Sustained intravitreal delivery of carnosic acid by RADA16 peptide hydrogel protects the retina in acute ocular hypertension Authors: Hao Hu, Yiyao Wang, Dadong Jia, Shi Luo, Yi Fan Deng, Ruwen Wang, Xiaoyu Liu, Xuelian Guo, Gang Wu, JingYi Sun, Liang Liang Published: 2026 ABSTRACT This study investigates the use of RADA16 hydrogels for delivering carnosic acid (CA) in glaucoma treatment. Rheological tests at 0.5% (w/v), 1% (w/v), and 2% (w/v) concentrations showed significant elasticity, with Young’s moduli of 30.18 Pa, 16.22 Pa, and 26.66 Pa, respectively. The 1% (w/v) concentration had the lowest stiffness, ideal for intravitreal injection. SEM revealed the hydrogel’s porous microstructure. The RADA16-CA system provided sustained CA release over 72 h, peaking at 6 h. In vitro, high CA concentrations (≥ 256 μg/mL) were cytotoxic, but RADA16 and RADA16-CA enhanced cell proliferation. RADA16-CA demonstrated superior efficacy and biocompatibility. In an acute ocular hypertension model, RADA16-CA improved retinal health in rats by reducing retinal nerve fiber layer thickness, ROS-positive cells, and pro-inflammatory cytokines, while boosting cell survival and antioxidant activity.These findings highlight the potential of the RADA16-CA sustained-release system as an optimized approach for glaucoma treatment.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1038/s41598-026-56818-6. Introduction Glaucoma is a major cause of irreversible blindness, with its global prevalence steadily increasing. By 2030, it is projected to affect 95.4 million people, rising to 111.8 million by 2040, placing significant psychosocial and economic burdens on individuals, families, and society1. The pathogenesis of glaucoma involves complex neurodegenerative processes, driven by multiple risk factors. This includes the loss of retinal ganglion cells (RGCs) and axonal degeneration, leading to irreversible visual impairment. Currently, lowering intraocular pressure (IOP) is the only proven strategy to slow disease progression; however, RGC injury and loss can still occur even with normal IOP levels2. The blood-ocular barrier complicates treatment by restricting drug penetration into the retina, hindering effective neuroprotection against RGC apoptosis3. Therefore, identifying neuroprotective agents and exploring efficient drug delivery methods are crucial clinical priorities. Carnosic acid (CA), a natural diterpenoid phenolic compound, exhibits multimodal neuroprotective effects through various mechanisms. It maintains OPA1 via the Parkin/IKKγ/p65-dependent pathway to promote mitochondrial fusion, inhibits NF-κB-mediated inflammation, activates the Nrf2 antioxidant pathway, and reduces the Bax/Bcl-2 ratio and caspase-3 activity4–8. In models of acute intraocular pressure elevation, diabetic retinopathy, and β-amyloid toxicity, CA significantly reduces RGC loss and alleviates retinal edema9–12. However, CA’s delivery is challenging: systemic administration fails to reach therapeutic intraocular concentrations due to first-pass metabolism and the blood-ocular barrier, while intravitreal injection requires frequent dosing due to CA’s short half-life in the vitreous. Repeated injections also carry risks such as endophthalmitis, retinal detachment, and intraocular hemorrhage12. Therefore, an efficient drug delivery system is urgently needed to enhance CA’s intraocular retention and bioavailability.Ocular controlled-release systems, which use functional carriers to regulate drug binding and release, offer a promising solution. These systems maintain local drug concentrations, preserve bioactivity, improve bioavailability, and minimize adverse effects13. Designing safe, stable, and efficient intraocular drug delivery platforms is a primary research focus. Among various carriers, self-assembling peptide (SAP) hydrogels, particularly RADA16, are ideal candidates for CA delivery. RADA16 demonstrates excellent biocompatibility and biodegradability, forms a vitreous-mimetic 3D network, and contains hydrophobic domains that enhance CA’s aqueous stability. Additionally, its sol–gel transition under physiological conditions allows for minimally invasive intravitreal injection and sustained drug release14–16. This addresses CA’s delivery limitations directly. Through rational molecular design, SAP hydrogels form hydrophobic-core/hydrophilic-shell structures that encapsulate both hydrophilic and hydrophobic drugs14. Their high water content, viscoelasticity, and tissue permeability extend vitreous retention time and improve bioavailability. In summary, as ocular controlled-release carriers, SAP hydrogels extend drug duration, enhance bioavailability, and reduce dosing frequency15. By mimicking tissue microenvironment characteristics and offering high drug loading capacity, they enable precise control over neurocompatibility, injectability, and release kinetics, providing an effective approach for optic nerve protection in glaucoma and optimizing drug delivery. RADA16, a prototypical hydrogel-forming SAP composed of 16 amino acids (Ac-RADARADARADARADA-CONH2), spontaneously forms β-sheets in physiological saline, assembling into nanofibrillar hydrogels. Its 3D structure closely resembles the extracellular matrix (ECM), supporting cell adhesion and growth16. The hydrophobic surfaces of RADA16 bind poorly soluble drug domains, forming peptide–drug complexes that enhance aqueous stability and enable SAP-mediated drug loading and sustained release. Maria Lucia17 demonstrated that RADA16 hydrogels serve as controlled-release carriers for lipophilic drugs, achieving sustained, controllable, and efficient drug release without disrupting peptide self-assembly. Additionally, the delivery performance can be modulated by incorporating functional motifs into the RADA16 backbone. For instance, Wang18 extended the angiogenic motif SVVYGLR to create RADA16-SVVYGLR, which promotes neural stem cell survival, proliferation, angiogenesis, and neuroregeneration at brain injury sites. In conclusion, RADA16’s ECM-mimetic structure, sol–gel transition under physiological conditions, and excellent biocompatibility and biodegradability make it an ideal drug delivery carrier. Given that oral administration and the blood-ocular barrier significantly reduce CA’s bioavailability, we have developed a RADA16-based sustained-release system for intravitreal CA delivery. This study aims to develop a delivery system that enhances CA’s stability and bioavailability, reduces complications associated with frequent injections, and improves therapeutic outcomes in glaucomatous optic neuropathy. Materials and methods Main reagents RADA16 peptide (purity ≥ 90% by HPLC; Wuhan Tiande Biotechnology Co., Ltd., Wuhan, China) and carnosic acid (CA, ≥ 98%; Sigma-Aldrich, St. Louis, MO, USA) were used to prepare the self-assembling peptide hydrogels. DMEM/F-12 medium, fetal bovine serum (FBS) and penicillin–streptomycin were purchased from Gibco, Thermo Fisher Scientific (Waltham, MA, USA). Phosphate-buffered saline (PBS; 1 × , pH 7.4; G4202-500ML; Servicebio, Wuhan, China) was used for cell washing and dilutions. The RGC-5 retinal ganglion-like cell line (authenticated by STR profiling and used at passages 4–12) was obtained from WanWu Biotech (Hefei, China). A Cell Counting Kit-8 (CCK-8; Dojindo Laboratories, Kumamoto, Japan) was used to assess cell viability. ELISA kits for interleukin-1β (IL-1β; detection limit 5 pg/mL), interleukin-6 (IL-6; 10 pg/mL) and tumor necrosis factor-α (TNF-α; 8 pg/mL) were purchased from Cloud-Clone (Wuhan, China). Primary antibodies against Brn3a (1:500), Bcl-2 (1:1000), NLRP3 (1:800) and β-actin (1:2000) were obtained from Bingcure (Wuhan, China), and HRP-conjugated secondary antibodies (1:5000) from Cell Signaling Technology (Danvers, MA, USA). Calcein-AM/PI staining kit, DAPI, total RNA extraction kit and PCR primers were purchased from Wuhan Servicebio Technology Co., Ltd. (Wuhan, China). Alexa Fluor 488 NHS ester was obtained from Thermo Fisher Scientific (Waltham, MA, USA). Ultrapure water was produced using a Milli-Q water purification system (Millipore, Burlington, MA, USA). The main instruments used in this study included a CO₂ incubator (E1911R, SIM, USA), a multifunctional microplate reader (Multiskan FC, Thermo Fisher Scientific, Waltham, MA, USA), a flow cytometer (Accuri C6, BD Biosciences, San Jose, CA, USA), an inverted fluorescence microscope (DMI3000B, Leica Microsystems, Wetzlar, Germany), a laser scanning confocal microscope (A1R + , Nikon, Tokyo, Japan), a UV–visible spectrophotometer (ZF6, Shanghai Jiapeng Technology, Shanghai, China) and a transmission electron microscope (JEM-2100, JEOL Ltd., Tokyo, Japan). Experimental animals One hundred SPF male Sprague–Dawley rats (180–220 g) with normal eyes were obtained from Three Gorges University’s Laboratory Animal Center (License No. SCXK 2022–0012). The study followed ARVO animal research guidelines and was conducted in the university’s Barrier Laboratory, with approval from the Animal Ethics Committee (Approval No. 2025030I). Establishment of an acute high intraocular pressure (HIOP) rat model and grouped drug administration Male SD rats (180–220 g) were assigned to five groups (n = 6 per group; total n = 30). The groups were: Group A: Normal control, no treatment. Group B: HIOP glaucoma model only. Group C: HIOP glaucoma model + RADA16 