Animal experiments were conducted in accordance with the Guide for the Care and Use of Laboratory Animals published by the US National Institutes of Health (NIH Publication No. 85-23, revised 2011) and the guidelines on the ethical use of animals of Fudan University. Adult male Sprague–Dawley (SD) rats weighing approximately 180 g (SLAC Laboratory Animal Co., Ltd., Shanghai, China) were used in the experiment. The rats were housed with a daily 12-h light/12-h dark cycle and had free access to food and water. A total of 45 SD rats were randomized into the normal control (NC) group (n = 10), the light damage (LD) group (n = 10), the LD + vehicle group (n = 10, daily intraperitoneal injection of 1 μl DMSO), and the LD + Pae group (n = 15, daily intraperitoneal injection of 1 μl Pae, 80 mg/kg). After one week of Pae intraperitoneal injection, the electroretinogram (ERG) function was detected 3 days after 24-h blue light exposure (wavelength of 400–440 nm). The rats were then sacrificed with excessive abdominal anesthesia of chloral hydrate (600 mg/kg), and eyeballs were harvested for either immediate or future use.

The eyeballs from each group were harvested, enucleated and immediately fixed in 4% paraformaldehyde for 24 h. Then the samples were embedded with paraffin, serially sectioned (5-μm/section) and stained with hematoxylin and eosin (Takara: C0105S). The sections were observed using a light microscope (Leica, Wetzlar, Germany), and the slices cutting cross-sectionally through the eyeball and the optic nerve were selected. The images photographed were all location matched, and the thickness of different retinal layers were measured at the points located approximately 250 μm from the optic nerve. Each selected section was measured five times and averaged to obtain the values for one sample.

4% Paraformaldehyde-fixed and optimal cutting temperature compound-embedded rat eyes of every groups were sectioned at a thickness of 10 μm and subjected to TUNEL assay (Takara: C1089) to detect apoptotic cells. TUNEL staining was performed in accordance with the manufacturer’s instructions (Zhu et al., 2010). After counterstaining with DAPI (1:2000; Life Technologies, Carlsbad, CA, United States) for 10 min, the sections were observed using a confocal microscope (Leica SP8, Hamburg, Germany).

A piece of posterior pole tissue (1 mm × 1 mm × 1 mm) of eyes from each group was rapidly dissected on ice after enucleation, and fixed via 2.5% glutaraldehyde in 0.1 M phosphate buffer at 4°C for 2–4 h. After washing in 0.1 M phosphate buffer for 3 times, the samples were post-fixed with 1% osmic acid at 4°C for 2 h and washed again. The samples were then dehydrated using an ascending alcohol series, infiltrated with a 1:1 mixture of resin and acetone and embedded in epoxy resin. After being polymerized in a 60°C oven for 48 h, 60–80 nm ultrathin sections were obtained by using an ultramicrotome (Leica, Leica UC7). The sections were double stained with lead citrate and uranyl acetate (each for 15 min) and examined under a transmission electron microscope (TEM; HITACHI, HT7700).

The ERG was measured to evaluate the RP function of rats before intraperitoneal injection and three days after retinal light damages separately and was recorded by an Espion Diagnosys System (Diagnosys, Littleton, MA, United States). As described in Miyai et al. (2019), after 24 h of dark adaptation, the rats were given intraperitoneal anesthesia, and their pupils were dilated with phenylephrine hydrochloride and tropicamide (0.5%). Two wire loop electrodes were placed on the corneal surface of the eyes and served as the ERG signal-recording electrodes. In addition, two subdermal needle electrodes were inserted into the base of the tail and nasal part and separately served as the ground electrode and the common reference electrode. Retinal responses were recorded for 30 min.

Light stimulation was performed using a white LED following the protocol described in Miyai et al. (2019). Dark and light adaptation was performed in four steps, and the light intensity was switched from weak to strong. Electroretinographic waveforms were recorded and sampled, and the data were analyzed by using a Diagnosys digital acquisition system. The waveforms of ERG were measured from trough to peak (Lazarou et al., 2015), and the values of ERG amplitudes were compared among the four groups.

A photoreceptor cell line, 661W cells, was cultured in DMEM (Invitrogen: 11965092) containing 10% fetal bovine serum and 1% penicillin/streptomycin under normal condition (5% CO₂, 37°C) as previously described (Birgisdottir et al., 2013). Cells of the passages 15–20 were used in the following experiments.

The 661W cells were seeded in 96-well plates at a density of 1 × 10⁴ cells per well, and grouped as follows: control cells, cells exposed to H₂O₂, cells treated with Pae, and cells incubated with both H₂O₂ and Pae. The cells were treated with different concentrations of H₂O₂ (0–200 μM) and Pae (0–200 μM) for different time periods (0, 24, 48, 72, 96 h). After the cells were grown to approximately 80–90% confluency, Cell Counting Kit-8 (CCK-8) (Dojindo: CK04) assay was performed following the manufacturer’s protocol. The absorbance of each well was read at 450 nm by a microplate reader (BioTek, United States).

