Topically Administered NOX4 Inhibitor, GLX7013114, Is Efficacious in Treating the Early Pathological Events of Diabetic Retinopathy

Chronic exposure to hyperglycemia and other causal factors, such as hypertension and lipid abnormalities, are believed to initiate a cascade of biochemical and physiological changes that ultimately lead to microvascular damage and retinal dysfunction observed in diabetic retinopathy (DR). Proliferative DR (PDR) remains the most common complication of diabetes and the leading cause of preventable blindness in the working-age population in developed countries. Its clinical hallmarks include the breakdown of the blood-retinal barrier (BRB), vascular microaneurysms, endothelial cell proliferation, and neovascularization (in the late stage). Mild nonproliferative DR (early-stage DR [ESDR]) precedes PDR and is characterized by retinal neurodegeneration, glial activation, vasculopathy, and leaky vessels (1,2).

The American Diabetes Association has defined DR as a highly specific neurovascular complication, referring to the diabetes-induced disruption of the homeostasis of the neurovascular unit, comprising retinal neurons, glia, and the vasculature (3). Α close interdependency exists among the members of this triad. Results of studies of experimental animals and patients with diabetes have suggested that neurodegeneration and neuroinflammation (i.e., imbalance of neurotrophic factors and proinflammatory cytokines) are critical components in the pathophysiology of DR and underlie the mechanisms leading to retinal cell death, suboptimal visual acuity, and blindness (1,2,4–7). Many of these biochemical defects are present as early events in DR, prior to the development of microangiopathy and neovascularization (6–9).

Currently, there are no pharmacological treatments for the early events of DR. Glycemic control, by insulin or hypoglycemic agents, is the only approach to prevent or delay the development of the advanced stages of DR. Fenofibrate reduced the progression of type 2 diabetes in clinical trials and has been recommended for the prevention of DR progression in patients with preexisting disease. However, even though fenofibrate treats neurodegeneration and microangiopathy, there are no clinical trials to support its use as treatment of the ESDR (2,6). The only approved medications, anti–vascular endothelial growth factor (anti-VEGF) agents and corticosteroids, are indicated to treat diabetic macular edema, a feature of the advanced stages of DR. These treatments are associated with adverse effects and do not reverse visual loss (2).

Hyperglycemia affects all members of the neurovascular unit that are responsible for the development and also the progression of DR. Therefore, more therapeutic strategies aiming at the treatment of the early events of DR are essential to prevent or compromise the appearance of PDR and vision loss.

ESDR models have been used to investigate the pharmacological profile of drugs, administered topically, in terms of protection or prevention of the early events of DR (9,10). Peptides, such as somatostatin and glucagon-like peptide, administered topically, prevented retinal neurodegeneration in diabetic rats (treated with streptozotocin [STZ]) and retinal inflammation and vascular permeability in the db/db mouse, respectively (10,11).We have recently reported that the small microneurotrophin molecule BNN27, administered intraperitoneally or topically, prevented retinal neurodegeneration and neuroinflammation in the STZ model of DR (12,13). These findings suggest that neurodegeneration, inflammation, and vascular permeability represent early events in DR, in accordance with studies of human patients with diabetes that have linked these pathologies with the development of the diabetic insult in retina (4).

The family of NADPH oxidases generates reactive oxygen species (ROS) (e.g., superoxide, hydrogen peroxide) that are responsible for the induction of oxidative stress (OS), which subsequently leads to neurodegeneration and neuroinflammation, manifestations of toxic events present in many brain and ocular pathologies (14–16).

The NOX family includes seven known isoforms, namely NOX1 through NOX5, DUOX1, and DUOX2 (16). The expression of these isoforms varies in the central nervous system (i.e., in brain and retina). It has been reported that upon ischemic stroke, NOX4, located in endothelial cells and neurons, caused the breakdown of the blood-brain barrier and neuronal cell death (17). In retina, NOX isoforms have been mainly associated with the development of vascular dysfunction (18,19) and inflammation (20,21). ΝΟΧ1, NOX2, and NOX4 are expressed in rat retinal ganglion cells, endothelial cells, pericytes, and macro- and microglia (19).

We recently investigated the role of NOX inhibitors in the in vivo retinal model of α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) excitotoxicity (22). We reported that the excitatory amino acid AMPA increased OS and the levels of activated macro- and microglia. The pan-NOX inhibitor VAS2870 and the NOX1 inhibitor ML171 reversed the AMPA-induced increase in macro- and microglia activation and protected brain nitric oxide synthetase (bNOS)–expressing amacrine cells. GLX7013114 (a NOX4-specific inhibitor) protected retinal OS and neuroinflammation damage but had no effect on neural dysfunction (22). These findings suggested that NOXs (1,2,4) are important players in the pathophysiology of retinal disease, whose pathophysiology involves excitotoxicity and OS that lead to alterations in retinal neurons and glia.

