-Cas9 knockout plasmids (Santa Cruz Biotechnology, Santa Cruz, CA) following established methods as previously described and according to the manufacturer’s protocol. All other experimental procedures remained consistent with those described earlier.

Protein extraction​

Frozen samples (about 100 mg) were quickly ground into fine and uniform powder in liquid nitrogen and then homogenized in 1 ml of phenol extraction buffer, after that 1 ml of saturated phenol with tris-HCl (pH 7.5) was added. After several times shake, the mixture was kept at 4°C for 30 min. The upper phenolic phase was separated from the aqueous phase by centrifugation at 7100g at 4°C for 10 min, transferred to a fresh tube, and mixed with five volumes of precold 0.1 M ammonium acetate–methanol. After being kept at −20°C overnight, the mixture was centrifuged at 12,000g for 10 min at 4°C to pellet precipitated protein. For the wash step, the pellet was resuspended twice with precold methanol and twice with ice-cold acetone. Following another round of centrifugation, the pellet was collected, air dried, and resuspended with 300 μl of lysate solution. After incubation of 3 hours at room temperature, the solution was centrifuged to remove any insoluble fraction, and the resulting supernatant contained the total extractable protein. The total protein concentrations were quantified by BCA assay.

Liquid chromatography–mass spectrometry​

The proteomic data analysis was performed by Shanghai OE Biotech Co. Ltd. (Shanghai, China). All analyses were performed by a timsTOF Pro mass spectrometer (Bruker) equipped with an EASY-Spray source (Thermo Fisher Scientific, USA). Samples were loaded by a C18 column (15 cm × 75 μm) on an EASY-nLC 1200 system (Thermo Fisher Scientific, USA). The flow rate was 300 nl/min, and linear gradient was set as follow: 0 to 20 min, 5 to 22% B; 20 to 24 min, 22 to 37% B; 24 to 27 min, 37 to 80% B; 27 to 30 min, 80% B. Ion mobility is set from 0.7 to 1.3 Vs/cm2, and the collision energy range from 20 to 59 eV. The tandem mass spectrometry spectra were recorded from 100 to 1700 mass/charge ratio.

Preparation of CM, CMG, SEVs@CMG, and FA-SEVs@CMG​

To safeguard the CDs from potential fluorescence quenching due to direct contact, we encapsulated MnO2 within cationic chitosan. Cationic MnO2 was mixed with negatively charged CDs at room temperature under dark conditions for 2 hours, followed by centrifugation at 16,000 rpm for 10 min to obtain CM. The CM was then resuspended and further mixed with Gox at room temperature under dark conditions for 2 hours, followed by centrifugation at 16,000 rpm for 10 min to obtain CMG.

FA was incorporated into SEVs through the self-assembly insertion method (58). Briefly, DSPE-PEG2000-FA was dissolved in dimethyl sulfoxide and then mixed with SEVs at a DSPE-PEG2000-FA:SEV protein weight ratio of 1:1. After 4 hours of incubation at 37°C, unincorporated (free) DSPE-PEG2000-FA was removed by ultrafiltration. Quantitative determination of free FA concentration was performed after ultrafiltration using Thermo Fisher Scientific U3000 High-Performance Liquid Chromatography System (Thermo Fisher Scientific Inc.) The conjugation efficiency was calculated as
Subsequently, CMG (with excess SEVs) was added and extruded using a liposome extruder equipped with 200- and 100-nm polycarbonate membrane filters, respectively. Last, centrifugation at 16,000 rpm for 10 min removed blank SEVs, obtaining FA-SEVs@CMG. The preparation of SEVs@CMG omitted the incubation process with DSPE-PEG2000-FA. The preparation of SEVs-CDs, Y79Es-CDs, and ProLs-CDs followed a similar procedure to SEVs@CMG.

Cellular uptake, uptake mechanism, and cytotoxicity​

Y79 cells cultivated on six-well plates at a primary density of 100,000 cells per well were incubated with CDs, CM, CMG, SEVs@CMG, and FA-SEVs@CMG (containing the same concentration of CDs) for 1 or 4 hours. After washing with 10 mM PBS three times, the cells were resuspended in 10 mM PBS, and the mean fluorescence intensity of CDs was tested by flow cytometer. Furthermore, the incubated cells were stained with a lysosomal green probe, washed with 10 mM PBS, and subsequently observed under CLSM for further analysis.

To identify probable internalization mechanisms of CDs, CMG, SEVs@CMG, and FA-SEVs@CMG, Y79 cells were preincubated at 4°C or treated with inhibitors at 37°C for 0.5 hours, followed by incubation with the respective formulations at 4°C or in the presence of inhibitors at 37°C for another 1.5 hours. The cellular uptake of the formulations at 37°C served as the positive control (100% uptake efficiency). Other culture procedures were the same as normal cell uptake tests. Before adding the formulations, chlorpromazine, sodium azide, filipin, mβ-CD, and amiloride were added into the six-well plates and incubated for 1 hour, respectively. Thereafter, CDs, CM, CMG, SEVs@CMG, and FA-SEVs@CMG were added and incubated for 4 hours. Flow cytometry was used to quantify the internalization amount of each group, and the ratio of each inhibitor group to the control was calculated.