administration. Group D: HIOP glaucoma model + CA administration. Group E: HIOP glaucoma model + RADA16–CA administration. Acute HIOP was induced by bilateral scleral venous cauterization. Anesthesia was induced and maintained with inhaled isoflurane (2–3%). Rats were placed in the prone position; the operative eye received topical 0.5% procaine hydrochloride, and the periorbital area was disinfected with povidone-iodine. After eyelid opening, the bulbar conjunctiva was incised 1 mm posterior to the corneal margin, the subconjunctival tissue was separated, and the superior scleral veins were exposed. Using microsurgical forceps under magnification, clusters of superior scleral veins were dissected, the main veins (typically at the 3, 6, 9, and 12 o’clock positions) were identified, and the vessels were gently lifted off the scleral surface. A microcautery tip was applied to each vessel for 1–2 s until occlusion was evident; excessive cauterization was avoided to prevent scleral perforation. Successful sealing was confirmed by congestion and distension of the proximal segment and pallor of the distal segment. Suturing and care: the conjunctiva was repositioned and closed with interrupted 8–0 absorbable sutures. Chloramphenicol eye drops were instilled, the rectus muscle suspension thread was removed, and the eyeball was repositioned. Animals were housed in single cages until full recovery from anesthesia. Drug administration began 1 day after modeling. Intravitreal injections were performed in both eyes using a 32G ophthalmic micro-injection needle (5 µl per eye, single dose). Group C received RADA16, Group D received CA, and Group E received RADA16–CA. Final evaluation was conducted 7 days after administration. The intravitreal injection site was located on the sclera 1.5–2 mm posterior to the limbus, with careful avoidance of the cornea, iris, and crystalline lens to minimize drug leakage and tissue injury. A 32G ophthalmic microsyringe needle was used to minimize scleral trauma and reduce postoperative intraocular pressure (IOP) fluctuations and drug reflux. Preparation of RADA16-CA Prior to Injection: The RADA16-CA composite was prepared at room temperature (RT, 20–25 °C). Temperatures above RT were avoided to prevent premature gelation, whereas temperatures below RT were avoided to preserve fluidity and injectability, thereby ensuring unobstructed injection and maintaining optimal viscosity. Aseptic Conditions: Hydrogel preparation and aliquoting were conducted in a laminar flow hood using sterile, pyrogen-free, and nuclease-free consumables throughout to prevent ocular infection. Intraocular injections were performed under an operating microscope; magnification and illumination improved procedural accuracy and reduced the risk of collateral injury to adjacent tissues. The intraocular injection procedure (including site localization, scleral puncture, drug delivery, needle withdrawal, and hemostasis) required approximately 5–8 min per eye. For a single animal receiving bilateral injections, the total surgical duration (including anesthesia induction, bilateral procedures, and immediate postoperative care) was approximately 15–20 min. Control Group Setting: Normal rats were assigned to the control group and subjected to identical experimental conditions, including synchronized anesthesia, to eliminate anesthesia-related confounding. Preparation of RADA16 hydrogel and RADA16-CA sustained-release system To prepare a 1% (w/v) RADA16 hydrogel, lyophilized RADA16 peptide (10 mg; molecular weight 1712.8 Da) was dissolved in 1 mL sterile deionized water and gently shaken for 48 h at 25 °C, followed by adjustment of the pH to 7.4. For construction of the RADA16–CA sustained-release system, CA (19.4 mg; molecular weight 332.4 Da) was added to 1 mL of 1% RADA16 hydrogel to achieve a theoretical feeding mass ratio of 10:1 (CA:RADA16). This ratio was selected based on preliminary experiments to balance drug loading capacity and hydrogel integrity, and the mixture was magnetically stirred at room temperature for 48 h to obtain a homogeneous RADA16–CA colloidal suspension. The encapsulation efficiency (EE) of CA in the hydrogel was determined by HPLC. After drug loading, the suspension was centrifuged and the supernatant was collected and diluted with DMSO. The concentration of non-encapsulated CA in the supernatant (W_free) was quantified using an external standard calibration curve. EE was calculated according to the following equation:EE (%) = [(W total – W free) / W total] × 100%, where W_total is the total amount of CA initially added to the system. Under these conditions, the EE of the 1% RADA16 hydrogel for CA was calculated to be 27.75%. In vitro release of CA from the RADA16–CA system was evaluated using a dialysis method. Equal volumes (600 μL) of RADA16–CA suspension or free CA solution were placed into dialysis bags (molecular weight cut-off [MWCO] 8–14 kDa; Spectrum Laboratories, Rancho Dominguez, CA, USA), equilibrated for 10 min, and immersed in 30 mL sodium hyaluronate solution at 37 °C with shaking at 100 rpm. At predetermined time points (0, 2, 4, 6, 8, 12, 24, 30, 36, 48 and 72 h), 3 mL of the external medium was withdrawn and replaced with an equal volume of fresh medium. CA concentrations were determined by UV spectrophotometry at 254 nm, and the cumulative release percentage was calculated. Rheological characterization of RADA16 hydrogel The viscoelastic properties of the RADA16 and RADA16–CA hydrogels were evaluated using a rotational rheometer (HAAKE MARS 40, Thermo Fisher Scientific, Karlsruhe, Germany) equipped with a cone–plate geometry (cone diameter 25 mm, cone angle 1°, gap 0.5 mm). Freshly prepared hydrogels (0.5%, 1% and 2%, w/v) were gently loaded onto the lower plate, and excess material at the edge was removed. All measurements were performed at 25 °C after an equilibration period of 5 min. First, a strain-sweep test was conducted at a fixed angular frequency of 10 rad/s with a strain range of 0.1–10% to determine the linear viscoelastic region. Subsequently, oscillatory frequency-sweep tests were performed within this region at a constant strain of 1%, with the angular frequency ranging from 0.1 to 100 rad/s. The storage modulus (G′) and loss modulus (G″) were recorded as functions of frequency. The apparent Young’s modulus (E) was calculated from G′ using the relation E = 2G′(1 + ν), assuming a Poisson’s ratio (ν) of 0.5 for soft hydrogels. Scanning electron microscopy (SEM) of hydrogel microstructure The microstructure of blank RADA16 and RADA16–CA hydrogels was examined by scanning electron microscopy (SEM). Hydrogels were allowed to fully gel and then frozen at − 80 °C, followed by lyophilization to obtain dry porous scaffolds. The freeze-dried samples were cut into small pieces, mounted on aluminum stubs using conductive carbon tape and sputter-coated with gold for approximately 30 s using an ion sputter coater (e.g., SC7620, Quorum Technologies, Laughton, UK). Samples were imaged with a field-emission SEM (e.g., SU8010, Hitachi, Tokyo, Japan) at an accelerating voltage of 5–10 kV. Representative images were acquired at different magnifications to observe fiber morphology and pore architecture of the self-assembling peptide network. Fourier transform infrared (FTIR) spectroscopy FTIR spectroscopy was performed using a Nicolet iS5 spectrometer (Thermo Fisher Scientific, Waltham, MA, USA) equipped with a deuterated triglycine sulfate (DTGS) detector. Before each measurement session, the instrument was allowed to warm up for at least 30 min to ensure stabilization of the infrared source and detector. FTIR samples were prepared using the potassium bromide (KBr) pellet method. Briefly, KBr powder was dried at 110 °C for 4 h to minimize moisture interference. Lyophilized hydrogel samples were finely ground in an agate mortar and pestle and mixed thoroughly with dried KBr at a mass ratio of 1:100 (sample:KBr). The mixture was further ground to obtain a uniform fine powder and then compressed under vacuum in a stainless-steel die at approximately 10 tons for 2 min to form a transparent pellet. The pellet was mounted on a magnetic sample holder and placed in the sample chamber for spectral acquisition. Spectra were recorded in the mid-infrared region from 4000 to 500 cm⁻1 at a resolution of 4 cm⁻1 with 32 accumulated scans per sample. A background spectrum was collected from a pure KBr pellet under identical conditions immediately before each sample measurement to correct for atmospheric CO₂ and H₂O interference. Background subtraction, baseline correction, and peak identification were performed using OMNIC software (version 9.2, Thermo Fisher Scientific). Each sample was analyzed in triplicate using independently prepared pellets, and representative spectra were presented. Characteristic absorption bands were assigned according to published literature on peptide secondary structures and phenolic compounds. RGC-5 cell culture and viability assays This section outlines aseptic preparation, thawing, maintenance, subculture, cryopreservation, and viability assessment of RGC-5 cells. All required consumables and instruments should be sterilized by autoclaving and thoroughly dried. Place them in the cell clean bench and expose to ultraviolet light for 15–20 min. Set the water bath to 37 ℃; prewarm complete medium to 37 ℃ and bring PBS to room