Light-induced Nrf2−/− mice as atrophic age-related macular degeneration model and treatment with nanoceria laden injectable hydrogel

Elevated oxidative stress and associated reactive oxygen species (ROS) accumulation are hallmarks in the induction and progression of age-related macular degeneration (AMD). By exposing nuclear factor erythroid 2-related factor (Nrf2) knockout (Nrf2−/−) mice to mild white light, we were able to generate a new dry-AMD like murine model to the study. This animal model developed phenotypes of photoreceptor degeneration, retinal function impairment, ROS accumulation, and inflammation reaction in a relatively shorter time. In the treatment of this animal model we utilized an antioxidative and water soluble nanoparticle known as glycol chitosan coated cerium oxide nanoparticles (GCCNP). The delivery of GCCNP protected retina against progressive retinal oxidative damage. Further combination of GCCNP with alginate-gelatin based injectable hydrogel provided synergistic antioxidant effects and achieved a more rapid recovery of the retinal pigment epithelium and photoreceptor cells. This combined treatment technique has the potential to translate into a clinical intervention for the treatment of AMD.


Result
Nrf2 −/− mice are susceptible to light-induced retinal lesions. Study has shown that the downregulation of Nrf2 expression leads to compromised regulation of oxidative stress, leaving the retina more prone to light induced oxidative damage 26 . To examine whether light exposure can lead to PRC or RPE cells atrophy, we exposed Nrf2 −/− mice on different intensity of light exposure at 4,500 lux for 2 h and 3 h, and 10,000 lux at 2 h. We found that Nrf2 −/− mice can be triggered by a 3 h, mild white light exposure (4,500 lux) at 2 months of age (P 2 m), resulting in permanent PRC/RPE impairment. At 10 weeks post-3 h exposure, histological studies showed that the outer nuclear layer (ONL) of the retina was thinned and nearly 50% PRC nuclei were absent after 3 h light exposure, indicating irreversible PRC degeneration. The structure of OS/IS were also found to be disrupted after light exposure (Fig. 1A). 2 h light exposure showed that ONL thickness appeared normal. The 10000 lux 2 h damage caused the most severe damage and the majority of ONL degenerated at the endpoint. The evident sign of damage or degeneration was absent from INL and GCL from histological sections, suggesting the major ROS elevation mainly exist in light-sensitive PRC.
The Funduscopy revealed retinal morphological changes after exposure to light (Fig. 1B). Apparently, 4500 lux 2 h light exposure caused lesions on the retina at 1 week post-2 h light exposure, however the lesions were recovered 10 weeks post exposure. 3 h light exposure induced more severe lesions as revealed by the increased accumulation of bright-yellow spot, and more lesion spots became visible with time passing. We measured the thickness change of the ONL post exposure by optical coherence tomography (OCT), and 3 h 4500 lux light could lead to GA-like phenotypes, such as degeneration of PRC and RPE. The naïve Nrf2 −/− mice have an average ONL thickness of approximately 75 µm at 6 weeks post-natal (Fig. 1C). 2 h 4,500 lux caused a ~20% reduction of ONL thickness (60 ± 10 µm) at the first week-posted light exposure, but the thickness of ONL was found recovered at the fifth week post light exposure (70 ± 8 µm). 3 h light exposure group had more severe damage than 2 h exposure at 4,500 lux on the first week (41 ± 5 µm, ~50% reduction in ONL thickness compared to naïve mice, P = 0.013). Although there was partial recovery after five weeks, the degeneration of ONL thickness could not be reversed (49 ± 6 µm, ~40% ONL thickness reduction compared to naïve mice, P = 0.004). The 2 h 10000 lux resulted in severe cataracts after light exposure, creating challenges to perform Fundus, OCT, and electroretinography (ERG). Therefore, we choose 4500 lux 3 h light exposure for further studies, where the damage mainly existed on ONL and RPE (Fig. 1A,B). the light-induced Nrf2 −/− mice have dry AMD-like retinal pathology, we performed fundus/OCT imaging, histology, TEM, and ERG analyses. We found that bright white-yellowish spots appeared on the Fundus images in the most light-exposed eyes (41 out of 50). OCT images suggested a drusen-like structures, which are located under RPE, disrupted the PRC and RPE structure. Retinal histology analysis identified drusen-like alterations of RPE as shown in Fig. 2A. However, most of the deposit under RPE was found in a basal linear formation, rather than a granular-shaped drusen as shown by others in advanced stage atrophic AMD 27 ( Fig. 2A). Other morphologic alterations include hemorrhage and vacuole, indicating the accumulated toxic products and cytokines responsible for RPE degeneration ( Fig. 2A). In addition, capillary changes can be found between choroid and RPE ( Fig. 2A), but there is no clear evidence that this model can generate choroidal neovascularization at the endpoint of experiments. TEM images displayed the accumulated autophagosome and lipofuscin in RPE (Fig. 2B) and a significant increase of the BrM thickness (1.12 ± 0.25 µm, P = 0.005) compared with un-treated controls (Fig. 2B). The detachment of PRC from RPE was also observed from TEM images.