Glutamate excitotoxicity and OS are two major defects in DR. Diabetes increases glutamate levels in humans and rats and alters glutamate uptake, thus interfering with retinal glutamate homeostasis, leading to retinal cell death (23–25). The relationship between OS and diabetes has also been addressed. Du et al. (26) reported that 2-month-old diabetic mice increased retinal superoxide production via various mechanisms that included NOX signaling. Apomycin (a nonselective NOX inhibitor) reversed superoxide levels in retina. In addition, an 8-month treatment with apomycin reduced the diabetes-induced increase in retinal superoxide levels and capillary degeneration.

NOX4 was overexpressed in db/db mouse retina, as well as in hypoxia and in high-glucose treated primary bovine retinal capillary endothelial cells (18). More recently, Ahmad et al. (27) reported that apomycin reversed the NOX4-induced oxidative damage in the diabetic rat retina and in hypoxic human retinal Müller cells via mechanisms involving the inhibition of survival molecules SIRT and BDNF or the activation of caspase-3. The involvement of the NOX4 gene in severe DR of type 2 diabetes was also supported by a recent genome-wide association study (28). GLX7013114 was reported to improve β-islet mitochondrial activity and survival in a model of short-term stress, suggesting that it may be a putative therapeutic for type 2 diabetes (29). The results of the aforementioned studies suggest that NOX4 inhibition is an important strategy to pursue for developing therapeutics for diabetes and DR.

In the present investigation, we used two paradigms of ESDR to assess the neuroprotective, anti-inflammatory, and antivascular actions of the novel NOX4 selective inhibitor GLX7013114 when administered topically as eye drops. NOX4 inhibition attenuated or blocked the following diabetes-induced actions: 1) OS (nitrative), 2) neuronal cell death, 3) activation of macro/microglia, 4) release of proinflammatory cytokines, and 5) vascular leakage. These results suggest that GLX7013114 eye-drop treatment is efficacious for the treatment of the early pathological events of DR.

Adult male and female Sprague-Dawley rats (200–300 g) were maintained at 22°C on a 12-h light/dark cycle and had ad libitum access to food and water. Euthanization was performed with CO₂ inhalation. All experiments were conducted following the Statement for the Use of Animals in Ophthalmic and Vision Research, in compliance with Greek national laws (Animal Act, P.D. 160/91), the European Union Directive for animal experiments (2010/63/EU), and the 3Rs principle (i.e., replacement, reduction, and refinement).The protocol was approved by the Animal Care Committee assigned by the local Veterinarian Authorities (project authorization 207608).

For the induction of diabetes, a single dose of STZ (70 mg/kg; Sigma-Aldrich, Steinheim, Germany), dissolved in citrate buffer (0.1 M, pH 4.7), was intraperitoneally administered to animals (STZ group) after an overnight fasting period. Control animals received 0.1 M citrate buffer. Based on the higher sensitivity to STZ and subsequent death of male rats, most of the STZ-treated rats were female (12). At 48 h after STZ administration, rats with blood glucose levels >300 mg/dL were considered diabetic (12).

Animals were divided into three separate groups: control, diabetic nontreated, and diabetic treated. Two different paradigms of the DR model were used to assess the early events of DR. In paradigm A (preventive treatment), the treatment was applied over 2 weeks. At 48 h after STZ administration (10), GLX7013114 (provided by Glucox Biotech AB, Stockholm, Sweden) was administered to diabetic rats topically (eye drops [10 mg/mL, 20 μL/eye] dissolved in DMSO) once daily for 14 days. Paradigm B (“treated”) was a 5-week model in which mice received GLX7013114 (10 mg/mL/eye) 4 weeks after STZ injection, for 7 consecutive days, once daily (eye drops [20 μL/eye] dissolved in DMSO) (12,13). In each case, the diabetic nontreated group and the control group received 20 μL of vehicle (DMSO) in each eye, as eye drops, once daily for the same duration as the treatment of the diabetic treated group (14 or 7 days, respectively).

At 24 h after the last day of treatment, animals were euthanized and their eyes enucleated, cryoprotected, frozen, and sectioned (12). Vertical serial sections (10-μm thick) were collected consistently from the central retina and were placed consecutively on six slides, each containing eight sections in total (30).

