Development of graphitic carbon nitride quantum dots-based oxygen self-sufficient platforms for enhanced corneal crosslinking

Keratoconus is a bilateral corneal ectasia that typically begins in adolescence and leads to progressive stromal thinning, irregular astigmatism, and potential vision loss. Standard corneal collagen cross-linking (C-CXL) uses riboflavin (RF) and low-intensity UVA (3 mW/cm² for 30 min) to stiffen the corneal stroma but is time-consuming. Accelerated protocols (A-CXL) increase UVA intensity to reduce treatment time; however, stromal oxygen (O2) is rapidly depleted, limiting cross-linking efficacy. Prior work shows O2 is exhausted within seconds at higher irradiances and that cross-linking outcomes are strongly O2-dependent. Existing O2 delivery approaches (contact/non-contact devices, anterior chamber O2) are limited by slow O2 diffusion and short A-CXL exposure times. Photocatalytic, biocompatible O2-generating nanomaterials activated by UVA could locally elevate stromal O2 during A-CXL. Graphitic carbon nitride (g-C3N4) has suitable band energetics for photocatalytic water splitting, absorbs in the UVA range (including 365 nm), and can generate ROS; g-C3N4 quantum dots (QDs) are dispersible and biocompatible. This study develops g-C3N4 QDs and RF@g-C3N4 QDs as oxygen self-sufficient photosensitizers for A-CXL, evaluates biosafety, efficacy under normoxic and hypoxic conditions, and explores mechanisms.

- Conventional CXL (RF + UVA, 3 mW/cm², 30 min) strengthens corneal stroma but is lengthy (Wollensak et al.). - A-CXL aims to shorten time using higher irradiance, but equivalence to C-CXL is debated due to oxygen dependence. Studies show rapid stromal O2 depletion at both 3 and 30 mW/cm² UVA (Kamaev et al.), and lower biomechanical outcomes under hypoxia (Richoz et al.). - Strategies to increase O2 during CXL include contact/non-contact O2 delivery and anterior chamber O2 injection; however, limited by diffusion and short exposure times, with mixed biomechanical benefits in oxygen-enriched environments. - g-C3N4 photocatalysts: suitable HOMO/LUMO alignments for water oxidation/reduction, demonstrated oxygen evolution and visible/UVA absorption; nano-sized g-C3N4 exhibits good biocompatibility and ROS generation and has been explored as an O2 donor for photodynamic therapy. - Prior to this work, g-C3N4 QDs had not been investigated as CXL photosensitizers or as UVA-driven intra-stromal O2 generators for A-CXL.

- Synthesis: g-C3N4 QDs prepared via alkali-assisted thermal treatment of melamine (with KOH/NaOH) at 330–370 °C, followed by aqueous dispersion, filtration (0.2 μm), and dialysis (5 days). RF@g-C3N4 QDs composite formed by mixing lyophilized g-C3N4 QDs (2.5 mg) with RF (0.1 mg) in 1 mL water. Bulk g-C3N4 prepared by annealing melamine at 550 °C. - Characterization: TEM/HRTEM, XRD, FTIR, XPS (C1s, N1s, O1s), UV-Vis diffuse reflectance, fluorescence excitation/emission, zeta potential, size distribution. - Photocatalysis: Dissolved oxygen (DO) measured in sealed conditions under 365 nm UVA irradiation (15 min). Water-splitting reactor assessed O2 evolution under UV-vis or 365 nm UVA with triethanolamine sacrificial agent. Singlet oxygen generation quantified by DPBF probe (absorbance at 410 nm) and ESR under 365 nm UVA (9 mW/cm²). - Stability: Dispersibility and particle size monitored in H2O, PBS, 10% FBS, and 100% FBS over 10 days at 4 °C. - In vitro biocompatibility: CCK-8 viability assays on HCEC, hRMEC, ARPE-19 cells with g-C3N4 QDs up to 800 μg/mL (24 h). Calcein-AM/PI live/dead staining (HCEC). Annexin V-FITC/PI flow cytometric apoptosis analysis (HCEC). - Animal model: Male New Zealand white rabbits (2.0–2.5 kg). Epithelial-off CXL after 30 min presoak with: PBS; RF (0.1% in PBS for A-CXL; 0.1% in 20% dextran for C-CXL); g-C3N4 QDs (2.5 mg/mL); RF@g-C3N4 QDs (2.5 mg/mL QDs + 0.1% RF). UVA 365 nm, 8 mm spot. Irradiance/time regimens: 3 mW/cm² 30 min; 6 mW/cm² 15 or 30 min; 9 mW/cm² 10, 15, or 20 min; 18 mW/cm² 5 min. Comparative A-CXL at 9 mW/cm², 10 min for PBS, RF, g-C3N4 QDs, RF@g-C3N4 QDs; C-CXL at 3 mW/cm², 30 min for RF. - Ex vivo efficacy: Collagenase II enzymatic digestion of 8 mm corneal buttons, serial imaging, residual area quantification (ImageJ), and 48 h dry weight. SEM for stromal ultrastructure. - Biomechanics: Ex vivo stress–strain testing of 10×3 mm corneal strips; Ultimate stress, Young’s modulus. In vivo with Corvis ST at D0, D3, D5, D7, D14, D21: A1L, A2V, HC-R, HC-DA; central corneal thickness (CCT) by Optovue RTVue OCT (D21, D30). - Ocular safety: Modified Draize test (slit-lamp, fluorescein staining) at 0, 6, 24 h after 50 μL of 2.5 mg/mL g-C3N4 QDs. Postoperative slit-lamp monitoring at 0, 1, 3, 7, 14, 30 days. - Endothelium: Specular microscopy (EM-4000) counts at D0, D1, D15, D30; Alizarin Red S + Trypan Blue staining of endothelium (± A-CXL) for morphology and viability. - Histology/apoptosis: Corneal H&E; TUNEL staining (± A-CXL). Systemic H&E of heart, liver, spleen, lung, kidney at D30. - Systemic safety: Hemolysis assay (up to 800 μg/mL). Rabbit bloodwork and serum biochemistry at D0, D1, D15, D30. - Hypoxia mechanism study: A-CXL under nitrogen-purged hypoxic conditions (9 mW/cm², 10 min), enzymatic digestion, residual area and dry weight comparisons among PBS, RF, g-C3N4 QDs, RF@g-C3N4 QDs. - Statistics: Mean ± SD; unpaired t-test, one-way or two-way ANOVA as appropriate.

