Roles for Redox Signaling by NADPH Oxidase in Hyperglycemia-Induced Heme Oxygenase-1 Expression in the Diabetic Retina

Abstract

Purpose.: The antioxidant response element (ARE)–mediated antioxidant pathway has an important role in maintaining the redox status of the retina. The expression of ARE-mediated antioxidants, such as heme oxygenase-1 (HO-1), remains unclear in the db/db mice. We evaluated the expression of HO-1 in the retinas of db/db mice and investigated a possible role for NADPH oxidase.

Methods.: Fresh retinas were harvested from 8-, 12-, and 20-week db/db or db/m mice. Reactive oxygen species were detected by dihydroethidium. The expression levels of HO-1, Nox2, and Nox4 were evaluated by immunohistochemistry and Western blotting. In vitro retina explants culture was used to assess the role of NADPH oxidase in high glucose-induced HO-1 expression.

Results.: The expression of HO-1 was increased in the retinas of 8-week db/db mice, while it was decreased in 20-week db/db mice compared to age-matched controls. Similarly, the activation of Nox4 was increased in the retinas at 8 weeks and returned to basal levels at 20 weeks in db/db mice compared to age-matched controls. The activation of Nox2 was increased in the retinas of 8-, 12-, and 20-week db/db mice compared to age-matched controls. The NADPH oxidase inhibitors apocynin and DPI significantly blocked the HO-1 expression that was induced by high glucose levels in cultured retina explants.

Conclusions.: The expression patterns of HO-1, Nox2, Nox4 in db/db mouse retinas, and the suppressive effects of NADPH oxidase inhibitors on the expression of HO-1 induced by high glucose levels in cultured retina explants suggest that the expression of HO-1 is, at least partially, mediated by NADPH oxidase in this diabetic animal model.

Introduction

Diabetic retinopathy (DR) is one of the leading causes of blindness worldwide. ^1,2 The pathologic changes in DR include retinal vascular damage, which includes vessel leakage, microaneurysms, ischemia, and neovascularization. However, mounting clinical evidence suggests that early stage DR injuries may be reflected not by vascular abnormalities, but rather by impaired retinal visual function as revealed by electroretinogram recordings, loss of color vision, and loss of contrast sensitivity. ^3,4 Thus, the onset of neuronal dysfunction occurs much earlier than the appearance of visible vascular lesions. ^5–7 The alterations of the visual function of retinal ganglion cells (RGCs) in the early stages of DR have been demonstrated in our previous study. ⁸  

Oxidative stress, which is induced by the increased accumulation of reactive oxygen species (ROS) and/or decreased antioxidant capacity, has an important role in the pathogenesis of DR. ^9,10 ROS are produced continuously as natural byproducts of the normal metabolism of oxygen and have important roles in redox signaling. ^11,12 Increases in the generation of ROS in the diabetic retina have been confirmed by several groups. ^7,13,14 The NADPH oxidase family of enzymes (Noxs) are one of the major sources of ROS in many cell types and tissues, including the retina. ^15–17 Among the seven subtypes of NADPH oxidase, NADPH oxidase-2 (Nox2) and −4 (Nox4) are two important modulators of redox signaling that are inducible in multiple cell types, including retinal cells, at the level of transcriptional expression. Increased activity and expression of Nox2 and Nox4 have been found in the retinas of type 1 and type 2 diabetic animal models, and have been related to increased oxidative stress in the diabetic retina. ^14,18–21  

Heme oxygenase-1 (HO-1) is an antioxidant regulated by antioxidant/electrophile response element (ARE/EpRE) that catalyzes the degradation of heme to iron, carbon monoxide, and biliverdin. Decreases in HO-1 gene expression and protein activity in the vasculature have been found in cases of diabetes. ^22–25 The overexpression of HO-1 in the retina can restore visual function in diabetic animal models. ^26,27 The induction of HO-1 by oxidative stress or inducers may reduce damage in cells by restoring the balance of antioxidants and pro-oxidants in the vasculature, ²⁸ suggesting that HO-1 represents an important adaptive mechanism of the tissue for moderating the severity of cell damage produced by oxidative stress. However, the expression of HO-1 in the retinas of db/db mice has not yet been studied to our knowledge. 

