Intravitreal injection of adenosine A2A receptor antagonist reduces neuroinflammation, vascular leakage and cell death in the retina of diabetic mice

Diabetic retinopathy is a major complication of diabetes mellitus and a leading cause of blindness. The pathogenesis of diabetic retinopathy is accompanied by chronic low-grade inflammation. Evidence shows that the blockade of adenosine A2A receptors (A2AR) affords protection to the retina through the control of microglia-mediated neuroinflammation. Herein, we investigated the therapeutic potential of an antagonist of A2AR in a model of diabetic retinopathy. Type 1 diabetes was induced in 4–5 months old C57BL/6 J mice with a single intraperitoneal injection streptozotocin. Animals were treated one month after the onset of diabetes. The A2AR antagonist was delivered by intravitreal injection once a week for 4 weeks. Microglia reactivity and inflammatory mediators were increased in the retinas of diabetic animals. The treatment with the A2AR antagonist was able to control microglial reactivity and halt neuroinflammation. Furthermore, the A2AR antagonist rescued retinal vascular leakage, attenuated alterations in retinal thickness, decreased retinal cell death and the loss of retinal ganglion cells induced by diabetes. These results demonstrate that intravitreal injection of the A2AR antagonist controls inflammation, affords protection against cell loss and reduces vascular leakage associated with diabetes, which could be envisaged as a therapeutic approach for the early complications of diabetes in the retina.

was treated with a selective A 2A R antagonist, SCH 58261, by an intravitreal injection every week, for four weeks, starting one month after the onset of diabetes. This experimental protocol allows the study of the therapeutic potential of the A 2A R antagonist, since the retina is already affected by the disease when the treatment is initiated.

Results
Retinal thickness is reduced one month after diabetes induction. All the animals used in this study were monitored for weight and glycemic before diabetes induction (baseline), and 1 and 2 months after the onset of diabetes. Hemoglobin A1C (HbA1C) was measured only in a subset of animals after 2 months of diabetes induction, randomly chosen. Biochemical data and weight of the animals from this sub-set are presented in Table 1.
Retinal thinning has been reported as an early alteration in diabetic animals 24 . Four weeks after the onset of diabetes, retinal thickness significantly (p < 0.0001) decreased in diabetic animals when compared with control animals (Fig. 1), a structural indication of alterations due to diabetes. Since the aim of this work was to investigate the therapeutic potential of a selective A 2A R antagonist for the treatment of diabetic retinopathy, we wanted to start the intravitreal injections after established retinal alterations. Therefore, the A 2A R antagonist was injected intravitreally thereafter every week, in a total of 4 injections.