In cytotoxicity evaluations, HCEC, ARPE-19, or Y79 cells at the logarithmic growth phase were cultivated on 96-well plates at a primary cell density of 2000 cells per well for 24 hours. Cells were then incubated with CDs, CM, CMG, SEVS@CMG, and FA-SEVs@CMG at concentration gradients for 4 hours, and PBS was used as the negative control. A total of 100 μl of CCK-8 working solution was then added, followed by another 2 hours of incubation. Quantification of the cell viability was achieved by measuring the absorbance with Tecan’s Infinite M200 microplate reader at 450 nm.

Bio-TEM​

The 0.1 M cacodylate buffer containing glutaraldehyde [2.5% (v/v)] solution was used to fix the seeded cells and then added 1% osmium tetroxide (w/v). After dehydration with methanol, the cells were infiltrated and lastly implanted in Epoxy resin. The ultrathin section was counterstained with 4% uranyl acetate (w/v), lead citrate, and lastly monitored under the electron microscope.

Intracellular •OH, MMP, LPO, GSH, glucose, and H​

Y79 cells (5 × 105 cells per well) were grown in six-well plates and incubated for 24 hours. The RPMI 1640 medium was then replaced with fresh medium containing PBS, CDs, CM, CMG, SEVs@CMG, and FA-SEVs@CMG, respectively. After 4 hours, the cells underwent two washes with PBS, followed by the addition of a 50 μM DCFH-DA solution and incubation for duration of 30 min. The cells were subjected to analysis using flow cytometry and CLSM. Furthermore, the Rhodamine 123 dye staining was used to detect the MMP. The intracellular GSH level was assessed using a reduced GSH assay. The intracellular LPO content was quantified using an MDA assay kit. The intracellular glucose content was detected by a Glucose Assay Kit with O-toluidine. The intracellular H2O2 level was measured by H2O2 assay kit.

Autophagy and apoptosis assays​

Y79 cells (5 × 105 cells per well) were cultured in six-well plates and incubated for 24 hours. Subsequently, the culture medium was replaced with medium containing CDs, CM, CMG, SEVs@CMG, and FA-SEVs@CMG, respectively. The cells were then stained with MDC for autophagosomes at 37°C for 30 min. After three washes with assay buffer, the cells were immediately observed under a fluorescence microscope.

The apoptosis of the cells was measured using the Annexin V–FITC Apoptosis Detection Kit. The cells were seeded in six-well plates (106 cells per well) and incubated for 24 hours. The cells were treated with CDs, CM, CMG, SEVs@CMG, and FA-SEVs@CMG for 4 hours, respectively. The cells were collected and incubated with annexin V and PI at room temperature for 15 min. Last, cells were assessed by flow cytometry analysis. The antibodies required for Western blot detection in Y79 cells include anti–Caspase-9 (1:1000; Servicebio, China, no. GB11053), anti–Caspase-8 (1:1000; Servicebio, China, no. GB11594), anti-LC3 (1:1000; Servicebio, China, no. GB11124), anti–Bcl-2 (1:1000; Servicebio, China, no. GB154380), anti-BID (1:1000; Abcam, USA, no. Ab32060), anti-GPX4 (1:1000; Servicebio, China, no. GB115275), and ACTIN (1:1000; Servicebio, China, no. GB15003).

In vivo studies​

Male and female 6- to 8-week-old nude mice were divided into six groups (n = . Ketamine (100 mg/kg) and xylazine (10 mg/kg) were administered intraperitoneally. To establish the RB xenograft model, 1 × 104 tumor cells suspended in a final volume of 1 μl were injected under the retina (59). Five days later, normal saline, CDs, CMG, SEVs@CMG, and FA-SEVs@CMG (each containing 1 mg/ml of CDs, administered at a dosage of 10 μl/day) were administered via topical ocular instillation. Tumor growth was monitored using bioluminescence imaging with a Caliper IVIS imaging system. Mice were infused with 150 mg/kg body weight of d-luciferin 10 min before imaging. At 6, 12, and 24 hours after the initial administration, mice were euthanized to collect eyeballs and tumors, in which CD content was quantified to calculate the percentage of tumor-penetrating CDs relative to the total administered dose at each time point.

Electroretinography​

Mice were adapted to the dark overnight, and their eyes were dilated with 1% tropicamide under dim red light. GENTEAL was applied to each eye, and recordings were made using gold electrodes that were placed on the cornea. The reference electrode was inserted subcutaneously behind the head, and the ground electrode was inserted in the tail. ERG data were collected using an Espion E2 system, and mouse body temperature was maintained at 37°C during the experiment.

TUNEL and P62 assay​

TUNEL was used to analyze cell death in the cornea after different treatments. Briefly, frozen slices of eye tissues were fixed with eyeball fixation solution for 15 min, washed three times with PBS, permeabilized in freshly prepared 0.1% Triton X-100, and then incubated with terminal deoxynucleotidyl transferase solution for 60 min at 37°C. Furthermore, the tumor slices were subjected to stain with autophagy-related markers (p62) and labeled with its secondary antibodies. Last, the samples were imaged by CLSM.