Briefly, retinal tissues and cultured cells were lysed in radio-immunoprecipitation assay buffer (ASPEN, China) containing protease inhibitor cocktail (ROCHE). All sample extracts were electrophoresed by SDS-PAGE (ASPEN, China) and electrophoretically transferred to PVDF membranes. The membranes were then blocked with 5% skim milk at room temperature for 1 h, incubated with primary antibody diluent at 4°C overnight and incubated with horseradish peroxidase-coupled secondary antibodies. The blots were developed using enhanced chemiluminescence (ECL) and the signals were captured in a dark room. The immunoreactive bands were analyzed in triplicate by ImageJ, with GAPDH being used as a loading control.

Primary antibodies were as follows: anti-LC3 (1:1,000, Cell Signaling Technology/CST: 3868), anti-Bax (1:1,000, CST: 2774), anti-Bcl-2 (1:1,000, CST: 9942), anti-casepase-3 (1:1,000, CST: 9662), anti-p62 (1:1,000, CST: 8025), anti-Beclin-1 (1:2000, CST: 3738), anti-PINK1 (1:2000, CST: 6946), anti-Parkin (1:2000, CST: 4211), anti-Optineurin (1:2000, CST: 58981), anti-DNP52 (1:2000, CST: 60732), anti-BNIP3L/Nix (1:2000, CST: 12396), anti-BNIP (1:2000, CST: 44060), and anti-GAPDH (1:1,000, CST: 8884).

The 661W cells were seeded in 6-well plates at a density of 1 × 10⁵ cells per well and treated with H₂O₂ (100 μM) and/or Pae (50 μM) for 24 h after the cells were grown to approximately 80% confluency. Mitochondrial reactive oxygen species (ROS) were measured with a ROS Detection Kit (Takara: S0033S) according to the manufacturer’s instructions.

All experiments were repeated more than three times. Data were presented as mean ± standard deviation. One-way ANOVA and Bonferroni’s multiple-comparisons test were performed to compare the between-group difference. All statistical tests were performed with SPSS version 20 (IBM, United States) and p < 0.05 were considered statistically significant.

Exposure to blue light successfully induced the LD model in SD rats. Compared with the control group, significant thinned outer nuclear layers (ONL) were observed in the LD group and LD + vehicle group, whereas the thickness of ONL was effectively preserved after the administration of Pae (Figures 1A,B). It has been reported that apoptosis and autophagy appear to markedly accelerate three days after light injury, thus apoptosis in retina was examined by TUNEL assay three days after LD modeling. Apoptotic activity was almost absent for the control group, while an increasing number of TUNEL-positive cells were detected in the retina of the LD group and LD + vehicle group, especially located in ONL. Similarly, Paeonol treatment effectively attenuated light-induced RPs apoptosis in the LD + Pae group (Figure 1C).

The ERG reflects broad-scale retinal function, with the a- and b-waves indicating photoreceptor and second-order cell responses, respectively. As shown in Figure 2, the decreased amplitudes of a- and b-waves on scotopic ERG caused by LD were significantly mitigated by Paeonol treatment.

The above results indicated that Pae exerts protective effects on RPs by ameliorating morphological and functional damage induced by blue light exposure.

The visualization of double-membrane compartments via TEM was considered as the gold standard for identifying autophagosomes (Pickford et al., 2008). As shown in Figure 3A, we found that cells from the NC group contained healthy mitochondria, which were easily recognizable in the normal cytoplasm. On the contrary, numerous double-membrane vacuoles accompanied by a few accumulated autophagic compartments were observed in cells of the retina subjected to 3 days after LD. Meanwhile, in Pae-treated retina, both newly formed and mature autophagosomes could be detected. Compared with increased autophagosome formation in the LD + Pae group, autophagy was inhibited in LD retina due to reduced autophagy induction, not by an increased autophagic flux.

As shown in Figure 3B, Western blot analysis was used to evaluate the apoptosis in retina. Cleaved caspase-3 levels in the retina of the LD and LD + vehicle groups were remarkably higher than those in the NC group, which could be reversed by Pae treatment. Meanwhile, the Bax expression levels in the LD and LD + vehicle groups were remarkedly higher than those in the NC group. In addition, Bcl-2 levels were significantly reduced in retinas of the LD and LD + vehicle groups compared with those in the NC group. Pae significantly inhibited the up-regulation of Bax and down-regulation of Bcl-2. These results indicated that Pae might exert anti-apoptotic effects in the retina. This conclusion was consistent with the results of TUNEL assay (Figure 1C).