Treatment with A 2A R antagonist decreases microglia reactivity in the retina of diabetic mice.
A large body of evidence demonstrate that inflammation contributes to diabetic retinopathy and microglia reactivity occurs early in the course of the disease 11 . Microglia were labelled with antibodies against ionized calcium-binding adaptor molecule 1 (Iba-1, a marker of microglia) and major histocompatibility complex class II (MHC-II, highly expressed in reactive microglia) ( Fig. 2A). Therefore, microglial cells (Iba-1 + cells) labeled with MHC-II (MHC-II + Iba-1 + cells) were considered reactive microglia 15 . As expected, microglia reactivity was increased in the retina of diabetic animals (16.1 ± 4.0% of total Iba-1 + cells; p < 0.05) when compared with the retinas of control animals (7.7 ± 1.8% of total Iba-1 + cells) (Fig. 2B). The intravitreal injection of SCH 58261, the A 2A R antagonist, significantly decreased microglial reactivity in the retinas of diabetic animals (6.3 ± 0.6% of total Iba-1 + cells; p < 0.01). The administration of SCH 58261 to non-diabetic animals did not significantly change the number of reactive microglia (5.9 ± 0.9% of total Iba-1 + cells). These alterations in microglia reactivity in diabetic animals were not due to changes in the total number of microglia (Iba-1 + cells) (Fig. 2C). The Translocator protein (18 kDa) (TSPO) is constitutively expressed in retinal microglia and in inflammatory conditions TSPO expression increases 25,26 . Therefore, the effect of SCH 58261 on microglia reactivity was further studied by evaluating the protein levels of TSPO (Fig. 2D). Diabetes significantly increased by 3.4-fold the protein levels of TSPO in the retina of diabetic animals (p < 0.01) when compared with the control condition (saline-injected retinas from non-diabetic animals) (Fig. 2D), and this effect was abolished by the A 2A R antagonist. The treatment of non-diabetic animals with the A 2A R antagonist did not change TSPO protein levels (Fig. 2D). in the retina were determined by Western blot and ELISA (Fig. 3). Several cytokines have been demonstrated to play a role in diabetic retinopathy 5,27 , including TNF and IL-1β. The A 2A R antagonist was reported to decrease the levels of these two cytokines in other experimental models, including in the retina 16,18,28 . Therefore, in the current work we focused on assessing the effect of SCH 58261 on the levels of TNF and IL-1β in the retinas of diabetic animals, although we do not exclude the contribution of other cytokines to the disease progression. The protein expression of TNF was significantly increased in the diabetic retinas when compared with control retinas, as assessed by Western blot (Fig. 3A, p < 0.001). The TNF levels were also quantified by ELISA (Fig. 3B), and a significant increase in the retinas of diabetic animals was found (386.8 ± 110.2 pg/mg protein, p < 0.01), when compared with control retinas (229.9 ± 78.0 pg/mg protein). While the treatment of non-diabetic animals with A 2A R antagonist did not modify the protein levels of TNF, the treatment of diabetic animals slightly reduced the protein levels of TNF. The IL-1β protein levels were also increased by 3.5-fold in the retinas of diabetic animals (p < 0.05) when compared with control retinas (Fig. 3C). Diabetes significantly increased the IL-1β levels in the retinas (2130.5 ± 1140.0 pg/mg protein; p < 0.001) when compared with the control (1492.9 ± 264.2 pg/mg protein) (Fig. 3D). The treatment of diabetic animals with A 2A R antagonist did not modify the levels of IL-1β.
Treatment with SCH 58261 decreases iNOS in the diabetic retinas. An increase in the protein levels of inducible nitric oxide synthase (iNOS) has been reported in the retina in diabetic conditions 29 . The protein levels of iNOS were significantly increased (3-fold change) in the retinas of diabetic animals (injected with saline) when compared with the control condition (p < 0.01, Fig. 4A). The treatment with the A 2A R antagonist significantly decreased the iNOS protein levels (p < 0.05) comparatively to the diabetic retinas treated with saline. The intravitreal injection of SCH 58261 to non-diabetic animals did not change iNOS protein levels. Despite the overall increase in iNOS protein in the retinas of diabetic mice and the decrease by A 2A R blockade (observed both by Western blot and by immunohistochemistry, Fig. 4), we were not able to identify a specific cell type responsible for these alterations since there were no changes in the distribution of iNOS in retinal vertical sections (Fig. 4B).

Administration of the A 2A R antagonist decreases cell death in the retina of diabetic animals.
Diabetes results in the death of retinal cells 30 . It is widely accepted that neuroinflammation triggers cell death, including in the diabetic retina 31 . Retinal cell death was evaluated by TUNEL assay in retinal sections (Fig. 5).
The number of TUNEL + cells significantly increased after 8 weeks of diabetes (15.2 ± 3.9 TUNEL + cells per field, p < 0.05) when compared with the saline injected control retinas (7.2 ± 1.5 TUNEL + cells per field) (Fig. 5A,B). The intravitreal injection of SCH 58261 to diabetic animals significantly decreased diabetes-induced retinal cell death (5.8 ± 0.9 TUNEL + cells per field, p < 0.05). The activation of caspase-dependent signaling apoptotic mechanisms has been associated with the loss of retinal cells in diabetic retinopathy, including retinal ganglion cells (RGCs) [32][33][34] . The presence of active caspases in the ganglion cell layer was evaluated using the CAS-MAP reagent, a probe to detect active caspases 1, 2, 3, 6, 8, 9, or 10 ( Fig. 5C). Diabetes significantly increased the number of apoptotic cells in the ganglion cell layer, as indicated A 2A R antagonist attenuates RGC loss in a mouse model of diabetic retinopathy. Previous findings in the STZ-induced diabetic mouse model have reported loss of RGCs 30,35 . RGCs were labeled with an antibody anti-Brn3a (Fig. 6), a transcription factor that is expressed by these cells 36 . There was a decrease in the number of RGCs after 8 weeks of diabetes (42.8 ± 3.5 Brn3a + cells/mm, p < 0.05) when comparing with control animals (60.2 ± 3.3 Brn3a + cells/mm, Fig. 6A,B). The A 2A R antagonist did not change the number of Brn3a + cells of non-diabetic animals (57.2 ± 2.2 Brn3a + cells per mm). However, the treatment of diabetic animals with the A 2A R antagonist promoted the survival of RGCs (53.6 ± 4.5 Brn3a + cells/mm, p < 0.05).