To investigate the antioxidant capacity of early stage type 2 diabetic retinas and the role of NADPH oxidases in the regulation of ARE-antioxidants, we analyzed the expression of HO-1 and NADPH oxidases in db/db mouse retinas, and the effects of NADPH oxidase inhibitors on HO-1 expression in ex vivo retinal explant cultures. Our data showed that the expression of HO-1 was increased adaptively in early diabetic retinas, but this expression decreased as diabetes developed. The similarity in the expression patterns of Nox4 and HO-1, and the inhibitory effects of NADPH oxidase inhibitors on high glucose-induced HO-1 expression suggested that NADPH oxidases, especially Nox4, are potential regulators of HO-1 expression in the diabetic retina. Therefore, these data demonstrated that early diabetic retinas have the adaptive antioxidant capacity to counteract increased oxidative stress, this capacity is diminished and further damaged as the disease develops, and NADPH oxidases are potential activators of the ARE-antioxidant pathway in early diabetic retinas. 

Materials and Methods

Animals

Totals of 45 C57BKS/J male db/db mice (8-week: 15; 12-week: 15; and 20-week: 15), 45 male age-matched heterozygous db/m litter mates (8-week-old: 15; 12-week-old: 15; and 20-week-old: 15), and 20 male C57BKS/J mice 8 weeks old (Jackson Laboratory, Bar Harbor, ME; strain background information is available in the public domain at http://jaxmice.jax.org/strain/000642.html) were used in our study. The animals were housed in a 12-hour light-dark cycle environment and fed on a normal diet. Animal blood glucose levels were measured with a blood glucose meter (Accu-Chek active; Roche Diagnostics Deutschland GmbH, Mannheim, Germany) at 9:00 AM every Monday and just before the experiment. The blood glucose levels and body weights of the animals are summarized in the Table. The animals used in our study did not receive insulin treatments. All experiments were performed in accordance with the Peking University guidelines for animal research, and the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. The experimental animal protocol used in our study was approved by Peking University Institutional Animal Care and Use Committee (IACUC). 

Detecting the Generation of ROS

The generation of retinal ROS was assessed with dihydroethidium (DHE; Invitrogen Molecular Probes, Eugene, OR), as described previously. ⁸ Briefly, fresh retinas were harvested and frozen quickly in liquid nitrogen for cryosection (Leica CM1950; Leica Microsystems Ltd, Wetzlar, Germany). Cryosections (10 μm) were washed with a warm PBS solution and then incubated with 5 μM dihydroergotamine (DHE) in PBS for 30 minutes at 37°C. DHE specifically reacts with superoxide anions and is converted to the red fluorescent compound ethidium. The sections were examined and photographed using an inverted fluorescent microscope equipped with a digital camera (Eclipex Ti-S; Nikon Instech Co., Tokyo, Japan) under identical exposure conditions, and the optical densities of the staining in the outer nuclear layer (ONL), inner nuclear layer (INL), and ganglion cell layer (GCL) were measured from randomly selected images. Five measurements were taken at 200 μm intervals from each image using a commercial software program (Photoshop CS5; Adobe Corp., San Jose, CA). 

In Vitro Retinal Cultures

Retinal explants were obtained from 8-week-old male C57BKS/J mice. The retinas were dissected according to the protocol described by Gaddini et al. ⁴ Briefly, the eyes of mice under deep anesthesia were enucleated and transferred to a Petri dish containing ice-cold sterile PBS. The anterior chamber of the eye was removed, and the retina was peeled away gently from the eyecup under a dissection microscope. The whole retina was dissected carefully from the sclera and was placed flat onto a PET microporous insert (Transwell; Corning Life Sciences, Amsterdam, The Netherlands), which allowed the tissue to be in contact with the culture medium on the apical and basal sides. The retinas then were cultured in Dulbecco's modified Eagle medium (DMEM; Gibco, Paisley, Scotland, UK) containing 10% fetal calf serum and antibiotics in a tissue culture incubator at 37°C with 5% CO₂. Retinal tissue cultures were exposed to normal glucose (NG; 5.5 mM), high glucose (HG; 25 mM), or mannitol (M; 5.5 mM glucose + 19.5 mM mannitol) as an iso-osmolar control, and were maintained for 24 hours with or without the NADPH oxidase inhibitors apocynin (100 μM) and DPI (10 μM). 