Primary antibodies raised against nitrotyrosine (NT), nerve fiber layer (NFL), bNOS, glial fibrillary acidic protein (GFAP), ionized calcium-binding adaptor molecule-1 (Iba-1), and cleaved caspase-3 were used. Sections were incubated overnight with each specific primary antibody, diluted in 0.1 mol/L Tris-buffered saline, containing 0.5% normal goat serum and 0.3% Triton-X-100, followed by incubation with secondary antibody, CF543 goat anti–rabbit IgG, for NT, bNOS, Iba-1, and caspase-3 immunoreactivity (IR); and CF488A goat anti–mouse IgG for NFL and GFAP IR, for 1.5–2 h at room temperature, and cover slipped with mounting medium containing DAPI (Biotium, Fremont, CA). All the antibodies used in the immunohistochemical (IHC) studies are listed in Supplementary Table 1.

Lysates of retinal tissues were prepared according to Ibán-Arias et al. (12). Briefly, retinal tissues were homogenized and sonicated, and lysates were analyzed by SDS-PAGE (12.5% acrylamide). Specific primary antibodies raised against VEGF and GFAP were used for incubation of the membranes overnight at 4°C. For the normalization of protein content in retinal lysates, the membranes were incubated with a specific antibody raised against GAPDH. Secondary antibodies were used for VEGF, GFAP, and GAPDH, and protein visualization was performed using LumiSensor Chemiluminescent HRP Substrate kit (Genscript, Piscataway, NJ). The optical density of the protein bands was quantified with ImageJ 1.44 software. Additional antibody information can be found in Supplementary Table 1.

Total RNA was isolated from retinal samples after the TRIzol (Invitrogen, catalog 15596026) extraction protocol and the subsequent Turbo DNase (Ambion, AM2238) treatment, and the total RNA was used for cDNA synthesis using the PrimeScript 1st Strand cDNA Synthesis Kit (Takara Bio, catalog RR037A), according to the manufacturer’s instructions. The expression levels of the different transcripts were quantified with the KAPA SYBR FAST qPCR Kit Master Mix (2×) (Kapa Biosystems, catalog KK4602) and a BIO-RAD CFX Connect Real-Time PCR System, normalized to β-actin, and calculated by the comparative cycle-threshold method. The sequences of the primers used in this analysis are listed in Supplementary Table 1.

TNF-α levels were determined in retinas of diabetic nontreated and diabetic treated rats using an ELISA kit (Abcam, ab100785), following the manufacturer’s instructions. Total protein concentration of each sample was measured using NanoDrop 2000 (Thermo Fisher Scientific) and duplicates of the samples were analyzed by an ELISA reader (BIO-RAD, model 680). TNF-α levels in each sample were normalized to the total protein of the sample.

The Evans blue (EB) assay was used for the evaluation of BRB permeability in the 5-week model of DR. Twenty-four hours after the last day of treatment, animals were anesthetized by intraperitoneal injection of a mixture of ketamine (100 mg/kg) and xylazine (10 mg/kg). EB dye was diluted in saline 0.9% (30 mg/mL), and the EB solution for each animal was prepared at a dose of 45 mg/kg and injected intravenously via the lateral tail vein. Two hours after EB injection, blood samples were collected by cardiac puncture with 19G needles, animals were perfused with 0.9% saline at 37°C to remove excess dye from the circulation, and retinas were collected and weighed. Retinal samples were incubated for 18 h in formamide (300 μL at 72°C) and then centrifuged for 1 h at 22,000g (4°C). Blood samples were centrifuged at 2,100g for 10 min at 4°C immediately after the isolation and plasma samples were diluted in formamide (1:1,000). Retinal extracts and plasma samples were used as duplicates and analyzed by a spectrophotometer at 620 nm (maximum absorbance of EB) and 740 nm (minimum absorbance of EB), as previously mentioned (31). BRB leakage was calculated in micrograms of EB using the following equation: BRB leakage = (retina EB concentration/plasma EB concentration)/retina wet weight.