Funduscopy revealed excess concentrations of oxidized lipoprotein in RPE at 4-month post light exposure groups. (Supplementary Materials S1). Lipofuscin can be detected using autofluorescence from 488 nm light source excitation. Study showed that lipofuscin/drusen can generate intense green fluorescence (Supplementary Materials S2). The hemorrhage under RPE was also observed on histology samples. Additionally, the random localized lesion and thickening of choroid layer was also found on 3 h light exposed mice PI-10 weeks (Supplementary Materials S3).
The ERG showed a significant reduced amplitude in PRC (Fig. 3). The scotopic a-wave represents the electric response from rod PRC cells, which are sensitive to dimmed light. The Nrf2 −/− mice loss approximately 50% response to lights (−107 ± 16 µV at 10 cd*s/m 2 ) of the scotopic a-wave in the first week after light exposure. Although the response was slightly recovered after five weeks, the PRC function was further compromised at the endpoint (−97 ± 13 µV at 10 cd*s/m 2 , scotopic a-wave). These results are directly correlated to the PRC thickness change results, indicating the loss of PRC functionality. The photopic b-wave readings indicate the response from cone PRC cells. Similar to scotopic a-wave, the impairment of cone cells was significantly reduced (P = 0.017 1-week, P = 0.010 5-week, and P = 0.004 10-week) compared to naïve mice. The overall function of the retina (mainly from bipolar cells) from the scotopic b-wave also reduced ~33% in 1 week post light exposure   . Naïve, 3 h 4500 lux light-exposed mice (One-way ANOVA with Tukey's multiple comparison test, n = 14, error bars indicate SEM *P < 0.05, **P < 0.01, ***P < 0.001). www.nature.com/scientificreports www.nature.com/scientificreports/ (−149 ± 19 µV at 2.25 cd*s/m 2 ). Although there were some recoveries after 5 weeks, the overall function of the retina was disrupted (−125 ± 18 µV at 2.25 cd*s/m 2 , ~40% reduction) after 10 weeks as direct result of accumulated ROS.
the Gccnp laden hydrogel is biocompatible and non-toxic. Our previous research not only demonstrated the biocompatibility of GCCNP and injectable hydrogel, but also assessed the wide safe dose range in which GCCNP can be used 28 . In addition, delivery of nanoceria to ARPE-19 cells via injectable hydrogel was evaluated in our study and showed that hydrogels were biocompatible and sustained nanoceria release for a duration of two months without causing any side-effect. From Fundus and OCT images of Nrf2 −/− mice with the absence of light damage, the hydrogels started to degrade from the first-week post-injection and lost some content by 5 weeks of injection. By the time of 10 weeks, the residue of hydrogel was absent from the image and the hydrogels were degraded (Supplementary Materials S4). Compared to naïve Nrf2 −/− mice, there is no significant sign of inflammation response or toxicity to PRC or RPE. We assume that the sustained release character of the hydrogel prevented the dose dumping of GCCNP upon injection, assisting GCCNP in reaching the RPE in a sustained manner. From the fundus and OCT images of blank and GCCNP laden hydrogels intravitreal injection, properly injected hydrogel is clear, and does not cause cataracts and other damage to the anterior segment (Supplementary Materials S4). There is no evident signs of retinal structure changes or abnormalities to PRC or RPE as shown from Fundus and OCT. The retina retained a healthy formation and no retinal detachment or cataracts were found after injection.
