The findings presented in the current study provide new evidence suggesting that topical administration of AM1710 (CB2R agonist) and SR141716 (CB1R antagonist), alone and as dual treatment displayed, in a differential manner, antioxidant, neuroprotective, anti-inflammatory, and vasculoprotective properties in the ESDR model. SR141716 (Rimonabant) is a widely used potent and selective CB1R antagonist [41], while AM1710 is a potent and selective CB2R agonist with very high functional selectivity for CB2R over CB1R [42]. SR141716 has been employed previously in a study investigating the role of CB1R on the development of vascular inflammation and cell death in a mouse model of DR and human retinal cell line [33]. This article was the first to show that CB1R activation plays an important role in the pathogenesis of DR and that CB1R antagonists may be beneficial in reducing diabetes-induced oxidative/nitrative stress and vasculopathy. To our knowledge, AM1710 has never been investigated in healthy or diseased retinas. Therefore, we chose to examine these two cannabinoids, and not others, in the ESDR model with the hope that the single and dual treatments of these compounds will result in novel findings regarding their beneficial use in the treatment of ESDR.

One of the major pathological components of DR is the progressive degeneration of neuronal cell types in the retina. Studies using various experimental models of DR have detected extensive loss of amacrine cells [43], photoreceptors [44], and retinal ganglion cells [45]. These findings have been substantiated by post-mortem data from diabetic human donors [8,46] or by measuring the thickness of retinal layers (OCT) in subjects [47] and in diabetic rats [8]. Numerous reports have shown the neuroprotective actions of CB2R agonists in the brain and their therapeutic potential in a plethora of neurological disorders [48]. In contrast, fewer studies have examined the role of CB2R activation in the retina.

In the present study, we report for the first time that CB2R activation by the selective agonist AM1710, administered topically as eye drops for two weeks, was able to block the diabetes-induced loss of a subpopulation of amacrine cells [nitric oxide synthetase (NOS) expressing].

Another key pathological feature of DR reported in this study was a marked decrease of NFL immunoreactivity, observed two weeks after diabetes induction, which was confirmed both by the quantification of the intensity and thickness of NFL stain. Similar alterations in the morphology of NFL were also reported in five-week [43] and three-month models [49] of STZ-induced DR. The reduction of NFL thickness has also been reported by OCT measurements in human DR patients [50,51]. AM1710 and SR141716, administered either alone or in combination, rescued ganglion cell axons from diabetes-induced damage in the retina [NFL-IR (thickness & intensity)]. To our knowledge, the current study is the first to indicate deficits in NFL as early as two weeks after the onset of diabetes. A number of human studies have reported a reduction in NFL thickness observed at the very early stage of DR, termed preclinical DR [52], or in diabetic patients with no DR [53]. Collectively, these data imply that changes in NFL thickness represent one of the earliest structural changes at the neuronal level of the diabetic retina, thus forewarning the future development of DR.

Despite the reduction in NFL thickness and intensity, as assessed by NFL IR, we did not observe a statistically significant reduction in the thickness of the GCL. Similarly, no difference was observed in the number of retinal ganglion cells (RGCs) between control and diabetic animals (H&E staining). Martin et al. [54] reported that the number of RGCs, as well as the relative thickness of GCL, remained unaffected two weeks post-STZ-induced diabetes in a murine model, in agreement with our findings. However, a statistically significant reduction was observed ten weeks post-diabetes initiation. These results suggest that NFL may be a useful marker to indirectly extract information about the fate of RGCs in optical neuropathies, including DR. Yi et al. [55] reported that axotomy of the optic nerve leads to progressive loss of ganglion cells (optic nerve crush model). The existence of a large number of RGCs, with no detectable retrograde Fluoro-Gold staining, was reported in a mouse model of glaucoma, suggesting that part of their axons was damaged, but RGCs remained functional and expressed various RGC genes [56]. The authors suggested that despite the degeneration of their distal part, the proximal portion of the axons remained unaffected one-month post-insult, thus contributing to RGCs’ survival [56]. These studies support our data and suggest that treatment of diabetes-induced NFL deficits in the ESDR will affect the viability of ganglion cells in a positive manner.

An increased number of TUNEL⁺ cells in all three nuclear retinal layers (ONL, INL, GCL) was reported as early as two weeks after diabetes induction [57]. AM1710 also provided neuroprotection in the ESDR model as suggested by the reduction of the diabetes-induced increase in cleaved caspase-3⁺ cells in the INL, compared to the diabetic untreated animals, and the abolishment of diabetes-induced reduction in INL thickness. These findings are in agreement with previous studies from our laboratory, where it was shown that the neuroprotective properties of the endocannabinoid 2-AG [25] and the synthetic cannabinoid (R)-WIN55,212-2 [20] are in part mediated via CB2R activation in the animal model of AMPA induced retinal excitotoxicity. Further proof regarding the neuroprotective properties of CB2R activation was reported by Imamura et al. [40], who showed that HU-308 (CB2R agonist) reduced degeneration of photoreceptors, both in vivo and in vitro, in an animal model of light-induced age-related macular degeneration.