The 5-week model (paradigm B) was chosen to assess pattern electroretinography (PERG) measurements of retinal ganglion cell (RGC) function. We performed a preliminary proof-of-concept study using a low volume (10 μL/eye) of formulation GLX7013114 (10 mg/mL), because the appropriate volume to be instilled as a single shot in the cul-de-sac of rat eye is between 10 and 20 μL. Twenty-four hours after the last day of treatment, PERG was performed in rats according to Porciatti et al. (32). Anesthetized Sprague-Dawley rats were transferred on a heating plate with the rat superior incisor teeth hooked to a bite bar and the head gently restrained by two ear knobs. The body was kept at a constant temperature of 37°C using a feedback-controlled heating pad (TCAT-2LV, Physitemp Instruments, Inc, Clifton, NJ). To prevent corneal dryness, topical balanced salt solution (10 μL) was applied. Simultaneous recordings of PERG response from both eyes were obtained using a common subcutaneous needle in the snout (Jorvec Corp, Miami, FL). To obtain PERG records, visual stimuli (black and white horizontal bars generated on light-emitting diode tablets) are presented independently to each eye at a distance of 10 cm (56° vertical × 63° horizontal field; spatial frequency, 0.05 cycles/degree; 98% contrast; 800 cd/m² mean luminance; left-eye reversal rate, 0.992 Hz; right-eye reversal rate, 0.984 Hz). Electrical signals recorded were averaged (>1,110 epochs), and PERG responses from each eye were isolated by averaging at stimulus-specific synchrony. PERG waveforms consist of a positive wave (defined as P1) followed by a slower negative wave with a broad trough (defined as N2). Therefore, each waveform is analyzed by measuring the peak-to-trough (P1-N2) amplitude defined as the PERG amplitude and the time-to-peak of the P1 wave defined as PERG latency (32).

Healthy control rats (n = 4 per each time point) were used in the GLX7013114 pharmacokinetic study. They were randomly divided in seven different groups, corresponding to the seven different time points that were included in the study, 0.25 h to 24 h. GLX7013114 (20 μL, 10 mg/mL) was administrated to each eye, and retinas were collected at the preassigned time points. Liquid chromatography–tandem mass spectrometry (LC-MS/MS) analysis was performed, and retinal GLX7013114 levels were analyzed in a time-dependent manner. Briefly, retinas were homogenized, centrifuged at 10,000g for 10 min at 5°C, and then the samples were diluted 1:1 with 50 μL acetonitrile/ to deionized water and 1:1 100 μL acetonitrile to /deionized water containing 100 nmol/L warfarin (internal standard). LC-MS/MS analysis was performed using a SciexQtrap 6500 (ESI negative mode) coupled to a Waters Acquity UPLC.

Light microscopy images were acquired using a Leica DMLB fluorescence microscope (HCX PL Fluotar, ×20/0.50 or ×40/0.75 lens; Leica Microsystems) with a Leica DC 300Fcamera. The number of bNOS⁺ amacrine cells was manually counted along the entire length of three different retinal sections. NT-, caspase-3-, NFL-, GFAP-, and Iba-1-IR were analyzed in two different images from central retina, three retinal sections per sample (six images per retina). The public domain ImageJ software (version 1.44) was used to measure 1) mean gray value of NFL-IR [ganglion cell layer (GCL) to inner plexiform layer (IPL)] and GFAP-IR [GCL to outer nuclear layer (ONL)] and 2) retinal thickness (GCL to IPL; three different sites per section). The total numbers of NT⁺ and caspase-3⁺ cells, as well as the number of the activated Iba-1⁺ cells, were manually counted in two images from three different sections for each retina. Cell counting was followed by normalization to the total counting area: NT-IR[GCL-retinal pigment epithelium (RPE)], caspase-3 and Iba-1-IR[GCL-inner nuclear layer (INL)]. In all studies, data were expressed as a percentage of the control samples (100%).

All data were expressed as the mean ± SD and were analyzed, using GraphPad Prism, version 5.0 (GraphPad Software, San Diego, CA), by two-tailed unpaired t test or one-way ANOVA, followed by Newman-Keuls or Tukey post hoc analysis. The level of statistical significance was P < 0.05.

All data generated or analyzed during this study are included in the published article and the Supplementary Material.

Nitration of tyrosine residues is a basic form of OS-induced damage of tissue proteins. Two weeks after the initial STZ administration, the number of NT⁺ cells was significantly increased (P < 0.001) in the diabetic retinas in the INL and GCL compared with the control retinas (Fig. 1A and B). NT⁺ cells were also observed in the RPE. Topical GLX7013114 (10 mg/mL, 20 μL, once per day for 2 weeks) treatment, beginning 2 days after STZ injection (paradigm A) significantly reduced the number of NT⁺ positive cells in the diabetic treated retinas (P < 0.001) compared with diabetic nontreated retinas (Fig. 1A and B). A similar effect was observed in paradigm B, where the 7-day treatment with GLX7013114, at 4 weeks after STZ administration, reduced the number of NT⁺ cells in rat retina (Supplementary Fig. 1).

Real-time PCR analysis revealed a statistically significant increase in the relative NOX4 mRNA expression in the diabetic nontreated rat retinas compared with control retinas (paradigm A; P < 0.05) (Fig. 1C). This finding suggests that the NOX4 isoform plays an important role in the diabetic retina as early as 2 weeks after STZ administration.