the Gccnp hydrogel protected Rpe and pRc from structural deformation in light-induced Nrf2 −/− mice. To investigate whether GCCNP and its injectable hydrogel can improve the morphological changes in the light-induced Nrf2 −/− mice model's RPE and PRC, we injected GCCNP and the injectable hydrogel (with or without GCCNP) to 3-day post light exposure Nrf2 −/− mice. No adverse events or inflammatory reactions were observed post-injection. At 10 weeks post-injection (PI), histological staining showed that standalone hydrogel didn't make any progress on RPE abnormality, but partially improved PRC recovery, although not reaching significance. In light induced Nrf2 −/− mice, both GCCNP and GCCNP laden hydrogel groups exhibited regeneration of PRC structure and morphology. These two groups also displayed an ONL thickness comparable to naïve Nrf2 −/− mice on both PRC and RPE (Fig. 4A).
From Fundus/OCT scans, the injectable hydrogel only group did not show any adverse events. At 1-week PI, no significant difference in ONL thickness was observed in GCCNP (60 ± 6 µm), hydrogel only (52 ± 10 µm), and GCCNP laden hydrogel (67 ± 7 µm) groups, compared to light-exposed group; although GCCNP laden hydrogel group achieved more notable ONL recovery as revealed by OCT measurement. The retina started to recover from PI-5 weeks in GCCNP hydrogel group (70 ± 10 µm) and lasting until the experiment endpoint (Fig. 4B). We did not find any obvious retinal thickness changes in the GCCNP standalone group until PI-10 weeks (58 ± 9 µm). The GCCNP hydrogel group showed significantly faster recovery compared to the GCCNP group (P = 0.049).
The GCCNP hydrogel reduced AMD-like atrophy in light-induce Nrf2 −/− mice. From the histology sections, results showed that both GCCNP and GCCNP hydrogel treatments improved the AMD-like phenotypes in light-exposed Nrf2 −/− mice. Apparent granule/basal linear drusen, vacuole, hyper/hypo-pigmentation, or neovascularization was not observed from those mice (Fig. 4).
The treatment using GCCNP and GCCNP hydrogel reduced the number of lipofuscin to the normal level (approximately 1-2 lipofuscin granules per cell), and the difference between treated groups and untreated groups is significant (P = 0.0005) (Fig. 5B). Compared to the untreated group, both GCCNP and GCCNP hydrogel achieved significant improvements (P < 0.0001). The BrMs of both treated mice were comparable to naïve eyes after the treatment (0.58-0.64 µm) (Fig. 5B). The electron-dense nanoceria was also observed in RPE cells with homogeneous distribution (Fig. 5A), as similar shown in our previous research and Supplementary Materials S3. From Fundus images of light-induced mice treated by GCCNP laden hydrogels, the green signal of lipofuscin was eliminated after the treatment (Supplemental Materials S2).
Hydrogel-mediated GCCNP delivery rescued retinal functions in light-exposed Nrf2 −/− mice. To determine whether GCCNP and GCCNP laden hydrogels can restore vision function in light-exposed Nrf2 −/− mice, ERGs were performed at PI-5 weeks and PI-10 weeks. When comparing GCCNP laden hydrogel mice to the results of standalone GCCNP treated mice (scotopic a wave and photopic b wave, Fig. 6) the difference in ERG retinal functions were significant (P = 0.047). GCCNP laden hydrogel groups showed earlier recovery periods, which could be facilitated by the GCCNP laden hydrogel treatment both scotopic (−173 ± 23 µV at 10 cd s/m 2 scotopic a-wave) and photopic (213 ± 13 µV at 2.25 cd s/m 2 , photopic b-wave) ERGs at PI-5 weeks. At PI-10 weeks, the ERG signal from GCCNP and GCCNP hydrogel treated eyes reached a similar level (Fig. 6). Compared with untreated and standalone GCCNP treated mice, both treatments achieved statistically significant changes in the ERG retinal functions.