Immunohistochemistry

We used immunofluorescence to examine the localization and expression of HO-1 and Nox4. Briefly, the eyes were enucleated, postfixed in 4% paraformaldehyde for 45 minutes, and embedded in OCT. Sections were cut transversely along the temporal-nasal axis of the eyeball. To ensure comparability, only sections that contained the optic nerve stump were used in this comparative study. Three retinal sections per animal were sampled to increase the reliability of the data, and the numbers obtained were pooled together to give the final number of cells with immunostaining in each retina. The cryosections (10 μm) were thawed, air-dried, and washed three times with 0.01 M PBS (pH 7.4). Tissue specimens were treated first with 3% BSA (Sigma-Aldrich Corp., St. Louis, MO) in 0.3% Triton X-100 for 20 minutes at room temperature and then incubated with one of the following primary antibodies: rabbit polyclonal antibody against HO-1 (Stressgen, Inc., San Diego, CA) or rabbit polyclonal antibody against Nox4 (Abcam, Cambridge, UK). Immunoreactivity was detected by an FITC-labeled secondary antibody (Abcam), and the cell nuclei were counterstained with 4′-6-diamidino-2-phenylindole (DAPI). The optical density of the staining in the GCL was measured from randomly selected retinal slide images. Five measurements of each image were taken at 200 μm intervals using a commercial software program (Photoshop CS5: Adobe Corp.). Specifically, using the tools of Photoshop software, optical density plots of a fixed area (100 × 100 pixels) were generated and, after background subtraction, were used as an index of immunostaining intensity. 

Western Blotting

The eyes were enucleated, and the retinas were collected and shock frozen at −80°C within 2 minutes of enucleation. The retinas later were ultrasonically homogenized at 4°C in 300 μL RIPA buffer containing 50 mM Tris (pH 7.4), 150 mM NaCl, 10 mM EDTA, 0.1% SDS, 1% NP-40, 0.5% sodium deoxycholate, 1 mM Na₃VO₄, 1 mM NaF, 1 mM EGTA, 1 mM phenylmethylsulfonyl fluoride, and a proteinase inhibitor. Protein concentrations were determined with a BCA protein assay to ensure equal protein loading, and 20 μg of protein in each lane were separated by 10 or 12% SDS-PAGE. Next, the proteins were transferred to a nitrocellulose membrane (Millipore, Billerica, MA), which then was blocked and probed with rabbit polyclonal anti–HO-1 (Stressgen Inc.), anti-Nox2, and anti–Nox4 (Abcam) antibodies. A peroxidase-conjugated anti–rabbit secondary antibody (PerkinElmer, Inc., Wellesley, MA) was used, and the blots also were probed for β-actin (Sigma-Aldrich Corp.) as a loading control. The protein bands were visualized using the Amersham Biosciences ECL Western blotting detection reagent (GE Healthcare Life Science, Uppsala, Sweden) according to the manufacturer's instructions. For quantification, blots from at least five independent experiments (5 animals per group) were used and quantified by software Image J. 

Statistical Analysis

Data are expressed as mean ± SEM. Analysis between multiple groups was performed by two-way ANOVA, where one factor was animal type (dm versus db) and the other factor was duration of diabetes (8, 12, and 20 weeks), followed by Bonferroni posttests. If only one factor was involved, one-way ANOVA was conducted and followed by Bonferroni multiple comparison posttests. P < 0.05 was considered statistically significant. 

Results

Development of Hyperglycemia in db/db Mice

Consistent with previous studies, ^7,8 the db/db mice used in the current study had hyperglycemia starting at 8 weeks of age (dm 7.4 ± 0.22, db 9.2 ± 0.6, P < 0.01). For details see the Table. 