In vivo fluorescence imaging​

Healthy nude mice were treated with eye drops of CD solution, CMG, SEVs@CMG, and FA-SEVs@CMG. In vivo fluorescence imaging was measured at different time points (1, 10, 20, 30, 40, 60, 90, 120, 150, and 180 min) using the IVIS Lumina LT system. In addition, RB-bearing nude mice were subjected to fluorescence imaging at 1, 10, 20, and 30 days posttreatment with each formulation.

In vivo safety study​

After 30 days of continuous administration, PBS, CDs, CMG, SEVs@CMG, and FA-SEVs@CMG side effects were also examined in mice and rabbits. The mice were euthanized, and the eyeball, heart, liver, spleen, lung, and kidney were harvested for H&E staining to assess tissue damage. OCT imaging was used to assess the structural integrity of retinal and choroidal morphologies across all therapeutic groups through comparisons with healthy controls. Whole blood samples were obtained for hematological parameter analysis using the Mindray veterinary automatic hematology analyzer (BC-2800Vet). In addition, the collected serum was measured for alanine aminotransferase, aspartate aminotransferase, creatinine, blood urea nitrogen, white blood cell, red blood cell, hemoglobin, platelets, mean corpuscular hemoglobin, and red cell volume distribution width using Auto Analyzer.

In addition, slit-lamp observation was used to assess rabbit corneal integrity, and tonometry was used to monitor IOP changes. The levels of TNF-α, TGF-β, IL-6, and IL-1β in the cornea were measured using ELISA kits. Immunohistochemical staining with CD4+ and CD80 markers on rabbit ocular sections was performed to evaluate immunogenicity across experimental groups.

Statistical analysis​

All statistical analyses were conducted using the Origin 2021 software. Measurement data were presented as mean ± SD. A two-tailed t test was used for comparisons between two groups, while a one-way analysis of variance (ANOVA) with a Tukey post hoc test was conducted for multiple group comparisons. All the data are presented as the means ± SDs. The quantitative assessment of fluorescence intensity was performed via ImageJ, a commercial image analysis program.

Acknowledgments:​

We are particularly grateful to G. Wei from Fudan University, Shanghai, and W. Wang from the University of Hong Kong for providing the fluorescently labeled tumor cells.

Funding:​

This work was funded by the National Natural Science Foundation of China (grant no. 82172086 to Y.Z.), the National Key R&D Program of China (grant no. 2020YFE0201700 to X.T.), and the Frontier Technology Platform Program of Educational Department of Liaoning Province (grant no. LJ232410163022 to Y.Z.). The funders had no role in study design, collection, analysis and interpretation of data, in the writing of the report, and in the decision to submit this article for publication.

Author contributions:​

J.Z.: Writing—original draft, conceptualization, investigation, writing—review and editing, methodology, resources, data curation, validation, supervision, formal analysis, software, project administration, and visualization. Y.Z.: Writing—original draft, conceptualization, investigation, writing—review and editing, methodology, resources, funding acquisition, data curation, validation, supervision, formal analysis, software, project administration, and visualization. T.Y.: Investigation. Y.D.: Investigation, data curation, and formal analysis. X.L.: Software. C.C.: Investigation, formal analysis, and software. X.B.: Resources. H.L.: Visualization. M.W.: Software. H.H.: Writing—review and editing. J.G.: Writing—review and editing. X.T.: Writing—review and editing, funding acquisition.

Competing interests:​

The authors declare that they have no competing interests.

Data, code, and materials availability:​

All data and code needed to evaluate and reproduce the results in the paper are present in the paper and/or the Supplementary Materials. Please contact the corresponding author for access to materials generated by this study (pharmzy@163.com).

Supplementary Materials​

The PDF file includes:​

Figs. S1 to S61
Tables S1 to S3
Legend for data S1
PDF attached to this post.

Other Supplementary Material for this manuscript includes the following:​

Data S1

Excel spreadsheet

REFERENCES​

1 R. Manukonda, R. V. L. Narayana, S. Kaliki, D. K. Mishra, G. K. Vemuganti, Emerging therapeutic targets for retinoblastoma. Expert Opin. Ther. Targets 26, 937–947 (2022).
GO TO REFERENCE CROSSREF PUBMED WEB OF SCIENCE GOOGLE SCHOLAR

2 M. Plousiou, A. De Vita, G. Miserocchi, E. Bandini, I. Vannini, M. Melloni, N. Masalu, F. Fabbri, P. Serra, Growth inhibition of retinoblastoma cell line by exosome-mediated transfer of miR-142-3p. Cancer Manag. Res. 14, 2119–2131 (2022).
GO TO REFERENCE CROSSREF PUBMED WEB OF SCIENCE GOOGLE SCHOLAR