temperature in advance. Remove the frozen cells from the -80 ℃ refrigerator and quickly place the cryotube in a 37 ℃ water bath, keeping the cap above the water level and the lower part immersed to avoid contamination. Gently agitate the cryotube so that the suspension thaws completely within ~ 1 min; do not vortex. Wipe the outside of the cryotube with 75% alcohol, transfer the contents into a 15 mL centrifuge tube containing 3–4 mL of prewarmed complete medium to dilute the cryoprotectant, then centrifuge at 1000 rpm for 3 min. After centrifugation, carefully aspirate and discard the supernatant. Resuspend the cell pellet in 3 mL of complete culture medium (DMEM/F12, 1% penicillin–streptomycin, 15% fetal bovine serum) by gentle pipetting. Seed the suspension into a 6 cm culture dish and add an additional 1 mL of complete medium to cover the surface evenly. Gently rock to mix and place the dish in a 37 ℃, 5% CO₂ humidified incubator. Cell Passage: Monitor cell morphology and confluence under an inverted microscope. When confluence reaches 80%–90% with minimal debris, proceed to passage. Bring required reagents to room temperature in advance and irradiate the ultraclean bench with UV for 30 min. After irradiation, ventilate the bench for 3 min before starting. Aspirate spent medium, rinse the monolayer with 2 mL PBS, and remove PBS completely. Add 1 mL of 0.25% trypsin–EDTA, incubate at 37 ℃ for 2–3 min, observe rounding and detachment, and tap the dish gently to assist release; neutralize immediately once cells detach. Add 3 mL complete medium to neutralize trypsin and gently pipette to obtain a single-cell suspension. Transfer to a 15 mL tube and centrifuge at 1000 rpm for 3 min. Discard the supernatant, add 3 mL complete medium, and resuspend thoroughly. Seed the desired volume (e.g., 1.5 mL) into a new 6 cm dish, supplement with fresh complete medium to an appropriate volume (e.g., 3–4 mL), mix gently, and return to the 37 ℃, 5% CO₂ incubator. Cell Cryopreservation: When adherent cells reach ~ 80% confluence (logarithmic phase), they are suitable for cryopreservation. Bring reagents to room temperature and irradiate the ultra-clean bench with UV for 30 min. After UV irradiation, ventilate for 3 min before starting. Aspirate culture medium, wash with 2 mL PBS, and remove PBS completely. Add 1 mL trypsin, incubate at 37 ℃ for 2–3 min, and stop digestion once cells detach. Add 3 mL complete medium to neutralize, gently pipette to suspend, transfer to a 15 mL tube, and centrifuge at 1000 rpm for 3 min. Discard the supernatant, resuspend the pellet in 3 mL serum-free cryopreservation solution, aliquot evenly into labeled cryovials (cell line, passage number, date), place vials in a programmed cooling container, and store at -80 ℃ (short-term; for long-term storage, transfer to liquid nitrogen). Cell Counting and Viability: After trypsinization and centrifugation, discard the supernatant and resuspend the pellet in an appropriate volume of culture medium. Pipette 100 μL of the cell suspension and mix with 900 μL PBS (for counting only) or 0.4% trypan blue (for viability assessment) (1:10 dilution). Clean the hemocytometer and coverslip with 75% alcohol and air-dry. Load 10 μL of the diluted suspension into each chamber by capillary action, avoiding bubbles, then count under a light microscope within 3 min. Calculate live cell concentration as: (N/m) × 10 × n × 104 cells/mL, where N is the total number of live cells counted across m large squares, and n is the dilution factor. Cell viability (%) = (number of viable cells / total cells) × 100. CCK-8 assay to assess the effect of different treatments on RGC-5 cell viability To evaluate the effects of different treatments on RGC-5 cell viability, a CCK-8 assay was performed. RGC-5 cells were seeded into 96-well plates at a density of 5 × 103 cells/well in 100 μL complete medium and cultured for 24 h at 37 °C in a humidified incubator with 5% CO₂ to allow attachment and recovery. Cells were then divided into four groups: CA (0–512 μg/mL), 1% RADA16 hydrogel (0–960 μg/mL), 1% RADA16–CA (0–512 μg/mL) and a medium-only control. For each series, the 0 μg/mL condition served as the untreated (vehicle-equivalent) dose point. Before adding CCK-8, ensure the final volume in each well is 100 μL; maintain equal volumes across treatments. After 24 h of treatment, 10 μL of CCK-8 solution was added to each well and incubated for 2 h at 37 °C in the dark. The 2-h incubation time was chosen according to the manufacturer’s instructions and confirmed by preliminary time-course experiments showing a stable linear relationship between absorbance and viable cell number within this time frame. Absorbance at 450 nm was measured using a microplate reader. Blank wells contained medium plus CCK-8 but no cells. Cell viability was calculated as: Cell viability (%) = (A treatment – A blank) / (A control – A blank) × 100%. Each condition was tested in triplicate wells, and the experiment was repeated three times independently. Live/dead staining assay Live/dead staining was performed to assess cell viability under different treatments. RGC-5 cells in 96-well plates were treated with CA, RADA16, RADA16–CA (512 μg/mL for each), or medium (Control) for 24 or 48 h. After treatment, cells were gently washed with PBS, then incubated with Calcein-AM/PI working solution (37 °C, dark, 30 min) according to the manufacturer’s instructions. Following staining, wells were washed once with PBS and imaged immediately under identical exposure settings using a fluorescence microscope. Live (green) and dead (red) cells were visualized using a fluorescence microscope. For quantification, images from randomly selected fields were analyzed in ImageJ,and the percentage of live/dead cells was calculated; each group included three independent replicates. IOP measurement IOP was measured using the iCare TONOLAB (model ic200, iCare Finland Oy, Helsinki, Finland) after a 10-day acclimation period. The probe was positioned perpendicular to the corneal center, and six valid readings were taken to calculate the average IOP per eye by the same operator at the same time of day to minimize circadian and operator variability. Measurements were made at the following time points: baseline (pre-HIOP), immediately post-HIOP induction, and 1, 3, and 7 days post-HIOP induction. Data were analyzed using Friedman ANOVA (nonparametric repeated-measures) with Dunn’s post hoc test. Retinal histology and immunofluorescence staining Ten eyes per group were enucleated, fixed in 4% neutral buffered formalin for 24 h, and embedded in paraffin. Serial Sects. (4 μm) were cut for histological and immunofluorescence analyses. Sections passing through the optic nerve head were preferentially selected to standardize anatomical location. For hematoxylin–eosin (H&E) staining, sections were deparaffinized in xylene, rehydrated through a graded ethanol series and stained using an H&E kit (G1120, Solarbio, Beijing, China) according to the manufacturer’s instructions. After mounting, retinal sections were scanned with a VS200 microscope (Olympus, Tokyo, Japan), and retinal layer thickness was measured using ImageJ software (version 1.8.0, NIH, Bethesda, MD, USA). Total retinal thickness and the thickness of GCL, INL, and ONL were measured at fixed distances (e.g., 500 μm and 1000 μm) from the optic nerve head in three nonoverlapping fields per section, and averages were calculated from three sections per eye. For retinal immunofluorescence staining, paraffin sections were deparaffinized, rehydrated and subjected to heat-induced antigen retrieval (10 mM citrate buffer, pH 6.0, 95 °C, 15 min). After blocking with 5% (w/v) bovine serum albumin (BSA) for 30 min at room temperature, sections were incubated overnight at 4 °C with anti-Brn3a primary antibody (e.g., 1:200). Following PBS washes (3 × 5 min), sections were incubated with FITC-conjugated secondary antibody at 37 °C for 40 min in the dark (e.g., 1:500) and counterstained with DAPI to visualize nuclei. Sections were mounted with an anti-fade medium before imaging. Brn3a-positive cells (surviving RGCs) were imaged using a fluorescence microscope and quantified with ImageJ. Brn3a + RGCs were counted per mm along the ganglion cell layer in three nonoverlapping fields (central retina near the optic nerve head) and averaged across three sections per eye by an observer masked to group allocation. Retinal reactive oxygen species (ROS) fluorescence staining Ten eyes per group were enucleated, frozen, and embedded in optimal cutting temperature (OCT) compound and stored at − 80 °C until sectioning. Sections (10 μm) were prepared, thawed, and incubated with 10 μM dihydroethidium (DHE) in HBSS for 30 min at 37 °C, protected from light. The sections were rinsed with PBS (3 × 5 min), mounted with anti-fade medium, and counterstained with DAPI for 10 min. ROS (red fluorescence) and cell nuclei (blue fluorescence) were observed under a fluorescence microscope using identical exposure settings across groups, and mean fluorescence intensity within the ganglion cell layer was quantified in ImageJ with background subtraction (no-DHE control).For each section, three non-overlapping fields were analyzed, and the averaged values were expressed as relative fluorescence intensity for statistical analysis. Western blotting Retinas were homogenized in ice-cold RIPA