Table. 

 

Animal Blood Glucose Levels

Age, wk	Species	N of Animals	Blood Glucose, mmol/L	Body Weight, g
8	db/m	15	7.4 ± 0.22	18.2 ± 0.9
	db/db	15	9.2 ± 0.6*	32.3 ± 1.2†
12	db/m	15	7.7 ± 0.19	19.1 ± 0.5
	db/db	15	26.2 ± 1.47‡	41.5 ± 1.2†
20	db/m	15	8.4 ± 0.3	23.1 ± 0.6
	db/db	15	28.3 ± 1.02‡	53.2 ± 1.1†

 Values are means ± SEM.

*   P < 0.01.

†   P < 0.0001.

‡   P < 0.001 versus db/m.

ROS Generation in the Retinas of db/db and db/m Mice

As a first step toward understanding the generation of ROS in the retinas of db/db mice, we identified ROS generation in freshly prepared retinal sections by DHE labeling. As shown in Figure 1, compared to 8-, 12-, and 20-week-old db/m mice, the db/db mouse retinas had significantly elevated ROS generation at 8 weeks of age, and this oxidative stress continued to progress to until 20 weeks of age (Fig. 1A). Quantitative analysis showed that 8-, 12-, and 20-week-old db/db mice had significantly higher levels of ROS in their retinas compared to db/m mice (8 weeks dm 9.67 ± 0.84, db 22.71 ± 4.74, P < 0.01, n = 5; 12 weeks dm 7.90 ± 2.2, db 22.50 ± 6.50, P < 0.01, n = 5, 20 weeks dm 8.11 ± 0.56, db 21.13 ± 3.71, P < 0.01, n = 5, Fig. 1B). 

Figure 1

 

Increased ROS generation in the retinal neurons of diabetic mice. Micrographs of retinal sections stained with DHE. (A) Representative micrographs of retinal sections stained with DHE. (B) Quantitative analysis of ROS levels in the whole retina. The fluorescent intensities of DHE-labeled neurons were quantified (means ± SEM, n = 5). Compared to db/m mice, the generation of ROS was increased significantly in the retinas of 8-, 12-, and 20-week-old db/db mice. In situ ROS generation was determined in freshly prepared retinal sections by immunohistochemical staining for DHE; the thickness of each retinal section was 10 μm. A quantitative analysis was performed using an image analysis software program (Adobe Photoshop CS5; Adobe Corp.). **P < 0.01 compared to db/m mice with the same age, two-way ANOVA with Bonferroni posttests, n = 5.

NADPH Oxidase Expression in Diabetic Retinas

The NADPH oxidase family of enzymes (Noxs) are one of the major sources of ROS in many cell types and tissues, including the retina. ^15–17 Among the seven subunits of NADPH oxidase, Nox2 and Nox4 have been implicated in type 1 and type 2 diabetic retinopathy. ^14,20,21 Thus, we examined the expression of Nox2 and Nox4 in db/db retinas. The retinal expression of Nox4 was determined by immunofluorescence staining and Western blotting, while the expression of Nox2 was determined only by Western blotting. As illustrated in Figure 2, Nox4 immunoreactivities in the 8- and 12-week diabetic retinas were significantly greater than that of age-matched littermates (8-week dm 10.80 ± 1.23, db 17.10 ± 5.37, P < 0.01, n = 5; 12-week dm 10.93 ± 2.30, db 16.35 ± 3.34, P < 0.05, n = 5), indicating that Nox4 was activated in retinal neurons from early diabetic mice. Furthermore, the immunoreactivity of Nox4 was located mainly in the GCL, indicating that Nox4 was activated primarily in this cell layer. These results were confirmed further through a Western blot analysis. As displayed in Figure 3, Nox4 protein levels in the whole retina of 8-week-old db/db mice were significantly higher than those of age-matched db/m mice (dm 0.61 ± 0.16, db 1.31 ± 0.29, P < 0.01, n = 5). However, the expression pattern of Nox2 was different from that of Nox4. As shown in Figure 3, the protein levels of Nox2 in the whole retina of 8-, 12-, and 20-week db/db mice were significantly higher than those of age-matched db/m mice (8-week dm 0.48 ± 0.08, db 0.77 ± 0.09, P < 0.05, n = 5; 12-week dm 0.54 ± 0.08, db 0.90 ± 0.22, P < 0.05, n = 5; 20-week dm 0.49 ± 0.15, db 0.83 ± 0.08, P < 0.05, n = 5). 