GCCNP laden hydrogels prevented production of excessive ROS and pro-inflammatory protein in light-exposed Nrf2 −/− mice. Additionally, we investigated potential antioxidative properties of GCCNP and its injectable hydrogels using free radical assay, real-time Polymerase chain reaction (RT-PCR) and Western blotting. To directly asses the ROS quenching in PRC and RPE, we conducted dichlorodihydrofluorescein diacetate (DCF-DA) assay and luminol assay ( Fig. 7A and Supplementary Materials Table 1). We found a substantial increase of fluorescence and luminescence signal in light-exposed Nrf2 −/− mice. Treatment with GCCNP and/or GCCNP injectable hydrogels significantly reduced the luminescence signal intensity to a close to normal value at PI-10 weeks (P = 0.012, n = 4). (2019) 9:14573 | https://doi.org/10.1038/s41598-019-51151-7 www.nature.com/scientificreports www.nature.com/scientificreports/ Western blotting revealed that the light exposure resulted in an accumulation of oxidative products of 4-hydroxynonenal (HNE), cellular response to hypoxia product of hypoxia-inducible factor (HIF)-1a), and products of AP-1 and NF-κB3 (p65) proteins that involved in cell apoptosis and inflammation 9,29 . From Fig. 7B, the light-exposed retina produced a large amount of 4-HNE and HIF-1α at 5 and 10 weeks after light exposure, and those can be significantly reduced by the treatment of GCCNP and its injectable hydrogels. The GCCNP and hydrogel showed synergistic effects in protein studies (P < 0.001). In comparison with the GCCNP treated mice, hydrogels with GCCNP further reduced the production of 4-HNE (P = 0.0453 at PI-5 weeks and P = 0.0245 at PI-10 weeks). AP-1 and NF-κB3 were also significantly downregulated at PI-5 weeks (P = 0.017 AP-1 and P = 0.045 NF-κB3) and PI-10 weeks (P = 0.013 AP-1 and P = 0.007 NF-κB3) with GCCNP laden hydrogel.

Synergistic antioxidative mechanism of Gccnp and injectable hydrogel.
To study the synergistic therapy mechanism of GCCNP and hydrogel, we performed the analysis of structural and dimensional changes of GCCNP released from hydrogel and compared with standalone GCCNP. The UV-Vis spectra showed absorption maximum of GCCNP at 295 nm and alginate-gelatin hydrogel at 260 nm, and the spectrum of GCCNP released from hydrogel combined maximum absorption values of both standalone GCCNP and alginate-gelatin hydrogel. It is hypothesized that chitosan and alginate formed polyplexes after hydrogel degradation 30 . To prove the hypothesis, GCCNP released from hydrogel and standalone GCCNP were imaged using TEM. It is clearly that GCCNP released from hydrogel has a much larger dimension compared to standalone GCCNP. The DLS size of standalone GCCNP is ~170 nm after sonication, but degraded hydrogel released GCCNP have a wider range of size 200-900 nm (Supplementary Materials S5). The zeta potential test showed opposite potential of alginate and standalone GCCNP. The alginate-gelatin hydrogel has ~−43 mV potential but GCCNP has over 30 mV potential. The released GCCNP reached ~−30 mV potential, which is the evidence that glycol chitosan and alginate formed polyplex and increased in size.

Discussion
In human atrophic AMD, the antioxidative gene, Nrf2's, expression is downregulated or compromised due to aging 31 . Deficiency or downregulation of Nrf2 gene expression led to failure in neutralization of excessive ROS and caused long-term oxidative stress on PRC and RPE cells 24,26 , thus the PRC and RPE cells will suffer from chronic oxidative stress related complication. Accumulated oxidized waste will accumulate and immunity (pro-inflammatory proteins, macrophages, and multiple complement factors) will be activated in response to both rising ROS and waste presence. Eventually, apoptosis and membrane attack complexes response will be triggered in the RPE and PRC cells, this cellular damage can result in higher order complications, including the loss of retinal function and inflammation.