CB1R blockade was reported to be neuroprotective in retinal disease models, such as DR [33] and MNU-induced retinal degeneration [36]. However, CB1R blockade did not reverse the diabetes-induced loss of bNOS expressing amacrine cells nor reduce the number of cleaved caspase-3⁺ cells in the INL. The inability of the CB1R antagonist to influence the attenuation of bNOS amacrine cells and the increase in the number of cleaved caspase 3⁺ cells may be due to the low density of CB1R expressed in nitric oxide synthetase-expressing amacrine cells that are detected (stained) by the bNOS antibody. Furthermore, the CB1R is located mainly in the GCL and less in the INL, where the cell bodies of amacrine cells are located [58]. However, the density of the CB1R in the different amacrine cells has not been reported. Our data are in agreement with a recent study reporting that the CB1R antagonist AM251 aggravated neurodegeneration in a rat model of light-induced retinal degeneration by increasing mRNA levels of pro-apoptotic markers (Bad, Bax) and activated cleaved caspase 3, as well as mRNA levels of the anti-apoptotic marker, Bcl-2. [59]. In the present study, CB1R expression was also examined in three groups (Control, Diabetic, Diabetic + SR141716). No statistically significant difference in CB1R IR was observed, suggesting that neither diabetes nor the treatment affects the expression of CB1R in the retina.

Retinal inflammation is closely linked to persistent macroglia and resident macrophage activation, which are present at the ESDR [60]. In accordance with these findings, we show that diabetes increased the number of Iba1 ⁺ cells and GFAP expression in Müller cells in the two-week model of DR. AM1710 attenuated the diabetes-induced increase in the number of reactive Iba1⁺ cells as well as the levels of TNFα in diabetic retinas, in a statistically significant manner. While the anti-inflammatory actions of CB2R agonists have been well described, this is the first report of its kind using an in vivo model of DR. CB2R activation has been shown to shift the pro-inflammatory M1 phenotype to the anti-inflammatory M2 by stimulating the release of anti-inflammatory cytokines [61]. We have also shown that CB2Rs are expressed in reactive macrophages in the retina, as assessed by their colocalization with Iba1. The synthetic cannabinoid (R)-WIN55,212-2 (CB1R/CB2R agonist) attenuated the AMPA-induced increase in the activation of these cells, acting via CB2Rs [20]. AM1710 also reduced macroglia activation in diabetic retinas, as shown by the reduction of GFAP IR. This result is consistent with data showing an up-regulation of CB2Rs in Müller cells [62], suggesting that CB2R plays a crucial role in the physiology of macroglia. Sun et al. [63] reported that the CB2R agonist JWH015 attenuated GFAP expression in astrocytes, an effect that was abolished with the pre-treatment of the animals with the CB2R antagonist AM630.

The mechanism via which CB2R agonists may mediate their anti-inflammatory actions has been reported in different studies. AM1710 activation of CB2R leads to adenylyl cyclase and protein kinase A (PKA) inhibition and attenuation of cAMP levels via Gi/o signaling, that suppress resident macrophage activation via the modulation of c-Jun N-terminal kinases (JNKs) [64]. CB2R activation also leads to the attenuation of NO/nitrative stress and pro-inflammatory cytokines TNF-a and IL-6 via a process that is dependent on ERK1/2 and Akt [65]. CB1R blockade has been suggested to afford anti-inflammatory properties in models of retinopathy, including DR [33,66]. In our study, blockade of CB1R did not attenuate the number of activated macroglia (Figure 5B) and resident macrophage (Figure 6B) nor TNFα levels (Figure 6C) in diabetic retinas, in contrast to our original hypothesis. In a previous study, SR141716 was shown to block diabetes-induced activation of Müller cells (GFAP IR) and NFκB, localized in the vascular layers of the retina, in a mouse STZ model of DR [33]. However, it has been reported that activation, and not a blockade, of CB1R by anandamide was able to attenuate microglia-mediated neuroinflammation in the brain [67]. Despite the ability of SR141716 to reduce nitrative stress and vascular leakage, it did not affect the inflammatory components. As mentioned earlier, this difference may be due to a lower density of CB1Rs located on glial cells.

In agreement with our observation, Solino et al. [59] reported that the application of the CB1R antagonist, AM251, in a rat model of light-induced retinal degeneration increased macroglia reactivity and the expression of TNFα. Another possible explanation for the inability of CB1R blockade to promote anti-inflammatory actions may be due to the existence of a sexual dimorphism effect. A recent study reported that the inflammatory response in the brain is dependent on CB1Rs on resident macrophage cells in male mice but independent of CB1R in female mice [68]. In our study, most of the animals employed were female.

In the model of ESDR, upon onset of diabetes, there was an increase in nitrotyrosine (NT) expression in the retina, a marker of nitrative stress. These findings are in agreement with Hernandez-Ramirez et al. [10], who reported increases in the expression of NO and 3-NT in an STZ-induced DR model seven days post-diabetes initiation. We report that blockade of CB1R and activation of CB2R by topical administration of SR141716 or AM1710, respectively, reduced nitrative stress, as indicated by the attenuation of the diabetes-induced increase in the number of NT+ cells. These findings are in agreement with studies showing CB1R blockade attenuated oxidative damage in various disease models [33,69].