Figure 2

 

Immunostaining for Nox4 in the retinas of db/m and db/db mice. Nox4 expression in the retinas of 8-, 12-, and 20-week-old db/m and db/db mice was determined by immunofluorescence staining with a specific Nox4 antibody; the thickness of each retinal section was 10 μm. (A) Representative micrographs of retinal sections that were stained for Nox4, DAPI was used to counterstain the nucleus. (B) Quantitative analysis of Nox4 expression in the retinal ganglion cell layer. The fluorescence intensity of Nox4-labeled neurons was quantified (means ± SEM, n = 5). Compared to db/m mice, Nox4 expression was increased significantly in the retinas of 8- and 12-week-old db/db mice. **P < 0.01, *P < 0.05 compared to db/m mice with the same age. Two-way ANOVA with Bonferroni posttests, n = 5. Other conventions are as in Figure 1.

Figure 3

 

Increased expression of NADPH oxidase in the retinas of db/db mice. Retinas were harvested freshly from 8-, 12-, and 20-week-old db/m and db/db mice according to the methods described in the article. Nox2 and Nox4 expression levels in the whole retina were determined by Western blotting. (A) Representative immunoblot of Nox2 and Nox4 expression in the whole retina. (B) Illustrates the densitometric analysis of Nox2 expression relative to the loading control (means ± SEM, n = 5). Compared to db/m mice, Nox2 expression was increased significantly in the retinas of 8-, 12-, and 20-week-old db/db mice. *P < 0.05, two-way ANOVA with Bonferroni posttests, n = 5. (C) Illustrates the densitometric analysis of Nox4 expression relative to the loading control (means ± SEM, n = 5). Compared to db/m mice, Nox4 expression was significantly increased in the retinas of 8-week-old db/db mice. *P < 0.05, two-way ANOVA with Bonferroni posttests, n = 5.

HO-1 Expression in Diabetic Retinas

To evaluate whether the diabetic retina has an adaptive antioxidant capacity to counteract the increased oxidative stress induced by NADPH oxidase, the expression levels of one of the ARE-mediated antioxidants, heme oxygenase-1, were examined in retinas of 8-, 12-, and 20-week mice by immunofluorescence staining and Western blot analysis. As shown in Figure 4, immunoreactivity of HO-1 was located mainly in the GCL. The intensity of HO-1 immunoreactivity in diabetic retinas from 8-week-old mice was significantly greater than that of age-matched db/m mice (dm 9.68 ± 1.17, db 17.35 ± 1.13, P < 0.001, n = 5). There were no significant differences in HO-1 expression in the retinas of 12-week-old diabetic and nondiabetic mice. Moreover, the immunoreactivity of HO-1 in 20-week db/db mouse retina was much less than that of age-matched db/m mice (dm 12.52 ± 1.34, db 8.21 ± 3.33, P < 0.05, n = 5). Consistent with the results obtained by immunofluorescence staining, the results of our Western blot analysis of the whole retina (Fig. 5) showed that the expression of HO-1 was increased adaptively in the retinas of 8-week-old db/db mice (dm 1.44 ± 0.25, db 3.42 ± 0.74, P < 0.001, n = 5), whereas HO-1 expression was decreased in the retinas of 20-week-old db/db mice (dm 2.30 ± 0.88, db 1.08 ± 0.26, P < 0.05, n = 5) compared to age-matched controls. 