Intravitreal administration of SCH 58261 inhibits retinal thinning in diabetic animals.
Retinal thinning can be a feature of retinal cell loss associated with diabetic retinopathy progression 37 . Retinal thickness was measured by OCT. Since OCT allows for longitudinal studies, retinal thickness was referred to the baseline and alterations to the baseline were documented (Fig. 7). The thickness of the retinas of control animals remained quite constant throughout the study. However, thinning was found in the retinas of diabetic (8 weeks) animals (p < 0.05 when compared with the baseline; Fig. 7A,B). The treatment of diabetic animals with SCH 58261 for 4 weeks did not statistically modify retinal thickness, when compared with the baseline values (Fig. 7A,B), which might be in accordance with a protective effect of the A 2A R antagonist.

Treatment with A 2A R antagonist reduces retinal vascular leakage in diabetic mice. BRB dys-
function is a hallmark of diabetic retinopathy and can be assessed by fluorescein angiography. Vascular leakage, as a measurement of BRB dysfunction, was assessed in vivo by fluorescein angiography (Fig. 8). In most of non-diabetic mice (79% of animals), fluorescein was maintained within the retinal vessels, indicating absence of vascular leakage. The angiograms of diabetic animals showed regions of vascular leakage and an overall diffuse green haze, indicating extravascular fluorescein, a sign of vascular leakage. The treatment of diabetic animals with the A 2A R antagonist reduced the number of animals with retinal vascular fluorescein leakage (Fig. 8A,B), suggesting that the A 2A R antagonist restored the BRB integrity lost in the course of diabetes. Furthermore, the analysis of the permeability coefficient, obtained from the images of the angiograms over time, shows a tendency www.nature.com/scientificreports www.nature.com/scientificreports/ to an increase in permeability in the retina of diabetic animals, which was not present in the eyes of the diabetic animals treated with the A 2A R antagonist (Fig. 8C). Indeed, the curve profile of the diabetic eyes treated with SCH 58261 was very similar to the control group (Fig. 8C). The area under the curve (AUC) (Fig. 8C), for each group suggested that the treatment with A 2A R antagonist may attenuate the effects of diabetes on vascular leakage. Moreover, the permeability coefficient of the diabetic animals was significantly higher than the control animals and SCH 58261 was able to significantly decrease the permeability coefficient of diabetic animals (at 300 seconds, two-way ANOVA using Fisher's LSD test, p < 0.05).