buffer supplemented with protease inhibitors, lysates were cleared by centrifugation (12,000 × g, 10 min, 4 °C), and protein concentrations were determined by BCA assay. Equal amounts (30 μg per sample) were separated on 10% SDS-PAGE gels (P0012A, Beyotime, Shanghai, China) and transferred onto PVDF membranes (IPVH00010, Millipore, Burlington, MA, USA). Membranes were blocked with 5% (w/v) non-fat milk (A600669, Sangon Biotech, Shanghai, China) for 1 h at room temperature and incubated overnight at 4 °C with primary antibodies against Bcl-2 (1:1000, ab32124, Abcam, Cambridge, UK), NLRP3 (1:1000, 15,101, Cell Signaling Technology, Danvers, MA, USA), and β-actin (1:5000, TA-09, ZSGB-BIO, Beijing, China). After primary incubation, membranes were washed with TBS-T (0.1% Tween-20) for 3 × 10 min. After washing, membranes were incubated with HRP-conjugated secondary antibody (A0208, Beyotime, Shanghai, China) for 1 h at room temperature. Protein bands were visualized using an ECL detection kit (Thermo Fisher Scientific, Waltham, MA, USA) and quantified with Image Lab software (version 6.1, Bio-Rad, Hercules, CA, USA). Band intensities were normalized to β-actin and expressed relative to the control group (set to 1.0); n = 3 independent biological replicates per group. Determination of retinal SOD activity and MDA content Retinal oxidative stress was assessed by measuring superoxide dismutase (SOD) activity and malondialdehyde (MDA) content in retinal tissue homogenates. At the end of the experiment, rats were euthanized and both eyeballs were enucleated. Under a dissecting microscope, the neurosensory retina was carefully separated from the RPE–choroid complex, weighed, and homogenized in ice-cold PBS (pH 7.4) at a ratio of 1:9 (w/v). All procedures were performed on ice to minimize oxidation. The homogenates were centrifuged at 2,500–3,000 rpm for 20 min at 4 °C, and the supernatants were collected for biochemical analysis. SOD activity and MDA content were measured using commercial assay kits (SOD and MDA Assay Kits, Nanjing Jiancheng Bioengineering Institute, Nanjing, China) according to the manufacturer’s instructions. SOD activity was determined by the xanthine oxidase method, and results were expressed as U/mg protein. MDA levels were quantified using the thiobarbituric acid reactive substances (TBARS) method and expressed as nmol/mg protein. Total protein concentrations were determined by a BCA protein assay, and all samples were measured in duplicate. Standard curves and kit-provided blanks/controls were run in parallel, and absorbance was read on a microplate reader at the recommended wavelengths; results are reported as mean ± SD. Reverse transcription-quantitative polymerase chain reaction (RT-qPCR) to detect retinal target gene transcription levels Total RNA was extracted from rat retinal tissue. After the animals were sacrificed at the end of the experiment, both eyeballs of each Sprague–Dawley rat were enucleated under deep anesthesia, and the neurosensory retina was carefully dissected from the eyecup under a stereomicroscope, immediately snap-frozen in liquid nitrogen and stored at − 80 °C until use. Retinal tissue from each eye was processed individually for RNA extraction. All procedures were performed on ice under RNase-free conditions. RNA was isolated and quantified, using TRIzol reagent or an equivalent column-based kit according to the manufacturer’s instructions, and reverse transcription was performed using 5 × SuperMix with gDNA Remover (1 μg total RNA in a 20 μL reaction) at 25 °C for 5 min, 50 °C for 30 min, and 85 °C for 5 min. RNA purity and integrity were verified before cDNA synthesis (A260/280 = 1.8–2.1; RIN ≥ 7). Quantitative PCR was carried out with 2 × SYBR Green Master Mix and gene-specific primers using the following program: 95 °C for 30 s, followed by 40 cycles of 95 °C for 15 s and 60 °C for 30 s. Each sample was run in technical triplicate; no-template and no-RT controls were included. A melting-curve analysis confirmed a single peak, and primer efficiencies were 90%–110% (R2 ≥ 0.99). Relative mRNA expression levels were calculated using the 2⁻ΔΔCt method, with GAPDH as the internal control. GAPDH stability across groups was verified prior to analysis, and results are presented as fold change relative to the control group (Table 1). Table 1Parameters of GAPDH, BCL—2 and NLRP3 primers.Primer nameSequence (5’—> 3’)LengthTmLocationGAPDH Forward PrimerACGGGAAACCCATCACCATC2060.03278–297GAPDH Reverse PrimerCTCGTGGTTCACACCCATCA2059.97483–464BCL-2 Forward PrimerAAGCCGGGAGAACAGGGTAT2060.62242–261BCL-2 Reverse PrimerGTATCCCACTCGTAGCCCCT2060.47329–310NLRP3 Forward PrimerTGCATGCCGTATCTGGTTGT2060.0436–55NLRP3 Reverse PrimerACCTCTTGCGAGGGTCTTTG2059.96426–407 Parameters of GAPDH, BCL—2 and NLRP3 primers. ELISA for detection of retinal inflammatory cytokine concentrations At the end of the experiment, retinas were dissected from enucleated eyes, weighed, and homogenized in ice-cold PBS (pH 7.4; typically 1:9, w/v). Homogenates were centrifuged at 2,000–3,000 rpm for 20 min at 4 °C, and the supernatants were collected. Total protein concentration was determined by BCA assay. A protease inhibitor cocktail was added to the homogenization buffer, and all steps were performed on ice to minimize protein degradation. Retinal levels of IL-1β, IL-6 and TNF-α were measured using commercial ELISA kits (Cloud-Clone, Wuhan, China) according to the manufacturer’s instructions. Briefly, standards and appropriately diluted samples were added to antibody-coated wells, followed by incubation, washing, addition of HRP-conjugated reagents and chromogenic substrates, and termination with stop solution. Absorbance at 450 nm was read with a microplate reader, and cytokine concentrations were calculated from standard curves. Samples were diluted to fall within the linear range; standard curves were fitted using a four-parameter logistic model (R2 ≥ 0.99), and blank-corrected absorbance values were used for quantification. All samples were assayed in duplicate, with at least three biological replicates per group, and results are expressed as pg/mg protein. Statistical analysis of experimental data Data are shown as mean ± SEM, with individual data points plotted where appropriate. Normality was checked with the Shapiro–Wilk test and homogeneity of variances with Levene’s test. One-way ANOVA and Tukey’s post hoc test were used for multiple group comparisons. If assumptions were violated, Welch’s ANOVA with Games–Howell post hoc or Kruskal–Wallis with Dunn’s post hoc was applied. A two-tailed p-value < 0.05 indicated statistical significance. Analysis was done with GraphPad Prism 10.0. Unless otherwise specified, n denotes the number of animals; when both eyes from the same rat were included, values were averaged per animal prior to group comparisons. Results Conformational and Rheological Properties of RADA16 hydrogels To study the 3D microstructure and rheological properties of RADA16 hydrogels for intravitreal use, three concentrations (0.5% (w/v), 1.0%, and 2.0% wt) were examined using SEM and rheological tests. SEM showed a porous network that became denser with higher concentrations. Rheological tests indicated gel-like behavior with the storage modulus (G′) higher than the loss modulus (G″). Using a Poisson’s ratio of 0.5, the Young’s modulus was calculated as 30.18 Pa for 0.5% (w/v), 16.21 Pa for 1.0%, and 26.66 Pa for 2.0% hydrogels. Although the modulus did not increase linearly with concentration, the 1.0% hydrogel showed the lowest stiffness among the three formulations, suggesting that factors such as network homogeneity and fiber packing may influence the mechanical behavior in addition to peptide concentration alone. Therefore, the 1.0% hydrogel was selected as the preferred formulation for intravitreal injection, as its viscoelastic properties are closer to the biomechanical range considered suitable for hydrogel-based vitreous substitutes19 (Fig. 1).Fig. 1Morphology characterization and oscilatory theology analysis of peptide-hydrogels.(a-f) HR-TEM images, respectively, for 0.5,1.0, and 2.0 wt% hydrogel from left to right. (g-i) The curves of storage modulus (G’) and loss modulus (G”) of hydrogels with concentrations of 0.5% wt (d), 1.0% wt(e),and 2.0% wt (f) from left to right, (j) The Young’s modulus of the 0.5, 1.0, and 2.0%wt hydrogels were 30.18 Pa、16.218 Pa、26.655 Pa, respectively. (k) Molecular structure of RADA16. Morphology characterization and oscilatory theology analysis of peptide-hydrogels.(a-f) HR-TEM images, respectively, for 0.5,1.0, and 2.0 wt% hydrogel from left to right. (g-i) The curves of storage modulus (G’) and loss modulus (G”) of hydrogels with concentrations of 0.5% wt (d), 1.0% wt(e),and 2.0% wt (f) from left to right, (j) The Young’s modulus of the 0.5, 1.0, and 2.0%wt hydrogels were 30.18 Pa、16.218 Pa、26.655 Pa, respectively. (k) Molecular structure of RADA16. FTIR spectroscopic characterization of hydrogel formulations The chemical composition and structural integrity of the hydrogel system were evaluated by FTIR spectroscopy. The spectrum of the 1% (w/v) RADA16 hydrogel (Fig. 2a) showed characteristic absorption bands at 3412.86, 2926.07, 2854.26, 1637.59, and 1540.95 cm⁻1, corresponding to N–H/O–H stretching, C-H stretching, and peptide-associated amide vibrations. Additional bands at 1452.33, 1384.43, 1202.97, and 1138.79 cm⁻1 were assigned to C-H bending and C-N/C-O stretching, confirming the formation of a chemically intact RADA16 hydrogel network.Fig. 2(a) FTIR spectrum of RADA16 peptide hydrogel.FTIR absorption spectrum of 1% RADA16 hydrogel showing characteristic peaks: 3412.86 cm⁻1 (N–H/O–H stretch), 2926.07/2854.26 cm⁻1 (C-H stretch), 1637.59 cm⁻1 (C = C, benzene ring), 1540.95 cm⁻1 (amide II), 1452.33/1384.43 cm⁻1 (C-H bend), 1202.97/1138.79 cm⁻1 (C-N/C-O stretch), and 802.27–624.52 cm⁻1 (long-chain CH₂ rocking). Peaks confirm RADA16 chemical integrity;(b): FTIR spectroscopic analysis of RADA16 and RADA16-CA hydrogels.