Figure 4

 

Immunostaining of HO-1 in the retinas of db/m and db/db mice. The HO-1 expression levels in the retinas of 8-, 12-, and 20-week-old db/m and db/db mice were determined by immunofluorescence staining using a specific HO-1 antibody; the thickness of each retinal section was 10 μm. (A) Representative micrographs of retinal sections stained with HO-1. (B) Quantitative analysis of HO-1 expression in the RGC layer. The fluorescence intensity of HO-1-labeled neurons was quantified (means ± SEM, n = 5). Compared to db/m mice, HO-1 expression was significantly increased in the retinas of 8-week db/db mice (***P < 0.001, by two-way ANOVA with Bonferroni posttests, n = 5), whereas expression was significantly decreased in the retinas of 20-week-old db/db mice (**P < 0.01, two-way ANOVA with Bonferroni posttests, n = 5).

Figure 5

 

Increased expression at 8 weeks and decreased expression at 20 weeks of HO-1 in the retinas of db/db mice. Retinas were harvested freshly from 8-, 12-, and 20-week-old db/m and db/db mice according to the methods described in the article. HO-1 expression in the whole retina was determined by Western blotting. (A) Representative immunoblot of HO-1 expression in the whole retina. (B) The densitometric analysis of HO-1 expression relative to the loading control (means ± SEM, n = 5). Compared to db/m mice, HO-1 expression was increased significantly in the retinas of 8-week-old db/db mice (**P < 0.01, by two-way ANOVA with Bonferroni posttests, n = 4), while it was decreased in the retinas of 20-week-old db/db mice (*P < 0.05, two-way ANOVA with Bonferroni posttests, n = 4).

Effects of High Glucose on HO-1 Expression in Cultured Retina Explants

To examine whether the increase in HO-1 expression in 8-week-old diabetic retinas was induced by hyperglycemia, we performed an in vitro retinal tissue culture experiment. A representative immunoblot of HO-1 expression in the whole retina is shown in Figure 6A, and the densitometric analysis of HO-1 expression relative to the loading control is displayed in Figure 6B. Treatment with a high concentration of glucose for 24 hours significantly increased HO-1 expression in the retina (ctrl 1.0 ± 0.21, 25G 1.56 ± 0.30, P < 0.05, n = 5). In contrast, the osmotic control, D-mannitol, had no effect on HO-1 expression. 

Figure 6

 

High glucose-induced increased HO-1 expression in cultured retinal explants. Freshly collected retinas were cultured for 24 hours in culture medium consisting of DMEM with 20 mM D-mannitol plus 5 mM D-glucose (25M) or 25 mM D-glucose (25G). (A) Representative immunoblot of HO-1 expression in the whole retina. (B) The densitometric analysis of HO-1 expression relative to the loading control (means ± SEM, n = 5). Compared to the osmotic control, high glucose significantly increased the expression of HO-1 in retinal explants (*P < 0.05, control versus 25G, one-way ANOVA with Bonferroni posttests, n = 5).

Effects of NADPH Oxidase Inhibitors on High Glucose-Induced HO-1 Expression in the Retina

Finally, we asked whether the increase in HO-1 expression in the retina was mediated by NADPH oxidase. The NADPH oxidase inhibitors apocynin and DPI were added to the culture medium along with a high concentration of glucose. Figure 7A illustrates a representative immunoblot of HO-1 expression in the whole retina, while Figure 7B shows the densitometric analysis of HO-1 expression relative to the loading control. Collectively, these data suggested that a 24-hour treatment of both NADPH oxidase inhibitors significantly abrogated the high glucose-mediated induction of HO-1 expression (25G 0.56 ± 0.16, DPI + 25G 0.28 ± 0.09, APO + 25G 0.28 ± 0.09, P < 0.01, n = 5). 

Figure 7

 

NADPH oxidase-mediation of high glucose-induced HO-1 expression in cultured retinal explants. Freshly collected retinas were cultured for 24 hours in culture medium consisting of DMEM with 20 mM D-mannitol plus 5 mM D-glucose (25M) or 25 mM D-glucose (25G) and in the presence of the NADPH oxidase inhibitors apocynin (APO, 100 μM) or DPI (10 μM). (A) Representative immunoblot of HO-1 expression in the whole retina. (B) Densitometric analysis of HO-1 expression relative to the loading control (means ± SEM, n = 5). The high glucose-induced increases in HO-1 expression within retinal explants were significantly attenuated by the NADPH oxidase inhibitors apocynin or DPI (*P < 0.05, 25G vs. 25M; #P < 0.05, 25G + inhibitors versus 25G alone, one-way ANOVA with Bonferroni posttests, n = 5).