- Photocatalysis: g-C3N4 QDs (350 °C synthesis) showed strongest UVA-driven O2 generation; after 15 min at 365 nm, DO increased by ~2 mg/L. In a water-splitting system, O2 evolution rate increased from 0.008 to 0.02 mmol g⁻¹ h⁻¹. Robust singlet oxygen production confirmed by DPBF decay and ESR under 365 nm UVA. - Optical properties: g-C3N4 QDs exhibited strong absorption 280–450 nm, excitation max ~370 nm, emission ~411 nm; concentration and annealing temperature affected slight spectral shifts. - In vitro biocompatibility: Cell viability in HCEC, hRMEC, ARPE-19 remained ≥100% up to 800 μg/mL (24 h). Live/dead staining showed minimal PI-positive cells up to 400 μg/mL. Flow cytometry: ~90% viable HCECs after 24 h at 400 μg/mL. - A-CXL optimization (g-C3N4 QDs): Compared to PBS (complete dissolution ~6 h), CXL with 3 mW/cm² was weak; 6 mW/cm² (15 min) markedly enhanced enzymatic resistance; optimal at 9 mW/cm² (10 min) with highest residual dry mass (~0.9 mg after 48 h). 18 mW/cm² (5 min) reduced effect, likely due to O2 imbalance. SEM showed increased fibril density and reduced interfibrillar spaces with g-C3N4 QDs A-CXL. - Comparative A-CXL efficacy: At 9 mW/cm², 10 min, g-C3N4 QDs and RF@g-C3N4 QDs groups had larger residual disc area (~28% of initial) and higher 48 h dry mass than RF 9 mW/cm² (~13%), and were comparable to conventional RF 3 mW/cm², 30 min. Stress–strain: g-C3N4 QDs and RF@g-C3N4 QDs significantly increased ultimate stress and Young’s modulus over PBS and RF 9 mW/cm², comparable to RF 3 mW/cm². - In vivo biomechanics (Corvis ST): g-C3N4 QDs A-CXL significantly improved parameters vs PBS and RF 9 mW/cm²: increased A1L, slower A2V, higher HC-R, lower HC-DA, approaching conventional RF 3 mW/cm² outcomes. CCT: Edema resolved by D21–D30 for g-C3N4 QDs and RF 3 mW/cm²; RF 9 mW/cm² and RF@g-C3N4 QDs showed more persistent edema at D21, with RF@g-C3N4 QDs normalizing by D30. - Ocular safety: Minimal acute irritation with g-C3N4 QDs instillation (no epithelial defects at 6–24 h). Postoperative signs (opacity, hyperemia) subsided faster in g-C3N4 QDs group; near-baseline by day 14. Endothelial counts: significant reduction in RF 3 mW/cm²; only minor decrease in A-CXL groups, especially g-C3N4 QDs. Corneal H&E: no overt damage across groups; endothelial staining showed healthy morphology without blue (dead) cells in g-C3N4 QDs groups ± A-CXL. TUNEL: negligible apoptosis signals after 9 mW/cm², 10 min. - Systemic safety: Hemolysis ~0.6% at 800 μg/mL. Blood/biochemistry: transient abnormalities (ALT, ALB, PLT) at D1/D15, normalized by D30; others not significantly altered. Organ H&E (heart, liver, spleen, lung, kidney) normal at D30. - Hypoxia study: Under N2 hypoxia, RF yielded weak A-CXL (complete dissolution ~8 h). g-C3N4 QDs produced strong resistance with residual disc area ~46% at 48 h; RF@g-C3N4 QDs showed similar performance to g-C3N4 QDs. However, residual dry weight under hypoxia was ~50% of that under normoxia for g-C3N4 QDs, indicating residual O2 limitation. Energy competition between RF and QDs likely reduces composite efficacy. - Overall: g-C3N4 QDs enable dual functionality—local O2 generation and photosensitization—improving A-CXL efficacy at ≥6 mW/cm², optimally 9 mW/cm² for 10 min, with favorable ocular/systemic safety.