(a) FTIR spectrum of 1% (w/v) RADA16 peptide hydrogel showing characteristic vibrational bands: 3412.86 cm⁻1 (N–H/O–H stretch), 2926.07/2854.26 cm⁻1 (C-H stretch), 1637.59 cm⁻1 (C = C, benzene ring), 1540.95 cm⁻1 (amide II), 1452.33/1384.43 cm⁻1 (C-H bend), 1202.97/1138.79 cm⁻1 (C-N/C-O stretch), and 802.27–624.52 cm⁻1 (long-chain CH₂ rocking). (b) FTIR spectrum of RADA16-CA composite hydrogel displaying additional features: split N–H/O–H peaks at 3414.57/3283.43 cm⁻1, amide I band at 1630.76 cm⁻1, amide II band at 1545.67 cm⁻1, amide III band at 1291.73 cm⁻1, and enhanced C-O stretching region (1181.14–1041.38 cm⁻1) indicating successful CA incorporation. Comparison of (a) and (b) demonstrates that CA loading preserves the RADA16 hydrogel network integrity while introducing characteristic carnosic acid functional groups. (a) FTIR spectrum of RADA16 peptide hydrogel.FTIR absorption spectrum of 1% RADA16 hydrogel showing characteristic peaks: 3412.86 cm⁻1 (N–H/O–H stretch), 2926.07/2854.26 cm⁻1 (C-H stretch), 1637.59 cm⁻1 (C = C, benzene ring), 1540.95 cm⁻1 (amide II), 1452.33/1384.43 cm⁻1 (C-H bend), 1202.97/1138.79 cm⁻1 (C-N/C-O stretch), and 802.27–624.52 cm⁻1 (long-chain CH₂ rocking). Peaks confirm RADA16 chemical integrity;(b): FTIR spectroscopic analysis of RADA16 and RADA16-CA hydrogels.(a) FTIR spectrum of 1% (w/v) RADA16 peptide hydrogel showing characteristic vibrational bands: 3412.86 cm⁻1 (N–H/O–H stretch), 2926.07/2854.26 cm⁻1 (C-H stretch), 1637.59 cm⁻1 (C = C, benzene ring), 1540.95 cm⁻1 (amide II), 1452.33/1384.43 cm⁻1 (C-H bend), 1202.97/1138.79 cm⁻1 (C-N/C-O stretch), and 802.27–624.52 cm⁻1 (long-chain CH₂ rocking). (b) FTIR spectrum of RADA16-CA composite hydrogel displaying additional features: split N–H/O–H peaks at 3414.57/3283.43 cm⁻1, amide I band at 1630.76 cm⁻1, amide II band at 1545.67 cm⁻1, amide III band at 1291.73 cm⁻1, and enhanced C-O stretching region (1181.14–1041.38 cm⁻1) indicating successful CA incorporation. Comparison of (a) and (b) demonstrates that CA loading preserves the RADA16 hydrogel network integrity while introducing characteristic carnosic acid functional groups. After incorporation of carnosic acid (CA), the FTIR spectrum of the RADA16-CA hydrogel (Fig. 2b) retained the principal peptide-associated bands while exhibiting several distinct spectral changes. In particular, peak splitting in the N–H/O–H stretching region (3414.57 and 3283.43 cm⁻1), a shift in the amide region, and the appearance of an amide III band at 1291.73 cm⁻1 suggested that CA loading altered the local intermolecular environment without disrupting the overall peptide framework. Moreover, enhanced signals in the 1200–1000 cm⁻1 region, including prominent peaks at 1181.14 and 1041.38 cm⁻1, were consistent with CA-associated C-O stretching vibrations, supporting successful incorporation of the drug into the hydrogel matrix. Importantly, the characteristic peptide-related bands remained preserved after CA loading, indicating that the self-assembled structure of RADA16 was maintained. Together, these findings demonstrate that CA was successfully incorporated into the RADA16 hydrogel, with spectral features consistent with intermolecular interactions, yielding a structurally stable composite formulation suitable for sustained drug delivery. Sustained release, drug permeability, and pH analysis of RADA16 hydrogels and CA In Fig. 3a, 19.4 mg of CA was added to 1 mL of 1% (w/v) hydrogel and incubated for 48 h. After dilution with DMSO, HPLC analysis showed a concentration of 14.07 mg/mL, with 5.33 mg of CA loaded and an encapsulation efficiency of 27.75% (w/v). For 2% (w/v) hydrogel, 38.8 mg of CA was used, resulting in a concentration of 31.6 mg/mL, 7.2 mg of CA loaded, and an encapsulation efficiency of 18.56%. For the 0.5% (w/v) hydrogel (1 mL), 9.7 mg of CA was added and incubated for 48 h. After dilution with DMSO, HPLC analysis showed a concentration of 8.1 mg/mL, with 1.6 mg of CA loaded and an encapsulation efficiency of 16.49%. As depicted in Fig. 3b, the 1% (w/v) RADA16-CA system reached its peak concentration (141–161 μg/mL) within 4–12 h, then stabilized. Figure 3c shows that both free CA and the 1% RADA16-CA hydrogel maintained near-neutral pH values of 7.06 and 6.95, respectively, indicating favorable physicochemical compatibility for intravitreal injection. Figure 3d illustrates that the RADA16-CA sustained-release system ensures continuous CA delivery and retention in the cytoplasm for up to 72 h.Cellular fluorescence microscopy showed that, unlike the free CA group, which lacked green fluorescence in the cytoplasm, the RADA16-CA group exhibited distinct green signals. This confirms the system’s effectiveness in delivering and retaining CA in the cytoplasm.Fig. 3Characterization of the RADA 16-CA sustained-release system. (a) Photos of the porous RADA 16 hydrogel loaded with Carnosic Acid(CA).(b) Drug release kinetics of the RADA16-CA sustained-relea-se system.(c) pH values of free CA and 1% RADA16-CA.(d) Cellular immunofluore-scence observatio-n of Carnosic Acid(CA) intracellular retention, scale bar: 100 μm. The CA drug is green fluorescent, and the nucleus is blue fluorescent (DAPI). Characterization of the RADA 16-CA sustained-release system. (a) Photos of the porous RADA 16 hydrogel loaded with Carnosic Acid(CA).(b) Drug release kinetics of the RADA16-CA sustained-relea-se system.(c) pH values of free CA and 1% RADA16-CA.(d) Cellular immunofluore-scence observatio-n of Carnosic Acid(CA) intracellular retention, scale bar: 100 μm. The CA drug is green fluorescent, and the nucleus is blue fluorescent (DAPI). CCK-8 cytotoxicity and cell viability assays As shown in Fig. 4a–c, the effects of CA, 1% (w/v) RADA16 hydrogel and the RADA16–CA sustained-release system on RGC-5 cell proliferation were assessed by CCK-8 after 24 h. In the CA group, cell viability remained stable at 0–128 μg/mL but decreased significantly at concentrations ≥ 256 μg/mL. In contrast, the RADA16 group showed stable viability at 0–60 μg/mL, with a significant increase at 120–960 μg/mL. The RADA16–CA group exhibited stable viability at 0–32 μg/mL, followed by a significant increase at 64–512 μg/mL. As illustrated in Fig. 4d–f and quantified in Fig. 4, live/dead staining at 512 μg/mL after 24 and 48 h revealed marked cell death in the CA group at 24 h, indicating pronounced cytotoxicity. The 1% (w/v) RADA16 group significantly improved RGC-5 cell viability compared with the control (p < 0.001), whereas the RADA16–CA group showed an even stronger pro-survival effect (p < 0.0001). At 48 h, cell activity slightly declined in all groups; CA remained highly cytotoxic (p < 0.0001), while 1% (w/v) RADA16 (p < 0.05) and RADA16–CA (p < 0.0001) continued to enhance cell survival relative to the control. Overall, as shown in Fig. 4, high-dose CA (512 μg/mL) was cytotoxic to RGC-5 cells, whereas 1% (w/v) RADA16 supported cell proliferation, and the RADA16–CA sustained-release system at the same CA concentration (512 μg/mL) exerted a more pronounced pro-proliferative effect (Fig. 4).Fig. 4Evaluation of RGC-5 cell viability and protective effects. (a-c) Cell viability of retinal ganglion cells (RGC-5) treated for 24 h with gradient concentrations of (a) Carnosic Acid (CA), (b) 1% RADA 16 hydrogel, and (c) the RADA 16-CA sustained-release system, as assessed by CCK-8 assay. (d-e) Representative live/dead staining images of RGC-5 cells after respective treatments, scale bar: 100 μm. Live cells are stained green (calcein-AM), and dead cells are stained red (propidium iodide). (f)Time-dependent effects of different treatments on retinal ganglion cell (RGC-5) viability.Statistics: All quantitative data were analyzed using one-way ANOVA with Tukey’s post hoc test, except for panel (f) which used two way ANOVA with Sidak’s post hoc test. Significance: ***p < 0.001, **p < 0.01, *p < 0.05, ns = not significant. Evaluation of RGC-5 cell viability and protective effects. (a-c) Cell viability of retinal ganglion cells (RGC-5) treated for 24 h with gradient concentrations of (a) Carnosic Acid (CA), (b) 1% RADA 16 hydrogel, and (c) the RADA 16-CA sustained-release system, as assessed by CCK-8 assay. (d-e) Representative live/dead staining images of RGC-5 cells after respective treatments, scale bar: 100 μm. Live cells are stained green (calcein-AM), and dead cells are stained red (propidium iodide). (f)Time-dependent effects of different treatments on retinal ganglion cell (RGC-5) viability.Statistics: All quantitative data were analyzed using one-way ANOVA with Tukey’s post hoc test, except for panel (f) which used two way ANOVA with Sidak’s post hoc test. Significance: ***p < 0.001, **p < 0.01, *p < 0.05, ns = not significant. Acute ocular hypertension model