3 M. R. T. Silva, T. M. Peng, X. Zhao, S. Li, M. Farhan, W. H. Zheng, Recent trends in drug-delivery systems for the treatment of diabetic retinopathy and associated fibrosis. Adv. Drug Deliv. Rev. 173, 439–460 (2021).
GO TO REFERENCE CROSSREF PUBMED WEB OF SCIENCE GOOGLE SCHOLAR

4 X. Qin, H. Shi, H. Li, B. Chu, J. Zhang, Z. Wen, X. Sun, H. Wang, Y. He, Wearable electrodriven switch actively delivers macromolecular drugs to fundus in non-invasive and controllable manners. Nat. Commun.16, 33 (2025).
GO TO REFERENCE CROSSREF PUBMED WEB OF SCIENCE GOOGLE SCHOLAR

5 J. Shen, H. Gao, L. Chen, Y. Jiang, S. Li, Y. Chao, N. Liu, Y. Wang, T. Wei, Y. Liu, J. Li, M. Chen, J. Zhu, J. Liang, X. Zhou, X. Zhang, P. Gu, Q. Chen, Z. Liu, Eyedrop-based macromolecular ophthalmic drug delivery for ocular fundus disease treatment. Sci. Adv. 9, eabq3104 (2023).
CROSSREF PUBMED WEB OF SCIENCE GOOGLE SCHOLAR

6 K. Jiang, X. Fan, Y. Hu, S. Yao, Y. Liu, C. Zhan, W. Lu, G. Wei, Topical instillation of cell-penetrating peptide-conjugated melphalan blocks metastases of retinoblastoma. Biomaterials 284, 121493 (2022).
CROSSREF PUBMED WEB OF SCIENCE GOOGLE SCHOLAR

7 K. Jiang, Y. Hu, X. Gao, C. Zhan, Y. Zhang, S. Yao, C. Xie, G. Wei, W. Lu, Octopus-like flexible vector for noninvasive intraocular delivery of short interfering nucleic acids. Nano Lett. 19, 6410–6417 (2019).
CROSSREF PUBMED WEB OF SCIENCE GOOGLE SCHOLAR

8 Y. C. Kim, M. D. Shin, S. F. Hackett, H. T. Hsueh, R. Lima e Silva, A. Date, H. Han, B. J. Kim, A. Xiao, Y. Kim, L. Ogunnaike, N. M. Anders, A. Hemingway, P. He, A. S. Jun, P. J. McDonnell, C. Eberhart, I. Pitha, D. J. Zack, P. A. Campochiaro, J. Hanes, L. M. Ensign, Gelling hypotonic polymer solution for extended topical drug delivery to the eye. Nat. Biomed. Eng. 4, 1053–1062 (2020).
GO TO REFERENCE CROSSREF PUBMED WEB OF SCIENCE GOOGLE SCHOLAR

9 Y. Tian, T. Zhang, J. Li, Y. Tao, Advances in development of exosomes for ophthalmic therapeutics. Adv. Drug Deliv. Rev. 199, 114899 (2023).
GO TO REFERENCE CROSSREF PUBMED WEB OF SCIENCE GOOGLE SCHOLAR

10 B. Mead, S. Tomarev, Extracellular vesicle therapy for retinal diseases. Prog. Retin. Eye Res. 79, 100849 (2020).
GO TO REFERENCE CROSSREF PUBMED WEB OF SCIENCE GOOGLE SCHOLAR
11 D. Wei, W. Zhan, Y. Gao, L. Huang, R. Gong, W. Wang, R. Zhang, Y. Wu, S. Gao, T. Kang, RAB31 marks and controls an ESCRT-independent exosome pathway. Cell Res. 31, 157–177 (2021).
GO TO REFERENCE CROSSREF PUBMED WEB OF SCIENCE GOOGLE SCHOLAR

12 G. van Niel, G. D’Angelo, G. Raposo, Shedding light on the cell biology of extracellular vesicles. Nat. Rev. Mol. Cell Biol. 19, 213–228 (201.
GO TO REFERENCE CROSSREF PUBMED WEB OF SCIENCE GOOGLE SCHOLAR

13 A. Parra, L. Padilla, X. Lucas, H. Rodriguez-Martinez, I. Barranco, J. Roca, Seminal extracellular vesicles and their involvement in male (in)fertility: A systematic review. Int. J. Mol. Sci. 24, 4818 (2023).
GO TO REFERENCE CROSSREF PUBMED WEB OF SCIENCE GOOGLE SCHOLAR

14 Y. Ma, Q. W. Ma, Y. Sun, X. F. Chen, The emerging role of extracellular vesicles in the testis. Hum. Reprod.38, 334–351 (2023).
GO TO REFERENCE CROSSREF PUBMED WEB OF SCIENCE GOOGLE SCHOLAR

15 L. Samanta, R. Parida, T. R. Dias, A. Agarwal, The enigmatic seminal plasma: A proteomics insight from ejaculation to fertilization. Reprod. Biol. Endocrinol. 16, 41 (201.
CROSSREF PUBMED WEB OF SCIENCE GOOGLE SCHOLAR