Hyperglycemia-induced oxidative and nitrative stress contribute to microvascular complications of DR by promoting DNA fragmentation and apoptosis of endothelial cells [70]. The reduction of NT levels has been shown to protect retinal micro-vasculature from the nitrative insult [33,71]. In the present study, we observed an increase of NT expression in the diabetic retina (Figure 7A,B) that was reversed by all three treatments. We also showed that NT is localized in endothelial cells, as observed by its co-localization with CD-31 (endothelial cell marker, Figure 8A), suggesting that blood vessels undergo nitrative damage. This finding is confirmed by a previous study showing that NT expression is prominent in the retinal vasculature of diabetic animals [72]. The nitrative stress-induced endothelial cell impairment is a possible explanation for the increased vascular leakage that is also reported in the present study (Figure 8B). We observed an increase in retinal vascular permeability two weeks post-STZ treatment. Studies in diabetic rats reported that increased vascular permeability in the retina was observed as early as six [73] or seven days [74] after STZ administration. AM1710 reduced vascular leakage in diabetic retinas (Figure 8B). In compliance with our results, CB2R activation was shown to reduce the expression of cell adhesion molecules in human retinal endothelial cell cultures and attenuate TNF-α induced retinal vascular permeability in vivo [75,76]. Similar to the actions of AM1710, SR141716 (CB1R antagonist) also blocked the diabetes-induced increase of vascular permeability, suggesting that the blockade of CB1R is efficacious in maintaining the integrity of BRB. Also, vascular protective actions were afforded by SR141716 in a mouse model of retinal degeneration induced by MNU (N-methyl-N-nitrosourea) [36] and a mouse model of DR [33].

The dual treatment (AM1710 and SR141716) was efficacious in attenuating nitrative stress (Figure 7B,C). However, its efficacy was not different from the single treatments on NFL (Figure 1B,C) or bNOS IR (Figure 3B). Also, it did not afford any statistically significant effect in reducing Müller cell nor resident macrophage activation (Figure 5B and Figure 6B, respectively). Taking into account the anti-inflammatory actions of AM1710 and the inability of SR141716 to affect macroglia and resident macrophage activation, we hypothesized that the dual treatment would mimic single CB2R agonist treatment. Studies in the brain have shown that CB1R and CB2R can form functional heterodimers in neuronal [77], as well as in Iba1⁺ cells. Increased expression of CB1R/CB2R heterodimer has been observed as a response to an inflammatory stimulus [78]. The most interesting part, which seems relevant to our study, is that CB1R/CB2R heterodimers display cross-antagonism in which administration of CB1R antagonists was able to block the effects of CB2R agonists [77,78].

Members of the ECS, endocannabinoids, and CB1/CB2 receptors, found in peripheral organs, are involved in the regulation of energy homeostasis and play an important role in obesity and metabolic disorders [79] Both CB1R antagonists and CB2R agonists modulate glucose metabolism. The synthetic cannabinoid SR141716 (Rimonabant), also employed in this study, was the first selective CB1R antagonist/inverse agonist to be investigated regarding its role in glucose metabolism in patients and animals [80]. Clinical trials [Rimonabant in Obesity (RIO) program] in overweight/obese non-diabetic and type-2 diabetes patients reported that Rimonabant reduced glucose tolerance and insulin resistance [81,82]. Similar results were reported in animal studies employing different CB1R antagonists and various models of obesity [80,83]. The efficacious effects of CB1R antagonists in energy metabolism were negatively affected by their CNS psychiatric adverse effects. In order to bypass this disadvantage, peripheral neutral CB1R antagonists have been developed and were shown to improve glycemic and insulin resistance [84,85]. 2-AG activation was reported to attenuate insulin secretion by pancreatic beta-cells of the mouse via CB2R activation [86] and to improve glucose homeostasis in the rat [87]. All in all, the peripheral endocannabinoid system plays an important role in glucose metabolism. The roles of CB1R antagonist and CB2 agonist in the retina (CNS) and periphery suggest that these members of ECS are prospective putative therapeutics for diabetes and diabetic retinopathy.

In closing, our findings suggest that topical administration of the three cannabinoid treatments, such as eye drops, provides protection to the diabetic retina in a differential manner against the four pathologies of ESDR. The actions of both CB2R activation and CB1R blockade in restoring ganglion cell axons (NFL-IR) in ESDR suggest that both agents may be effective in retarding RGC death. AM1710 is efficacious as an antioxidant, anti-inflammatory, neuroprotective and vasculoprotective agent and, thus, a promising new therapeutic for ESDR. Further advancement of retinal imaging to screen and identify the early events in DR, such as neurodegeneration in diabetic patients, is crucial for selecting neuroprotective drugs and implementing personalized treatments. As our findings clearly implicate the endocannabinoid system, the therapeutic benefits of this class of compounds should also extend to patients with diabetic nephropathy and cardiopathy/stroke since DR has been associated with the development of these diseases [2].