Cleaved caspase-3-IR was used to assess the antiapoptotic properties of NOX4 inhibitor. An increase in the number of caspase-3⁺ cells was observed mainly in the INL of diabetic retinas in paradigm A and in the INL and GCL in paradigm B that was statistically significant compared with control retinas (P <0.001) (Fig. 2A–C). GLX7013114 (10 mg/mL; 20 μL) reversed the diabetes-induced increase in caspase-3-IR in diabetic retinas in both paradigms (paradigm A: P < 0.05 compared with the control, P < 0.01 compared with diabetic nontreated group [Fig. 2B]; paradigm B: P < 0.001 compared with diabetic nontreated group [Fig. 2C]) (Supplementary Fig. 2).

IHC studies were performed using antibodies raised against NFL and bNOS, markers of retinal ganglion cell axons and NOS containing amacrine cells, to assess the neuroprotective actions of GLX7013114 on neural elements. Diabetes attenuated NFL-IR in the 2- and 5-week models (Fig. 3A and B), and these effects were blocked by GLX7013114 (paradigms A and B: P < 0.001 compared with the control and with the diabetic nontreated group) (Fig. 3C and D). Quantification of the area corresponding to NFL immunostaining in layers GCL-IPL is also shown (Fig. 3D and F). In both models, diabetes was linked with a significant reduction of the area corresponding to NFL-IR, an effect reversed by GLX7013114 (for the 2- and 5-week models: P < 0.001 and P < 0.01, respectively, compared with control; and P < 0.05 and P < 0.01, respectively, compared with the diabetic nontreated group) (Fig. 3D and F).

PERG, a sensitive measure of RGC function, was also used in paradigm B. PERG amplitudes of the control, the STZ group (diabetic nontreated), and the GLX7013114 group (10 μL/eye, 10 mg/kg diabetic treated group) were compared (Fig. 4). The average value of control PERG amplitude was 16.2 μV, in agreement with previous studies in rats (33). The average PERG amplitude in retinas of diabetic rats was significantly reduced compared with control retinas (>50%, P < 0.001, P < 0.0001 compared to control), whereas PERG amplitude values of retinas of GLX7013114-treated rats were significantly higher when compared with the diabetic nontreated retinas (Fig. 4) (P < 0.01 compared with diabetic nontreated group), suggesting a protection of RGC function (Fig. 4B). No statistically significant changes were observed in the latency values among the three groups (P < 0.05 compared with the control) (Fig. 4C). On the basis of these positive findings and in consideration of the 3Rs principle, no further studies were carried out at higher volume of formulation.

Diabetes also attenuated bNOS⁺ somata located in the INL and processes in the IPL. NOX4 inhibition afforded partial protection to the bNOS⁺ amacrine cells against the diabetes insult in both models of DR (in the 2- and 5-week models: P < 0.01 and P < 0.001 compared with control; and P < 0.05 and P < 0.01, respectively, compared with the diabetic nontreated group) (Supplementary Fig. 3).

Retinal activated micro- and macroglia are major sources of inflammatory mediators leading to neuroinflammation (34). To ascertain whether GLX7013114 has the pharmacological profile of an anti-inflammatory agent useful for the treatment of DR, we initially examined its effect on the diabetes-induced activation of micro- and macroglia and, subsequently, the release of proinflammatory cytokines in both paradigms of DR.

Iba-1 (a microglial marker) was used to assess the effect of NOX4 inhibition on microglia in the diabetic retina. Diabetes induced an increase in the number of activated Iba-1⁺ cells in rat retina, an effect that was attenuated by GLX7013114 in both paradigms (for paradigms A and B: P < 0.05 and P < 0.001, respectively, compared with control; and P < 0.001 compared with the diabetic nontreated group) (Fig. 5A–C and Supplementary Fig. 3).

Diabetes also increased GFAP expression. GLX7013114 attenuated this effect in a statistically significant manner (paradigm A: P < 0.001 compared with control, and P < 0.01 compared with the diabetic nontreated group) (Fig. 6A and B). In agreement with the IHC results, Western blot analysis also showed that GFAP expression was reduced in the presence of GLX7013114 (P < 0.01 compared with control and compared with diabetic nontreated retinas) (Fig. 6C). A similar observation was observed in paradigm B, in which 7-day treatment with GLX7013114 reduced activated macroglia (GFAP-IR) (P < 0.05 and P < 0.001 compared with control; P < 0.01 compared with the diabetic nontreated group) (Fig. 6D and Supplementary Fig. 5).