Discussion

Our study analyzed the expression of one of the ARE-mediated antioxidants, HO-1, in the retinas of db/db mice and examined the possible role of NADPH oxidase in HO-1 activation. Our data showed that the generation of ROS is increased in the diabetic retina and is accompanied by a parallel upregulation of HO-1 in the early stages of diabetes. However, the expression of HO-1 was attenuated as the disease progressed. Interestingly, one of the NADPH oxidase subtypes, Nox4, also showed a similar pattern of activation. To the best of our knowledge, our study is the first to investigate the expression of ARE-regulated antioxidant pathways and addresses a possible role for NADPH oxidase in the mediation of HO-1 expression in the retina of type 2 diabetic mice. 

Our previous study showed that oxidative stress has a pivotal role in mediating neuronal degeneration within diabetic retinas. ⁸ Increased oxidative stress can occur through increased generation of ROS and/or decreased antioxidant capacity. ¹ Consistent with our previous data, our study demonstrated that an increase in the generation of ROS may have an essential role in the neural degeneration of diabetic retinas. Moreover, our study also revealed that a decrease in antioxidant capacity during the late stages of diabetes also may contribute to the neural degeneration of diabetic retinas. 

As shown previously ^7,8,29 and confirmed in our study, db/db mice have hyperglycemia at 8 weeks of age. Hyperglycemia-induced oxidative stress has been documented by numerous studies. ^1,30,31 In our current study, the hyperglycemia-induced generation of ROS was increased in the retinas of db/db mice at ages of 8, 12, and 20 weeks compared to age-matched db/m mice. In the early stage of diabetes, hyperglycemia-induced oxidative stress is associated with ARE-mediated gene transcription. There is a general consensus that transient and relatively low levels of ROS can act as second messengers in cells. ^17,32 Consistently, HO-1 was temporarily upregulated by hyperglycemia/high glucose in our study. The increase in HO-1 expression at 8 weeks of age implies that early diabetic retinas have an adaptive antioxidant capacity that is sufficient to counteract the increased oxidative stress induced by hyperglycemia. However, as the disease progresses, this antioxidant capacity is attenuated. Recent studies have shown that aging per se led to a progressive increase in oxidative stress and a decrease in the Nrf2 target genes, including HO-1. ³³ Consequently, aging in combination with persistent hyperglycemia may lead further to an even higher propensity for diabetic-induced retinal neurocircuitry injury in the human adult population due to the downregulation of the ARE pathway. 

Our previous data have shown that, at 8 weeks of diabetes, RGCs do not show functional changes, although these mice have hyperglycemia and show an increase in ROS generation at this age. However, at 20 weeks of age, the visual function of RGCs has deteriorated substantially. ⁸ The adaptive activation of HO-1 in the 8-week-old diabetic retina may contribute, at least partially, to maintaining the normal function of RGCs at this stage. The increased induction of HO-1 consequently may counteract the increased oxidative stress induced by hyperglycemia and eliminate the deleterious effects of ROS on RGCs, thus maintaining normal visual function. However, as the impact of hyperglycemia accumulates, by 20 weeks of age, the ARE-mediated antioxidant pathway in the retina becomes damaged, as demonstrated by the decrease in HO-1 expression observed in our study, and results in a consequent loss of RGC function. Our findings are consistent with a recent study in streptozotocin (STZ)-induced diabetic rats in which the induction of HO-1 protected RGCs from further damage. ³⁴ ARE regulation is one important antioxidant pathway that helps maintain the redox status of tissues ^35,36 ; HO-1 is one of the antioxidants regulated by ARE. ³⁷ Decreased activation of HO-1 has been found in tissues of diabetic patients. ^22–24 The therapeutic effect of insulin on DR is mediated by the overexpression of HO-1 in the retina, ²⁷ and the induction of HO-1 via viral delivery in the retina attenuates photoreceptor apoptosis in an experimental model of retinal detachment. ²⁶ Together, these findings suggest that the induction of HO-1 can restore retina function. Thus, it is plausible that a pharmaceutical induction of HO-1 may be an efficient method of protecting RGCs in type I and type II diabetic retinopathy. 