LC3-II and Beclin-1 are two major autophagy markers, and the PINK1/Parkin pathway is one of the important pathways in autophagy. The expression of LC3 and Beclin-1 in the retina was measured to evaluate the alteration of autophagy (Figure 4). In the LD and LD + vehicle groups, the expression levels of LC3-I and LC3-II were notably downregulated. In particular, the conversion rate from LC3-I to LC3-II in both groups was significantly decreased compared with that in the NC group, suggesting the absence of LC3-II and LC3-I. The ratio of LC3-II to LC3-I was higher in LD + Pae group comparing with it in LD + vehicle group. PINK1 and Parkin were remarkably downregulated in the LD and LD + vehicle groups compared with the NC group. Correspondingly, as the key regulatory protein for autophagy, Beclin-1 was also reduced. Moreover, the alterations of autophagy markers and autophagy pathway were significantly mitigated by Pae treatment. The above results indicated that LD damage in retina caused the resistance of autophagosome formation, the decreased autophagic flux and the block of autophagy, which fortunately Pae can improve to a much better extent.

To determine the optimal dosage of H₂O₂, 661W cells were cultivated with 0, 25, 50, 100, or 200 μM H₂O₂, and the cell viability was measured by using a CCK-8 kit. Compared with the control group, cell viability was dramatically decreased by incubating with 50, 100 and 200 μM H₂O₂, while no significant decrease of cell viability was observed in cells treated with 25 μM H₂O₂ (Figure 5A). Thereafter, we determined the optimum time period of cell culture. We used different concentrations of Pae (0, 25, 50, 100 and 200 μM) to stimulate cells for 24, 48, or 72 h, and accessed cell viability to determine the adequate time of experiment. As shown in Figure 5B, we selected 48 h as the cultivation time to detect changes in the viability of cells treated with different concentrations of Pae. Thus, we measured the cell viability employing 100 μM H₂O₂ with different concentrations of Pae (0, 25, 50, 100 and 200 μM) treatment to sort out the moderate amounts of Pae. After 48 h in vitro culture, the cell viability was notably increased by 50–100 μM Pae and decreased by 200 μM Pae compared with H₂O₂ stimulation (Figure 5C). The results indicated that cell viability impaired by H₂O₂ could be effectively improved by Pae treatment in 661W cells.

Then, we examined the apoptosis-associated proteins including cleaved casepase-3, Bax and Bcl-2. The expression levels of pro-apoptotic cleaved caspase-3 and Bax were upregulated, whereas those of anti-apoptotic Bcl-2 were downregulated by H₂O₂ stimulation compared with the control group. These apoptosis-related alternations could be reversed by Pae treatment (Figures 5D,E). The results demonstrated that apoptosis induced by H₂O₂ could be significantly alleviated by Pae treatment in 661W cells.

Subsequent experiments were conducted to determine whether Pae affected oxidative stress and autophagy in H₂O₂-stimulated 661W cells. By using Mito-Sox assay to detect the mitochondrial ROS, we found that the fluorescence intensity was significantly higher in the H₂O₂ group than the control group, and the fluorescence intensity noticeably decreased in the H₂O₂+Pae group, suggesting the antioxidative effect and protective effect of Pae on mitochondria in 661W cells (Figure 6A).

Furthermore, the expression of major autophagy marker proteins was measured (Figure 6B,C). After H₂O₂ stimulation, the expression levels of LC3-I, LC3-II and Beclin-1 were significantly downregulated, and the expression level of p62 were elevated compared with the control group (all p < 0.01). However, these changes of expression level could be reversed by the administration of Pae (all p < 0.01), which implied that Pae could activate autophagosome formation and enhance the autophagic flux in H₂O₂-stimulated 661 W cells. Significantly, the ratio of LC3-II to LC3-I was higher in Pae group comparing with in Control group and was higher in H₂O₂ + Pae group than H₂O₂ group. In combination with the expression of p62 and Beclin-1, it is possible that the autophagy overactivation was occur.

We speculated that Pae might regulate autophagy in 661W cells based on evidence obtained from above experiments. Thus, we measured the PINK1/Parkin dependent and BNIP3L/Nix dependent pathways of autophagy, respectively.

The results of Western blot analysis showed that the expression of PINK1 and Parkin were significantly reduced after H₂O₂ stimulation (both p < 0.01), but were notably augmented after Pae treatment (both p < 0.01). It was worth noting that PINK1 and Parkin were dramatically upregulated by Pae treatment alone when compared with the control group (both p < 0.01) (Figure 7A). We further detected the expression levels of critical downstream proteins in PINK1/Parkin pathway and found H₂O₂-induced decreases of DNP52 and Optineurin were significantly reversed by Pae (both p < 0.01). In accordance with findings of PINK1 and Parkin, Pae treatment alone significantly augmented Optineurin (Figure 7A). The results showed that Pae could activated autophagy via the PINK1/Parkin pathway and might act on Optineurin directly.

In addition, we evaluated the BNIP3L/Nix dependent pathway of autophagy. Significantly, the expression levels of BNIP3 and BNIP3L/Nix were both decreased in the H₂O₂ group (both p < 0.01) but increased after Pae treatment (both p < 0.01) (Figure 7B), which indicated that Pae could activate autophagy via BNIP3L/Nix dependent pathway of autophagy in 661W cells as well.