16 L. Wehrli, I. Galdadas, L. Voirol, M. Smieško, Y. Cambet, V. Jaquet, S. Guerrier, F. L. Gervasio, S. Nef, R. Rahban, The action of physiological and synthetic steroids on the calcium channel CatSper in human sperm. Front. Cell Dev. Biol. 11, 1221578 (2023).
CROSSREF PUBMED WEB OF SCIENCE GOOGLE SCHOLAR

17 R. Sharma, A. Agarwal, G. Mohanty, R. Jesudasan, B. Gopalan, B. Willard, S. P. Yadav, E. Sabanegh, Functional proteomic analysis of seminal plasma proteins in men with various semen parameters. Reprod. Biol. Endocrinol. 11, 38 (2013).
CROSSREF PUBMED WEB OF SCIENCE GOOGLE SCHOLAR

18 Q. Chen, S. Sun, H. Lin, Z. Li, A. Wu, X. Liu, F. G. Wu, W. Zhang, Supra-carbon dots formed by Fe3+-driven assembly for enhanced tumor-specific photo-mediated and chemodynamic synergistic therapy. ACS Appl. Bio Mater. 4, 2759–2768 (2021).
CROSSREF PUBMED WEB OF SCIENCE GOOGLE SCHOLAR

19 P. Muhammad, S. Hanif, J. Li, Carbon dots supported single Fe atom nanozyme for drug-resistant glioblastoma therapy by activating autophagy-lysosome pathway. Nano Today 45, 101530 (2022).
CROSSREF WEB OF SCIENCE GOOGLE SCHOLAR

20 X. Wu, M. Xu, S. Wang, K. Abbas, X. Huang, R. Zhang, A. C. Tedesco, H. Bi, F,N-doped carbon dots as efficient type I photosensitizers for photodynamic therapy. Dalton Trans. 51, 2296–2303 (2022).
GO TO REFERENCE CROSSREF PUBMED WEB OF SCIENCE GOOGLE SCHOLAR

21 L. H. Fu, Y. Wan, C. Qi, J. He, C. Li, C. Yang, H. Xu, J. Lin, P. Huang, Nanocatalytic theranostics with glutathione depletion and enhanced reactive oxygen species generation for efficient cancer therapy. Adv. Mater. 33, e2006892 (2021).
CROSSREF PUBMED WEB OF SCIENCE GOOGLE SCHOLAR

22 T. Xiao, M. He, F. Xu, Y. Fan, B. Jia, M. Shen, H. Wang, X. Shi, Macrophage membrane-camouflaged responsive polymer nanogels enable magnetic resonance imaging-guided chemotherapy/chemodynamic therapy of orthotopic glioma. ACS Nano 15, 20377–20390 (2021).
GO TO REFERENCE CROSSREF PUBMED WEB OF SCIENCE GOOGLE SCHOLAR

23 C. Ji, Y. Zhou, R. M. Leblanc, Z. Peng, Recent developments of carbon dots in biosensing: A review. ACS Sens. 5, 2724–2741 (2020).
GO TO REFERENCE CROSSREF PUBMED WEB OF SCIENCE GOOGLE SCHOLAR

24 R. Wu, Y. Chong, G. Fang, X. Jiang, Y. Pan, C. Chen, J. J. Yin, C. Ge, Synthesis of Pt hollow nanodendrites with enhanced peroxidase-like activity against bacterial infections: Implication for wound healing. Adv. Funct. Mater. 28, 1801484.1–1801484.11 (201.
WEB OF SCIENCE GOOGLE SCHOLAR

25 C. Liu, L. Tai, W. Zhang, G. Wei, W. Pan, W. Lu, Penetratin, a potentially powerful absorption enhancer for noninvasive intraocular drug delivery. Mol. Pharm. 11, 1218–1227 (2014).
CROSSREF PUBMED GOOGLE SCHOLAR

26 D. Thacharodi, K. P. Rao, Development and in vitro evaluation of chitosan-based transdermal drug delivery systems for the controlled delivery of propranolol hydrochloride. Biomaterials 16, 145–148 (1995).
GO TO REFERENCE CROSSREF PUBMED WEB OF SCIENCE GOOGLE SCHOLAR

27 E. B. Peixoto, C. B. Collares-Buzato, Protamine-induced epithelial barrier disruption involves rearrangement of cytoskeleton and decreased tight junction-associated protein expression in cultured MDCK strains. Cell Struct. Funct. 29, 165–178 (2005).
GO TO REFERENCE 1CROSSREF PUBMED WEB OF SCIENCE GOOGLE SCHOLAR

28 T. H. Yeh, L. W. Hsu, M. T. Tseng, P. L. Lee, K. Sonjae, Y. C. Ho, H. W. Sung, Mechanism and consequence of chitosan-mediated reversible epithelial tight junction opening. Biomaterials 32, 6164–6173 (2011).
GO TO REFERENCE CROSSREF PUBMED WEB OF SCIENCE GOOGLE SCHOLAR