As a result of diabetes-induced increase of macro- and microglia activation, we observed alterations in the expression levels of proinflammatory cytokines TNF-α, IL-1β, and IL-6 in the rat diabetic retina. Two weeks after STZ administration, retinal protein levels of TNF-α in the diabetic untreated animals were upregulated. GLX7013114 reduced the diabetic insult statistically significantly (P < 0.05 compared with the diabetic nontreated group) (Fig. 7A). In the 5-week model, real-time PCR analysis revealed a statistically significant increase in IL1-β and IL-6 mRNA levels in the diabetic nontreated retinas compared with control (P < 0.001 and P < 0.05, respectively, compared with the control) (Fig. 7B and C). GLX7013114 attenuated the diabetes-induced increase in IL1-β and IL-6 mRNA levels (P < 0.001 and P < 0.05, respectively, compared with the diabetic nontreated group) (Fig. 7B and C).

Western blot and real-time PCR analysis was also used to assess the effect of NOX4 inhibition on VEGF protein and mRNA expression, respectively, in both paradigms. Two weeks after STZ administration, diabetes caused a small but statistically significant increase in protein levels of VEGF in the retina, an effect that was not altered by GLX7013114 (P < 0.05 compared with control) (Fig. 8A). However, in paradigm B, 7-day GLX7013114 treatment, 4 weeks after STZ administration, statistically significantly attenuated the protein levels of VEGF (P < 0.05 compared with the control and with the diabetic nontreated group) (Fig. 8B). The latter finding is in agreement with data obtained with real-time PCR analysis that support GLX7013114 attenuation of VEGF mRNA levels (P < 0.05, P < 0.01 compared with the control; P < 0.05 compared with diabetic nontreated group) (Fig. 8C).

The EB assay was used for the evaluation of BRB permeability in the 5-week model of DR. Diabetes induced a significant (P < 0.01) retinal vascular leakage increase, as indicated by the increase in EB dye levels (Fig. 8D). GLX7013114 administration significantly reversed this effect (P < 0.01 compared with the control retinas and with diabetic nontreated retinas) (Fig. 8D).

The pharmacokinetic data show that GLX7013114 (10 mg/mL), administered as eye drops (20 μL/eye) to control rats reached the retina as soon as 15 min. Its highest concentration in retinal tissue was observed 30 min after the initial administration and was equal to 17.97 (SD ±13.53) pmol/mg tissue (approximately 1 μmol/L), followed by a quick reduction of the detectable levels (8 h; 1.25 pmol/mg retina or 63 nmol/L; P < 0.01 compared with the concentration at 30 min) (Supplementary Fig. 6).

NOXs have emerged as novel targets for the development of new therapeutics for the treatment of neurodegenerative disease (i.e., in brain and retina). In the present study, we show that GLX7013114, a novel NOX4 inhibitor, has the pharmacological profile of a putative therapeutic for the treatment of the early events of DR.

GLX7013114 is characterized as a small molecule, a selective NOX4 inhibitor (IC₅₀, 0.3 μmol/L), with no affinity for other NOX1, NOX2, and NOX5 isoforms present in retina (35). It has no scavenging or assay interference properties in comparison with other NOX inhibitors reported in the literature. GLX7013114 appears to be the only compound tested so far that fulfills all the criteria that dictate the specificity of NOX inhibitors (35,36).

In two variations of the in vivo STZ rat model of DR, namely the 2-week (preventive treatment; paradigm A) and 5-week (4 weeks after onset of diabetes and 1 week treatment; paradigm B), diabetes induced an increase in nitrative stress. The colocalization of the NT stain with the nuclear marker DAPI suggested that part of the OS damage is neuronal. However, NT-IR stain may also be located in vessels because NOX4 has been reported in retinal endothelial cells in addition to neurons, glia, and pericytes (37). We were unable to observe colocalization of NT and CD31 (a marker of endothelial cells) for technical reasons. We also observed increased neuronal apoptotic cell death (caspase-3 expression) in the INL, and less in the IPL and GCL, in both paradigms. However, increased caspase-3 expression in the IPL and GCL, 2 weeks after STZ administration, was also reported (38).

Taken together, these findings suggest that GLX7013114, topically administered, blocked the early events of DR in respect to nitrative OS and apoptotic cell death and concur with the antioxidant and neuroprotective effect of GLX7013114 in the retinal AMPA excitotoxicity model (22). In agreement with our results, Ahmad et al. (27) reported that NOX4 inhibition reversed the diabetes effects on OS, ROS levels and caspase-3 expression in hypoxic human retinal Müller cells and diabetic rat retina. Also observed was the attenuation of the diabetes-induced increase in NOX4 mRNA and OS in diabetic (i.e., STZ-treated) retinas by GKT13783(a dual NOX1 and NOX4 inhibitor) (39).