Nrf2 is a transcription factor that regulates the expression of antioxidant proteins which serve to protect against oxidative damage triggered by injury and inflammation 10 . Nrf2 −/− mice have shown AMD-like retinal pathology including drusen-like deposits and RPE/BrM degeneration at their late age (12 months) 23 . In our newly generated murine model, we have shown that the AMD-like retinal pathology can be induced by a mild white light exposure in Nrf2 −/− . Compared to aged (12 months) Nrf2 −/− AMD-like mouse model, this model saves valuable time and allows for easier access in the evaluation of atrophic AMD treatment. AMD pathology can be triggered by a 3 h, mild white light exposure (4500 lux) in lieu of commonly studied 10,000 lux at 2 months of age and result in permanent retinal functional loss. Visible light with a wavelength shorter than 470 nm may cause acute damage to RPE and PRC cells through elevation of ROS 24 . With the absence of antioxidative Nrf2 pathway activation, the oxidative stress in the Nrf2 gene knockout model is difficult to neutralize. Thus, the light exposure could trigger localized oxidative stress, resulting in the oxidation of lipids and accumulation of inflammatory proteins/nucleic acids.
The raising ROS levels can result in a wide range of vision impairment-related complications 23,26 , including inflammation and angiogenesis. The large vacuole of autophagy, which remove oxidized proteins/lipids, is a significant symbol of dry AMD, and this ROS driven process is minimized with GCCNP treatment 23 . The largely produced pro-inflammatory proteins, (HIF-1α, IL-6, and TNF-α) caused by ROS elevations, are determined from this murine model through multiple experiments, and can be well controlled under antioxidant treatment 32,33 . Moreover, our model has the potential to develop wet AMD-like phenotypes. In a 4-month time period our murine model exhibited abnormal changes among choroidal capillaries, which is an indicative sign of neovascularization.
The sustained and auto-regenerative antioxidative properties of GCCNP were evaluated in our previous publication. Preliminary studies using GCCNP and alginate-gelatin hydrogel on ARPE-19 cell lines proved GCCNP's anti-oxidation and anti-inflammation feasibility and determined the safe dosage range for injection 16,28 . The treatment using GCCNP rescued both structure and function of the retina. The combined treatment of the alginate-gelatin injectable hydrogel with GCCNP was able to achieve a rapid recovery of RPE and PRC within 5 weeks. This combined treatment proved to be superior to the standalone GCCNP which took up to 10 weeks www.nature.com/scientificreports www.nature.com/scientificreports/ for results of the treatment to prevail. Previous studies displayed not only the robust synergistic antioxidative properties of alginate-gelatin injectable hydrogel, but also its ability to reduced inflammation and angiogenesis 28 . As a sustained releasing vehicle, the hydrogel prevented initial burst release of GCCNP, reducing complication of GCCNP overdose (e.g. cytotoxicity) and allowing for a more homogeneous distribution under the retina. After mixing with hydrogel, GCCNP formed polyplexes with alginate moiety of hydrogel, and trafficking of GCCNP allowed alginate to enter the PRC and RPE cells. The negative charge of the alginate moiety of the polyplex is believed to aid in the transportation of GCCNP 34 . The alginate moiety can also contributes to the scavenging of free radicals 35 . Therefore, GCCNP facilitates the transportation of alginate and performs synergistic antioxidation.
As an effect of the alginate treatment, PRC were shown to have an earlier recovery compared to standalone GCCNP treatment from ERG and WB study. In addition, the efficacy of GCCNP in wet AMD treatment was shown in separate anti-angiogenesis studies using laser damaged murine models 16 .
Thus, the combination of GCCNP and injectable hydrogel is highly promising for AMD treatment, and proposes a feasible treatment plan for other elevated oxidative stress associated ophthalmic diseases (e.g. diabetic retinopathy) 36 . However, the mismatch of fast novel hydrogel formulation development and rigorous FDA approval procedure, prevents most hydrogels from being approved in a drug delivery purpose. Furthermore, the long-term safety and effectiveness of nanoceria still requires further studies and may not be available for clinical trials in the near future. Long-term evaluation of GCCNP and hydrogel are still requested before clinical trial on dry AMD patients.