29 K. P. Xu, X. F. Li, F. S. Yu, Corneal organ culture model for assessing epithelial responses to surfactants. Toxicol. Sci. 58, 306–314 (2000).
GO TO REFERENCE CROSSREF PUBMED WEB OF SCIENCE GOOGLE SCHOLAR

30 Y. Wang, Y. Hu, J. An, H. Zhang, X. Liu, X. Li, Z. Zhang, X. Zhang, Liposome-based permeable eyedrops for effective posterior segment drug delivery. Adv. Funct. Mater. 34, 2403142 (2024).
GO TO REFERENCE CROSSREF WEB OF SCIENCE GOOGLE SCHOLAR

31 D. Huang, Y. S. Chen, I. D. Rupenthal, Overcoming ocular drug delivery barriers through the use of physical forces. Adv. Drug Deliv. Rev. 126, 96–112 (201.
GO TO REFERENCE CROSSREF PUBMED WEB OF SCIENCE GOOGLE SCHOLAR

32 B. Chiang, J. H. Jung, M. R. Prausnitz, The suprachoroidal space as a route of administration to the posterior segment of the eye. Adv. Drug Deliv. Rev. 126, 58–66 (201.
GO TO REFERENCE CROSSREF PUBMED WEB OF SCIENCE GOOGLE SCHOLAR

33 C. Zihni, C. Mills, K. Matter, M. S. Balda, Tight junctions: From simple barriers to multifunctional molecular gates. Nat. Rev. Mol. Cell Biol. 17, 564–580 (2016).
GO TO REFERENCE CROSSREF PUBMED WEB OF SCIENCE GOOGLE SCHOLAR

34 S. Moztarzadeh, M. Y. Radeva, S. Sepic, K. Schuster, I. Hamad, J. Waschke, A. García-Ponce, Lack of adducin impairs the stability of endothelial adherens and tight junctions and may be required for cAMP-Rac1-mediated endothelial barrier stabilization. Sci. Rep. 12, 14940 (2022).
CROSSREF PUBMED WEB OF SCIENCE GOOGLE SCHOLAR

35 M. Takeichi, Dynamic contacts: Rearranging adherens junctions to drive epithelial remodelling. Nat. Rev. Mol Cell Biol. 15, 397–410 (2014).
CROSSREF PUBMED WEB OF SCIENCE GOOGLE SCHOLAR

36 L. Li, J. Xin, H. Wang, Y. Wang, W. Peng, N. Sun, H. Huang, Y. Zhou, X. Liu, Y. Lin, J. Fang, B. Jing, K. Pan, Y. Zeng, D. Zeng, X. Qin, Y. Bai, X. Ni, Fluoride disrupts intestinal epithelial tight junction integrity through intracellular calcium-mediated RhoA/ROCK signaling and myosin light chain kinase. Ecotoxicol. Environ. Saf. 257, 114940 (2023).
GO TO REFERENCE CROSSREF PUBMED WEB OF SCIENCE GOOGLE SCHOLAR

37 M. L. Chen, F. M. Tsai, M. C. Lee, Y. Y. Lin, Antipsychotic drugs induce cell cytoskeleton reorganization in glial and neuronal cells via Rho/Cdc42 signal pathway. Prog. Neuropsychopharmacol. Biol. Psychiatry 71, 14–26 (2016).
GO TO REFERENCE CROSSREF PUBMED WEB OF SCIENCE GOOGLE SCHOLAR

38 S. Sayedyahossein, A. Rudkouskaya, V. Leclerc, L. Dagnino, Integrin-linked kinase is indispensable for keratinocyte differentiation and epidermal barrier function. J. Invest. Dermatol. 136, 425–435 (2016).
GO TO REFERENCE CROSSREF PUBMED WEB OF SCIENCE GOOGLE SCHOLAR

39 K. P. Xu, J. Yin, F. S. Yu, SRC-family tyrosine kinases in wound- and ligand-induced epidermal growth factor receptor activation in human corneal epithelial cells. Invest. Ophthalmol. Vis. Sci. 47, 2832–2839(2006).
GO TO REFERENCE CROSSREF PUBMED WEB OF SCIENCE GOOGLE SCHOLAR

40 G. B. Park, D. Kim, Cigarette smoke-induced EGFR activation promotes epithelial mesenchymal migration of human retinal pigment epithelial cells through regulation of the FAK-mediated Syk/Src pathway. Mol. Med. Rep. 17, 3563–3574 (201.
GO TO REFERENCE PUBMED WEB OF SCIENCE GOOGLE SCHOLAR

41 R. R. Rigor, Q. Shen, C. D. Pivetti, M. H. Wu, S. Y. Yuan, Myosin light chain kinase signaling in endothelial barrier dysfunction. Med. Res. Rev. 33, 911–933 (2013).
GO TO REFERENCE CROSSREF PUBMED WEB OF SCIENCE GOOGLE SCHOLAR

42 E. T. Barfod, A. L. Moore, B. G. Van de Graaf, S. D. Lidofsky, Myosin light chain kinase and Src control membrane dynamics in volume recovery from cell swelling. Mol. Biol. Cell 22, 634–650 (2011).
GO TO REFERENCE CROSSREF PUBMED WEB OF SCIENCE GOOGLE SCHOLAR