Discussion
In this work, we demonstrate the therapeutic potential of an A 2A R antagonist for the treatment of the retinal complications in diabetes. Importantly, the treatment of STZ-induced type I diabetic mice with the A 2A R antagonist by intravitreal injection was initiated 4 weeks after the onset of diabetes, when the retina is already compromised 6,29,38 .  Fig. 4 and 42 ). We and others demonstrated that the blockade of A 2A R affords protection to the retina and brain against noxious conditions involving microglia-mediated neuroinflammation control [15][16][17][18][19][20]28,43,44 . The intravitreal administration of SCH 58261 in diabetic animals decreased retinal microglial reactivity. Previously, we demonstrated that blocking A 2A R in microglia prevents the reactivity of these cells and retinal neural cell loss 28 , suggesting that the effects observed herein occur through the modulation of microglia phenotype. One limitation of this study is the use of only one drug to infer on the therapeutic potential of A 2A R   www.nature.com/scientificreports www.nature.com/scientificreports/ Diabetic retinopathy is a low-grade chronic inflammatory disease and a large body of evidence supports the role of inflammation in BRB dysfunction 1,4,51,52 . Diabetes increased the retinal levels of TNF and IL-1β, but the treatment with SCH 58261 only slightly attenuated the expression of TNF. The differential modulation of the levels of proinflammatory cytokines by the A 2A R antagonist might be dependent on the pathological conditions or the animal species being evaluated. In the current work, the treatment with the A 2A R antagonist was initiated four weeks after the onset of diabetes. This may explain the lack of effects of SCH 58261 on reducing the levels of TNF and IL-1β since at four weeks of diabetes the inflammatory environment is already established 29 . In a previous work, we reported that a single intravitreal injection of SCH 58261 prevents the increase in the expression of IL-1β induced by transient retinal ischemia in Wistar rats, without altering the levels of TNF 16 . In contrary, the mouse model of transient retinal ischemia promotes an increase in TNF without interfering with IL-1β retinal levels 15 .
Also, microglial reactivity and neuroinflammation were shown to contribute to retinal degeneration in diabetic retinopathy [9][10][11][12][13] . Caspases are family of cysteine proteases known to be involved in the initiation and execution of apoptosis and have been implicated in the death of RGCs 34 , including in diabetic retinopathy 32,53 . Our results show an increase in the number of cells with active caspases within the GCL. The GCL is composed not only by RGCs but also astrocytes and displaced amacrine cells, and it is estimated that in the rodent retina RGCs comprise 50% of the total GCL population 54,55 . We found a reduction in the number of RGCs, suggesting that diabetes is causing caspase-dependent RGC death. Nevertheless, displaced amacrine cells may also degenerate in diabetes 56 . Moreover, we do not rule out the possibility that active caspases are also present in other retinal layers, since other cells, like photoreceptors are also affected by diabetes [57][58][59][60] . Interestingly, TUNEL staining was mainly observed in the ONL. Several mechanisms orchestrate cell death in the retina during diabetes, and cell death does not occur at the same time window. The absence of TUNEL staining in the GCL may be due to the efficient phagocytosis by microglial cells that become reactive in the course of diabetes 61 and actively phagocyte RGC debris 62 . The A 2A R antagonist afforded protection to the retinas of diabetic animals, namely to RGCs, probably the controlling microglia reactivity, as demonstrated in other noxious conditions 15,16,18,28,63 .
The protein levels of iNOS were increased in the retinas of diabetic animals, as reported previously 29,64,65 . Once expressed, iNOS is constantly active thus producing high amounts of NO 66 , with detrimental effects to the retina, such as increased leukostasis and BRB breakdown 29 . The cells expressing iNOS were not identified but one can speculate, based on previous reports, that iNOS expression occurs mainly in microglia and astrocytes in the retina of diabetic mice 13,67 . In fact, the A 2A R antagonist prevents iNOS increase in microglia triggered by noxious inflammatory conditions 16,18 . In the current work the treatment with the A 2A R antagonist reduced the diabetes-induced expression of iNOS. Taking the role of iNOS to the disruption of BRB in mouse models of diabetic retinopathy 29,64 and our data on the control of microglia-mediated neuroinflammation by A 2A R antagonist, we might postulate that the blockade of the A 2A R improves BRB properties by decreasing microglia reactivity. Nevertheless, it was demonstrated that A 2A R activation mediates pathological angiogenesis in the retina by promoting HIF-1α accumulation and increased glycolysis that lead to endothelial cell proliferation and sprouting 68 , and therefore we cannot discard that A 2A R antagonist may also be targeting endothelial cells.
Our data support the therapeutic potential of A 2A R antagonists for the treatment of diabetic retinopathy. We can suggest the control of microglia-mediated neuroinflammation as the axis that will contribute to a decrease in proinflammatory mediators and nitrosative stress, maintenance of BRB integrity and neuroprotection. The A 2A R emerges as a candidate to be further tested in the context of diabetic retinopathy.

Materials and Methods
Animals. C57BL/6 J mice were housed in a controlled environment (21.8 ± 0.1 °C and 67.6 ± 1.6% relative humidity, 12 hours light/12 hours dark cycle) with free access to water and standard rodent chow. All procedures were approved by the Animal Welfare Committee of the Faculty of Medicine of University of Coimbra (ORBEA:  and conducted in accordance with the European Community directive guidelines for the use of animals in laboratory (2010/63/EU) transposed to the Portuguese law (Decreto-Lei 113/2013) and were also in agreement with the Association for Research in Vision and Ophthalmology statement for animal use.

Immunohistochemistry.
Retinal sections were immunolabeled for Iba-1 and MHC-II as previously described 15 using the antibodies listed in Table 2. Retinal sections were imaged using an inverted fluorescence microscope (Zeiss Axio HXP-120, Zeiss, Oberkochen, Germany) with a 20x objective (Plan Achromat 20 × /0.8 M27). The number of cells immunoreactive to Iba-1 was counted and the number of cells immunoreactive to both Iba-1 and MHC-II (reactive microglia) was normalized to the % of total microglia.