establishment, intravitreal administration, and IOP results After induction, an ophthalmic exam showed vein congestion and dilation, indicating successful model establishment. Baseline intraocular pressure (IOP) was measured, and a drug was injected into the eye on Day 1 post-surgery. IOP was monitored daily for 7 days. Before induction, IOP levels were similar across groups. From Days 1 to 7, the model group’s IOP averaged 31.5 mmHg, significantly higher than the control group’s 14.4 mmHg, confirming model success. RADA16-CA reduced IOP to 17.0 mmHg, while RADA16 alone had no effect. CA alone also lowered IOP to 20.5 mmHg. RADA16-CA’s effect was greater than CA alone (Fig.5). Ocular phenotype after glaucoma surgery modeling and changes in intraocular pressure(IOP). (a-b) Representative images showing the ocular phenotype in mice following glaucoma surgery modeling. Note the congested and dilated proximal veins (arrow) and the blanched distal veins (arrowhead) after the modeling procedure. (c-d) The line chart and bar graph showing the changes in mean intraocular pressure (IOP) of mice in each group before modeling (Day 0) and on Days 1, 3, 5, and 7 after modeling(n = 6 per group).Statistical analysis: Two-way analysis of variance (two-way ANOVA) was used to evaluate the effects of treatment, time, and their interaction on intraocular pressure, followed by Sidak’s post hoc test for pairwise comparisons between groups. Significance levels were indicated as: ****p < 0.0001, ***p < 0.001, **p < 0.01. Retinal RNFL thickness, ROS DHE staining, and RGC immunofluorescence in rats In the control retina (NC), RGCs had normal morphology and the RNFL was structurally regular and of normal thickness. In the elevated IOP model (Model), RGCs were swollen and deformed, and the RNFL was edematous and significantly thicker. The RADA16 intervention (Model + RADA16) did not improve RGC swelling or RNFL thickness compared to the Model group. However, in the CA intervention (Model + CA), RGC morphology improved significantly with reduced edema, and RNFL thickness decreased. In the RADA16-mediated sustained-release CA group (Model + RADA16-CA), some RGCs showed apoptotic features (less than in the Model group), and RNFL thickness was significantly reduced, indicating the greatest structural improvement. Quantitative analysis showed that the NC group had the thinnest RNFL, while the Model group had a significant increase (p < 0.0001 vs. NC). The Model + RADA16 group was similar to the Model group (p > 0.05). Both Model + CA and Model + RADA16-CA groups had significantly reduced RNFL thickness compared to the Model group (both p < 0.0001), with the Model + RADA16-CA group showing even lower thickness than the Model + CA group (p < 0.05), indicating CA’s effectiveness in reducing edema. The RADA16-CA sustained-release system further decreased RNFL thickness (Fig. 6a-b).Fig. 6Histological evaluation, oxidative stress assessment, and retinal ganglion cell (RGC) survival analysis. (a-b) Representative images of H&E staining of retinal sections and quantit ative analysis of retinal nerve fiber layer (RNFL) thickness. (c-d) Representative images of DHE staining showing superoxide production in retinal sections and its quantitative analy-sis.(e–f) Representative images of immunohistochemical staining for RGC-specific markers and quantitative analysis of surviving RGCs.For all quantitative analyses, statistical significance was determined by one-way analysis of variance (ANOVA) followed by Tukey’s multiple comparison test, with significance levels indicated as: ***p < 0.001, **p < 0.01, *p < 0.05, and “ns” indicating no significant difference. Histological evaluation, oxidative stress assessment, and retinal ganglion cell (RGC) survival analysis. (a-b) Representative images of H&E staining of retinal sections and quantit ative analysis of retinal nerve fiber layer (RNFL) thickness. (c-d) Representative images of DHE staining showing superoxide production in retinal sections and its quantitative analy-sis.(e–f) Representative images of immunohistochemical staining for RGC-specific markers and quantitative analysis of surviving RGCs.For all quantitative analyses, statistical significance was determined by one-way analysis of variance (ANOVA) followed by Tukey’s multiple comparison test, with significance levels indicated as: ***p < 0.001, **p < 0.01, *p < 0.05, and “ns” indicating no significant difference. Panels c–d show DHE staining in retinal sections. Images were acquired at 200 × magnification to maximize tissue coverage. ROS accumulation was quantified in three fields per section using ImageJ (version 1.53) as relative fluorescence intensity rather than cell density. The Model group exhibited significantly higher ROS fluorescence intensity than the NC group (p < 0.0001).The Model + RADA16 group did not differ significantly from the Model group (p > 0.05). In contrast, the Model + CA group showed lower ROS fluorescence intensity than the Model group (p < 0.01), and the Model + RADA16-CA group exhibited the lowest ROS fluorescence intensity among all treatment groups (p < 0.05 versus Model + CA), indicating the strongest suppression of oxidative stress.Panels e–f illustrate Brn3a-positive RGC counts. RGCs were imaged at 200 × magnification, and the number of surviving RGCs per field was quantified using ImageJ (version 1.53).The NC group had the highest RGC counts, whereas the Model group showed a marked reduction (p < 0.0001). Both the Model + RADA16 group (p < 0.01) and the Model + CA group (p < 0.001) demonstrated higher RGC counts compared with the Model group, suggesting protective effects of RADA16 hydrogel and CA, respectively. The Model + RADA16-CA group showed the greatest preservation of RGCs (p < 0.0001 versus Model) and outperformed the Model + CA group (p < 0.05), supporting an additive neuroprotective effect of CA when delivered via the RADA16 sustained-release system. Results: oxidative stress, inflammatory cytokines, and molecular/protein analyses. Oxidative stress assessment Oxidative stress was evaluated by measuring retinal SOD activity and MDA content, where reduced SOD activity and elevated MDA levels indicate increased oxidative damage (Fig. 7a-b). SOD activity was highest in the NC group, whereas the Model group showed a significant decrease (p < 0.0001). The Model + RADA16 group did not differ significantly from the Model group (p > 0.05). In contrast, the Model + CA group exhibited a significant increase in SOD activity compared with the Model group (p < 0.05), and the Model + RADA16-CA group showed an even greater increase (p < 0.05). For MDA, the Model group had the highest levels and the NC group the lowest (p < 0.0001). The Model + RADA16 group was not significantly different from the Model group (p > 0.05), whereas the Model + CA group had significantly lower MDA levels than the Model group (p < 0.05), with a further significant reduction in the Model + RADA16-CA group (p < 0.01). These findings indicate that CA attenuates oxidative stress–induced damage, and that the RADA16-CA sustained-release system provides superior antioxidative efficacy.Fig. 7Analysis of oxidative stress, inflammatory cytokine levels, and gene/protein expression. (a-b) Assessment of oxidative stress levels. The activities of antioxidant enzyme superoxide dismutase(SOD) and the content of lipid peroxidation product malondialdehyde (MDA) were measured in retinal tissues. (c-e) Evaluation of pro-inflammatory cytokine expression. The protein levels of interleukin-1β (IL-1β), interleukin-6 (IL-6), and tumor necrosis factor-α (TNF-α) in retinal tissues or serum were quantified by enzyme-linked immunosorbent assay (ELISA). (f-g i-j) Analysis of mRNA and protein expression. The expression levels of target genes were detected by quantitative real-time PCR (qPCR).(h)the corresponding protein levels w-ere confirmed by western blot (WB) analysis. All quantitative data are presented as mean ± standard deviation (mean ± SD). Comparisons between groups were performed using one-way analysis of variance (one-way ANOVA) followed by Tukey’s post hoc test. *p < 0.05, **p < 0.01, and ***p < 0.001 indicate statistically significant differences between groups. Analysis of oxidative stress, inflammatory cytokine levels, and gene/protein expression. (a-b) Assessment of oxidative stress levels. The activities of antioxidant enzyme superoxide dismutase(SOD) and the content of lipid peroxidation product malondialdehyde (MDA) were measured in retinal tissues. (c-e) Evaluation of pro-inflammatory cytokine expression. The protein levels of interleukin-1β (IL-1β), interleukin-6 (IL-6), and tumor necrosis factor-α (TNF-α) in retinal tissues or serum were quantified by enzyme-linked immunosorbent assay (ELISA). (f-g i-j) Analysis of mRNA and protein expression. The expression levels of target genes were detected by quantitative real-time PCR (qPCR).