43 N. Worakajit, S. Satitsri, T. Kitiyakara, C. Muanprasat, Myosin light chain kinase-mediated epithelial barrier dysfunction as a potential pathogenic mechanism of afatinib-induced diarrheas: A study in human colonoid model. Eur. J. Pharmacol. 987, 177174 (2025).
GO TO REFERENCE CROSSREF PUBMED WEB OF SCIENCE GOOGLE SCHOLAR

44 J. Zhou, Y. Liu, J. Chen, N. Xiong, D. Yi, Kaempferol suppresses glioma progression and synergistically enhances the antitumor activity of gefitinib by inhibiting the EGFR/SRC/STAT3 signaling pathway. Drug Dev. Res. 84, 592–610 (2023).
GO TO REFERENCE CROSSREF PUBMED WEB OF SCIENCE GOOGLE SCHOLAR

45 D. P. Wu, Y. Zhou, L. X. Hou, X. X. Zhu, W. Yi, S. M. Yang, T. Y. Lin, J. L. Huang, B. Zhang, X. X. Yin, Cx43 deficiency confers EMT-mediated tamoxifen resistance to breast cancer via c-Src/PI3K/Akt pathway. Int. J. Biol. Sci. 17, 2380–2398 (2021).
GO TO REFERENCE CROSSREF PUBMED WEB OF SCIENCE GOOGLE SCHOLAR

46 J. Wu, M. Q. Barkat, J. Su, F. Wu, D. Tan, T. Shen, Q. He, M. Qu, M. Lu, J. Cai, X. Wu, C. Xu, Inhibition of non-muscular myosin light chain kinase accelerates the clearance of inflammatory cells by promoting the lysosome-mediated cell death. Biomed. Pharmacother. 170, 115986 (2024).
GO TO REFERENCE CROSSREF PUBMED WEB OF SCIENCE GOOGLE SCHOLAR

47 L. Gao, J. Zhuang, L. Nie, J. Zhang, Y. Zhang, N. Gu, T. Wang, J. Feng, D. Yang, S. Perrett, X. Yan, Intrinsic peroxidase-like activity of ferromagnetic nanoparticles. Nat. Nanotechnol. 2, 577–583 (2007).
GO TO REFERENCE CROSSREF PUBMED WEB OF SCIENCE GOOGLE SCHOLAR

48 L. Lin, Q. Zhu, A. W. Xu, Noble-metal-free Fe-N/C catalyst for highly efficient oxygen reduction reaction under both alkaline and acidic conditions. J. Am. Chem. Soc. 136, 11027–11033 (2014).
GO TO REFERENCE CROSSREF PUBMED WEB OF SCIENCE GOOGLE SCHOLAR

49 S. Guan, A. Munder, S. Hedtfeld, P. Braubach, S. Glage, L. Zhang, S. Lienenklaus, A. Schultze, G. Hasenpusch, W. Garrels, F. Stanke, C. Miskey, S. M. Johler, Y. Kumar, B. Tümmler, C. Rudolph, Z. Ivics, J. Rosenecker, Self-assembled peptide–poloxamine nanoparticles enable in vitro and in vivo genome restoration for cystic fibrosis. Nat. Nanotechnol. 14, 287 (2019).
GO TO REFERENCE CROSSREF PUBMED WEB OF SCIENCE GOOGLE SCHOLAR

50 S. Chakraborty, P. Nandi, J. Mishra, Niharika, A. Roy, S. Manna, T. Baral, P. Mishra, P. K. Mishra, S. K. Patra, Molecular mechanisms in regulation of autophagy and apoptosis in view of epigenetic regulation of genes and involvement of liquid-liquid phase separation. Cancer Lett. 587, 216779 (2024).
CROSSREF PUBMED WEB OF SCIENCE GOOGLE SCHOLAR

51 G. Mariño, M. Niso-Santano, E. H. Baehrecke, G. Kroemer, Self-consumption: The interplay of autophagy and apoptosis. Nat. Rev. Mol. Cell Biol. 15, 81–94 (2014).
CROSSREF PUBMED WEB OF SCIENCE GOOGLE SCHOLAR

52 A. Monji, Y. Bando, M. Aoyama, T. Mitsui, T. Murohara, Glucose depletion is essential for the calorie-restriction-mediated cardiac angiogenesis via PKA/AMPK-dependent autophagy. Eur. Heart J. 34, 5870–5870 (2013).
GO TO REFERENCE CROSSREF WEB OF SCIENCE GOOGLE SCHOLAR

53 W. Pfister, T. Blick, S. F. Kok, M. Waltham, 6P antioxidant N-acetylcysteine inhibits cancer-induced osteolysis in mouse models of breast-to-bone metastasis. Ann. Oncol. 23, ii16 (2012).
GO TO REFERENCE CROSSREF GOOGLE SCHOLAR