This study addresses the central limitation of A-CXL—rapid stromal oxygen depletion—by introducing g-C3N4 QDs that, under 365 nm UVA, both generate oxygen in situ and act as photosensitizers. The results demonstrate that appropriately chosen irradiance/time (e.g., 9 mW/cm², 10 min) allows the QDs to balance O2 production and consumption, overcoming hypoxia-induced inefficiencies and yielding biomechanical strengthening comparable to conventional RF C-CXL but in a shorter time. The superior enzymatic resistance, ex vivo mechanical metrics, and in vivo Corvis ST parameters, along with recovery of CCT, collectively indicate enhanced cross-linking efficacy. Under hypoxia, g-C3N4 QDs still facilitate substantial cross-linking, confirming that photocatalytic O2 generation can mitigate oxygen limitations inherent to A-CXL. The decreased benefit of combining RF with g-C3N4 QDs is consistent with competition for excitation energy and limited O2 availability for two photosensitizers. Importantly, the study underscores that excessively high irradiance with too short exposure (e.g., 18 mW/cm² for 5 min) can reintroduce oxygen imbalance and reduce efficacy. Safety assessments indicate minimal ocular surface irritation, preserved endothelium with A-CXL regimens, negligible apoptosis, and acceptable systemic safety, supporting translational potential. Overall, g-C3N4 QDs offer a promising route to oxygen self-sufficient A-CXL, with practical guidance on irradiance/time optimization.

The authors developed g-C3N4 QDs and RF@g-C3N4 QDs as oxygen self-sufficient photosensitizers for accelerated corneal cross-linking. g-C3N4 QDs exhibit strong UVA-driven oxygen and singlet oxygen generation, excellent dispersibility, and biosafety. In rabbit corneas, A-CXL with g-C3N4 QDs at 9 mW/cm² for 10 min markedly improved enzymatic resistance and biomechanical properties, outperforming RF-based A-CXL and approaching conventional RF C-CXL. Under hypoxia, g-C3N4 QDs maintained significant cross-linking efficacy, validating the oxygen-generating mechanism. Safety was supported by ocular exams, endothelial cell analyses, corneal histology/apoptosis assays, systemic bloodwork, and organ histology. The work suggests g-C3N4 QDs can enable shorter, effective, and safer A-CXL and may be applicable to corneal ectasias. Future research could optimize nanoparticle dose and UVA parameters to further balance O2 generation/consumption, refine formulations for stromal penetration and retention, evaluate long-term transparency and functional outcomes, and assess efficacy in diseased corneas and clinical trials.

- Despite in situ oxygen generation, hypoxia still diminished A-CXL efficacy versus normoxia; residual dry mass under hypoxia was ~50% of that in normoxia for g-C3N4 QDs. - Excessively high irradiance with short duration (18 mW/cm², 5 min) reduced efficacy, indicating a narrow optimization window for irradiance/time to balance O2 dynamics. - Adding RF to g-C3N4 QDs offered limited benefit, likely due to competition for excitation energy and insufficient oxygen to support two photosensitizers. - Efficacy and safety were evaluated in a rabbit model with short- to mid-term follow-up; human clinical data and longer-term outcomes were not assessed in this study.