(h)the corresponding protein levels w-ere confirmed by western blot (WB) analysis. All quantitative data are presented as mean ± standard deviation (mean ± SD). Comparisons between groups were performed using one-way analysis of variance (one-way ANOVA) followed by Tukey’s post hoc test. *p < 0.05, **p < 0.01, and ***p < 0.001 indicate statistically significant differences between groups. Inflammatory cytokine detection Glaucoma-related inflammation was assessed by measuring retinal levels of the pro-inflammatory cytokines IL-1β, IL-6 and TNF-α (Fig. 7c–e). Compared with the NC group, cytokine concentrations were markedly increased in the Model group (p < 0.0001). The Model + RADA16 group did not differ significantly from the Model group. In contrast, the Model + CA group showed significantly lower levels of IL-1β, IL-6 and TNF-α than the Model group (p < 0.01), and the Model + RADA16-CA group exhibited an even greater reduction (p < 0.05). These data suggest that CA suppresses inflammatory cytokine production, and that incorporation into the RADA16 hydrogel further enhances this anti-inflammatory effect. Molecular-level qPCR analysis To investigate molecular mechanisms underlying neuroprotection and inflammation, retinal mRNA expression of the anti-apoptotic gene BCL2 and the inflammasome component NLRP3 was quantified by qPCR (Fig. 7d-g). BCL2 expression was significantly lower in the Model group than in the NC group (p < 0.0001), indicating reduced anti-apoptotic capacity. No significant difference was observed between the Model and Model + RADA16 groups. In the Model + CA group, BCL2 expression was significantly higher than in the Model group (p < 0.05), and the Model + RADA16-CA group showed a further increase compared with the Model + CA group (p < 0.05). NLRP3 expression was significantly elevated in the Model group relative to the NC group (p < 0.0001), reflecting enhanced inflammasome activation. Again, the Model + RADA16 group did not differ significantly from the Model group, whereas the Model + CA group showed a significant reduction in NLRP3 expression compared with the Model group, with an even greater inhibitory effect in the Model + RADA16-CA group. Taken together, these results indicate that CA modulates the BCL2/NLRP3 axis by upregulating anti-apoptotic BCL2 and downregulating pro-inflammatory NLRP3, and that the RADA16-CA system further amplifies this molecular protection. Protein-level western blot analysis Protein levels of BCL-2 and NLRP3 were further examined by Western blotting (Fig. 7h–j). Consistent with the qPCR results, BCL-2 protein expression was significantly lower in the Model group than in the NC group (p < 0.0001). The Model + RADA16 group showed no significant change compared with the Model group. In contrast, BCL-2 levels were significantly higher in the Model + CA group than in the Model group (p < 0.05), and the Model + RADA16-CA group exhibited a further increase compared with the Model + CA group (p < 0.01). NLRP3 protein expression was markedly increased in the Model group relative to the NC group (p < 0.0001). The Model + RADA16 group did not differ significantly from the Model group, whereas the Model + CA group showed significantly lower NLRP3 levels (p < 0.001). The greatest reduction in NLRP3 expression was observed in the Model + RADA16-CA group (p < 0.01 versus Model + CA). These protein-level findings corroborate the transcriptional data and further support the notion that the RADA16-CA sustained-release system enhances CA-mediated neuroprotection by coordinately regulating BCL-2 and NLRP3. Discussion This study presents the development and characterization of an injectable RADA16 peptide hydrogel system for sustained intravitreal delivery of carnosic acid (CA) in glaucoma treatment. The 1% (w/v) RADA16 hydrogel demonstrates optimal mechanical properties for intravitreal injection, with a Young’s modulus of 16.22 Pa that provides sufficient structural integrity while minimizing mechanical stress on delicate retinal tissues. The hydrogel’s ECM-mimetic nanofibrillar network (see Fig. 1) supports cellular interactions and enables sustained drug release over 72 h, addressing the major limitation of free CA’s rapid clearance from the vitreous humor. Beyond its biomechanical suitability, the molecular compatibility between the hydrogel matrix and CA is also critical for sustained intraocular delivery. The FTIR data further support the structural compatibility of the RADA16 hydrogel as a carrier for CA. In the spectrum of blank RADA16 hydrogel, the amide II band at 1540.95 cm⁻1 together with the broad N–H/O–H stretching band at 3412.86 cm⁻1, indicative of an extensive hydrogen-bonding network, represent characteristic signatures of the self-assembled peptide structure and are consistent with previously reported spectra of RADA16 and other ionic-complementary peptide hydrogels20. CA contains catechol and polyphenolic hydroxyl groups, which are closely associated with its antioxidant activity21. When CA is incorporated into the RADA16 matrix, drug-peptide interactions are expected to influence loading efficiency, release kinetics, and overall formulation stability22. In the present study, FTIR analysis showed splitting of the N–H/O–H stretching region after CA loading, suggesting the formation of hydrogen-bonding interactions between the phenolic hydroxyl groups of CA and the amide moieties of the peptide network. Such non-covalent interactions are likely to promote stable dispersion of the drug within the hydrogel matrix and provide a structural basis for controlled release: moderate hydrogen bonding may reduce burst release while still allowing gradual dissociation under physiological conditions, thereby supporting sustained drug release22. Importantly, the amide I, II, and III bands remained clearly detectable after drug loading, without marked disruption of the characteristic peptide-associated features, indicating that the β-sheet secondary structure of RADA16 was largely preserved23. This structural preservation is particularly important because the β-sheet architecture is central to the mechanical strength, self-healing behavior, and overall stability of the hydrogel. Disruption of this conformation would be expected to weaken the gel network or even cause structural collapse, thereby compromising the rheological properties required for intravitreal injection24. Taken together, the spectroscopic findings suggest that CA is incorporated into the RADA16 hydrogel mainly through physical encapsulation rather than by insertion that disrupts inter-sheet β-sheet stacking. This is consistent with a physical loading mode that preserves the functional integrity of the drug-delivery system. Drawing on previous studies, we selected a 24-h incubation period for the CCK-8 assay instead of extending it to 72 h (consistent with the duration of drug release), for the following reasons. First, 24 h is sufficient to evaluate acute cytotoxicity and the initial therapeutic response, which meets the requirements for safety assessment. For RGC-5 cells, the standard protocol widely reported in the literature involves seeding the cells in 96-well plates at a density of 5 × 103 cells per well, followed by a 24-h adherence and recovery period, and subsequent incubation with CCK-8 reagent for 2 h25. Second, prolonging the incubation time to 48–72 h in the CCK-8 assay may introduce significant interference due to cell proliferation and medium depletion, thereby reducing the comparability of the results. In contrast, the 72-h drug release test is designed to simulate the sustained administration window required in clinical practice, which is more aligned with the demand for long-term exposure in the management of chronic diseases. RADA16 demonstrates tissue-repair properties26–28, lipophilic drug-loading ability29, and anti-biofilm activity30, making it an ideal hydrogel carrier for CA in glaucoma-related retinal ganglion cell (RGC) protection. This study further confirms that RADA16 hydrogel supports the growth of RGC-5 cells and primary rat RGCs, showcasing excellent biocompatibility and sustained myristic acid loading and release. The 1% (w/v) RADA16 alone showed a surprising pro-survival effect on RGC-5 cells,which may be attributed to the peptide’s inherent bioactivity and ECM-mimetic properties that provide a supportive microenvironment. As a natural phenolic diterpenoid, CA significantly reduces key inflammatory mediators such as IL-1β, IL-6, and TNF-α. In a rat model of aflatoxin B1-induced toxicity, CA decreased IL-6, IL-1β, and TNF-α levels to control-group equivalents, supporting its anti-inflammatory properties6,31. Elevated levels of these cytokines in the aqueous humor and tear fluid of various glaucoma types indicate inflammation’s role in disease onset and progression, underscoring the potential of anti-inflammatory strategies32. Our acute ocular hypertension model showed significantly increased IL-1β, IL-6, and TNF-α levels, which CA reduced, protecting RGCs. The RADA16-CA system exhibited superior therapeutic efficacy, likely due to RADA16’s ability to maintain stable intraocular CA concentrations for over 72 h (see Fig. 3b), providing sustained cytokine inhibition, unlike the transient effect of free CA. Patients with glaucoma exhibit a significant imbalance in oxidative stress. Multiple independent clinical studies consistently show that, compared with healthy controls, superoxide dismutase (SOD) activity is decreased and malondialdehyde (MDA) levels are increased in both peripheral blood and aqueous humor33–35. These findings corroborate the presence of systemic oxidative stress and implicate these markers in disease pathophysiology. Consistent with the clinical observations, the same trend—reduced SOD activity and elevated MDA levels—has been observed in retinal tissues of rat glaucoma