Terminal deoxynucleotidyl transferase (TdT)-mediated dUTP nick end labeling (TUNEL) assay.
Cell death was determined in retinal sections by DeadEnd ™ Fluorometric TUNEL assay following the manufactur-  Western blot. Preparation of retinal extracts and Western blot was performed as previously described 69 .
In order to avoid biased matched comparisons, samples from both retinas of each animal were loaded, side by side in the gel, assuming separated membranes for control and diabetic animals. Nevertheless, membranes were developed at the same time and comparisons were made to the β-actin ratio. The membranes were incubated with the antibodies described in Table 2  www.nature.com/scientificreports www.nature.com/scientificreports/ Enzyme-linked immunosorbent assay (ELISA). The retinas were collected in ice-cold ELISA buffer (20 mM imidazole-HCl, 100 mM KCl, 1 mM MgCl 2 , 1 mM EGTA, 1 mM EDTA, 1% Triton X-100), supplemented with phosphatase inhibitors (10 mM NaF and 1 mM Na 3 VO 4 ) and complete mini protease inhibitor cocktail tablets (Roche, Sigma-Aldrich, St Louis, MO, USA). The tissue was homogenized by sonication, centrifuged at 10000 × g for 5 min at 4 °C and the supernatant was collected. Protein concentration was determined using BCA (Pierce Biotechnology, Rockford, IL, USA). The levels of TNF and IL-β were quantified in accordance to the manufacturer's instructions (PeproTech, London, UK), using a microplate reader (Synergy HT; Biotek, Winooski, VT, USA). The results were normalized to the total amount of protein in each sample.
Optical coherence tomography (OCT). Retinal thickness and structure were analyzed followed Spectrum Domain-OCT (coupled to a Phoenix Micron IV Retinal Imaging Microscope, Phoenix Technology, Pleasanton, CA, USA), as previously described 70 , with minor modifications as follows. Briefly, the animals were anesthetized using ketamine (80 mg/kg; Imalgene ® 1000) and xylazine (5 mg/kg; Ronpum ® 2%), the pupils were dilated using tropicamide (Tropicil Top ® , 10 mg/ml) and oxybuprocaine (Anestocil ® ) was applied in the cornea for topical anesthesia. Corneal hydration was maintained using carmellose sodium (Celluvisc ® ). OCT was performed at baseline (before diabetes induction), one month after the onset of diabetes and at the end of the study (2 months after diabetes onset). Eye fundus images were obtained using the Micron IV Retinal Imaging Microscope for image guidance. The thickness of retinal layers was analyzed after OCT scan data segmentation with InSight (Phoenix Technology). For each eye, 4 B-scans were analyzed (2 images above and 2 images bellow the optic nerve). The results were normalized to the baseline, as follows: x xi x 100% − × x i represents baseline retinal thickness; x represents retinal thickness at the defined timepoint.
Fluorescein angiography. The animals were anesthetized as described for OCT and the eyes were treated equally. Animals were injected with 5% fluorescein (100 μl, i.p.), as described previously 71,72 . Fluorescein angiography images were acquired with the Micron IV Retinal Imaging Microscope (Phoenix Technology, Pleasanton, CA, USA) sequentially until saturation. For this experiment, only one eye per animal was analyzed (SCH 58261 treated and non-treated diabetic retinas are from different animals) in order to keep the acquisition time equivalents between conditions. The images were classified to the presence of leaky areas by three independent researchers that were masked to the experimental condition.
The permeability coefficient was estimated as previously described 73,74 , as follows. The fluorescence intensity of a major vessel and of the interstitial space was determined with ImageJ for 4 time points (30 seconds, 180 seconds, 300 seconds and 480 seconds after injection).
Fluorescence intensity ratio F was calculated to obtain the slope of the linear equation: F represents the fluorescence intensity in a given time point; ΔC represents the fluorescence difference between a major vessel and the interstitial space; t is the time after injection in seconds. The permeability coefficient P was calculated applying the formula: slope F represents the slope obtained from the fluorescence intensity over time; ΔC represents the fluorescence difference between the major vessel and the interstitial space at a given time; A represents the area of the region of interest.
Statistical analysis. Data is presented as mean ± SEM. The results were analyzed using Graphpad Prism software version 6.01 for Windows. The OCT data was analyzed using IBM SPSS statistics software version 24 for Windows. The distribution of the data was evaluated by Shapiro-Wilk normality test. The data with normal distribution was analyzed with parametric one-way ANOVA followed by Sidak's multiple comparisons test, and for data without Gaussian distribution the statistical significance was determined with Kruskall-Wallis test followed by Dunn's multiple comparison test or Mann-Whitney test, as indicated in the figure legends. For the analysis of OCT data, the Wilcoxon matched-pairs signed rank test was used. Fluorescein angiography was analyzed by two-way ANOVA using Fisher's LSD test and the area under the curve (AUC) of the permeability coefficient was calculated setting the baseline at Y = 0. Statistical significance was defined for p < 0.05.