The antioxidant effects of HO-1 are mediated by the generation of carbon monoxide, biliverdin, and its metabolite bilirubin, and the sequestration of redox active iron by ferritin. ²⁸ The regulation of HO-1 gene expression occurs on multiple levels and is inducer-specific. ^38–40 At the transcriptional level, HO-1 is mediated by the transcription factor Nrf2. ^37,41 Under physiologic conditions, Nrf2 is sequestered in the cytosol by Keap1 and is targeted for proteasomal degradation. ^42,43 In the presence of electrophiles or ROS, Nrf2 is released from Keap1 and then translocates into the nucleus, activating the transcription of target genes, such as HO-1. NADPH oxidase-derived ROS contribute to the AGE-induced adaptive activation of HO-1 in the endothelium ³¹ and bradykinin-induced HO-1 activation in the brain. ⁴⁴ Among the seven subunits of NADPH oxidase, Nox2 and Nox4 have been demonstrated to be involved in oxidative stress in type 1 and type 2 diabetic retinopathy, ^14,20,21 which is consistent with the results demonstrated in our study. Notably, and as shown in our study, the inhibition of NADPH oxidase abrogates high glucose-stimulated HO-1 expression in cultured retina explants, suggesting that NADPH oxidase is involved in HO-1 activation in the retinas from db/db mice. Moreover, the activation pattern of Nox4, but not Nox2, was similar to that of HO-1 in the diabetic retina, suggesting that the increase in HO-1 expression was potentially mediated by NADPH oxidase 4. Nox4 may have a protective role in retinal neurons by mediating the activation of the ARE-antioxidant pathway, whereas Nox2 may have a detrimental role in retinal dysfunction because it was activated continuously in the diabetic retina. This conclusion is supported strongly by similar findings in other tissues. Studies have shown that Nox4 produces mainly hydrogen peroxide instead of superoxide, which is different from other Nox isoforms, such as Nox1 and Nox2, ^45–47 and Nox4-derived hydrogen peroxide has pro-survival effects to retina-derived cells. ⁴⁷ The induction of Nox4 in the vasculature has been documented to be an adaptive stress response to pathophysiologic insults that induce the activation of Nrf2-mediated ARE-antioxidants, ^32,48 whereas Nox2 has been identified as a mediator of vascular dysfunction. ^15,20,49 However, further studies using specific NADPH oxidase inhibitors, such as small interfering RNAs, for different subunits of NADPH oxidase are essential to forming final conclusions as to whether Nox4 and/or other NADPH oxidase subunits are involved in the adaptive activation of HO-1 in early diabetic retinas. 

Apart from HO-1, other Nrf2/ARE-mediated antioxidants, such as NAD(P)H:quinone oxidoreductase 1 (NQO1) and γ-glutamylcysteine synthetase, have been reported to be involved in oxidative stress-related retinal diseases. ^50,51 It is plausible that these factors may have a different temporal response than HO-1 in trying to protect the retina. Thus, it would be necessary to investigate the expression patterns of these antioxidants in db/db mouse retinas. 

In summary, our findings suggested that the expression of HO-1 is upregulated in the early diabetic retina in response to hyperglycemia, and this upregulation is mediated partially by NADPH oxidase. However, the persistence of hyperglycemic insults diminished the expression of HO-1, suggesting that long-term hyperglycemia leads to an increase in ROS generation and decreased antioxidant capacity. 

Acknowledgments

Supported by National Basic Research Program of China (973 Program, 2009CB320900 [MP] and 2011CB510206 [MP]), and National Science Foundation of China grants (30831160516 [MP] and 81200691 [MH]), and Ministry of Education of China (20090001120075 [MH]). 

Disclosure: M. He, None; H. Pan, None; C. Xiao, None; M. Pu, None