Diabetes neuronal damage was evident in both paradigms of ESDR. GLX7013114 reversed the diabetes-induced attenuation of the number of bNOS-expressing retinal amacrine cells (INL), thus support the neuronal staining of NT in the INL (NT-DAPI colocalization). The depletion of NOS-expressing amacrine cells in the diabetic retina has been reported previously (40) and is in agreement with our present and previous findings in the STZ-DR model (12,13). A similar effect of GLX7013114 was also observed in the reversal or blockade of the diabetes-induced attenuation of the NFL-IR, as well as the thickness or shrinkage of the area corresponding to NFL-IR (GCL and IPL).

PERG has been used as a sensitive measure of RGC function (41). PERG analysis was performed in all groups of paradigm B. The average value of control PERG amplitude was 16.2 μV, which is in agreement with previous studies conducted with rats (33). Animals treated with GLX7013114 (10 mg/mL, 10 μL/eye) exhibited a statistically significant increase in the average value of PERG amplitude compared with the diabetic nontreated group, supporting GLX7013114’s efficacy in the protection of RGC function. On the basis of these positive findings and in consideration of the 3Rs principle, no additional studies were carried out at higher volume of formulation.

All of the aforementioned findings of this study suggest that GLX7013114 provides neuroprotection to retinal neurons (amacrine and RGCs) and RGC axons against the diabetic insults of ESDR.

As the resident inflammatory cells in the retina, microglia modulate inflammatory processes. Microglia exist as two morphological types: resting and activated. Several reports addressed the morphology and activation of microglia (amoeboid state) after the onset of hyperglycemia (42,43). Activated microglia increase the levels of proinflammatory cytokines and caspase-3 expression (44,45). Zeng et al. (43) reported that the number of activated microglia was increased in the IPL of the diabetic retina as early as the first month of diabetes and in the IPL and outer plexiform layer (OPL) at 4 months. On the basis of these data, it was suggested that a paracrine mechanism was involved in the regulation of microglia in the inner retina, where NFL, photoreceptor, and other retinal cell damage was responsible for the increase in the number of activated microglia in the IPL and ONL (42).

Inhibition of the activation of microglia affords neuroprotection to the diabetic retina (42). We report that GLX7013114 attenuated activated microglia, in the diabetic retina (GCL, IPL, INL) of both paradigms of DR (as labeled by white arrows in Fig. 5 and Supplementary Fig. 4). The number of resting microglia was very low and not marked in the relevant images. Our findings are in agreement with the published studies mentioned above and suggest that GLX7013114 reverses the activation of microglia and thus has putative anti-inflammatory properties.

Retinal Müller cells (macroglia) also play an important role in the inflammation process induced by diabetes and are an important source of proinflammatory cytokines, TNF-α, IL-1β, IL-6, and VEGF (34). In our study, NOX4 inhibition attenuated the diabetes-induced increase in GFAP expression in the 2-week diabetic-retina model, as established by IHC and Western blot analysis. Similarly, GLX7013114 attenuated the expression of GFAP in the 5-week model (Fig. 5D and Supplementary Fig. 5).

Our data confirm the allegation that both micro- and macroglia are a source of anti-inflammatory cytokines. GLX7013114 inhibited the diabetes-induced increase of TNF-α protein levels in the 2-week model and mRNA levels of IL-1β and IL-6 in the 5-week paradigm. Other reports have also shown that NOX inhibition by apomycin and GKT13783 (a dual NOX1 and NOX4 inhibitor) attenuated the expression of retinal inflammatory proteins in 2-month-old diabetic mice (26) and the hypoxia-induced upregulation of NOX1 and NOX/4 mRNA levels in primary cultures of rat retinal microglia, Müller and ganglion cells and ROS levels, respectively (21). In the latter study (21), GKT13783 attenuated ROS and VEGF levels released by neuron and glial cell populations [oxygen-induced retinopathy (OIR) rat model].

VEGF is responsible for vascular permeability and neovascularization in the diabetic retina (34). VEGF mRNA expression has been reported in retinal neurons, glia, endothelial cells, pericytes, and RPE (34). An increase in VEGF mRNA expression was reported in glial cells and the optic nerve in human patients with nonproliferative DR (46). VEGF is believed to have a dual role in DR, namely as an initial neuroprotectant of retinal neurons against the hyperglycemia-induced OS and as a facilitator or enhancer of the progression of DR (34,47).