conclusion
Nrf2 −/− mice can generate the phenotype of dry AMD after moderate light exposure at their early age. These phenotypes included inflammation, drusen-like formation, compromised retinal function, degeneration of PRC and RPE, and accumulation of oxidized products. This model saved time and displays similar phenotypes to that of their aged counterparts, providing new tools to study dry-AMD drug intervention. GCCNP delivery www.nature.com/scientificreports www.nature.com/scientificreports/ via intravitreal injection reduced ROS caused damage in the mouse retina. The GCCNP laden hydrogel further improves the treatment's efficacy compared to standalone GCCNP. The combination of GCCNP and injectable hydrogel have potential to translate to the clinical field for treatment of dry AMD.
intravitreal injection of Gccnp and hydrogel. 3 days after light exposure, 10 µL 1 mg/mL GCCNP was loaded into 10 uL nanofil syringe (World Precision instrument) with a 33 G blunt end needle. The Nrf2 −/− mice were anesthetized, penetrated under limbus with 27 G needle, and 1 uL GCCNP was injected via intravitreal method using micro syringe pump at 20 uL/min infusion rate. The injection of hydrogel required the same apparatus. 1 µL hydrogel precursor (with 0 or 1 mg/mL GCCNP) was injected intravitreally, 0.5 uL CaCl 2 was infused immediately at the same infusion rate and the same position. 1 µL 0.9% PBS was injected as injection control group. fundus and oct. Mice were anesthetized and pupils were dilated with 1% tropicamide. Fundus images were captured using a Micron IV retinal imaging system (Phoenix Research Labs). Corneas were moistened with Genteal lubricant eye gel (Novartis) and positioned with the Micron eyepiece in direct contact with the eye through the gel. OCT images were captured using the full-scan setting as an average of 10 frames/scan. Three scans were taken per eye in the superior/inferior and/or nasal/temporal cardinal directions relative to the optic nerve. Retinal thickness measurements were analyzed using InSight OCT segmentation software (Voxeleron). The thickness of ONL and total retinal thickness (between and including the ganglion cell layer and the IS (inner segment)/ OS (outer segment junction)) were measured at 0.2 mm intervals from the optic nerve head (ONH) 38 . eRG. ERG was recorded and analyzed as suggested in previously published protocols 39 . Briefly, mice were dark-adapted overnight, and eyes were dilated with 1% tropicamide under dim red light. Genteal gel 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 in the tail. ERG was collected using an Espion E2 system (Diagnosys), and mice body temperature was maintained at 37 °C during the experiment 38 . The light stimulus proceeded with 0.1, 1, 2.25, 4, and 10 cd*s/m 2 white light under dim light with 10 s intervals and each stimulus was recorded five times for rod cells and bipolar cells response. Next, mice were light adapted with 10 cd*s/m 2 white light for 5 min, then proceed with 2.25 cd*s/m 2 for cone cells response.
Histology. The Haemotoxylin and Eosin (H&E) staining of eye tissues were carried out by the UNC Center for Gastrointestinal Biology and Disease Histology Core Facility. Briefly, fixed tissues were embedded in paraffin and sectioned into 10-mm sections for eye tissues. Sections were stained with H&E, according to conventional staining protocols, and visualized using bright-field microscopy.
Rt-pcR. The levels of IL-6, TNF-α, IFN-γ and caspase-3 obtained from homogenized retina and RPE were assessed using RT-PCR. RNA was extracted from retinas following the TRIzol protocol (Life Technologies) according to the manufacturer's instructions. Samples were subjected to RT-PCR and normalized to β-actin to verify the efficiency of RNA extraction and for the presence of cytokines. Primers targeting the cytokines were used 32 . free radical assay. The DCF-DA and luminol assay was adopted from previous publication 16,40 . Briefly, isolated fresh retinal tissue was homogenized in 0.5 mL PBS-HEPES buffer (0.5 M PBS containing 20 mM HEPES, pH 7.2) and transferred to a 96-well plate. DCF-DA solution was added at 0.05 mM final concentration and cultured at 37 °C for 30 min. Luminol (5-amino-2,3-dihydro-1,4-phthalazinedione, Sigma-Aldrich) DMSO solution was added at 0.2 mM final concentration. Fluorescence and chemluminescence were quantified using plate reader (BMG Omega). Fluorescence intensity was measured with ex488 nm/em532 nm filter. Chemluminescience counts were obtained at 30 second intervals and the final results were represented as the area under curve (AUC) for 5 min.
Microscopy. The imaging of cells was conducted using the Zeiss LSM-700 confocal fluorescence microscope and observer D1 fluorescence microscope with Zeiss AxioCam MRC camera. The images were processed by ZEN software.