models36–38. In diabetic retinopathy rat models and MDI-induced adipocyte oxidative stress models, carnosic acid (CA) exhibits dual antioxidant actions by directly suppressing MDA production39 and upregulating endogenous antioxidant enzymes (e.g., SOD)40. In the acute ocular hypertension rat model of this study, CA reversed the decline in SOD activity and the elevation in MDA levels, thereby alleviating retinal tissue injury. Moreover, leveraging the sustained-release property of RADA16, the RADA16–CA system conferred stronger protection than free CA by prolonging intraocular retention, reducing rapid clearance, and thereby sustaining SOD activity while lowering MDA levels. Alterations in Bcl-2 expression directly influence apoptosis. Specifically, the Bcl-2 family regulates mitochondrial pathway–mediated apoptosis, and its dysregulation is associated with retinal ganglion cell (RGC) axonal degeneration and neurodegeneration41. In ocular tissues, the NLRP3 inflammasome is expressed and, upon activation, promotes the maturation and release of interleukin-1β (IL-1β). IL-1β is elevated in the peripheral blood of glaucoma patients, contributing to neurotoxic inflammation that leads to axonal degeneration and RGC death42. Consistent with these pathways, carnosic acid (CA) has been shown to upregulate Bcl-2 and inhibit NLRP3 activation, conferring protective effects in high-fat diet–induced brain injury and subarachnoid hemorrhage models43. In the acute ocular hypertension model used in this study, CA attenuated retinal damage by upregulating Bcl-2 and downregulating NLRP3 expression. Compared with free CA, the RADA16-CA sustained-release system was associated with stronger modulation of the Bcl-2/NLRP3 axis, as reflected by higher Bcl-2 expression and lower NLRP3 expression in retinal tissue. This enhanced effect may be related, at least in part, to the sustained-release behavior of the hydrogel, which likely prolonged intraocular drug exposure and improved the overall neuroprotective efficacy of CA. Consequently, it partially bridged the “therapeutic gap” caused by the limited bioavailability of free CA (see Figs. 3d and 6e–f). It should be noted that our qPCR and western blot (WB) data are correlative, reflecting transcriptional and protein changes rather than establishing causality. Previous studies have shown that Bcl‑2 overexpression can indirectly inhibit NLRP3 activation and alleviate inflammation by stabilizing mitochondrial function, suppressing ROS production, and regulating autophagy44–46. Consistent with this, in a human retinal endothelial cell model, Bcl-2 overexpression attenuated NLRP3 activation and conferred protection47. supporting involvement of this axis in our observations. Future studies should incorporate gain- and loss-of-function and pathway validation experiments to confirm direct regulation within the Bcl-2/NLRP3 axis. The enhanced therapeutic efficacy of RADA16-CA over CA alone can be attributed to several factors. First, the sustained-release profile maintains therapeutic drug concentrations for extended periods, preventing the rapid decline seen with free drug administration. Second, the hydrogel provides mechanical support and creates a favorable microenvironment for RGC survival. Third, the peptide hydrogel may enhance drug penetration into retinal tissues and improve cellular uptake, as demonstrated in Fig. 3d. These advantages position the RADA16-CA system as a promising advancement over conventional free drug administration. Compared with other drug delivery systems for glaucoma treatment, our RADA16-CA system demonstrates distinct advantages. Unlike polymeric nanoparticles, the RADA16 hydrogel integrates more naturally with the vitreous matrix, supporting ocular biocompatibility and promoting sustained intravitreal retention and homogeneous drug distribution46. In contrast to liposomal formulations with often limited and unstable drug-loading capacity48, peptide hydrogels such as RADA16 offer high biocompatibility, more stable drug loading, structural stability, tissue-supportive properties, and reduced toxicity. Furthermore, our system is injectable, enabling minimally invasive and repeatable drug administration, in contrast to surgically implanted devices that necessitate operating-room procedures. Taken together, these features underscore the translational potential of the RADA16–CA platform for glaucoma therapy.This study has several limitations that warrant acknowledgment. First, the Young’s modulus of the vitreous humor varies with age and interindividual factors, and reliable quantitative datasets for human and animal vitreous are lacking. The Young’s modulus of the RADA16 hydrogel used in this study (approximately 16–30 Pa) does not precisely match that of the native vitreous humor. This low modulus facilitates injection and conformability, but its long-term structural integrity and retention time in the vitreous humor remain uncertain. Future in vivo biodegradation and in situ rheological assessments are needed to optimize the balance between mechanical properties and treatment duration. Second, atomic force microscopy (AFM) or microrheology methods were not used to evaluate the stiffness of the vitreous matrix, limiting insight into the time-dependent mechanical compatibility of RADA16-CA. Future work will apply AFM/microrheology for longitudinal, quantitative characterization. Third, the 72-h in vitro release test may not capture the dosing window required in chronic glaucoma. We will extend release monitoring to 14 days and pair it with in vivo vitreous/retinal drug quantification to more accurately depict sustained CA exposure and therapeutic levels. Fourth, this study used an acute intraocular hypertension model and lacked chronic-model and long-term efficacy data. Future studies will incorporate microbead occlusion and DBA/2 J genetic models with extended follow-up to enhance clinical relevance. Fifth, although FTIR analysis supported the structural integrity of the RADA16-CA composite and suggested intermolecular interactions between the hydrogel and CA, more advanced spectroscopic approaches such as NMR may be needed in future studies to further clarify the exact molecular binding mode. Sixth, this study lacked in vivo pharmacokinetic data, thereby limiting our ability to quantify retinal CA exposure over time; we will establish a pharmacokinetic–pharmacodynamic (PK–PD) relationship using LC–MS/MS quantification and aqueous humor/vitreous microdialysis. Seventh, functional evaluation of RGCs (such as pattern electroretinography [pattern ERG] or visual behavioral tests) was not conducted, hence whether morphological protection translates into functional benefit remains uncertain; future studies will include evaluations such as pattern ERG, optokinetic response (OKR), and pupillary light reflex (PLR). Eighth, although RGC-5 cells are widely used, they do not fully recapitulate human RGC biology; future work will include human induced pluripotent stem cell (iPSC)-derived RGCs to improve translational relevance. Despite these limitations, this biocompatible hydrogel system shows promise for reducing dosing frequency and thereby improving adherence and therapeutic outcomes. The translational potential of this system may extend beyond glaucoma to other retinal diseases requiring sustained intraocular drug delivery (e.g., age-related macular degeneration and diabetic retinopathy). Future work will prioritize optimization of the delivery platform by incorporating targeting ligands to enhance retinal cell specificity, tuning crosslinking density to modulate release kinetics and mechanical properties, and conducting validation in large-animal models with ocular anatomy more comparable to humans. Additionally, whether the hydrogel is remodeled by retinal glia (astrocytes, Müller cells) and integrates with endogenous extracellular matrix (ECM) components—with consequent effects on residence time and bioactivity—warrants investigation. Ultimately, the goal is to develop a clinically translatable sustained-release system that provides long-term neuroprotection and reduces the treatment burden for patients with glaucoma. Conclusions The 1% (w/v) RADA16 hydrogel is a biocompatible carrier that optimally matches the vitreous body’s mechanics, allowing for ≥ 72-h cytosolic delivery of CA and enhancing its antioxidant, anti-inflammatory, and neuroprotective effects in vitro and in vivo. RADA16 hydrogels form nanofiber networks with viscoelastic properties at 0.5% (w/v), 1% (w/v), and 2% (w/v) concentrations, with the 1% (w/v) formulation being ideal for intravitreal injection. This system overcomes free CA’s limitations, providing sustained release and efficient cytoplasmic delivery. In RGC-5 cells, 1% (w/v) RADA16 supports proliferation, while CA boosts antioxidant defense by reducing ROS and apoptosis, with the RADA16-CA combination showing superior efficacy at equivalent doses.In a SD rat model of acute ocular hypertension, 1% (w/v) RADA16 promoted RGC growth. CA provided protection by lowering IL-1β, IL-6, and TNF-α levels, boosting SOD activity, reducing MDA, and decreasing intraocular pressure, while increasing Bcl-2 expression and reducing NLRP3 inflammasome activity. The 1% (w/v) RADA16-CA system enhanced these neuroprotective effects. Supplementary Information Supplementary Information. Supplementary Information.