54 A. Ashkenazi, W. J. Fairbrother, J. D. Leverson, A. J. Souers, From basic apoptosis discoveries to advanced selective BCL-2 family inhibitors. Nat. Rev. Drug Discov. 16, 273–284 (2017).
GO TO REFERENCE CROSSREF PUBMED WEB OF SCIENCE GOOGLE SCHOLAR

55 W. A. Siddiqui, A. Ahad, H. Ahsan, The mystery of BCL2 family: Bcl-2 proteins and apoptosis: An update. Arch. Toxicol. 89, 289–317 (2015).
GO TO REFERENCE CROSSREF PUBMED WEB OF SCIENCE GOOGLE SCHOLAR

56 M. Song, T. Liu, C. Shi, X. Zhang, X. Chen, Bioconjugated manganese dioxide nanoparticles enhance chemotherapy response by priming tumor-associated macrophages toward M1-like phenotype and attenuating tumor hypoxia. ACS Nano 10, 633–647 (2016).
GO TO REFERENCE CROSSREF PUBMED WEB OF SCIENCE GOOGLE SCHOLAR

57 L. Su, S. Qin, Y. Cai, L. Wang, W. Dong, G. Mao, S. Feng, Z. Xie, H. Zhang, N-doped carbon dot nanozymes with acid pH-independence and substrate selectivity for biosensing and bioimaging. Sens. Actuat. B. Chem. 353, 131–150 (2022).
GO TO REFERENCE CROSSREF WEB OF SCIENCE GOOGLE SCHOLAR

58 E. S. Choi, J. Song, Y. Y. Kang, H. Mok, Mannose-modified serum exosomes for the elevated uptake to murine dendritic cells and lymphatic accumulation. Macromol. Biosci. 19, e1900042 (2019).
GO TO REFERENCE CROSSREF PUBMED WEB OF SCIENCE GOOGLE SCHOLAR

59 R. Gao, R. N. Mitra, M. Zheng, K. Wang, J. C. Dahringer, Z. Han, Developing nanoceria-based pH-dependent cancer-directed drug delivery system for retinoblastoma. Adv. Funct. Mater. 28, 1806248 (201.
GO TO REFERENCE CROSSREF PUBMED WEB OF SCIENCE GOOGLE SCHOLAR

TL;DR:

+

=

 

[Harry Dresden]

i wonder if pig semen can make scientists use normal words in their study titles

 

[AirMail]

kits were bought from Elabscience Biotechnology Co. Ltd. (Wuhan, China).

China doing some amazing stuff, but makes me a bit nervous tbh.

 

[Otterly]

Are they implying you could use a drop delivery system to deliver genetic type payloads to the posterior compartment rather than the current injection route? That’s both very interesting and a little bit frightening.

Why frightening? The eye is immune privileged like the brain is, you can’t just drug your way in there. This implies you could create an aerosolised delivery mechanism for breaching the eye, and this could be extrapolated to the blood brain barrier, and thus this paper is basically BIOWEAPONS

 

[pot of sneed]

I guess muslims can't have this treatment lol

 

[Nero Fiddled]

I think we're missing the concerning question here.

Who woke up one morning and thought "I've had a really good idea, I'm gonna get some pig cum, spit it into a syringe and inject it into a mouses eye"?

 

[Dog-O-Tron 5000v5.0]

Now that is a very specific kink

 

[Blewberry Nausea]

Are they implying you could use a drop delivery system to deliver genetic type payloads to the posterior compartment rather than the current injection route?

You’re the educated one I am not 99% of this shit is far over my head. It did say something about this method being useful for bypassing some barrier around the retina

Here’s a more layman article about it

Scientists have used pig semen to develop eye drops that can stop tumour growth in the retina and preserve vision, a study<a href="https://archive.ph/jozl5#ref-CR1">1</a> in mice shows. The work is published today in Science Advances.
It is hoped that the drops could be developed to treat children with retinoblastoma, a cancer of the retina. The condition is typically treated with injections of drugs into the eye, chemotherapy or laser therapy, all of which can damage non-cancerous parts of the eye. Yu Zhang and colleagues at Shenyang Pharmaceutical University in China wanted to develop a method that could penetrate the barrier around the retina and deliver the drugs without causing any unnecessary damage.
The eye drops contain exosomes — tiny particles released by almost every type of cell — engineered to carry molecules that kill cancer cells. Mice treated with the eye drops showed signs that the tumours had stopped growing. They also had comparable eyesight to control animals without tumours.
The team showed that it was possible for exosomes to penetrate the retina. The technique could improve drug delivery across other barriers that are similarly difficult to breach, such as the blood–brain barrier — to treat conditions including Alzheimer’s disease — or the mucosal barrier, says Chunxia Zhao, a researcher studying drug delivery and nanomedicine at Adelaide University in Australia.

https://archive.ph/jozl5

https://www.nature.com/articles/d41586-026-00982-2

Eye drops made from pig semen deliver cancer treatment to mice​

Minuscule particles with the ability to cross hard-to-penetrate barriers can be loaded with drug treatments to target intractable diseases.

 

[REGENDarySumanai]

I guess muslims can't have this treatment lol

I came in here to say this as well. I guess they have to deal with it then.