VEGF was reported to be constitutively expressed in Müller cells and to enhance the survival of 1) Müller cells (an autocrine role) and 2) photoreceptors and RGCs (47). Μany reports suggest that neighboring cells, such as neurons and vessels, may also release VEGF and increase its levels, aiding the progression of the disease to microangiopathy. The importance of an autocrine VEGF signaling for vascular homeostasis and the mechanism involved have been investigated. According to Rossino et al. (48), OS instigates a VEGF autocrine loop in Müller cells via a mechanism involving transcription factors responsible for the transcription, translation, and release of VEGF and VEGFR2. The authors suggested that OS not only triggers the VEGF autocrine loop but also sustains it and, thus, is responsible for the early and late stages of DR. VEGF thus appears to be regulated via autocrine and paracrine mechanisms. However, apart from OS, there are other important mechanisms that cause upregulation of VEGF during the development of DR, including hypoxia, chronic hyperglycemia, advanced glycation end products, and proinflammatory cytokines (2).

In the present study, diabetes induced an increase in VEGF levels in both early-stage paradigms of DR. However, GLX7013114 attenuated the diabetes-induced increase in VEGF levels in a statistically significant manner only in the 5-week paradigm. Increased VEGF levels in rat retina have been observed in other studies at 1 week (31) or 2 weeks (49) after STZ administration. In agreement with our data, Li et al. (18) reported a NOX4-dependent upregulation of VEGF in retinas of db/db mice 6 weeks after the initiation of their study. However, other studies have also used the db/db mouse model for the investigation of new treatments for ESDR (9,11). The results of the EB assay in our 5-week paradigm revealed a significant increase of the vascular leakage (EB leakage) in the diabetic nontreated retinas compared with the control. A similar increase was reported after 6 (31) or 7 (50) days after STZ injection. In our model, ocular topical administration of GLX7013114 attenuated vascular leakage in a statistically significant manner (P < 0.01), suggesting a protective role of this NOX inhibitor against diabetes-induced disruption of BRB integrity (Fig. 8D). These findings suggest that GLX7013114 may play a role in the attenuation of VEGF levels and the vascular leakage produced in ESDR.

In the present study, we also focused on the topical route of administration and the efficacy of GLX7013114 as a neuroprotective agent against ESDR damage, when administered via a less-invasive route. Our results support previous studies showing that different agents, topically administered in experimental models of ESDR, are efficacious in attenuating apoptosis and glial activation (9–11,13).

The pharmacokinetic study substantiated that GLX7013114 reached the retina when administered as eye drops and was detectable up to 8 h. These data justify the efficacy of GLX7013114 and suggest that a proper dose regimen may be beneficial for the treatment of DR.

In summary, the findings of this study demonstrate that the novel NOX4 inhibitor, GLX7013114, administered topically as eye drops, was efficacious in treating the pathological events observed in ESDR. Additional preclinical studies are essential to support the significance of GLX7013114 as a new candidate for the treatment of DR.

Acknowledgments. The authors thank Sofia Papadogkonaki, Department of Pharmacology, School of Medicine, University of Crete, for her help in different parts of this project.

Funding. This work was supported by a grant from the University of Crete (ELKE KA 10752 to K.T.); the Christina Spyraki award to S.D.; the Graduate Program of Neurosciences, School of Medicine University of Crete, and a scholarship from the State Scholarships Foundation (no. 2022-050-0502-52089) to D.S. Glucox Biotech AB supported this study by generating, characterizing, and providing NOX inhibitors and by funding a small part of the project.

Funding from Glucox Biotech AB does not alter the company’s adherence to policy on sharing data and materials.

Duality of Interest. On 9 May 2018, P.W. and E.W. submitted European patent application no. 18171556.6, protecting the Nox4 selective compound GLX7013114. The funder Glucox Biotech AB provided support in the form of salary for P.W. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. S.D. performed and designed experiments; acquired, analyzed, and interpreted the data; and contributed to the writing of the manuscript. P.W. and E.W. contributed study materials. C.B., G.L.R., and V.M. conducted the pattern electroretinography study and contributed to data acquisition, analysis, and interpretation. R.S. conducted the pharmacokinetic study and liquid chromatography–tandem mass spectrometry analysis. D.S. performed the EB assay and contributed to data acquisition, analysis, and interpretation. N.M. was responsible for animal breeding, and ELISA data acquisition and analysis. S.G. and P.V. contributed to real-time PCR data acquisition and analysis. K.T. conceived and designed experiments, interpreted data, supervised the project, and contributed to writing the manuscript. All authors read and commented on drafts of the manuscript and the final version. K.T. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this work were presented as posters at the Federation of European Neuroscience Societies (FENS) Virtual Forum, 11–15 July 2020; the Hellenic Society of Neuroscience Meeting, 4–6 October 2019, Heraklion, Crete, Greece; and the Panhellenic Meeting of Basic and Clinical Pharmacology, 7–9 October 2022, Arta, Greece.